Comparability of Reporting
Between the Birth
Certificate and the National
Natality Survey

 

U.S. DEPARTMENT OF
HEALTH, EDUCATION, AND WELFARE
Public Health Service
Office of Health Research, Statistics, and Technology
National Center for Health Statistics

 
 

 

 

Library of Congress Cataloging in Publication Data

Querec, Linda J.
Comparability of reporting on birth certificates and national natality survey question-
naires.

(Vital and health statistics : Series 2, Data evaluation and methods research ; no. 83)
(DHEW publication ; no. (PHS) 80-1357)

Includes bibliographical references.

Supt. of Docs. no.: HE 20.6209:2/82

1. Registers of births, etc.—United States. 2. Birth certificates—United States. 3. United
States—Statistics, Vital. 4. Social surveys—United States. I. Title. II. Series: United States.
National Center for Health Statistics. Vital and health statistics : Series 2, Data evaluation
and methods research ; no. 83. III. Series: United States. Dept. of Health, Education, and
Welfare. DHEW publication ; no. (PHS) 80-1357.
RA409.U45 no. 83 [HA38] 312'.07'23s 79-607801
ISBN 0-8406-0178-6 [312.07'23]

 

 

For sale by the Superintendent of Documents, U.S. Government Printing Office, Washington, D.C. 20402
DATA EVALUATION AND METHODS RESEARCH Series 2
Number 83

Comparability of Reporting
Between the Birth
Certificate and the National
Natality Survey

Describes comparability of reporting of selected items common to
the birth certificate and National Natality Survey questionnaires.
Items include age and education of parents, number of previous
children, number of fetal deaths, plurality, birth weight, length of
pregnancy, and receipt of prenatal care.

DHEW Publication No. (PHS) 80-1357

 

U.S. DEPARTMENT OF HEALTH, EDUCATION, AND WELFARE
Public Health Service
Office of Health Research, Statistics, and Technology
National Center for Health Statistics
Hyattsville, Md. April 1980
NATIONAL CENTER FOR HEALTH STATISTICS

DOROTHY P. RICE, Director

ROBERT A. ISRAEL, Deputy Director
JACOB J. FELDMAN, Ph.D., Associate Director for Analysis
GAIL F. FISHER, Ph.D., Associate Director for the Cooperative Health Statistics S ystem
ROBERT A. ISRAEL, Acting Associate Director for Data Systems
ALVAN O. ZARATE, Ph.D., Acting Associate Director for International Statistics
ROBERT C. HUBER, Associate Director for Management
MONROE G. SIRKEN, Ph.D., Associate Director for Mathematical Statistics
PETER L. HURLEY, Associate Director for Operations
JAMES M. ROBEY, Ph.D., Associate Director for Program Development
GEORGE A. SCHNACK, Acting Associate Director for Research
ALICE HAYWOOD, Information Officer

DIVISION OF VITAL STATISTICS

JOHN E. PATTERSON, Director
ALICE M. HETZEL, Deputy Director
ROBERT L. HEUSER, M.A., Chief, Natality Statistics Branch
JOSEPH D. FARRELL, Chief, Programming Branch
MABEL G. SMITH, Chief, Statistical Resources Branch

Vital and Health Statistics-Series 2-No. 83

 

DHEW Publication No. (PHS)80-1357
Library of Congress Catalog Card Number 79-607801
CONTENTS

Introduction ..

 

Sources and Limitations of Data

 

 

Selected Findings

Age of Mother

 

Age of Father

 

Number of Children Born Alive and Still Living

 

Number of Children Born Alive and Now Dead

 

 

Live-birth Order

Number of Fetal Deaths

 

Education of Parents

 

 

Plurality

Birth Weight

 

 

Length of Pregnancy

Month of Pregnancy Prenatal Care Began

 

Number of Prenatal Visits

 

Comparability of Reporting by Race of Child and Age and Education of Mother .......ccceeeeeererrcnereecsasnnns

References

 

List of Detailed Tables

 

Appendixes
I. Technical Notes
II. 1972 National Natality Survey Source Documents

 

 

LIST OF TEXT TABLES

A. Percent of cases reporting identical age of mother and percent reporting age of mother within the
same age group on the birth certificate and National Natality Survey questionnaires, by age of
mother reported on the birth certificate, 1972

 

B. Percent of cases reporting identical age of father and percent reporting age of father within the
same age group on the birth certificate and National Natality Survey mother’s questionnaire,
by age of father reported on the birth certificate, 1972

 

C. Percent of cases reporting identical education of mother and percent reporting education of
mother within the same education group on the birth certificate and National Natality Survey
mother’s questionnaire, by education of mother reported on the birth certificate, 1972 .............

10

11

12

14

15

28
53
Percent agreement of selected items on the birth certificate and National Natality Survey ques-
tionnaires, by race of child reported on the birth certificate, 1972 ..auiinurnimanissnssmsnises

Percent agreement of selected items on the birth certificate and National Natality Survey ques-
tionnaires, by age of mother reported on the birth certificate, 1972

Percent agreement of selected items on the birth certificate and National Natality Survey ques-
tionnaires, by education of mother reported on the birth certificate, 1972 ....cccvverrvrerrivreniueennnne

 

SYMBOLS

Data not available--------------- domme --

Category not applicable---------eseeememmceaaeeeee |

 

Quantity zero
Quantity more than 0 but less than 0.05--—-- 0.0

Figure does not meet standards of
reliability or precision---------------ceceeeeeeeeee

 

 

 
COMPARABILITY OF REPORTING BETWEEN
THE BIRTH CERTIFICATE AND
THE NATIONAL NATALITY SURVEY

Linda J. Querec, M. A., Division of Vital Statistics

INTRODUCTION

For nearly two decades more than 99 per-
cent of all live births in the United States have
been registered!; however, little information has
been available on the quality of birth certificate
reporting. One method of examining quality of
reporting is by comparison with vital record fol-
lowback survey data. This report was designed
to measure the extent of agreement between the
responses provided on the birth certificate and
responses provided on a mailed questionnaire
from the last National Natality Survey. Al-
though the last National Natality Survey was
conducted in 1972, there is little reason to be-
lieve that comparability of reporting changes
enough from year to year at the national level
for the findings of this report to be substantially
affected.

Some comparability studies exist,2# but
most are limited in scope and often are con-
fined to medical conditions of the mother
and/or the child. One exception is a study?
conducted in New York State in 1972 in which
specific information on the birth certificates was
compared with the same information on the
hospital records. This study contains many of
the same items that are compared in this report.
Two important similar studies were conducted
by the U.S. Bureau of the Census to assess the
accuracy of the 1970 census reporting. One
study matched responses to items on the census
questionnaire with responses to the 1970 Cur-

rent Population Survey.® The second study used
reinterviews to evaluate census responses.’

The items considered in this study are com-
mon to both the birth certificate and the survey
questionnaires and include the following: age
and education of parents, plurality, birth weight,
length of pregnancy, month of pregnancy that
prenatal care began, number of prenatal visits,
number of children born alive and still living,
number of children born alive and now dead,
live-birth order, and number of fetal deaths. Al-
though it cannot be determined from this study
which information is correct when the two doc-
uments differ on a particular item, in some in-
stances it is possible to hypothesize which in-
formation is more accurate.

SOURCES AND LIMITATIONS OF DATA

The National Natality Survey (NNS) con-
sisted of a 1-in-500 sample of births drawn from
the microfilm file of birth certificates received
by the National Center for Health Statistics
(NCHS) for births occurring in the survey year.
The original sample consisted of 6,505 births,
816 of which were eliminated because they were
either reported as or inferred to be out-of-
wedlock births. (See appendix I for method of
inference.) The remaining 5,689 births were in-
cluded in the survey. Each mother in the survey
received a mother’s (M) questionnaire that re-
quested health and demographic information.
The name and address of the attending phy-
sician and the hospital where the delivery oc-
curred are listed on the birth certificate, thereby
making it possible to obtain additional informa-
tion from these sources. This information in-
cludes a pregnancy history, information about
prenatal and postpartum care, and information
concerning the delivery episode. If the address
of the attending physician was the same as that
of the hospital where the delivery occurred, then
one questionnaire, the long hospital (HL) ques-
tionnaire, was sent to the hospital. If the phy-
sician’s address was different from that of the
hospital where delivery occurred, then a physi-
cian (P) questionnaire was sent to the physician
and a short hospital (HS) questionnaire was sent
to the hospital. The HS and P questionnaires to-
gether contain the same information as the HL
questionnaire. Information on the sample design
and collection of data as well as samples of the
questionnaires and the U.S. Standard Certificate
of Live Birth are included in the appendixes.

Limitations of the data centered, for the
most part, on various forms of nonreporting.
The basic form of nonreporting was failure to
respond to the questionnaire (unit nonresponse).
Of survey mothers, 71.5 percent responded to
the questionnaire; among physicians who re-
ceived the P questionnaire, 72.2 percent re-
sponded; and among hospitals that received
either the HL or HS questionnaire, 85.4 percent
responded (table III). Because the questionnaires
differ in the type of data requested, some but
not necessarily all items may be available for a
particular case.

As shown in table IV, mothers of white
births had a much higher response rate (73.6
percent) than mothers of all other births (56.0
percent). Response rates for mothers varied
greatly by age of mother, color of child, and
live-birth order. They ranged from a high of 82.7
percent among 25-29-year-old mothers of white
first births to a low of 39.1 percent among
20-24-year-old mothers of third or higher order
births other than white.

Questionnaires that are returned may have
one or more unanswered questions or an impos-
sible or illegible response, all of which are classi-
fied as item nonresponse. Although missing data
were imputed for other reports from the NNS,
no imputation was done for this report, because
the purpose of this study was to evaluate the

items as they were reported. Appendix I con-
tains a discussion of procedures that were used
to improve response rates.

Although a birth certificate was available for
each case, not all States use the U.S. Standard
Certificate of Live Birth and some certificates
do not contain all items on the standard birth
certificate. Thus certain data may be available
on the survey questionnaires, but not from all
the birth certificates. Table II shows the report-
ing areas for these selected items. One item, the
number of fetal deaths, is not reported consist-
ently from State to State—some States include
all fetal deaths and others include only those
occurring after a specified period of gestation.

As with the questionnaires, an item con-
tained on the birth certificate may have been
left either unanswered or may have had an im-
possible or illegible response. Unlike the ques-
tionnaires, no followback could be done by
NCHS to complete or correct the birth certifi-
cate item. Table V contains item nonresponse
rates for the birth certificate and questionnaires.

Responses to selected items on the question-
naires were matched with the corresponding
data from the birth certificate. The extent of
agreement for each item was determined only
for those births in the sample for which there
was a response to the item on both the certifi-
cate and the questionnaire. For each item, both
the text tables and the appendix tables should
be consulted for an indication of the extent of
exclusion of cases due to nonreporting and non-
response. The agreement rate was obtained by
computing the percent of cases for which an
identical response was provided on both the
questionnaire and the birth certificate by using
the birth certificate and its response as the
base.

Because not all births were utilized in the
comparison, the results reported here may not
be completely representative of all births to mar-
ried women. It is not possible to determine
whether accuracy of reporting among births
omitted from this study either because of occur-
rence in nonreporting States or because of non-
response to the survey differs from accuracy
among those included in the survey. Thus com-
parability of reporting might differ from that
shown in this report, but the direction and mag-
nitude of these differences are unknown.
SELECTED FINDINGS

A great deal of variation was found in com-
parability of reporting for items common to
the birth certificate and the National Natality
Survey questionnaires. Comparability ranged
from excellent to poor. Items that had an excel
lent level of agreement (90 percent or better)
were age of mother (mother’s questionnaire),
plurality, number of children born alive and still
living, number of children born alive and now
dead, and live-birth order. Those items with
good agreement (80 to 89 percent) included age
of mother (long/short hospital questionnaire),
age of father, birth weight, length of pregnancy,
and number of fetal deaths. Comparability of
reporting of parents’ education was fair (70 to
79 percent), and reporting of prenatal care was
poor (less than 50 percent).

A discrepancy of *1 may be significant for
certain items and not for others. For example, a
discrepancy of one unit would not make a
notable difference in the reporting of either pre-
natal visits or birth weight, but would make a
significant difference in the reporting of either
live-birth order or plurality. Although agreement
of prenatal visits was poor, the discrepancy was
often only one or two visits. If cases with a dif-
ference of one or two visits were included with
those that have an identical number of visits,
then agreement increased markedly, to 56 and
66 percent on the mother’s and hospital/physi-
cian questionnaires, respectively.

The impact of discrepancies is reduced
further in tabulations in which the data are pre-
sented in grouped form. Many items are most
useful when tabulated in this manner. Age of
mother and age of father are usually tabulated in
5-year age groups. Although reporting of identi-
cal age of mother was found for 90.7 percent of
the comparison cases, age of mother was re-
ported within the same 5-year age group for
96.9 percent of the cases. For age of father,
agreement increased from 84.5 percent to 94.6
percent when 5-year-age-group tabulations were
used. Data are shown in grouped form for other
variables as well. Agreement of education of
mother increased from 77.2 to 85.5 percent for
grouped data. For birth weight, agreement
within 500-gram weight groups was found for
96.0 percent of the comparison cases.

The results of this study compare favorably
with those obtained from similar studies con-
ducted by the Bureau of the Census and the
New York State Department of Health. For
example, agreement of age of mother between
the birth certificate and the hospital record was
89.7 percent in the New York State study,
which was very close to the 87.8 percent found
in this study. In addition, agreement of report-
ing of age of mother within 5-year age groups
was better in this study than in a study of com-
parability of reporting between the 1970 census
and the 1970 Current Population Survey (CPS)—
96.9 percent in this study compared with 93.3
percent in the census study.

Other items included in this and other stud-
ies were age of father, live-birth order, number
of fetal deaths, education of parents, birth
weight, gestation, and month of pregnancy that
prenatal care began. Generally, comparability
of reporting for these items was very similar in
each of these studies.

In summary, this study shows that (1) the
comparability of reporting of most items is
good, (2) the impact of discrepancies is min-
imized further when data are presented in tabu-
lations of grouped data, and (3) these findings
are similar to those of other studies.

AGE OF MOTHER

Age of mother was available from the
mother’s questionnaire, the long/short hospital
questionnaire, and the birth certificate. The
birth certificate and the HL/HS questionnaire
asked for the age of the mother at the time of
the birth, that is, the age at her last birthday
preceding the birth. The M questionnaire re-
quested the mother’s date of birth and from this
date her age at time of delivery was determined.
It is reasonable to expect that the mother’s age
that is based on a date of birth would be more
accurate because the birth date is constant and
therefore should be easier to recall. This also
eliminates any tendency to report age as of the
nearest birthday.

When data were available from both the
birth certificate and the M questionnaire, an
identical age was reported for 90.7 percent of
the mothers (table 1). In cases where disagree-
ment was found, it was most likely to be
Table A. Percent of cases reporting identical age of mother and percent reporting age of mother within the same age group on the birth
certificate and National Natality Survey questionnaires, by age of mother reported on the birth certificate, 1972

 

 

 

 

 

 

mr Age of mother reported on birth certificate
A t of f th T
§reainent oF age of mother © 11 Under 20 | 20-24 | 25-29 | 30-34 | 35 years
years years | years | years | and over
Mother's questionnaire
Certificate same as QUESTIONNAITE ...........ceervviuiuiureeeaeeesrrenreneeeeeeessesranaaeaes 90.7 91.6 90.8 91.7 89.4 86.5
Within same age group on certificate and questionnaire...............cccceeevvvnnnne 96.9 96.1 97.3 97.6 96.6 94.0
Long/short hospital questionnaire

Certificate Same as QUESHIONNBIFR w.....iviiririreriesrssisissrssrsassesenssunssransennsrrers 87.8 93.6 88.2 85.6 87.0 83.2
Within same age group on certificate and questionnaire.............c..cceeevvvvnnnne 96.2 98.1 96.8 95.7 95.1 93.1

 

 

 

 

 

 

 

because a lower age was determined from the
birth certificate than from the M questionnaire:
6.6 percent reported a younger age at delivery
on the birth certificate compared with 2.7
percent reporting a younger age on the M
questionnaire. In most cases the discrepancy was
only 1 year, with 2.5 percent having a discrep-
ancy of 2 years or more.

The amount of agreement of the age item
varied only slightly with the age of the mother
that was reported on the birth certificate.
Agreement was lowest (86.5 percent) among
mothers aged 35 years and over; among younger
mothers it was approximately 90 percent.

As shown in table 1, agreement between the
birth certificate and the HL/HS questionnaire
was only slightly lower. The comparison, limited
to the 84.0 percent with response on both
records, showed that 87.8 percent reported an
identical age on both records. Unlike the com-
parison of the birth certificate and the M
questionnaire, the discrepant cases usually re-
ported a higher age on the birth certificate than
on the HL/HS questionnaire (7.9 percent with a
higher age on the birth certificate compared
with 4.3 percent with a higher age on the
questionnaire). Only 2.9 percent showed a dis-
crepancy of 2 years or more. The proportion of
cases with exact agreement ranged from 93.6
percent among mothers under 20 years old to
83.2 percent among mothers aged 35 years and
older. The New York State study? found that
89.7 percent reported an identical age on both

the birth certificate and the hospital record—
comparable to the 87.8 percent in this study.

Since natality data are generally tabulated
by 5-year age groups for analytical purposes, an
error of 1 year on the birth certificate would
result in a difference in the tabulations only
when the correct age fell within another age
interval. The proportion of mothers with re-
ported age within the same 5-year age group on
both the birth certificate and the mother’s ques-
tionnaire was 96.9 percent and varied by age as
shown in table A. Agreement between the birth
certificate and the long/short hospital question-
naire by 5-year age groups was similar, with age
reported within the same 5-year age category for
96.2 percent of the cases.

These results compare favorably with those
obtained from a study® conducted by the
Bureau of the Census which matched mothers’
ages that were reported in the 1970 census with
those reported in the 1970 CPS. This study
found that 93.3 percent of the respondents
reported their age to be within the same 5-year
age group in both the 1970 census and the 1970
CPS.

AGE OF FATHER

Survey data on age of the father are available
only from the mother’s questionnaire. The
questionnaire asked for father’s date of birth,
but the birth certificate asked for father’s age at
time of delivery. Agreement between these
Table B. Percent of cases reporting identical age of father and percent reporting age of father within the same age group on the birth
certificate and National Natality Survey mother’s questionnaire, by age of father reported on the birth certificate, 1972

 

 

 

Age of father reported on birth certificate
Agreement of age of father Total Il ynder 20 | 20-24 | 25-29 | 30-34 | 35-39 | 40 years
years years years | years years | and over
Certificate same as QUESTIONNAINe..........ceuueeeenesrerinierinsi irre. 84.5 81.8 86.7 86.9 82.8 78.8 75.9
Within same age group on certificate and questionnaire................ 94.6 94.4 95.3 95.8 93.6 92.3 90.7

 

 

 

 

 

 

 

 

 

documents was not as good for the father’s age
as for the mother’s age. Of the comparison cases,
84.5 percent showed exact agreement for this
item (table 2). Only 3.7 percent had a discrep-
ancy of 2 years or more. A larger proportion
(9.1 percent) reported a younger age on the
birth certificate than on the M questionnaire
(6.4 percent). Similar results, with 88.6 percent
reporting identical age, were found for the New
York State study.

The proportion of cases showing exact agree-
ment varied with the age of the father as
determined from the birth certificate—increasing
from 81.8 percent for fathers under 20 years of
age to 86.9 percent for fathers 25-29 years and
then decreasing with age to 75.9 percent for
fathers 40 years and over. Among fathers under
35 years, the discrepancy was most likely the
result of a lower age given on the birth certifi-
cate than the one determined from the question-
naire; . among fathers 35 years and over the
discrepancy was more likely to be in the oppo-
site direction.

As noted previously, natality data most
often are tabulated by 5-year age groups and
small discrepancies generally will not distort
these tabulations. Father’s age was reported to
be within the same 5-year age group for 94.6
percent of the sample cases for which this
information was provided on both documents.
Agreement by age of father is shown in table B.

NUMBER OF CHILDREN BORN
ALIVE AND STILL LIVING

The number of previous children “born alive
and still living” (BASL) was available from the
birth certificate, the mother’s questionnaire, and
the long/short hospital questionnaire. The birth

certificate simply requested the number of
children still living from previous deliveries. The
HL/HS questionnaire asked for the number of
children now living, and a check box was pro-
vided for those cases requiring the response
“none.” The mother’s questionnaire, which
probably obtained the most accurate question-
naire response, asked for the total number of
children (including the present birth) born to
the mother and the name, sex, and dates of birth
and death of any deceased child to determine
the number still living.

This item and the number of children born
alive, but now dead, are used to determine
live-birth order from the certificate. Although
the correct response in cases where there had
been no previous children would be “0” or the
word “none,” an X, dash, or blank sometimes
was used. These may have been used to indicate
that the answer was unknown; however, in cases
where a numeric response was given for one item
and an X, dash, or blank given for the other, the
non-numeric entry was treated as “00.” In addi-
tion, if an X or dash response was given to both
items, then the response was considered “00”
for both. The format for the mother’s and the
long/short hospital questionnaires, which dif-
fered from that of the birth certificates, made
such an edit unnecessary for the survey data.

No discrepancy was found for 97.0 percent
of the cases with responses on the certificate and
the M questionnaire (see table 3), and less than 1
percent disagreed by more than one child.
Agreement was greatest among those reporting
no children BASL on the birth certificate (98.2
percent) and least among those reporting five
children or more BASL (91.7 percent or less).
Among those reporting two children or more
BASL on the birth certificate, deviation was
more likely to be that a higher number (most
often a deviation of one) was reported on the
birth certificate than on the M questionnaire.
Mothers who did not carefully read the instruc-
tions on the M questionnaire may have excluded
the present birth and included only previous
births. This could account for the discrepancy.

Comparability was slightly lower between
the birth certificate and the HL/HS question-
naire (92.1 percent) than it was between the
certificate and the M questionnaire (97.0 per-
cent). Only 1 percent disagreed by more than
one child (table 3). Agreement ranged from 78.3
percent for those reporting eight children or
more BASL on’ the birth certificate to 96.8
percent for those reporting no children BASL.
Among those reporting two children or less
BASL on the birth certificate, disagreement was
most likely the result of fewer children being
reported on the birth certificate; among those
reporting three children or more, most often
fewer children were reported on the
questionnaire.

NUMBER OF CHILDREN BORN
ALIVE AND NOW DEAD

The mother’s questionnaire, the long/short
hospital questionnaire, and the birth certificate
all contained an item for determining the num-
ber of previous children “born alive and now
dead” (BAND).

As evidenced in table 4, comparability was
very high with 98.3 percent of the certificates
and the M questionnaires showing exact agree-
ment. This was largely a result of the high
proportion of agreement among those reporting
“none” on. the birth certificate (99.1 percent).
This category accounted for 96.5 percent of all
births. However, agreement was not as good
among those reporting one or more previous
children BAND (77.5 percent). For this category
there was a definite bias toward a greater
number reported on the birth certificate than on
the M questionnaire (21.7 percent). This is
surprising because the M questionnaire probed
further than the birth certificate, asking name,
sex, and dates of birth and death of any children
BAND, thereby encouraging the reporting of
some cases on the questionnaire that the birth
certificate might miss.

Agreement between the HL/HS question-
naire and the birth certificate was similar to that
between the M questionnaire and the birth
certificate (table 4). Identical reporting was
found for 97.5 percent of the cases, mainly a
result of 99.0-percent agreement among those
reporting no children BAND on the birth certifi-
cate—the category in which the vast majority of
cases fell. There was less consistency with those
reporting one child or more BAND on the birth
certificate when compared with the HL/HS
questionnaire than when compared with the M
questionnaire—only 57.3 percent were in exact
agreement. As with the M questionnaire, there
was a bias toward a greater number reported on
the birth certificate (42.0 percent).

LIVE-BIRTH ORDER

Live-birth order is defined as the total
number of children born alive to the mother,
including those still living, those now dead, and
the current birth. Reporting of live-birth order
showed a high degree of comparability between
the mother’s questionnaire and the birth certifi-
cate (96.1 percent) as shown in table 5. Fewer
than 1 percent were discrepant by more than
one birth. Comparability was greatest among
births reported as first and second order on the
birth certificate (98 percent). Among higher
order births agreement ranged from 73.3 to 94.8
percent. When a discrepancy was found, the
birth certificate generally reported a higher birth
order than the questionnaire, as expected on the
basis of the direction of difference for the two
components (BASL and BAND).

Comparability for this item was slightly
lower between the long/short hospital question-
naire and the birth certificate (91.1 percent)
than it was between the M questionnaire and the
birth certificate. Even so, only 1.3 percent
showed a discrepancy of two births or more.
Agreement was less than 90 percent for all birth
orders except the first. Among discrepant cases
of fourth and higher orders, the birth certificate
was much more likely to report a higher birth
order than the questionnaire.

The New York State study found similar
results: 94.8 percent of the sample cases showed
no discrepancy in reporting live-birth order. A
study’ done by the Bureau of the Census
comparing census responses with responses to a
reinterview survey found comparability to be
lower with 89.8 percent reporting the same
number of live-born children.

NUMBER OF FETAL DEATHS

Each of the survey questionnaires as well as
the birth certificate contained a question asking
for the number of previous fetal deaths. Unfor-
tunately, the wording of the item differed
among the data sources and even though all
birth certificates required reporting of previous
fetal deaths, the reporting requirements were
not consistent for all areas. Although 42 States
required reporting of previous fetal deaths at
any gestational age, 1 State required reporting
only after 16 weeks, and an additional 7 States
and the District of Columbia required reporting
only after 20 weeks. The States with limited
reporting accounted for 28 percent of all births
in 1972. Thus some discrepancies would be
expected because of these differences.

A comparison can be made of the response
on the birth certificate item concerning the
number of fetal deaths with the responses given
both on the mother’s and the long/short hospital
questionnaires. The M questionnaire used two
questions to obtain the number of fetal deaths;
the first asked for the number of stillbirths, and
the second asked for the number of miscarriages.
Although a stillbirth was defined as a baby born
dead, there was no definition of miscarriage, but
only an indication that any previously included
stillbirths should be excluded. Although the two
questions do not clearly indicate which gesta-
tional ages apply to which category, the sum of
stillbirths and miscarriages should be a good
indication of the total number of fetal deaths.

Agreement was found for 89.2 percent of
the comparison cases (table 6). The discrepant
cases were more likely to have fewer fetal deaths
reported on the birth certificate than on the
questionnaire (9.7 percent). About 90 percent
of those reporting no fetal deaths on the birth
certificate (the category containing the great
majority of cases) also reported none on the
questionnaire.

Birth certificates that reported one previous
fetal death were in agreement with the survey
questionnaire 80.1 percent of the time. Among

those reporting two or three or more fetal
deaths on the birth certificate, the agreement
rates were 86.8 and 75.0 percent, respectively.
Within categories, there was no consistent bias in
either direction in those cases where a discrep-
ancy was found.

The HL/HS questionnaire was much more
specific in its request for the number of fetal
deaths by asking for all pregnancies that did not
end in a live birth, including all miscarriages,
abortions, stillbirths, and so forth.

As shown in table 6, exact agreement among
matched cases was 88.7 percent. Discrepancies
were more likely the result of fewer fetal deaths
reported on the birth certificate than on the
questionnaire. This is due solely to the large
number of cases reporting no fetal deaths on the
certificate for which this is the only possible dis-
crepancy that could occur.

In cases where the birth certificate reported
no fetal deaths, agreement was 90.3 percent,
which is almost identical to that between the M
questionnaire and the birth certificate for that
category. Among cases reporting two fetal
deaths or more on the birth certificate, there
was a greater amount of discrepancy between
the birth certificate and the HL/HS question-
naire than between the birth certificate and the
M questionnaire. Of those reporting one, two, or
three or more fetal deaths on the birth certifi-
cate, 77.2, 73.4, and 55.2 percent, respectively,
reported consistently on the two forms. Each of
these categories showed a bias toward a greater
number reported on the birth certificate than on
the HL/HS questionnaire. The direction of this
bias might be unexpected because in some
reporting areas, fetal deaths of early gestational
age do not have to be reported on the birth
certificate.

Agreement for this item in the New York
study was 81.0 percent, somewhat lower than
the 88.7 percent agreement between the HL/HS
questionnaire and the birth certificate.

EDUCATION OF PARENTS

The mother’s questionnaire was the only
source of survey data on educational attainment
of parents. The questionnaire requested the
highest grade of regular school completed and
that any specialized training such as beauty-
barber college and hospital schools be listed

separately. The standard birth certificate re-

quested the highest grade of regular school
completed with no provision for specialized
training. Inclusion of specialized training with
the years of regular school reported on the birth
certificate could result in an upward bias in the

data. In 1972, 39 States followed the standard

birth certificate and required the reporting of
educational attainment of parents. Generally,
the item was worded on the State birth certifi-
cates the same as on the standard birth certifi-
cate except for the Illinois certificate that had a
provision similar to the M questionnaire for
reporting specialized training.

As shown in table 7, among mothers for
whom this information was provided on both
sources, 77.2 percent reported an identical
amount of education on both forms. An addi-
tional 15.6 percent showed a discrepancy of
only 1 year. A wide variation in agreement rates
was found when the data were tabulated by
years of education recorded on the birth certifi-
cate. The proportion with identical education re-
ported on the two forms varied from 55.4 per-
cent among those with 13-15 years of school to
90.0 percent among those with 12 years. Be-
cause completion of 12 years is usually marked
by receipt of a diploma, this category is prob-
ably one of the easiest to identify and recall;
therefore, it is not surprising that a high rate of
agreement was found.

The large proportion of disagreement found
for the category 13-15 years might be the result
of an upward bias on the birth certificate caused
by adding specialized training to the regular
schooling by those with 12 years of education.
For example, a mother who had finished high
school and later attended business school may
have included the latter with her regular school-
ing on the birth certificate, in which case she
would be erroneously included in the category
13-15 years. However, because the mother’s
questionnaire contained a separate category for
reporting specialized training, only regular
schooling would be included, and her education
would be correctly reported as 12 years. The
type of discrepancies found support this conten-
tion: 35.0 percent of those included in the
category 13-15 years reported a greater number

of years of schooling on the birth certificate
than on the M questionnaire compared with
only 9.6 percent reporting a greater number on
the M questionnaire. Because natality tabula-
tions for the most part show education data in
grouped form, a difference of only 1 year (and
sometimes more) often would not result in
misclassification in the tabulation. In tabulations
of grouped data, 12 years of schooling was the
only category not combined with the other
years—this was the category with the least dis-
crepancy (10.0 percent). Although 77.2 percent
of the records reported identical education on
both forms, 85.5 percent would be in agreement
in tabulations of grouped data (table C).

As shown in table 8 a bias existed in the
reporting of education of the father, with the
greatest nonresponse to the questionnaire (43.1
percent) among those reporting the least educa-
tion on the birth certificate and the least
nonresponse (14.0 percent) among those report-
ing the most education on the certificate.

Comparability was slightly lower for educa-
tion of father than of mother; 72.3 percent
reported an identical number of years of school-
ing for fathers on both forms compared with
77.2 percent for mothers. An additional 17.7
percent was discrepant by 1 year for education
of father. The discrepancy was equally divided
between a higher and lower number reported on
the birth certificate when compared with the
questionnaire.

After classifying the data by the amount of
education of father reported on the birth certifi-
cate, it was found that comparability was
greatest among high school and college graduates
(85.8 and 74.7 percent, respectively). Agree-
ment for each of the other three categories was
low, around 55 percent. For the category 13-15
years, there was a slight bias toward reporting
more education on the birth certificate than on
the questionnaire, but this bias was smaller than
that found among mothers in this education
category. Discrepancies in other categories were
not biased in a particular direction except for
the category 0-8 years where 31.7 percent of
the sample mothers reported the fathers as
having less education on the birth certificate
than on the questionnaire when compared with
the 13.5 percent who reported more. In this
category a substantial proportion (29.6 percent)
Table C. Percent of cases reporting identical education of mother and percent reporting education of mother within the same education
group on the birth certificate and National Natality Survey mother’s questionnaire, by education of mother reported on the birth

certificate, 1972

 

Education of mother reported
on birth certificate

 

 

i f
Agreement of education of mother Total 0-8 9-11 12 13-15 | 16 years
years | years | years | years | or more
Cartificatn Same 8 QUESTIONINBING scsiirrimssrrmssiitsiusmistssesiniastmvia si ieemse sie 77.2 || 65.7 | 66.1 90.0 55.4 74.7
Within same education group on certificate and questionnaire ..........ccccceveeiinrniens 85.5 78.4 85.3 90.0 70.4 90.7

 

 

 

 

 

 

 

differed by 2 years or more. However, only a
small proportion of comparison cases (6 per-
cent) was included in the category 0-8 years.

These findings for comparability of report-
ing of parents’ education were better than in the
CPS-Census Match Study in which 65.2 percent
of the respondents reported the same amount of
education on both the census and CPS. How-
ever, they were lower than the results in the
New York State study: for mother’s and father’s
education, 95.2 percent and 94.5 percent, re-
spectively, had identical education reported on
both forms.

PLURALITY

Reporting of plurality on the birth certifi-
cate, as required by all registration areas, has a
response rate of nearly 100 percent. The long/
short hospital questionnaires were the only
survey questionnaires that requested this infor-
mation. Among comparison cases 99.8 percent
reported an identical number at birth on both
data sources. Of those births reported as single
births on the birth certificate, 99.9 percent were
also reported as single births on the HL/HS ques-
tionnaire with the remainder reported as twin
births (table 9). When a birth was classified as
twin on the birth certificate, it was also classi-
fied as twin on the questionnaire in 94.4 percent
of the cases. The remaining 5.6 percent were
reported as single births. Perhaps the disagree-
ment occurred in cases where only one twin was
live born and the existence of the second twin
was not transcribed from the hospital record to
the survey questionnaire. Both cases of triplets
or higher plurality were reported as such on the
questionnaire.

BIRTH WEIGHT

Reporting of birth weight was required by
all registration areas’ and included on the long/
short hospital questionnaires. Birth weight was
reported either in grams or pounds and ounces
on both forms, but all weights were converted to
grams for comparison.

As shown in table 10, there was no discrep-
ancy in the reported birth weight for 86.5
percent of the comparison cases. Only 2.4
percent differed by 250 grams (8.8 ounces) or
more. Because birth-weight data are usually
tabulated in 500-gram intervals, many of the
discrepant cases would be correctly classified in
the grouped data tabulations. There was agree-
ment within 500-gram intervals for 96.0 percent
of the births.

The level of exact agreement was lower
(77.3 percent) among births determined from
the birth certificate to be low birth weight
(2,500 grams or less) than among heavier weight
births (87.3 percent). When a discrepancy ap-
peared, births determined as low birth weight
from the birth certificate were more likely to
have a higher weight reported on the HL/HS
questionnaire than on the birth certificate,
whereas heavier weight babies were more likely
to have a lower weight reported on the ques-
tionnaire.

The New York State study found a similar
comparability—91.7 percent had identical birth
weight (within 1 once) reported both on the
birth certificate and on the hospital record.

LENGTH OF PREGNANCY

The length of pregnancy was determined
from the date the last menstrual period (LMP)
began as reported on both the birth certificate
and the long/short hospital questionnaires. In
1972, 39 States and the District of Columbia
included the required LMP date on their birth
certificates. The remaining States asked for
length of pregnancy in completed weeks. This
results in considerable heaping at 40 weeks,
probably because many babies weighing 6-9
pounds would be considered full term and were
reported as 40 weeks’ gestation.

Nonresponse to the questionnaire seemed to
be biased toward infants reported as premature
(less than 37 weeks’ gestation) on the birth
-certificate (table 11). Among births for which
data were provided on the birth certificate but
not on the questionnaire, 11.0 percent were
classified as premature, compared with 9.2
percent among cases for which gestation was
available from both sources.

Comparability of response could be deter-
mined for less than half of the survey cases. Of
these, 85.8 percent reported identical length of
gestation on both forms; an additional 6.8
percent differed by only 1 week.

Births determined to be premature from the
birth certificate were less likely to have identical
gestation reported on the questionnaire than
were those of longer gestation (74.8 percent
compared with 86.9 percent). There appeared to
be a bias toward the reporting of longer gesta-
tion on the questionnaire among those births
that were reported as premature on the birth

certificate with nearly 14 percent reporting a

gestation period 4 weeks or more longer on the
questionnaire. Among births with gestation re-
ported as 37 weeks or more on the birth
certificate, only 3.0 percent differed by 4 weeks
or more in either direction on the questionnaire.

MONTH OF PREGNANCY PRENATAL
CARE BEGAN

In 1972, 40 States and the District of Colum-
bia required reporting of the month of preg-
nancy that prenatal care began on the birth cer-
tificate. The information was obtained for the
birth certificate by querying either the mother
or the attending physician. The mother’s re-
sponse was most likely based on recall. Although

10

the physician has access to medical records con-
cerning the care that he provided, he might be
unaware of any additional care provided by
other physicians and could report care as begin-
ning later than it actually did.

A comparison can be made between this
information on the birth certificate with that on
the long hospital and physician’s questionnaires.
However, the questionnaire, unlike the birth
certificate, specifically asked that care reported
be limited to that provided by the attending
physician, causing discrepancies in cases where
more than one physician was consulted.

On the birth certificates where a comparison
could be made, 42.9 percent reported care
beginning in the same month of pregnancy both
on the questionnaire and the birth certificate
(table 12). Another 36.9 percent disagreed by
only 1 month. Discrepancies of 1 month often
do not affect classification by the trimester care
began, the form in which the data are generally
most useful, further minimizing disagreement.
Thus, the proportion showing agreement when
tabulated by trimester (including those reporting
no care) rather than by single month was in-
creased to 75.7 percent.

The rate of agreement among those report-
ing no care on the birth certificate was 72.7
percent. Among those reporting some prenatal
care, a wide variation in agreement existed by
the month that the care began as reported on
the birth certificate. The HL/P questionnaire
corroborated only 10.2 percent of the birth
certificates where care was reported to have
begun in the first month. However, a large
proportion of these cases (55.3 percent) re-
ported care beginning in the second rather than
the first month on the questionnaire, which is
still in the first trimester. Among the other
categories, agreement ranged from 28.9 to 53.2

© percent.

In cases where differing months were re-
ported, it was more likely that the birth certifi-
cate had care as beginning earlier than the
questionnaire did; 38.4 percent reported earlier
care on the birth certificate in comparison with
18.6 reporting earlier care on the questionnaire.
Thus it is tempting to surmise that many
discrepancies arise from cases where more than
one physician was consulted, and that these
discrepancies result from the inclusion of early
care by other physicians on the birth certificate.
However, for certificates reporting care begin-
ning in the third month or later, there was a
tendency, and in some cases quite large, toward
reporting an earlier date on the questionnaire
than on the birth certificate (see table 12),
which would refute this hypothesis.

The New York State study also compared
reporting of month prenatal care began and
found comparability to be better than the
present study (53.9 compared with 42.9
percent).

NUMBER OF PRENATAL VISITS

The birth certificate, the mother’s question-
naire, and the long hospital/physician question-
naire each contained an item on the number of
prenatal visits. Information on the birth certifi-
cate was probably provided by the mother or
the attending physician. The source of data for
the M questionnaire was the mother who was
queried several months after the birth of her
child. The questionnaire was structured to ascer-
tain additional sources of prenatal care and the
number of visits to these sources which may
have helped to elicit an accurate response. How-
ever, heaping at an even number of visits seems
to indicate that the response was often a guess.
Because data for the HL/P questionnaire were
obtained directly from the hospital’s or the
attending physician’s records, visits to additional
sources of prenatal care may not have been in-
cluded. Heaping was not as great among re-
sponses on the HL/P questionnaire.

Among cases where data were available from
both the birth certificate and the M question-
naire, only 15.6 percent reported an identical
number of visits on both sources (table 13).
Approximately 56 percent agreed within 2 visits;
however, nearly 24 percent were discrepant by 5
visits or more.

Classification of data by the number of
prenatal visits reported on the birth certificate
reveals an agreement rate ranging between 7 and
19 percent for all categories except the no
prenatal care category for which 66.7 percent
agreement was found; however, the number of

cases in this category was small. Among cate-
gories 1-2 through 11-12, there was a large bias
toward a lower entry on_the birth certificate
than on the mother’s questionnaire. The size of
this bias generally decreased as the number of
visits increased, from 76 percent for those
reporting 1-2 visits to 43 percent for those
reporting 11-12 visits. A large proportion of
these (between 22 and 59 percent) reported at
least three fewer visits on the birth certificate.

With the exception of the category 15-16
visits, agreement was lowest (approximately 8
percent) among those reporting 13 visits or more
on the birth certificate. Unlike those reporting
fewer prenatal visits, there was a bias toward a
greater number of visits reported on the birth
certificate than on the questionnaire.

As shown in table 13, there was a somewhat
greater consistency of response between the
birth certificate and the HL/P questionnaire.
The number of prenatal visits reported on the
HL/P questionnaire was identical to the number
reported on the birth certificate 25 percent of
the time and agreed within 2 visits 66 percent of
the time. Although the M questionnaire and the
birth certificate differed by 5 visits or more
almost 24 percent of the time, the HL/P
questionnaire differed from the birth certificate
by this amount less than 16 percent of the time.

Consistency of reporting between these two
documents was, for the most part, inversely
related to the number of visits as reported on
the birth certificate, decreasing from 88.2 per-
cent agreement at no visits to 2.9 percent at 19
visits or more. For most of these categories,
between 62 and 72 percent agreed within 2
visits. Among those reporting 10 visits or less,
there were more likely to be fewer visits
reported on the birth certificate than on the
HL/P questionnaire. However, this bias was not
nearly as great as that found in the comparison
of the birth certificate with the mother’s ques-
tionnaire, which may have been a result of the
mother’s knowledge of visits to other physicians.
Among those reporting 11 visits or more on the
birth certificate, a larger proportion reported
more visits on the birth certificate than on the
HL/P questionnaire. Although the number re-
porting no prenatal care on the birth certificate
was small, agreement was good (88.2 percent).

1
COMPARABILITY OF REPORTING
BY RACE OF CHILD AND AGE
AND EDUCATION OF MOTHER

To determine differences in comparability
among subgroups of the population, the percent
agreement of selected variables was computed
by race of child and by age and education of
mother as reported on the birth certificate.

As shown in table D, comparability was
generally better for white than for black births.
The only exception was the number of prenatal
visits (as reported on the long hospital/physician
questionnaire) for which the percent of cases
with exact agreement was found to be slightly
higher among black than among white births
(28.1 vs. 24.4 percent, respectively). However,
there was a higher proportion of white than of
black births for which reporting differed by no

more than two visits (66.1 and 63.3 percent,
respectively).

Only for the prenatal care items reported on
the HL/P questionnaire was there a relationship
between comparability of reporting and age of
mother (table E). Accuracy of reporting the
month of pregnancy that prenatal care began in-
creased with the age of the mother—from 37.0
percent for mothers who were under 20 years of
age to 47.1 percent for mothers 35 years of age
and older. The same relationship was found by
trimester that care began. The percent of cases
reporting prenatal care starting within the same
trimester ranged from 65.7 percent for the
youngest age group to 80.2 percent for the old-
est age group.

For the number of prenatal visits, a small
variation in comparability by age of mother was
found: the percent with exact agreement was

Table D. Percent agreement of selected items on the birth certificate and National Natality Survey questionnaires, by race of child
reported on the birth certificate, 1972

 

 

 

 

 

Race of child
reported on
: Question- 2 the birth
Selected item naires Total ceroticals
White | Black
Education of mother

INUIMDEY OF Cases OOIMIDBIE uvirusmims si amir a aR Tr eA STS ANE A NAR AAR SE Fr 08 Lem ERR Sere ad M 2,833 || 2,610 223
Percent agreement by:

SINIOIE VORES OF SUITION c.ccicsiirisineinsrunmrsmnronsiissssmarirs serra sia san ss ee ETT EAT Sa rae ER ESS Ee RRA M 76.4 77.4 65.0

Grouped years OF SOND)... cwuiiimiiiimmssvormmmisniimiisssiis saris sist ss sisastemirives M 85.6 86.3 77.6

Birth weight
NUMDbEr Of CASES COMPETE .......cccevviuuuuiirieirrrnrsiiereerrarsrarasssesseeessessssssssssesesssesssassssssssssnessessasssnne HL/HS 4,693 || 4,247 446
Percent agreement within 500-Gram interval ..........ccceeeiiiiuueuieeririereeeesiinsesesessssnsesessssssesssssnnsseeens HL/HS 96.0 96.6 90.6
Month of pregnancy prenatal care began

INUMDBEE OF CASES COINMBAIEN. ...cnvevinsassssernnnnunsnssssmsssinssnniosoosssmsivashussinsssssnsi sin siessss irs s smn rs E SHE HL/P 2,950 || 2,734 216
Percent agreement by:

SIMIC PIONTI wo errrrvrrrresvisrees sr ST IT ER EA SA TE SR SRST TAT A ST as tsa don suh dR mmr An E ORL HL/P 42.9 43.5 35.6

BPHIIBSIE. .. ce rrsiesoverinmernnesssmmsrssvesansnessis rena s es SEATS A SOAS SRS A TT TS TT STR r ea TS RRA HL/P 75.7 76.3 68.1

Number of prenatal visits

INUTRDEE OF GOSHS COMPAL . vx revunrerersnunrisrsrrreusnevsersrssrstrrestbesssars snk esmi isms mass Ses arr AT AAR RR HL/P 2,221 || 2,025 196
Percent agreement by:

SINQIB VISE covsvvvrismsinvsmmrsvermrsisssisssssssssrssasssssssssssassansensans ATR RTA A TE As ARR EA SARA SAREE AE HL/P 24.7 24.4 28.1

HE VIBES corriss senssmmmsmnsnsnenarsrvanstbocsses sya sc ase SERA A SE TT ETA EEE Ee RT EERO RET HARE rae HL/P 65.8 66.1 63.3

 

 

 

 

 

1M refers to mother’s questionnaire; HL/HS refers to long/short hospital questionnaire; HL/P refers to long hospital/physician

questionnaire.
2Total includes white and black races only.

12
Table E. Percent agreement of selected items on the birth certificate and National Natality Survey questionnaires, by age of mother
reported on the birth certificate, 1972

 

 

 

 

 

Age of mother reported
y on the birth certificate
Selected item Buesinn Total
Under 20 | 20-34 | 35 years
years years and over
Education of mother

Number of Cases COMPAred .........cceiiueereiririnneessniires sss essere assesses sessseasees 2,869 325 | 2,360 184
Percent agreement by:

* SINGIE YOars OF SCNOD)..ciuiisrnmsirmsinmrsimmmisiin sires srsienirs sien SETS M 76.2 74.8 76.6 73.4

Grouped years Of SCHOO «uuuimisivisrmsisssnamssssien rss sats asst si III 48 TIRE RY M 85.5 85.8 85.7 83.2

Birth weight
INUMDET Of CASESICOMPBIBL .ovorurisrisinaricivrrmmmeniussinse ii rs SAREE SEETHER SRI SOR HL/HS | 4,769 687 | 3,794 288
Percent agreement within 500-gram interval .........cccccvevrireereeenienieesssessssssnsssnsesnen HL/HS 96.0 95.5 96.1 95.1
Month of pregnancy prenatal care began

INUIMDEr OF C583 COMPBIET ....vvivivsssinismmrissasssnisstsnsnrrrerssrssrensuassvsnensssetis st sas SESSA ARS HL/P | 3,002 432 | 2,398 172
Percent agreement by:

Single month HL/P 42.9 37.0 43.7 47.1

TT EIRIBBLBY. ..ccxunemesnernmmrsuanaiebbsssiissnssssa ines Tr SEA ms TAR ERAT SS EAA SS SAA ARSE SAS 18H HL/P 75.8 65.7 77.3 80.2

Number of prenatal visits

NUMDbBr OF CaSES COIMPATBH cicciiinrirmsrmrismisinss rsa ress TITS TRY HL/P | 2,258 313 | 1,814 131
Percent agreement by:

SINGS VISI. sce rrssesnermmsnesssnmmarsmsrsssnas iris ss STB A TE EAE ARI EEE ITH 3 ear a key HL/P 249 28.4 24.0 29.0

EDVIEIE. oo vee diennress rT STRAT ATE ARR ss ER TAS OHSS AR RRR ams SRSA RA TAR SAAT ATE HL/P 65.9 64.5 65.5 73.3

 

 

 

 

 

 

1M refers to mother’s questionnaire; HL/HS refers to long/short hospital questionnaire; HL/P refers to long hospital/physician

questionnaire.

lower among mothers 20-34 years of age (24.0
percent) than among both younger and older
mothers (28.4 and 29.0 percent, respectively).
Agreement within 2 visits showed a much larger
age differential that increased with age from
64.5 percent among mothers under 20 years of
age to 73.3 percent among mothers 35 years of
age or older.

Table F shows that there was only a slight
variation in percent agreement by educational
attainment of mother. For birth weight, com-
parability of reporting between the birth certifi-
cate and the long/short hospital questionnaire
increased slightly with education, ranging from
95.1 percent among mothers with less than 12
years of schooling to 96.8 percent among those
with 13 years or more.

The prenatal care items showed more varia-
tion by education than did birth weight, but no

consistent pattern was found. For the month of
pregnancy prenatal care began, mothers with the
most education had a slightly higher percent of
agreement than mothers with less education (ap-
proximately 46 percent vs. slightly over 43 per-
cent, respectively). However, education had a
considerable effect on agreement rates by tri-
mester, with the magnitude of agreement in-
creasing as education increased. Agreement rates
by trimester ranged from 70.5 percent for
mothers with less than 12 years of education to
81.4 percent for mothers with 13 years or more
of education.

Exact agreement for number of prenatal vis-
its ranged from 23.8 percent for those with 12
years of education to 27.4 percent for those
with less than 12 years of education. Agreement
within two visits showed a very slight inverse re-
lationship to education.

13
Table F. Percent agreement of selected items on the birth certificate and National Natality Survey questionnaires, by education of
mother reported on the birth certificate, 1972

 

 

 

 

 

 

 

Education of mother reported
g on the birth certificate
Selected item Question Total
naire Less than 13 years
oss 12 years Y
12 years or more
Birth weight
Number Of Cases COMPATEQ...........cocueieiueeeiiereeinreiinesiseeessssesssssersseesssssssssesssnes HL/HS | 3,275 853 1,617 805
Percent agreement within 500-ram interval ...........ccccveeeeriirneeesesenesesrsseeseseennns HL/HS 96.0 95.1 96.1 96.8
‘Month of pregnancy prenatal care began
Number Of Cases COMPArEd............ceeeeeeeeereiiierissisressseessesesssseeeesemsessssssssssssssssns HL/P | 2,291 562 1,138 591
Percent agreement by:
Single month HL/P 43.9 43.4 43.1 45.9
THANIBELIBE cov tvmensmnamsssvs spins tonsmvsssstns iiss sesnesmsssssrensrsmsesnsv ars Kevenssss sss amsramvens HL/P 76.6 70.5 71.2 81.4
Number of prenatal visits
Number of cases cComPared.............cccceerreereiirinninenrseesssseesssss esse sssssssssssssssssens HL/P | 2,135 552 1,018 565
Percent agreement by:
SINQIBVISIT sisivsesrmmrisiorinssmmmissssssiamspessssssasrasssestintnsns sR ERo na HR AOTS SHA ASSITRAAR ENS HL/P 249 27.4 23.8 24.4
TD WASHER. . ciinromssm torsmsss amis asa Er SAI SEAT TS sis aaa a srs Si bnnsa ties snes ss ass saws sbe HL/P 65.7 66.5 66.1 64.1

 

 

 

 

 

 

1HL/HS refers to long/short hospital questionnaire; HL/P refers to long hospital /physician questionnaire.

REFERENCES

1U.S. Bureau of the Census: Test of birth registration
completeness 1964 to 1968. Census of Population and
Housing, 1970, Evaluation and Research Program. Series
PHC(E)2. Washington. U.S. Government Printing Office,
Mar. 1973.

2National Office of Vital Statistics: Evaluation of
obstetric and related data recorded on vital records and
hospital records: District of Columbia, 1952, by E. Op-
penheimer et al. Vital Statistics-Special Reports, Vol. 45,
Nos. 1-14. Public Health Service. Washington, D.C.,
Nov. 20, 1957. pp. 363-416.

3Montgomery, T. A., Lewis, A., and Hammes, L.:
Live birth certificates—evaluation of medical and health
data in California. California Med. 96(3):190-195,
Mar. 1962. :

4Lilienfeld, A. M., et al.: Accuracy of supplemental
medical information on birth certificates. Pub. Health
Rep. 66(7):191-198, Feb. 16, 1951.

14

5Carucci, P.: Reliability of statistical and medical
information reported on birth and death certificates.
New York State Department of Health Monograph No.
15. New York State Department of Health, Albany,
N.Y., May 1979.

6U.S. Bureau of the Census: Evaluation and research
program—1970 Census of Population and Housing.
Accuracy of Data for Selected Population Character-
istics as Measured by the 1970 CPS-Census Match. Series
PHC(E)-11. Washington. U.S. Government Printing
Office, Jan. 1975.

7U.S. Bureau of the Census. Evaluation and research
program—1970 Census of Population and Housing.
Accuracy of Data for Selected Population Character-
istics as Measured by Reinterviews. Series PHC(E)-9.
Washington. U.S. Government Printing Office, Aug.
1974.
10.

11.

12.

13.

LIST OF DETAILED TABLES

Comparability of reporting age of mother between the birth certificate and the National Natality Survey questionnaires, by
age of mother reported on birth certificate, 1972 .......cccericenninnnimnininnnsninnesessssnsemn. AeA EAT SRA SAAT

Comparability of reporting age of father between the birth certificate and the National Natality Survey mother’s question-
naire, by age of father reported on Dirth Certificate, 1972 ...causminmtisiminmissmrism semicon mirsirrsiiissirsern

Comparability of reporting number of previous children born alive and still living between the birth certificate and the Na-
tional Natality Survey questionnaires, by number of previous children born alive and still living reported on birth certificate,
O72 cr rreeuratsernenisnvenmeits sos ER EET Ee ELE SE I EE AT ETT A EEA er EE EE SAE AAA HENS tS ATTRA REA URES LET STATE ORIEN

Comparability of reporting number of previous children born alive and now dead between the birth certificate and the Na-
tional Natality Survey questionnaires, by number of previous children born alive and now dead reported on birth certificate,
1 IIT eI DR ho Ee Er

Comparability of reporting live-birth order between the birth certificate and the National Natality Survey questionnaires, by
live-birth order reported on birth Cortificate, TOTZ wcrc simmemmin sss ris seers sist sess sa IA Er IAA SARA CRRA RET SERS

Comparability of reporting number of fetal deaths between the birth certificate and the National Natality Survey question-
naires, by number of fetal deaths reported on birth certificate, 1972 ....uraisssisimissirmssisirmiimnis semanas nies

Comparability of reporting education of mother between the birth certificate and the National Natality Survey mother’s
questionnaire, by education of mother reported on birth certificate, 1972 .........ccccumiinnniniimieii sean

Comparability of reporting education of father between the birth certificate and the National Natality Survey mother's
questionnaire, by education of father reported on birth certificate, 1972 ..........cccvminrinininiiiiinnn en

Comparability of reporting plurality between the birth certificate and the National Natality Survey long/short hospital ques-
tionnaire, by plurality reported on birth certificate, 1972 ....cismierrsisesrseisssiermssmssarseissisissessssi iets sossssvessgesssrsturisrivesssnss sain

Comparability of reporting birth weight between the birth certificate and the National Natality Survey long/short hospital
questionnaire, by birth weight reported on birth certificate, 1972 ........cccvnniniiiniiminni i ee,

Comparability of reporting length of pregnancy between the birth certificate and the National Natality Survey long/short
hospital questionnaire, by length of pregnancy reported on birth certificate, 1972 .........cccvviuiiininininniiinnnn.

Comparability of reporting month of pregnancy prenatal care began between the birth certificate and the National Natality
Survey long hospital/physician questionnaire, by month of pregnancy prenatal care began reported on birth certificate,
12 7 1 SN ON CO SRNL NL URC TO NN OL TERE Nur SOV VN 0 DOI CIE DOAN M0 TO MY oy cc

Comparability of reporting number of prenatal visits between the birth certificate and the National Namality Survey question-
naires, by number of prenatal visits reported on birth certificate, 1972 ..ciuwmmssiisssssissrsmsisrsserisssvtarssesmrssessisssisens

16

17

18

19

21

, 22

23

24

24

25

25

15
Table 1. Comparability of reporting age of mother between the birth certificate and the National Natality Survey questionnaires, by age
of mother reported on birth certificate, 1972

 

 

 

 

 

 

 

 

 

 

 
 

 

 

 

 

 

 

 

 

 

 

 

 

 

Age of mother reported on
All birth certificate
Comparison of certificate with questionnaires COFURCHISS
Under 20-24 | 25-29 | 30-34 | 35 years No
20 years | years years years | and over | response
Number
All SaMPIe CASES ...coureireeiernerirneerineeeeiaeeersaeessnnes 5,689 I 831 2,136 1,680 691 346 5
Mother's questionnaire Percent
C388 EXCINABM.....ccnsrsirrsrssinsiimnississsssrarinrssiossersnsesansrssanmrene 29.8 I 44.3 | 30.4 24.3 | 26.5 23.1 100.0
Number
CB508 COMPANET..cccvimevmsmirmrsssmrsssssssssrsssissserssemsinsrssnsinses 3,996 I 463 | 1,487 | 1.272 508 266 -
Percent distribution
CASS COMPArBT. .ccsuirissssirrmrivsssinrimssnsssmorssssisnsosiitinasessssiuass 100.0 | 100.0 | 100.0 | 100.0 | 100.0 100.0 .
RR ESS. =
Certificate less than questionnaire:
2 years or more 1.4 2.2 1.4 1.3 0.6 15
5.2 4.5 5.6 4.3 6.9 4.9
Certificate same as qUESTIONNAITe ........c.ccceeeeerernveeerennns 90.7 91.6 90.8 91.7 89.4 86.5
Certificate greater than questionnaire:
V WBE ciussricssnsrsssssninimsssssnmressasisiisos sms eissesta cis ars sass 1.6 1.1 1.7 1.3 1.8 2.6
2 VRAIS OF TOT .0vvirsorsvrsmsssnrinsssnanstrsassssstiassesssrsianssss 1.1 0.6 0.5 1.3 1.4 4.5
Long/short hospital questionnaire Percent
Ca30s eXCICB..civrerivssstrrmisesssarrsssses UN "16.0 I 17.7 16.0 14.6 16.8 15.9 100.0
Number
Cases COMPAred.........ccervreereecrnnerresrosenneesssssansessssnesssssssnnns 4,779 I 684 1,794 1,435 575 291 | -
Percent distribution
Cases COMPAIt..c.uursissimnimrsrinsssrinssrissraistmnsrssnion we 100.0 100.0 | 100.0 | 100.0 | 100.0 100.0 z
Certificate less than questionnaire:
2 YOOBIS OF TIIOIC covsscssnssssrmsssinssssirmsssnsissionssinsesnnarsaninn 1.4 0.6 1.4 1.7 2.1 0.3
V WOOL rrinirrrimsiss ris as A RASTER EI SATS S FRSA RRR 29 3.7 23 3.1 31 3.8
Certificate same as qUESTIONNAITe ..........ccceerrricnnnreresssnne 87.8 93.6 88.2 85.6 87.0 83.2
Certificate greater than questionnaire:
6.4 1.9 7.0 7.5 5.9 8.9
1.8 0.3 19 2.1 19 3.8

 

 

 

 

 

 

 

 

 

16
Table 2. Comparability of reporting age of father between the birth certificate and the National Natality Survey mother’s ques-
tionnaire, by age of father reported on birth certificate, 1972

 

w |

Age of father reported on birth certificate

 

 

 

 

 

 

 

 

 

 

 
 

 

 

 

Comparison of certificate with
mother’s questionnaire certificates Under 20-24 | 25-29 | 30-34 | 35-39 | 40 years No
20 years | years | years | years | years | and over | response
Number
All Sample CaSes.......uuvurerrrrerernnranencnenes 5,689 I 278 | 1,681 1,926 | 990 | 451 | 338 | 25
Percent
00388 BRCINTB cecrrsrrsesssrisssssinrmnssssnnsrisrsassensrnsass 324 | 48.6 | 38.6 28.3 | 25.4 31.0 | 29.9 | 100.0
Number
Ca30S COMPBrE coccerrerinrarsmrnissasssssrssssssrsssseiasionsone 3,843 I 143 | 1,032 | 1,381 739 | 31 | 237 | -
Percent distribution
CASAS COMPBIOU ..covesssvinrvririssnrrrsissrssssrssssisssssasniag 100.0 100.0 | 100.0 | 100.0 | 100.0 | 100.0 100.0
Certificate less than questionnaire:
2 YOOIS OF MNIOIBucirecrrrrsrrinsrsrsarmsssirsminasasnsstn 1.9 1.4 2.3 1.4 1.9 2.6 2.5
Y WOT rarsrsisrssrsssnsa sat EEA AVES HERES TARAS 7.2 11.9 6.7 6.0 8.9 7.7 7.6
Certificate same as questionnaire ..........cceeeuus 84.5 81.8 86.7 86.9 82.8 78.8 75.9
Certificate greater than questionnaire:
4.6 4.2 3.5 4.3 4.6 71 8.0
2 years or more..... 1.8 0.7 0.8 1.4 1.8 3.9 5.9

 

 

 

 

 

 

 

 

17
Table 3. Comparability of reporting number of previous children born alive and still iiving between the birth certificate and the National Natality Survey
questionnaires, by number of previous children born alive and still living reported on birth certificate, 1972

 

All

Number of previous children born alive and still living

reported on birth certificate

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 
 

 

 

 

 

 

 

Comparison of certificate
with questionnaires certificates 8 N
None 1 2 3 4 5 6 7 of. 0
more | response
Number
All sample cases..... 5,689 I 2,110 | 1,746 | 845 | 425 184 | 100 | n | 32 | 34 | 142
Mother's questionnaire Percent
C308 eXCINURD ....covciissnsirssnsmmmmmrassmrmisie 304 Il 30.2 | 26.1 | 25.2 325 | 29.3 | 40.0 | 324 | 46.9 | 356.3 | 100.0
Number
Cases compared............ccvcrvnrnmnnnenmmiesenno 3,960 Il 1,473 | 1,291 | 632 | 287 | 130 | 60 48 17 | 22
Percent distribution
Cases COMParetl.. qu innsumiinnsiiamens 100.0 || 100.0 | 100.0 | 100.0 | 100.0 | 100.0 | 100.0 | 100.0 | 100.0 | 100.0 |
Certificate less than questionnaire:
3 children or more 0.1 0.2 0.1 - - 0.8 - - - -
0.2 0.2 0.1 0.2 0.3 - - . . .
1.0 1.4 0.9 0.3 0.8 33 4.2 - -
Certificate same as questionnaire .... 97.0 98.2 979 | 954 96.5 96.2 91.7 85.4 82.4 77.3
Certificate greater than questionnaire:
1.3 1.0 3.5 21 1.5 5.0 6.3 5.9 9.1
0.2 0.6 1.0 - - 21 - -
0.2 - 0.8 - 2.1 11.8 13.6
Long/short hospital questionnaire : Percent
C8883 XOINCEN .ovvoeiirsvmrsvirscinmosrissimisstiiinemins 20.4 I 11.3 | 18.6 | 18.1 | 18.8 23.4 | 20.0 | 19.7 28.1 | 32.4 | 100.0
Number
Cases compared............eucinmmermersnnssnnssisinsssniis 4,527 ll 1,745 1,421 | 692 345 | 141 80 57 23 | 23 | -
Percent distribution
C903 COMPATE... coves cimsinivisssrmsmmnsssiniernin 100.0 || 100.0 | 100.0 | 100.0 | 100.0 | 100.0 | 100.0 | 100.0 | 100.0 | 100.0
Certificate less than questionnaire:
3 children or more 0.1 0.2 0.1 0.1 - . - - - -
2 children 0.3 0.6 0.1 0.3 0.3 - - - - -
1 child 4.5 24 7.2 4.6 3.8 4.3 3.8 3.5 13.0 8.7
Certificate same as questionnaire .... 92.1 96.8 89.4 91.2 88.7 88.7 85.0 84.2 65.2 78.3
Certificate greater than questionnaire:
23 3.2 3.0 4.1 5.0 10.0 105 | 13.0 4.3
0.2 Fr 0.7 1.7 - - - - .
0.3 poate 1.4 2.1 1.3 1.8 8.7 8.7

 

 

 

 

 

 

 

 

 

 

 

 

 

 

18
Table 4. Comparability of reporting number of previous children born alive and now dead between the birth certificate and the
National Natality Survey questionnaires, by number of previous children born alive and now dead reported on birth certificate, 1972

 

 

 

 

 

 

 

 

 

 

 

 

 

Number of previous children born alive and
now dead reported on birth certificate
Comparison of certificate with questionnaires a 1 or more
None 2 response
Total | 1 SITIO
Number
A) SAMNPIG CBIBS..ccvssstvsvsrrressssrserrsrrrsssvssmsssssssssssssssssasnsssnseries 5,689 I 5,285 | 200 fl 176 24 204
Mother's questionnaire Percent
£as5 @XCIUARD ......uuuureereeieiririieieeieeeeeeeeeree cess sssessssssssssasesssssessessssnssesesenn 30.9 I 28.2 | 31.0 i 29.5 | 41.7 100.0
Number
CASES COMPAIE......cececeee eevee sess ssssssss senses ssssssssssssseseresss assesses 3032 ll 3708 | 138 ll 124 14 | .
Percent distribution
CASES CBIVIDATEN.... crx 00:1 svnssserstnsas a FATA HER FEA ERS REE TCR RI Ses Bp R ars THs ea eens 100.0 II 100.0 | 100.0 || 100.0 | 100.0 |
Certificate less than QUEStIONNAITE ..........cccceiriinrrrrnrenieneeeeeeeereeineenas 0.9 0.9 0.7 0.8 -
Certificate Same as QUESTIONNAIT. .......cccuuuererrrrerueeneerereeeseeeeeeeanaeanns 98.3 99.1 775 79.8 57.1
Certificate greater than QUEeStiIONNAIre........c.cceeivvreuuiiireieeeernineiireneens 0.8 21.7 19.4 42.9
Long/short hospital questionnaire Percent
CASBSONBINTIOY vrimsrssssromsnssensinnssnessebosvssssbelssssss sib entra a sade AVS ITERATE 21.4 I 18.3 | 21.5 ll 21.6 20.8 100.0
Number
CHSBE COTHPBIET i... cow renmrnrsrsunuussnrsnimsnbirionssosssrsmmmenstssiiiarmensnsint sossonsonnnnves 4,473 I 4,316 | 157 I 138 | 19 =
Percent distribution
yy I 100.0] 100.0 | 100.0 || 100.0 100.0
Certificate less than QUESTIONNAITE..........c ieee issrsmismasssssnirsenrnie 1.0 1.0 0.6 - 5.3
Certificate same as QUESTIONNAINE........cccceeeieiiireeerererssresrnsensssssseseenns 97.5 99.0 87.3 58.0 52.6
Certificate greater than QUESTIONNAITE..........ccceeerririirrneninnereneeneennennes 1.5 42.0 42.0 42.1

 

 

 

 

 

19
Table 5.

reported on birth certificate, 1972

Comparability of reporting live-birth order between the birth certificate and the ational Natality Survey questionnaires, by live-birth order

 

 

Live-birth order reported on birth certificate

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Comparison of certificate All
with questionnaires certificates 1st 2d 3d 4th 5th 6th 7th 8th or No
higher | response
Number
All sample Cases .............ccccuvevriveeriinnns 5,689 I 2,054 1,707 850 427 | 197 | 98 75 | 76 205
Mother's questionnaire Percent
Cases excluded... 31.1 I 30.1 | 26.1 | 24.4 | 326 31.0 38.8 34.7 | 40.8 | 100.0
Number
Cases COMPAred..........coceeveeeeeeeeeieiiiiceeeeee eens 3,918 I 1,435 | 1,262 | 643 288 | 136 | 60 | 49 | 45 |
Percent distribution
Cases COMPARED. .........curissnsinisssinsstissnrommassmnnsanens 100.0 || 100.0 | 100.0 | 100.0 | 100.0 | 100.0 | 100.0 | 100.0 | 100.0
Certificate less than questionnaire:
2 births or more 0.3 0.3 0.2 0.2 0.3 1.5 1.2 - -
1.3 1.8 0.7 0.9 0.7 15 5.0 2.0 4.4
Certificate same as questionnaire ................... 96.1 97.9 98.0 23.3 94.8 91.9 83.3 87.8 73.3
Certificate greater than questionnaire:
1.8 1.0 5.0 2.8 3.7 10.0 6.1 8.9
0.5 0.6 1.4 1.56 - 4.1 13.3
Long/short hospital questionnaire Percent
Cases eXCIUAR..........conuninmmisimmmmimsstrmmsesrmsssverrsns 21.4 I 172.2 | 18.9 176 | 19.7 | 23.9 | 20.4 18.7 | 28.9 | 100.0
Number
Cases COMPAred........coccrvvinssisivivisissssvsssimmsmmmnrmasers 4,471 ll 1,701 | 1,384 700 343 150 | 78 61 | 54
Percent distribution
Cases compared... 100.0 || 100.0 | 100.0 | 100.0 | 100.0 | 100.0 | 100.0 | 100.0 100.0
Certificate less than questionnaire:
2 DIRS OT IMO ..crvrrssssvnmsarsesssssranion 0.7 0.7 0.5 0.4 0.9 0.7 1.3 - 1.9
1 DI, cise 4.7 25 73 53 2.9 4.7 26 6.6 9.3
Certificate same as questionnaire ................... 91.1 96.8 88.9 89.6 85.7 86.0 82.1 78.7 63.0
Certificate greater than questionnaire:
3.0 3.3 3.9 7.0 6.7 11.5 13.1 14.8
2 births or more. 0.7 0.9 35 2.0 2.6 1.6 1.1

 

 

 

 

 

 

 

 

 

 

 

 

20

 
Table 6. Comparability of reporting number of fetal deaths between the birth certificate and the National Natality Survey
questionnaires, by number of fetal deaths reported on birth certificate, 1972

 

 

Number of fetal deaths reported
on birth certificate

 

 

 

 

 

 

 

 

 
 

 

 

Comparison of certificate with questionnaires All
certificates N
None 1 2 Sor 5
more | response
Number
AN SAINIDIGTICESBE . ..cvinnssvsrirnsrssssnisrrsssssttrssssssss bisa sms dt EEm TIS RRR REISS 5,689 I 4,804 | 440 | 89 | 34 | 322
Mother's questionnaire Percent
CaS EXOIUNAR coocinrursssseimssmrinsnisrossissiintsssannunssss SEERA FEARLESS ASTER 32.9 I 29.2 | 25.9 | 23.6 | 41.2 | 100.0
Number
CASES COMPBIEU covvsvrerssonvsssuvmnsnsssosssssssssinss is ans sh sess rseuinsves sass simsinssussssasnns 3,817 ll 3,403 | 326 68 | 20 | -
Percent distribution
C585 COMPABTEU .vriurrerrarremermsrisissirsmmitrarsississirsirssmssisssmsresssesmsnessassrsen 100.0 Il 100.0 | 100.0 | 100.0 | 100.0 |
Certificate less than questionnaire:
2 fetal deaths or more 2.6 2 1.8 1.6 5.0
1 fetal death 7 7.1 7.1 5.9 5.0
Certificate same as QUESTIONNAITE ........uuueeiereeerreeruniierseeeesssnnsnssressesssnns 89.2 90.2 80.1 86.8 75.0
Certificate greater than questionnaire:
TY £818) OBER ovina Tiina ssa Sr ss sR en ARR Cnn RR RAS 1.0 11.0 29 -
2 F210) COIS OF INOIC..cricecsssrisrsrmrarrsivsssssssanrissssnasnssisss sss bORaeniNeTIIAS 0.1 29 15.0
Long/short hospital questionnaire Percent
C0388 OROINRU. coosinsrsssvisnsssssssrsimmsnienntnsnssniasssnese casts asta AEs ARIS EH ITEIIRREERVRSRS 21.9 I 17.6 14.3 11.2 14.7 100.0
Number
CHSRS COMNPAIEL .cvivvmmmnssismmssivmvprsonssnisasiisssmmpensssstsss sss ran ta TIS STII CPO UHERY 4,443 I 3,958 | 377 79 | 29 -
Percent distribution
Cases COMPAred civ sisssressssssssenss ATR RAEI SAR HY 100.0 || 100.0 | 100.0 | 100.0 | 100.0
Certificate less than questionnaire:
2 1010] deaths OF INOID.....cvvsisimsimsaisss tr itriris asses sirasatssss sat sssss iss 22 2.2 29 1.3 -
} FOUR) QBREN wovivinraisns sn rr ST EA RE EI A TS AA ARR RTT RE RR cw ome 7.3 7.5 5.6 5.1 6.9
Certificate same as QUESHIONNAING ....cccvsrarsesrrsssssisinmssssstssssssasstonrsssasn 88.7 90.3 77.2 73.4 55.2
Certificate greater than questionnaire:
Total doth .o.oivismmsmnsiisnssessammnsssnssthbasss sas ies 1.5 14.3 11.4 6.9
2 fetal deaths or more 0.4 8.9 31.0

 

 

 

 

 

 

 

 

21
Table 7. Comparability of reporting education of mother between the birth certificate and the National Natality Survey mother’s
questionnaire, by education of mother reported on birth certificate, 1972

 

 

 

 

 

 

 

 

 

 

  
 

 
  

 

Education of mother reported on birth certificate Not on
Comparison of certificate with All ey
mother’s questionnaire certificates 0-8 9-11 12 13-15 | 16 years No cote
years | years | years | years | or more | response
Number
All SaMPIO CASES .ocviivmriimrorssersivsssssnsnians 5,689 I 245 | 808 | 1,879 | 558 | 397 | 69 } 1.733
Percent
Cases excluded ...........o.oveeeeeeeeeeieeee seer ereeesseeeenns 49.5 i 45.3 | 37.9 243 | 15.9 13.4 | 100.0 100.0
Number
Cases COMPAred.........ccecuveeeeeeereeerersrssssssssssssssseseseees 2,871 I 134 | 502 1,422 469 | 344 - | -
Percent distribution
Ca388 COMPArEU....coichemmmissinimmimsnitssssissnanssssionnnsronse 100.0 | 100.0 | 100.0 100.0
Certificate less than questionnaire:
3 years or more... 1.4 24 0.4 -
2 years.......... 1.8 1.1 1.7 -
1 YBB,. cou cisnmmroniniins sss sans aes Aaa RT 14.3 23 7.5 9.3
Certificate same as questionnaire ..........cc......... 66.1 90.0 55.4 74.7
Certificate greater than questionnaire:
1 year... 13.7 24 26.0 8.1
2 Years .c.covniss a 1.6 1.0 4.7 2.6
3 YEAS OF MOF .....ceevrrrrenererereieerreeeenensnsenes 1.0 1.1 4.3 5.2

 

 

 

 

 

 

 

 

 

1A large proportion of the excluded cases occurred in the 11 States and the District of Columbia that did not include the item on the

birth certificate. See table II in appendix I.

22
Table 8. Comparability of reporting education of father between the birth certificate and the National Natality Survey mother’s ques-
tionnaire, by education of father reported on birth certificate, 1972

 

Education of father reported on birth certificate

 

 

 

 

 

 

 

 

 

 

 

 

: 'e . Not on
Comparison of certificate with All cortifi-
mother’s questionnaire certificates 0-8 9-11 12 13-15 | 16 years No Cate
years | years | years | years | or more | response
Number
All SAMPI@ CASES ......vveervrerrrrrerrnnesrnnenes 5,689 I 299 625 | 1,687 606 | 662 87 | 1,733
Percent
Cases excluded] ........cvevveueirieeiniseieseeenisseens senses 50.0 I 43.1 | 41.9 25.1 | 20.0 | 14.0 | 100.0 | 100.0
Number
Ca383 COMPAIBL...coscsrrirrinsrisesnersrrrsssserersessssnssssseses 2,842 I 170 | 363 1,263 | 485 561 -
Percent distribution
Cases COMPAred.........cocuvieierrerreresininnnneesninsenseesnsnes 100.0 | 100.0 | 100.0 100.0
Certificate less than questionnaire:
3 Y@ars OF MOE ....ccceeciiurnreeiiirnneeeeiniineneenins 2.8 26 0.8 -
4.1 1.7 3.1 -
YY OBY sisi AS 16.2 3.4 14.4 10.7
Certificate same as questionnaire ...........cccoueees 55.1 86.8 53.6 74.7
Certificate greater than questionnaire:
NYO rrsetissirirernmneteitinseverrnssepnneseneroresd 15.4 3.1 19.2 10.0
5.5 1.6 6.0 1.6
1.9 1.7 2.9 3.0

 

 

 

 

 

 

 

 

 

 

1A large proportion of the excluded cases occurred in the 11 States and the District of Columbia that did not include the item on the

birth certificate. See table II in appendix I.

23
Table 9. Comparability of reporting plurality between the birth certificate and the National Natality Survey long/short hospital ques-
tionnaire, by plurality reported on birth certificate, 1972

 

 

 

 

 

 

 

 

 

  
 

Plurality reported on birth certificate
" ’ n All
P
lurality reported on the questionnaire Ed — ] Triplet No
Single | Twin ;
or higher | response
Number
Al} SBINPIS CBIBB..rassssssrisinmsserssssinssssirssrsisesssissssteossrarss sess aREO IRE EIS IEITS 5,689 Il 5,576 108 | 2 | 3
Percent
C2383 EXCITE. ..ccrmnsnnrirrrmssssnssssnsssinssssnnissnnnssstssssseses ssn snssssve sa ens sensssss poms ss nesists 15.8 I 15.8 | 16.7 - | 100.0
Number
Cases compared ............. RT Ae IIH A EATS S VAR ASA S ARES TSAR ERRATA RAIA 4,789 I 4,697 90 | 2 -
Percent distribution
CBS2S COMMIPABL cir sss sas Es FAs soa TTR Soa eI Sasa ras sro SRmenahs sae namenamnsmssA esas 100.0 Il 100.0 | 100.0 100.0 THs
CT 98.1 99.9 5.6 - 5a
TWilcsermaeens 1.9 0.1 94.4 - wor
Triplet or higher .........ccocvenenenes 0.0 - - 100.0 we

 

 

 

 

 

Table 10. Comparability of reporting birth weight between the birth certificate and the National Natality Survey long/short hospital questionnaire, by birth weight
reported on birth certificate, 1972

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Birth weight reported on birth certificate
i ae i All
Comparison of certificate with long/ Sg 5,001
; : v certifi- 501- 1,001- | 1,501- | 2,001- | 2,601- | 3,001- | 3,501- | 4,001- | 4,501- :
Shor: hoshiBl questionnaire cates || 9590 | 1.000 | 1.500 | 2,000 | 2,500 | 3.000 | 3,500 | 4,000 | 4,500 | 5,000 | 92ms | No
grams | grams | grams | grams | grams | grams | grams | grams | grams | grams ho Yssponse
Number
All sample cases..........courevinrens 5,689 I k} | 23 | 26 | 93 | 251 | 982 | 2,204 1,491 | 486 | 87 | 16 | 27
Percent
Cases excluded on marinsmninimimiis 16.5 I - 8.7 | 26.9 16.1 11.6 18.7 16.6 15.5 12.3 16.1 25.0 | 100.0
Number

426 | 73 | 12 -

Cases compared. 4,750 I 3 21 19 78 | 222 798 1,838 | 1,260

 

Percent distribution

Cases COMPAFEL.......ouinnninsininis 100.0 || 100.0 | 100.0 { 100.0 | 100.0] 100.0 | 100.0 | 100.0 | 100.0 | 100.0 | 100.0 | 100.0

 

Certificate less than questionnaire:

 

 

500 grams or more 0.5 33.3 - - 3.8 2.7 0.5 0.4 0.2 0.2 - -
250-499 grams. 0.8 - 4.8 - 2.6 1.8 0.5 1.2 0.6 - - -
1-249 grams... 3.8 - 9.5 5.3 12.8 6.3 4.6 3.8 29 2.8 14 -

 

Certificate same as questionnaire. 86.5 33.3 76.2 78.9 71.8 79.7 86.6 86.9 87.5 89.4 87.7 83.3

 

Certificate greater than questionnaire:

 

 
 

 

 

 

 

 

 

 

 

 

 

 

 

1-249 grams... 71 33.3 9.5 15.8 17 9.5 73 7.0 71 5.4 5.5 8.3
250-499 grams. 0.5 - - - 1.3 - 0.3 0.4 0.7 0.9 1.4 -
500 grams or more ... 0.6 hen - - - - 0.3 0.3 1.1 1.2 4.1 8.3

 

 

 

24
Table 11. Comparability of reporting length of pregnancy between the birth certificate and the National Natality Survey long/short hospital questionnaire, by
length of pregnancy reported on birth certificate, 1972

 

 

 

 

 

 

 

 

 

 

 

 

 

 

  

 

 

Length of pregnancy reported on birth certificate
i - ; All Not on
Comparison of certificate with long/ certifi- || Under certifi-
short hospital questionnaire 20 20-27 | 28-31 32-35 36 37-39 40 41-42 | 43-44 No a
cates weoks weeks | weeks | weeks | weeks | weeks | weeks | weeks | weeks | response ca
Number
All sample Cases.........ccceererunnnna 5,689 I 1 1 39 163 113 1,299 790 815 178 | 974 | 1,306
Percent
C828 OROIIBH oon ess nriseriirivrsiscmunisnissrins 53.0 [ - 9.1 | 28.2 | 29.4 | 18.6 224 | 20.5 20.0 | 21.4 100.0 | 100.0
Number
Cass COMPBred ou rmmnivisimisammmses 2,674 ll 1 | 10 | 28 | 115 92 | 1,008 628 | 652 140 - |
Percent distribution
Cases compared... cuuiicinmrivsinsrermnn 100.0 |] 100.0 | 100.0 | 100.0 | 100.0 | 100.0 | 100.0 | 100.0 | 100.0 | 100.0
Certificate less than questionnaire:
5 weeks or more. 1.0 100.0 10.0 71 9.6 5.4 0.6 - - -
4 weeks .. 1.0 - - 7: 7.0 4.3 1.1 0.5 - -
3 weeks 0.3 - - - 0.9 1.1 0.5 0.2 - -
2 weeks 1.0 - - - - 22 1.9 0.8 0.3 -
1 week... 4.0 | - 40.0 3.6 35 4.3 4.4 4.8 26 1.4
Certificate same as questionnaire......... 85.8 - 50.0 75.0 73.9 79.3 86.7 88.2 87.4 79.3
Certificate greater than questionnaire:
1 week... 2.8 - - 71 35 - 26 24 35 3.6
2 weeks... 1.0 - - - - - 0.4 1.0 21 29
3 weeks... 0.9 - - - - - 0.7 0.6 1.2 29
11 - « - 0.9 1.9 0.4 0.5 15 71
14 - - - 0.9 22 0.8 1.1 1.2 29

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1gee table II in appendix I for a listing of the 11 States that did not include the date last normal menstrual period began on the birth certificate.

Table 12. Comparability of reporting month of pregnancy prenatal care began between the birth certificate and the National Natality Survey long hospital/physician
questionnaire, by month of pregnancy prenatal care began reported on birth certificate, 1972

 

Month of pregnancy prenatal care began reported on birth certificate

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Comparison of certificate with long i fie No pre- Noten
hiospital/hysicisn questionnaire cates || 1st | 2d | 3d | 4th | 5th | 6th | 7th | 8h | 9th | natal oe cate!
care
Number
All sample Cases.......ccovvrvrruriurnnns 5,689 I 448 1,731 1,271 494 | 258 142 102 48 22 38 | 373 | 762
Percent
Cases exXCIUAB. iv iiniminmmmisnisaniis 44.0 I 30.1 29.5 27.9 35.0 326 31.7 225 45.8 18.2 42.1 100.0 100.0
Number
Cases compared 3,187 I 313 | 1,221 916 321 | 174 97 | 79 26 18 | 22 - -
Percent distribution
Cases COMPArEd ...cwisvinisuunenmimimiri 100.0 || 100.0 | 100.0 | 100.0 | 100.0 | 100.0 | 100.0 | 100.0 | 100.0 | 100.0 100.0
Certificate earlier than questionnaire:
2 months or more... 15.8 34.5 15.4 139 11.8 14.9 9.3 7.6 1.5 or
1 month 226 55.3 27.0 11.9 11.2 19.0 11.3 22.8 1.5 33.3
Certificate same as questionnaire... 42.9 10.2 53.2 44.1 40.2 33.9 28.9 39.2 38.5 389 727
Certificate later than questionnaire:
14.3 4.3 27.9 24.9 14.4 25.8 11.4 19.2 1.1 45
2 months or more... 4.3 2.2 11.8 17.8 24.7 19.0 19.2 16.7 22.7

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1See table II in appendix I for a listing of the 10 States that did not include the item on the birth certificate.

25
Table 13. Comparability of reporting number of prenatal visits between the birth certificate and the National Natality Survey questionnaires, by number of prenatal visits reported
on birth certificate, 1972

 

Number of prenatal visits reported on birth certificate

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 
 
  
 

 

    

 

  

 

 

. Te All Not on
Comparison of certificate fi ii
ith questionnaires ils 19 or No certill,
wi cates || None | 1-2 34 5-6 7-8 | 910 | 11-12 13-14 | 15-16 | 17-18 cate
more | response
Number
All sample Cases ...........ocourviviivininnns 5,689 I 31 | 61 181 269 442 756 | 809 | 355 | 202 | sol as 268 | 2,230
Mother's questionnaire Percent
| 710] 26] coal seal sao! sar] 22 mal 252] a0ol anol
Cases oxBIUtEd ....cxiviinasirmnansmnno 63.6 71.0 52.5 60.8 46.8 38.9 33.7 28.2 | 25.4 25.2 40.0 40.0 100.0 100.0
Num
£8808 COMPAIET..c..srcrirsmmmircismmsnsrsrscrirsemiris 2,071 I 9 29 7 143 270 501 581 265 151 2 27 | - |
Percent distribution
C8305 COMPAIBU....ocnnsnrsnssirnmmremmnrssmmimmss 100.0 || 100.0 | 100.0 | 100.0 | 100.0 | 100.0 | 100.0 | 100.0 | 100.0 | 100.0 | 100.0 | 100.0 | |
Certificate less than questionnaire:
6 visits or more 15.4 22.2 44.8 324 29.4 19.6 16.0 1.2 7.9 1.3 11.4
4 visits. 5.1 1.1 - 7.0 13.3 10.7 4.2 4.0 23 0.7 -
3 visits. 6.3 - 13.8 7.0 7.0 8.1 8.4 71 1.5 0.7 4.2 -
2 visits. 10.9 6.9 18.3 126 11.9 15.6 4 10.2 6.0 12.5 11:1
1 visit.. 11.5 - 10.3 9.9 13.3 13.0 10.8 13.6 8.7 9.3 4.2 14.8
Certificate same as questionnaire ............. 15.6 66.7 12.2 14.1 9.1 16.3 16.8 18.6 8.3 17.9 8.3 7.4
Certificate greater than questionnaire:
1 visit. — 9.7 6.9 8.5 6.3 7.0 9.4 9.8 12.7 73 - 7.4
2 visits. 84 - 1.4 23 5.9 5.0 1.9 17.7 6.0 125 -
3 visits. 6.5 - 21 2.2 5.0 5.3 8.7 13.2 16.7 3.7
4 visits. 35 1.4 24 0.4 1.4 3.1 75 11.9 12.5 3.7
5 visits or more 8.3 vu 28 4.8 76 8.4 9.4 15.9 29.2 40.7
Long hospital /physician questionnaire - Percent
Casesexciuded....c.ummmmmmmsmm iver 60.4 I 45.2 | 36.1 41.4 | 34.2 31.0 | 26.3 28.6 | 28.2 24.8 | 25.0 | 24.4 100.0 | 100.0
Number
Cases COMPAIBL.. cvviverermimsvensmrnrrsmisnios 2,250 I 17 39 106 177 306 | 557 578 255 | 162 | 30 | 34 | - | -
Percent distribution
Cass COMPAIRD..icvs nv isrimirmmssgerirmras mums 100.0 || 100.0 | 100.0 | 100.0 | 100.0 | 100.0 | 100.0 | 100.0 | 100.0 | 100.0 | 100.0 | 100.0
Certificate less than questionnaire:
5 visits or more 5.5 11.8 12.8 14.2 7.3 8.9 6.6 2.8 20 20 3.3 -
4 visits. 3.9 - 77 75 4.0 6.6 45 3.1 24 0.7 - -
3 visits. 5.6 - 7.7 3.8 6.8 7.2 7.4 5.4 2.0 3.3 3.3 29
2 visits, 8.1 7.7 6.6 6.2 9.2 9.7 10.4 5.5 33 3.3 -
1 visit., 13.9 - 7.7 6.6 215 13.4 16.4 13.8 145 11.2 10.0 - .
Certificate same as questionnaire ............... 26.0 88.2 48.7 20.2 26.4 30.6 248 24.9 20.8 12.6 16.7 29 .
Certificate greater than questionnaire:
11.8 5.1 16.1 9.6 12 11.0 14.5 17.6 12 13.3 8.8
7.3 26 4.7 4.0 6.9 5.6 8.0 11.4 11.2 16.7 8.8
48 rin 6.6 28 1.3 4.1 5.5 nM 9.9 10.0 5.9
3.7 5.7 1.7 0.7 3.1 3.8 43 11.8 33 11.8
6 visits or more. .......... 10.3 10.7 8.2 7.9 7.8 125 27.0 20.0 58.8

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1See table II in appendix I for a listing of the 14 States that did not include the item on the birth certificate.

26
II.

II.

III.

IV.

APPENDIXES

CONTENTS

Technical Notes
Sample Design
Birth Certificate and Questionnaires
Collection of Data
Response Rates

 

 

 

 

 

1972 National Natality Survey Source Documents
Certificate of Live Birth
National Natality Survey Mother Questionnaire
National Natality Survey Hospital Long Form

 

 

 

 

LIST OF APPENDIX TABLES

Number of live births in the United States and number of births and sources of information in
the 1972 National Natality Survey ..... EE

Areas reporting educational attainment of parents, date last normal menstrual period began
(LMP), month of pregnancy prenatal care began, number of prenatal visits, and marital status
of mother on birth certificates: Each State, 1972

 

Response rates by type of respondent and stage of data collection: 1972 National Natality
Survey

 

Response rates for mothers by age of mother and color of child: 1972 National Natality Survey...

Item nonresponse rates for selected items, by selected respondent sources: 1972 National Natal-
ity Survey

 

28

30

31

31

32

27
APPENDIX |
TECHNICAL NOTES

Sample Design

The sampling frame for the 1972 National
Natality Survey was a file on microfilm of live-
birth certificates received each month by the
National Center for Health Statistics from the
54 birth registration areas of the United States.
These birth registration areas included the 50
States, the District of Columbia, and the cities of
New York, Baltimore, and New Orleans, which
had independent registration systems. Each
registration area assigned a file number to each
birth certificate, and these file numbers run
consecutively from the first to the last birth oc-
curring during the year in that area. The sample
for the survey was based on a probability design
that used these birth certificate numbers. Each
500 consecutive records from each area consti-
tuted a primary sampling unit, and one record
from each primary unit was selected at random.
Thus the sample of selected birth certificates
represented 1/500th of the live births occurring
in the 54 areas during 1972.

Sampled records for infants who were re-
ported or inferred to be out of wedlock were ex-
cluded from the survey, and no questionnaires
were mailed to any of the respondent sources.
Thus the statistics presented in this report per-
tain only to births to married women occurring
in the United States in 1972. In the registration
areas having an item on the birth certificate to
identify out-of-wedlock births, 555 sampled
records were excluded because the birth was re-
ported to be out of wedlock. In the 12 areas
that did not have this item, a birth was inferred
to be out of wedlock under the following condi-
tions: (1) the name of the father of the child
was omitted; (2) the mother’s surname as stated
in the “informant” or “mailing address’ section

28

was the same as her maiden name and was dif-
ferent from the father’s surname; (3) the
mother’s surname was different from her maiden
name, but also was different from the father’s
and the baby’s surname; (4) the mother’s sur-
name was missing from both the “informant”
section and the “mailing address’ section of the
certificate and the baby’s surname was different
from the father’s surname. Using these criteria,
261 sample records were inferred to be for out-
of-wedlock births. The 816 reported and in-
ferred out-of-wedlock cases were excluded from
the survey, and no questionnaires were mailed to
the mother, physician, or hospital named on
those birth certificates.

Table I shows the number of births in the
United States in 1972, the number of mothers in
the original sample, and the number of mothers,
physicians, and hospitals to whom question-
naires were mailed.

Table I. Number of live births in the United States and number
of births and sources of information in the 1972 National
Natality Survey

 

 

Item Number
Live births in the United States............ceeeeerrnnnneennnns 3,258,411
Births selected in the sample ............cccceeevivinerrvnnnnnnns 6,505
1
Out-of-wedlock births excluded from survey...... 816
In-wedlock births included in survey .................. 5,689
Mothers mailed a questionnaire..........cc.ccouve 5,676
Mothers not mailed questionnaire because
mother not U.S. resident ........cccceveeveennenenns 13
Births for which questionnaire was mailed
10 NOSPILA chris iinmimsmmmnininssmrimmsinssssany 5,647
INONNOSPION DIPTINS. cu iciiicisinriorssinanmineanisassiinmnes 42
Births for which questionnaire was mailed
10 PIVYSICIAN Jor vnivismsinssvississrorsssrsvissnpsanses 4,415

 

 
Birth Certificate and Questionnaires

Facsimiles of the U.S. Standard Certificate
of Live Birth and the questionnaires used in the
survey are shown in appendix II. The short
hospital (HS) and physician (P) forms are not
shown, but together their content was the same
as the long hospital (HL) form, which is shown.
Although most of the registration areas’ birth
certificates include the same basic information,
the standard certificate is not used by all regis-
tration areas. The item used to identify out-of-
wedlock births is omitted by 12 areas, and this
information was inferred for their records as de-
scribed previously. Educational attainment of
parents, date last normal menses began (used for
computing length of gestation), month of preg-
nancy that prenatal care began, and total num-
ber of prenatal visits are other items used in this
report that are not present on all the areas’ cer-
tificates. Table II shows the reporting areas for
each of these items.

Collection of Data

Data for the National Natality Survey were
collected primarily by mail by using the ad-
dresses given on the birth certificates. No ques-
tionnaires were mailed when the birth was re-
ported or inferred to be illegitimate.

If there was no response to the original mail-
ing of the questionnaire within certain time
limits, followup procedures were instituted as
follows:

1. If mothers did not return the original
questionnaire within 16 days, they were
sent a second questionnaire. If the fol-
lowup questionnaire elicited no response
after an additional 21 days, then an in-
terview by telephone or in person was
attempted. Incomplete or inconsistent
items on questionnaires from mothers
were followed up by telephone or per-
sonal interview. The mother of the in-
fant was the only person from whom in-
formation for the mother’s questionnaire
was accepted. In the telephone and per-
sonal interviews, no proxy respondents
were accepted.

2. If physicians did not return the original
questionnaire within 22 days, a postcard

reminder was sent. If after an additional
14 days there was still no response, then
a followup questionnaire was mailed. No
interviews were attempted with physi-
cians.

3. If hospitals and clinics did not return the
original long or short hospital question-
naire within 22 days, a followup ques-
tionnaire was sent. No further followup
attempts were made. Any items left
blank by physicians or hospitals were as-
sumed to mean that the respondent had
no knowledge of that aspect of the
mother’s or baby’s health care or
history.

Response Rates

In any survey where the participation of the
subjects is not mandatory, there will be some
subjects who do not respond to the survey ques-
tionnaire. Mothers, physicians, and hospitals
were all informed, both on the printed question-
naires and by the telephone and personal inter-
viewers, that they were under no legal obligation
to participate in the survey and that their partic-
ipation was completely voluntary.

The final response rates were 71.5 percent
from the mothers, 72.2 percent from the physi-
cians, and 85.4 percent from the hospitals. Table
III shows the number of persons or institutions
queried and the percent responding at the differ-
ent stages of data collection. The response rates
from the mothers varied with age—the younger
mothers had lower response rates. At each age,
the response of white mothers was higher than
that of all other mothers. The number of births
in the sample and the response rates by age of
mother and color are shown in table IV.

Table V shows selected variables used in this
report and their item nonresponse rates. Al-
though not every item has been included in
the table, there was no variable with a higher
nonresponse rate than those shown. The figures
do not include cases where no questionnaire was
returned, where an item did not appear on a
State’s certificate, or where a “don’t know’ re-
sponse was allowed.
Table Il. Areas reporting educational attainment of parents, date last normal menstrual period began (LMP), month of pregnancy
prenatal care began, number of prenatal visits, and marital status of mother on birth certificates: Each State, 1972

 

Area

Educational
attainment
of parents

Date last
normal
menstrual
period
began
(LMP)

Month of
pregnancy
prenatal
care began

Number of
prenatal
visits

Marital
status
of
mother

 

Alabama...
Alaska..
Arizona
Arkansas. 2
CAEONNID oi civvmnrisciirimisimevimmsssssivitsms ss RA STRSTR STIS RSS

 

COIOTBOO scons sssssrsscmuninsosstin sets ps isss snes a SAT SRI SERA
Connecticut ...............
DeIBWare......msmivisisenss
District of Columbia... :
IOI sasrem mss Ta Tamas arr wmmrss ann ammuesrss sm mnsspsnrunserenyesnnmnnelie

BOOT os: ssmmun min insrvrs rss ass ck TR rvs st sini En eran ra dams ee ns Finmrs aman thasaye
Hawaii...

 
 

 
 
 
 

Kentucky .
Louisiana.

PABPY RIND. cosmmisiinssrmsmsassanssnsiontmminsnsviine saan AAT EAA ARR T to oo
Massachusetts. —_—
IVCIIGAI oosrinmnsmnsininsiniinnissminsssmvivensesspvviin issn irnsmnsin is ionsnsomy
IVININBSORD ou raimsmsnnsmsmsmsminrisssssniinmssss sees tas ssa Bese HASTA ROSATO AAA
NA SSIES) rerrrssasdtencumranvesersmmsesntsnssssnssssnnssansssssenrsarisessassnrspsiross pansnss

 

IVISBOMINY 2c rsnnnunnsinmssnaiimensnmrssssensaanssssssinmssnsossasiysssassssnnsosssssenmsg sss uss
MONEE. 21sec iinsnsssrimammsmmsms rrr TT TR sa arse s9
Nebraska..

  
 
 
  

INOW JBFSBY ....ocoincmmnimmmsmnminressassvanssmsmssvssmnis sim wns s asa A ERR SASSER S33 55
New Mexico...
New York.........
North Carolina.. wens
NOIth DaKOta.......uuueeeiirieiiieriieieiereee se srsree see sreesaees ss sseas esse sssssaneanen

Oklahoma...
Oregon........c..
PeNNSYIVANIA cei cee cece crane aera ananassae aren aaa
BROCE 181800... ivommmmsnnirte sss ssnssss ssi essa AE EERE RENE ATER

 

SOU Car ING ..oioscsssrrimsmsssivimssassst snes asses toss 13s EER FRSA ERRATA FRI RR RETR
SOU DAKOTS,..co0inenimmannisasinnisiomssstirmsssan iss ors saa SRS HE SOS

 

VBHITIONT ..cvucins ins sssprmmsssssssesmnrsmmsssessssssssssress busses sssesmssnsssstarniasisetvisnnss

  
  

Virginia.......
Washington ....
West Virginia..
Wisconsin....... ire
WYOMING ccc crusssnisinmmimmsasnsrrisssestasss sas sass eras sass estan smss sis has svs aR ARRAS A

 

 

 

 

x| xX

x

XI X| XIX

 

 

 

 

x

 

 

 

 

X|X[X [X |X

XXX [XX |X

X[X| X

X|X| Xx

 

 

x

x

x

x

 

 

 

 

 

 

 

 

XXX] X|X |X| X

XI X[X[X|X |X] X]

X|X|X|X[X |XX |X

 

X | X[X]|X|X[X [XIX

 

 

 

 

 

 

 

 

 

 

XIX X[X|X[X | X[X|X

xX [XX] |X] |X |X[X|X

 

 

 

 

XX

X|X

 

 

 

 

 

 

 

 

 

 

XIX |X| X|X |X| X|X|X

 

XX [X[X]X IX] XXX [X[X|X[ [XO X]X[X[X]X | X]|X|X| [X [X]|X]|X[X|X |X|X

 

 

 

 

 

 

XXX] XXX] [X[X[X |X] [X|X[X [X]|X|X]| |X |X[X|X|X[X |X|X[X{x

 

xX| |X|X

 

XIX[X|X] [XXX] X[X|X |X] [X|X]|X [X]|X[X] [X [X|X]|X]|X[X |[X]|X|X

 

XIX[X|X[XIX |X| [X[X[X [Xl [|X|X

 

XIX[X|X|[X

 

30
Table 111. Response rates by type of respondent and stage of data collection: 1972 National Natality Survey

 

 

 

 

 

 

 

 
 

 

 

 

 

 

 

 

 

 
 
  

Respondent
Response status ;
Mothers Private Hospitals
physicians
Number
TOU HD SUEY sssurisssnirssnsirimsbisssmstsssstss soa ris ES ERE SR SHIRA wn 5689 | 4416 | 5647
Percent distribution
TOTAL IN SUPVBY ..rovcrisiorsrsssssissssasrmssmrsmss ises sess ist iaass va VRE ERT SURE ETSY TAINS ATARI CATR AIP RITT IS 100.0 | 100.0 100.0
Total response .......cuvveinne 85.4
Response to first mailing ...... 76.6
Response to second mailing.. 8.8
FROSPIONSE 10 TIVE OIVIBW vu coesus setts sears rE ES EET FSA TA EET REF RRR EFT REH I ETE A RASS Te ARR RHA RARE FEI RARER SSIS
OLA NONTBEPOIEE oc vsvrrranssnssrsssrisrsestssenrssostes tars san HEIRS ATARI TSAR T SEE TE CREE TARR TITLE I RPE aA 14.6
Table IV. Response rates for mothers by age of mother and color of child: 1972 National Natality Survey
Total T White All other
Ageiof mother Number Percent Number Percent Number Percent
in sample | responding || in sample | responding | in sample | responding
AUBIRS .ovvossssimimiimmimnsnssiinsismsmsnssssasnnes 5,689 71.5 5,007 73.6 682 56.0
MInar 20 YoarS. .....iunimmsmmstim ssi ss rriss soy 833 57.5 708 58.9 125 49.6
20-24 years ......... a 2,137 70.7 1,899 73.5 238 48.7
25-29 years... 1,681 76.9 1,517 78.8 164 59.1
30-34 years sie 692 74.7 586 75.6 106 69.8
35 VRAIS aN VBI ..oocusvssssivsssismarivsssssnssssrisssirsisessainnsss 346 77.2 297 78.8 49 67.3

 

 

 

 

 

 

 

 

 

31
Table V. Item nonresponse rates for selected items, by selected respondent sources: 1972 National Natality Survey

 

 

 
 
  
 
 
 
 
 
 
 
 
  

Percent
Respondent :
Item SOUTES item
nonresponse
BOR OF NOTICE occurs sssnrsssinmsssnmssstinss sesss shia SAE 404 45TH RR A SRR SES HATS SERA SARS 4 Ee FOS LAREN THEE RR RFR RR USERS A RRS LRTI BC 0.1
APB OF MONBE vuvivivrivansisssimvsmasisi visitas mss . M 1.2
Age of mother... ’ HL/HS 1.4
AGB OF TOBE .iiviriisrsrsensscsmsiinesssesasssssssensenssssshssnseassssrssn res stats svs tS IIIS ER EERIE SRS EARS SA SHAS ERS SRR R AP HURTS SRE R PE SR RSRAES BC 0.4
AOR OF TAGE ovine tints a oss ee EE EAT TRE A SEE ARS EA SAA EAB Sass Aas RRA RP RSA RE RAR ORAS SRA RR SR FERRARIS M 3.7
Education of mother..... . BC 1.2
Education of mother..... . M 0.2
Education of father.... BC 1.8
Education of father... M 1.0
Birth weight ............... 3 BC 0.4
BIE WOIONL «oi crinniiscansmsssssesrassssstonsssssessaionsis sass smssssstnasmessssssstassessasiamssssssmmm : HL/HS 1.3
Date iast menstrual Period Bega cor imnimriimmrsississmissmasmssissssis son . BC 15.0
Date last menstrual period began............... . HL/HS 12.4
Month of pregnancy prenatal Care BEGAN ............ccciierirrueneeirrnriiiree ert ee ates ener eessssssssssssssssseessreseesesssessanasses BC 5.2
Month of pregnancy Prenatal Care BEGAN ...........ceeeiiiireereiiireiereeii irre tease crassa esessssserssssssessssssssessesssnnessssnnnes HL/P 1.0
Number of prenatal Visits ..........ccceeereiinnnes BC 4.7
Number of prenatal Visits ........cccceceereniereerieninenns HL/P 1.7
Number of prenatal visits to attending physician... M 6.0
Number of prenatal visits to other physician......... M 4.6
Number of previous children born alive, still living BC 1.2
Number of previous children born alive, now dead................ : BC 23
Number of previous children born alive, still living................ . M 0.1
Number of previous children born alive, now dead . HL/HS 2.17
Number of stillbirths. .......ccccecveniiiiicninniiicsinninienn _— M 0.5
Number of miscarriages........... re, M 0.4
NUMDET OF PrOVIOUS DrOBIIBNGIES. covvrirrirssrsrsrnriss sssss sissssiasssrassssssssssvssnavns svssss ves 43 FETS MEREP IER ARS RARE RF HSS FTN ERR RS HL/HS 2.6

 

 

 

1BC refers to birth certificate; M refers to mother’s questionnaire; HL/HS refers to long/short hospital questionnaire; HL /P refers to
long hospital /physician questionnaire.

 

000

32
APPENDIX 11
1972 NATIONAL NATALITY SURVEY SOURCE DOCUMENTS

Certificate of Live Birth

 

FORM APPROVED

 

  
    
  
     
    
   
 
 
 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

   
 
  
 
  

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

  

 

 

 

 

   

 

 

 

   

 

 

 

 

BUDGET BUREAU NO. 63-R1900 r - U.S. STANDARD r a
TYPE, OR PRINT IN LOCAL FILE NUMBER CERTIFICATE OF LIVE BIRTH BIRTH NUMBER
PERMANENT INK rrr “WIDOLE nT TI
CHILD NAME inst MiDoll LAST TMONTH, DAY, YEAR | UR
SEE HANDBOOK FOR DATE OF BIRTH
a INSTRUCTIONS \ 2 2 M
2 SEX THIS BIRTH — SINGLE, TWIN, TRIPLET, ETC. IF NOT SINGLE BIRTH —80RN FIRST, SECOND, COUNTY OF BIRTH
2 (SPECIFY) THIRD, ETC. (SPECIFY)
3 3 “ |» so
INS) TY LIMIT —
x CITY, TOWN, OR LOCATION OF BIRTH Ing Io 3 y Lm ATS HOSPITAL — NAME (IF NOT IN HOSPITAL, GIVE STREET AND NUMBER )
<
T 5b. 5c 54
8 MOTHER — MAIDEN NAME FIRST MIDDLE ast AGE LL TIME OF (STATE OF BIRTH (IF NOT IN U.S.A., NAME COUNTRY)
THIS BIRTH)
«
: MOTHER Lic STATE I T T a CITY LIMIT :
& ! COUNTY | CITY, TOWN, OR LOCATION oh Sins STREET AND NUMBER
3 | 7c 74 In
2 — NAME Fast MIDDLE Gast AGE (AT TME OF [STATE OF BIRTH (IF NOT IN U.S.A. NAME COUNTRY |
3 [ACT] THIS BIRTH)
| ; » x
u TINFORMANT RELATION TO CHILD
:
n ».
x DATE SIGNED (MONTH, DAY, YEAR) ATTENDANT —m D., D.0., MIDWIFE, OTHER
= STATED ABOVE. ( SPECIFY )
T 100. SIGHATURE 100 10¢
g CERTIFIER — NAME (TYPE OR PRINT) MAILING ADDRESS (STREET OR R.F.D. NO., CITY OR TOWN, STATE, ZIP)
a
S
1 10d _ 100
wl REGISTRAR — SIGNATURE DATE RECEIVED BY LOCAL REGISTRAR
Sa MONTH DAY ear
§ 2 Ile 11h
=
os CONFIDENTIAL INFORMATION FOR MEDICAL AND HEALTH USE ONLY
« = RACE — FATHER EDUCATION — SPECIFY HIGHEST GRADE COMPLETED PREVIOUS DELIVERIES — HOW MANY OTHER CHILDREN
3 WHITE, NEGRO, AMERICAN INDIAN, ETC ELEMENTARY HIGH SCHOOL COLLEGE ARE NOW LIVING (WERE BORN ALIVE | WERE BORN DEAD
g (SPECIFY) 01,20, ons 023, 00 4) 11,2,.4, 00 § +1 NOW DEAD weit DEAN A Ay TIME
3 12 [FX Mo. * «
- RACE —MOTHER EDUCATION — SPECIFY HIGHEST GRADE COMPLETED DATE OF LAST LIVE BIRTH | DATE OF LAST FETAL DEATH
x EE amr mmm eer ree ey eer Sr eeanyy MONTH DAY YEAR | MONTH DAY Year
- WHITE, NEGRO, AMERICAN INDIAN, ETC ELEMENTARY HIGH SCHOOL COLLEGE
- (SPECIFY) 9,1,2.0,8, on 1,34, 0054+
5 1s W 1s
5 DATE LAST NORMAL MENSES BEGAN [MONTH OF PREGNANCY PRENATAL PRENATAL VISITS TOTAL NUMBER [LEGITIMATE
z DEATH MONTH Dav YEAR | CARE BEGAN (IF NONE, SO STATE) (SPECIFY YES OR NO)
¥ UNDER ONE YEAR [FIBST, SECOND, THIRD, ETC. ( SPECIFY )
&
OF AGE
Seren stare rue u 1% 19 0
8 ach 2am [COMPLICATIONS RELATED TO PREGNANCY WRITE "NONE ) | BIRTH INJURIES TO CHILD NONE)
THIS CHILD.
122 2
s | COMPLICATIONS NOT RELATED TO PREGNANCY (DESCRIBE OR WRITE "NONE | | CONGENITAL MALFORMATIONS OR ANOMALIES OF CHILD (DESCRIBE OR WRITE "NONE )
1 MuLnete siRTHS |
| ENTER STATE FILE [24 2
3 Mumpth 10 [COMPLICATIONS OF LABOR TOESCRIBE OR WRITE NONE 1
LIVE BIRTH(S)
°
$ be.
1
£ FETAL DEATH(S)

 

 

 
National Natality Survey Mother Questionnaire

34

 

DEPARTMENT OF HEALTH, EDUCATION, AND WELFARE
PUBLIC HEALTH SERVICE

HEALTH SERVICES AND MENTAL HEALTH ADMINISTRATION
ROCKVILLE, MARYLAND 20852

NATIONAL CENTER FOR
i" HEALTH STATISTICS

The Public Health Service is conducting a national survey of medical
care provided to mothers who have babies during 1972. We are trying

to learn more about the medical care mothers received during the period
before and after the birth of the child. Past studies have shown that
medical care is related to the health of a mother and her baby. The
information which mothers throughout the country give us will greatly
aid in planning better medical care programs for all American women.

You are one of a small sample of mothers being selected to represent
all mothers having babies in 1972. Because of this you play an im-
portant role in telling us about the medical care you received before
and after the birth of your child.

All information you give us, as well as that provided by medical per-
sonnel and facilities listed by you in the questionnaire will be held
strictly confidential. No information will be released to any other
person or agency.

In giving answers to the first part of the form, please name every
doctor, hospital, or clinic from which you received any care related
to your pregnancy during the period specified in the question. It

is necessary that we obtain as complete and accurate a picture as pos-
sible of all the medical care you received before and after the birth
of your baby. If you do not know an exact answer to any of the ques-
tions in the form, give your best estimate. Please complete the form
and return it within the next few days in the enclosed postage-free
envelope.

Thank you for your cooperation.
Sincerely yours,

Yel @ Jorae 2 —

Robert A. Israel
Director, Division of Vital Statistics

 

NAME OF CHILD DATE OF BIRTH

 

 

 

 
 

 

ASSURANCE OF CONFIDENTIALITY — All information which would permit identification of an individual, or of an establishment, will be held confidential, will be
used only by persons engaged in and for the purpose of the survey, and will be protected against disclosure in accordance with provisions of .i2 CFR Part I.

 

 

VOLUNTARY PARTICIPATION — Completing this form is voluntary; you are under no legal obligation to do so.

 

NATIONAL BIRTH SURVEY

 

PART |. SOURCES OF MEDICAL CARE

This part is concerned with persons or places which provided medical care to you. If you do not know a complete address, please give us as much informa-

tion as you can.

 

1. (a) List the name and address of the doctor, midwife, or other person
who delivered your baby.

 

NAME (First) (Last)

 

ADDRESS (Number) (Street)

 

(City or Town) (State) (Zip Code)

 

(b) How many times were you seen for medical care by this person dur-
ing the year before the baby was born? (DO NOT include the deliv-
ery episode).

Number

 

3. Did you see a doctor, midwife, or other person for any medical care re-
lated to your pregnancy within THREE MONTHS AFTER THE BIRTH

2. Were you seen by any other persons or places (hospitals, maternity
clinics, etc.) for prenatal care (care related to your recent pregnancy)?

[JYes
{

[JNo (Go to question 3)

List the names and addresses of all persons or places which provided
prenatal care to you.

 

A. NAME (First) (Last)

 

ADDRESS (Number) (Street)

 

(City or Town) (State) (Zip Code)

 

How many times were you seen for prenatal care by the above?

 

 

 

 

 

 

 

 

 

 

(Number)
OF YOUR BABY? :
B. NAME (First) (Last)
[J Yes [J No (Go on to Part II)
ADDRESS (Number) (Street)
List the names and addresses of all persons who provided medical care
related to your pregnancy within three months after the birth of your
baby. (City or Town) (State) (Zip Code)
mAMC (Fis) Yast) How many times were you seen for prenatal care by the above?
(Number)
ADDRESS (Number) (Street) .
If more space is needed, continue below.
(City or Town) (State) (Zip Code)
NAME (First) (Last)
ADDRESS (Number) (Street)
(City or Town) (State) (Zip Code)

If more space is needed, continue below.

 

 

HSM-254-1 (PAGE 2)
REV. 7-62

(GO ON TO PART II)

0.M.B. No. 68-5 72088
Approval expires: 12-31-73

35
 

PART Il. INFORMATION ON HEALTH INSURANCE

 

In this part, we are interested in any health insurance you or your husband may have had during the TWELVE MONTHS before the baby was born.

 

1. Did you have any kind of health insurance for hospital or doctor bills at any
time during the twelve months before your baby was born?

[J Yes

[J No (Go to question 6)

 

2. Did you have any kind of health insurance at the time your baby was born?

[J Yes [JNo

 

3. (a) Did health insurance pay for any part of the medical care you received
during your pregnancy PRIOR TO the delivery?

[Yes [JNo

[7] No medical care bill during pregnancy

(b) What part of the medical bills during pregnancy did your insurance pay?

[(J1/4 or less
[Jover 1/40 1/2

[Jover1/2to 3/4
[Jover 3/4

5S. (a) Did health insurance pay any part of the doctor’s bill for
delivering your baby?

[yes [J No [7] No doctor's bill

(b)

What part of the doctor's bill did your insurance pay?

[J 1/4 or less
[Jover 1/4 to 1/2

[Jover1/2 to 3/4
[Jover 3/4

 

 

4. (a) Did health insurance pay any part of the hospital bill when your baby
was born?

[J Yes [J No [J No hospital bill

(b) What part of the hospital bill did your insurance pay?
[J 1/4 or less [Jover1/2 to 3/4
[Jover1/4t0 1/2 [Jover 3/4

 

6. (a) Did any organization or agency (such as the Armed Forces, Medi-
caid, welfare, lodges, unions, etc.),pay for or provide any part of

the medical services connected with pregnancy or birth?

[Yes [J No

(b;

What part of the medical services were paid for or provided by
the organization or agency?

[J 1/4 or less
3 over 1/4 to 1/2

[Jover 1/2 to 3/4
[J over 3/4

(c) What is the name of the agency or organization?

(GO ON TO PART III)

PART lll. INFORMATION ABOUT YOU AND YOUR CHILDREN

We are interested in the outcomes of all the pregnancies you have ever had, even if they occurred before your present marriage. Please INCLUDE the

child listed on the front of the questionnaire.

 

1. How many children have you ever had? (Count all those that were born
ALIVE to you AT ANY TIME.) TTI

2. Have any of these children died? (DO NOT count miscarriages or babies that
were born dead.)

[J Yes

[TJNo (Go to question 3)

Please list below, the name, sex, and dates of birth and death of each such
child.

 

(First)

NAME OF CHILD SEX
(Middle) |M|F

DATE OF BIRTH
Mo. | Day |Year

DATE OF DEATH
Mo. | Day | Year

3. Were any of your children living away from you when the child listed
on the front of the questionnaire was born? (For example, usually
living with relatives, adopted by someone else, in the Armed
Forces, etc.) Do not include children who were away at school or
college.

[7] Yes [CJNo (Go to question 4)

Please list below the name, sex, and date of birth of each such
child.

 

DATE OF BIRTH
Mo. | Day Year

NAME OF CHILD SEX

(First) (Middle) M| F

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

4. (a) Have you ever had a stillbirth? (That is, a baby that was born dead)
[J Yes [CINo (Go to question 5)

 

Number
A [eee]

 

 

(b) How many have you ever had?

 

(c) Please give the date of your last stillbirth.

(Mo. Day Year)

 

HSM-254-1 (PAGE 3)

(Part Ill continued on Page 4)
REV, 7-72
 

PART lll. Continued

5. (a) Have you ever had a miscarriage? (DO NOT include any stillbirth
counted in Question 4)

TJ Yes [TJ No (Go to question 6)

 

Number

 

 

 

(b) How many have you ever had?

 

(c) Please give the date of your last miscarriage.

(Mo. Day Year)

 

6. Thinking back, just before you became pregnant with your new baby, did
you want to become pregnant at that time?
I wanted this pregnancy at an earlier time, as well as at that time.

[7] 1 wanted to become pregnant at that time.

[CJ 1 did not want to become pregnant at that time, but I wanted another
child sometime in the future.

OO ! did not want to become pregnant at that time, or at any time in the
uture.

 

7. Do you expect to have more children?

[] Definitely yes

[] Probably yes

How many more children do you think
you will probably have?

 

Number

 

 

 

 

[7] Probably no
[7] Definitely no

(GO ON TO PART IV)

PART IV. INFORMATION ABOUT YOU AND YOUR HUSBAND

 

(Check ONE box only)

1. Is this your first marriage?

[] Yes —==——s-Please give the date of your marriage

Wo. Day Year
Please give the date of your first marriage
Mo. Day Year
[Neo
Please give the date of your present marriage
Mo. Day Year

 

2. (a) What is the highest grade of regular school (elementary school, high
school, two year or four year college or university) that you COM-
PLETED? (DO NOT include business or trade schools, or other
specialized training)

(Circle the highest grade of regular school completed)

0 12345678 9 10 11 12 13 14 15 16 17 18+
y University or Graduate
None Elementary school High school college school

(b) Other specialized training;
C- Yis [CNo

Specify:
(trade schools, beauty-barber college, hospital schools, etc.)

 

Circle years completed

Less than one 1 2 3 or more

 

3. (a) What is the highest grade of regular school (elementary school,
high school, two-year or four-year college or university) that
your husband COMPLETED? (DO NOT include business or
trade schools or other specialized training.)

(Circle the highest grade of regular school he completed)

0 12345678 91011 12 13 14 15 16 17 18+
; University or Graduate
None Elementary school High school college school

(b) Other specialized training;
[Yes

t
Specify:
(trade schools, beauty-barber college, hospital schools, etc.)

[No

 

Circle years completed

Less than one 1 2 3 or more

 

(GO ON TO PART V)

 

HSM-254-1 (PAGE 4)
REV, 7-72

37
 

PART V. INFORMATION ABOUT YOUR FAMILY

 

In this part information is asked about all relatives living with you when the baby listed on the front of the questionnaire was born.

 

1. List below all relatives who usually lived with you at the time of your recent delivery. Be sure to list yourself, your baby, your husband (if he lived at
home), as well as any of your children and other relatives living with you. Include children who were away at school or college. DO NOT include rela-
tives who lived somewhere else (for example, relatives in the Armed Forces). Also, DO NOT include relatives who were only staying in your home

temporarily when the baby was born.

 

NAME

Enter your name on the first line; enter the names of every
other relative who lived with you on the following lines. Be
sure to include the baby

(First Name) (Last Name)

For YOURSELF and EACH RELATIVE, provide the information requested below.

 

RELATIONSHIP TO YOU

(Husband, daughter, son,
father, father-in-law, nephew,
stepson, adopted daughter,
etc.)

DATE OF BIRTH

 

Mo.

Day

Year

MARITAL STATUS

Single (never married)
Married

Separated

Widowed

Divorced

 

YOURSELF

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2. Who was the head of this family? (This person must be you or one of the relatives who is listed above.)

[J Your husband
[] Yourself

[] Another relative ——— Name of head

 

HSM-254-1 (PAGE 5)
REV. 7-72

38

(GO ON TO PART VI)
 

PART Vi. FAMILY INCOME

 

The following questions refer to the money income of members of your family during the TWELVE MONTHS before the baby was born. Include all incomes
of members of the family whom you listed even if they were not living together during part of the twelve months. Include all income from wages, salaries,
investments, property, Social Security, welfare, unemployment compensation, help from relatives, etc.

 

1. What was the income (total income before deductions for taxes, bonds, 2. Whatwas the total family income (before deductions for taxes, bonds, dues,

dues, insurance, etc.) received by YOUR HUSBAND from all sources
during the twelve months before the baby was born? (This income should
include money from wages, salaries, commissions, bonuses, tips, own
business, professional practice, farm, unemployment compensation, etc.)

insurance, etc.) received by YOURSELF, YOUR HUSBAND, and ALL
OTHER LISTED FAMILY MEMBERS from all sources during the twelve
months before the baby was born? (This income should include money
from wages, salaries, commissions, bonuses, tips, own business, pro-

If exact amount is not known, please check your best estimate. fessional practice, farm, unemployment compensation, etc.) If exact

amount is not known, please check your best estimate.

(Check one) (Check one)

[] none or under $1,000
[7] 81,000 to $1,999
[] $2,000 to $2,999
[7] $3,000 to $3,999
[] $4,000 to $4,999

[7] $5,000 to $6,999
[7] $7,000 to $9,999
[7] $10,000 to $14,999
[7] $15,000 to $24,999
[7] $25,000 or more

[7] none or under $1,000
[C7] $1,000 to $1,999
[C7] $2,000 to $2,999
[C7] $3,000 to $3,999
[C7] $4,000 to $4,999

[7] $5,000 to $6,999
[7] $7,000 to $9,999
[C7] $10,000 to $14,999
[7] $15,000 to $24,999
[7] $25,000 or more

 

(GO ON TO PART VII)

PART VII. PERSON COMPLETING THIS FORM

 

NAME

 

ADDRESS

 

(Number) (Street)

 

(City or Town) (State) (Zip Code)

 

TELEPHONE NO. DATE OF COMPLETION

 

(Month, day, year)

 

NOTES AND COMMENTS

 

H SM-254-1 (PAGE 6)

If more space is needed continue on back.
REV. 7-72

GPO 930-960

39
National Natality Survey Hospital Long Form

40

 

DEPARTMENT OF HEALTH, EDUCATION, AND WELFARE
PUBLIC HEALTH SERVICE
HEALTH SERVICES AND MENTAL HEALTH ADMINISTRATION
ROCKVILLE, MARYLAND 20852

NATIONAL CENTER FOR

~ HEALTH STATISTICS

Your assistance is needed in a national birth survey being conducted by
the Public Health Service with the approval of your State Health Depart-
ment. We are seeking information on the amount of medical care provided
to expectant mothers, and in the case of live births, to their newborn
infants. This information is being collected on a sample of approximately
7,400 mothers of live births who represent the nearly 3.7 million women
having deliveries during 1972. Each of the mothers included in the
sample is being sent a questionnaire asking about her recent pregnancy.

According to our records, the mother named below was seen or treated at
your facility at some time during the survey period given at the bottom
of this page.

Since the 1972 national birth survey is based on only a small sample of
mothers, it is particularly important that we receive as much informa-
tion as possible on all mothers in the sample.

Please be assured that all information which you report about the mother
and the newborn will be kept completely confidential. No identifying
information will be disclosed to any person or any other agency. The
data we collect will be used for statistical purposes only.

Please complete the questionnaire and return it within two weeks. Your
cooperation in this study is deeply appreciated.

Sincerely yours,

Ad Corpo

Robert A. Israel
Director, Division of Vital Statistics

 

 

 

 

 

 

 

Name of Mother Name /Sex

Address Date of Delivery
City (town), State, Zip Code Survey Number
PERIOD COVERED BY THIS SURVEY: From To

 

 
 

 

ASSURANCE OF CONFIDENTIALITY. AH information which would permit identification of on individual, or of an establishment, will be held confidential, will be used only
by persons engaged in end for the purpose of the survey, and will be protected against disclesure in accordance with provisions of 42 CFR Part |.

 

 

VOLUNTARY PARTICIPATION — Completing this form is veluntary; you are under ne legel obligation te de se.

 

NATIONAL BIRTH SURVEY
PART I. PREGNANCY HISTORY OF MOTHER

 

(DO NOT include this delivery)

Total number of previous pregnancies ........oeeeeenn.. or [_] None

Number of pregnancies not ending in
live birth (include all miscarriages,

abortions, stillbirths, @16.) « vuvssvesssonsrenrsvsne or [J] None
Number of live births ves sess vensconnssnersnnsennes or [[] None
Number now living ve csvessnssinnrasnssnnsrnnns or [_] None
Number now. dead «...cvcrevrsnsnnsssvnseansnnse or [] None

(GO ON TO PART Il)

PART Il. DELIVERY EPISODE

 

In this part we are interested in the condition of the mother and child and the care which the mother and child (if live born) received in the

hospital from the time of delivery to discharge.

 

SECTION A. THIS DELIVERY

 

Mo. Day | Year

 

 

 

 

 

 

 

 

1. Deve ok admins) t moth 9. Complications of this pregnancy (Check all those which apply)
. Date of admission of mother

: [J Urinary infection [J Anemia
2. Date of discharge of mother [2 Hypertension C1 Resello

(a) Was mother discharged [J Alive [J Dead [CO] Toxemia preeclampsia [C] Embolism
[C] Eclampsia [C] Obesity
. i a 3 Oth
3. Age of mother at time of this delivery (ageatlastbirthday) years O (Specify) [J Nene
Me. Day | Yeor 10. Underlying medical conditions existing during this pregnancy

 

 

 

4. Date last normal menses began

 

5. Total duration of labor (If precise answer is not known, give your best
estimate) hours (or) [] Not known

 

6. Type of Anesthetic for delivery (Check all those which apply)

[] Inhalation [J Local
[J Spinal and epidural [CT] None
[J] Other

(Specify)

 

 

7. Type of delivery

(Check all those which apply) [C] Obesity

[CO] Anemia
[J] Diabetes {T] Cardiovascular-renal
[] Varicosity disease
[] Asthma
[CJ] Congenital heart disease J Other chronic pulmonary
[C] Thyroid condition [C] Orthopedic condition
[J] Other [J Nene

(Specify)

 

 

Were any complications tomother’s health noted afterdelivery?

[J Yes, specify

 

 

[J] Spontaneous [] Cesarian Section
[J] Forceps [C] Breech [J No
[CJ Other

(Specify) 2

. Was any operation performed which will prevent future

 

 

8. Complications of labor (Check all those which apply)

[TJ] Inadequate pelvis [CJ] Unusual bleeding
[CJ] Transverse lie [CJ Prolonged labor

[CJ] Multiple birth [CT] Anesthesia reaction
| Abnormal position of [2 Placenta abruptio

placenta or cord

[C] Premature rupture of membranes [_] None

[C] Other

preg ies? [] Yes [J No

 

 

 

. Condition of infant at delivery

[C] Born alive [C] Born dead (Go to Part 111)

APGAR rating; give scores

 

At one minute or [_] Not Done

At five minutes or [_] Not Done

 

 

(Specify)
HSM-254-5 (PAGE 2) O.M.B. No. 68.5 72088
REV. 7-72 Approval Expires : 12-31-73

41
 

PART Il. SECTION A (Continued)

 

14. Were any unusual resuscitative efforts required?
[] Yes, specify
[J Ne

 

 

15. Congenital malformations or anomalies noted at delivery

 

18. Infant named on the front of this questionnaire was,
[] Single (Go to Section B)
[J Twin

[] Triplet, or other plural

 

 

 

(or) [_] None
16. Birth injuries noted at delivery 19. If not single birth, was infant born
(or) ([] None [7 First
17. Weight of infant at birth [J Second
pounds, ounces, (or) grams. [J Third or higher

 

 

SECTION B. NEWBORN

 

1. Age when first examined

 

3. Were any congenital malformations or anomalies noted before

 

 

 

 

 

 

 

outside of delivery room hours (or) days discharge?
[J Yes, specify
2. Were any birth injuries noted before discharge?
[J Ne
[J Yes, specify
4, Were any other illnesses noted before discharge?
[CJ Ne i
[] Yes, specify
{J Ne
5. Was infant discharged alive?
[] Yes [CJ] No === Age at death days, hours
| Cause of death (Go to Part Il)

(a) Date of discharge

ont, ear

ay

 

(b) Was infant given a discharge examination?

[] Yes [J Ne

[CJ RESIDENT (or) [_] PRIVATE

 

NAME OF PHYSICIAN

 

(c) Place infant was discharged to
[] Family home
[[] Medical care facility

O Other

 

NAME

 

(GO ON TO PART Ill)

HSM-254-5 (PAGE 3)
REV. 7-72

42
 

PART lil. PRENATAL AND POSTPARTUM CARE OF MOTHER

In this part we are interested in all the prenatal and postpartum care the mother received from the period TWELVE MONTHS before
this delivery to THREE MONTHS after.

 

 

3. Was the mother referred to any other medical facilities or persons

1. Did the mother make any visits for prenatal care to facilities operated
for prenatal care?

by this hospital?

 

 

 

 

 

 

 

 

 

 

 

 

 

: Yes [CI Ne (Go to question 4)
[1 Yes [CJ No (Go to question 2.)
(a) Month of pregnancy prenatal care began (Check one)
NAME
1st [J 2nd [J 3rd [7 4th [CO] 5th
[J 6th [7th [CO 8th [9th ADDRESS (Number) (Street)
(b) How many visits did she have?
(Number) (City or Town) (State) (Zip Code)
(c) Were any complications or unusual conditions noted during the
mother’s prenatal care period? NAME
[Yes [CI No (Ge to question 2)
ADDRESS (Number) (Street)
List below any complications or unusual conditions which were noted. {City or Tom) (Ste) {Zip Code)
Trimester of Visit Complications or Unusual Conditions
NAME
ADDRESS (Number) (Street)
(City or Town) (State) (Zip Code)

 

 

4. Did the mother make any visits to the hospital for postpartum care

 

during the period between her discharge from the hospital and three
months after?

 

[Yes [CJ No (Go to question 5)

 

How mony visits did she have?

 

(Number)

 

2. Was the mother referred to this hospital by a physician or by another Date Condition or Reason

hospital or clinic?

O Ye [CJ No (Ge to question 3)

 

 

 

 

NAME
ADDRESS (Number) (Street)
(City or Town) (State) (Zip Code]

Did the above provide any prenatal care to the mother? [] Yes [| No

 

 

 

 

5. Did the mother receive family planning information,

[Yes [CI Ne
[J Yes [J No
[J Yes [Ne

[C] Don’t Know
[C] Don’t Know
[CJ Don’t Know

(a) during her prenatal period?
(b) during her delivery episode in the hospital?
(c) during her postpartum period?

 

6. (a) If yes to any of the above, did the mother agree to use the family planning information provided?
Yes [No [CT] Don’t Know

 

(b) What method of contraception did she agree to use?

 

HSM-254-5 (PAGE 4)
REV. 7-72

(GO ON TO PART IV)

43
44

 

PART IV. PERSON COMPLETING THIS FORM

 

 

 

 

 

 

NAME
ADDRESS

(Number) (Street)

(City or Town) (State) (Zip Code)
TELEPHONE NO. DATE OF COMPLETION (Mo. Day Yr.)

 

 

NOTES AND COMMENTS

 

HSM-254-5 (PAGE S)
REV. 7.72

#U.S. GOVERNMENT PRINTING OFFICE: 1980

GPJ 930.088

311-240/14

1-3
Series 1.

Series 2.

Series 3.

Series 4.

Series 10.

Series 11.

Series 12.

Series 13.

Series 14.

Series 20.

Series 21.

Series 22.

Series 23.

VITAL AND HEALTH STATISTICS Series

Programs and Collection Procedures.—Reports which describe the general programs of the National
Center for Health Statistics and its offices and divisions and data collection methods used and include
definitions and other material necessary for understanding the data.

Data Evaluation and Methods Research.—Studies of new statistical methodology including experi-
mental tests of new survey methods, studies of vital statistics collection methods, new analytical
techniques, objective evaluations of reliability of collected data, and contributions to statistical theory.

Analytical Studies.—Reports presenting analytical or interpretive studies based on vital and health
statistics, carrying the analysis further than the expository types of reports in the other series.

Documents and Committee Reports.—Final reports of major committees concerned with vital and
health statistics and documents such as recommended model vital registration laws and revised birth
and death certificates.

Data From the Health Interview Survey.—Statistics on illness, accidental injuries, disability, use of
hospital, medical, dental, and other services, and other health-related topics, all based on data collected
in a continuing national household interview survey.

Data From the Health Examination Survey and the Health and Nutrition Examination Survey.—Data
from direct examination, testing, and measurement of national samples of the civilian noninstitu-
tionalized population provide the basis for two types of reports: (I) estimates of the medically defined
prevalence of specific diseases in the United States and the distributions of the population with respect
to physical, physiological, and psychological characteristics and (2) analysis of relationships among the
various measurements without reference to an explicit finite universe of persons.

Data From the Institutionalized Population Surveys.—Discontinued effective 1975. Future reports from
these surveys will be in Series 13. )

Data on Health Resources Utilization. —Statistics on the utilization of health manpower and facilities
providing long-term care, ambulatory care, hospital care, and family planning services.

Data on Health Resources: Manpower and Facilities. —Statistics on the numbers, geographic distri-
bution, and characteristics of health resources including physicians, dentists, nurses, other health
occupations, hospitals, nursing homes, and outpatient facilities.

Data on Mortality. —Various statistics on mortality other than as included in regular annual or monthly
reports. Special analyses by cause of death, age, and other demographic variables; geographic and time
series analyses; and statistics on characteristics of deaths not available from the vital records based on
sample surveys of those records.

Data on Natality, Marriage, and Divorce.—Various statistics on natality, marriage, and divorce other
than as included in regular annual or monthly reports. Special analyses by demographic variables;
geographic and time series analyses; studies of fertility; and statistics on characteristics of births not
available from the vital records based on sample surveys of those records.

Data From the National Mortality and Natality Surveys. —Discontinued effective 1975. Future reports
from these sample surveys based on vital records will be included in Series 20 and 21, respectively.

Data From the National Survey of Family Growth.—Statistics on fertility, family formation and dis-
solution, family planning, and related maternal and infant health topics derived from a biennial survey
of a nationwide probability sample of ever-married women 15-44 years of age.

For a list of titles of reports published in these series, write to: Scientific and Technical Information Branch

National Center for Health Statistics
Public Health Service
Hyattsville, Md. 20782
SHINIEH-108 lied Sotiog

SSVIS YIP pup ity
op wr suonongnd 0g

00€$ '3SN ILVAIHd HO4 ALTYN3d
SS3INISNEG 1V121440

SSVI QHIHL :
Z8L0Z puejAsepy ‘ajjiasneAn

96 M3IH AemybiH 1sapm-15e3 00LE

31151181 Y1|aH JO) 131Ua] |euOnEN

Lo - ABojouyda pue ‘sonsnelS ‘Yoleasay Yijeay JO ad10
M'3'H 40 LN3IWLHV43A 'S'N 221A135 Y1|eay 21nd

adivd S334 ANV 39V1S0Od 3HV4TI3IM ANY 'NOILYINAI 'HLIVIH 40 INIWLHVL3A SN

SHON

   

£8 "ON -Z sauag
LSEL-08 (SH) "ON uoneaiignd M3HA

 
byw
al

7

0) a
4 2
NCHS
a Le]
« &

R/4
CASIN

(TR REID

0) (CT TT TET FL
FSET

U.S. DEPARTMENT OF
HEALTH AND HUMAN SERVICES

Public Health Service
Office of Health Research, Statistics, and Technology

National Center for Health Statistics

 

 

 
 

 

 

 

 

Library of Congress Cataloging in Publication Data

Lunde, Anders Steen
The person number system of Sweden, Norway, Denmark, and Israel.

(Vital and health statistics : Series 2, Data evaluation and methods research ; no. 84)
([DHHS publication] ; (PHS) 80-1358)

Supt. of Docs. no.: HE 20.6209:2/84

1. Scandinavia—Statistical ‘services. 2. Identification numbers, Personal—Scandinavia.
3. Privacy, Right of—Scandinavia. 4. Israel—Statistical services. 5.Identification numbers,
Personal—Israel. 6. Privacy, Right of—Israel. I. Title. II. Series: United States. National
Center for Health Statistics. Vital and health statistics : Series 2, Data evaluation and meth-
ods research ;no. 84. III. Series: United States. Dept. of Health and Human Services
DHHS publication ; (PHS) 80-1358.
RA409.U45 no. 84 [HA37.82] 312'.07'23s 79-607800
ISBN 0-8406-0179-4 [001.4'22]

 

 

For sale by the Superintendent of Documents, U.S. Government Printing Office, Washington, D.C. 20402

 
DATA EVALUATION AND METHODS RESEARCH Series 2
Number 84

The Person-Number Systems
of Sweden, Norway, Denmark,
and Israel

DHHS Publication No. (PHS) 80-1358

 

U.S. DEPARTMENT OF HEALTH AND HUMAN SERVICES
Public Health Service
Office of Health Research, Statistics, and Technology
National Center for Health Statistics
Hyattsville, Md. June 1980
NATIONAL CENTER FOR HEALTH STATISTICS

DOROTHY P. RICE, Director

ROBERT A. ISRAEL, Deputy Director
JACOB J. FELDMAN, Ph.D., Associate Director for Analysis
GAIL F. FISHER, Ph.D., Associate Director for the Cooperative Health Statistics System
ROBERT A. ISRAEL, Acting Associate Director for Data Systems
ALVAN O. ZARATE, Ph.D., Acting Associate Director for International Statistics
ROBERT C. HUBER, Associate Director for Management
MONROE G. SIRKEN, Ph.D., Associate Director for Mathematical Statistics
PETER L. HURLEY, Associate Director for Operations
JAMES M. ROBEY, Ph.D., Associate Director for Program Development

GEORGE A. SCHNACK, Acting Associate Director for Research

ALICE HAYWOOD, Information Officer

OFFICE OF INTERNATIONAL STATISTICS

ALVAN O. ZARATE, Ph.D., Acting Associate Director

Vital and Health Statistics-Series 2-No. 84

 

DHHS Publication No. (PHS) 80-1358
Library of Congress Catalog Card Number 79-607800
FOREWORD

For some years, the proposal to establish a person numbering system in the
United States has been a matter of public policy debate.

The use of a number to uniquely identify an individual, and its application
universally for commercial, governmental and military identification purposes has
obvious advantages from a data processing, epidemiological and health research,
and record linkage point of view; and personal advantages to the individual as
well. Person numbering systems have existed in many parts of the world under a
variety of systems of governments. However, many persons view the establish-
ment of such a system as a real or potential threat to the traditional freedoms en-
joyed by Americans.

At its May 1976 meeting, the United States National Committee on Vital and
Health Statistics recommended that the National Center for Health Statistics pre-
pare a report describing the operation and use of these systems.

This report describes the techniques used in these numbering systems. No at-
tempt has been made to assess the health, social, or economic impact of these sys-
tems. The report is intended to be factual, and makes no recommendations either
for or against the use of person numbering systems in the United States.
ACKNOWLEDGMENTS

This report was made possible through the cooperation of officials of the cen-
tral bureaus of statistics of Sweden, Norway, and Denmark, and the Director of
the Ministry of the Interior of Israel. The associate authors who contributed
manuscripts are:

Ms. Judith Huebner, Deputy Director General, Ministry of the Interior,
Hakirya, Jerusalem, Israel

Ms. Gerd S. Lettenstrom, Chief, Population and Health Statistics, Central
Bureau of Statistics, Oslo, Norway

Mr. Stig Lundeborg, Department for Statistics on Individuals, National Cen-
tral Bureau of Statistics, Stockholm, Sweden

Mr. Lars Thygesen, Chief, Public Registers Section, Danmarks Statistik,
Copenhagen, Denmark

Dr. Anders S. Lunde, Consultant to the National Center for Health Statistics,
wrote the introduction and the last chapter, and edited the final report. The offi-
cials of the countries mentioned contributed to the final chapter.
CONTENTS

 

 

 

 

 

 

 

PP OTRWOTt seerssssnsnsiss tsar sas sa as srs HEIR APPEAL SEA EEA RAPER ROPER SASSO SIRES LE SORA RRR RAY
Acknowledgments
Chapter I. Development of the Person-NUmMbeEr SYStEM ...ccccccesssrnearesicssssscsscssssnsnsssnsassnsssesssssssssssns
Introduction
The Person Number and Its Use
Construction of the PN
Selection of the Countries
Chapter II. The Personal Identification Number in Sweden

Chapter III.

Chapter IV.

Chapter V.

Technical and Other Problems
Future Anticipated Use of the PN System

The Person-Number System in Norway

 

 

Introduction
Operation of the PN System

 

Integration in the Numbering System

 

Use of the PN

 

Social Security Insurance System

 

Health Services

 

Vital Statistics

 

Health Statistics

 

sesessee

 

Solving Problems

 

 

 

Introduction

 

Characteristics in the System

 

 

Organization of the Files
Operation of the System ...

 

Updating of the Central Population Register

 

Use of the Person Number

 

Administrative Use

 

Use in the National Insurance System

 

Use in Vital and Health Statistics

 

Use in Research

 

Problems in the System

 

Future Use

 

Plans for Additional Studies and Research

 

Introduction

Personal Identification Numbers and Population Statistics in Denmark

 

Structure of the Central Population Register

 

General Application of the PN

 

 

Population Statistics
Other Applications

 

Future Plans Regarding the Utilization of Registers

Introduction

#essessssssscscenencsnsressastttssttstssstrsnanene

The Population Registry and the Use of Personal Identity Numbers in Israel ...ccccoeereeenne

 

A Note on the Israeli Use of a Number System

 

Development of the Existing PN

 

 

Use of the PN System
Registration of Births

 

Voters’ Ledger

 

Border Control Card File (Personal) ........

 

—
=
=

=

00 ND

—_ OWN OOOO

—
 

Problems and Difficulties with the System
Plans for the Future

 

Chapter VI. ~~ Advantages and Limitations of the Person-Number System ........ccceesueennee AFTER rI TERRA
Sweden
Norway
Denmark ......
Israel

 

 

 

 

 

Chapter VII. Confidentiality and Privacy
Sweden
Norway
Denmark
Israel

 

 

 

 

Chapter VIII. Other Person-Number Systems in the World Today
Advanced Systems
Finland
France
Iceland y
The Netherlands
Developing Countries
Argentina
Chile
Colombia
Peru
Uruguay
Jordan
Health, Social Security, and Insurance Systems

 

 

 

 

 

 

 

 

 

 

 

 

 

 

United Kingdom
Australia
Other Countries
Other National Systems

 

 

 

 

Federal Republic of Germany

 

 

 

 

 

 

Japan
Portugal
United States

References

Bibliography

Appendixes

I. Construction of the PN in Sweden
II. Construction of the PN in Norway
III. Structure of the PN in Denmark
IV. Construction of the PN in Israel
V. Items in the Swedish Cancer-Environment Register
VIL. Items in Study “Outcome of Successive Pregnancies for Norwegian Women 1967-1979” ...... ioe
VII. Population, Birth, and Death Data

 

 

 

 

 

 

LIST OF TEXT TABLES
A. Examples of Person Numbers: Sweden, Norway, Denmark, and ISrael ....cccceeevneerresseneesssssssessesnsene
B. Characteristics maintained in the Central Population Register Situation File .............. verse

C. Items in Person Numbers in planned or operational national data SYStEMS ......ccceeeeseeererseecsansessnseses

vi

12

40
 
 

SYMBOLS

Data not available------=-reeeeeeememmme cece ---

Category not applicable-----e-eeeemereeeeeeeeeeenee oe

 

Quantity zero
Quantity more than 0 but less than 0.05-—- 0.0

Figure does not meet standards of
reliability or precision----------se-esseeceeeeeeo. —

 

 

 

viii
THE PERSON-NUMBER SYSTEMS OF SWEDEN,
NORWAY, DENMARK, AND ISRAEL

Anders S. Lunde, Ph.D., Stig Lundeborg, Gerd S. Lettenstrom,
Lars Thygesen, and Judith Huebner?

CHAPTER |
DEVELOPMENT OF THE PERSON-NUMBER SYSTEM

INTRODUCTION

By the beginning of the 20th century many
western European nations had established popu-
lation registers that involved the continually
updated systematic registration of the popula-
tion primarily by name, residence, age, sex, and
marital status. Originally, these were local regis-
ters administered at the municipal, county, or
province level. By the middle of the 20th cen-
tury, a movement to centralize the registers had
begun and, while the local systems were kept
intact, data from the local register offices were
provided to Central Population Registers
(CPR’s), which were usually maintained at the
national statistical office.

The primary reason for establishing popula-
tion registers was to maintain reliable informa-
tion for administrative purposes, particularly for
program planning, budgeting, and taxation
records. The registers also were useful for
developing voting and education registers, main-
taining social insurance and welfare files, and

aNational Center for Health Statistics, National Cen-
tral Bureau of Statistics (Sweden), Central Bureau of
Statistics (Norway), Danmarks Statistik (Denmark), and
Ministry of Interior (Israel), respectively.

supporting police and court references. The
registers eventually proved useful for the devel-
opment of population statistics, especially for
studies of internal migration, but also for re-
search in many areas, including health statistics.
Although the product of the early local
registers could be accumulated for national
administrative purposes and statistics, this task
was difficult and time consuming. The punch-
card system, used for large-scale periodic statis-
tics, was often operated through regional
punch-card centers. The development of the
CPR’s coincided with the methodologies pro-
vided by electronic data processing (EDP),
which greatly increased the potential for data
handling, storage, and retrieval, and provided for
national magnetic data-tape registers of the
population. The CPR’s also simplified certain
problems of control and data quality because
the management of the system was centralized.
At first, individual registration forms and the
resultant lists of persons were coded by various
coding systems using name and residence. Num-
bers assigned either randomly or sequentially by
date of entry into a file were also used. Local
registers and other institutions developed sepa-
rate file codes. When CPR’s were created, their
efficient operation required that each individual
be defined by a single, universal code. As com-
puters were used, a universal numbering system
was developed and each individual was given a
Person Number (PN).

Numerals are preferred as the primary refer-
rents or standard identifiers in data-base com-
puter systems, because computer operations are
based on a numerical system and any other
codes must be translated into a number for com-
puter applications. If more than one code is ap-
plied to the records of a single individual, the
chances of error are increased. A unique and uni-
versal number is efficient and easy to handle for
comparing and sorting, storage capacity is in-
creased, and record matching and linking are
considerably improved.

THE PERSON NUMBER AND ITS USE

The term ‘Person Number” refers to the
entire personal identification number used in the
national statistical systems of Sweden, Denmark,
Norway, and Israel. This number will lessen the
confusion caused by using either a “birth num-
ber,” a “personal identity number,” an “‘individ-
ual identification number,” or a “CPR number.”
The commonly used “birth number,” for ex-
ample, may be most confusing because it may
refer either to a random number provided at
birth or to a number that includes a birth date.
The Scandinavian countries also differentiate
between a six-digit number based on birth date
and a three-digit birth number (“individual”
digits). /

The PN, as a computer file key, has the
characteristics of uniqueness, permanence, reli-
ability, and universality. It is unique and inde-
pendent because it is the sole identifier for one
particular individual and for that person alone.
It is permanently assigned, therefore, no dupli-
cation should occur. Although several numbers
relating to an individual might be linked in a file,
each additional item and operation adds to pos-
sible error. The PN is reliable in the sense that
the system provides for such quality controls
that the correct number appears on the named
person’s records. It is universal because it is one
identifiable unit in a universe of numbers.

In a typical population numbering system,
the PN is placed on all pertinent documents in
the national data system. As a basic source, or
data base, the system may act largely as a census
in which persons are defined by a select number
of characteristics or events. This data base is
continually updated by the addition of births,
subtraction of deaths, addition of immigrants,
and the removal of emigrants—a classic demo-
graphic formula. The data base also can provide
information on marriage and divorce and other
personal particulars or events. These data are
required for national assessment and program
planning.

The PN is placed on documents and becomes
a file reference in many clerical operations and
statistical systems in public administration.
These systems often develop their own data files
and, using the PN, they either may update their
information from the CPR reports, or they may
attain direct integration with the CPR to com-
bine the information in their files with informa-
tion from the CPR. Data from censuses and vital
statistics registration systems may be integrated
to achieve current population statistics. Some
countries are planning central registers and
archives with the capacity for integrating many
forms of data for public administration. The
technology to achieve these goals has been
available for some time.

Although the PN systems have become
indispensable for public administration in the
countries in this report, their potential for
important research has not been fully realized.
Sweden, in designating a ready-made survey
frame by selecting the records of all persons
born on the 15th of the month, has been able to
take surveys of socioeconomic importance. Nor-
way and Denmark are engaged in significant
occupational mortality studies by PN linkage to
death, medical, and census records. The National
Institute of Child Health and Human Develop-
ment is supporting a study, remarkable in its
details, entitled “The Outcome of Successive
Pregnancies for Norwegian Women 1967-1976”
(see appendix VI). A unique numbering system
affords the possibility of automatically linking
records from the Medical Birth Registry, which
gives details on pregnancy experiences and births
as well as family histories of diseases, with the
death files, which provide data on the causes of
death of infants under 1 year of age, and the
census file, which provides statistical background
data on the mother and father. No reference to
individuals is made in these studies.

Government agencies and central statistical
bureaus are well aware of the possibilities for
linking statistical information to obtain new in-
sights into areas of fertility, mortality, morbid-
ity, and health services (see appendix VII), but
they are prevented from pursuing these lines of
inquiry because of budget and priority re-
straints. University and institute research groups
undoubtedly will increasingly refer to CPR data
and data from various subsystems. Government
regulations generally permit the use of statisti-
cal (not individual) data for research.

CONSTRUCTION OF THE PN

The PN is constructed in accord with stan-
dard computer practice for checking and correc-
ting errors in the number, which includes the
addition of check digits. The check digits are
calculated by means of Modulus 10 or Modulus
11 algorithms computed on the basis of the
other digits. With one check digit, 1 out of
2,000 errors will pass through the central sys-

tem; with two check digits, 1 out of 100,000 er-
rors will pass through. In the Scandinavian coun-
tries, the PN is constructed with an initial six
digits for day, month, and year of birth; in
Sweden, the order is reversed. Three file num-
bers or individual digits, numbered 001-999,
follow. Odd numbers stand for males and even
numbers for females. In Denmark, the first digit
represents the century of birth. In Norway, per-
sons born in the 19th century are assigned num-
bers 749-500 and those in the 20th century
499-000. Israel has a PN of nine numbers, as-
signed at random and including one check digit.
This number has no cue for age or sex.

Table A gives examples of PN’s. (See also
appendixes I through IV.)

For complete coverage of the population,
the PN is provided at birth and remains with the
individual until death. Even after death refer-
ence can be made to data concerning that per-
son, for example, in a retrospective study of the
outcome of disease. Israel has had the unique
experience of providing a PN to each individual
in a new population and to thousands of immi-
grants since 1948, and it has served as a key to
needed social services. However, the Scandi-
navian countries centralized their numbering
systems only in the last decade. Remarkably,

Table A. Examples of Person Numbers: Sweden, Norway, Denmark, and Israel

 

 

 

 

 

 

 

Country, number, and identification Date of birth Individual Check
digit digit
Sweden
450470-1488 (female Born Apr. 10, THB) ......commnsimmnsriniiorssnmiisssismsmiisisnas 45 04 10 148 8
(year) | (month) | (day)
Norway
260597-65131 (male born May 26, 1897).........ccccivcverririrnnennnnsnsessses ses snseensnns 26 05 97 651 31
(day) | (month) | (year)
Denmark
030836-1171. (male DOrn Jung 3, TO3B) ....ccuviiisivnmissmissrssmmissimmisinssrssas 03 06 36 117 1
(day) | (month) | (year)
Israel!
0871188383 (rogistarot) POISONY....cccvscssismmisrmssessssssssmsissssrssasormssertess crsneassnshiasane 3

 

 

 

1Random number is assigned.
in less than 10 years, their systems have become
firmly established and completely basic to pub-
lic administration.

SELECTION OF THE COUNTRIES

The Scandinavian countries and Israel are
forerunners in the development and application
of a unique PN system in national statistical
systems. In these countries the PN has not only
been used for civil administration, but also has
been extended into several research areas. Nor-
way, Sweden, and Denmark are now engaged in
research in health and related areas. Israel has
used the PN system differently from the north-
ern countries, and its plans for the system are
extensive. Because of these developments, Nor-
way, Sweden, Denmark, and Israel were selected
for study.

Considerable international interest in a PN
system exists, and many other countries have

either established such systems in part or plan to
do so. The question of whether to establish a PN
system has emerged in all developed countries,
and some developing countries have even estab-
lished the system in the laws of the land. The
status of these systems is discussed in chapter
VIII.

The report deals individually with the ex-
periences of Norway, Sweden, Denmark, and
Israel. Although the data processing aspects of
the numbering systems are very similar, the
discussion of each country will differ because
each author emphasized matters peculiar to his
or her country. These areas of emphasis collec-
tively provide a broad overview of numbering
systems operations.

Chapters on the four countries are followed
by evaluational chapters in which the advantages
and limitations of numbering systems, the prob-
lems of confidentiality and privacy, and the
status of numbering systems throughout the
world are discussed.
CHAPTER II
THE PERSONAL IDENTIFICATION NUMBER IN SWEDEN

INTRODUCTION

In Sweden, the administrative functions of
the Government are largely delegated to central
administration boards, each representing differ-
ent areas of competence. One such authority,
the National Tax Board, is concerned with popu-
lation registration and the administration of the
Person Number. Actual registration procedures
are carried out by the Church of Sweden, which
is the state church, by using a system that is
approximately 300 years old. The local clergy-
man is the registrar, and his registration district
is the parish. There are 2,500 parishes in
Sweden. The clergyman also acts as the local
census official. In the local population registra-
tion system, individuals are listed according to
their place of residence in relevant ledgers and
registers. A separate record is kept for each indi-
vidual, is updated annually, and follows a per-
son throughout his life. The records of deceased
persons are kept in a historical file.

Every person living in Sweden on January 1,
1947, was assigned a personal identity number
(PN) in the county of registration. Since then,
each of the 24 county administrations has kept a
register of the resident population. In 1967,
these records were transferred from metal plates
to magnetic tape (personal-data tape). The
county registers contain a person’s name, per-
sonal identity number, birth date, place of birth,
parochial registration locality, and address. The
data also include information on marital status,
date of change of marital status, citizenship,
legal incapacity, and membership or nonmem-
bership in the Church of Sweden. There are also
codes relating to pensions, conscription, mer-

chant navy register, and income tax. Local
parochial registration offices send personal-data
cards containing pertinent changes to the county
administration. The county administration also
supplies data to other bodies such as the local
tax authorities, the Social Security Administra-
tion, and the defense authorities.

OPERATION OF THE PN SYSTEM

The main function of the National Tax
Board is to supervise and organize the work of
the local and regional authorities. It is responsi-
ble for systems design, programming, and the
development of the various routines to be
administered by the regional and local authori-
ties. The National Tax Board issues birth num-
bers ranging from 001 to 929 to the county
administrations to be assigned to newborn chil-
dren. Birth numbers from 930 to 999 form a
reserve series used by the National Tax Board.
The first set of birth numbers is usually suffi-
cient to cover the number of births per day in
each county but, if it is not, numbers can be
taken from the reserve series. When a county
administration is notified of a live birth by the
local parochial registration office, it assigns the
first vacant birth number from the county
administration digit series to the child for the
day of birth. Odd numbers are used for boys and
even ones for girls.

Not only the newborn, but also other
categories can be assigned a PN by the National
Tax Board. One example is the immigrants who
have registered in a parish, who have not
previously resided in Sweden, and who have not
already been given a number. When an immi-
grant registers at the parochial registration of-
fice, the latter informs the National Tax Board
and requests a number. The board checks to de-
termine whether the applicant has been given a
number and, if not, it assigns one. A person who
has been assigned a number at any time in the
past will be reassigned his original number.
Since 1969, PN’s have also been assigned to per-
sons who have come from abroad to work
but who have not stayed long enough to be reg-
istered at the parochial office as living in Swe-
den. The National Tax Board will assign such a
person a number either when it issues him a pre-
liminary income tax return or when he enrolls
in the social insurance scheme. There are other
minor categories of people who may also be
given PN’s for special registration, for instance,
Swedish citizens living abroad who apply for
a Swedish passport and who have not pre-
viously been given a PN.

INTEGRATION IN THE NUMBERING
SYSTEM

The magnetic-tape file of the population
kept at the county administration office is ar-
ranged according to PN’s with the oldest number
issued first and the newest number issued last;
this file is updated once a week. Data concerning
members of a family are linked; a husband’s rec-
ord will contain a note with his wife’s PN and
vice versa. The record of a person under age
18 will contain a note with the PN of the head
of the household. This method enables a combi-
nation of data concerning members of a family
living together. The record for every person in
the file contains an identification number for
the county, the municipality, the parish, and the
real-estate unit where he resides. This number
relates to a special magnetic-tape register of all
the real-estate units in the county. The popula-
tion files and the real-estate files can be
matched, and the information integrated. A list
arranged by all the real-estate units and the
persons inhabiting them is compiled periodically
so that the exact legal domiciles of each individ-
ual can be established for the coming year.

The population file is also used as the basis
for the tax file, which is kept by the county

administratien and contains all significant infor-
mation about taxes and their collection. A
taxpayer’s identity number is the same as his PN
plus the two-digit code of the county where he
lives. In this way the population file and the tax
file can be integrated. For various official
purposes full integration can be achieved be-
tween the population file, the tax file, and the
real-estate file via the PN and the real-estate
identity number.

USE OF THE PN

In Sweden, the personal identity number is
used extensively in public administration. At
the outset, the PN was used primarily as a more
efficient way to identify individuals. Since the
introduction of computers in population regis-
tration and taxation applications, the numbering
system has become a necessary, fundamental
element in the maintenance of an accurate and
efficient public administration system. The iden-
tity number has become the key file number in
various subsystems and is placed on all pertinent
records of the individual. Specific uses are as
follows:

A. General
1. Population registration
Taxation
Military service (Military Register)

Civil defense

ov nN

Social insurance (National Social In-
surance Board)

Ss

Education/school register (local edu-
cation committees, school boards)

7. Health services (hospitals, patient
medical records)

Passports (National Police Board)

Registers of motor vehicles and driv-
ing licenses (National Car Registra-
tion Board)

10. Insurance companies, banks, and
other organizations
In Sweden, other official or non-

official bodies use the PN in their
register of employees, clients,
policy holders, and so forth. In
some cases, the central adminis-
tration permits notification of
changes of address to these
organizations.

B. Health
1. Vital statistics

2. Health statistics

Social Security Insurance System

The social security insurance system
(chiefly, health insurance, parents’ insurance,
national retirement pensions, and supplementary
pensions) is administered by the National Social
Insurance Board and the local social insurance
offices. A comprehensive advanced data process-
ing (ADP) system has been developed for this
administration and is used not only for social
security insurance but also for certain other
allowances and benefits. The ADP system, and
consequently the social security insurance ad-
ministration, is based on the fact that the in-
sured can be identified by their PN’s.

The personal identity numbers serve as keys
or links among different registers and with PN’s
as identifiers, these registers can be brought up
to date quickly and simply. The local social
security insurance offices are continually in
touch with current information in the central
registers via display terminals at approximately
500 central and local insurance offices. In the
terminal communication system, the PN is used
as the identifier.

With this number, data can also be ex-
changed with other authorities, and social secur-
ity benefits can be transmitted simply and
safely.

Health Services

Various patient booking systems that are
now being established in the health services are
based on the PN. When a patient visits a hospital
for the first time, he is issued a patient-
attendance card, which is a metal plate stamped
with his PN, name, and other information. This

plate is used to identify the patient for all
hospital services. The PN is also used as the
identifier on a patient’s medical record—a docu-
ment that the health services are legally obliged
to keep. The PN is also frequently used as an
identifier in various kinds of health services staff
administration registers such as those of physi-
cians, nurses, and paramedical personnel.

Vital Statistics

Every week the county administrations no-
tify the National Central Bureau of Statistics
(SCB) of changes in their population register
concerning births, migration, deaths, and marital
status, and PN’s are used as the means of
identification.

The SCB gathers information in a register
system called the Register of the Total Popula-
tion (RTP) which includes all persons entered on
parochial registers in Sweden, and which is
updated at regular intervals. Apart from PN’s, it
not only contains the majority of the county
administrations’ population data, but also con-
tains certain data on income and taxes. The
main object of this register is to serve as a base
for the production of statistics concerning indi-
viduals. If the Data Inspection Board has given
its permission then it is possible, by using the PN
as a matching device, to carry out the joint
processing of statistical material within the area
and to reduce the number of variables that have
to be collected in certain other sets of statistical
material.

This register also provides a framework for
statistical samples. The individuals selected in
the sample can be linked by their PN’s to any
other available information pertinent to the
survey.

Sweden’s population statistics are based on
the Register of the Total Population system.

The PN serves as the key to matching and
duplicating in the following manner:

To check duplication of notifications
To check notifications of multiple births

To compile tables showing rates for
emigrants according to the duration of
their stay in Sweden (the immigrant

register is matched against the emigrant
register)

To compile tables showing rates for
those who have received Swedish citizen-
ship according to the duration of their
stay in Sweden (the immigrant register is
matched against the register of those
who have received Swedish citizenship)

To compile migration statistics for mu-
nicipalities relating to 5-year periods in
which those who have moved over a
municipal boundary once are recorded
separately and those who have moved
both in and out are recorded separately
(in-migrants are matched against out-
migrants)

To produce migration statistics, for ex-
ample, for 1970 and 1971 from data
from the 1970 Census of Population and
Housing on items such as occupation,
type of economic activity, education,
and income (the 1970 and 1971 migra-
tion register was matched against the
1970 Census of Population and Housing)

The PN’s have been used in censuses of popula-
tions since 1960, primarily as a matching device.
Their use has steadily increased as more infor-
mation from the censuses has been gathered
from the register material. Consequently, the
number of facts collected directly from the
general public has been reduced.

The PN’s are used as matching devices in
other areas of population statistics in Sweden,
for example, the SCB’s Register of Deaths and
the Register of Persons Born on the 15th of the
Month. The latter register comprises about 3
percent of the population. The register contains
demographic data (including migration) pertain-
ing to those born on the 15th and to the
husband or wife and children living with the
individual. In addition, it contains data relating
to income and economic activity (according to a
rough breakdown) for the individual and his or
her spouse. The register has been used chiefly
for sample surveys of fertility, migration, and
changes in income at both an individual and a
family level, as well as a sampling frame for
interview surveys.

The Register of Deaths contains data from
the 1960 Census of Population and Housing and
from the annual mortality data for 1961-70.
Data on demographic features, occupation, eco-
nomic activity, educational level, and so forth,
from the census are linked to mortality data
such as place of residence at the time of death,
date of death, and causes of death. This linkage
enables both the combination of data and the
study of mortality in any population that can be
defined by means of census data. The Register
of Deaths is expected to be of great interest in
epidemiological studies.

Health Statistics

The PN’s are used as identifiers in a number
of registers in the public health sector. They are
used as matching devices in the various registers
to enable data supplied on different occasions
regarding the same individual to be interrelated.
The registers have many uses including statistics
in special surveys. The National Board of Health
and Welfare is primarily responsible for Swe-
den’s central health statistics. The PN’s are used
in the following major registers among others:

1. Statistics on patients

2. Notifications on the medical aspects of
births

3. The Cancer Register

4. The Register of Gynecological Medical
Examinations

5. The Register of the Side-Effects of Me-
dicinal Drugs

Statistics on patients.—Statistics on patients
are compiled mainly to obtain information for
administrative planning. The primary data con-
sist not only of the patient’s PN, but also of
particulars concerning medical diagnoses and
operations when the patient was in the hospital.
The PN’s permit (1) the identification of the
data record so that supplementary material may
be obtained and particular items can be checked
and corrected, (2) the computerized linkage of
several different periods when a patient received
medical attention, and (3) the computerized
supplementation of data from other registers.
Furthermore, the existence of PN’s increases the
opportunities for the use of statistics in special
surveys, for instance, epidemiological research.
The statistics on patients have another function
regarding the side effects of medicines: to draw
attention to the adverse reactions to drugs
because information is accessible from patients’
medical records. The PN’s facilitate such a
search.

Birth Notification.—The medical aspects of
births are registered partly to provide a founda-
tion for medical statistics of deliveries, as a
planning base. This form of registration also
enables retrospective use of data to detect
groups among the newborn that are at risk. The
register contains numerous medical data con-
cerning the mother’s state of health during
pregnancy, the course of the delivery, and the
state of health of the newborn child. The
mother is identified in the register by means of
her PN. The newborn child’s number is not
added until the register is updated which is
about 8 months after the register year has
expired.

Cancer Register.— The Swedish Cancer Regis-
ter (see appendix V) was established in 1958. In
1978, permission to supplement the register
with data from the Population and Housing
Census was granted by the Data Inspection
Board. Persons registered in the 1961-73 Cancer
Register and the 1960 census were identified in
both registers by their PN’s, and the Cancer-
Environment Register was compiled by merging
both registers.

Although the Cancer-Environment Register
is primarily intended for research, it also has
another function, which is to point out potential
health hazards. This register has been widely
used as a gateway to various research projects.
Recently it has proven to be of major impor-
tance in investigations of the relationship be-
tween the occupational environment and the
origin of tumor diseases.

Data on a cancer case are collected from
three different sources: (a) clinics, (b) patholo-
gists/cytologists, and (c) mortality statistics. The
PN is not always available on records from the
first two sources. Compilations of data from
clinics and pathologists/cytologists are made
manually with the individual’s birth date and
name as identifiers. The PN’s are later checked

and supplemented by comparing them with a
microfiche register of the total population of
Sweden. The data are then registered by com-
puter, and the data from the mortality statistics
are transferred by joint computerized processing
with the PN as the identifier. The register is
updated continuously when there are additions
or corrections. The PN is used as the identifier in
updating the register. For identification pur-
poses all or part of a name is also registered. The
name is necessary, for instance, in cases where it
has not been possible to obtain complete per-
sonal identity numbers.

Gynecological Medical Examinations Regis-
ter.—To supplement the Cancer-Environment
Register, the National Board of Health and
Welfare keeps a register containing data from
gynecological examinations under the aegis of
the county councils. Those asked to submit to
these examinations are selected from the popula-
tion registration system. The use of PN’s in these
examinations enables monitoring a woman’s
state of health regarding cancer and making fol-
lowup assessments.

Medicinal Drug Register.—The Register of
the Side-effects of Medicinal Drugs should pro-
vide information speedily about side-effects that
may have been caused by a certain medicine. To
enable patients to be followed scientifically,
their PN is necessary.

The PN’s are also used in registers and
surveys carried out by bodies other than the
National Board of Health and Welfare, such as
the Tuberculosis Register, which is kept by the
National Association for Cardiac and Pulmonary
Diseases, and the registration by the National
Bacteriological Laboratory of notified cases of
diseases that are a danger to the general public.

TECHNICAL AND OTHER PROBLEMS

Apart from the problem that the system is
not entirely self-checking, no serious technical
problems are apparent. However, immigration
from countries having an imperfect population
registration has given rise to a problem for a
system with PN’s based on dates of birth. If an
immigrant does not know his date of birth, he is
assigned a PN based on an estimated date of
birth. If such a person emigrates from Sweden
and then immigrates into the country again,
there is the danger that he may be assigned a
new PN. Another problem is that people from
countries with imperfect population registration
systems who do not know their date of birth
often say they were born on January 1st. The
result may be both confusion between individ-
uals and an uneven distribution in registers of
individuals indexed by PN’s.

The problems that arise in the use of the PN
are those related to the statistical processing of
material where PN’s have been used for identifi-
cation. Such use of PN’s became firmly estab-
lished in various registers and surveys during the
1960’s. One reason was that these numbers
made it possible for various sets of material to
be processed together and, therefore, for better
utilization of information. Respondents no
longer had to be asked for the same information
repeatedly. The production of statistics could be
made more efficient.

Three elements surfaced when this new
method of statistical production began to be
implemented: (1) members of the general public
either did not know their own PN’s or they gave
the wrong or incomplete number due to a
memory lapse; (2) the quality of the number
was not satisfactory; and (3) not everyone had
been assigned a PN, for example, foreign stu-
dents studying for a short period in Swedish
schools.

In the 1960 Census of Population and
Housing, PN’s were used, and errors sometimes
arose that prevented a correct linkage between
data from the national registration and data
from the special Census of Population form.
Consequently, the national registration data for
one individual might be entered with the popula-
tion data of another. Similar errors arose when
these data were being processed jointly with
other sets of material. These errors were a
combination of circustances: an individual,
when filling in his statistical form, might copy
his PN incorrectly or errors might remain from
1947 when PN’s were introduced for the popula-
tion of Sweden. At that time the check-digit
system had not yet been introduced.

Over the years, the majority of the errors
occurring in PN’s were corrected by the popula-
tion registration authorities, and the check-digit

10

system was introduced. Because of the wide-
spread use of PN’s in administrative material,
members of the general public have become
increasingly aware of the value of remembering
their number and of being able to look it up
easily. Consequently, errors in these numbers are
now very rare compared with the 1960’s.

Solving Problems

Regarding problems that may arise in a
statistical survey due to errors in PN’s, such
errors can now be rectified in various ways. The
manner of discovery of the error must be
determined. Most errors are discovered by the
check-digit system when the material is being
registered by computer. Others may be dis-
covered in the actual survey process, for in-
stance, in an interview survey where the survey
population and certain basic data are gathered
from previously collected material using PN’s.
When matched in the computer, the data of
those selected for the survey are supplemented
by an individual’s name and address from a
register that is kept by the population registra-
tion authorities. The particulars are then sent to
the various addresses. If an individual finds he
has been sent incorrect data, he will usually
inform the sender. The identity of the person
who was originally selected can be traced by
analyzing the data in the survey where the
selected individual gave the wrong PN. If that
survey contains additional particulars to identify
this individual, he can be found. This work,
however, is time consuming and may not always
be feasible because of a shortage of resources.
Therefore, the errors may, instead, reduce the
number of individuals in the survey by leading
to nonresponse. Nevertheless, this problem is
not insurmountable.

A different way of solving the error-problem
is to use additional identification devices when
processing the material. However, this course is
not taken frequently in statistical surveys; it is
chiefly used in administrative contexts where
errors in the joint processing of registers may
have unfortunate effects in personal affairs, for
instance, if children’s allowances, pensions, and
so forth are sent to the wrong person.
FUTURE ANTICIPATED USE OF
THE PN SYSTEM

The National Tax Board and the National
Office of Organization and Management (see
chapter VII for a more detailed explanation of
the National Tax Board and the role of the Data
Inspection Board) are developing proposals for a
new ADP system for the registration of the
population, taxation records, and tax collection
in the future. Parliament has decided that these
proposals should be implemented. The new
system is to be adopted in stages. The first stage,
which is concerned with taxation records and
tax collection, was ‘introduced -in 1979. A
committee to examine the question of a future
census for population and housing was estab-
lished and charged to examine the feasibility of
using more extensively data found in various
administrative sectors such as the administration

of taxes and the registration of real estate.
Because there are many registers with the same
PN in the country, it was thought possible to
base the collection of data on the information
available in them. The committee completed its
work at the end of 1978 and made recommenda-
tions regarding the use of registers concerning
the 1980 census. Although the supply of infor-
mation from the existing registers could be
considerably expanded, it was not sufficient to
provide all data needs. In the spring of 1979,
Parliament agreed on the plans for the 1980
census by largely following the recommenda-
tions of the committee.

The work of studying the feasibility of using
registration data will continue. The National
Central Bureau of Statistics has been charged
with determining ways to obtain data other than
through the collection of census forms, such as
through population registers and data linkage.

13
CHAPTER III
THE PERSON-NUMBER SYSTEM IN NORWAY

INTRODUCTION

As in Sweden, local recording of births,
deaths, and marriages has been undertaken in
the parishes of Norway since the 17th century.

Local population registration offices were estab-

lished only in the 20th century; the first was in
Oslo in 1906. The principle functions of these
offices were to administer taxes and to process
information on where people lived and where
they moved. During World War II, on March 1,
1943, the occupation authorities ordered all
municipalities to establish registration offices.
After the war, the Act of November 15, 1946
(revised in 1970), maintained the practice of
universal registration. On January 1, 1965, the
registration offices were centralized under the
tax authorities and administered by the Central
Bureau of Statistics.

During the 1950’s the growth of electronic
data processing (EDP) led to an increased
interest in developing computerized statistical
file systems. Public institutions developed such
systems and frequently used personal identifica-
tion numbers as key identifiers. These systems
were not uniform, and the reference numbers
differed. The need for a standardized national
numbering system for efficient public adminis-
tration was recognized. The Central Bureau of
Statistics in 1961 was asked to establish and
maintain a national identification numbering
system, and the Central Population Register
(CPR) was established October 1, 1964.

At first, the register was composed of all
residents of Norway on November 1, 1960, the
date of the 1960 census, and all persons who
immigrated or were born between November 1,
1960, and October 1, 1964. Since the latter

12

date, all children born in Norway and all persons
immigrating to Norway for the first time are
assigned a PN. :

Norway maintains another major register,
the Register of Establishments and Enterprises,
started in 1956 and based on a 1953 census of
industrial establishments. Two series of identifi-
cation numbers of six digits and one check digit
are used in the files. Partial registers for indus-
trial establishments and statistics of accounts are
produced from the central register, as well as,
statistics of wholesale and retail trade.

CHARACTERISTICS IN THE SYSTEM

The CPR is kept on magnetic tape. The
record length is fixed. Certain characteristics are
kept in the situation file (see table B).

Table B. Characteristics maintained in the Central Population
Register situation file

 

 

 

 

  
   
  
 

Number
Characteristic of
digits
Identification nUMber........c.ccoevviiicinniiniinne nes i
Municipality wa 4
NBIC sicninirismrisssmmimsmmmsisstiersress ss stsssivsss satassaiasans 26
Address (name of street, road, etc., and number)..... 30
POSTE AISERIC . covniianninuimanissssmmmmmnsasts serosa REE 42 4
Type of registration (resident, deceased, etc.) ... 1
Marital status .......ccocevenrnrreeeieenennnenn. 1
Identification number of mother .... 11
Identification number of father ...... 1
Date for type of registration ........... - 6
Identification number of spouse ..... views 1
Family Umber cousins evi 11
Date of removing day..... oie 6
Nationality......cceeeeriereccnnnnes ioe 3
Migrated FrOM/L0 COUNTLY ...ccororemresrsessnmsssssermsssrsnssens 3

 

 
ORGANIZATION OF THE FILES

The CPR is organized into five files:

1. The situation file contains the acutal
values of the characteristics for each
individual on a given date.

2. The report file contains all entries about
birth, migrations, changes, or corrections
in characteristics after a given date. It
also contains the dates of these events.

3. The chronology file contains the most
recent value of each characteristic as well
as the date of the latest change in the
characteristic.

4. The history file contains the old values

of characteristics that have changed as
well as the dates of these events, and has

the same formal structure as the report
file.

5. The statistic file contains data from
various reports that are used to produce
statistics. Some data are taken from the
chronology file, but most are taken
directly from the reports.

OPERATION OF THE SYSTEM

The registration office keeps a written card
for each resident in the municipality. This card
contains the same data as the CPR. When a
person moves from one municipality to another,
he or she has to notify the authorities within 8
days. The move is registered in the municipality
to which the person has moved, and the registra-
tion office in this municipality requests this
person’s card from the registration office in the
municipality from which he or she has moved.

The registration offices get reports from
hospitals, clergymen, local authorities, and indi-
viduals about births, deaths, marriages, migra-
tion, and other events. The registration offices
update their own registers, supply the reports
with necessary identification numbers, and send
reports to the Central Bureau of Statistics once a
week. The registration offices also keep a sepa-
rate register on magnetic tape. This register
contains some characteristics that are not in-

cluded in the CPR such as information on
occupied houses. The registration offices have a
number to identify every house in the munici-
pality. This number provides information used
in local and regional planning. This local register
is updated once a week by separate reporting
routines.

UPDATING OF THE CENTRAL
POPULATION REGISTER

Each week the Central Bureau of Statistics
receives reports from the local registration offi-
ces concerning the following items:

1. Births

. Marriages

. Deaths

. Internal migration

. Movements within the municipality
. Immigration

. Emigration

. Inquiry about registration problems

. Missing persons

OS WOW 00 uN OO Ot op» WW N

—

. Paternity to children not born in mar-
riage

11. Name

12. Adoption

13. Separation/divorce

14. Changes in the family number

15. Correction of characteristics in the CPR
16. Alteration in names of streets/roads

17. Naturalization

18. Changes in marital status

These reports are divided into three different
categories: new units in the CPR, alterations in
the value of characteristics in the CPR, and
corrections of erroneous characteristics. When
receiving these reports, the Central Bureau of
Statistics makes a brief check on the material,
and any revisions are keypunched. Then, various
logical controls are performed on the material.
Once a month a chronology control is per-
formed, and the CPR is updated.

USE OF THE PERSON NUMBER

The national PN system is used in adminis-
trative routines, in the production of statistics,
and in research projects.

Administrative Use

Several public authorities and institutions
use or have reference to the numbering system.
Examples of such organizations are the taxation
authorities, the Health Administration, the
National Insurance Administration, the Director-
ate for Seamen, the Road Administration, the
Defense Department, and the Administration of
Elections. Also, some private institutions
(mostly banks and insurance companies) are
permitted to use the PN system.

Use in the National Insurance
System

The PN is now used in all parts of the
National Insurance System. This system com-
prises old age pensions, disability pensions,
benefits for occupational injuries and rehabilita-
tion, benefits during sickness, survivor’s benefits,
and family allowances. This system is financed
partly by the national budget and partly by pre-
miums paid by employers and employees. The
insured person pays his premium through taxa-
tion, and the amount stipulated is a certain per-
centage of his income. The amount paid to the
insured either as a pension or as benefits during
sickness is related to his income.

In the old age pension and disability pension
plans, a person is given points for each work
year based on his income. Points are stored over
years and identified by the PN. At retirement,
the pension is fixed on the basis of points for
the 20 most favorable years. For persons not
attaining points for 20 years, separate rules
apply. Separate rules also relate to disabled
persons.

On July 1, 1978, a revised system was
introduced for payment of benefits during sick-
ness. Regulations state that the employer shall

pay for the first 2 weeks of sick leave. If the
duration of sickness (absence from work) ex-
ceeds this period, the insurance office shall pay.

Control of cases when benefits during sick-
ness are given is possible because a register has
been established comprising employers and em-
ployees. Employees are identified by the PN in
this register, and employers are given a separate
employer’s number.

Benefits during sickness are taxable income
and are also regarded as pension-producing
income. Notifications sent from the insurance
system to the taxation office are identified by
the PN.

Use in Vital and Health
Statistics

Reports used for updating the CPR are also
used in the production of statistics, especially
vital and health statistics. Most of these data are
taken directly from the reports after they have
been processed in the register system and have
been transferred to the statistical file. In addi-
tion, the reports are supplied with data ex-
tracted from the register by record linkage.

In internal migration statistics, marital status
and municipality from which the person has
migrated are taken from the CPR.

Extensive use of the birth-number system is
made in producing birth and death statistics. For
births and deaths, two sets of reports are
collected: one contains information primarily
for civil administrative use, and another contains
medical information. Although civil reports on
births are first sent to the local population
register and then to the Central Bureau of
Statistics, the medical reports are sent to the
Medical Birth Registry established in 1967. To
supply this medical register with the birth
number for infants and parents, records on tape
from the CPR are linked each month with the
medical reports related to the mother and child.
In addition to transferring identification num-
bers to the medical register, this linkage also
reveals cases where civil or medical reports are
lacking. This control is important for birth
statistics because dead fetuses and live-born
infants who die shortly after birth are not
always reported in the civil registration system.
In the production of death statistics, medical
reports and civil reports are linked by birth and
death dates. Through this process, lists are
prepared mechanically on nonmatching death
records. In addition, duplicate records are listed.
After checking the lists (and correcting the
errors), medical reports that are missing are
obtained to complete the data on the causes of
death. The CPR is also supplied with deaths that
are not reported on the population register. With
this process, a complete file on deaths can be
produced as an accurate basis for death statis-
tics. The complete file on deaths is linked with
the CPR to obtain the name of the deceased.
This register file is used for production of a na-
tional death index that can be used for death
clearance by all institutions.

Once a year, a situation file of the CPR is
used to produce tables showing population by
sex, age, and marital status in each administra-
tive unit. By using the PN, sex and age can be
directly obtained. Furthermore, tables are pro-
duced separately for the foreign population
based on the characteristic for citizenship. The
PN, which also is the basis of the family-number
system, makes it possible to link persons with
family nuclei. Family statistics have been pro-
duced twice (1974 and 1977) and will now be
produced every other year. The birth number
will also be used to add data on education and
income to the family records.

Moreover, the PN is used in the registers
established for physicians and dentists. By link-
ing these registers with the CPR, data on marital
status, residence, and other information can be
added and used for statistics. These two registers
are also linked with the taxation file to produce
income statistics for physicians and dentists.

Use in Research

The purpose of establishing a history file was
to provide research units with the possibility of
developing studies, such as those based on
cohorts, showing how demographic characteris-
tics have influenced demographic events.

The Central Bureau of Statistics has started
work not only on projects concerning migration,
but also on projects monitoring changes in
marital status for population projections. Fur-
thermore, cohorts of marriages have been fol-

lowed regarding the number and the spacing of
births.

More frequently, the PN system has been
used in research projects based on record linkage
of data from different sources. The population
census of 1970, which includes very detailed
information on education, occupation, housing
conditions, and other specifics, has been linked
to demographic as well as economic data.

In the field of vital and health statistics,
several items were analyzed on the basis of
record linkage.

Divorces for the period 1971-73 were linked
with characteristics of the married couple at the
time of the census and with data from the
taxation file to show the influence of occupa-
tion, housing conditions, education, and income
on divorce. The results were published by the
Central Bureau of Statistics in 1975.1

Cause of death from the censuses from 1970
to 1973 inclusive was linked with occupational
data. Standardized rates showed marked differ-
ences in mortality according to occupational
group. The results were published by the Central
Bureau of Statistics in 1976.2

To extend the basis for calculating mortality
for occupational groups, mortality data were
linked with occupational data for both 1960 and
1970. Deaths among persons economically ac-
tive in 1960 but not in 1970 were included. For
special groups, such as persons who remain in
the same occupational group for a long period,
more reliable data will be presented in the
future. The tables processed will be analyzed
and published under the title “Occupational
Mortality.”3

Record linkage with census data has also
been undertaken for research units outside the
Central Bureau of Statistics (see appendix VI).
In these cases the linkage is made in the Central
Bureau of Statistics and only tabular results are
available. For instance, data from the Cancer
Registry have been linked to occupational data.
A recent analysis was made by linking medical
birth register data (perinatal deaths, preterm
births, and congenital malformations) to census
data (occupation and education of mother and
father). Possible connections between working
conditions and related factors cencerning pa-
rents and conditions during pregnancy, at deliv-
ery, and future developments in the infant were

15
traced. The results will be published in one of
the official statistical series in 1980. Because the
PN is included in the 1960 and 1970 censuses,
record linkage has been undertaken to follow
the resident population at both censuses for
changes in characteristics.

Problems in the System

Technical problems.—When the Central Bu-
reau of Statistics was asked to establish a
national PN system it was assumed that the
number would contain the date of birth as a
memory aid. However, in establishing the regis-
ter, a number of errors in the date of birth were
introduced. When these errors were discovered,
the persons involved were given new numbers.
This occurrence caused some technical problems
when matching files from different periods.

The PN per se contains three individual
digits. Three digits can provide 999 individual
numbers on a specific date, of these, 250 are
used for persons born in the 19th century and
250 are stored for future use. Because a large
number of immigrants give their date of birth as
the 1st of a month, in particular January 1, the
available individual numbers may be too few.

A general problem connected with storing
data—the problem of converting data from one
system to another—has caused additional work.
In the CPR, the family number for each family
nucleus is the birth number of the head of the
family. Thus, when the value of the characteris-
tics in this birth number changes, the value of
the characteristics in the family number also
changes.

Other problems.—A better CPR is increas-
ingly needed. Working continuously to improve
the quality of the characteristics data is one way
to meet the demand for better information.
Another is to improve the means of updating the
register. This method has proved difficult be-
cause the comprehensive control system to
update the reports is overly complex. The CPR
should be updated once a week instead of once a
month.

One special problem refers to migration.
Everyone is required to report a migration to the
local registration office within 8 days. However,
some people do not report them.

16

The time from the occurrence of an event to
the time the report is received in the CPR is
often too long. Simplified and rational report
routines are being developed to make this
interval shorter. Also, possibilities for better
EDP-routines are being considered.

The CPR contains only one concept of
address, that is, the address where the person
lives. Some authorities and institutions use a
postal address such as a postal box number,
which can differ from the home address. Techni-
cally, several addresses can be established as
characteristics in the CPR.

Training the staff of the local registration
offices is needed. The work in these offices asso-
ciated with the CPR is growing more compli-
cated, and the responsibility of each employee
has expanded. To solve these problems the Cen-
tral Bureau of Statistics regularly organizes
training meetings with the staff of the local
registration offices. Over 4 years, all the local
registration offices will have participated in
these meetings.

All the problems in the CPR can be solved if
resources are available. Plans for solving the
problems are set up as part of the long-term
planning program.

Future Use

A broader use for the PN system will depend
on three conditions: the development of the
CPR, the expansion of the statistical file system,
and the rules that the political authorities will
make for the collection and use of personal data.
The development of the CPR can proceed in two
ways: first, by establishing completely new
characteristics in the register, and second, by
technically developing the register (e.g., putting
the CPR on direct-access media and providing
terminals to the local registration offices).

For the expansion of the statistical file
systems, the central authorities are planning a
new countrywide file system and its core will
consist of the following three registers: a central
estate register, a central building register, and a
central address register. This register system will
be linked to other central registers, such as the
CPR. The main purpose of this statistical file
system is to create an information system for
local and regional planning that will give new
and wider opportunities for the use of the CPR.

More private and public institutions are
expected to use the PN. This usage will provide
the increased possibility of linking registers and
will create further demands regarding the quality
of characteristics in the CPR.

Plans for Additional Studies
and Research

The establishment and development of regis-
ters based on the PN system will certainly

increase the possibilities, not only for more and
better statistics, but also for studies based on
record linkage. A register on health personnel is
being implemented. This register now comprises
physicians, authorized nurses, auxiliary nurses,
and some special groups of nurses. All categories
are identified by a PN. In information systems
on patients in health institutions, the use of a
PN will make it possible to follow persons
through the health service system.

When the population census of 1980 is
completed, a followup of the analysis of mor-
tality and occupation will be undertaken.

17
CHAPTER IV

PERSONAL IDENTIFICATION NUMBERS AND POPULATION
STATISTICS IN DENMARK

INTRODUCTION

In 1924, for reasons similar to those in
Sweden and Norway, namely for the collection
of taxes, police work, registration of electors,
and population statistics, Denmark’s Parliament
passed the Act on Local Population Registers.
All municipalities were mandated to establish
local population registers, which would be files
containing information about all persons resid-
ing in the municipality indexed by name, occu-
pation, date and place of birth, residence, family
circumstances, and citizenship. The municipal-
ities were required to keep the files continually
updated by information from various agencies
on births, deaths, marriages, and divorces, and
individuals were responsible for notification of
changes of address. The establishment of a
central register was debated, but it was not
enacted. Eventually the population registers
developed other functions; tax authorities and
health insurance funds, for example, were noti-
fied about migrations of persons. These agencies
also kept their own registers.

Electronic data processing (EDP) was intro-
duced in 1968, and a Central Population Regis-
ter (CPR), with a national magnetic data-tape
register of the Danish population was established
and included all the local population registers in
one administrative system. The process involved
the introduction of the personal identification
number (PN). The CPR numbering system was
extended to all sectors of public administration
and replaced all other systems. In 1972, the
system was extended to include Greenland. The
CPR therefore contains information about all

18

present and past residents in Denmark since
1968 and Greenland since 1972.

The CPR also contains information about
nonresident persons who have some connection
with Denmark, such as payment of taxes. The
CPR only holds current information. Thus for
changes of address, information related to the
old address is deleted from the register. How-
ever, persons who either emigrate or die remain
on the register with the data that were recorded
either at the time of their departure or their
death. A special historical register is also main-
tained and consists of about 6 million persons.
Technically, the CPR today is stored in a
data-base system with direct access to individual
information. Keys for direct access are either the
PN or the address. The register is updated 4 days
a week.

The CPR is administered by the Secretariat
for Personal Registration in the Department of
Interior. One responsibility of the Secretariat is
to instruct the administrators of the 294 local
population registers on questions concerning
updating and communicating with the CPR.
Also, the Secretariat is responsible for keeping
the technical system operational. The register is
located at I/S Datacentralen, which is a publicly
owned data-processing center. The responsibility
for the coordination of the total statistical
production rests with Danmarks Statistik, the
Central Bureau of Statistics of Denmark. This
office cooperates in the production of data with
several authorities such as the Secretariat for
Personal Registration, the National Health Serv-
ice, local tax authorities, research institutes,
and various government agencies.
STRUCTURE OF THE CENTRAL
POPULATION REGISTER

The CPR system is composed of several sub-
registers including the person and the address
registers. The former is the nucleus of the
system, containing information on the pre-
viously mentioned 6 million persons identified
by the PN. Personal data include name, address,
place of birth registration, citizenship, as well as
references to relatives (children, parents, wife, or
husband) by their PN. Person Numbers are
assigned to newborn children and corrections of
wrong numbers are made automatically by
updating the CPR with the correct date of birth
and sex. Immigrants entering Denmark for the
first time are provided with a PN. If they do not
know their date of birth, they are assigned a
year of birth, and the day and month of their
registration are added. Between 1,000 and 2,000
persons disappear each year; most are foreigners
who are presumed to have returned to their
native lands. Others are mainly Danish citizens
who will probably appear at some later time.

GENERAL APPLICATION OF THE PN

The PN is used in public administration as a
file number that serves as an accurate identifica-
tion of individual persons. It is used in connec-
tion with all matters when the public authorities
are involved: tax cases, payment of social
benefits, hospitalization, admission to schools
and institutions, purchase of real estate, military
service, and so forth. By means of the PN the
tax authorities, the social administration, and
other offices can retrieve general personal data
from the CPR system. Many branches of the
administration have constructed special EDP
registers that, apart from the CPR data, also
contain personal data relevant to the various
fields of administration. These special registers
also use the PN as the file number, and the
registers are constructed so that the general
personal data are updated by automatic reports
from the CPR. It is of special importance that
the CPR can supply information about increases
and decreases within the population group with
which the administrative authority is concerned.

The special registers contain data different
from that of the CPR because they are not
meant for general use in administration and the
specialized data seldom have the same degree of
reliability. The more important public registers
are the Tax Register, the Pension Registers, the
Unemployment Registers, the Social Registers,
the Hospital Registers, the Registers of Persons
Receiving Education, the Registers of Real
Property, and the Central Register of Enterprises
and Establishments. Within the health sector, a
number of registers make use of the PN.
Hospital registers, just mentioned, store informa-
tion about patients and their medical histories.
In some counties, these registers are on-line
systems used directly by the hospital staff in
connection with patient treatment.

The Sickness Insurance Register, also kept
by the counties, records information on insured
persons (all adult residents in Denmark) and
their health services. The National Health Serv-
ice keeps registers on legal abortions, births,
and deaths, largely used in medical statistics.
Two specialized medical registers are the follow-
ing: (1) the Cancer Register, where %2 million
living and dead cancer patients are recorded, and
(2) the Heart Register, where patients with acute
coronary thrombosis are registered. These re-
search registers are almost exclusively used for
statistical analysis.

- The PN, serving as the key to all systems,
makes it possible to link data from different
sources and subsystems. Only the public authori-
ties can acquire information about the PN’s and
receive personal data from the CPR. However,
the employers must know the PN’s of their
employees because they deduct withholding
taxes. Another exception is that banks register
the CPR numbers of all their deposit account
customers. This legality makes it possible for the
authorities to exercise an effective control over
the taxpayers’ information to the tax authorities
about bank deposits and interest income.

Population Statistics

The basis of the statistical use of the CPR is
the current population statistics, which are the
statistics on population size, composition, and

19
changes in births, deaths, migrations, and so
forth.

Surveys of the total population and its
composition are normally made once a year on
January 1. The basis of the surveys is a complete
copy of the CPR’s register of persons and
addresses compiled at the turn of the year by
electronic data processing at Danmarks Statistik.

However, the delay in reporting births,
deaths, and migrations is usually about 3 weeks
from the time the change occurred to the time
the information is properly recorded in the
register. The population register will, therefore,
never correspond to the resident population.
These delays are taken into account because the
population count recorded in the local popula-
tion register on January 1, is adjusted regarding
births, deaths, migrations, and so forth, that
have taken place before January 1, but which
were reported within the first 1% months of the
new year. By this period it is ensured that all
reports, which are not subject to extraordinary
delays, are included in the statistics.

The statistics of population changes are
based on magnetic tapes with update records
that Danmarks Statistik receives from the CPR.
These change extracts must be subjected to an
initial processing at Danmarks Statistik, when
the change must be established in each individ-
ual case. It must be decided whether a change
has actually taken place, or whether it is only a
correction of wrong data. When processing
change extracts, delays in the reports must be
noted. In the current statistics, changes are
transferred to the next period (year or quarter
of a year) if they are reported later than 1%
months after the end of the period.

The birth statistics, except for the CPR data,
are based on midwives’ reports that are com-
pleted for each live birth in Denmark. The
midwives’ reports are not systematically pro-
vided with the child’s PN, but both reports are
given the mother’s PN, which serves as a
secondary identification so that all items of
information in the midwives’ reports can be
linked automatically with the changes reported
from the CPR.

Accordingly, the death statistics are pre-
pared by a combination of data from two
sources, namely CPR data and the original death

20

certificates completed by physicians. The PN of
the deceased is stated on both sources and is
used for an automatic linkage of the data.

Certificates and reports are included as a
supplementary basis because medical informa-
tion about birth (for example, the inducement
of labor) or about death and its cause are not
contained in the CPR. The demographic vari-
ables such as sex, residence, and occupation
theoretically could be taken from the certifi-
cates and reports, but by applying the demo-
graphic variables from the CPR, data handling
can be reduced. Also, the birth and death
statistics are coordinated with the other CPR-
based population statistics. This system is most
desirable because the number of deaths either by
geographical area or by special industries should
be seen in relation to the size and composition
of the population in that same group.

Other Applications

The PN is extensively used for statistical
production where personal data from various
sources are combined. As previously mentioned,
such matchings are currently conducted in popu-
lation statistics, but the possibilities of person-
oriented matchings are numerous. Studies mak-
ing use of this technique are described as
follows.

Occupational mortality.—This study demon-
strated differences in the mortality between
various occupational and industrial groups. In-
formation about occupation at the time of death
is not very useful for such a study. Persons often
leave the occupation that they held throughout
their economically active life because of old age
or poor health. These occupations, however,
may have influenced their state of health.
Mortality probably is higher in occupations that
are strenuous and injurious to health.

The study, therefore, followed the popula-
tion that was registered at the census in 1970
whose employment status was recorded. In
addition to the population census material,
including PN’s, the systematized CPR material
is computed into the population statistics and
provides information about decreases in the
population resulting from deaths, migrations,
and other causes during 1970-75. Medical state-
ments from death certificates were also used in
this study to calculate the cause-specific mortal-
ity rates for the various industrial groups.

Census population statistics.—In Denmark,
population and housing censuses are tradition-
ally taken every 5th year, and accordingly a
population census should have been taken in
1975.

Because of the cost and inconvenience to the
population, the census was not taken. Instead
planning data corresponding to population cen-
sus statistics were obtained from existing regis-
ters.

The reference date of these statistics is July
1, 1976, and the basic source is the Central
Population Register. The population is grouped
both by various geographical criteria, for ex-
ample, parishes, according to the CPR’s address
register, and by urban areas as defined by the
United Nations. Maps from the various munici-
palities combined with population data from the
CPR are used to code urban areas. Through a
complicated linking process involving several tax
registers and a Central Register of Enterprises
and Establishments, information about an indi-
vidual’s occupation and industry can be pro-
vided. Finally, the various items of information
have been combined to produce information for
family nuclei. This combination was accom-
plished through address data and family refer-
ences from the CPR.

Sample surveys.—As indicated, existing regis-
ter data can be used in the production of
population, social, and health statistics.

This method of production is, of course, far
from satisfying all statistical requirements. For
example, satisfactory information about factors
of employment and their effect on the state of
health cannot be compiled solely on the basis of
data already registered. A closer analysis of this
subject would require the collection of special
information from the population, perhaps from
special groups of the population. This collection
may be done by interviews or mailed question-
naires, and it could, moreover, be necessary to
carry out clinical and laboratory tests. The data
collection should be based on some form of
randomly selected representative samples. The
selection of a representative sample would re-
quire a complete list of all persons in the
population under review, and therefore, the CPR

serves as a suitable basis. The population that is
the subject of the study can be printed out from
the CPR provided it can be defined by means of
the data contained in this register.

Danmarks Statistik and Socialforsknings-
instituttet (The National Institute of Social
Research) currently conduct various sample sur-
veys on the basis of the CPR, such as surveys of
Danmarks Statistik’s labor force that are accom-
plished by mailed questionnaires sent to 1%-5
percent of the population. The samples used for
these surveys are persons selected by their dates
of birth, and the selection is made solely on the
basis of the PN’s. The questionnaires never ask
for background data that can be obtained from
the CPR register.

Health statistics.—Linkage techniques using
the PN are employed increasingly in the health
area. The scope for such research is very broad
as evidenced by an analysis of the recurrence of
legal abortions among Danish women in
1975-77, which was based on the Abortion
Register and compiled by the National Health
Service. Here, the PN’s are used for linking to
create a history of each woman. Another survey
that is being prepared concerns the use of
medicine and drugs. Prescriptions for medicines
must always contain the PN of the patient. In
this survey, all prescriptions issued on one
particular day have been registered on magnetic
tape, therefore, the use of medicine can be
distributed according to the social circumstances
of the patients. The prescription tape then will
be linked with personal data on income and
occupation, on the basis of tax registers and the
Central Register of Enterprises and Establish-
ments.

FUTURE PLANS REGARDING THE
UTILIZATION OF REGISTERS

The future development of population and
social statistics will to some degree be deter-
mined by the development of the registration
systems. Moreover, the requirements of statisti-
cal production are also taken into consideration
when reorganizations and extensions of the
registers are being planned.

During the next few years expansions of the
register systems will make it possible to comply

21
with urgent needs of the users of statistics. Two
of these expansions will follow. The chain of
registers important for population statistics has
recently been extended by a new link: a national
basis register of all buildings and dwellings is
about to be established. This register will con-
tain information about, among other things,
location, size, age, and installations of the
dwellings, and the dwellings will be identifiable
by means of a specified address code. The
address code system constitutes a national and
unambiguous system of identification of all
buildings and dwellings, and the codes will, in
the same way as the PN’s, be used in all
pertinent registers listing a residence in the CPR
as well as listing -a business establishment’s
address in the Central Register of Enterprises
and Establishments.

The address code system will make it possi-
ble to combine information about dwellings and
residents. This combination means that the data
about dwellings can be used as background
variables in the various population statistics, for
example, health statistics. Statistics on dwellings
and residents are, moreover, essential to public
and private planning, and traditionally such sta-
tistical data could only be obtained in connec-
tion with the regular population and housing
censuses. In the future, analyses can be made
more frequently and can be more current.

Denmark is obligated to produce both popu-
lation and housing census statistics in 1981, and
it plans to collect the statistical data from the
registers directly without inquiries to the popu-
lation; however, it may be necessary to supple-
ment the census with questionnaire surveys
based on samples regarding data that cannot be
retrieved from the registers.

The other planned expansion of the register
system is registration of establishments, such as
local business units. The existing Registers of

22

Enterprises and Establishments contain informa-
tion only about enterprises that are legal busi-
ness units comprising more than one establish-
ment.

The expansion, which is taking place largely
for statistical purposes, will make it possible to
specify the number of establishments within a
certain geographical area. Moreover, it will be
possible to combine information about the
location of the establishment and the residence
of the employee to determine the extent of
commuting. Also, the location of employment
can be used as a basis variable in population and
social statistics. Finally, the possibilities of
analyzing the population by industry, will be
improved. These improvements are to be real-
ized by a project based on data from the tax
authorities’ register of employers’ notifications
of employee’s wages, salaries, taxes paid, and
other information.

This register contains references partly to
the employees by their CPR numbers, and partly
to the enterprises by the employer code num-
bers, but the project requires that individual
establishments be identified. At present the
project is still in an initial stage but the first
statistics compiled will be for 1979. The out-
come of this project is essential to fulfilling all
the possibilities of the register-based census of
1981.

In conclusion, with the existing registers
many possibilities for conducting important
statistical surveys have not yet been explored.
Some of the surveys discussed can be expected
to elicit new statistical projects. For instance,
the basic data developed for the study of
occupational mortality (employment and death
statistics) can also be used for other special
studies such as the mortality of selected popula-
tion groups.
CHAPTER V
THE POPULATION REGISTRY AND THE USE OF PERSONAL
IDENTITY NUMBERS IN ISRAEL

INTRODUCTION

The State of Israel was established May 15,
1948. The fledgling Government immediately
took steps to conduct a census that, in addition
to enumerating the population, was to serve as a
voter’s register in the elections of January 1949,
and was also to be the basis for a new permanent
population register. The census was taken on
November 8, 1948, under the direction of the
Central Statistical Bureau in cooperation with
the Ministry of Interior. At the top of each
census form a personal identity number of six
digits was printed; this number became the PN
of the persons registered in the census. Eventu-
ally these forms were transferred to local regis-
tration offices throughout the country, and they
became the “Register of Residents.”

A card index for each person was also main-
tained. Every resident aged 16 and over received
an identity certificate; a child under 16 was en-
tered on the parents’ identity card. Each card
was identified by the same number that appeared
on the census form. Immigrants and persons
born after the census were added to the Register
of Residents and were provided with identity
numbers. If a person emigrated or died, his form
was removed from the subdistrict office files to
a central removals archive. In February 1949 an
Act, the Registration Ordinance of Inhabitants
of 1949, legalized the census registration forms,
and the Population Registry Law of 1965
established the present basic registration proce-
dures.

According to that law, the following items
shall be included in the Population Register:

1. Surname, first name, and previous name

2. Name of parents

. Date and place of birth
Sex

. Ethnic group

. Religion

NOY Ov Ws Oo

. Personal status (single, married, divorced,
or widowed)

8. Name of spouse

9. Names, dates of birth, and sex of chil-
dren

10. Past and present nationality, nationali-
ties

11. Address
12. Date of entry into Israel

13. Date of becoming a resident

In addition to the previous information from
census forms and the Ordinance of 1949, data
was compiled on occupation, additional profes-
sions, employers, language in daily use, and
ability to read and write.

Those responsible for the establishment of
the Central Population Register in 1948 prob-
ably did not completely understand the impli-
cations and potential of the registration system
as an administrative tool. It was initially con-
ceived as a device for the identification of
individuals, for the enumeration of the popula-
tion, for the distribution of food, and for the
avoidance of fraud in elections. Other registers
established by the early Government were origi-
nally completely independent. These included
Registers of Birth and Death (Ministry of
Health); Record of Immigration, Naturalization,
and Passports (Ministry of Immigration); and

23
Records for Food Rationing (Ministry of Com-
merce and Industry). A resident sometimes had
to apply to four or five different agencies to
obtain documentation and assistance.

In 1952 the Ministry of Interior transferred
information from report forms to keypunch
cards. One index used was the PN. In 1966, the
content of the cards was transferred to magnetic
tape. Under both methods, the Population Reg-
ister was used more extensively as a tool for
public information and administration. In 1952,
for example, almost all of the functions of the
Ministry of Immigration were combined with
those of the Population Register. The resident’s
visa form became a card in the population
register system, with the PN as a link. In 1953,
the function of issuing ration books to the
population was transferred from the Ministry of
Supplies and Rationing to the Population Regis-
ter Office, Department of Interior. Again the PN
was the link. The censuses of 1961 and 1972,
which also used the PN, referred to the register
to check on the complete enumeration of the
population. Various Government agencies pres-
ently use the PN as an identifier in their records
systems and have a tie-in with the Population
Register.

A NOTE ON THE ISRAELI USE
OF A NUMBER SYSTEM

Several countries, mainly in Europe, had
reservations about using a personal identity
number that reminded them of wars and concen-
tration camps, and, therefore, thought that they
should not attach a number to a person but
rather use only his name. It is a paradox that the
State of Israel introduced a Person-Number
system first. It has now been in use for 30 years,
and is generally accepted, possibly because the
resident’s name is primarily used with the PN to
assist in identification and integration.

Adopting another number consisting of
meaningful digits, such as the birth date that is
used in the Scandinavian countries, has been
suggested. However, as an immigration State,
Israel would experience many errors in the
registration of divergent peoples, who register in
haste and cannot give their birth dates. Further-

24

more, the Israeli system is too well established
for any fundamental changes to be considered.

DEVELOPMENT OF THE EXISTING PN

As stated earlier, the first PN was printed on
the 1948 census forms, and the first number
issued was 000001. To avoid an excess of high
numbers, a series prefix in the form of a letter
was affixed when the six-digit number would no
longer suffice, and a series prefix was used for
special populations (e.g., an “a” series block was
given to new immigrants who moved directly to
immigrant homes from a ship). The new series
began with “a 000001” and continued through
“g.” When the punch-card system was intro-
duced, the letter prefix was translated to a
number: “a” became ‘1.” Thus the original
six-digit number became a seven-digit PN. Many
errors occurred as a result of the duplication of
numbers. Because some agencies did not rou-
tinely use the prefix, duplication was sometimes
appalling. To minimize error at the time of
input into the computer, the Ministry of Interior
decided in 1977 to add one control digit to the
PN using the Modulus 10 system. At the same
time an additional digit was added to the PN to
enable the numbering system to accommodate
future needs above 10 million. Thus the PN
today is composed of nine digits, including one
check digit.

USE OF THE PN SYSTEM

Israel has been interested in developing a
computer terminal network for the extended
integration of statistical systems, but this devel-
opment has been delayed. Only one terminal
exists at the Department of Interior that links
their mechanization unit (an input-output unit)
with the computer at the Office Mechanization
Center for answering various location requests,
for correcting errors, and for clarifying input.
Current updates from the Central Population
Register are provided to the following agencies:

1. The Ministry of the Interior—from the
main office to the field offices
2. The police

. The Ministry of Defense and all branches
of the Army

0

. The National Insurance Institute
. The Bank of Israel

. The Ministry of Foreign Affairs
. The Ministry of Health

. The Ministry of Education

C0 3 OO Ot Wp»

9. The Ministry of Absorption
10. The Central Statistics Bureau
11. The Broadcasting Authority
12. The Parliament’s Elections Committee

13. All the municipal authorities in the
country

Most of the previously mentioned agencies
also get constant periodic service and also
various ad hoc services.

Registration of Births

An example of the process by which integra-
tion of data between ministries is achieved can
be demonstrated in the efforts to standardize
the registration of births. The problem was one
of timeliness and duplication, and various re-
porting forms were used for this effort. The
Ministry of Health, on the basis of the hospital
patient forms for mothers, punched the data of
the newborn children in the hospitals. It also
prepared immunization cards that were sent to
the newborn child’s parents. The Ministry of
Interior registered newborn children in the
population registry, prepared birth certificates,
and entered the childrens’ names in the parents’
identity certificates. The Central Statistics Bu-
reau handled about 100,000 birth forms and,
coded, punched, and absorbed the information
for its own purposes. The National Insurance
Institute received ongoing reports of births, but
independently gathered personal data about the
newborn children and their parents to pay the
childrens’ allowance to families beginning with
the birth of the first child. The Institute needs

this data within a few days of occurrence to
begin payments.

In October 1975, a joint task force was
created to recommend a common birth form
and a coordinated flow process. The original of
the prepared form goes to the registration center
of the Department of Immigration and Registra-
tion, which is now set up for combined coding
with the Central Statistics Bureau, and for
processing by the Office Mechanization Center.
The first copy goes immediately to the National
Insurance Institute and is designed to provide
the basis for the payment of the children’s
allowance. The second copy is given to the
parents for entry on the identity certificate by
an employee at the immigration and registration
office (in the parents’ presence). The third and
last copy remains at the hospital.

This process is to be experimentally tested,
and after a trial period the Ministry of the
Interior and the National Insurance Institute
may share the coding and punching. The Minis-
try of Health and the Central Statistics Bureau
will get a copy of magnetic tapes with data from
the Ministry of the Interior. Both the Ministry
of Health and the Central Statistics Bureau will
then release the birth notification copies and in
return will receive the required data in a
mechanized form from the Ministry of the
Interior. The preparation at the Ministry of the
Interior that is necessary for issuing a birth
certificate, and the process of entering the
information on the parents’ identity certificates
will remain as they are.

Voters’ Ledger

The voters’ register both for the Parliament
and for local government is based on the PN and
the records of the population registry. If neces-
sary, the Ministry of the Interior can prepare an
updated voters’ register 2 months prior to
Parliamentary elections. According to law, elec-
tions in the State take place every 4 years, but
there are cases when, for various reasons, they
are held earlier.

Border Control Card File (Personal)

The border control card file was under the
sole charge of the Israeli police, but has now

25
been transferred to the Department of Immigra-
tion and Registry. It has two types of handwrit-
ten cards, one concerning the traffic of tourists
and the other the traffic of residents.

A mechanical parallelization between this
card file and other collections, such as the
population registry collection or a collection of
individuals insured in the National Insurance, is
impossible for obvious reasons. The plans for the
mechanization of the border control system
were completed recently and their execution has
just now begun. The integration between the
border control collection and other registers can
be developed with the PN as the linking factor.
Obviously this integration applies only to the
part of the records that concerns residents.

PROBLEMS AND DIFFICULTIES
WITH THE SYSTEM

Israel’s peculiar position as an immigration
State with a heterogeneous population creates a
situation where initial registration can be a
source of error. Registration takes place at three
main centers: border posts, hospitals, and regis-
tration offices. No uniform authority supervises
all registration. Border guards, for example, are
under the jurisdiction of the police as well as the
Department of Immigration and Registration. At
times, the bulk of the border registration must
be carried out at night under duress. New
immigrants are often incapable, because of
language difficulties, of answering questions
properly, and therefore, inaccurate information
can be provided and must be corrected later.
Letting immigrants enter without the PN assign-
ment was considered but was vetoed because
new immigrants require many vital services
immediately and these services are programmed
through the numbering system.

A more confusing problem that still exists is
in the construction of the PN. The series letter
prefix led to the duplication of numbers. Also,
when the system of data handling was changed
from a manual to a mechanical card index in
1952, the series letter was translated into a digit
extending the number from six to seven posi-
tions. The phenomenon of duplicate PN’s for a
time damaged the reputation of the Central

26

Population Register and deterred some agencies
from participating in the system. All the dupli-
cate numbers were removed from the computer
and half of the owners of these numbers were
summoned to the register offices to eliminate
the duplication. Some residents refused to co-
operate, and others protested, so this problem
has not been completely resolved.

The type of control digit used by the
Department of Interior was introduced because
other ministries had already adopted that sys-
tem. A better system than the Modulus 10, with
different weights, could have been introduced
but would have disrupted the existing systems
and would have been costly. However, a reliable
control exists because the digits input program
checks the initials of the person’s name as well
as the PN.

Another problem has emerged with the
addition of the control digit plus an additional
digit to provide a number over 10 million. All
the registrants (about 3,500,000 are in the
Population Register) have to be notified that
they have a control digit. Notifications were
planned for the time of elections. All other
Ministries that have a relation to the system and
have used a seven-digit format also must be
made aware of this change. Extensive informa-
tional activities, arrangement of joint seminars,
and periodic checks are planned to cope with
this problem.

PLANS FOR THE FUTURE

Israel plans to attain the following objectives
within the next 5 years:

1. Expansion of integration, with all the
Government agencies that handle popu-
lation matters as well as with all public
institutions that require data from the
system. This expansion will involve max-
imal development of the data base to
include all information in the personal
file and any additional data.

2. Addition of the control digit to the PN’s
on all pertinent documents and the
elimination of all other types of numbers
in public administration (such as Army
serial numbers).

Development of uniform standards for
all documents, including both format
and content.

Establishment of a network of terminals
in all immigration and registry bureaus

that will be connected to a central
computer.

5. Establishment of a central archive.

Israel also expects to mechanize the handling
of passports, develop a registry of firearms, and
generally improve the quality of data.

27
CHAPTER VI

ADVANTAGES AND LIMITATIONS OF THE
PERSON-NUMBER SYSTEM

Although the advantages of using a single
referent, a PN number, in statistical computer
systems may be readily apparent, and such a
system is in practice in business, banking, and
Government enterprises, the advantages, as well
as the limitations, as they are understood by the
countries in this study, should be examined.

SWEDEN

The PN’s are now being used extensively in
Sweden as a means of identification in the na-
tional registration and taxation systems, in the
rest of the public sector, and in the private
sector. The use of these numbers is widespread
because they have considerable advantages and
few technical disadvantages. Their greatest ad-
vantages are stability, theoretical safety, ease of
use, and ease of memorization. The disadvan-
tages are that they are not entirely self-checking
(reversal is possible), and they are carriers of
information.

The high stability of PN’s over time is
inherent in their constituent parts. An individual
is assigned a number at birth or, in the case of an
immigrant, at his first contact with the popula-
tion registration authorities, and this number
remains unchanged throughout his life. A num-
ber can only be altered if it is discovered that
errors occurred when it was assigned, for in-
stance, if more than one person was given the
same birth number, or if the wrong sex digit was
added. Changes in the year, month, or day of an
individual’s birth are extremely rare except for
immigrants from countries with imperfect popu-

28

lation registration systems. Nor should there be
any change in the three digits of the supple-
mented birth number. The last digit, the check
digit, is only altered if one of the other digits is
altered.

The check digit can be constructed accord-
ing to various systems. A system is chosen by
weighing the pros and cons of various factors,
such as theoretical safety, ease of use, availabil-
ity of technical aids and so forth. The Swedish
system was designed to facilitate the establish-
ment of the register on the ADP. The system,
which was planned in the 1960’s, was based on
the data registration equipment then available
via keypunch cards, when every position was
costly. This fact must be considered in any
assessment of the checking system when one
digit is chosen for the PN.

A PN is easy to remember because people
usually remember their own date of birth, and
they only have to learn four other figures.
Moreover, information is available to aid the
memory.

The disadvantage of Sweden’s PN system is
that it is not entirely self-checking. Because of
the advantages this disadvantage is tolerated.
The check digit does not single out all the
reversals of digits within the number. However,
such errors most likely are only minor in extent.
Probably these errors only play a minor part
when compared with the errors occurring in the
practical use of PN’s in the various registers,
such as manually copying errors. Instead, the
essential question is whether or not it is prefer-
able to have an open information-carrying num-
ber, as Sweden does now with the date of birth,
or a system that does not disclose any informa-
tion.

The PN’s are firmly integrated in various
official registers throughout Sweden; many peo-
ple think that the use of these numbers has
become too prevalent both in private and in
official registers. Many, too, dislike the fact that
these numbers reveal a person’s date of birth.
This fact is one disadvantage of an easily run
system. A system with a number that does not
disclose any information is a much more ardu-
ous one to administer—especially the process of
assigning the numbers. Furthermore, such a
system would not have been possible at the end
of the 1940’s when the present system was
started.

NORWAY

The main reason for the establishment of the
PN system was a need for simplified routines
concerning reports on personal matters that
were sent to the public administration. The
system has solved this problem. The establish-
ment of the PN system created other possibili-
ties. First, it has improved the quality of
personal characteristics in two ways: by increas-
ing the value of every characteristic and by
shortening the time from the occurrence to the
recording of an event. Second, it has provided
the potential for the linkage of different public
registers. Finally, it has increased the data
available for use in statistics, planning, and
research.

The establishment of the PN system is in
part the establishment of a statistical file system.
Such a system is based on the idea that data
from different units can be stored in such a way
that they can be extracted when needed. This
storage is possible with the identification num-
bers. The role of the register in this system is to
assign and maintain the identification numbers.
By these identification numbers, information
from different sources may be linked together.
Furthermore, individuals may be followed over
time.

The Central Bureau of Statistics is building
up a statistical file system where the files
contain all data about persons, establishments,

and enterprises having identification numbers.
These data are collected through surveys, cen-
suses, and administrative sources. The basis of
the system is the CPR and the Central Register
of Enterprises and Establishments.

DENMARK

Experience from the first decade with the
CPR numbers confirms that the use of one
common and accurate PN within public adminis-
tration highly facilitates the administrative
work, especially in connection with EDP. The
administrative registers can communicate by the
CPR numbers, and therefore a given personal
data item, for example an address, only needs to
be recorded at one place in the total register
system. This communication also ensures that
the best and most updated data can be used by
all registers. For the same reason, the distribu-
tion of public registration records from several
authorities is only a matter of practical organiza-
tion because the data about the individual
technically can be combined with great accu-
racy, although they are obtained from different
registers. However, the highly efficient use of
the registers results in a fear of registration;
therefore, the problems concerning accessibility
of data are accentuated.

Because the PN contains information about
the date of birth and sex, it is relatively easy for
the individual to remember it rather than a
randomly assigned reference number that would
have to consist of at least seven digits to include
the entire Danish population. By the input of
data into the various register systems, the check
digit of the CPR number is automatically con-
trolled, and therefore, it is a safeguard against
errors both in the writing and the registration of
the number.

The use of CPR numbers has had a very
great impact on the production of different
forms of population statistics. Coordination of
the various fields of statistics into one consistent
system is now possible. The concept of popula-
tion statistics can be used as a common frame-
work for references of social statistics, income
statistics, health statistics, educational statistics,
and so forth. Four data-processing modules

29
intended for common use in the various popula-
tion statistics are the following:

1. A module for distribution of the popula-
tion by residence. (Any geographical
division may be applied.)

2. A module for distribution of the popula-
tion by industry and occupation within a
specified period (1 year).

3. A module for assessment of the size of
the population at any given time.

4. A module for distribution of the popula-
tion by family nuclei, such as groups of
persons consisting of married couples or
single persons with children, if any,
staying at the same address.

The possibilities of coordinating population
statistics are still far from being fully utilized,
but this utilization clearly will result in a much
better service to users of the statistics.

As previously mentioned, the use of CPR
numbers and administrative registers has meant
an improvement in the rationalization of many
surveys. Consequently, it has been possible to
reduce the processing period and publish statis-
tics that are more current than those previously
published. This ability has resulted in a much
better service to the users—especially to re-
searchers and planners. To conduct surveys
based on questionnaires sent to a representative
sample of a population group is efficient because
the CPR, on which the samples are based, is
frequently updated and contains data of a very
high quality; of special importance are the qual-
ity of address data in the CPR. These factors re-
duce the risk of nonresponse which is a source
of error in all questionnaire surveys.

The CPR number has made possible certain
types of studies that were previously considered
impracticable, such as life-cycle analyses where
the development of a given population is fol-
lowed during a rather long period. This study is
previously described in the correlation between
mortality rates and occupation.

ISRAEL

Every resident in the State has at the office
of immigration and registration in his subdistrict

30

a personal file that contains a registration card
with his personal characteristics: surname, first
name, parents’ names, date and place of birth,
nationality, religion, ethnic group, personal sta-
tus, names of spouse and children, date of entry
to Israel, and PN. The personal files are arranged
by consecutive PN’s. A resident who changes his
address from one subdistrict to another can have
his file transferred. A resident is bound by law
to give notice of any change regarding marriage,
divorce, change of address, and so forth. If a
resident leaves the country for good or has died,
his file is transferred from the living archive to
the removals archive. An obvious advantage is
that comprehensive centralization makes it pos-
sible for a resident to get service at one agency
and theoretically within a single day. He can go
to one of the immigration and registration
offices and accomplish the following: give notice
of a change of address, request a change of
name, give notice of a change in his personal
status, have a newborn child registered, request a
birth certificate, get an extract from the register
of residents, request a document of Israeli
nationality, request and renew a passport
(transit certificate), get information about
another person’s address, check whether he or
his family members appear in the voters’ regis-
ter, and request a death certificate for a person
who died in the country. The system hinges on
an identity certificate with an identity number
whose use has become well rooted in public
administration. In Israel an identity certificate is
a fusion of the simple identification card system
and the family book system, which is best
known in Western European countries. The
identity certificate contains not only informa-
tion relating to the owner of the certificate, but
also contains information concerning the spouse
and any children who are still minors, and this
information is based on the PN with notations.
However, with the advantages of this system cer-
tain disadvantages have surfaced:

1. Generally the resident can get service
only in the subdistrict of residence
where the personal file is located.

2. Considerable migration of files occurs
because of change of address from one
10.

subdistrict to another, with all the un-
desirable implications.

. The dependence on the personal file for

rendering service necessitates maintain-
ing many archives that use a great deal of
expensive room, encumber work, and
require considerable manpower.

. The collection of residents is kept se-

quentially in the computer on magnetic
tapes by consecutive PN’s and, therefore,

it is impossible to carry out updates or

to pull out occasional data, and all the
material in each run must be viewed.

. When the computer is inaccessible, many

alphabetical and identification lists must
be kept for ongoing clarification and
location.

. The inability to trace family relation-

ships is a disadvantage because the exist-
ing registration system is an individual,
rather than a family, system.

. When the content of the card file is

transferred to the computer and not all
of the information kept in the file is
entered on the record in the computer,
then it is impossible to give service
directly from the computer, and this
problem increases the dependence on the
personal file.

. Integration with other systems is diffi-

cult to achieve.

. Documentation is often prepared manu-

ally and is not efficient.

Advanced methods in the area of auto-
matic data processing developed in the
last decade have not been adopted.

In 1970 the Ministry of the Interior initiated

1.

an extensive survey to propose a more advanced
method for maintaining an automatic popula-
tion registry based on a PN, which would
remove the difficulties and adapt the system to
modern methods. The following decisions re-
sulted:

An effort to expand the record in the
computer will be undertaken. This ex-

pansion will be achieved by opening
every personal file with PN’s, comparing
its content with the content of the
computer record, and correcting and
completing the computer record. The
record then will contain all the current
data in the file, and thus the great
dependence on the personal file will
automatically be discontinued.

At the end of this operation all personal
files will be transferred to a central
archive, because use of the files will be
reduced to a minimum.

A magnetic data pool (central data base)
will be established to replace magnetic
tapes with sequential records. The main
capacities of the base will be

a. Constant updating, even relatively
small numbers of records in the
collection, efficiently and inexpen-
sively

b. Pullout by various keys through
standard service programs and pull-
out programs that will be specially
prepared for different purposes

c. Change and expansion of the content
of the registers without the need for
large-scale planning and program-
ming

d. Connecting with a communication
network without the need for large-
scale replanning (preparation for an
on-line)

The establishment of a data base has the
following advantages:

1.

The ability to furnish institutions and
individuals with one-time or periodic
information concerning various groups
and sections, in accord with their re-
quests.

The ability to issue documentation that
presently is given manually and is based
on the personal file directly from the
computer. Such documents include iden-
tity certificates, birth certificates, and
passports.

31
32

3.

4,

The capacity for integration with exter-
nal systems by the use of terminals.

The establishment of a network of termi-
nals from the subdistrict bureaus of the
Ministry of the Interior to the Office

Mechanization Center or to another cen-
ter, and the rendering of services directly
to the public through these terminals at
any office regardless of the permanent
place of residence of the applicant.
CHAPTER VII
CONFIDENTIALITY AND PRIVACY

The confidentiality of personal information
contained in public records systems and the
protection of individual privacy are matters of
concern in every country. These subjects are
discussed at all international statistical and
population conferences. The introduction of the
computer and the use of a PN as a key to related
files created some public concern that these
mechanisms may make invasion of privacy easy.
Most governments have taken steps to insure
confidentiality and privacy in the handling of
public records and tapes by establishing legal
limitations and procedures. Access to and con-
trol over records are strictly limited in most
countries. Some western countries that would
find it relatively easy to develop PN systems
have refused to do so. In one country, the
recollection of invasion during war time is a
factor; in another, an extensively integrated
numbering system was debated in Parliament for
more than 5 years because legislators were not
satisfied that privacy could be maintained. The
idea that stringent control over the population is
an aim of such a system is refuted by the fact
that the numbering systems are most advanced
in the Western democracies. This report provides
a description of the situation in these countries.

SWEDEN

From 1947 to the end of the 1960’s the
general public in Sweden accepted the use of
PN’s, deeming them to be a useful and easily
manageable instrument by which individuals
could be safely identified. However, during the
taking of the general Census of Population and
Housing in 1970, the question of privacy be-

came eminent. Fears were voiced that the new
system, with the rapid spread of ADP records,
would lead to an undue infringement of privacy.
As an outcome of these discussions, a committee
was appointed to study the problem; its report
led to the passing of the Data Act and to the
establishment of a supervisory authority, .the
Data Inspection Board, which came fully into
operation on July 1, 1974.

The Data Act protects the individual from
any “undue infringement of personal integrity”
that may ensue from the increasing use of ADP
in personal registration. According to the Data
Act, no ADP register, file, or other record
containing information about an individual may
be established or kept without a special permit
from the Data Inspection Board. The latter must
examine each register from the viewpoint of
personal integrity before deciding whether to
grant a permit. When doing so, the board may
also impose rather far-reaching directives con-
cerning how the register should be kept and
managed to obviate any undue invasion of the
privacy. If those responsible for a register
infringe these directives, then their permit may
be withdrawn by the board.

Special rules apply to registers established by
governmental or parliamentary decision. The
Data Inspection Board has no jurisdiction over
such registers; however, it must always monitor
the establishment of such registers and, as stated
in the preamble of the Data Act, its views must
be given serious consideration. The board is also
to insure that the directives issued by the
Government or by Parliament regarding the
management of registers are complied with.

For example, the Cancer-Environment Regis-
ter is, in principle, available on application to all

33
scientists and investigators interested in cancer
research. The use of the register is, however,
regulated by the Data Act with the directives
issued by the Data Inspection Board, by the
Secrecy Law pertaining to the right of access to
public records, and by an agreement between
the National Board of Health and Welfare and
the National Central Bureau of Statistics. Direct
responsibility lies with a special committee for
the Cancer-Environment Registry, appointed by
the Director General of the National Board of
Health and Welfare.

The Data Act is divided into five sections.
The first section provides definitions of terms
that occur in the text. No mention is made of
personal registers as such except for an example
of an ADP register containing personal data that
can be linked to individuals identifiable either
by their name, by a PN, or by some other
means. The second section contains provisions
concerning permits, directives relating to the
purpose and content of the registers, and data
that may only be registered in exceptional cases.
The third section contains provisions relating to
the obligations of those responsible for the
registers of individuals. Specifications are stated
regarding how and when corrections are to be
made in the registers, how a registered person is
to be informed, if he requests it, of the facts
registered about him, and when data are not to
be made available. Another provision states the
obligation of those responsible not to disclose
information in the registers. The fourth section
is devoted to the Data Inspection Board’s
controlling and supervisory powers. Finally, the
fifth section contains rules about any violations
of the directives issued by the Data Inspection
Board and the consequences of any violation
such as penalties, damages, or fines.

A bill proposing the establishment of a CPR
in Sweden was submitted to Parliament in 1972.
The main purposes of a CPR were to facilitate
information storage in records by various sectors
of public administration, and to reduce the
storage needs to a minimum. Furthermore, the
use of PN’s would make it easy to retrieve
information from the CPR. Consequently, the
number of particulars such as name and address
could be reduced in other registers. In fact, this

34

procedure has to some extent become a reality
because the county administration now supplies
data to other computerized registers. However,
the bill was not passed; Parliament, with the
problem of privacy in mind, deemed it advisable
to wait for the passing of a Data Act.

In 1976, Parliament passed a bill setting up a
reduced central register of individuals, called the
Coordinated Register of Individuals and Ad-
dresses (SPAR). Parliament also established what
the register was to contain and that its contents
were not to be disclosed without the permission
of the Data Inspection Board. The SPAR register
is updated from information held in the county
administrations’ notification tape of population
registration data. The register must be updated
with these particulars once a week. Further-
more, the register is to contain particulars
concerning assessed income and the ownership
of property. These particulars should be updated
once a year. If a person does not want to receive
advertising matter directly, the register is to be
marked to this effect. The SPAR register is now
being constructed and is expected to be in
operation in the spring of 1978; it will be devel-
oped and operated by the National Data Centre
for Administration Data Processing (DAFA).

In the discussions of privacy during the past
few years, the use of PN’s has been a focal
question. Although these numbers facilitate in-
tegration between different registers, the exist-
ence of such an efficient key to all kinds of
information involves great hazards. It would be
illogical to eliminate an efficient tool in public
administration just because this tool is open to
the danger of misuse. The question therefore
remains: What precautions must be taken to
prevent misuse? Privacy problems must obvi-
ously be handled by including adequate safe-
guards in ADP systems, as established in the
Data Act and in the regulations issued by the
Data Inspection Board. Technically, PN’s can be
prevented from acting as keys to registers when
special signs are given to prove the legitimacy of
an inquiry. Another possibility is releasing only
the data required specifically in an inquiry. Con-
sequently, only selected categories of inquiries
will be able to obtain information from the
registers, and to some of them only limited data
will be made available. Certainly, the use of PN’s
in Sweden will continue, and the increasingly
widespread problems concerning privacy will not
prove insurmountable.

NORWAY

For many years, the Central Bureau of
Statistics has been collecting information about
persons, and is now establishing a statistical file
system that contains several registers. The
amount of collected data about persons is
steadily increasing. In addition, the local regis-
tration offices keep much information about
persons stored on magnetic tape in regional
computer centers. The increasing amount of
data about persons has brought the problems of
confidentiality into focus. These problems are
no longer only technical problems, but are also
political problems.

In 1975, a report was presented by a
committee formed to study problems of “pri-
vacy” in connection with the establishment
and use of public administrative data banks and
statistical data archives. Another report con-
cerning personal data in private enterprises was
presented in 1974.5 In June 1978, a Data
Act was passed, but has not yet been enforced.
Chapter 1 of the law states that the regulations
refer to personal registers and use of personal
data in central and local governmental services,
in private enterprises, in associations, and in
other institutions.

According to chapter 2, a Data Inspection
Board shall be established as a supervisory au-
thority. The Data Inspection Board shall, among
other things, establish and keep a list of all per-
sonal registers for which a permit must be ob-
tained. As stated in chapter 4, personal registers
using ADP records cannot be established with-
out special permit from the King. In addition, a
permit is necessary for other personal registers,
when they include data such as race, political or
religious views, health conditions, and sexual
attitudes.

In chapter 3 general rules are given regarding
the rights that persons have concerning informa-
tion stored about them. Generally, all data shall

be available, but there are some exceptions
regarding information that can do harm to the
person or his family. Exceptions are also stated
for registers used only for statistics, research,
and general planning.

The Central Bureau of Statistics is concerned
about data protection. This question not only
refers to the CPR, but also to the entire activity
of the Central Bureau of Statistics. Therefore,
strict rules for data processing, data storage, and
data use have been introduced. Only selected
groups of persons can require physical data, and
no single person has access to the information.
During the year, the Central Bureau of Statistics
receives many inquiries for use of personal data.
These inquiries are forwarded from private and
public institutions, individuals, research workers,
and many other sources. The Central Bureau of
Statistics collects data pursuant to a special sta-
tistical act and an act of population registration.
In general, the Central Bureau of Statistics is
bound to secrecy about the data. However, per-
sonal data given to the CPR according to the
latter act, can be delivered to other data users.
Pursuant to the act of personal registration, the
public authorities (but not all public institu-
tions) have the right to obtain personal data
needed in their work. Also, many research insti-
tutions, private banks, insurance companies, and
so forth, under certain conditions, get birth
numbers, names, and addresses from the CPR
for their employees, policy holders, and
customers.

The local registration offices keep informa-
tion about persons on written cards. However,
the rules for handling the data are strict, and the
staff must sign a declaration of secrecy. The reg-
istration offices have the authority to approve
some standardized use of personal data by the
public authorities. The Central Bureau of Statis-
tics has given detailed specifications for imple-
mentation, and has established a list of local,
regional and central authorities that are allowed
to receive such data. The Central Bureau of Sta-
tistics has the responsibility to inspect the regis-
tration offices, and routinely does so to make
sure that all comply with the regulations.

The Central Bureau of Statistics has made an
agreement with the regional computer centers

35
about storing, use, and destruction of personal
data. This agreement includes the conditions
that were established to allow the regional com-
puter centers to handle the data. These condi-
tions comprise the rules that the computer cen-
ters must follow to protect data and prevent
abuse. The computer centers register all use of
personal data. The agreement makes it possible
to inspect the centers and all their personal data
routines. The staff at the regional computer cen-
ters also must sign a declaration of secrecy.

DENMARK

During the period since the introduction in

1968 of the PN’s, the public debate concerning

registration of personal information and safe-
guard of privacy has gradually become intensi-
fied. Occasionally, the problem is presented as if
it were a matter of being for or against PN’s.
This attitude, of course, is misleading, because
the problems concerning the protection of per-
sonal data predate establishment of the CRP.
The introduction of the PN, however, has made
the operation of registers more efficient, and the
accuracy of personal identification has made it
easier to combine information so that many
items of information about an individual can be
integrated.

As indicated, only public authorities are en-
titled to demand knowledge of the PN. Accord-
ing to Danish Law the public authorities must
not give any personal information to individuals
or business establishments. However, a person
can, within certain limits, insist on knowing
what information is registered about him and
has free access to the local population registers
for information about the addresses of individ-
uals that the inquirer can identify. The registers
are not permitted to pass on individualized in-
formation on a large scale to private persons.
The staff within the public administration and in
the EDP centers, dealing with information for
the public, are subject to the general rules of
nondisclosure.

If the public authorities use information on
the register in a way that is considered by the
population as a violation of the right of integ-
rity, this use may result in an opposition against

36

supplying necessary information to them,
including the PN. In the case of widespread
opposition, the existence of the population
registers could be threatened. However, only a
few, if any, examples of abuse in connection
with the use of registers are known.

In the last few years a certain reluctance
against registration has been observed, although
the manifestation has perhaps not been so pro-
nounced as in other countries.

In 1978, the Danish Parliament enacted two
bills on registers—one concerns Government and
municipal registers. The Public Registers Act,
which became effective January 1, 1979, clearly
stresses the rules of nondisclosure that were
already in effect as well as the rules of accessi-
bility of register data. Moreover, it provides the
establishment of a Register Board, that shall
supervise the registers and approve regulations
concerning their operation with a special empha-
sis on safety measures for the protection of in-
formation. The board has the power to exercise
control by direct inspection of the installations
where the registers are kept. The board must
be notified in each case of matching of registers
and the board may establish the conditions of
how such matchings shall be executed.

Although the primary aim of the Act is to
regulate the use of registers in public administra-
tion, the regulation also applies to registers that
are used only for statistical and scientific pur-
poses. However, the rules applying to such
registers are different. Thus the linking of
information from different registers can be
administered without notification to the board
when the purpose is strictly statistical, and the
right of a person to know what the registers hold
on him is not extended to statistical registers.
The reason is that statistics cannot threaten
privacy, because information about individuals
cannot be identified from the statistical tables,
and because individual information on statistical
registers must not be used for administrative
purposes.

The problem of protection of personal infor-
mation has long been recognized as important to
the production of statistics. The population’s
confidence that information given for statistical
purposes is not disclosed to public authorities or
to individuals is vital to the activities of Dan-
marks Statistik. Consequently, information must
be treated as confidential and must be safe-
guarded so that no information that may be
related to individuals is published.

The practice of Danmarks Statistik has been
most restrictive, implying that no identifiable
information may be transmitted—even to institu-
tions solely concerned with research work. With
the new Act, this practice has become part of
Danish Law. The Act makes a concession for
research and statistical productions that is not
previously made by Danmarks Statistik. With
the board’s consent in each case, data from
statistical registers may be transferred to another
public agency that will use the data for research
purposes only.

The debate on the question of integrity in
connection with the PN has only in one case had
consequences that were detrimental to the pro-
duction of statistics on social benefits. Under
The Social Assistance Act (Bistandsloven),
Danmarks Statistik must collect data. identified
by the social clients’ PN’s from the municipali-
ties. A few local governments have refused to
give the PN’s because of their clients’ right to
anonymity. This refusal is not acceptable to the
government as it conflicts with the Act of
Danmarks Statistik. The dispute has not yet
been concluded. The problem may be solved by
the previously mentioned Register Act.

ISRAEL

The Ministry of Interior is acutely aware of
the current opinions as well as fears concerning
the danger of violating the individual’s privacy
by ‘“‘overintegration” when important concen-
trated information is located at one center and
might be available to unauthorized persons who
will use it in a harmful way. Therefore the
Ministry established these conceptual and practi-
cal limits:

1. The data base must not contain informa-
tion beyond personal logistical data spec-
ified in the Population Registry Law.

2. An employee of the Ministry is forbid-
den to ask questions beyond the 14
questions in the law and is absolutely
forbidden to record anything in the file

beyond a person’s answers to the legiti-
mate questions.

3. No particulars in addition to those men-
tioned will be noted in the data base,
consequently no possibility exists for
passing it to any external element. The
data base must therefore not contain
information such as hospitalization,
mention of a criminal past, or data about
the resident’s income. This information
is essential to other agencies and should
only be filed in those agencies. The data
base will update only the entries in the
other agencies.

What is the common practice for giving
information from the resident’s collection? Sec-
tion 29(b) of the Population Registry Law
states: “Any person may receive information
concerning the name and address of any other
person registered in the registry.” This Section
then states that, without examining the purpose
and without permission from the person regis-
tered, another party may be given information
that appears either in the voters’ register, which
is open to all, or in any telephone directory.
However, section 29(c) of the law also states:
“A person who is prima facie interested may
also receive information concerning the date of
birth and particulars of other registrations deter-
mined by the regulations of a person registered
in the registry.” Recently additional particulars
that may be disclosed were determined by
regulations that were approved by the parlia-
mentary committee of the Constitution, Legisla-
tion, and the Judicial Committee, and they are:
parents’ names, place of birth, personal status,
sex, ethnic group, date of entry to Israel, and
PN. Information about religion and nationality
was not included and must therefore not be
disclosed.

A serious restriction in this section states
that information can be given only ‘“‘to one who
is prima facie interested.” The Attorney General
has clarified these words indirectly by deciding
that the information should not be given if it is
needed “for ideological, political, or commercial
purposes.” This statement implies that every
legitimate purpose is permitted. Therefore, in-
formation can be given to Government agencies
and public institutions such as local Government

37
institutions, the National Insurance Institute,
research institutions, and universities to aid
them in the execution of their duties provided
that they need the information, state why it is
needed, and sign a declaration affirming that
they will not pass the information on to others.
In summary, information is not given if the
information is requested for a political purpose,
for a business purpose, or for a purpose that
opposes the public interest (e.g., solicitation not
to enlist in the Army).

In addition to the preceding instructions, a
very strong mechanical security system at the
Office Mechanization Center prevents any at-
tempt to obtain information from the data base.
Pullout and access are permitted only by a code
(key) that is constantly replaced and is only
known to a restricted number of employees.
Even the Minister of the Interior does not know
the code.

In brief, the rules are:

1. The population register is “sterile” and it
may contain only those particulars of
registration that are specified in the
Population Registry Law.

2. Information may be channeled to other
systems, either governmental or public,
only after those who request it have
demonstrated that it is required for the

38

proper execution of their functions and
for a legitimate purpose. This approach
is very useful because it releases the
public agencies from the necessity of
updating the logistic registration particu-
lars, therefore, they can devote them-
selves more efficiently and economically
to their assigned tasks. For example, the
Ministry of Health, which is interested in
a followup of people who contracted
cancer, will get reports about the facts of
their death without the population regis-
try employees knowing why the infor-
mation is needed and what use is made
of it.

The authorization for giving information
can be given by the Minister of Interior
or the Director of the Mechanization
Department only. Information cannot be
obtained without an explicit authoriza-
tion by one of these two.

The extent of information concentrated,
the process of channeling information,
and the particulars that may be disclosed
are determined in the law and its regu-
lations.

Technically, an almost hermetic security
system guards against extracting infor-
mation from the computer.
CHAPTER VIII

OTHER PERSON-NUMBER SYSTEMS
IN THE WORLD TODAY

Several types of national data systems use
PN’s as identifiers. Some countries have systems
similar to those in the Scandinavian countries
although they may not have the same extensive
applications. Developing countries, despite the
lack of advanced subsystems (e.g., vital statistics
systems), skilled manpower, adequate financing,
and other resources have adapted the PN system
to meet their needs. Some PN systems apply
only to a particular agency or operation (e.g.,
vital and health statistics, national insurance, so-
cial security, and pension fund), and others are
still in the planning stage. Examples of some of
these systems follow. See table C for items in
Person Numbers in national data systems.

ADVANCED SYSTEMS

Finland

PN’s have been used in Finland since 1964;
in 1970 the structure was confirmed by law. The
program, the Personal Identification Code (PIC),
is administered by the Population Register Cen-
ter. This system is used primarily by the
administrative and statistical agencies of the
Government and in a limited way in the private
sector. The PIC is particularly useful for the
construction of population registers and for
record linkage of various types.

The PN is a 10-digit figure: 6 birth digits, 3
serial digits, and 1 check digit. The PN has been
the identifier in the population census, house-
hold surveys, examination registers, university
student registers, manpower surveys, as well as
in judicial, criminal, election, and vital statistics.

The PN is the identifier in the Central Popula-
tion Register, registers of the tax authorities,
and registers of the National Pension Institute.
The data collected by the Central Statistical
Office are used only for statistical purposes;
strict rules of confidentiality are guaranteed by
the law.

France

Numbering systems for individuals as well as
for industrial and commercial organizations have
been in existence in the French Republic for
many years. In 1941 the Demographic Service,
the forerunner of the Institut Nacional de la
Statistique et des Etudes Economique (INSEE),
introduced a 13-digit number that included
codes for sex, month and year of birth, and
geographic codes for place of birth. This system
has formed the basis for an individual identifica-
tion index that is kept current by reports from
civil registration offices. In 1970, the records
were placed on computer tape; two check digits
were added for computer adaptation. Local
administrators send notifications of birth and
death as well as changes in civil status to INSEE,
and consulates advise INSEE of the deaths of
French citizens abroad. The new system became
operational January 1, 1973, and is extensively
used by the Social Security Administration, the
Election Registration, the Ministry of National
Education, and the Ministry of Economy and
Finance.

The computerized system has permitted the
development of new research studies. One study
on occupational mortality is based on popula-
tion samples from the censuses of 1954 and

39
Table C. Items in Person Numbers in planned or operational national data systems

 

 

 

 

 

 

 

 

 

 

 

 

 

  

 

  

 

 

Date of birth ; Place
Serial Check Random
Country Total Pg Sex of
Day | Month | Year number digit birth pumber
General population systems Number of digits
Europe:
DD RIINBI RK nuscrcsiosiverrmiinssptierssrreseeammirinkeirinekeisess 10 2 2 2 3 1 (1) -
Finland......... 10 2 2 2 3 1 - -
Franta ec cmnmsservisrsisersrsvin 15 - 2 2 3 2 1 5
Federal Republic of Germany .. “| 212 2 2 2 2 1] 3 -
1CRIANG sovriinsrvinismsssssrvenissiirniompsssnninmssisssvesinsonsns 9 2 2 2 2 1 - - i
NEthEFIANDS ....eurveirsiveirirssessnssssiss esses ens sessassans 310 . . . - 2 . . 8
Norway 1 2 2 2 3 2 | 4 -
Portugal a 1 2 2 2 54 1 Ek -
SWEAEN.....vorvvversiessessssnssessssensssss sissies sens snsises 10 2 2 2 3 1] 4 -
South America:
AFGENTING 1.vovvererersseseeerssssessasssessessssssses seins = 8 . - | = 8 + | em) -
Chile.....ouven. 13 . . 2 5 2 - | 722
Colombia... 2 11 2 2 2 3 1 1 - '
PEF Usuruniisersesisessanssansssnssnses cerns crs 813 - - 2 5 1{ 4 hl
Uruguay .uveseeniss ATARI TREE A ares 3g - - - - 1 - - 7
Asia:
39 - - - - 1 - - 8
13 - - 3 5 ; 1 | 22(2)
National health systems
United KiNGAOM.....u.cevveivevecinssnsssensessessssessessesissans 10g . . 2 3 . . 3
AUSralia...c ns 10 2 2 2 3 1 - -
Other
UNHET Stata. ccimmrmsmsinmnissscssinsiisitisssmnsninssns 1114 - - 2 6 - - 3

 

 

 

 

 

 

 

 

 

 

 

1Sex indicated in control or check digit.
2This special Person Number, for insurance purposes, contains 2 digits for a regional number, and 1 digit for the initial letter of the
given name.
Random number used.
4Sex indicated in serial number.
SSerial number includes code for nationals and foreigners.
6For check purposes, 2 digits may be added for year of birth and 1 digit for sex (1—male, 2—female).
Contains codes for province of birth and legal office.
80ne additional digit (3) indicates Peru; 4 digits represent location of registration (department, province, or registration office).
9Two digits represent parents’ place of birth.
10Five letters and 3 numbers (LMNOP 123) include codes for subdistrict of birth, year of birth registration, and quarter of
registration.
11The U.S. birth number, used by State vital statistics registration offices, has 3 digits for country and State of birth, 2 digits for year
of birth, and 6 digits for State file number. A birth in North Carolina in 1980 could have the following birth number:

132-80-000428

The first digit represents the United States (1). The next two digits represent the State of North Carolina (32), in a system that begins
with Alabama (01) and continues through Alaska (50) and Hawaii (51). The District of Columbia is 08. There are additional codes for
Puerto Rico and the Virgin Islands.

1975. Another study directed by the National the population drawn from the census of 1968
Institute for Health and Medical Research, the and followed up with the census of 1975) forms

Institut National de Santé et Recherché Médical, a special sampling frame for socioeconomic
follows tuberculosis and cancer patients’ records studies. Changes in occupation, marriages, births
until death. A special file of persons born on of children, deaths, and other events are inte-

October 1, 2, and 4 (approximately 1 percent of grated through linkage by means of the PN.

40
The PN is also used for compiling statistics
of annual wages by wage records sent from
employers to the Social Security Administration
and linked through the PN. This compilation is
done for a sample of wage earners that includes
all persons born in October in the even years.

The advantages of the system, according to
the reports, lie in the easy access to files, the
availability for statistical or accounting tabula-
tions, the certainty of identification of docu-
ments, and the cost-effectiveness.

Iceland

A population register, The National Regis-
try, is directed by the Statistical Bureau of the
Republic of Iceland. This register, which became
operative in 1953, is based on a PN provided at
birth. This number originally consisted of eight
digits, but in 1964 it became a nine-digit number
when a Modulus 11 check digit was added, with
six digits for the date of birth, two serial digits,
and one check digit.

The PN has been principally used as the
identification code number in the internal opera-
tion of the register. It has not been used
extensively for administrative or other purposes.
In 1966, a registry of all school children was
established (a general continuous pupils regis-
ter).

Another numbering system is also in use,
based on the Nafnnummer or Name Number,
related to the name of a person. Introduced in
1959, it is given to every individual who reaches
age 12. The Nafnnummer, seven digits plus a
check digit, indicates the person’s place in the
alphabetical sequence of the population by
name. A seven-digit Nafnnummer is sufficient
because the population is approximately
220,000. Internally, the Nafnnummer is used for
linking persons in the same family.

The Netherlands

In the Netherlands a system of continuous
accounting for population statistics was intro-
duced in 1850 and used the census of 1849 as a
starting point. The system is decentralized: each
municipality keeps its own population register
under the Netherlands Inspectorate of Popula-
tion Registers. Before 1940, the family was the

registration unit but that year, the individual be-
came the registration unit and the population
register in all municipalities contains a person’s
register consisting of standard Personal Cards
(PC). The PC’s are issued at birth and immigra-
tion, and follow the person while in the country,
throughout his lifetime. All changes in civil sta-
tus are recorded on the PC and the card of a
head of a family is used for a survey of the
nuclear family. After death, the PC is removed
from the register and placed in a central file kept
by the Central Bureau of Geneology. Because
the municipal files refer only to the resident
population, special procedures exist for the re-
moval of emigrants and the addition of immi-
grants. The efficient operation of the system has
provided the Netherlands with timely annual
data compiled by the Central Bureau of Statis-
tics, such as total population; population by age
and sex; and population by sex, year of birth,
and marital status.

Because the municipalities vary in data-
handling methods (some have computers, and
others use punchcards or metal plates in addi-
tion to hand-administered files), efforts have
been made to standardize and improve tech-
nical processing. In 1979, a standard municipal
computerized system was operating in more

- than 200 municipalities. A Central Population

Register is planned. A bill will be presented
to Parliament at the end of 1979 and will pro-
vide for, among other things, a PN. Various
types of personal numbers had been discussed,
but for explicit technical and political reasons,
including the matter of confidentiality, random
numbers have been recommended.

Actually, such numbers have been in use
since 1968 by those municipalities that com-
puterized their population registers. They were
furnished by the central Government for com-
puter use only, and consist of eight digits and
two check digits. In 1979, the number appeared
on the PC’s of more than 8 million persons, but
it was used only within the municipal admini-
stration, and was not known by any individual.
The new legislation will enable the Netherlands
to produce population and social statistics more
efficiently.

At present, statistics on national social insur-
ance, sickness benefit funds, and hospitals do

41
not use the municipal numbers in their records.
If a central system is introduced, the use of the
number may be extended to these records.

DEVELOPING COUNTRIES

Developing countries, in coping with the in-
creasing problems of data management, have
considered establishing national data systems for
program planning and Government administra-
tion. Many countries have shown a special inter-
est in the PN systems of the Scandinavian
countries, and have sent administrators and
statisticians to study them.

In Latin America, legislation establishing
national PN systems has been passed in some
countries, but the concept has not had broad
application, nor has it led to any extensive inte-
gration of data from subsystems. The PN is fre-
quently used as a means of identification in
addition to an identity card, photograph, and
fingerprint. It is also used as a file number on
various administrative records, such as social
security and civil registration records.

Argentina

A population numbering system has been in
~ operation in Argentina for a number of years.
Under this system each person on reaching 18
years of age was registered and issued a seven- or
eight-digit number. Males were given a Libreta
de Enrolamiento and females were given a
Libreta Civica that contained their registration
number. In 1968, the registration age was
lowered to 16 years of age and the documents
were unified in a Certificado Nacional de Identi-
dad. In 1970, the name of the identification
document was changed to Documento Nacional
de Identidad (DNI).

Although this activity focused on the con-
tinued registration of adult persons, a corre-
sponding program in 1968 began the registration
of all persons at birth. Every newborn child is
issued a PN of eight digits. At 8 years of age each
person’s DNI is updated with a photograph and
fingerprints. At 16 years of age, the individual’s
signature is recorded and information such as
marital status, occupation, education, languages,
involvement in the practical arts and sports is

42

added. This information is updated periodically
throughout the lifetime of the individual. Also,
beginning in 1968, all persons born before that
year, on reaching 16 years of age, are issued a
PN and DNI. By 1984 this procedure will be
completed and in that year the entire population
will be included in the unified PN system.

Foreigners may also be given PN’s when they
become residents; a minimum of 1 year of
permanent residence is required. If they become
Argentinian citizens, then the standard PN and
DNI are issued.

Under the law, the purpose of the system is
to provide information on the country’s human
resources for defense and development: also
anticipated is the eventual development of a
continuous population register.

Chile

Although recognizing the necessity that
reliable information must flow to Government
agencies for development and planning, Chile
was also concerned about the duplication of
administrative information systems. A national
integrated population information system invol-
ving two factors was planned: (a) the assignment
of a unique number under the Rol Unico Na-
cional (RUN) to each person and (b) the crea-
tion of computer files or population data banks
known as the Régistro Nacional de Pobldcion
(RNP). The information system would stand-
ardize identifiers in all institutions such as social
security and pension institutions, the tax author-
ity, and the electoral register. Records from the
Civil Registration and the Identification Service
would be incorporated, and the files would be
continually updated by information on birth
and death as well as reports of changes in civil
status.

Most of the adult population had already
been issued an identity number in the tax sys-
tem under the Rol Unico Tributario (RUT),
which consisted of seven digits and one check
digit. The RUT number is still used for identifi-
cation in tax offices and in Government agen-
cies. Since 1975, a new identity number, a PN of
11 digits and 2 control digits, is issued at birth
and follows a person throughout his life. Those
born prior to 1975 will continue to use the old
number. The new number consists of two digits
for province of registration, legal office, and
year of registration; and a five-digit serial num-
ber, plus two check digits. The basic file will
contain the PN (RUN), name, date of birth, sex,
identity number, and possibly nationality and
residence. The new number is to be used as the
standard identifier for all records related to
health, education, and welfare, and as a research
tool in socioeconomic studies.

Between 1976 and 1978, the massive project
of assigning numbers to the unnumbered popula-
tion was undertaken. In 1979, an administrative
reorganization of the Civil Register and Identifi-
cation Service was initiated. The possible revi-
sion of the numbering system is one of the
topics for consideration. Furthermore, a plan to
create a new central file using the RUN number,
which will integrate data from all separate files
related to the Social Security System, was ready
for implemention in 1979. It will go into ef-
fect with the reorganization of the civil registry
service.

Colombia

In the Republic of Colombia, the National
Registration Service (SNI) of the National Ad-
ministrative Department of Statistics (DANE)
was established in 1968, The SNI instituted a
new Civil Registry System in 1971, and the key
to this system was a unique number issued at
birth. The SNI considered using a unique iden-
tifier for an individual and his socioeconomic
characteristics. This identifier could help solve
some registration problems and also could elim-
inate duplication among various civil administra-
tions. The number consists of 11 digits: 6 digits
to indicate year, month, and day of birth; 1 digit
to indicate sex (1l-female, 2-male); 3 digits to
indicate a serial number, plus 1 check digit.

The number was first assigned only to babies
born in Bogota; then it was assigned to births in
other large cities as well as other areas. The
present file contains 5 million registrations of
the 25 million citizens of the country. The
registration number covers documents related to
birth, death, marriage, identification, and driver’s
licenses. Various other procedures are presently
under study. Officials from other Latin Ameri-
can countries, notably Ecuador and Costa Rica,
have recently visited Colombia to study the sys-

tem for possible adoption in their countries.

No plans to link records exist. The number-
ing system is used to check files and correct
them, to prevent duplication of the records, and
to calculate vital statistics estimates. Hopefully,
in time, all civil records will bear the PN.

Peru

To obtain demographic data to plan proj-
ects, allocate resources, and evaluate programs,
the Government of Peru in 1975 established the
National Statistical System (INE) designed to
gather, tabulate, and analyze statistical data.
Furthermore, INE was given the responsibility
for developing and improving the civil registra-
tion/vital statistics system. For technical support
INE requested the collaboration of the Office of
International Statistics at the U.S. National Cen-
ter for Health Statistics through its Vital Statis-
tics Improvement Project (VISTIM). A Model
Vital Statistics System was implemented in
demonstration areas throughout the country.
Among the new techniques used was the event
identification number (EIN) assigned to all birth
certificates. With the development of a PN, INE
plans to add a check digit.

The EIN is uniquely constructed. The first
digit (3) represents Peru as defined by an inter-
national convention on vital statistics numbering
cosigned by Canada, Peru, and the United
States. The next four digits represent a location
code identifying a registration office (including a
department, province, and official or registration
institution). The next two digits represent year
of registration. The serial number of five digits
that represents the birth number follows. With
the check digit the PN will consist of 13 digits.
Presently, the total number of digits appears
flexible, and depends on the actual number of
digits used to indicate the place of registration
and serial number of births. Because the system
is still evolving a final appraisal cannot be made.
The major objective is to introduce successful
aspects of the model system into the National
Vital Statistics System.

Uruguay

In 1974, Uruguay established a PN system
based on a complicated coded combination of

43
seven digits and four letters. The first six digits
constituted the birth date, and the seventh digit
was a code indicating sex and citizenship. The
four letters were the first initials of the first and
family names. Two check digits make the total
13. Obvious problems with this complex identi-
fier existed, and in 1978 Uruguay simplified the
system by developing a random PN number of
seven digits plus one check digit. This number is
issued to all persons over age 12. Provisions are
made for the registration of aliens and emigrants
and for all persons entering the legal registration
age. Each registered person is issued an identity
card containing the PN, name, residence, civil
status, date and place of birth, a photograph,
and fingerprint. The cards are subject to renewal
at ages 20 and 60.

The system is administered by the National .

Directorate of Civil Identification under consul-
tation with the Honorary Technical Advisory
Commission. The number is now largely used for
personal identification and recordkeeping in
public and private institutions such as hospitals
and schools, and will be extended to national
administration. Registration at death is recorded
by the PN, and registration at birth is recorded
by the parents’ PN’s. When the system is fully
operational, the record linkage potential will be
explored to develop an alternative to the census
of population. While developing the PN system,
Uruguay also established a Register of Enter-
prises and Entrepreneurs identified by unique
and permanently assigned numbers.

Jordan

Perhaps the most recent application of a PN
in a national system is in the Hashemite King-
dom of Jordan.

The Department of Civil Status registers Jor-
danians in a special “Civil Register,” which is a
family register based on the head of the family.
The detailed registration form provides informa-
tion about all family members, including educa-
tion, occupation, and marital status. In addition,
each head of the family receives a 36-page Fam-
ily Book where every member has a page to
record personal details and changes. When a fam-
ily member reaches age 16, he is provided an
identity card that facilitates travel in Jordan and
the Arab common market countries.

44

Under the new system, a PN is issued at
birth and remains with the individual through-
out his life. This number appears in the family
register, in the Family Book, on the I. D. card,
and on all official documents. The 13-digit
number is constructed as follows:

2 digits — place of birth

2 digits — parents’ place of birth

3 digits — year of birth

1 digit — sex (1—male, 2—female)

5 digits — serial number

As an example, the number 12219552 45894,
would identify a woman born in 1955 at
Amman, whose parents were born at Naour. Pre-
sumably, she was the 45,894th child born in
Amman in that year. The PN has been used for
administration and identification and, therefore,
has improved civil registration. The development
of a more complex numbering system is not con-
templated now.

HEALTH, SOCIAL SECURITY, AND
INSURANCE SYSTEMS

National health records systems using a PN
as an identifier exist only in a few countries and
are highly restricted. Considerations of privacy
and confidentiality limit access to hospital and
physicians records even for statistical research.
Two systems of special interest are the fully op-
erational system of the United Kingdom and the
proposed system of Australia.

United Kingdom

The initial use of a PN in the United King-
dom coincided with the wartime establishment
of the National Register in September 1939.
After a census enumeration of the population,
identity cards were issued to all recorded per-
sons by the Central National Registration Office.
To record births, a unique code of four letters
and a serial number (1-500 in each subdistrict),
was developed. The system of national registra-
tion ended in 1952, but prior to that, the Na-
tional Health Service (NHS) needed an identify-
ing number for each patient and the obvious
choice was the National Registration code. In
1965, a new code was established, which con-
sisted of five letters and a birth register entry
number. Three letters indicated the subdistrict
of birth; the fourth letter, the year of registra-
tion (M=1979); and the fifth letter, the quarter
of registration. The last is useful because births
are collated and indexed on a quarterly basis.

The birth number serves only as a file num-
ber in the NHS and it has no statistical use be-
cause it is a key to any form of general purpose
population register. Considerations of confi-
dentiality impose restrictions on use outside the
NHS system.

Australia

The Australian experience differs from that
in other countries because discussion of a num-
bering system began with a plan to develop med-
ical record linkage throughout the country. A
working group of the Computer Committee of
the Hospital and Allied Services Advisory Coun-
cil, selected in 1971, established criteria for the
identification key to medical records, studied
identification systems in various countries, made
a computer analysis of existing files, and arrived
at a combined Letter-Number indicator. This in-
dicator consisted of the first four characters of
the surname, the first two characters of the first
forename, the second initial, sex, date of birth,
and a check digit. Plans to establish a National
Population Index for identifying populations at
risk, which would be located at a National Index
Centre, were recommended but were not im-
plemented.

Changes in Government in 1972 and 1975
and in the health insurance programs altered the
plans regarding medical record linkage. Health
insurance prior to September 1976 had been
administered by a single Government commis-
sion; after that, private insurers provided this in-
surance. This change required some form of
unique identifier, and a health insurance number
of 10 digits was chosen. This number consisted
of six digits for date of birth and sex, three
digits for a serial number, and one for a check
digit. However, continuing controversy has
delayed implementation.

Other countries

Other limited national numbering systems
exist, mostly in the insurance and social security
systems where a personal account number ap-
plies only to (or primarily to) the records of that
system.

Czechoslovakia.—An Insurance Number is
used, which refers to the beneficiaries file in the
pension administration and consists of nine dig-
its, six for birth date and three for the serial
number.

Switzerland.—An Insurance Number is also
used for systems such as registration, tax, and
sickness funds. The number has 11 digits: 3 for
the surname, 2 for the year of birth, 1 for quar-
ter of the year of birth, 2 for day of birth, 2 for
serial number, and 1 check digit.

Austria.—A Social Account Number of 10
digits is implemented: 3 are for the serial num-
ber, 4 for birth day and month, 2 for decade,
and 1 check digit.

United States.—A Social Security Account
code of nine digits is employed (three for area
number, two for group number, and four for
serial number). In some countries, the codes for
social security and other accounts could be
made into national population PN’s, but the ex-
tension of the existing numbers might be diffi-
cult. In the United States, for example, an esti-
mated 4 million duplicate account numbers
exist.

OTHER NATIONAL SYSTEMS

The PN system has been discussed exten-
sively in all regions of the world. Japan tested
several types of PN’s in urban areas but did not
apply them nationally. In the Federal Republic
of Germany, an extensive system was proposed
but was not approved by Parliament. Inquiries
regarding the PN systems and their applications
to health services have been received by the Na-
tional Center for Health Statistics from such
diverse areas as Thailand and Brazil. In the
United States, discussion of a national number-
ing system has been ongoing for almost 40 years.

Federal Republic of Germany

In 1971, the Federal Ministry of the Interior
had submitted to Parliament a draft of a Federal

45
Registration Law that provided for the introduc-
tion of a uniform personal identification num-
ber. Parliament refused to introduce a PN for
registration and administration.

A special personal number, an “insurance
number,” has been issued to members of the
statutory old-age insurance group as well as to
employees entitled to obligatory health and/or
unemployment insurance (wage earners, salaried
employees, and persons undergoing occupational
training) who are not subject to obligatory old-
age insurance.

This special 12-digit PN is composed of a
regional number for the old-age insurance funds
(2 digits), the date of birth (6 digits), initial
letter of the birth name (1 digit), the serial num-
ber and distinction by sex (2 digits), and the
check digit (1 digit). This PN is used only for
maintaining accounts of the insured, handling
queries within the scope of statutory social in-
surance, and compiling statistics on employed
persons.

Protection regarding privacy and confiden-
tiality is established by law because the secrecy
provisions of the Social Code permit the disclo-
sure of microdata to third parties when either
the individual agrees to or when legal obligation
demands disclosure. Other agencies of social
administration receive microdata subject to the
secrecy provisions within the scope of admini-
strative assistance, that is, when data are neces-
sary for the completion of their tasks. A linking
of statutory old-age insurance data with other
registers or files has not been provided for, par-
ticularly because other institutions (e.g., health
insurance) have numbering systems of their own
for their members and the linking of data is
doubtful under the data protection law.

Japan

In 1971 and 1972, Japan tested various
types of PN’s in five different cities. One PN was
a 10-digit code, including 1 check digit. Another
was a 14-digit number with the first 6 numbers
representing the birth date, the next 4 repre-
senting the area of residence, the next 3 repre-
senting a serial number, and the last one a check
digit. The PN activity never advanced past the
testing stage.

46

Portugal

The numbering system of Portugal is admin-
istered by the Bureau of National Registration
of the Ministry of Justice. Since 1957, a perma-
nently assigned number has been issued sequen-
tially and placed on identification cards. The
identification card is not mandatory for the
population; however, it is required for attending
high school, obtaining a driver’s or a marriage
license, and declaring taxes. It has been custom-
ary to obtain a card after age 12. The identifica-
tion number is used for many administrative
purposes such as tax control, health care and
insurance, and social security.

The National Register was established by law
in 1973. The objective was to create two central
files, the Central File of Population and the Cen-
tral File of Organizations. The population file
was to be based on a new national number pro-
vided after birth, which would consist of 11 dig-
its; 6 for date of birth, 4 for the file number (the
first included a code for Portuguese nationals
and foreigners, and the last distinguished men
from women), and 1 check digit. The implemen-
tation of the population file was suspended after
the revolution of April 1974. The new Constitu-
tion of the Portuguese Republic of April 1976 is
very explicit in the use of Person Numbers:
“Citizens shall not be given all-purpose national
identification numbers” (Article 35). The Gov-
ernment is preparing drafts for a law on privacy.
The Constitution is to be revised in 1980, and
apparently the Central File of Population will be
discussed. In the meantime, the file on civil
identification and the number on the identifica-
tion cards are being kept in force.

The Central File of Organizations includes
enterprises, associations, agencies, official bod-
ies, and self-employed individuals. A nine-digit
number (type of organization, serial number,
and check digit) is the key to this file.

United States

The concept for the PN system originated
during World War II when State and local vital
statistics offices were deluged with requests for
certifications of birth, for proof of citizenship
and age. In 1941 at the urging of the Association
of State and Territorial Health Officers
(ASTHO), a Vital Records Commission was
appointed to investigate the system and suggest
improvements. The Commission suggested a
fixed identification number for each person in a
national registration. In 1947, the Council of
Vital Records and Vital Statistics suggested that
each State adjust its birth certificate number to
conform to a uniform numbering plan for the
entire country. A uniform number in three seg-
ments—three digits for country and State of
birth, two digits for year of birth, and six digits
for State file number—was suggested. This sug-
gestion was cleared with the ASTHO and was
subsequently ratified by 35 out of 54 registrars.
For various reasons, the concept dropped

out of sight until the midsixties. In 1966, the
Study Group on Record Linkage of the U.S.
Public Health Conference on Records and Sta-
tistics recommended that the birth number, as
suggested in 1948, be placed on all State certifi-
cates by January 1, 1968. The U.S. National
Committee on Vital and Health Statistics also
endorsed State use of the number. Although the
“universal” number now appears on State birth
certificates, the Federal government does not
use it. Inquiries concerning it have been made by
the Social Security Administration. States use
the number for filing and for exchanges of cer-
tificates between States. For the U.S. birth num-
ber, or any other PN, to become truly opera-
tional, a Federal law establishing it throughout
the country has to be passed by the Congress.

47
REFERENCES

ICentral Bureau of Statistics: Divorces 1971-1973.
Statistiske Analyser No. 16. Oslo. Central Bureau of Sta-
tistics, 1975.

2Central Bureau of Statistics: Occupational Mortality
1970-1973. Statistiske Analyser No. 21. Oslo. Central
Bureau of Statistics, 1976.

3Central Bureau of Statistics: Occupational Mortal-
ity. RAPP 79/19. Oslo. Central Bureau of Statistics,
1979.

4Ministry of Justice: Offentlige persondatasystemer
og personvern (Public personal data systems and protec-

48

tion of [personal] privacy.) NOU 1975:10. Oslo. Min-
istry of Justice, 1975.

5Ministry of Justice: Persondata og personvern (Per-
sonal data and protection of privacy.) NOU 1974:22.
Oslo. Ministry of Justice, 1974.

6U.S. Department of Commerce, Bureau of the
Census: World Population 1977: Recent Demographic
Estimates for the Countries and the Regions of the
World. Washington. U.S. Government Printing Office,
1978.

7United Nations: Demographic Yearbook 1977. New
York. United Nations, 1978.
BIBLIOGRAPHY

Aurbakken, E., and Bjerve, P. J.: The role of regis-
ters and the linking of data from different sources.
Bulletin of the International Statistical Institute, Vol.
XLV, Book 3, 1973. pp. 169-184.

Bacci, R., Baron, R., and Nathan, G.: Methods of
record linkage and applications in Israel. Bulletin of the
International Statistical Institute, Vol. XLII, Book 2,
1967. pp. 766-785.

Bakketeig, L. S., and Hoffman, H. J.: Perinatal
mortality by birth order within cohorts based on sibship
size. Brit. Med. J. 2:693-696, 1979.

Bakketeig, L. S., Hoffman, H. J., and Sternthal,
P. M.: Obstetric service and perinatal mortality in Nor-
way. Acta Obstet. Gynecol. Scand., Suppl. 77, 1978.

Bakketeig, L. S., Seigel, D. G., and Sternthal, P. M.:
A fetal-infant life table based on single births in Norway,
1967-1973. Am. J. Epidemiol., 107(3):216-255, 1978.

Bolander, A. M.: Linkage of census and death rec-
ords to obtain mortality registers for epidemiological
studies in Sweden. Excerpta Medica: Proceedings of the
11th International Cancer Congress. Florence, Italy,
1974.

Central Bureau of Statistics: The Use of the Nether-
lands System of Continuous Population Accounting for
Population Statistics, by J. C. van den Brekel. Voorburg
(The Hague). Department for Population Statistics,
1976.

Dalenius, T.: The invasion of privacy problem and
statistics production—an overview. Stat. Tidsskr. No. 3:
213-225, 1974.

Departamento Administrativo Nacional de Estadis-
tica. (DANE). Manuel de instrucciones para effectuar el
registro de nacimiento. Bogota. Direccion General de
Procesamiento de Datos, Servicio Nacional de Inscrip-
tion, Registro de Personas, 1973.

Departamento Administrativo Nacional de Estadis-
tica (DANE). Ventajas del nuevo sistema de registro
civil. Bogota. Servicio Nacional de Inscription, 1973.

Department of Commerce: Computers, Health Rec-
ords, and Citizen Rights. National Bureau of Standards
Monograph 157. Washington. U.S. Government Print-
ing Office, 1976.

Department of Health, Education, and Welfare:
Records, Computers, and the Rights of Citizens. DHEW
Pub. No. (OS) 73-94. Washington. U.S. Government
Printing Office, July 1973.

Desabie, J.: Organismes: L’INSEE entreprend
d’automatiser le repertoire des personnes. Econ. Stat.,
No. 10:69-71, Mar. 1970.

Dunn, H. L.: Record linkage. Am. J. Pub. Health
36:1412-1416, Dec. 1946.

Dunn, H. L.: A national identity registration system
to synthesize social statistics. Estad.: J. Interam. Stat.
Inst. X1(40):605-615, Sept. 1953.

Federal Security Agency: Historical Review of the
Birth Number Concept. National Office of Vital Sta-
tistics, Public Health Service, Washington, D.C., 1948.

Hansen, M. H.: The role and feasibility of a national
data bank, based on matched records and alternatives.
Report of the President’s Commission, Washington,
D.C.: Fed. Stat. 11(1-61): 1971.

Hoffman, H. J., Bakketeig, L. S., and Stark, C. R.:
Twins and perinatal mortality: a comparison between
single and twin births in Minnesota and Norway 1967-
1973, in Twin Research: Biology and Epidemiology.
New York. Allan R. Liss, Inc., 1978. pp. 133-192.

Institut National de la Statistique et des Etudes
Economiques (INSEE): Automatisation du repertoire
d’identification des personnes physiques. Paris. INSEE,
Conception Generale, Feb. 1970.

Institut National de la Statistique et des Etudes
Economiques (INSEE): Le numero national d’identite
description. No. 160/429. Paris. INSEE, Jan. 19, 1971.

Karlsen, K., and Skaug, H.: Registers in the Central
Bureau of Statistics. Pub. No. 22. Oslo. Statistiske Sen-
tralbyra, 1968.

Lunde, A. S.: The birth number concept and rec-
ord linkage. Am. J. Pub. Health 65(11):1165-1169,
Nov. 1975.

Lunde, A. S.: National data systems and record link-
age. International Statistical Institute: Contributed
Papers (Warsaw), Sept. 1975. pp. 536-541.

Lunde, A. S.: Problems in the Establishment of
National Data Systems. Proceedings of the Social Sta-
tistics Section (135th Annual Meeting, Atlanta), Ameri-
can Statistical Association, 1975. pp. 544-547.

Ministry of Interior: Population Register of Israel.
Jerusalem. Ministry of Interior, 1972.

Nathan, G., and Baron, R.: The Israeli Population
Register as a Framework for Sample Surveys. Interna-
tional Symposium on Automation of Population Regis-
ter Systems, Jerusalem, Sept. 25-28, 1967.

National Institute of Statistics: INE-VISTIM Vital
Statistics Improvement Final Project Plan. (translation
from Spanish). Lima. National Institute of Statistics,
1978.

Nordbotten, S.: Purposes, problems, and ideas re-
lated to statistical file systems. Bulletin of the Interna-

49
tional Statistical Institute, Vol. XLII, Book 2, 1967. pp.
734-750.

Ohlsson, I.: Merging of data for statistical use. Bul-
letin of the International Statistical Institute, Vol. XLII,
Book 2, 1967. pp. 750-764.

Selmer, E. S.: Registration numbers in Norway:
some applied number theory and psychology. J. Roy.
Stat. Soc. (Series A), 130, pt. 2, 225-231, 1967.

Statistiches Bundesamt, Federal Republic of Ger-
many: Betriftt: Personen Kennzeichen. Bonn. Statis-
tiches Bundesamt, 1972.

Statistisk Sentralbyra: The Central Registers of the
Central Bureau of Statistics of Norway. Working Papers
of the Central Bureau of Statistics, Oslo, 1974.

50

Statistisk Sentralbyra: Fetal and Infant Mortality
1969-1972. Statistical Analyses No. 15. Oslo. Statistisk
Sentralbyra, 1975.

Tryggveson, R.: Population Registration and Its Use
for Statistical Purposes. International Symposium on
Automation of Population Register Systems, Jerusalem,
Sept. 25-28, 1967.

United Nations: Methods of Collecting, Organizing,
and Retrieving Social Statistics to Achieve Integration:
Report of the Secretary General. (E/CN.3/516). New
York. United Nations Economic and Social Council,
1978.

Widen, L: Registret over femtondefodda. Stat.
Tidsskr., No. 5:366-372, 1969.
II.

IIL.

VIL

VIL

APPENDIXES

CONTENTS

Construction of the PN in Sweden

 

Construction of the PN in Norway

 

Structure of the PN in Denmark

 

Construction of the PN in Israel

 

Items in the Swedish Cancer-Environment Register

 

Items in Study “Outcome of Successive Pregnancies for Norwegian Women 1967-1976" ......cccceunee

Population, Birth, and Death Data

 

APPENDIX FIGURE

Construction of Person Number for a male born on June 3, 1936: Denmark .....ccccerresncescsanssccsenes

APPENDIX TABLE

Total population and number of births and deaths with rates of Sweden, Norway, Denmark,
Israel, and the United States of America

 

52
53
54
55
56
57
59

54

59

51
APPENDIX |
CONSTRUCTION OF THE PN IN SWEDEN

The PN’s were adopted in 1947 when the
regional registers were set up at county admin-
istration levels. Originally, these numbers con-
sisted of nine digits which included the date of
birth (six digits) and a birth number (three
digits). Date of birth is indicated by two digits
for the year, two for the month, and two for the
day. For example, the date of birth of a person
born April 25, 1938, is written 380425. The use
of this reverse order—year, month, and day—was
farsighted at the time; recently the International
Standards Organization declared that numerical
dates should always be expressed in that order.
The birth number consists of three digits rang-
ing from 001 to 999. Odd numbers stand for
males and even ones for females.

When the existing county registers, which
were printed from metal plates, were replaced in
1967 by magnetic tape registers, a check digit
had to be added to the date of birth and the
birth number. This check digit was arrived at by
using Modulus 10 algorithms. Therefore, a PN
now contains 10 digits. The date of birth and
the birth number/check digit are separated by
a hyphen (e.g., 380425-6653). The year a per-
son reaches 100 years of age the plus sign
supersedes the hyphen.

In the ADP system, the Modulus 10 algo-
rithm is used for the calculation of the check

digit, and the weights 2 and 1 are used. This
enables an identity number to be checked
automatically when data are registered. The
method of calculation is illustrated as follows:

Identity number 450410 149
Weights X 212121 212
Products 850420 2418

Sum of products 8+5+0+4+2+4+1+8 = 34
(N.B.: 18 is read as 1+8.)

The last digit of the sum is subtracted from 10
(i.e., 10-4=6), and the answer becomes the
check digit. The complete identity number be-
comes 450410-1496.

The identity number is checked when

recorded.

Identity number 4504101496
Weights X 2121212121]
Products 85042024186

Sum of products 8+5+0+4+2+0+2+4+1+8+6=40

If the last digit of the sum of the product is not
0, then an error must obviously have occurred;
this is indicated by the output.

000

52
APPENDIX II
CONSTRUCTION OF THE PN IN NORWAY

The birth number system includes both the
identification number and the routines to assign
and maintain the identification numbers. The
Norwegian identification number is based on the
date of birth and the sex of each person. The
number consists of 11 digits and is constructed
as follows:

Date of Birth Individual Check
Day Month Year digits digits

26 05 97 651 31

The last individual digit indicates sex. If the
person is a woman, the number is even. If the
person is a man, the number is odd.

Persons born in the 19th century are as-
signed numbers within the 749-500 range, and
those born in the 20th century are assigned
numbers in the 499-000 range. Numbers are as-
signed consecutively in descending order.

The last two digits are check digits. The
first digit is calculated by weighting the date of
birth and the individual digits with standard
weights.

 

Birth number 2 605 9 7 6 51
Weights X 3 761 8 9 4 5 2
Products 6 42 0 5 72 63 24 25 2
Sum of

products 6+42+0+5+72+63+24+25+2 = 239

The sum is divided by 11 (i.e. 239 +11 =21
with a remainder of 8 where 7; =the remainder)
The first check digit is K; =11-7 (3 in pre-
ceding sample).

If; =0, the k; =0.If »; = 1, then the num-
ber is rejected and the next individual digit in
the range is assigned.

The second digit is calculated in the same
way; however, IBM standard weights are used:

 

Birthnumber 2 60 53 9 7 6 51 3
Weights X5 43 2 7 6 5 432
Products 10 24 0 10 63 42 30 20 3 6

Sum of
products 10+24+0+10+63+42+30+20+3+6 = 208

The sum is divided by 11 (i.e., 208 +11 =18
with a femeinig of 10). The second check digit
is kg =11- 179 =1. The complete PN becomes
26 05 97 651 31. The five last digits of the birth
number, that is, the individual digits and the
check digits, are called the personal number.

Every PN is calculated by a special auto-
matic EDP routine. This routine checks which
numbers are in use and which are not. A birth
number that once has been used will never be
used again.

The Central Bureau of Statistics has the re-

- sponsibility of assigning birth numbers twice a

month. Lists comprising the assigned birth num-
bers are then sent to the various local registra-
tion offices.

000

53
APPENDIX 111
STRUCTURE OF THE PN IN DENMARK

The Person Number is purely numerical and
consists of 10 digits.

Figure I shows how the number is con-
structed.

The check digit is computed by means of the
Modulus 11 algorithm on the basis of the other
nine digits which are weighted by a set of con-
stants as follows:

 

Identity 030 6 3 611 7
Constants X4 39 16 5493 2
Products 09042 18 30 4 3 14
Sum of

products 0+9+0+42+18+30+4+3+14 = 120

The sum is divided by 11 (i.e, 120+11=10
with a remainder of 10). The check digit is
found by subtracting the remainder from 11
{ie,11-10=1),

By employing this method certain combina-
tions of the first 9 digits give a check digit of
10; these combinations cannot be used. If the
result of the calculation is a check digit of 11
(i.e., 11 - a remainder of 0), the check digit is
given the value of 0.

When the check digit is odd, as in this case,
the number is assigned to a man. When it is even,
it is assigned to a woman.

 

 

Reference number
Date of birth

 

pipe. |
[3 % 3 nn
Day of birth I LI. |
Month of birth
Year of birth
Century of birth
Check digit, sex

 

 

 

 

Figure I. Construction of Person Number for a male born
June 3, 1936: Denmark

000

54
APPENDIX IV
CONSTRUCTION OF THE PN IN ISRAEL

The original PN of Israel, developed in con-
nection with the census of 1948, consisted of six
digits as in the number 114888. Later, a series
prefix was added using letters “a” through “‘g’’;
with a “d” prefix the number would become
d/114888. These letters were often assigned to
designated groups; for example, an “a” series
block was given to immigrants arriving by ship,
and series ‘‘¢”’ was provided to maternity hospi-
tals. The introduction of a mechanical card
index system in 1952 required that the letter
prefix be changed to a number; “a” became 1
. and so on, with the result that the PN became a
seven-digit number (in our example, “d” became
4, and the number was 4114888). In 1966 the
content of the mechanized residents card index
was transferred to magnetic tape and, subse-

quently two changes were introduced. First, the

base number was extended by one digit to ac-
commodate the needs of a future population of
10,000,000 or more; this created an eight-digit
PN (04114888). Second, a check digit was intro-
duced, calculated on the Modulus-10 algorithm,
which provided a nine-digit PN. The method of
calculation is as follows:

Identity number 04114 88 8
Weights X 12121 21 2
Products 08124163816

Sum of digits 0+8+1+2+4+1+6+8+1+6 = 37

When the last digit of the sum is subtracted
from 10, the answer (3) becomes the check digit
and the complete PN reads 041148883.

O00O0

55
APPENDIX V
ITEMS IN THE SWEDISH CANCER-ENVIRONMENT REGISTER

The Swedish Cancer-Environment Register is
administered by Planning Division 3, Statistics:
National Board of Health and Welfare, S-106 30,
Stockholm, Sweden. Data were integrated from
the Cancer Register and the 1960 Census of
Population and Housing.

Data from the Cancer Register (persons
registered in 1961-73) include:

Person Number (PN)

Tumor serial number

Treatment

Died from cancer

Diagnosis incidentally at autopsy
Benign

Causes of death

Date of death

Age at diagnosis

Age at death

Sex Survival time

Name

Occupation Data from the 1960 Census of Population
Civil status and Housing include:

Domicile

Date of diagnosis
Date of death
Hospital, department

Record-card number and year of record
entry

Pathologist/cytologist

Specimen number and year specimen taken
Site of tumor

PAD (Pathological diagnosis after death)
Diagnostic criterion

Metastasis

(oN oNe

56

Domicile

Type of activity

Occupational status

Occupation

Economic activity

Gainful employment

Location of place of work

Main occupation during the year
Country of birth

Higher education

Person Number (PN)
APPENDIX VI

ITEMS IN STUDY “OUTCOME OF SUCCESSIVE PREGNANCIES
FOR NORWEGIAN WOMEN 1967-1976"

This groundbreaking research is being con-
ducted by Dr. Leiv S. Bakketeig, Institute of
Community Medicine, University of Trondheim,
Norway. It is receiving support from the Na-
tional Institute of Child Health and Human De-
velopment, National Institutes of Health, U.S.
Public Health Service.

Variables are integrated from three sources:
the Norwegian Medical Birth Registry (main-
tained at the University of Oslo), the Norwegian
Death Files (Central Bureau of Statistics), and
the Census File.

Data from the Medical Birth Registry in-
clude:

A. Variables relating to the mother:

1. Maternal age—by year, date of birth

2. Maternal parity—number of previous
births and among these number of

stillbirths

3. Maternal marital status—unmarried,
married, divorced, separated,
widowed

4. Time of marriage—by calendar year

5. Place of residence—by municipality
(444 in the country), characterized
by county, urban, mal, industrializa-
tion

6. Duration of pregnancy—days based
on date of birth minus date of LMP

7. Mother’s health before pregnancy—
maximum 3 conditions, 3-digit ICD
code

8. Mother’s health during pregnancy—

maximum 3 conditions, 3-digit ICD
code

B. Variable relating to the father:

1.

Paternal age—by year, date of birth

C. Variables relating to the family:

1.

Family relations between parents—
1st cousins, 2d cousins, etc.

Family history of disease—serious,
inheritable disease among relatives—
maximum 3 conditions, 3-digit ICD
code, 4th digit indicating relation-
ship of the newborn baby to the
family member suffering from the
disease

D. Variables relating to the confinement:

1.

Place of birth—by municipality and
by institution, classified by level of
obstetric service

. Time of birth—day, hour, and

minute

Induction of labor—special code/
classification 1-9

Fetal presentation—special
classification 1-9

code/

. Complications during labor—special

code, maximum 4 classifications

Intervention during labor—special
code

57
7. Intervention performed—by physi-
cian, by midwife

8. Information on amniotic fluid, pla-
centa, umbilical cord—special code

E. Variables relating to the birth:

1. :Sex
2. Plurality
3. Status—fetal death/stillbirth, died be-

fore onset of labor, during labor or
within unknown time of death; live-
births dying during first 24 hours,
1-6 days, 7-27 days, 28 days to 1
year, end year, 3d year or later; live-
births still alive

Birth weight—nearest 10 grams
Length
Asphyxia—special code

NO Ov Wb

Congential malformations, birth in-
juries, and diseases—maximum 3 con-
ditions, 4-digit ICD codes

Data from the Norwegian Death Files (Cen-

tral Bureau of Statistics) include:

58

. Place of residence at time of death—

municipality

000

NO Ov Wb 00 IN

Place of death—special code
Sex

Date of birth
Date of death
Type of death—accident or disease

Causes of death—maximum 4 diagnoses,
4-digit ICD code

For dead children less than 2 years of
age—marital status of mother

Data from the Census File include:

A. Information on the mothers:

1. Type of income/support

2. General education—Norwegian stand-
ard classification

3. Highest education—occupational edu-
cation, academic degree, and so forth

4. Occupational  activity—housewife,
student, employee, and so forth

5. Industry—standard International In-
dustry classification (ISIC)

Information on the fathers:

Same as above, except for “housewife”
classification
APPENDIX VII

POPULATION, BIRTH, AND DEATH DATA

Table |. Total population and number of births and deaths with rates of Sweden, Norway, Denmark, Israel, and the United States of

America

[ Midyear population estimates for 1977; births and deaths with rates for 1976]

 

 

 

  
 

 

 

 

 

 

Total Number of Birth Number of | Death

Country population? births? rates? deaths? rates?
BWBTIBIY 1. cvissinnsissivsssmmmmnnesnissssssanss bn shnnsnaessssnssssnhhesenenins avs vastunehunuananiius shes 8,255,000 98,345 12.0 90,677 11.0
INGIEWEY « cnvansimmmrsvasumisisdssssmmin ie sass sass ss AS aan ARI SAA A AER EE CHE CARTS TTA EARS 4,044,000 53,474 13.3 40,216 10.0
Denmark.. _— 5,089,000 65,267 12.9 54,001 10.6
Ce PR OOO LL AE RUON 3,611,000 97,469 27.6 23,856 6.8
UNIBd States oc. miiimmmnninsmmrinimnmissisnaasneamsmmanis 1 216,817,000 3,165,000 14.7 1,912,000 8.9

 

SOURCES: 1U.S. Department of Commerce, Bureau of the Census: World Population 1977: Recent Demographic Estimates for the

Countries and Regions of the World. Washington. U.S. Government Printing Office, 1978 (reference 6).

2United Nations: Demographic Yearbook 1977. New York. United Nations, 1978 (reference 7).

000

#U.S. GOVERNMENT PRINTING OFFICE: 1980 311-240/20 1-3

59
 
 

 

Series 1.

Series 2.

Series 3.

Series 4.

Series 10.

Series 11.

Series 12.

Series 13.

Series 14.

Series 20.

Series 21.

Series 22.

Series 23.

VITAL AND HEALTH STATISTICS Series

Programs and Collection Procedures.—Reports which describe the general programs of the National
Center for Health Statistics and its offices and divisions and data collection methods used and include
definitions and other material necessary for understanding the data.

Data Evaluation and Methods Research.—Studies of new statistical methodology including experi-
mental tests of new survey methods, studies of vital statistics collection methods, new analytical
techniques, objective evaluations of reliability of collected data, and contributions to statistical theory.

Analytical Studies.—Reports presenting analytical or interpretive studies based on vital and health
statistics, carrying the analysis further than the expository types of reports in the other series.

Documents and Committee Reports.—Final reports of major committees concerned with vital and
health statistics and documents such as recommended model vital registration laws and revised birth
and death certificates.

Data From the Health Interview Survey.—Statistics on illness, accidental injuries, disability, use of
hospital, medical, dental, and other services, and other health-related topics, all based on data collected
in a continuing national household interview survey.

Data From the Health Examination Survey and the Health and Nutrition Examination Survey.—Data
from direct examination, testing, and measurement of national samples of the civilian noninstitu-
tionalized population provide the basis for two types of reports: (I) estimates of the medically defined
prevalence of specific diseases in the United States and the distributions of the population with respect
to physical, physiological, and psychological characteristics and (2) analysis of relationships among the
various measurements without reference to an explicit finite universe of persons.

Data From the Institutionalized Population Surveys. —Discontinued effective 1975. Future reports from
these surveys will be in Series 13.

Data on Health Resources Utilization. —Statistics on the utilization of health manpower and facilities
providing long-term care, ambulatory care, hospital care, and family planning services.

Data on Health Resources: Manpower and Facilities. —Statistics on the numbers, geographic distri-
bution, and characteristics of health resources including physicians, dentists, nurses, other health
occupations, hospitals, nursing homes, and outpatient facilities.

Data on Mortality. — Various statistics on mortality other than as included in regular annual or monthly
reports. Special analyses by cause of death, age, and other demographic variables; geographic and time
series analyses; and statistics on characteristics of deaths not available from the vital records based on
sample surveys of those records.

Data on Natality, Marriage, and Divorce.—Various statistics on natality, marriage, and divorce other
than as included in regular annual or monthly reports. Special analyses by demographic variables;
geographic and time series analyses; studies of fertility; and statistics on characteristics of births not
available from the vital records based on sample surveys of those records.

Data From the National Mortality and Natality Surveys.—Discontinued effective 1975. Future reports
from these sample surveys based on vital records will be included in Series 20 and 21, respectively.

Data From the National Survey of Family Growth.—Statistics on fertility, family formation and dis-
solution, family planning, and related maternal and infant health topics derived from a biennial survey
of a nationwide probability sample of ever-married women 15-44 years of age.

For a list of titles of reports published in these series, write to: Scientific and Technical Information Branch

National Center for Health Statistics
Public Health Service
Hyattsville, Md. 20782

 
 

 

  

VITAL and HEALTH STATISTICS

ATA EVALUATION AND METHODS RESEARCH

 

 

LOTTE TC RET TC CLE LC OTE

U.S. DEPARTMENT OF

HEALTH AND HUMAN SERVICES

Public Health Service

Office of Health Research, Statistics, and Technology
National Center for Health Statistics

 
 

 

 

Library of Congress Cataloging in Publication Data

Woolsey, Theodore D.
Toward an index of preventable mortality.

(Vital and health statistics: Series 2, Data evaluation and methods research; no. 85)
(DHHS publication; (PHS) 81-1359)

Includes bibliographical references.

Supt. of Docs. no.: HE 20.6209:2/85

1. Mortality—United States—Statistical methods. 2. Death—Causes—Statistical methods.
3. Health status indicators—United States—Statistical methods. 4. United States—Statistics,
Medical. 5. United States—Statistics, Vital. I. Title. II. Series: United States. National Center
for Health Statistics. Vital and health statistics: Series 2, Data evaluation and methods re-
search; no. 85. III. Series: United States. Dept. of Health. DHHS publication; (PHS) 81-
1359. [DNLM: 1. Health surveys—United States. 2. Mortality—United States. W2 A N148vb
no. 85] g
RA409.U45 no. 85 312'.0723s [312'.2]
ISBN 0-8406-0189-1 80-607087

 

 

Wm en Wort Min A RR er EA DR mn Be BE WON asi mani Hime WI CA min mt a i AE IR ao A a PRA 5 RRA RE iain mpl pn

 
DATA EVALUATION Series 2
AND METHODS RESEARCH Number 85

Toward an Index of Preventable Mortality

This report presents a research study on the development of an
index that reflects the extent to which an area’s mortality rates ex-
ceed the lowest possible rates that could be achieved at this epoch
in the United States. The index is applied to mortality rates in 19
selected health service areas during 1969-71.

DHHS Publication No. (PHS) 81-1359

 

U.S. DEPARTMENT OF HEALTH AND HUMAN SERVICES
Public Health Service
Office of Health Research, Statistics, and Technology
National Center for Health Statistics
Hyattsville, Md. May 1981
NATIONAL CENTER FOR HEALTH STATISTICS

DOROTHY P. RICE, Director

ROBERT A. ISRAEL, Deputy Director
JACOB J. FELDMAN, Ph.D., Associate Director for Analysis and Epidemiology
GAIL F. FISHER, Ph.D., Associate Director for the Cooperative Health Statistics System
GARRIE J. LOSEE, Associate Director for Data Processing and Services
ALVAN O. ZARATE, Ph.D., Assistant Director for International Statistics
E. EARL BRYANT, Associate Director for Interview and Examination Statistics
ROBERT C. HUBER, Associate Director for Management
MONROE G. SIRKEN, Ph.D., Associate Director for Research and Methodology
PETER L. HURLEY, Associate Director for Vital and Health Care Statistics
ALICE HAYWOOD, Information Officer

DIVISION OF ANALYSIS

JOEL C. KLEINMAN, Ph.D., Director

Vital and Health Statistics-Series 2-No. 85

 

DHHS Publication No. (PHS) 81-1359
Library of Congress Catalog Card Number 80-607087

td i
CONTENTS

Introduction
Background of the Research
New Measures of Health Status and ‘‘Synthetic Estimates”
Premises of the Research

 

 

 

 

Use of Mortality Statistics to Measure Health Status

 

General Plan of the Research

 

Achievable Target Death Rates
Why Such a Standard?
First Considerations
Two Methods That Proved Unsatisfactory
Use of Geographic Variation
Methodology and Results

 

 

 

 

 

 

Alternative Forms of the Index
Desiderata
Forms of Index Examined and Their Sampling Variability

 

 

 

Computing and Testing the Indexes
Health Service Areas Selected for Testing and Computation Methods ......ceeereresnccsncsansncsssesnesssnsnnes
Tests of Sensitivity
Conclusions

 

 

 

Future Research
Improvement of the Achievable Target Death Rates
The Form of Indexes and Their Random Variability
Geographic Variability
“Market Testing” Mortality Indexes for Health Service Areas

 

 

 

 

 

References

 

List of Detailed Tables

 

TEXT FIGURE

1. Illustration of smoothing process for achievable target death rates: bronchitis, emphysema, and
asthma

 

LIST OF TEXT TABLES

A. Estimating achievable target death rates for males aged 1-85 years and over for bronchitis, em-
physema, and asthma: United States, 1969-71 and 1976, and geographic divisions, 1969-71.........

 

B. Achievable target death rates by sex, cause of death, and age

C. Ratio of achievable target death rates to U.S. death rates in 1976, by sex, cause of death, and
age

 

© nN NN ==

10

11

12
Number of population, and percent and number of selected characteristics of 19 health service
areas used in the illustrative computations, by selected health service areas...............ccoeerveessneesenseses

Selected rank orders and range of coefficients of variation (CV) for three forms of mortality
index (R;, Rf, and R;") for males, by cause-of-death group: 19 health service areas, 1969-71 ........

Highest and lowest value, modified range (MR), mean standard error (§;) and ratio for three forms
of mortality index (Rj, Rj", and SMR;) for males, by cause-of-death group: 19 health service
areas, 1969-71

 

 

SYMBOLS

Data not available------=-----cnemeeeeeemeeeeeceeeeee ---

Category not applicable------------csceeeeeecaacacnee

 

Quantity zero -
Quantity more than 0 but less than 0.05--- 0.0

Figure does not meet standards of
reliability or precision-------ceeeeeeeeeeecccceeee

 

 

 

22
TOWARD AN INDEX OF PREVENTABLE MORTALITY

Theodore D. Woolsey, Health Statistics Consultant

INTRODUCTION

Background of the Research

A few words in legislation enacted in 1974
have provided a strong stimulus to research in
the development of indexes of health status for
small-area populations. It has become almost a
standard practice in papers describing the need
for or results of this research to call attention at
the outset to Public Law 93-641 and, in particu-
lar, to Section 1513(b), which requires health
systems agencies to compile statistics on the
health status, health service utilization, and
health resources of the populations they serve.
Other legislation, such as that dealing with med-
ically underserved areas, has had a similar effect.

Enactment of these laws and general growth
in the field of health policy research have led to
a plenitude of papers about the measurement of
health status, many of which describe new ap-
proaches to this old subject. The following four
sources are helpful to those who keep track of
the research in progress: 14

1. A conference held at Tucson, Arizona,
1972,1 conducted by staff of the journal
Health Services Research.

2. A conference held at Phoenix, Arizona,
1976,2 conducted by staff of the Na-
tional Center for Health Services Re-
search, an agency of the U.S. Public
Health Service.

8. The Clearinghouse on Health Indexes,® a
continuing summary of work operated
by the Division of Analysis of the Na-
tional Center for Health Statistics.

4. Health Planning,* a weekly newsletter
published by the National Technical In-
formation Service. This abstracting serv-
ice, in collaboration with the National
Health Planning Information Center, re-
ports on papers and technical works on
health needs measurement and on other
health planning subjects.

Despite the increased research activity evi-
denced in these sources, there is no consensus on
which methods or sources of health status data
are suitable for use in the health planning work
of the health systems agencies. Agency reports
to the Bureau of Health Planning leave the im-
pression that there is as yet little consistency
among the agencies’ approaches to the measure-
ments of health status.’ Their reports express
goals in terms of statistical quantities, and some-
times in terms of health status measures—partic-
ularly the infant mortality rate—but no common
strategy for assessing the health problems of the
community has emerged.

The failure of the research to lead to any
widely accepted practice on the part of the
agencies is largely the result of the following two
causes in combination:

1. The methods being developed for col-
lecting new statistics are too expensive
and, even if applied, would leave the
agency with too little opportunity to
compare its own measures of health sta-
tus with those of other communities.

2. Existing data are either believed to be or
actually are inappropriate to the meas-
urement of health or are not analyzed in
such a way that they can be used for
health measurement.

The research described here is intended to
remedy the second cause of failure.

New Measures of Health Status and
“Synthetic Estimates’

There is no question that new statistics, spe-
cifically designed and collected, could eventually
provide more comprehensive and relevant meas-
ures of health status than could ever be derived
from existing data sources. There is also little
doubt that better health status measures would
improve the effectiveness of health planning.
However, the cost of such new collection would
be very high. This is because the collection
would probably have to be based on new sample
surveys of households, or on surveys of provid-
ers of care, or both. The surveys would have to
be conducted by every one of some 200 health
systems agencies and would be useless unless
they were repeated at least twice a decade. (If
one estimates an absolute minimum of $200,000
direct costs per survey, the annual budget for
such data collection would be about $200,000
X 200 X 1/5, or $8 million per year.) Further-
more, it would be years before the expertise
could be summoned to mount such local area
surveys on a national basis, and even longer be-
fore an analyzable body of data could be
gathered.

The availability of comparable statistics on
health status over time and space is of critical
importance to the health planner. Without the
ability to compare statistics for the planner’s
Jurisdiction with statistics of other jurisdictions
of similar demographic makeup, or to compare
current figures with those for earlier periods, the
planner is severely hampered in making intelli-
gent use of health status indicators. Conversely,
if comparable statistics could be made available,
the planner would have powerful analytical tools
to identify emerging health problems, to estab-
lish priorities for attacking them, and to measure
success in dealing with them. In particular, eval-
uation of program success or failure in terms of

2For example, the sickness impact profile® requires
a household sample survey.

health outcomes, especially given the almost un-
avoidable absence of truly experimental condi-
tions, must depend upon “before versus after”
and “with versus without” comparisons for
controls.

It is needs of this sort, coming from the
principal users of health status indicators for
local areas, that argue against “synthetic esti-
mates: sophisticated, computerized estimates for
local areas based on health statistics from na-
tional sample surveys, plus detailed population
characteristics of the area from the U.S. Bureau
of the Census or other sources. Synthetic esti-
mates are incapable of reflecting changes in
health status that do not occur simultaneously
in communities with similar population charac-
teristics. Therefore, they are useless for program
evaluation.

Premises of the Research

Such considerations led to several premises,
and it is on the following that the present re-
search is based:

1. For the time being and for some time to
come, the health planner must rely on al-
ready available data for constructing in-
dexes of health status.

2. The data must be available not only for a
particular health planning jurisdiction to
measure health at the present time, but
must also be available in comparable
form for all U.S. health planning jurisdic-
tions. The data also should cover the
past (for at least a decade), as well as
permit following trends into the future.

3. Perhaps self-evidently, the indicator
must be perceived by the health planner
to be appropriate to the measurement of
health and “sensitive” to the problems
with which the community is concerned.

USE OF MORTALITY STATISTICS
TO MEASURE HEALTH STATUS

The figures on numbers of deaths classified
by sex, age at death, and other characteristics of
decedents have been used to measure the health
of communities for centuries.”*9 Several recent
reports in the series Statistical Notes for Health
Planners, published by the National Center for
Health Statistics (NCHS), have reviewed the uses
of mortality statistics and some of the precau-
tions that must be observed in their use.10-13

Despite this history, there is a commonly
held opinion that mortality statistics are no
longer adequate measures of health status, prin-
cipally because of their alleged “insensitivity.”14
Shapiro argues, however, that whether those
who are involved with local health planning
should use mortality statistics “as one of the
measures of the health status of local area popu-
lations” is a “non-issue.”’15

No evidence on problems of quality of the
data or on inadequacy of the information to
identify significant health deficits and their
correlates can override the unique character-
istics of mortality statistics. Simply put,
they represent the only continuous source of
information on an unequivocal manifesta-
tion of health status that dates back many
years and is assured of continuity into the
foreseeable future, and the data can be ex-
amined on a geographically disaggregated
level often down to subareas within a city,
for example, or aggregated across civil sub-
divisions for medical market analysis.15

The question for the health planner, accord-
ing to Shapiro, is “how to maximize the utility
of this resource.”!® The author of this report
cannot express more cogently than that why the
research described herein has been launched, but
for further background the reader is referred to
Statistical Notes for Health Planners, No. 3 and
No. 6.11.13

Statistics on numbers of deaths by cause, age
at death, sex, race, and residence of the dece-
dent are available nationally through NCHS and
locally through State health departments and
health departments of larger cities. NCHS pub-
lishes annual volumes containing these figures in
summarized form. (For example, the complete
detail, by cause and age, is not shown for each
sub-State area.) NCHS also makes microdata
tapes available from which totals in almost any
usable detail can be obtained. The national sta-
tistics and those produced by the States do not

completely agree, but each year the discrepan-
cies (resulting chiefly from differences in coding
practices) are fewer and smaller as a result of
cooperative activities between the jurisdictions.

GENERAL PLAN OF THE RESEARCH

Two phases of the research are presented in
this report. They are as follows:

1. The rationale and the methods used
for determining a set of standard death
rates, specific for age, sex, and cause of
death, are discussed. The standard death
rates are intended to be the lowest pos-
sible rates that could be achieved at this
epoch in the United States assuming: (a)
the best scientific knowledge now avail-
able about methods of prevention and
treatment of disease and injury, and (b)
the successful application of this knowl-
edge in an optimum system of health
care, accessible to everyone. These stand-
ard death rates will be referred to as
“achievable target death rates.”

2. Experiments in using achievable target
death rates and actual mortality rates in
health service areas (HSA’s) as the basis
for various forms of mortality indexes
are detailed. Tests of these forms are ex-
amined to learn whether any one form
has advantages over another in terms of
“sensitivity,” a term to be defined later.

Finally, some recommendations for the
course of future research and for needed tabula-
tions of mortality data by cause of death are
presented. It is hoped that these will facilitate
further research and permit HSA’s to begin
“market testing” of the indexes developed.

ACHIEVABLE TARGET
DEATH RATES

Why Such a Standard?

Why should the effort be made to arrive at a
standard set of death rates such as those con-
ceived here? Up to this point, it has been cus-
tomary in this country to use death rates in the
United States as a whole as the normative stand-
ard for local area indexes.!!

The objection to using national death rates
as a standard is that statistical comparisons with
them, either in the form of absolute differences
or ratios, tend to give insufficient weight to
those categories of deaths that the health care
system can and should be attempting to reduce.
For example, one reason that infant mortality
rates in the United States are not as low as they
could be is that mortality among black infants is
currently (1976) about 75 percent higher than it
is among white infants. Yet there is no intrinsic
reason why that should be so. A community’s
health problems should be measured in terms of
how far its experience differs from what could
be achieved, not in terms of how far it differs
from a standard that itself reflects failures.

Furthermore, the use of a set of death rates
that represents the lowest ones achievable, given
successful application of present knowledge,
provides a standard that only needs to be modi-
fied at intervals of a decade or more. Thus the
resulting index yields more valid comparisons
over time.

The use of a standard classified by cause of
death, as is proposed here, reflects a conviction
that at least some degree of disaggregation is
needed, according to the type of health prob-
lem, in order to provide health status indicators
that can do more than just satisfy the curiosity
of the health planner. Does it really help to
know that the all-causes standardized mortality
ratio for Alabama HSA 01 is about 3 percent
above what it would be if the national death
rates by age, color, and sex were being experi-
enced? The answer is probably, “Yes, but it does
not help much.” To know how causes of death
related, for example, to hypertension, to air pol-
lution, to alcoholism, or to emergency health
services in that same community compare with
some achievable standard would appear to have
far more immediate applicability.

First Considerations

Having defined what the standard set of
death rates is intended to represent (see “Gen-
eral Plan of the Research”), the following ques-
tions needed to be answered:

1. In what demographic detail should the
standard set of death rates be expressed?

2. Specifically, what cause-of-death cate-
gories should be used?

3. Most critically, how should the mini-
mum achievable levels be determined?

The first question was easily answered. It has
long been recognized that age at death is a criti-
cal variable when mortality statistics are being
used to compare the health status of communi-
ties. Because of the very steep increase in death
rates from all causes combined with advancing
age, and owing to the heavy influence of age on
death rates for almost all cause-of-death groups,
it is almost essential that the experience
in different age groups be examined separately,
or that age be held constant in making compari-
sons between areas or over time. The latter
method is the basic element of most indexes of
mortality that have been devised and used in the
past!

NCHS uses several standard age-at-death
classifications, among which is an 11-group clas-
sification starting with “under 1 year of age,”
followed by “1-4 years,” and then eight 10-year
groups including “75-84 years,” and finally “85 °
years and over.” This classification was adopted
for the target achievable death rates.

It was also decided that rates should be deter-
mined separately for males and females but not
for the color dichotomy, white people versus. all
other people, so frequently used in categorizing
vital statistics in the United States. The reason
for omitting this dichotomy has perhaps already
been made clear: It was believed there should be
no different standard for minority groups. Of
course, this does not mean that in calculating
mortality indexes one would not examine the -

situation in the minority populations whenever

possible; in fact, such analysis often might help
to pinpoint health problems in the specific
community.

The disaggregation by cause of death ob-
viously introduces problems that are not en-
countered when deaths from all causes are com-
bined. One problem is that of small numbers of
deaths in the time period and geographic area of °
concern. These small numbers result in unstable
indexes arising from chance variation. This ques-
tion will be examined in some detail in later sec-
tions in which tests of the indexes are presented.

A related problem is the organization of the
cause-of-death categories into meaningful and
useful groups without introducing too much dif-
ficulty in the way of small numbers. The units
of the groups are the 4-digit rubrics of the Inter-
national Classification of Diseases, Adapted. In
this report all the data used are in terms of the
Eighth Revision International Classification of
Diseases, Adapted (ICDA-8).16 Furthermore,
the groups have purposely been kept consistent
with one of the NCHS standard recodes. This is
known as the “69-cause list.” (Adaptation to
the Ninth Revision, the classification now in use,
should not pose any particular difficulties.)
However, the degree of detail in the 69-cause re-
code is far too great for the purposes of this re-
port. Some cause-of-death groups are each no
more than 0.00001 of the total. It was arbi-
trarily determined that for local areas the size of
HSA’s, no cause-of-death group could be used
that amounts (for all ages, both sexes, in the
United States as a whole) to less than 1 percent
of the total.

In arriving at decisions about cause-of-death
groups, direct relevance to potential health prob-
lems in the community argues for greater detail,
but the avoidance of overly small numbers of
deaths and the need to keep from producing a
bewildering array of statistics when all indexes
for a community are compiled weigh the scale
heavily toward lesser detail. The compromise
reached in initial effort was 16 categories that
add to all deaths. However, this particular com-
promise was not entirely satisfactory, and in
subsequent stages of the work somewhat more
detail will be used.

The specifics of the groups used are a matter
of major importance in maximizing the useful-
ness of mortality data as indicators of health sta-
tus problems that normally concern a commun-
ity. As has been pointed out elsewhere,13 there
are some kinds of problems (e.g., mental disease
and mental retardation, problems of undernutri-
tion or overweight, problems of sensory impair-
ments, and the disabilities resulting from arthri-
tides) for which death statistics are so poor as to
be nearly useless indicators. But an intelligent

grouping of the causes of death, specifically de-
signed to meet the needs of community health
planning, can, nevertheless, greatly help to spec-
ify the particular problems of the population of
the area.

The selection of cause-of-death categories to
determine an acceptable set of rates was also
strongly influenced in the initial stages by the
past practices of NCHS in reporting vital statis-
tics. It had been decided by this author that the
calculations of indexes would be demonstrated
using the 1969-71 death statistics for a sample
of HSA’s defined geographically exactly as in
the NCHS report ‘“‘Standardized Mortality Ratio
and Years of Life Lost Index: State and Health
Service Areas, 1969-71,” Statistical Notes for
Health Planners, No. 3, Data Supplement.12 For
practical reasons, the numbers of deaths by
cause for the HSA’s were obtained by summariz-
ing already available statistics for counties.
These statistics had already been tabulated using
the aforementioned 69-cause recode. Any com-
bination of ICDA-8 codes inconsistent with that
recode would have required retabulating a large
data base to obtain material for the demonstra-
tion. It might also have resulted in establishing
standards for categories not ordinarily used by
NCHS.

An alternative basic data source for illustra-
tive material that was considered and rejected
was a data tape created at the Univeristy of Mis-
souri by Professor Herbert I. Sauer? The tape
contains numbers of deaths and death rates for
selected causes of death by sex and age for each
U.S. county for the 4-year period, 1968-72. The
reason for not using this otherwise valuable data
source was that the cause-of-death groups did
not correspond to those being published each
year by NCHS.

It would also have been preferable to calcu-
late the indexes for a more recent period of time
but, as will be seen, local population data are
needed, and the most recent point for which
these were available in the detail required was

bSee p- 7 of reference 13 for a description of the
data tape. Professor Sauer can be addressed at the Uni-
versity of Missouri, 111 Professional Bldg., 909 Univer-
sity Ave., Columbia, Mo. 65201.
the time of the 1970 census enumeration. These
considerations and the need to adhere to the cri-
terion that no category should include less than
1 percent of all deaths led to the following

grouping:
Cause-of-death group ICDA-8 codel®

Malignant neoplasms of digestive organs

AN PETTIONEUIN 1 ovusersiversannsanirisanmssirsing 150-159
Malignant neoplasms of respiratory

BYSLCIN sv orisssssissssmssrrsenssssssssssorinnassnssssssss 160-163
Malignant neoplasms of breast....................... 174

Malignant neoplasms of genital organs....180-187
All other malignant neo-

PASI .cvvrvirssasrnnrsenirson Remainder of 140-209
Diabetes mellitus ..cveisecssasssscsrrssrsssesrrrsessssers 250
Diseases of the heart.............. 390-398, 402, 404,

410-429

Hypertension and stroke ............... 400-401, 403,

430-438
Diseases of the arteries, arterioles, and

CAPINALIES oo cviinrersrssmmssnserssnirissnesnsannnsorons 440-448
Acute bronchitis, influenza, and

PREWNONIA crrossrsarmissssvers 466, 470-474, 480-486

Bronchitis, emphysema, and asthma....... 490-493
Major digestive diseases, except cirrhosis
of the liver.....531-533, 540-543, 550-553, 560,

574-575
Cirrhosis Of the IVEY, .ccruisisinmsbrimcssmns 571
Congenital anomalies and diseases of early
INLANCY wo iramserrsinnrsnusnemernnns 740-759, 760-778
All other diseases............. Remainder of 000-799
Accidental injuries and other
BEOMIB cot cnsnunenivacseesenennivbombattedannnine E800-E999

The most obvious shortcomings of this
grouping are the lumping together of All trau-
mas (E800-E999 in ICDA-8 classification) and
the combining of Infective and parasitic diseases
(000-136) with All other diseases. Each of the
classifications Motor vehicle accidents (E810-
E823), All other accidents (E800-E807, E825-
E949), Suicide (E950-E959), and Homicide
(E960-E978) constitutes more than 1 percent of
all deaths, and they are health problems of such
disparity that it is inappropriate to combine
them in the standard. Infective and parasitic
diseases, on the other hand, make up a trifle less
than 1 percent of all deaths in the country as a

whole. In problem areas, however, they might
exceed that by a considerable amount, and they
clearly represent a distinct kind of health prob-
lem, one which requires different solutions.

During the next stage of the research it
would also be desirable to revise the grouping of
cardiovascular diseases. The Ischemic heart dis-
eases (410-413) should have a separate standard,
and Hypertensive heart diseases (402, 404)
might better be combined with Hypertension
and stroke. Another change that would be an
improvement would be to group Chronic ob-
structive lung disease, now coded separately,
with Bronchitis, emphysema, and asthma (490-
493), because of the increasing use of the more
generalized term on death certificates. This
change will become possible with the beginning
of use of the Ninth Revision of the International
Classification of Diseases (ICD-9). There is no
obstacle (other than the need for special tabula-
tions) to such changes as these, but they should
be considered in the light of the changes needed
to convert the system to ICD-9.

Two Methods That Proved
Unsatisfactory

Before describing the method finally used to
estimate the achievable target death rates for the
11 age groups and 16 cause-of-death groups for
males and females (352 values), it is essential to
an understanding of the rationale for the choice
of method to describe in some detail two meth-
ods that were investigated and rejected.

It was first supposed that expert judgments
about achievable target death rates could be
obtained as a byproduct of another study. This
was the study of economic costs of diseases and
illnesses conducted for the National Institutes of
Health by the Georgetown University Public
Services Laboratory.!7 As a part of that study, a
forecast of U.S. death rates by age and sex for
the major cause-of-death groups in the year 2000
was needed. Instead of relying entirely on projec-
tions, as the Social Security Administration has
done in the forecasts it uses,!8 the Public Serv-
ices Laboratory called upon individual experts—
statisticians, epidemiologists, and medical spe-
cialists—to forecast the death rates by making

“use of trend data through 1974. No less than

five experts were used for each cause-of-death
group. For Malignant neoplasms the panel in-
cluded cancer epidemiologists and oncologists;
for Cardiovascular diseases, epidemiologists who
had worked in that field and cardiologists were
included; and so forth. The figures actually used
in the study were based upon arithmetic aver-
ages of the forecasts for each cause-age-sex cell,
with the most optimistic and most pessimistic
extremes excluded. For some cause groups, such
as Accidental injuries, no panels were involved;
the Social Security Administration’s estimates
for the year 2000 were substituted instead.

Although this task was clearly not directly
comparable to the one of making judgments
about achievable target death rates, it seemed to
involve many of the same thought processes and
had the advantage of bringing to bear the exper-
tise of a number of extremely knowledgeable
people. Hence, the first effort made in the pres-
ent research was to adapt the results of the
Public Services Laboratory study to this new
purpose.

Very shortly it became evident that the rates
of actual events were lower than the expert pre-
dictions in a number of the disease groups. Par-
ticularly in some major areas of the country,
death rates in 1969-71 throughout the age range
were consistently below those forecast for the
year 2000. Furthermore, death rates for some
causes, especially Hypertension and stroke, were
falling very rapidly in the mid-1970’s. If the
predicted rates were to be used as targets, the
population of many HSA’s would have been
found to have progressed beyond the targets
even before they began to be applied. Hence this
set of data was not used any further.

An idea that conceptually came much closer
to that of the achievable target death rate was
being developed by the Working Group on Pre-
ventable and Manageable Diseases in collabora-
tion with NCHS. This group listed the diseases in
which the occurrence of a single case of disease
or disability or a single untimely death would
justify asking, “Why did it happen?”’19:¢ In par-
ticular, the list includes the ICDA-8 Revision,

€The tables of reference 19 were revised as of
9/1/77. Reprints of the article with the revised tables
can be obtained from Dr. Rutstein at the Countway
Library, Harvard Medical School, 10 Shattuck St.,
Boston, Mass. 02115.

code number, and title of all conditions, cases of
which could have been prevented or managed “if
everything had gone well.”

The idea of using an event of this sort as a
warning signal is not new, as the Working Group
points out, and Rutstein has strongly advocated
the use of such “sentinel events’ as the basis of
a guidance system for national health care.20
The Working Group also foresees the use of the
list as “a tool to measure the baseline state of
health and of comparative health status
measurements in health service area (HSA)
populations.”21

Beginning efforts to build the counts of sen-
tinel events into a quantitative index of health
care quality are now being seen.22 That an index
of health care quality can also serve as an index
of health status, or vice versa, is made possible
by the nature of this particular method of meas-
uring quality, in which “quality” is defined as
“the effect of care on the health of the individ-
ual and the population.”!?

Despite the apparent congruence of the
ideas, the list of conditions from which deaths
are deemed to be unnecessary was rejected as a
basis for constructing achievable target death
rates. The reasons for this were principally the
three following:

1. Death from a particular disease condi-
tion was labeled “unnecessary” by the
Working Group if the condition was
preventable or manageable. In some in-
stances all cases of the disease were
covered; in other instances only deaths
occurring under a certain age or cases re-
sulting from a particular risk factor were
considered preventable or manageable.
To determine from these latter qualifica-
tions what proportion of all deaths
ascribed to that disease could have been
prevented requires a great deal of data
that are not available. For example, if
death from thyroid carcinoma is prevent-
able when the carcinoma resulted from
radiation exposure, what proportion of
all deaths from this cause are unneces-
sary?

2. Many numerically important causes of
death were not listed at all. For ex-
ample, nothing is said about Ischemic
heart disease, nor about Malignant neo-
plasms of the breast, nor about Automo-
bile accident injuries. Although these
may not be suitable sentinel events, one
certainly cannot say that none of these
deaths are unnecessary. The result of the
omission of these major causes of death
from the list is that the fraction of all
deaths considered to be unnecessary is
quite small. (Different workers have
reached different conclusions about this
proportion, depending upon what as-
sumptions are made about those items
qualified by risk factors. The estimates
range from 3 percent, calculated as a
part of this research, to 14 percent.23)

3. Although age is introduced as a qualify-
ing factor for a number of disease condi-
tions, it might well have been a consid-
eration in many others had age been
routinely considered as a variable.
Deaths from the cerebrovascular diseases
offer an appropriate example. Roughly
10 percent of all deaths were classified in
this group of diseases in the United
States in 1977, but nearly two-thirds of
these occurred at age 75 years and over.
Although only rarely could death from
stroke at an age beyond 75 have been
prevented “if everything had gone well,”
the same cannot be said of the 6 percent
of this important group that occurred at
ages under 55 years. For a quantitative
measure of unnecessary deaths to be-
come credible and useful, age at death
must be routinely introduced.

Use of Geographic Variation

In 1967, Guralnick and Jackson showed how
geographic variation in mortality by cause of
death could be used for establishing an index of
unnecessary deaths.24 They pointed out that the
idea had originated with Dr. William Farr, as did
so many commonsense ideas. Beginning in 1839,
Dr. Farr was superintendent of the Statistical
Department of the Office of the Registrar
General of England, and he made contributions
to the use of mortality statistics for 40 years. He

used mortality in the districts of England in
which sanitary conditions were least unfavorable
as a standard against which to measure the
health of residents of other areas. =~
Guralnick and Jackson applied the same
principle to compute “excess” deaths by cause
of death and age in two illustrative States. They
ranked the State cause-specific death rates in
each age group from 1 to 75 years and averaged
the lowest five rates. The resulting set of death
rates by age and cause was used to estimate “‘ex-
pected deaths” in the two test States. The dif-
ference between observed and expected deaths
was taken to be the measure of unnecessary
deaths, and the proportion of all deaths that
these constituted was the unnecessary death

‘index or UDI.

The advantages of this method of establish-
ing a standard are that it is completely reproduc-
ible, it requires no individual judgments except |
as to the details of the method, and it involves
only one easily understood assumption: Mortal-
ity achieved somewhere in a particular age group
and cause-of-death group can also be achieved in
other areas. THESE ly AEE] ;

The assumption, although simple, is not
necessarily true. The target death rates are
intended to be achieved through the successful
application of all present-day knowledge of pre-
vention and treatment. On the one hand, sup-
pose that the lowest death rate from diabetes
among middle-aged males is found to be in area
A. This is almost certainly not the lowest that
could be achieved in area A by application of
the best prevention and treatment. Thus the fig-
ure may represent an overestimate of the target
as it is conceived.

On the other hand, there may be factors,
such as the genetic composition of the popula-
tion, that are partly responsible for the low rate
in area A but which area B cannot possibly con-
trol by any application of prevention or treat-
ment. This has the result of providing a target
that is too low in relation to the intended one.25

Nevertheless, the advantages cited, particu-
larly that of being completely reproducible, are
believed to outweigh the disadvantages. A stand-
ard determined in this way can at least be used
in experimentation and ‘‘market testing” of
indexes.
Methodology and Results

The variability of mortality from one geo-
graphic area to another was investigated using a
tabulation of death rates by cause and age for
white males and white females in the nine geo-
graphic divisions of the United States in the
period 1969-71. The geographic divisions are the
standard ones used for tabulations by the U.S.
Bureau of the Census and throughout the Fed-
eral statistical system. Mortality of white per-
sons, rather than total mortality, was taken as
the basis for the standard because it is generally
lower than total mortality and represents an
achievable target for minority group mortality as
well. The cause-of-death groups were those al-

ready presented, except that the five subcate-
gories of malignant neoplasms were not sepa-
rately analyzed by geographic division. The
target rates for all malignant neoplasms in each
age-sex group were split into the site subcate-
gories in the same proportions as they are in the
United States as a whole. This was a temporary
device adopted because suitable data for analyz-
ing each site separately were unavailable.

The first step in analyzing a cause-sex group
was to rank the death rates in the white popula-
tion for the 9 geographic divisions for each of
the 11 age groups. The second step was to apply
an arithmetic adjustment for trend between the
period 1969-71 and 1976 to the lowest of the
death rates. This adjustment was made only for

Table A. Estimating achievable target death rates for males aged 1-85 years and over for bronchitis, emphysema, and asthma:
United States, 1969-71 and 1976, and geographic divisions, 1969-71

 

 

 

 

 

 

 

 

 

 

 

 

 

85
A Under 1-4 5-14 | 15-24 | 25-34 | 35-44 | 4554 | 55-64 | 65-74 | 75-84 | years
rea and year
1year| years| years| years | years | years | years | years | years | years and
over
United States:
1080-77 sicrsrsnissrrssssssrasnssssnsrnsine 3.9 0.7 0.2 0.3 0.5 24 12.7 60.3 | 172.6 | 278.7 282.0
N78 sos siimstsinisscrssrsnsistonaninansss 25 0.5 0.1 0.2 0.4 11 7.3 346 | 113.2 | 215.2 245.8
New England:
108971 crrcsrsrmnesnrssssssersrorsnnnnssss 3.9 0.7 0.2 0.3 0.6 2.2 10.7 52.0 | 160.6 | 266.8 302.2
Middle Atlantic:
1968-7 o..covvssenmsssnmrasisissnisssnnmssns 3.6 0.9 0.3 0.3 0.5 2.1 9.4 455 | 137.7 | 229.2 249.8
East North Central:
ROBD-7 ceri v0 srrrssrrseitersisinnssinsnss 3.4 0.7 0.2 0.4 0.5 25 12.0 61.4 | 177.6 | 292.7 289.2
West North Central:
1989-71 ..covocicrsrrirsssarssssinssssinsanes 3.1 0.8 0.2 0.4 0.4 24 12.8 57.6 | 170.0 | 265.8 260.1
South Atlantic:
VOBOTZY ..rcvivnissssinrresssiiinssrsisianss 3.7 0.4 0.3 0.3 0.5 2.6 15.7 724 | 1749 | 268.8 260.0
East South Central:
VOODETY .oveaivessinsanisnssensssssnaimanmss 4.2 0.9 0.2 0.4 0.5 25 16.2 70.9 | 181.4 | 252.7 257.8
West South Central:
NO69-77 i: eivsrseinsvivmssinenssrrssscesass 5.0 1.0 0.2 0.4 0.6 2.2 14.2 58.8 | 175.4 | 295.7 281.4
Mountain:
1080-71 c..cvmnsivnmsivissmnnnisanssnsnsssss 3.1 0.1 0.2 0.7 0.8 2.8 19.7 97.5 | 267.7 | 401.7 338.7
Pacific:
YOBG:ZY .ocivrensrssssnmnssnsrsnsivssismanin 4.8 0.6 0.2 0.3 0.5 2.6 121 60.8 | 191.56| 324.8 339.8
Preliminary achievable target death rates!
United States:
1969-71 .ooceeeerrrrecreneraessssssennns 20) oa}. 01) 02] o3l 110] 53). 261] s03| 1770] 277
Smoothed values?
United States:
ABBY ......oo stiri rssimisserissimsiiansd 17] 0.1] 0.1] 0.1] 0.3 | 0.8 | 55 | 27.0 | 89.4 | 174.3 |

 

 

10btained by multiplying the lowest rate in 1969-71 by U.S. rate for 1976 and dividing by U.S. rate for 1969-71.
For smoothed values read ratios from curve in figure 1 and multiply by U.S. rate for 1976.
rates that had declined during this interval. If
the rate had increased, no adjustment was made.
The reasoning was that if the lower rate had
been reached at the earlier period, it should be
possible to reach it again. Thus, if the national
death rate among white people for a particular
age-sex-cause group had declined 10 percent in
that period, the lowest rate among the geo-
graphic divisions was assumed to have declined
10 percent also, but'if it had increased, the low-
est geographic rate for 1969-71 was used with-
out change. This resulted in a set of preliminary
target rates for the 11 age groups in the cause-
sex group.

The third step was the introduction of a
smoothing process to remove some of the effects
of random variation in the adjusted lowest death
rates. It was carried out by plotting the ratio of
the preliminary target rate to the 1976 national
death rate for each of the 11 age groups for
each cause-sex category. The smoothing was per-
formed on ratios of the preliminary target rates
to the most current national death rates then
available on the grounds that such ratios should
have a more stable relationship to age at death
than the achievable target death rates them-
selves. It was hypothesized that, unless the data
strongly indicated otherwise, the ratios should

 

 

 

 

AGE

 

1.00 — a a
0.90 0
w
=
=
E os80 2
5
o
g
é 070
=
2
<
«© 0.60
x
=
=
o
& 0.50
2
=
w
-
2
GC 0.40
¥
oO
<
>
<
z 0.30
=
-
w
«
a
w
o 0.20
o
E
<
j .
“0.10
op 1 1 L 1 1 1 1 1 1 ad
1-4 5-14 15-24 25-34 35-44 45-54 55-64 65-74 75-84 86 years
Le years years years years years years years years years and over
year

 

 

Figure 1. Illustration of smoothing process for achievable target death rates: bronchitis, emphysema, and asthma

10
increase with age because for most causes of
death the proportion preventable should decline
with age. Through the 11 plotted points a
smooth curve was run using a flexible ruler. The
procedure followed was to allow no more than
two inflection points in the curve. A rule like
this is needed in fitting what is essentially a free-
hand curve to prevent the curve from passing
through, or close to, all data points. The ratios

to two significant figures were read off the plot-
ted curve, and the product of these and the cor-
responding national rates were the final target
rates.

In performing these steps, the age group 85
years and over was omitted. This resulted from
an arbitrary decision that no deaths at age 85
years or over would be considered preventable.
Hence, the target achievable death rate for that

Table B. Achievable target death rates by sex, cause of death, and age

[Rates per 100,000 population]

 

 

 

 

  
 

 

 

 

 

 

 

 

 

  
    
 

 

 

 

 
 

Sex, cause-of-death group, and ICDA-8 code’ Under 1-4 514 | 15-24 | 25-34 | 35-44 | 45-54 55-64 65-74 75-84
1year | years| years | years | years | years | years years years years
Male
All cause! 1,466 | 55.3| 32.2 | 1226 122.1 | 214.8 | 610.0 | 1,574.1 | 3,565.3 8,038.6
All malignant NEOPIASMS ......cccceerureeirneeisreenrieesssaessinsesseessanens 1.8 4.7 5.6 7.3 125 37.7 | 149.2 4171 876.0 1,481.1
Malignant neoplasms of digestive organs and
PEritONBUM ......ccceuerrerrerrssrisssnessssnssnnsns 150-159 0.1 0.1 0.1 0.2 1.4 6.8 31.8 94.5 217.3 392.1
Malignant neoplasms of respiratory system.. 0.1 0.1 0.1 0.2 1.4 13.5 70.8 172.5 299.9 353.7
Malignant neoplasms of breast .............. 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
Malignant neoplasms of genital organs.. vo 0.1 0.1 0.1 1.0 1.2 1.0 23 17.2 92.6 273.3
All other malignant neoplasms....... Remainder of 140-209 1.5 4.4 5.2 5.8 8.5 16.4 44.3 132.9 266.2 462.0
Diab mellitus 250 0.0 0.0 0.0 0.2 1.3 3.0 6.6 19.3 48.3 13.3
Diseases of the heart 390-398, 402, 404, 410-429 8.7 0.9 0.5 21 8.1 61.6 | 258.5 705.2 | 1,579.0 3,687.9
Hypertension and stroke. <.rr....400-401, 403, 430-438 25 04 0.4 0.9 25 7.0 20.9 73.7 277.8 994.0
Diseases of the arteries, arterioles, and capillaries ... “A445 0.3 0.0 0.1 0.2 0.5 1.5 6.8 31.3 110.2 342.4
Acute bronchitis, influenza, and pneumonia............cccuus :
470-474, ties: 46.8 26 0.8 1.2 1.8 4.2 9.8 26.3 81.1 356.9
Bronchitis, emphysema, and asthma ..........cccceeurnnnnas 490-493 1.7 0.1 0.1 0.1 0.3 0.8 5.5 27.0 89.4 174.3
Major digestive diseases, except cirrhosis of the
IVEY icresersserssires 531-533, 540-543, 550-553, 560, 574-575 12.7 0.3 0.2 0.2 0.4 1.6 4.3 11.5 31.7 773
Cirrhosis Of the HIVEF ...umunusnnnimmmsinmsmnmssnsinins 571 0.1 0.0 0.0 0.1 1.7 10.0 25.1 38.6 40.0 20.8
Congenital anomalies and diseases of early
infancy 948.8 13 1.7 1.3 12 1.0 1.5 21 20 21
All other diseases. 1909 | 144 5.0 10.4 12.9 23.8 58.8 142.9 321.7 674.6
Accidental injuries and other trauma ............cceeuuee E800-E999 323 | 241 123 98.6 78.9 62.6 63.0 79.1 108.1 206.5
Female
All 959.7 | 42.4) 19.1 41.0 65.0 | 122.3 | 327.8 782.7 | 1,786.8 5117.3
All malignant NEOPIASMS ..........ccvvveerrisivirsnsssnseerssnsssnsssnssssnens 21 4.2 3.7 4.7 12.7 456 | 142.3 302.4 490.3 814.6
Malignant neoplasms of digestive organs and
peritoneum 0.1 0.1 0.1 0.1 1.1 5.9 20.2 568.1 133.6 281.1
Malignant neopla: atory system.. 160-163 0.1 0.0 0.0 0.1 0.6 6.0 23.7 45.6 53.0 62.0
Malignant neoplasms of breast...........ccovvereverssenneninnnn 174 0.0 0.0 0.0 0.1 3.1 11.2 329 56.1 76.9 103.8
Malignant neoplasms of genital organs.. 180-187 0.1 0.1 0.1 0.4 1.6 6.2 17.4 35.3 55.8 80.7
All other malignant neoplasms....... Remainder of 140-209 1.8 4.0 3.5 34 6.2 16.3 48.0 107.3 171.0 287.1
Diab mellitus 250 0.0 0.0 0.0 0.2 1.0 21 5.8 17.3 46.3 118.6
Diseases of the heart................... 390-398, 402, 404, 410-429 6.0 0.7 0.5 1.1 3.0 14.7 62.1 222.9 683.9 2,273.1
Hypertension and stroke.... 400-401, 403, 430-438 1.9 0.3 0.3 0.7 2.6 8.2 21.3 57.6 203.6 889.7
Diseases of the arteries, arterioles, and capillaries ...... 440-448 0.5 0.0 0.0 0.1 0.3 1.0 34 11.1 42.7 2121
Acute bronchitis, influenza, and pneumonia..........c..cceunus 466
470-474, 480-486 39.2 2.3 0.8 1.0 1.4 27 6.2 129 37.4 194.9
Bronchitis, emphysema, and asthma...........ccecuernnenns 490-493 1.1 0.1 0.1 0.1 0.3 1.0 3.6 10.5 20.5 33.2
Major digestive diseases, except cirrhosis of the :
VBE srcemnercivsnised 531-533, 540-543, 550-553, 560, 574-575 9.4 0.2 0.1 0.1 0.3 0.9 25 6.5 16.8 54.0
CILTNOSIS OF TE HIVBE 10 iracsscrssmssrsensmrirssmmrssmsissrimisrsomad 571 0.1 0.0 0.0 0.0 0.7 45 11.8 16.9 15.7 13.1
Congenital anomalies and diseases of early
infancy 470-759, 760-778 744.7 8.0 1.7 1.0 0.9 1.0 1.4 1.6 1.6 1.8
All other diseases Remainder of 000-799 126.6 9.8 3.9 6.9 11.4 19.9 41.4 91.4 181.6 397.0
Accidental injuries and other trauma...........cccoueu. E800-E999 28.1 16.7 7.9 25.1 20.4 20.7 26.0 31.6 46.4 115.2

 

 

 

 

 

 

 

 

 

 

 

 

1From the National Center for Health Statistics: Eighth Revision International Classification of Diseases, Adapted for Use in the United States. PHS Pub.
No. 1693. Public Health Service. Washington. U.S. Government Printing Office, 1967.

11
age group is whatever death rate is actually
experienced. Table A shows a typical worksheet
for the computations, and figure 1 shows the
corresponding curve for smoothing. In table’ B

the final achievable target death rates are dis-
played, and in table C these are shown again in
the form of ratios to the U.S. national death
rates in 1976. In table C the denominators are

Table C. Ratio of achievable target death rates to U.S. death rates in 1976, by sex, cause of death, and age

 

 

 

 

 

 

  

Sex, cause of death group, Under 1-4 5-14 | 15-24 | 25-34 | 3544 | 45-54 | 55-64 | 6574 | 75-84
and ICDA-8 code! 1year | years| years | years years years years years years years
Male
AY CAUSES i..o..rrosssssssrsstsrossssinprrsssns 0.71 0.71 0.76 0.73 0.64 0.65 0.73 0.79 0.82 0.85
All malignant neoplasms ................. 140-209 053 | 0.84| 0.97 0.91 0.89 0.81 0.79 0.80 0.83 0.83
Diabetes mellitus.............. 0.0 0.0 0.0 0.50 0.65 0.65 0.66 0.68 0.70 0.76
Diseases of the heart
404, 410-429 0.33 | 0.43 0.56 0.62 0.68 0.77 0.81 0.84 0.85 0.87
Hypertension and stroke........ 400-401, 403,
430-438 054 | 0.50| 0.67 0.64 0.68 0.58 0.59 0.71 0.80 0.88
Diseases of the arteries, arterioles, and
CADINARIBS .iveesrsecniirinisernsrssssassnansnnd 440-448 0.38 0.0 1.00 0.50 0.56 0.65 0.77 0.86 0.92 0.96
Acute bronchitis, influenza, and
pneumonia............ 466, 470-474, 480-486 063| 0.58| 0.80 0.71 0.62 0.62 0.65 0.71 0.79 0.88
Bronchitis, emphysema, and
SEIN sr vsinicsenirarssnnrmonsasenisnsssinecrd 490-493 | 0.61 0.20| 0.50 0.33 0.60 0.62 0.73 0.79 0.83 0.86
Major digestive diseases, except cirrhosis
of the liver .........ccceuuuee. 531-533, 540-543,
550-553, 560, 574-575 064 | 0.60| 0.67 0.67 0.57 0.70 0.77 0.84 0.86 0.90
Cirrhosis of the liver.........ccccecvceeeniiinnnns 571 0.07 0.0 0.0 0.33 0.34 0.44 0.52 0.57 0.62 0.67
Congenital anomalies and diseases of
early infancy ........cceeeune 740-759, 760-778 0.79 | 0.88| 0.89 0.81 0.80 0.71 0.79 0.81 0.77 0.68
All other diseases ..... Remainder of 000-799 0.51 0.79 0.81 0.70 0.58 0.56 0.66 0.72 0.72 0.69
Accidental injuries and other
APOUINIA srvsasinermsasnnsensasssinnnrsancs E800-E999 059 | 0.65| 0.69 0.73 0.63 0.57 0.59 0.68 0.80 0.88
Female
AN CBUTES so iraiursnisrnnssiinsssin sid 068| 0.69] 0.72 0.70 0.67 0.68 0.73 0.78 0.81 0.85
All malignant neoplasms ................. 140-209 068| 0.84] 0.90 0.92 0.85 0.81 0.81 0.83 0.85 0.88
Diabetes IMBHIItUS...ccrisssssrinsssssssiessssionssss 250 0.0 0.0 0.0 0.50 0.63 0.64 0.60 0.61 0.65 0.73
Diseases of the heart. .............. 390-398, 402,
404, 410-429 0.30| 0.44| 0.50 0.55 0.59 0.64 0.69 0.76 0.80 0.83
Hypertension and stroke........ 400-401, 403,
430-438 044 | 043| 0.60 0.64 0.74 0.58 0.69 0.75 0.83 0.90
Diseases of the arteries, arterioles, and
COPINAIIOS viouarririssssrvissmsasnnnissssineind 440-448 0.50 0.0 0.0 0.50 0.60 0.63 0.74 0.82 0.87 0.94
Acute bronchitis, influenza, and
pneumonia............ 466, 470-474, 480-486 064| 0.56| 0.80 0.83 0.64 0.63 0.70 0.74 0.80 0.88
Bronchitis, emphysema, and
ASIN orvvissnssvansssssisinsissesvansssrnnines 490-493 0.61 0.25| 0.50 0.50 0.60 0.71 0.77 0.80 0.83 0.85
Major digestive diseases, except cirrhosis
Of the lIVer ......ceccenrnrain 531-533, 540-543,
550-5563, 560, 574-575 0.68| 0.50 0.50 0.33 0.50 0.75 0.71 0.83 0.85 0.91
Cirrhosis of the HVer........cccirsiescesrsssessss 571 0.13 0.0 0.0 0.0 0.29 0.40 0.52 0.57 0.61 0.65
Congenital anomalies and diseases of
early infancy .........ce...... 740-759, 760-778 0.76 | 0.84 0.81 0.83 0.75 0.77 0.82 0.80| 0.70 0.67
All other diseases ..... Remainder of 000-799 0.44 0.74 0.78 0.71 0.67 0.65 0.74 0.80 0.80 0.78
Accidental injuries and other
RERUN i a iaesinionssnssistanasat E800-E999 0.61 064| 0.65 0.68 0.62 0.60 0.66 0.74 0.86 0.96

 

 

 

 

 

 

 

 

 

 

 

l1From the National Center for Health Statistics: Eighth Revision International Classification of Diseases, Adapted for Use in the
United States. PHS Pub. No. 1693. Public Health Service. Washington. U.S. Government Printing Office, 1967.

12
the rates for the total population, not the white
population alone, because the purpose is to
show what proportion of the total mortality is
considered nonpreventable if one accepts these
achievable target death rates as standards.

By and large, the ratios increase with age, as
one might expect, since the preventability of a
death from most causes drops as one grows
older. However, using geographic variability as a
gauge of preventability produces some odd and
unanticipated results in terms of age. It can be
seen in table C, for example, that for all causes
of death the lowest ratios for males (and hence,
the highest proportion of deaths presumed pre-
ventable) occur not in childhood but at the
young adult ages. The geographic variability of
trauma death rates, which predominate at these
ages among the males, may account for this
result.

There is also the curious result that the low-
est ratios from ages 25 to 85 are seen for cirrho-
sis of the liver (table C). This rubric, included
specifically as an indicator of alcoholism preva-
lence, showed great geographic variability. While
some of this may be a genuine sign that the
disease can be prevented, it may also reflect a
varying failure in different regions to include
mention of this disease on the death certificate.
The completeness of reporting cirrhosis of the
liver as the underlying cause of death should be
studied. If it is still as poorly reported as it was
in Westchester County, New York, in 1934,26
steps should be taken to remedy the situation.

ALTERNATIVE FORMS OF
THE INDEX

Desiderata

In using mortality statistics to reflect the
health of a jurisdiction, such as a health service
area (HSA), it is a practical necessity to have
some form of summary index. As pointed out
earlier, the display of death rates by age, sex,
and cause of death for the jurisdiction’s popula-
tion and comparison of this set with a set of
achievable target death rates is too complex for
ready comprehension. Furthermore, a summari-
zation is needed in order to reduce effects of
random variation.

There is an infinite variety of ways in which
such a set of data can be summarized, but the
one selected for use must have the following
three properties, which limit the choices very
substantially:

1. It must be sufficiently easy to compute
so that the potential user will not be dis-
couraged by the time and expense of
computing it.

2. It must be readily comprehended and
must make sense in the context of its
use.

3. It must be as ‘“‘sensitive” as possible,
consistent with the other properties.

In this report, “sensitivity” is the degree to
which the true, underlying variability in the
phenomenon being measured shows up through
the “noise” of random variation. The mortality
in a particular population during a specified in-
terval of time is a finite sample of deaths pro-
duced at that time and place by the underlying
forces of mortality characteristic of that popula-
tion at that time. Every community differs from
every other in this characteristic set of underly-
ing mortality forces, and within a community
every time period differs from every other one.
Of course, the extent of the differences from
place to place or time to time may be large or
very small. The only way of measuring these dif-
ferences is to classify and count the particular
deaths that express the force of mortality. These
deaths are enumerated and classified as com-
pletely and accurately as possible, but the partic-
ular deaths, though counted completely, can be
treated as a random sample because of the very
large number of influences determining which
persons actually died. It is a sample generated by
a set of underlying death rates that represent the
mortality “universe” for that time and place.

Note that the concept of sensitivity as the
relationship between underlying variability and
random variation has nothing to do with the de-
gree to which mortality successfully measures
what is thought of as the health of the commu-

nity. The relevance of mortality to that purpose

is assumed, but with certain recognized limita-
tions. It is assumed that some of the deaths from
certain causes at certain ages were preventable

13
and, therefore, that estimates of these numbers
constitute a measure of correctable failures in
the health care system.

Forms of Index Examined and
Their Sampling Variability

Four different forms of summary index were
examined and examples were calculated. Each
index‘ was intended to summarize the mortality
of a local population over the whole of the age
range for males or females for a cause-of-death
group. Thus mortality for the population resid-
ing in the area (in this case, an HSA) during a
particular period of time (in the examples used,
1969-71) was to be expressed in the form of 24
numbers (12 cause-of-death groups for males
and 12 for females) plus “all causes” for males
and females.

The notation used in setting down the forms
of index is as follows:4

A = Subscript for 3-way age classification:
under 35 years, 35-64 years, and 65
years and over (coded I, II, III).

a = Subscript for 11l-way age classifica-
tion: under 1 year, 1-4 years, 5-14
years, ..., 75-84 years, and 85 years
and over (coded 1, 2, 3,..., 10,11).

d4 = Number of deaths in the local area

(HSA) in the 3-year period 1969-71 in
age group 4.

d, = Number of deaths in the local area
(HSA) in the 3-year period 1969-71 in
age group a.

pa = Population of the local area (HSA) in
1970 in age group 4.

Population of the local area (3) in
1970 in age group a.

I

Pa

Py ='1970 U.S. population in age group 4,
P, = 1970 U.S. population in age group a.

dto simplify the notation, subscripts are omitted
for the HSA, the sex, and, on the right side of the equa-
tion, for the cause-of-death group as well. Hence, in the
algebraic expression for the index R;, population data
are assumed to be for a given sex, and death rates or
numbers of deaths are assumed to be for the same sex
and for the cause-of-death group i.

14

m4 = Death rate per 100,000 population in
the local area (HSA) on an annual
basis in age group 4 in 1969-71.

m, = Death rate per 100,000 population in
the local area (HSA) on an annual
basis in age group a in 1969-71.

M, = Achievable target death rate per
100,000 population in age group 4.

M, = Achievable target death rate per
100,000 population in age group a.
The first form of index examined, called R;,

is defined as:

A

 

P
R; = 2 (1-14)
>,
in which
My, My (105),
Tr I me tn
4 my dy/3
dy = 3(1073)M ps
l-ry sm —————

dy

The values of M, are derived from the
values of M, very simply:

Mp, + Mypy + Mgps + Mypy + Mypy

 

 

My = b
1
: Mgpg + Map, + Mgpg.
MS
Loren :
- Mgpg + Mypyo +My 1p1
My = b TH
III
but
My1p41 =dyy

because the achievable target death rate at 85
years and over is considered to be identical to
the actual death rate in that local area.
Note that M4 is almost always less than m4,
so the ratio 74 is almost always less than 1 and,
of course, is positive. Hence 1-74 is almost
always positive and less than 1. R can never
exceed 1 but may optimally or by chance be
negative. A condition that always must be met is
dy #0. If dg should happen to be zero, as oc-
curred in age group I for some causes of death
in some of the HSA’s used as examples, the age
group must be collapsed with a neighboring one
and the index computed as a special case.

For additional brevity, write:

Py
Wy = ———
3.2
Hence
A
R;= Yuw,(l-74)
and
(8/105 )M4p4
ry = —
dy

Under certain reasonable assumptions the
variance of this weighted sum can be approxi-
mated as:

var(R;) = x4 wj var(1- ra)}

5 w? (3/10%)2M% p% var(1/d,)}-

The variable d4 is generated by a process
that is essentially Poissonian. Hence the vari-

© ©Actually, the conditions for a Pojssongan variability
. .are only closely met in’ the: generation of dy values, that
. is, deaths in thé age group under 35 years of age. The

variance of the index R; is. dominated, however, by ‘the

variance in the youngest age group in all the forms of R;
investigated here. For a single cause-of-death group and
sex, the number of deaths under 35 years of age in a
3-year period in a typical HSA is often quite small. For
example, in one California area, with a population in
1970 of about 1 million, the value of dj for males is
less than 100 in 9 of the 12 cause groups. In 7 of the
groups it is less than 50.

For a discussion of the Poissonian process see Sec-
tion 3.2 in Introduction to Stochastic Processes in
Biostatistics.

ance of d4 can be taken to be d4 . The variance
of 1/d4, however, is somewhat more complex.
As dy increases,

var(l/d)—> 1/d3 .

But for values of d, less than about 200
deaths, a better approximation is needed.
Table 1 (generously calculated for the purposes
of this report by Dr. Benjamin I Tepping)
shows the results of a Taylor’s series expansion,
providing approximations of E(1/d4) and
var(1/d4 ) for values of E(d4 ) from 2 to 200.

At E(d4) = 200, the value from the table is:
1.2885 X 10-7

as compared with:

1.25 X 1077 obtained from 1/2003.

The form of R; has certain desirable fea-
tures, and it was the form initially favored for
use with the achievable target death rates. The
statistic 74 is the ratio of the deaths that would
be expected from a particular cause in age group
A in the HSA if the community was to experi-
ence the achievable target death rate to the
actual number of deaths experienced. It is, thus,
the proportion that might be considered nonpre-
ventable. Hence 1 - r4 is the proportion of the
actual deaths that were preventable, and R; is a
weighted average of these ratios giving weight to
each age group in proportion to the fraction of
the male or female population in that age group
in the entire United States.

As pointed out in an earlier investigation of
nna indexes,?® the disadvantages of using

- “standard population” to weight the age..

i are: much less serious if the quantities

being weighted. do not tend to vary greatly .
from age group to age group, as the age- specific
death rates do.

The index R; is readily understood and
makes sense. It is also easily computed, but be-
cause of the condition that zeros must not occur
in the denominators of the terms, it is necessary
to collapse the number of the age groups from
the standard 11 to 3.

As will be seen later, however, R; is heavily
penalized by low sensitivity.

15
In the second form of index, called R;, the
statistic 74 is inverted and called 7 . In other
words,

ped dy

ry =—=

4 (3/10%)M4p4

and

4
bo ’
Ri=Y wary .

In this form, zero values of d4 give no
trouble. As noted in table B, certain values of
M, are zero because the deaths from that cause
in that age group are considered wholly prevent-
able. However, for the three-group age classifica-
tion, no zero values of M4 are encountered.

Most of the ry values are greater than unity
and, of course, all are positive. Hence a value of
R} <1 is optimal or a chance event.

Using the same reasonable assumptions as
for R; and again postulating the Poissonian be-
havior of d 4, we have the following approximate
expression for the variance of Rj:

A w?
phe | GNP

Here, the ratios of which a weighted sum
forms R} are a measure of the relative excess of
the actual deaths over the nonpreventable
deaths. This is somewhat analogous to the
“weighted average of relatives’ used in price
indexes, the weights being the average propor-
tions of the population experiencing the excess
instead of the amounts of that commodity pur-
chased. Also, the form of R} seems simple to
understand and is very little trouble to compute.
Its sensitivity will be considered in the follow-
ing section.

Both R; and R; have the characteristic that
the influence an age group’s mortality experi-
ence has on the index is roughly proportional
to the numbers of persons living in that age
group (in the United States, not in the HSA).
Conceptually, that seems to be a distinct ad-
vantage over other mortality indexes now in
common use. In the standardized mortality

16

ratio (SMR) and the comparative mortality fig-
ure (CMF)!1 the age group’s mortality experi-
ence has an influence approximately propor-
tional to the number of deaths occurring at
those ages, thus tending to emphasize the older
ages—ages at which deaths tend to be much less
preventable.

This is an important reason for the alleged
lack of sensitivity of mortality measures. It is
difficult to bring about improvement of the
index by any means within the power of the
community to apply.

On the other hand, the ability of the pre-
ventable mortality of a local population to show
through the “noise” of random variability may
not be as great in these weighted sums of ratios.
This is the central problem that must be ad-
dressed in the construction of a maximally use-
ful index.

The third form of index investigated, called
R}',is not a weighted sum of ratios but a single
ratio in which preventable deaths are summed
over all age groups and divided by the sum of
actual deaths in the local population for all ages
combined. It is expressed as

A
2 da - (3/10°)Mapa}
= = _—
2.d4
A
(8/10°) M4 pa
A
2 da
For this index the approximate variance is
4 2
(3/10°)2 (Mapa)
A 3 PATE
(£da)
Although the achievable target death rates

are relatively closer to actual death rates at the
older ages,

n
:

var(R{')=

dy - (3/10%)M 4p,

is considerably larger, per year of age, in the
older age groups (except that by definition it is
zero at ages 85 years and over) than it is at ages
under 35 years. Hence, this index suffers from
the same disadvantage as the SMR and the CMF.
It also has one of the basic disadvantages that
is found in the crude death rate: It is not inde-
pendent of the age distribution of the local area
population. HSA’s with larger proportions of
older people would therefore tend to have
higher values of R}' for that reason alone. Hence
R}' was soon dropped from further considera-
tion as a summary index of mortality.

The final version examined was a variant of
the ratio R;’, which, though not independent of
the age distribution, gives such heavy weight to
the younger ages in comparison to the older ages
that only an HSA with a very exceptional age
distribution would, by this fact alone, show up
with an advantage or disadvantage. This index,
called R;" here, is based on the same principle
as the years of life lost index of mortality that
has been described by a number of authors. 29-31
It has been used for comparing the seriousness
of causes of death®! and, for all causes com-
bined, as a measure of the health of counties.30
There have been a number of variations in the
way this years of life lost index has been com-
puted, but it appears that the simplest way gives
results that differ in no significant respect from
results attained by using more sophisticated
methods. The number of deaths in a given age
group is multiplied by the difference between
the midpoint of that age group and a fixed age
that is chosen to represent the end of “produc-
tive” life (meaning, it appears, the point at
which most individuals cease to contribute to
the economy). This product is summed over all
age groups for the population whose health is to
be measured, and the result is divided by a cor-
responding figure that is calculated using deaths
that would be expected if the death rates had
been those of the Nation, or some other stand-
ard. The choice of the upper age has differed
among different authors. In this report it is
taken as 75 years.

For deaths under 1 year of age, therefore,
the number of deaths is multiplied by 74.5 to
approximate the years of productive life lost by
persons dying at that age. In each succeeding age
group of the 11 standard groups the multiplier
is smaller, until, in the age group 75-84 years, it
vanishes. Hence, only deaths at under 75 years

of age contribute to the index. Used with the
achievable target death rates as the standard, the
equation is

a
dala
Blew

22(3/10°Mypals)

in which
2 = (75 - midpoint of age group a)

The variance is approximated by the follow-
ing expression:

1
var(R}") =

grit ; dl)? .
Eno mp = :

The structure of this index gives a death in
the age group under 1 year nearly 15 times the
weight that a death in the age group 65-74 years
receives. For most causes of death the number
of deaths in the latter group outnumber those
in infancy by equally large or much larger
factors. Consequently, the effects of differences
in population distribution tend to be diminished
by the weighting system, but the effects of
chance variation are increased. The assumption
of Poissonian variation in the number of deaths
is that the variance increases directly as the num-
ber of deaths increases, but the weights attached
to the age groups are squared to obtain the vari-
ance of the sum.

Since the standard mortality ratio (SMR)
index will be used for comparison with the
indexes presented in this section, its estimating
equation and variance using the same notation
and assumptions are shown here.

>d,

SMR; = ———5———
R 3/10% 22 (Mapa)
3d,
VA(SMR,) = free engi +
pi {310° Ep}?

17
The customary standard death rates in the
SMR are the national death rates. Hence in these
equations M, refers to the 1969-71 U.S. death
rate for cause 7 in age group a, instead of to the
corresponding achievable target death rate.

COMPUTING AND TESTING
THE INDEXES

For convenient reference in the discussion
that follows, the various forms of index exam-
ined can be described as follows:

R; = Weighted average of proportions of
actual deaths in each age group con-
sidered preventable, using U.S. pop-
ulations as weights.

R; = Weighted average of ratios of actual
deaths in each age group to “ex-
pected” deaths, that is, deaths con-
sidered nonpreventable, using U.S.

populations as weights.

R}' = Ratio of sum over all ages of deaths
considered preventable in each age
group to total actual deaths.

R}"' = Ratio of actual years of life lost to
expected years of life lost if only
nonpreventable deaths occurred in
each age group.

SMR; = Standardized mortality ratio, that is,
ratio of actual deaths at all ages to
expected deaths at all ages if U.S.
death rates were experienced in each
age group.

In each instance, of course, the index is for
cause-of-death 7 and for one or the other of the
sex groups.

Health Service Areas Selected
for Testing and Computation
Methods

In order to illustrate and test the various
forms of index, a group of 19 HSA’s was

18

selected. This sample was chosen with certain
objectives in mind. First and foremost was hold-
ing down the expense of aggregating county
mortality data for the 3-year period, 1969-71,
by cause of death, age, and sex into HSA totals.
To accomplish this, a sample of 10 States was
chosen to provide geographic variability. HSA’s
within the States were selected to illustrate
highly urban populations and areas with higher
proportions of rural and farm populations. Two
other considerations were that not too many
counties make up the HSA in order to save labor
(the numbers of counties ranged from 1 to 14
and totaled 117), and that the HSA not contain
too small a population (the 1970 populations
ranged from about a half million to 7 million).
Table D presents the HSA designation, location,
and several demographic characteristics of the
19 HSAs.

The mortality counts for 1969-71 by county
for the 10 selected States, classified in the Na-
tional Center for Health Statistics (NCHS) stand-
ard 69-cause recode, and by age, sex, and color
were supplied through the NCHS Office of Sta-
tistical Research. These counts were keyed and
summarized into the HSA totals and the broader
cause groups that were used for the achievable
target death rates. The keying was carried out by
Data Enterprises, Inc., and the programming and
computer time were donated by Westat, Inc.,
both firms in Rockville, Maryland. All index
computations were carried out by the author
using a handheld programmable calculator; the
programs, repeatedly used weights, populations,
and death rates were stored on magnetic strips
between calculating sessions. Numerical checks
on the data entry were built into the programs,
but the computations are certainly not com-
pletely error free.

For males, R;, R;, and R;"" were computed.
Tables 2-5 show the 1969-71 results for each
of the 12 cause-of-death groups, and for all
causes combined, in the 19 HSA’s. Some indexes
for males in the United States as a whole are in-
cluded for comparison (tables 2-4). For females,
only R;"" was computed. Examinations of sensi-
tivity were performed on the indexes for males
only.
LL6T-9L61—S33mS Pat) —y3 oof] :$901AISE UBWNY PU Yi eaH JO Juswisedaq oy) Wosj ele;
“SNSUSD OLGT UY) WO) Je $2InBy NIV “LLGT OOF BID A31D Pup A1uno) :snsus) oy) JO neaIng 'S’f) SY) WO TIECly

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

8't oy Ty ol 8'9 9C L's z'9¢ So o0C 8'9 088 SL's freee $81IUNOD AQieau pue sex nemiIpy
‘ZO uisuodsip
Ty €'s T's SEL 1'9 oy 69 Lee €'S Lee 60 L'8s 6SL $813UNCD AQJESU PUB UOSIPRY
110 ursuodsIp
Ly Te Ls £1 Sv ve z9 S'€Z si oz 6c ris svi'c |Tv $S1IUNOD IsamylIou O| pue spIeag
110 uoiBuiysem
ry 0's 9g £0 z9 cz 89 ez vy got viL 126 800'Z | eens. sanunod
UISYLIOU { PUR (X8853) YIEMBN
120 Assser man
8c Ty ve 9'6 £8 Ly 06 §'se £T eri [4 J zy 069 |” S31UNOD AIS-PIW p| pue Meuibes
:90 ueBiyaw
6v ry ov 90 SL 6'l S'9 L'9¢ Le €eT sol 6°06 sLL'y ee *S81IUNOD AQIBaU pur 101
110 veBiyp
TY ov 6v 60 Le £T S'8 [4°74 09 £1 Lee ze $0 peer "$81IUN0D AqJesu pus asounjeg
$0 puejAsen
gc zh ez oL ve ze vs 8 1'6p eit rst Log QLL |r s2nuN0d
Aquesu pue Ajuno) sebioe) adullg
‘60 puejAseyy
be zz £6 vo \'e re ot er ote zz I'v z68 €2G |e Ano AswoBiuoyy
120 PusjAsepy
vey op oy 90 vl cp o91 spl ze re £6 oes Pr TI aihesreverere seis iln soysiied
1320LIN0S | | PUB SUBSIO MBN
110 sueisino}
TY v's oe 1'0 zeL 8 oll po sz L'6€ vl £16 0ZE'L | reesei Ano)
80JuOW PUB ‘AlUnOD speQ ‘IweIN
: 60 epLIoj 4
6S I's v9 el gL gy g'€L 60L o'sy vil 6z1 29 188 |e s8nuN0d
ISIMYINOS 0} PuUe SIBAWN “34
‘90 epuOo} 4
re se v's vz g's vy £8 rez v8 ri pL vee BG | reesei, ag aimug
110 8semei8Q
€t Le [4] So 6's 1c gv 0’te v6 Lee 1'9 o0'0o8 ooze fet $83UNOY pue||O) PUR PJOJIBH
HO INDRIBUUOD
oz st 09 10 5g vz op 1'se 60L o'ee ve z'98 EBL | reese Aino pieyiiey
: 10 IN21I28UL0D
vy sy 9 00 z8 zt zs cz cv 06 60L 186 PT sajebuy $07)
11 ®uioed
sy St ov 0's i Lr 3 43 oot £0- 9'0Z oy S0L EVOL Often $213UNO [BAUS § PUB OUsSAL 4
60 Bus0pe)
oy €'s 68 10 L's 8c 8'9 8cl Ly Sot 98 £86 sy | $813UNOD AQIEsu pue 03sKOURI J UBS
$0 elusope)
0's £v zo 0's 08 1z Lot LL 8C 9tL or vy ey TTT $S1IUNOI UIBIOU | pus e38in3
110 elus0318)
zisquinN JuUsdieg 1equingy
SL-¥L6L go, oo "
‘SyINg eal) vi6l €L61 suos quin|
000° Jed ‘uonejndod | ‘uvoneindod | uone; | -iad asow inoyum oo Sunn) 0L-0961 uoRe| uone) | uone} oi
Aysow | 0001 sed | 00004 sed | -ndod | so 1071 syun SOAR) | <CBJTMELY | WOR. | eBdel | iq il nang > uopecc) pue
je8uosy pag si0100p | wuey | yumsiun | Buyemp | MOA | uiedioy | sib 520d | weg | uean | uoneindog Sa
sod jenuue |  jeadsoq |e1pony Bujemp | paidnoog | SeHIMEd | J09En BN | -uBieioy d
ebeseny paidnaoQ
seaJe 8JIAI8S

y3jeay paidsjes Aq ‘suoiieindwod 8AIIRIISN||I BY Ul PsN SEaJe 8DIAI8S YI[eay G| JO SOIISII910RIEYD Paldsjes JO Jequinu pue juadiad pue ‘uonejndod jo JequinN 'Q 8jqeL

19
Tests of Sensitivity

Two measures of sensitivity have been used
for comparing different forms of the mortality
index. One is the median and range of the coef-
ficient of variation (CV), which are simply the
median and range of the ratio of the estimated
standard error of the index to the index. The
other is a contrived measure consisting of the
ratio of the “modified range” among the 19
HSA'’s to the mean of the the 19 standard errors.
Modified range means the difference between
the arithmetic averages of the three highest
values and the three lowest values of the index.
This latter statistic was chosen over alternatives
requiring squared differences from the mean of
the HSA’s because it involves no assumptions
about the form of the universe of HSA’s from
which this nonprobability sample of 19 was
drawn, and also because it gives less emphasis to
extreme values.f The modified range is a meas-
ure of the variability as exhibited by these par-
ticular local areas, and the mean of the standard
errors is intended here as simply a typical esti-
mated standard error.

These parameters were calculated for the en-
tire set of HSA’s and for 9 of the 12 cause-of-
death groups as well as for all causes. The
indexes compared were R;, R;, R}", and SMR;
(the standard mortality ratio for cause-of-death
group 7). Table E shows the range and certain
rank orders, including the 19-area median, of
the coefficient of variation for R;, R;, and R;".
At this point in the analysis, R; was dropped
from further consideration. There were three
reasons for this elimination. First, the zero order
correlation coefficients between R; and R} were
consistently high, suggesting that these two
indexes were measuring the same features of
each HSA’s mortality. Note the correlation coef-

fA more direct measure of underlying geographic
variability, which does require the squaring of differ-
ences from the mean, is provided by the geographic com-
ponent of variance. This is the method used b
Kleinman et al. in analyzing infant mortality rates.
For comparison, the square root of the geographic com-
ponent of variance was also computed for the 9 causes
of death in the 19 HSA’s. The zero-order correlation
between this and the modified range was extremely
high.

20

ficients for the following cause groups for males
in the 19 areas:

All malignant neoplasms 92
Diseases of the heart 98
Hypertension and stroke 92
Bronchitis, emphysema, and asthma .82
Cirrhosis of the liver .80
Congenital anomalies and diseases

of early infancy 94
Accidental injuries and other

trauma 84
All causes 97

Second, a comparison of the coefficients of
variation of R; and R; for 9 cause groups in the
19 areas showed that in a majority of the paired
comparisons the former exceeded the latter (and
sometimes by substantial amounts). The follow-
ing are the results:

CV(R;)> CV(R;)<

 

Cause-of-death group ~~ CV(R;)  CV(R])
All malignant neoplasms 19 0
Diabetes mellitus 12 7
Diseases of the heart 15 4
Hypertension and stroke 14 5
Acute bronchitis, influ-

enza, and pneumonia 5 14
Bronchitis, emphysema,

and asthma 8 11
Cirrhosis of the liver 10 9
Congenital anomalies and

diseases of early infancy 18 1
Accidental injuries and

other trauma 15 4
Total 116 55

Furthermore, the median value of CV(R;)
was greater than the median of CV(R;) for all
the nine cause groups except the two respiratory
disease groups, and the range was greater in
every group. (See table E.)

A third reason for eliminating R; from the
comparisons was that, despite consolidation
from 11 age groups to 3 and the use of mortality
Table E. Selected rank orders and range of coefficients of variation (CV) for three forms of mortality index (R;, R;, and R;

reer

) for

males, by cause-of-death group: 19 health service areas, 1969-71

 

Cause-of-death group and coefficient of variation

5th 10th 15th

rank rank rank Range

 

All malignant neoplasms:
CV(R))......

0.148| 0.201 | 0.290 0.364

 

EVA csusamcomsmsioumuirnsssissasimmssans
CV(R;")

 

 

Diseases of the heart:
CV(R;)
CV(R;)...

Nr

Hypertension and stroke:

OVARY crosses onion a ————,

 

  

0.049| 0.066 | 0.082 0.072
0.021| 0.031 | 0.038 0.037

0.192| 0.310 0.721 1.7
0.177| 0.268 | 0.311 0.362
0.072] 0.113 | 0.124 0.163

0.069| 0.229 | 0.345 1.28
0.070| 0.105| 0.126 0.119
0.015] 0.021 | 0.025 0.025

0.101| 0.220 | 0.368 1.62
0.102] 0.150] 0.186 0.177
0.038| 0.053 | 0.069 0.077

 

 

0.027| 0.045| 0.089 0.237
0.064| 0.081 | 0.106 0.116
0.059| 0.077 | 0.104 0.107

0.107| 0.191 | 0.722 18.13
0.202| 0.295| 0.363 0.384
0.068| 0.080 | 0.107 0.125

 

Cirrhosis of the liver:
CV(R;)
CV(R;)
cv( R;")

 

Accidental injuries and other trauma:

OVARY orrevmmssiin reat

CVI(R;)

 

  

 

0.038| 0.236 | 0.750 8.57
0.130] 0.196 | 0.317 0.295
0.044| 0.060 | 0.089 0.099

0.109] 0.291 | 0.494 2.12
0.075] 0.103 | 0.137 0.119
0.033| 0.047 | 0.057 0.059

wsksEsReIR Rs SRse RRs Rs Te ane nn sR RES 0.021| 0.037 | 0.079 1.63

 

0.019| 0.022 | 0.032 0.036

 

CVIR;")

 

0.022 0.025 0.036 0.041

 

 

 

 

 

statistics for a period of 3 years combined, there
were instances when the actual number of
deaths for a cause group in the age group ‘“‘under
35 years” was zero. This happened in only one
of the 19 areas for 2 of the cause-of-death
groups, but since d4 appears in the denominator
of R;, the zero values had to be ruled out. This
was done in those two instances by computing
the ratio for all ages combined, omitting the
standard population weighting. However, in a
number of other age-cause cells there was only
a single death in the 3-year period, which led to
extreme values of the estimated variance of both

By this process of elimination there is left
only the comparison of the sensitivity among

R;, R}", and SMR;, using the second measure of
sensitivity described earlier, the ratio of the
modified range to the mean standard error. This
comparison is presented in table F.

By and large the modified range of both R;
and R;" is larger than that of the SMR; for the
same cause group. For seven of the nine cause-
of-death groups, not counting all causes com-
bined, the modified range of R; is the largest and
the spread of the SMR; index is the smallest.
However, the mean standard error is consistently
lowest for the SMR; and, almost without excep-
tion, is lower for R;" than it is for R;. Hence, if
the ratio of the former quantity to the latter is
taken as the indicator of the extent to which the
information contained in the index shows

21
Table F. Highest and lowest value, modified range (MR), mean standard error (5j) and ratio for three forms of mortality index (R;, R;",

"er

and SMR;) for males, by cause-of-death group: 19 health service areas, 1969-71

 

 

. Acute bets Congenital | ni
Highest and lowest . Hyper- ex Bronchitis, ¥ + Accidental
value, modified range All 2 Diabetes Disegsss tension omnis; emphysema, Ci titia aon es injuries
(MRY. mean standard || ‘causes || ™® nen mellitus 9 the and in shee, and ; the an } igeesas and other
error (s;) neoplasms ears stroke an 3 asthma iver 0 arly trauma
pneumonia infancy
Highest value:

2.128 1.474 3.526 2.305 3.442 3.936 6.929 6.095 2131 2.725
2.017 1.570 3.241 1.867 3.061 3.844 3.621 4.785 2.212 2.878
1.199 1.251 1.783 1.202 1.169 1.112 1.687 2.262 1.234 1.447
1.143 1.181 1.218 1.172 1.428 1.554 0.434 1.1256 1.108 1.048
1.117 1.044 1.128 1.077 1.320 1.570 0.930 1.122 1.292 1.018
0.783 0.870 0.668 0.792 0.705 0.619 0.686 0.571 0.725 0.625
0.710 0.254 1.845 0.919 1.179 2.012 4.656 3.927 0.776 1.432
0.669 0.411 1.605 0.560 0.966 1.818 2.113 2.644 0.732 1.510
0.280 0.328 0.955 0.348 0.333 0.386 0.650 1.1286 0.415 0.753
0.022 0.087 0.516 0.148 0.268 0.219 0.728 0.515 0.159 0.042
0.020 0.035 0.188 0.027 0.092 0.202 0.180 0.138 0.076 0.046
0.008 0.019 0.066 0.012 0.025 0.041 0.052 0.058 0.043 0.023

32.3 2.9 3.6 6.2 4.4 9.2 6.4 7.6 49 34.1

33.5 11.7 8.5 20.7 10.5 9.0 11.7 19.2 9.6 32.8

35.0 17.3 14.5 29.0 13.3 9.4 125 19.4 9.7 32.7

 

 

 

 

 

 

 

 

 

 

 

 

through the noise of random variation, the
standardized mortality ratio form of mortality
index is, for most cause groups, the most sensi-
tive of the three forms compared, though the
indicator is only slightly better for five of the
nine independent groups and slightly poorer
for a sixth group.

For all causes combined, the modified range
of the SMR; values in the 19 HSA’s is 35 times
the typical standard error; for the index based
on years of life lost it is 33% times the typical
standard error.

Conclusions

From these results the following conclusions
can be digwm :

Tides in a ferns of Petite] a

of ratios for individual age groups have
inherent advantages, namely, the weight-
ing of excess mortality according to the
approximate proportions of the popula-
tion experiencing the excess, the expres-
sion of excesses in a form that is inde-
pendent of the numbers of deaths at the
ages at which they occur, and an appar-

22

ent greater range of magnitude in the
values from one HSA to another as com-
pared with other commonly used in-
dexes. However, the random variation of
the weighted averages of ratios, at least
those forms included in this experiment,
is sufficiently higher than that of the
other forms that the underlying area-to-
area differences in mortality by cause of
death in a 3-year period may be
obscured.

As an alternative for use with an index in
which achievable target death rates by
cause of death, age, and sex are used as
the standard, the years of life lost form

- has nearly as high a sensitivity as the

. older single-ratio form of‘index in which’

the total ‘of actual deaths is divided by

the computed expected number; assum-
ing the standard mortality is experi-
enced. The years of life lost form has the
conceptual advantage that mortality at
the younger ages, considered much more
amenable to correction efforts, is
weighted a great deal more than is mor-
tality at advanced ages.
FUTURE RESEARCH

Improvement of the Achievable
Target Death Rates

Some areas of needed improvement in con-
structing the set of death rates used as achiev-
able targets in this report have already been
pointed out. The principal question remaining is
the soundness of the underlying assumption
that the lowest achieved mortality in some part
of the United States is a reasonable basis for
achievable target death rates as that concept has
been defined here. Can all deaths measured by
death rates as in excess of these justifiably be
labeled “preventable”? Or do these lowest
achieved rates understate what modern medical
science and an optimum health care system
could bring about? If geographic variability is to
be used as the basis for the standard, as Gural-
nick and Jackson suggested,?4* how should it be
analyzed? Some preliminary work done dlring
the course of this project suggests that using the
lowest State rate rather than the lowest geo-
graphic division rate might have produced quite
different results. It has also been suggested that
the mortality in the HSA that exhibited the low-
est death rates for a particular cause should be
used as the standard.

Of greatest value, it seems, in arriving at a
tenable target set of rates would be the further
use of the type of expert judgment that
Rutstein and the members of the Working
Group on Preventable and Manageable Diseases,
and the scientists called upon by the Public Serv-
ices Laboratory of Georgetown University,19-21
have brought to bear. However, such judgment
should be directed specifically to the question of

- the proportion of present mortality.as recorded. .
by ‘cause of death; age, and sex that could be: °

prevented given optimal circumstances.
© It is quite possible that small differences in
the values used for the target set of death rates

. would have little effect on the relative differ- -

ences in the index for a particular cause-of-death
category among the HSA’s. The sensitivity of
such comparisons to changes in the achievable
target death rates deserves further investigation.
Nevertheless, if the premises set forth in the
introduction to this report are correct, and if

mortality by cause of death can be used to meas-
ure the health of local area populations, such as
HSA’s, then the development of a widely ac-
cepted set of achievable target death rates can
prove extremely useful for measuring commu-
nity health and, hence, for health planning.

The Form of Indexes and
Their Random Variability

Variation in death rates from area to area
can be thought of as originating partly from dif-
ferences in the underlying forces of mortality,
which are largely a product of health conditions
in the population, and partly from differences
that could be expected to occur if one were able
to repeat many times the experience of the
identical set of forces acting on the identical set
of people during the same period of time. The
usefulness of small-area mortality indexes de-
pends heavily upon one’s ability to determine
the general magnitude of the latter, that is, the
random component of variability so that the
former component can be estimated.

The estimation of sampling error for these
indexes is based upon the theoretical work done
by Chiang2?7:33 and others, plus some simplify-
ing assumptions that are justified but that tend
to overestimate the sampling error slightly. For
the purpose of comparing the usefulness of dif-
ferent algebraic forms of an index number, the
crude methods used here would lead to no dif-
ferent conclusions than more sophisticated esti-
mates. Nevertheless, it would provide a sounder
basis for evaluating the indexes and computing
confidence intervals if real experiments of the
Monte Carlo type# could be conducted using
computer simulation of the- process by which

the deaths occur.

Geographic Variability 1

A further question in examining the data by -

“cause of death for these 19 areas is: What is the

8A Monte Carlo type experiment is an experiment
in which one attempts to duplicate the chance variation
occurring in nature by generating random numbers,
usually on a computer, in order to repeat artificially a
phenomenon that does not repeat itself under identical
conditions in nature.

23
significance of the apparently wide differences
in the amount of geographic variability exhib-
ited by indexes for different cause-of-death
groups? Already the National Cancer Institute of
the National Institutes of Health has made wide
use of the leads presented by county-to-county
differences in 10-year summaries of cancer mor-
tality. The leads are systematically being fol-
lowed up by planning pointed epidemiological
studies to test hypotheses. Yet the data pre-
sented in this report suggest that a number of
other causes have wider underlying geographic
variability than cancer does.

“Market Testing” Mortality Indexes
for Health Service Areas

The most direct method of determining
whether indexes of the type presented here can
be useful to planning staffs in the HSA’s is to
make the statistics available to the HSA’s and
work with the Bureau of Health Planning in the
Health Resources Administration, Public Health
Service, to investigate the use made of those
statistics, along with more commonly used in-

dexes, such as infant mortality, in setting goals
and guiding efforts.

If this is to be done, NCHS should routinely
tabulate 5-year aggregations of deaths by cause,
age, sex, and color for each HSA for the periods
centering on the years of the decennial and mid-
decade censuses. Although the 3-year aggrega-
tions used in this report can yield usefully pre-
cise measures for most HSA’s, small numbers are
still a problem. Health conditions do not change
so rapidly from one 5-year period to the next
that the value of 5-year averages is destroyed. In-
creasing the numbers of deaths on which the
HSA index is based by a factor of 5/3 reduces
the typical standard error, and, therefore, the
width of the confidence interval by a factor of
roughly 22% percent, a highly worthwhile gain.

The start in 1985 of what are hoped to be
regular mid-decade censuses, or, at least, large
sample surveys that will make possible improved
estimates of populations by county, age, race,
and sex, makes this an opportune time to con-
sider a program of 5-year summaries of deaths
by cause to facilitate the measurement of health
for areas smaller than States.

000

24
REFERENCES

1 Berg, R. L., ed.: Health Status Indexes. Proceedings
of a conference conducted by Health Services Research
in Tucson, Arizona, Oct. 1-4, 1972. Chicago. Hospital
Research and Educational Trust, 1973.

2Health Services Research: Health Status Indexes—
Work in Progress. Proceedings of a conference conducted
by National Center for Health Services Research in
Phoenix, Arizona, Oct. 25-28, 1976. Chicago. Health
Sery, Res. (Special issue) 11(4), Winter 1976.

3 National Center for Health Statistics: Clearinghouse
on Health Indexes. Washington. U.S. Government Print-
ing Office. Quarterly publication of the Division of
Anglysh, National Center for Health Statistics.

4National Technical Information Service: Health
Planning. Springfield, Va. National Technical Informa-
tion Service, U.S. Department of Commerce. Weekly
publication on behalf of the National Health Planning
Information Center, Bureau of Health Planning and Re-
Sovfees Development.

5 Faulkner, M. J.: Bureau of Health Planning and Re-
Sousces Development: Personal communication.

6Gilson, B. S. et al.: The Sickness Impact Profile:
Development of an outcome measure of health care.
Am. J. Public Health. 65:1304, Dec. 1975.

"Pearl, R.: Some landmarks in the history of biosta-
tistics, in Introduction to Medical Biometry and
S ajsiics. Philadelphia. W. B. Saunders Co., 1940.

8Greenwood, M.: Medical Statistics from Graunt to
Farr. London. Cambridge University Press, 1948.

9 Farr, W.: Methods for comparing local with stand-
ard death-rates, in Vital Statistics, Part IV-Deaths.
London. Office of the Sanitary Institute, 1885. pp.
128-130.

10National Center for Health Statistics: Infant mor-
tality, by J. C. Kleinman. Statistical Notes for Health
Planners, No. 2. Washington. U.S. Government Printing
Office, July 1976.

INational Center for Health Statistics: Mortality, by
J. C. Kleinman. Statistical Notes for Health Planners,
No. 3. Washington. U.S. Government Printing Office,
Feb. 1977.

12National Center for Health Statistics: Standardized
mortality ratio and years of life lost index: State and
health service areas, 1969-71. Statistical Notes for
Health Planners, No. 3, Data Supplement. Washington.
U.S. Government Printing Office, May 1977.

I3National Center for Health Statistics: Cause-of-
death data, by T. D. Woolsey. Statistical Notes for

Health Planners, No. 6. Washington. U.S. Government
Prigting Office, Feb. 1978.

4Elinson, J.: Insensitive health statistics and the
dilemma of the HSAs. Am. J. Public Health. 67(5):417-
4s May 1977.

18Shapiro, S.: A tool for health planners. Am. J.
Pobis Health. 67(9):816-817, Sept. 1977.

6 National Center for Health Statistics: Eighth Revi-
sion International Classification of Diseases, Adapted
for Use in the United States. PHS Pub. No. 1693. Public
Health Service. Washington. U.S. Government Printing
Office, 1967.

7 Mushkin, S.J. et al.: Cost of disease and illness in the
United States in the year 2000. Public Health Rep. (Spe-
“ial Supplement} 93(5):493-588, Sept.-Oct. 1978.

18RBayo, F., Shiman, H. W., and Sobus, B. R.: United
States Population Projections for OASDHI Cost Esti-
mates, Actuarial Study No. 77. Office of the Actuary,
Social Security Administration. Washington. U.S. Gov-
ernment Printing Office, June 1978.

19Rutstein, D. D. et al.: Measuring the quality of
medical care—a clinical method. N. Engl. J. Med. 294:
5% 588, Mar. 1976.

20Rutstein, D. D.: Blueprint for Medical Care. Cam-
bridge. Massachusetts Institute of Technology Press,
19 Id. (See particularly chapter 13.)

21working Group on Preventable and Manageable
Diseases: New Quantitative Method for Measuring the
Quality of Medical Care and Related Economic Indices.
Progessed document. Nov. 25, 1977. p. 4.

22Chen, M. K.: The K index: A proxy measure of
health care quality, in Health Status Indexes—Work in

. Progress. Proceedings of a conference conducted by the

National Center for Health Services Research in Phoenix,
Arizona, Oct. 25-28, 1976. Chicago. Health Serv. Res.
(Special issue) 11(4):452-463, Winter 1976.

23Elinson, J.: Letter to the editor. Am. J. Public
Health. 67(10):986, Oct. 1977.

24 Guralnick, L., and Jackson, A.: An index of unnec-
essary deaths. Public Health Rep. 82(2):180-182, Feb.
1967.

25Woolsey, T. D.: An investigation of low mortality
in certain areas. Public Health Rep. 64(29):909-920,
July 1949.

26Njcholl, M., and Bellows, M. T.: Effect of a confi-
dential inquiry on the recorded mortality from syphilis
and alcoholism. Am. J. Public Health. 24(8):813-820,
1934.

25
27 Chiang, C. L.: Introduction to Stochastic Processes
in Biostatistics. New York. John Wiley & Sons, Inc.,
1968. i

28Woolsey, T. D.: Adjusted death rates and other
indices of mortality, in F. E. Linder and R. D. Grove,
Vital Statistics Rates in the United States, 1900-1940.
Washington. U.S. Government Printing Office, 1943.

29Haenszel, W.: A standardized rate for mortality in
units of lost years of life. Am. J. Public Health. 40:
17-26, 1950.

30Kleinman, J. C.: Age-adjusted mortality indexes
for small areas: Applications to health planning. Am. J.
Public Health. 67:834-840, 1977.

31Romeder, J.-M., and McWhinnie, J. R.: Potential
years of life lost between ages 1 and 70: An indicator
of premature mortality for health planning. Int. J.
Epidemiol. 6:143-151, 1977.

32Kleinman, J. C., Feldman, J. J., and Mugge, R. H.:
Geographic variations in infant mortality. Public Health
Rep. 91(5):423-432, Sept.-Oct. 1976.

33 National Center for Health Statistics: Standard error
of the age-adjusted death rate, by C. L. Chiang. Vital
Statistics Special Reports. 47(9):275-285, Aug. 17,
1961.

000

26
LIST OF DETAILED TABLES

Results of a Taylor's series expansion: £(1/d4) and var(1/d4) for 2 < da < 200, assuming Poissonian distribution ................

Values of mortality index R; for males, using achievable target death rates as standard, for 12 cause-of-death groups and all
causes: United States and 19 health service areas, 1969-71, and United States, 1976 ........cccccccverviiiiiiriinnnreenneniieniisssssssssneeenee

Values of mortality index R} for males, using achievable target death rates as standard, for 12 cause-of-death groups and all
causes: United States and 19 health service areas, 1969-71, and United States, 1976 ............ccocerrenineninnenenninnenssncs scenes

Values of mortality index AR} for males, using achievable target death rates as standard, for 12 cause-of-death groups and all
causes: United States and 19 health service areas, 1969-71, and United States, 1976 ........c..cccceriririiiiinnerissincsinineenscssneeensnnnes
Values of mortality index R}" for females, using achievable target death rates as standard, for 12 cause-of-death groups and
all causes: United States and 19 health service areas, 1969-71, and United States, 1976 ............cceccveiriicnriniinsnininieeecnsnnsnnennan

29

30

32

34

36

27
&
i

2

Fa
A

A

hE in

mith oon

 

”
Table 1. Results of a Taylor's series expansion: E(1/d4) and var (1/d4) for 2 < da < 200, assuming Poissonian distribution

[Courtesy of Benjamin J. Tepping]

 

ll

 

 

 

 

var(1/dg) var(1/d4)
da E(1/dg) da E(1/da)
Mantissa | Exponent Mantissa | Exponent
0.57659 9.0702 -2 0.03337 4.1559 -5
0.43268 7.2527 -2 0.03229. 3.7505 -5
0.32963 49112 -2 0.03128 3.3962 -5
0.25777 3.0481 -2 0.03033 3.0853 -5
0.20779 1.8168 -2 0.02944 2.8112 -5
0.17249 1.0734 -2 0.02860 2.5688 -5
0.14689 6.4325 -3 0.02780 2.3535 -5
0.12776 3.9705 -3 0.02705 2.1616 -5
0.11302 2.5484 -3 0.02634 1.9901 -5
0.10135 1.7067 -3 0.02566 1.8363 -5
0.09190 1.1916 -3 0.02274 1.2651 -5
0.08407 8.6398 -4 0.02042 9.0844 -6
0.07749 6.4720 —-4 0.01853 6.7424 -6
0.07187 4.9834 -4 0.01695 5.1414 -6
0.06702 3.9263 -4 0.01563 4.0098 -6
0.06279 3.1534 -4 0.01450 3.1875 -6
0.05906 2.5739 -4 0.01266 2.1108 -6
0.05575 2.1301 -4 0.01124 1.4693 -6
0.05280 1.7839 -4 0.01010 1.0636 -6
0.05014 1.5095 -4 0.00918 7.9453 -7
0.04774 1.2891 dh 0.00840 6.0907 -7
0.04556 1.1100 -4 0.00775 4.7713 -7
0.04357 9.6275 -5 0.00719 3.8071 -7
0.04175 8.4056 -5 0.00671 3.0861 -7
0.04007 7.3831 -5 0.00629 2.5363 -7
0.03852 6.5206 -5 0.00592 2.1097 -7
0.03709 5.7878 -5 0.00559 1.7736 -7
0.03576 5.1612 -5 0.00529 1.5053 -7
0.03453 4.6222 -5 0.00503 1.2885 -7

 

 

 

 

 

 

 

 

 

 
Table 2. Values of mortality index R; for males, using achievable target death rates as standard, for 12 cause-of-death groups and all
causes: United States and 19 health service areas, 1969-71, and United States, 1976

[Standard error (s;) shown for certain cause-of-death groups]

 

 

Health service areas, 1969-71

 

 

 

   

 

 

 

 

 

 

United United
Cause-of-death group, Rj, and s; States, States, California Connecticut
1969-71 1976
01 04 09 11 01 04
All causes:
0.3703 | 0.2725 0.4108 | 0.3370 | 0.4039 | 0.3495 0.2264 | 0.2259
--- --- 0.011 0.008 0.007 0.003 0.015 0.014
0.2329 | 0.1322 0.2641 | 0.2524 | 0.1973 | 0.2560 0.2670 | 0.1840
Slim --- --- 0.060 0.037 0.045 0.015 0.048 0.058
Diabetes mellitus:
BR deter intima tins ids Eh Eres Eanes a 0.5414 | 0.4350 0.5169 | 0.3436 | 0.4571 | 0.3584 0.4668 | 0.3984
8] 1sarasnnsnnerineiaminsan ss sas esa ARORA RETATIA A TS A022 4 TA RRRRRR RAE - --- 0.109 0.192 0.240 0.069 0.276 0.287
Diseases of the heart: :
BE sritessntensiscissiaias bie see ists rs Era SATs 0.4089 | 0.3264 0.1547 | 0.1992 | 0.2414 | 0.3358 0.2629 | 0.2856
i --- --- 0.207 0.069 0.079 0.021 0.087 0.065
0.5037 | 0.3501 0.4475 | 0.3516 | 0.4617 | 0.3739 0.3818 | 0.3756
--- --- 0.165 0.098 0.063 0.038 0.165 0.115
Diseases of the arteries, arterioles,
and capillaries:
0.3910 | 0.3797 0.5028 | 0.1394 | 0.1444 | 0.3698 | -0.0594 | 0.1652
Acute bronchitis, influenza, and
pneumonia: 4
0.5411 | 0.3007 0.6300 | 0.6493 | 0.5945 | 0.5513 0.5437 | 0.4447
--- --- 0.033 0.016 0.024 0.011 0.038 0.060
0.6568 | 0.4416 0.7252 | 0.5762 | 0.7967 | 0.5942 | -0.0397 | 0.4502
--- --- 0.125 0.092 0.021 0.030 0.721 0.247
Major digestive diseases, except cirrhosis
of the liver:
0.5608 | 0.3021 0.3939 | 0.5370 | 0.5504 | 0.4554 0.4155 | 0.3540
Bovis cnssninnsmmssesnannssiinsramsnsissiiutnsssrsasidrstiinsssnsns 0.6246 | 0.6230 | -0.1615 | 0.7716 | 0.6823 | 0.7437 0.5212 | 0.5868
8 e043048 400 RTE Eas A ETI IA TIRANA RAE T SEER SA RR SRR RRR RRS --- --- 0.948 0.029 0.072 0.011 0.256 0.156
Congenital anomalies and diseases
of early infancy:
Binsin tristan simsiesis isin 0.3645 | 0.1869 0.1477 | 0.2495 | 0.4366 | 0.3582 0.3259 | 0.3467
8] sarsnsveiisrirorsinriinisanasvs spss tobi resssrsarsed ae eth avi vers --- --- 0.325 0.096 0.083 0.028 0.161 0.115
All other diseases:
Bs iinsinstenininiemmsmaistimstarisrsssvisiss irae 0.3805 | 0.3060 0.1163 | 0.1412 | 0.1648 | 0.0809 0.1926 | 0.3014
0.4144 | 0.3271 0.5917 | 0.4561 | 0.5654 | 0.4230 0.0908 | 0.0940
--- --- 0.012 0.011 0.009 0.005 0.033 0.030

 

 

 

 

 

 

 

 

 

 

INot adjusted to U.S. age distribution.

NOTE: In everyday use for health planning, only two significant figures should be carried.

30
Table 2. Values of mortality index R; for males, using achievable target death rates as standard, for 12 cause-of-death groups and all

causes: United States and 19 health service areas, 1969-71, and United States, 1976—Con.

[Standard error (s;) shown for certain cause-of-death groups]

 

 

Health service areas, 1969-71

 

 

 

 

 

Dela- i Loui- ka New Wash- z 3
Wore Florida sions Maryland Michigan Jersey ington Wisconsin
01 06 09 01 02 03 04 01 06 02 01 01 02
0.3751 | 0.4958 | 0.4031 | 0.4899 0.1214 | 0.3026 | 0.4271 | 0.3877 0.3468 | 0.3722 0.2869 0.2291 0.2419
0.013 0.007 0.007 0.006 0.021 0.012 0.006 0.004 0.011 0.006 0.007 0.014 0.010
0.2271 | 0.2266 | 0.2113 | 0.2315 0.2501 | 0.3187 | 0.2493 | 0.2849 0.2718 | 0.2448 0.1745 0.1374 0.1535
0.078 0.077 0.052 0.053 0.048 0.042 0.038 0.019 0.048 0.036 0.035 0.058 0.045
0.6725 | 0.3364 | 0.3047 | 0.6485 0.1616 | 0.4731 | 0.5549 | 0.6238 0.2163 | 0.5802 0.5278 0.5621 0.4801
0.132 0.247 0.363 0.114 0.288 0.241 0.108 0.042 0.350 0.060 0.059 0.146 0.149
0.3875 | 0.4203 | 0.3789 | 0.5452 0.1965 | 0.1636 | 0.4915| 0.3620 0.2961 | 0.4742 0.2067 0.1426 0.3557
0.089 0.058 0.043 0.025 0.125 0.120 0.023 0.025 0.095 0.027 0.059 0.136 0.042
0.3134 | 0.5540 | 0.3942 | 0.6886 0.3494 | 0.4387 | 0.5330 | 0.5130 0.2402 | 0.4909 0.4280 0.3943 0.4187
0.321 0.109 0.094 0.024 0.183 0.097 0.041 0.029 0.399 0.048 0.056 0.131 0.082
0.0962 | 0.6329 | 0.0684 | 0.2640 0.4214 | 0.5508 | 0.2798 | 0.3539 | -0.0335 | 0.0107 | -0.1498 | -0.3421 0.3208
0.6130 | 0.6583 | 0.5955 | 0.6662 0.3971 | 0.4716 | 0.5636 | 0.5975 0.4962 | 0.6063 0.4319 0.2937 0.5348
0.034 0.022 0.026 0.015 0.083 0.055 0.023 0.011 0.044 0.016 0.038 0.075 0.024
0.7321 | 0.6861 | 0.4235 | 0.7276 | 10.2339 | 0.6114 | 0.5516 | 0.7004 0.0979 | 0.4439 0.6307 | -0.0254 0.2947
0.078 0.131 0.306 0.033 0.025 0.170 0.017 0.248 0.720 0.196 0.079 0.079 0.383
0.5233 | 0.3822 | 0.3515 | 0.6650 0.2610 | 0.3316 | 0.5609 | 0.5373 0.4725 | 0.5434 0.4794 0.5569 0.4205
0.6523 | 0.5575 | 0.5248 | 0.6299 0.4198 | 0.4028 | 0.7785 | 0.7667 0.4017 | 0.7104 0.4650 0.3022 | -0.0891
0.120 0.132 0.195 0.049 0.243 0.371 0.012 0.010 0.301 0.018 0.098 0.282 0.764
0.3366 | 0.4410 | 0.5210 | 0.5042 | 10.2196 | 0.2841 | 0.3385 | 0.4087 0.2271 | 0.3495 0.3800 0.3178 0.4022
0.222 0.191 0.044 0.073 0.018 0.189 0.099 0.038 0.315 0.089 0.050 0.108 0.044
0.2454 | 0.5220 ( 0.3959 | 0.4593 0.0857 | 0.3056 | 0.5551 | 0.3611 0.0914 | 0.3631 0.2728 0.0987 0.0742
0.3842 | 0.6066 | 0.4750 | 0.5216 0.0302 | 0.2954 | 0.3982 | 0.3813 0.3999 | 0.2705 0.3364 0.2884 0.2039
0.022 0.010 0.011 0.010 0.046 0.023 0.011 0.010 0.018 0.015 0.012 0.023 0.019

 

 

 

 

 

 

 

 

 

 

 

 

 

31
Table 3. Values of mortality index Rj for males, using achievable target death rates as standard, for 12 cause-of-death groups and all
causes: United States and 19 health service areas, 1969-71, and United States, 1976

[Standard error (57) shown for certain cause-of-death groups]

 

 

Health service areas, 1969-71

 

United United
Cause-of-death group, Rj}, and s} States, | States, California
1969-71 1976

 

01 04 09 1

 

All causes:

1.6039 | 1.3835 | 1.743 | 1.617 | 1.732| 1.555
--- ---] 0.039 | 0.021 | 0.026 | 0.009

 

 

1.3295 | 1.1636 | 1.395 | 1.343 | 1.276 | 1.370

 

--- ---| 0.137 | 0.074 | 0.087 | 0.035

i Sn sitet Bs drin se iss se eR EES sandr ad aT hans Sharsrs os ders Rms daa Earn 22287 | 1.8184 | 3.526 | 1.571 | 1.872] 1.562

BRED SECT CR Se ERR al SE --- ---]1 1.169 | 0.360 | 0.502| 0.163
Diseases of the heart

{RT JOE, © RS ET Lol SE. 1.7273 1 1.85711 1.192 1.2681 1:1.3231 1.825

; --- ---| 0.185 | 0.108 | 0.139 | 0.057

2.0757 | 1.5613 1.857 | 1.549 | 1.987 | 1.601

 

 

--- ---| 0.418 | 0.199 | 0.299 | 0.093
16715} 1.7001 | 2.075 | 1.185 | 1.199| 1.625
--- ---| 0989 | 0.368 | 0.447 | 0.222
2.3040 | 1.4491 | 3.015 | 3.339 | 2.859 | 3.529
--- ---| 0.358 | 0.223 | 0.231 | 0.105
A IEC OR SRE Lp EL ER IE J rR CE Lest LOC Eo 3.5108 | 2.0920 | 3.728 | 2.865 | 6.929 | 2.751
Bers nrre asain tebe eo Re a Fe a bas REA RAS 3 nA RAR AREA --- ---| 1.355 | 0.754 | 1.400| 0.322
Major digestive diseases, except cirrhosis of the liver
REL RE ie i en So ROI SN, I 2.3999 | 1.4563 | 1.650 | 2.181 | 2.235| 1.849
8} sean ATi aN ans SAAR SR A SSH Rea sa SEER HAA Ceasar aaa as se sar en --- ---| 0.494 | 0.347 | 0.388 0.137
Cirrhosis of the liver
A CL A EER SEER EE ot. JN 2.9339 | 2.9862 | 1.148 | 4.437 | 3.192 | 4.142
le Ee EO OS PERE SU Le --- ---| 0.385 | 0.519 | 0.607 | 0.258
Congenital anomalies and diseases of early infancy
R; SAR ad ee TES RAS LR eA Ee Hess dars ss 1 4A BEE ae Heras aaa AAA SERA TAs 1.6028 | 1.2367 | 1.290 | 1.378 | 1.795| 1.559
7 --- ---| 0.185 | 0.130 | 0.184 | 0.063
BBY ins dd rt a dr i ads ens ss Be A Re GT 1.6519 | 1.4669 | 1.151 | 1.166 | 1.202| 1.090
fans HE Se ron EE arse ETE ex MES Ep sve np yew Era Tree E eg ET AEE eR eR Tr A EEA A rs FrA Fe Fes --- --- | 0.090 | 0.050 | 0.059 | 0.022
Accldental injuries and other trauma:
R; Se re rrr ry Bs AAT ar rsa Resaupe des ears SA Ree ET ET 1.7347 | 1.5030 | 2.495 | 1.877 | 2.330| 1.739
AL em Ec NE Ne NEB Ae Sr] --- ---| 0.074 | 0.036 | 0.050| 0.016

 

 

 

 

 

 

 

NOTE: In everyday use for health planning, only three significant figures should be carried.

32
Table 3. Values of mortality index R;} for males, using achievable target death rates as standard, for 12 cause-of-death groups and all
causes: United States and 19 health service areas, 1969-71, and United States, 1976—Con.

[Standard error (sp) shown for certain cause-of-death groups]

 

 

Health service areas, 1969-71

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Connecticut Dela- , Florida Loui- Maryland Michigan New Ww ash Wisconsin
ware siana Jersey | ington
01 04 01 06 09 01 02 03 04 01 06 02 01 01 02
1.294 | 1.294 | 1.604 | 2.128 | 1.715 | 1.978 | 1.143 | 1.437 | 1.755 | 1.646 | 1.544 | 1.608 1.405 | 1.301 | 1.320
0.027 | 0.024 | 0.033 | 0.043 | 0.025 | 0.024 | 0.030 | 0.024 | 0.018 | 0.012 | 0.029 | 0.019| 0.016 | 0.025 | 0.017
1.395 | 1.227 | 1.300 | 1.302 | 1.276 | 1.336 | 1.472 | 1.474 | 1.347 | 1.404 | 1.402 | 1.326 1.218 | 1.189 | 1.181
0.110 | 0.089 | 0.113 | 0.125 | 0.078 | 0.071 | 0.140 | 0.098 | 0.058 | 0.042 | 0.112 | 0.065| 0.057 | 0.097 | 0.063
1.877 | 1686 | 3.189 | 1.642 | 1.439 | 2.885 | 1.218 | 1.924 | 2.262 | 2664 | 1.407 | 2.580 | 2.334 | 2.498 | 1.938
0.584 | 0.496 | 0914 | 0.629 | 0.393 | 0.510 | 0.563 | 0.505 | 0.360 | 0.266 | 0.448 | 0.458 | 0.388 | 0.698 | 0.391
1.377 | 1.443 | 1.634 | 1.881 1.684 | 2.305 | 1.265 | 1.213 | 2.044 | 1.582 | 1.422 | 1.984 1.261 | 1.172 | 1.578
0.175 | 0.160 | 0.206 | 0.263 | 0.156 | 0.171 | 0.197 | 0.127 | 0.128 | 0.072 | 0.174 | 0.133 0.088 | 0.142 | 0.122
1628 | 1.624 | 1.489 | 2.338 | 1.691 3.442 | 1.612 | 1.798 | 2.222 | 2.090 | 1.428 | 2.021 1.761 | 1.663 | 1.741
0.303 | 0.276 | 0.312 | 0.457 | 0.246 | 0.341 | 0.379 | 0.281 | 0.212| 0.135 | 0.252 | 0.218 | 0.179 | 0.297 | 0.196
1.040 | 1.201 1.133 | 4.113 | 1.145 | 1.418 | 1.808 | 2.396 | 1.407 | 1.612 | 1.054 | 1.047 0.959 | 0.764 | 1.487
0.461 0.510 | 0.592 | 1.561 | 0.401 | 0.435 | 0.896 | 0.775 | 0.357 | 0.277 | 0.480 | 0.311 0.247 | 0.414 | 0.433
2368 | 1.830 | 2.922 | 3.936 | 2.718 | 3.651 1.721 | 1.920 | 2.402 | 2.819 | 2.670 | 2.995 1.804 | 1.554 | 2.671
0.252 | 0.184 | 0.322 | 0.406 | 0.209 | 0.232 | 0.252 | 0.190 | 0.146 | 0.107 | 0.278 | 0.179 | 0.117 | 0.197 | 0.180
1.003 | 1.838 | 4.758 | 3.398 | 1.800 | 4.868 | 0.434 | 2.688 | 2.319 4.046 | 1.403 | 1.897 2.808 | 1.149 | 1.418
0.503 | 0.687 | 1.567 | 1.287 | 0.553 | 1.022 | 0.067 | 0.839 | 0.530 | 0.486 | 0.508 | 0.509 0.554 | 0.461 | 0.419
1.779 | 1.655 | 2.129 | 1.690 | 1.665 | 3.044 | 1.489 | 1.605 | 2.296 | 2.188 | 2.029 | 2.211 1.948 | 2.307 | 1.742
0.442 | 0.350 | 0.525 | 0.453 | 0.293 | 0.395 | 0.497 | 0.322 | 0.281 | 0.175 | 0.407 | 0.290 | 0.254 | 0.452 | 0.261
2.091 2427 | 3.100 | 2.858 | 2.149 | 3.105 | 1.865 | 1.789 | 6.095 | 5.040 | 1.679 | 4.469 1.877 | 1.549 | 1.125
0.552 | 0.528 | 0.834 | 0.920 | 0.422 | 0.560 | 0.665 | 0.385 | 0.646 | 0.382 | 0.533 | 0.582 | 0.296 | 0.504 | 0.211
1.499 | 1556 | 1.557 | 1.888 | 2.131 | 2.086 | 1.108 | 1.537 | 1.557 | 1.707 | 1.406 | 1.616 1.627 | 1.595 | 1.716
0.171 | 0.171 | 0.218 | 0.204 | 0.170 | 0.183 | 0.176 | 0.211 | 0.111 | 0.079 | 0.170 | 0.103 0.122 | 0.218 | 0.145
1.241 | 1.446 | 1.327 | 2.221 1.697 | 1.874 | 1.161 | 1.443 | 2.323 | 1.616 | 1.105 | 1.655 1.392 | 1.116 | 1.081
0.073 | 0.072 | 0.085 | 0.115 | 0.070 | 0.065 | 0.090 | 0.070 | 0.062 | 0.033 | 0.068 | 0.055 | 0.045 | 0.067 | 0.044
1.100 | 1.105 | 1.652 | 2.725 | 1.960 | 2.123 | 1.048 | 1.467 | 1.696 | 1.628 | 1.674 | 1.373 1.542 | 1.421 | 1.274
0.040 | 0.036 | 0.057 | 0.078 | 0.042 | 0.043 | 0.047 | 0.046 | 0.030 | 0.019 | 0.052 | 0.028 | 0.027 | 0.045 | 0.028

 

33
Table 4. Values of mortality index R}"’ for males, using achievable target death rates as standard, for 12 cause-of-death groups and all

causes: United States and 19 health service areas, 1969-71, and United States, 1976

"ne

[Standard error (s;') shown for certain cause-of-death groups]

 

 

Health service areas, 1969-71

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

United United
.Cause-of-death group, R}", and s}"' States, | States, California
1969-71 1976
01 04 09 1"
All causes: .

BE tiiasavinTAs aatoniAis iyi REA 1.614 | 1.367 | 1.691 | 1.512 | 1.660 | 1.533
P aieaseeneseReRe ORR RRA TERNS RAR RONNIE RIESE AAT --- ---| 0.029 | 0.016 | 0.021 0.008
1.246 1.221 1.212 | 1.2N 1.131 1.229

--- ---| 0.047 | 0.028 | 0.035| 0.014

1.972 1.524 | 1.530 | 1.339 | 1.751 1.571

--- ---| 0.270 | 0.130 | 0.197 | 0.069

1.457 1.229 | 1.321 1.208 | 1.253 | 1.366

--- ---| 0.034 | 0.020 | 0.026 | 0.011

2.020 1.460 | 1.531 1.599 | 1.564 | 1.678

--- ---| 0.109 | 0.070 | 0.093 | 0.035

1.428 1.229 | 1.755 | 1.366 | 1.423 | 1.440

--- ---| 0.184 | 0.096 | 0.130 | 0.053

2.929 1.494 | 2908 | 3.230 | 2.821 | 3.474

--- ---| 0.302 | 0.192 | 0.218 | 0.100

2.281 1.343 | 3.462 | 1.692 | 3.621 2.001

--- ---| 0.277 | 0.131 | 0.286 | 0.074

2.059 1.366 | 1.823 | 2.068 | 2.197 | 1.803

--- ---| 0.331 | 0.216 | 0.285 | 0.099

2121 2.004 | 1.964 | 4.785 | 2.930 | 3.307

-.- ---| 0.175 | 0.165 | 0.166 | 0.067

1.774 1.262 | 1.466 | 1.308 | 1.739 | 1.566

--- ---1 0.111 | 0.062 | 0.077 | 0.028

1.593 1.567 | 1.109 | 1.174 | 1.207 | 1.102

-.- ---| 0.076 | 0.043 | 0.055 | 0.020

1.697 1.496 | 2.542 | 1.838 | 2.334 | 1.747

--- ---| 0.084 | 0.040 | 0.055| 0.018

 

 

 

 

 

 

 

 

34

NOTE: In everyday use for health planning, only three significant figures should be carried.
Table 4. Values of mortality index R}’ for males, using achievable target death rates as standard, for 12 cause-of-death groups and all
causes: United States and 19 health service areas, 1969-71, and United States, 1976—Con.

"

[Standard error (s;') shown for certain cause-of-death groups]

 

 

Health service areas, 1969-71

 

 

 

 

 

Connecticut Dela- Florida Loui- Maryland Michigan New | Wash- Wisconsin
ware : siana Jersey ington
01 04 01 06 09 01 02 03 04 01 06 02 01 01 02
1.285 | 1.308 | 1.646 | 1.896 | 1.659 | 2.017 | 1.117 | 1.477 | 1.801 | 1.655 | 1.520 | 1.591 1.419 | 1.304 | 1.350
0.020 | 0.020 | 0.029 | 0.028 | 0.018 | 0.021 | 0.025 | 0.025 | 0.015 | 0.010 | 0.025 | 0.015 | 0.014 | 0.022 | 0.014
1.208 | 1.202 | 1.378 | 1.255 | 1.338 | 1.570 | 1.074 | 1.393 | 1.502 | 1.346 | 1.234 | 1.304 | 1.138 | 1.044 | 1.178
0.040 | 0.037 | 0.053 | 0.040 | 0.030 | 0.037 | 0.052 | 0.053 | 0.027 | 0.018 | 0.046 | 0.025 | 0.024 | 0.041 | 0.026
1.764 | 1.645 | 2650 | 1.581 | 1.401 | 3.241 | 1.128 | 1.684 | 2.572 | 2.793 | 1.792 | 1.913 | 1.870 | 1.864 | 1.815
0.200 | 0.194 | 0.328 | 0.196 | 0.129 | 0.231 | 0.228 | 0.241 | 0.159 | 0.109 | 0.218 | 0.138 | 0.147 | 0.242 | 0.142
1.192 | 1.187 | 1672 | 1.296 | 1.329 | 1.867 | 1.077 | 1.426 | 1.665 | 1.456 | 1.417 | 1.604 | 1.302 | 1.290 | 1.381
0.028 | 0.027 | 0.043 | 0.029 | 0.022 | 0.031 | 0.036 | 0.040 | 0.022 | 0.014 | 0.035 | 0.021 0.019 | 0.031 | 0.021
1.557 1.479 1.762 1.928 1.565 | 3.061 1.320 | 1.787 | 2.176 | 2.070 | 1.872 1.842 1.733 | 1.572 | 1.900
0.096 | 0.093 | 0.139 | 0.102 | 0.072 | 0.126 | 0.130 | 0.137 | 0.077 | 0.050 | 0.112 | 0.068 | 0.066 | 0.108 | 0.073
1.423 | 1.276 | 1.332 | 1.685 | 1.340 | 1.722 | 1.349 | 1.969 | 1.591 | 1.323 | 1.472 | 1.318 | 1.349 | 0.891 | 1.421
0.143 | 0.138 | 0.165 | 0.154 | 0.098 | 0.139 | 0.186 | 0.254 | 0.100 | 0.063 | 0.163 | 0.089 | 0.087 | 0.106 | 0.100
2.336 | 1.898 | 2983 | 3.371 | 2.710 | 3.844 | 1.767 | 2.012 | 2.490 | 2.803 | 2.649 | 2902 | 1.926 | 1.570 | 2.614
0.224 | 0.174 | 0.318 | 0.301 | 0.175 | 0.236 | 0.241 | 0.211 | 0.137 | 0.099 | 0.271 | 0.157 | 0.116 | 0.192 | 0.168
1.194 | 1.701 | 2.796 | 2.664 | 2.035 | 2.860 | 0.930 | 2.226 | 1.969 | 2.475 | 2.113 | 1.557 | 2.420 | 1.624 | 1.481
0.131 | 0.177 | 0.323 | 0.218 | 0.138 | 0.238 | 0.151 | 0.297 | 0.130 | 0.103 | 0.213 | 0.109 | 0.142 | 0.173 | 0.109
1.492 | 1.511 | 2.109 | 1.819 | 1.881 | 2912 | 1.172 | 1.618 ; 2.356 | 2.234 | 2.038 | 2.131 1.883 | 2.220 | 1.826
0.262 | 0.252 | 0.401 | 0.278 | 0.201 | 0.299 | 0.322 | 0.293 | 0.202 | 0.132 | 0.295 | 0.191 0.189 | 0.335 | 0.192
2.164 | 2.461 | 2417 | 1.657 | 25678 | 2.379 | 1.301 | 2.338 | 3.674 | 3.391 | 1.496 | 2.378 | 1.778 | 1.122 | 1.668
0.149 | 0.166 | 0.207 | 0.153 | 0.129 | 0.142 | 0.155 | 0.185 | 0.136 | 0.086 | 0.144 | 0.105 | 0.088 | 0.123 | 0.091
1.537 | 1.577 | 1.753 | 2.212 | 2.010 | 2.039 | 1.292 | 1.543 | 1.734 | 1.825 | 1.693 | 1.893 | 1.546 | 1.464 | 1.490
0.090 | 0.081 | 0.107 | 0.127 | 0.083 | 0.073 | 0.099 | 0.076 | 0.057 | 0.037 | 0.090 | 0.061 0.051 | 0.083 | 0.056
1.216 | 1.447 | 1.345 | 2312 | 1.651 | 1.881 | 1.064 | 1.438 | 2.284 | 1.623 | 1.096 | 1.553 | 1.425 | 1.114 | 1.093
0.061 | 0.066 | 0.082 | 0.095 | 0.057 | 0.062 | 0.078 | 0.075 | 0.055 | 0.031 | 0.066 | 0.046 | 0.045 | 0.066 | 0.042
1.11% 1.094 1.614 2.878 1.991 2.104 1.018 | 1.369 1.666 1.617 1.715 1.397 1.474 1.397 1.236
0.045 | 0.041 | 0.063 | 0.088 | 0.046 | 0.047 | 0.052 | 0.047 | 0.033 | 0.022 | 0.059 | 0.032 | 0.030 | 0.050 | 0.031

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

35
Table 5. Values of mortality index R}’ for females, using achievable target death rates as standard, for 12 cause-of-death groups and all

causes: United States and 19 health service areas, 1969-71, and United States, 1976

* [Standard error (sj'") shown for certain cause-of-death groups]

 

 

Health service areas, 1969-71

 

 

 

 

 

 

 

 

United United
Cause-of-death group, R;", and s;"' States, States, California
1969-71 1976
01 04 09 1
All causes:
| EA I en er BL Sy NRE NEL J 1.651 1.379 1.676 | 1.628 | 1.613 | 1.644
Breit ns rs Tse Epi eT em eR ee ere reese --- ---| 0.040 | 0.022 | 0.029 | 0.011
All malignant neoplasms
BB sires ithiomsssnysb into ioe bees ems si a Hamas eA ome Soe ot AE nA 1.246 | 1.207 1.222 | 1.381 | 1.199 | 1.301
consis msasesnimabinssa ss rassnsus sei AER RE ers Te STROH REARS SEBS AAA TETD --- ---| 0.055 | 0.032 | 0.040 | 0.015
Diabetes mellitus:
irs a con Er ere tere re EV pres a ons nowt STATS bees opt 2.240 1.644 1.705 | 1.256 | 1.908 | 1.719
7 --- ---] 0.257 | 0.131 | 0.212| 0.070
1.691 1.377 1.437 | 1.301 1.421 1.589
--- ---| 0.062 | 0.034 | 0.046| 0.018
1.858 1.362| 1.360 | 1.708 | 1.335| 1.646
--- ---| 0.114 | 0.076 | 0.086 | 0.035
1.599 1.329| 2.545 | 1.908 | 1.496 | 1.580
--- ---| 0.373 | 0.201 0.214 | 0.086
rice ptr n Se tesnuier Enno De EE A tee ese 2.846 | 1.457| 3.617 | 3.111 | 3.260| 3.177
7 --- ---| 0.446 | 0.237 | 0.297 0.111
1.869 1.360 4.239 | 1.730 | 3.841 | 2.160
--- ---| 0.648 | 0.210 | 0.503 | 0.119
2.143 1.427 2.335 | 1.911 2.514 | 2.142
--- ---| 0.506 | 0.264 | 0.378 0.140
Te rae es rs TS ar ree Brees 2.459 2112) 3.309 | 5.339 | 2813| 3.762
A --- ---| 0.360 | 0.251 | 0.235| 0.103
BY soissieriniivivmisvisinminssmisssisisnsssvteti srt sass Gav a ES 1.722 1.307] 1.390 | 1.327 | 1.662] 1.528
; --- ---| 0.123 | 0.071 | 0.084 | 0.032
1.683 1.518] 1.291 | 1.279 | 1.404 | 1.284
ew ---| 0.094 | 0.054 | 0.074| 0.026
1.729 1.524 | 2.856 | 2.426 | 2.239 | 2.238
--- ---] 0.154 | 0.078 | 0.093| 0.035

 

 

 

 

 

 

 

 

NOTE: In everyday use for health planning, only three significant figures should be carried.

36
Table 5. Values of mortality index R}'for females, using achievable target death rates as standard, for 12 cause-of-death groups and all
causes: United States and 19 health service areas, 1969-71, and United States, 1976—Con.

[Standard error (s;

1

) shown for certain cause-of-death groups]

 

 

Health service areas, 1969-71

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Connecticut Dela- Florida Loui- Maryland Michigan New Wash- Wisconsin
ware siana Jersey | ington
01 04 01 06 09 01 02 03 04 01 06 02 01 01 02
1.368 1.365 1.690 1.729 1.676 2.016 1.132 | 1.531 1.816 | 1.734 | 1.572 1.678 1.461 1.259 | 1.449
0.028 | 0.027 | 0.040 | 0.035 | 0.023 | 0.028 | 0.033 | 0.035 | 0.021 | 0.014 | 0.035 | 0.020 0.019 | 0.029 | 0.020
1.183 1.285 1.333 1.044 1.338 1.394 1.144 | 1.246 | 1.370 | 1.277 | 1.256 1.386 1.199 | 1.184 | 1.245
0.041 0.043 | 0.057 | 0.038 | 0.032 | 0.038 | 0.054 | 0.053 | 0.029 | 0.019 | 0.052 0.028 0.027 | 0.048 | 0.030
1.659 1.298 3.446 1.880 1.647 | 4.887 1.034 | 2.058 | 3.272 | 3.108 | 2.404 2.425 1.715 | 1.834 | 2.336
0.191 0.158 | 0.349 | 0.231 0.143 0.268 | 0.251 0.279 | 0.167 | 0.119 | 0.267 0.146 0.140 | 0.249 | 0.178
1.256 1.310 2.142 1.416 1.483 2.628 1.093 | 1.618 | 2.107 | 1.834 | 1.556 1.982 1.331 1.203 | 1.414
0.046 | 0.046 | 0.082 | 0.051 0.037 | 0.061 0.063 | 0.074 | 0.041 0.025 | 0.060 0.037 0.031 0.048 | 0.034
1.559 1.601 1.842 1.773 1.615 2.565 1.417 | 1.352 | 1.943 | 2.134 | 1.698 1.655 1.741 1.425 | 1.850
0.101 0.100 | 0.153 0.104 0.075 0.109 | 0.141 0.125 | 0.075 | 0.053 | 0.126 0.066 0.071 0.102 | 0.077
1.284 1.486 1.664 1.356 | 1.387 2.047 1.021 1.902 | 2.140 | 1.476 | 1.867 1.213 1.518 | 1.782 | 1.544
0.201 0.216 | 0.311 0.212 0.158 | 0.216 | 0.252 | 0.329 | 0.191 0.100 | 0.285 0.129 0.153 | 0.274 | 0.189
1.761 1.860 | 3.219 | 3.071 2.348 | 3.886 | 1.184 | 1.913 | 2.031 | 2.801 2.469 2.759 1.764 | 1.368 | 2.330
0.234 | 0.206 | 0.400 | 0.350 | 0.204 | 0.281 0.234 | 0.254 | 0.148 | 0.120 | 0.312 | 0.188 0.137 | 0.217 | 0.194
1.282 1.439 1.5687 2.396 2.003 2.281 1.227 | 1.595 | 1.637 | 1.960 | 1.758 1.485 2.567 | 1.526 | 1.431
0.222 0.253 0.432 | 0.378 | 0.214 0.315 | 0.336 | 0.386 | 0.200 | 0.144 | 0.360 0.171 0.240 | 0.337 | 0.218
2.131 1.382 2.120 1.902 1.383 3.472 1.346 | 1.415 | 1.857 | 2.016 | 2.608 2.393 2.255 | 1.296 | 1.705
0.442 | 0.303 0.539 | 0.443 0.203 0.442 | 0.411 0.348 | 0.236 | 0.159 | 0.513 0.271 0.281 0.283 | 0.235
2.724 | 3.042 | 2.161 3.021 3.554 | 2.603 | 1.627 | 3.530 | 4.261 3.487 | 1.626 | 3.063 2.444 | 1.382 | 2.039
0.252 | 0.264 | 0.286 | 0.291 0.217 0.222 0.237 | 0.358 | 0.215 | 0.128 | 0.232 0.180 0.153 | 0.224 | 0.158
1.610 1.527 1.738 1.827 1.889 1.960 1.219 | 1.596 | 1.741 1.812 | 1.619 1.944 1.338 | 1.279 | 1.527
0.104 | 0.089 | 0.123 | 0.131 0.091 0.081 0.109 | 0.090 | 0.066 | 0.042 | 0.100 | 0.070 0.055 | 0.089 | 0.065
1.305 | 1.316 | 1.425 | 2.226 | 1.722 | 1.997 | 0950 | 1.689 | 2.209 | 1.777 | 1.302 1.542 1.505 | 1.105 | 1.284
0.077 | 0.072 | 0.100 | 0.114 | 0.067 0.076 | 0.086 | 0.099 | 0.065 | 0.039 | 0.088 0.053 0.055 | 0.076 | 0.054
1.263 1.045 1.505 2.630 2.040 1.628 1.072 1.303 1.513 1.618 1.929 1.269 1.776 | 1.379 1.439
0.082 | 0.067 0.106 0.142 | 0.079 0.071 0.089 | 0.080 | 0.054 | 0.037 | 0.109 0.051 0.058 | 0.087 | 0.058
000

37
) ed
i, a Tole
a ni
rl fi ui
Lh
} ml al

 

i i dh
hl

i ful

: he ok ¢
ge JR A fh
be La A is
Br ahh

jo i A

a x
=o x ;
pag

. i

set
| Tie
by

erg fy Np

es A
¢
A
Vie=S rt =
31 28d
ae
By E i .
3
gp— a = SCSd
=r
mea =
) £1 Fic &
a

oF = ay

 
da Rl a i ican

seals omnia

 

 

Series 1.

Series 2.

Series 3.

Series 4.

Series 10.

Series 11.

Series 12.

Series 13.

Series 14.

Series 20.

Series 21.

Series 22.

Series 23.

VITAL AND HEALTH STATISTICS Series

Programs and Collection Procedures.—Reports which describe the general programs of the National
Center for Health Statistics and its offices and divisions and data collection methods used and include
definitions and other material necessary for understanding the data.

Data Evaluation and Methods Research.—Studies of new statistical methodology including experi-
mental tests of new survey methods, studies of vital statistics collection methods, new analytical
techniques, objective evaluations of reliability of collected data, and contributions to statistical theory.

Analytical Studies.—Reports presenting analytical or interpretive studies based on vital and health
statistics, carrying the analysis further than the expository types of reports in the other series.

Documents and Committee Reports.—Final reports of major committees concerned with vital and
health statistics and documents such as recommended model vital registration laws and revised birth
and death certificates.

Data From the Health Interview Survey.—Statistics on illness, accidental injuries, disability, use of
hospital, medical, dental, and other services, and other health-related topics, all based on data collected
in a continuing national household interview survey.

Data From the Health Examination Survey and the Health and Nutrition Examination Survey.—Data
from direct examination, testing, and measurement of national samples of the civilian noninstitu-
tionalized population provide the basis for two types of reports: (1) estimates of the medically defined
prevalence of specific diseases in the United States and the distributions of the population with respect
to physical, physiological, and psychological characteristics and (2) analysis of relationships among the
various measurements without reference to an explicit finite universe of persons.

Data From the Institutionalized Population Surveys. —Discontinued effective 1975. Future reports from
these surveys will be in Series 13.

Data on Health Resources Utilization. —Statistics on the utilization of health manpower and facilities
providing long-term care, ambulatory care, hospital care, and family planning services.

Data on Health Resources: Manpower and Facilities. —Statistics on the numbers, geographic distri-
bution, and characteristics of health resources including physicians, dentists, nurses, other health
occupations, hospitals, nursing homes, and outpatient facilities.

Data on Mortality. —Various statistics on mortality other than as included in regular annual or monthly
reports. Special analyses by cause of death, age, and other demographic variables; geographic and time
series analyses; and statistics on characteristics of deaths not available from the vital records based on
sample surveys of those records.

Data on Natality, Marriage, and Divorce.—Various statistics on natality, marriage, and divorce other
than as included in regular annual or monthly reports. Special analyses by demographic variables;
geographic and time series analyses; studies of fertility; and statistics on characteristics of births not
available from the vital records based on sample surveys of those records.

Data From the National Mortality and Natality Surveys.—Discontinued effective 1975. Future reports
from these sample surveys based on vital records will be included in Series 20 and 21, respectively.

Data From the National Survey of Family Growth.—Statistics on fertility, family formation and dis-
solution, family planning, and related maternal and infant health topics derived from a biennial survey
of a nationwide probability sample of ever-married women 15-44 years of age.

For a list of titles of reports published in these series, write to: Scientific and Technical Information Branch

National Center for Health Statistics
Public Health Service
Hyattsville, Md. 20782
 

DHHS Publication Number (PHS) 81-1359
Series 2-No. 85

NCHS

U.S. DEPARTMENT OF HEALTH AND HUMAN SERVICES
Public Health Service

Office of Health Research, Statistics, and Technology
National Center for Health Statistics

3700 East-West Highway

Hyattsville, Maryland 20

OFFICIAL BUSINESS
PENALTY FOR PRIVATE USE, $300

For listings of publications in
the VITAL AND HEALTH
STATISTICS series, call 301-
436-NCHS.

POSTAGE AND FEES PAID
U.S. DEPARTMENT OF H.H.S.
HHS 396

 
 

 

 

Computer-Assisted
Spirometry Data Analysis

for the National

Health and Nutrition
ETFO TA ETE

U.S. DEPARTMENT OF

HEALTH AND HUMAN SERVICES

Public Health Service

Office of Health Research, Statistics, and Technology
National Center for Health Statistics

 

 
 

 

 

 

 

Library of Congress Cataloging in Publication Data

United States. National Center for Health Statistics.
Computer assisted spirometry data analysis for the Health and Nutrition Examination
Survey, 1971-1980.

(Vital and health statistics : Series 2, Data evaluation and methods vesearch 3 no. 86)
(DHHS publication ; no. (PHS) 81-1360)

Written by David P. Discher et al.

Includes bibliographical references.

1. Spirometry—Data processing. 2. Health and Nutrition Examination Survey. I. Dis-
cher, David P. II. Title. III. Series: United States. National Center for Health Statistics,
Vital and health statistics : Series 2, Data evaluation and methods research ; no. 86.
IV. Series: United States. Dept. of Health and Human Services. DHHS publication ; no.
(PHS) 81-1360.

RA409.U45 no. 86 [RC784.565] 812' 0723s 80607930
ISBN 0-8406-0195-6

 

 

 
Data Evaluation and Methods Research Series 2
Number 86

Computer-Assisted
Spirometry Data Analysis

for the National
Health and Nutrition

Examination Survey, 1971-80

The equipment, procedures, and data reduction methods employed
in the National Health and Nutrition Examination Survey for the
collection and analysis of spirometric data are described. Data vari-
ability and testing methodology are discussed, as well as the influ-
ence of milieu and technician training. The computer programs that
drive the data reduction and calibration are detailed, as are the algo-
rithms used in the calculation of various spirometric parameters.
The algorithms chosen for the determination of certain critical
parameters are documented and validated.

DHHS Publication (PHS) 81-1360

 

U.S. DEPARTMENT OF HEALTH AND HUMAN SERVICES
Public Health Service
Office of Health Research, Statistics, and Technology
National Center for Health Statistics
Hyattsville, Md. October 1980
NATIONAL CENTER FOR HEALTH STATISTICS

DOROTHY P. RICE, Director

ROBERT A. ISRAEL, Deputy Director
JACOB J. FELDMAN, Ph.D., Associate Director for Analysis
GAIL F. FISHER, Associate Director for Cooperative Health Statistics Systems
ROBERT A. ISRAEL, Acting Associate Director for Systems
ALVAN O. ZARATE, Ph.D., Acting Associate Director for International Statistics
ROBERT C. HUBER, Associate Director for Management
MONROE G. SIRKEN, Ph.D., Associate Director for Mathematical Statistics
PETER L. HURLEY, Associate Director for Operations
JAMES M. ROBEY, Ph.D., Associate Director for Program Development
GEORGE A. SCHNACK, Ph.D., Associate Director for Research
ALICE HAYWOOD, Information Officer

DIVISION OF HEALTH EXAMINATION STATISTICS

ROBERT L. MURPHY, Director
JEAN ROBERTS, Chief, Medical Statistics Branch
KURT MAURER, Acting Chief, Survey Planning and Development Branch

COOPERATION OF THE U.S. BUREAU OF THE CENSUS

Under the legislation establishing the National Health Survey, the Public Health Service is
authorized to use, insofar as possible, the services or facilities of other Federal, State, or private
agencies.

In accordance with specifications established by the National Center for Health Statistics,
the Bureau of the Census, under contractual agreement, participated in planning the survey and
collecting the data.

Vital and Health Statistics-Series 2-No. 86

 

DHHS Publication (PHS) 81-1360
Library of Congress Catalog Card Number 80-607930
CONTENTS

Introduction

 

Background
Spirometry Data Variability
Testing Methodology

 

 

 

Instrumentation
Spirometer and Support Electronics
Calibrators

 

 

 

Data Acquisition System

 

Spirometry Data Analysis: Program Description
Calibration Factor Computation (First Stage)
Volume and Flow Rate Signal Data Corrections (Second Stage)
Spirometer Trial Parameter Computation (Third Stage)

 

 

 

 

‘Spirometry Data Analysis: Validity of Alternative Algorithms
Methodology
Zero-Time and FEV ( Calculations
End-of-Trial, FVC, and FEF95.759 Calculations

 

 

 

 

Summary and Conclusions

 

References

 

List of Detailed Tables

 

Appendixes
I. Glossary
II. General Spirometric Test Procedures Used by NHANES

 

 

LIST OF FIGURES

 

1. Typical subject flow-volume curve set
2. Sample spirogram demonstrating the short baseline procedural error .....ceeesreenee ITA TIA
3. Sample spirogram demonstrating the no end-of-test plateau procedural CTTOL cosovussvomsorinsvsssannssin sons
4. Sample oscilloscope tracing demonstrating the premature termination artifact procedural error ....
5. Sample oscilloscope tracing demonstrating the inhalation artifact procedural €ITOT veueereessosessenees

6. Sample oscilloscope tracings, one normal ana one demonstrating the Venturi artifact procedural
error

 

7. A depiction of the mechanics of a spirometric Venturi artifact

 

8. Sample oscilloscope tracing demonstrating the low peak flow artifact procedural eITor ......ccesseensee

9. Sample oscilloscope tracing demonstrating the hesitation artifact nracedural error .....oeeeeeens shoes

45
£7

© © oo oo
10.
11.
12.
13.
14.
15.
16.
13.
18.
19.
20.
21.

22.

23.

Schematic of the NHANES spirometry system

 

Spirometer data—calibration sine wave

 

Digital data array from a calibration curve showing trough (B) and peak (C) .....cceurenecncesassancsananes
Schematic of flow-volume curve showing the relation of zero time to time of peak flow .....ccceeusues
Data output sheet for a normal subject showing a set of 5 SPITOGrams .....c.ccceescescrsnccsensessrsssansasesnes
Data output sheet for an abnormal subject showing a set of 5 spirograms .......... erfiviessearastinsiontytasis

Example of a normal time-volume spirogram

 

Normal spirogram (time-volume and flow-volume curves) with noise signals superimposed ............
Abnormal spirogram (time-volume and flow-volume curves) with noise signals superimposed ........
Manual calculation of ¢( showing a reproduction of the 4 alternative methods .......ccceeerueereraesencanes

Example of an abnormal time-volume spirogram

 

Manual calculation of end of trial showing the 4 alternative methods ............ wsteAeISR AAR RSS Ass RASS

FEV) ( analysis—means and standard deviations (30) of paired differences from triangular ex-
trapolation (method 2) measurements

 

FVC analysis—means and standard deviations (30) of paired differences from negative flow
(method 3) measurements

 

LIST OF TEXT TABLES

Procedural error codes and their definitions

 

Methods for zero-time and end-of-trial determinations

 

 

SYMBOLS

Data not available ceeeeeee  -- =
Category not applicable——-—-----ceomeeconeeee |

 

Quantity zero .
Quantity more than 0 but less than 0.05-—- 0.0
Figure does not meet standards of

reliability or precision--———----eeeeeereeeeeeen *

 

 

 

10
12
14
15
17
18
20
22
23
26
27

30

32

33

21
COMPUTER-ASSISTED SPIROMETRY DATA ANALYSIS
FOR THE NATIONAL HEALTH AND NUTRITION
EXAMINATION SURVEY, 1971-80

David P. Discher, M.D 2; Alan Palmer, PhD? ; Gregory Hibdon®; Terence A. Drizd, MSPHJI

INTRODUCTION

The wide acceptance of the Forced Expira-
tory Spirogram pulmonary function test in respi-
ratory epidemiologic studies is evidenced by
recent efforts of the Division of Lung Disease of
the National Heart, Lung and Blood Institute
and the American Thoracic Society to bring
greater precision to this important test.!»2 Spi-
rometry provides both medical practitioners and
epidemiologists with a simple yet objective
method of following the course of chronic ob-
--structive lung disease from its early inception to
its more advanced states, thereby permitting the
application of intervention measures and the
monitoring of results. Furthermore, epidemio-
logical studies can indicate early changes in func-
tion that can be related to various aspects of
environmental pollution thus permitting devel
opment of control strategies to mitigate further
degradation of function.

Unfortunately, spirometric testing is ham-
pered by a lack of sound and sensitive data ob-
tained from rigorous testing procedures on gen-
eral population groups. These data are necessary
for derivation of performance standards.

 

3Chairman, Department of Industrial and Environmental
Medicine, San Jose Medical Clinic.

bSenior Epidemiologist, Center for Community Health
Studies, Stanford Research Institute International.

CComputer Application Analyst, Stanford Research Insti-
tute International.

Statistician, Division of Health Examination Statistics,

National Center for Health Statistics.

The National Health and Nutrition Examina-
tion Survey of the National Center for Health
Statistics is the largest ongoing examination sur-
vey in the world. Thus the National Health and
Nutrition Examination Survey offers an oppor-
tunity to collect lung function data on various
population groups representative of all socioeco-
nomic groups, races, ages, sexes, and geographic
areas. Because additional data are also collected
on examinees that may be significant variables
for spirometric function, this survey will lead to
research on other variables.

Aware of the limitations of existing spirom-
etry data, the staff of the Division of Health Ex-
amination Statistics, which conducts the Na-
tional Health and Nutrition Examination Survey,
have undertaken an extensive review of the
existing spirometry data collection procedures
and computer processing program criteria to
ensure that data sensitivity is maximized.

This report details each of the steps taken to
ensure the collection of optimal data. An identi-
fication of the multiple source of variability
known to reduce the sensitivity of the data, a
description of the subsequent operating proce-
dures to minimize each of these sources of vari-
ance, a review of spirogram measurement criteria
as currently used in the National Health and
Nutrition Examination Survey program, and a
comparative analysis of various alternative algo-
rithms for increasing the accuracy of the meas-
urements are presented. The development of
alternative spirogram measurement techniques
was undertaken to further validate those tech-
niques suggested in the recent National Heart,
Lung and Blood Institute report! and, most im-
portant, to provide testable, documented logic
for the National Health and Nutrition Examina-
tion Survey (NHANES) criteria used in quality
control calibration, and measurement pro-
cedures.

These documented measurement criteria
should provide a foundation for the analysis of
current and future NHANES-collected data
from which new regression equations will be
developed for prediction on normative values.

BACKGROUND

Spirometry testing has been an integral part
of the National Health Examination Survey
(NHES) since 1963. During NHES Cycle II
(1963-65), spirometry data were obtained by a
Collins water-sealed spirometer, using the stand-
ard operating test procedures recommended in
the spirometer instruction manual. Generally,
technicians had little training in the theory and
physiological meaning of spirometry. Measure-
ments were made manually at great expense in
time and money, and the limitations of this level
of data collection became obvious? During
NHES Cycle III (1966-70), a spirometry testing
module that used computerized data collection
techniques was developed. Rigid standard oper-
ating procedures (SOP’s) were developed, and
concurrent technician and data surveillance pro-
grams were run to control for procedure and test
data variability. Data were analyzed using the
spirometry computer program*® developed by
the Public Health Service (PHS).

Further refinements were made in the spi-
rometry data collection module in 1970 before
the beginning of the National Health and Nutri-
tion Examination Survey (NHANES I). The data
acquisition hardware system that was used in
NHANES I to collect spirograms is described in
this report. Digital tape equipment was installed
to replace the analog data systems used in NHES
III, and general refinements of the SOP’s were
made to reflect the current methodology (e.g.,
the use of a standard set of five trials to ensure

maximal values). While NHANES I data were
being collected, the latest version of the PHS
computer spirometry program was reevaluated
and extensive program changes were made in
calibration and quality control procedures and
the logic used to define and compute the various
spirometric measurements. Recently, new cri-
teria have been developed and adopted by the
National Heart, Lung and Blood Institute
(NHLBI) to standardize the criteria used to
compute and analyze spirometric data in epi-
demiologic studies.] The NHLBI criteria are
comparable with those used by the NHANES
programs except in ‘“zero-time” and the “end-
of-test” computations. These methods are com-
pared and their strengths and weaknesses are
documented.

The initial discussion in this report relates to
nonsampling data errors that are caused by the
host of variables that the NHANES planning
group delineated as obstacles to collecting opti-
mal data. This discussion is followed by a de-
scription of the instrumentation and the quality
control programs that were developed to control
for these errors and of the test procedures used
during NHANES I. Finally an analysis of the test
procedures is presented and various alternative
methods for obtaining spirometric measure-
ments are compared.

Spirometry Data Variability

In establishing testing uniformity, the vari-
ables that must be considered include sélection
and training of technicians, testing techniques,
testing environment, spirometry equipment se-
lection, data measurement and computation,
and quality control.8:7 Each of these areas is a
potential cause of nonsampling error that di-
minishes or obscures any differences being
sought in epidemiological studies as well as the
validity of spirometry as a clinical-diagnostic
tool.

Examinee sources of variance.—Submaximal
expiratory effort during the performance of the
Forced Expiratory Spirogram (FES) is attribut-
able to a variety of factors. A common cause of
poor test data is failure of the subject to com-
prehend the test instructions; in children this
problem is often referred to as testing imma-
turity. This condition is a behavioral-social phe-
nomenon exemplified by a lack of school readi-
ness; the commands “Sit down and be quiet,”
“Raise your hand when you want to speak,”
“Pick up your pencil and copy the picture and
the words in your book” all require understand-
ing, willingess, and enough self-control for the
pupil to perform properly and effectively.
Older children and adults of various ethnic and
socioeconomic groups can also present problems
of language and comprehension, and these fre-
quently combine to frustrate meaningful data
collection. Examinees with such problems are
often performing the spirometry maneuver for
the first time and this situation, coupled with
anxiety regarding any medical procedure or its
implications, often results in an unacceptable
test despite the best efforts of the technician.

Technician sources of variance.—Spirometry
testing requires maximum subject participation
and an astute technician. Current practice dic-
tates that vigorous verbal encouragement be
given to the subject to stimulate maximal effort.
An experienced technician is a combination of
bully and cheerleader as he or she strives to elicit
this maximal response from the subject. The
technician must first explain the test, demon-
strate the procedure, cheer on or goad the
subject into putting forth his or her best effort,
and evaluate the degree of cooperation obtained.

The methods used to administer the test not
only vary from one technician to another, but
also vary from trial to trial with the same
subject. Not all technicians have equal abilities
to perform all tasks well. Some work well only
under supervision; if supervision is varied, tech-
nician performance also can vary.?

Any individual who is well motivated, inter-
ested, and reasonably intelligent and who has
the equivalent of a high school education can be
trained in spirometry.” Only 2 weeks of inten-
sive training are required to learn how to
administer the spirometry test, handle and cali-
brate the instruments, and perform the calcu-
lations. However, learning to obtain the best
possible performances from examinees of all
types and ages takes much longer—at least 6
months and perhaps a year. Such experience
develops the many approaches necessary to
instruct the examinee in a series of unfamiliar

maneuvers, such as, taking in the deepest breath
possible, inserting the mouthpiece and keeping
the lips tightly around it, and exhaling into the
spirometer as quickly, forcibly, and completely
as possible.

The most important quality of a pulmonary
function technician is the motivation to perform
the very best test on every examinee. Initial
enthusiasm after a while may turn into lack of
interest. The intellectual ability of the techni-
cian becomes particularily important in discern-
ing performance deficiencies of examinees and
correcting these errors in maneuver.

The qualifications of personnel being hired
to do spirometry are difficult to judge. This
process may be accomplished though a personal
interview with the prospective employee in
which previous and related work experiences are
reviewed and discussed. Each new technician
should be evaluated to determine the level of
training that will be required; and, if further
training is needed, it should be done under the
guidance of an experienced physician or pulmo-
nary physiologist in a laboratory where ample
testing is being performed with the highest
standards of accuracy and quality control.

Equipment sources of wvariance.—Spirom-
eters—much data are available on pulmonary
function sensors that point to a basic set of de-
sirable characteristics. The spirometers should be
accurate and precise, have linear volume and
flow rate response, be electronically (in elec-
tronic models) and pneumatically calibratable,
have a frequency response of the signal being
recorded (FES, 15 Hz), and have low inertia
without oscillatory fluctuations.?

Portability and compactness, although de-
sirable, should not be considered at the expense
of any of the preceding characteristics.

Automation.—Hand measurements of spiro-
metric data have been shown to be less precise
than automatic systems. Studies have shown
that when two trained pulmonary technicians
analyzed a number of spirograms, interobserver
differences were statistically significant.!0 Epi-
demiologic studies often require the combined
efforts of two or more observers for the study of
a large population; thus should one observer be
more precise than the other, the quality of the
better effort is diluted when the results are
pooled. The ability of measurements to discrimi-
nate between a normal and abnormal population
is vitiated under such circumstances.

The expenditures of time and people for
routine computations is no longer justifiable.
The use of automated techniques conserves
time, improves accuracy and precision, increases
work capacity, and reduces cost. Through these
means, the professional and technical staff be-
come free to pursue more challenging activities.

Testing Methodology

The need for calibration.—Although spirom-
etry equipment is extremely accurate, even the
best equipment requires both careful attention
and routine maintenance. For the electronic sig-
nals generated by moving the piston in the spi-
rometer to be related to known volumes and
known flows of air, the technician must perform
periodic calibration checks. To detect minor sig-
nal fluctuations between pneumatic and elec-
tronic calibrations, the technician must perform
a calibration as required by the SOP’s.11 Precise
adjustments of the equipment are made that
alter the volume-to-voltage relationship which
are based on observations that the technician
makes by using the pneumatic calibrations. The
technician becomes aware of the need for elec-
tronic service to the equipment when the elec-
tronic calibrations show wide fluctuations of the
standard electronic signal. Thus the first consid-
eration in obtaining valid data on forced expira-
tory maneuvers by electronic spirometry is an
understanding of the electronic principles inher-
ent in calibrations and maintenance.

The need for technician-examinee rapport.—
The second concept that the technician must
understand is the requirement that the forced
expiratory maneuver be correctly performed by
the subject under the close observation and
guidance of the technician. The technician can
enhance this communication by developing an
initial rapport, performing a good demonstration
of the maneuver, and clearly stating the standard
test instructions. The technician’s skill is mani-

fested by the subject’s comprehension of the

initial standard instructions, motivation to pro-
vide a maximal effort on a minimum of two of
the five expiratory trials, and correct notation of

procedural errors and redirection of test instruc-
tions accordingly. A number of barriers to a
successful test can be identified as follows:

® Testing immaturity—the subject cannot
follow directions.

® Inability to communicate—the subject
cannot speak the language or dialect of
the technician or any available
interpreter.

® Pain or disability—the subject cannot
take in a deep breath and/or rapidly
exhale down to full expiration.

® Voluntary refusal—the subject will not
participate because of fear or other
reasons.

Instructions to subjects, therefore, are stand-
ardized for the initial trials and follow standard
variations for subsequent trials depending on
observations of the technicians—observations
made by watching the subject perform the
maneuver and by monitoring oscilloscope dis-
plays of the flow and volume signals of all
completed trials for that subject. The skilled
technician quickly perceives difficulties from
these two sources and redirects the subject to
perform a correct maneuver. A number of
examinees tested in NHANES I did exhibit pain
or discomfort while performing the test or
indicated the presence of an upper respiratory
infection. With the assistance of the resident
physician, such subjects were disqualified from
the examination. Regarding those who refused
to take the test, their reasons were fully docu-
mented and will be examined for nonresponse
bias.

In summary, the test requires both a
technician-spirometer interaction to achieve ac-
curate and reliable signals and a technician-
subject interaction to achieve subject compre-
hension and motivation. These two interaction
areas define technician skill. A review of techni-
cian performance in the field, however, revealed
occasional drift in performance; therefore, re-
training procedures were routinely implemented
to reduce this source of error.

Spirometry data quality control.—The at-
tending technician is responsible for spirometric
data quality: Direct observations can be made
during the performance of the test and observed
errors can be corrected during the procedure.
Clearly, the technician has the cardinal role in
data quality control because he or she provides
clear and concise test instructions, coaches the
examinee to perform a maximal expiratory
maneuver, and provides an initial judgment of
the acceptability of the data obtained.

The technician can carry out this role by
proper use of the monitoring equipment and
careful observation of the subject. A memory
oscilloscope with an X-Y axis is regarded as a
reasonably precise tool for monitoring patient’s
spirometric effort. Flow is registered on the Y
(vertical) axis, and volume is measured on the X
(horizontal) axis. Each respiratory effort results
in a flow-volume curve, which is displayed on
the oscilloscope and compared with subsequent
curves (figure 1). The technician can thus
monitor discreet changes in patient effort and
cooperation by observing the shape of the curve
and the height of the peak flow deflection. This
monitoring information must be integrated with
subject performance observations. Appendix I is
a glossary of terms relating to this technician
function and includes a diagram of the three
phases in a normal spirogram trial (appendix
figure I).

The following paragraphs describe the cur-
rent criteria used by the NHANES technicians to
judge data quality.!!

Procedural error detection.—As a matter of
conscientious workmanship, a technician ex-
amines each trial within a test set during its
recording to identify the presence of any partic-
ular procedural error. Errors are a signal to the
technician that the examinee is experiencing
some problem with the test instructions either
because the instructions were unclear or compre-
hension was inadequate. When a procedural
error is identified, such as the absence of a ter-
minal decay curve (as seen on both the flow and
volume signal), the subject is reinstructed, with
emphasis on that part of the instruction where
the problem occurred, and a clear demonstration
of the test procedure is given.

The common procedural errors that alert the
technician to the possibility of an invalid trial
are described below.!! The best trials are those

with the largest forced vital capacity (FVC) ac-
companied by the highest flow rates. Procedures
for identifying the best trial are described within
the section entitled “Reliability Error Detec-
tion.”

Short baseline.—A short baseline can result
when the technician starts the recording equip-
ment too late, thereby not permitting establish-
ment of a sufficient baseline, or when the sub-
ject initiates expiration before instructed to do
so, thus obviating the baseline. A short baseline
cannot be observed on the flow volume display
but it is evident on the strip chart, as shown in
figure 2.

No end-of-test plateau.—During the test pro-
cedure, subjects who have large vital capacities
coupled with low terminal flow rates continue
to increase their expired volumes beyond the
preset recording time (9.19 seconds after the
technician initiates the NHANES I recording
system). This phenomenon typically occurs in
subjects with chronic obstructive lung disease
(COLD), although it can occur in subjects with
no known disease. The strip chart, not the visual
display, shows this phenomenon because the
former is a 9.19-second record whereas the latter
is a flow volume display that is independent of
time (figure 3). The recording equipment de-
scribed here does not have a manual override to
permit recording volumes beyond 9.19 seconds;
thus, the presence of a terminal flow is referred
to as premature termination by the recorder.

Premature termination artifact.—The prema-
ture termination artifact is manifested by the
absence of a typical phase III morphology of the
spirometric curve, that is, a slow decay curve
until residual volume is reached. Unlike a prema-
ture termination by the 9.19-second recorder,
this phenomenon occurs within 9.19 seconds
and is due to premature termination of the
effort by the subject; thus the phenomenon is
found on both the visual display and the strip
chart (figure 4).

Inhalation artifact.—Inhalation artifacts are
identified either by the flow-volume loop mor-
phology depicted in figure 5 or by review of
the flow signal on the recording paper and
observing that the flow rate decreases below the
baseline, which is followed by an increase of
flow greater than 1 liter per second (11 per
 

 

FLOW

¢
H
[4
4
!
4
i)

 

VOLUME

 

Figure 1. Typical subject flow-volume curve

 

 

VOLUME

 

 

 

TIME

 

Figure 2. Sample spirogram demonstrating the short baseline procedural error

 
 

 

 

VOLUME

 

 

 

TIME

 

 

Figure 3. Sample spirogram demonstrating the no end-of-test plateau procedural error

 

 

Sudden cessation
of flow rate

FLOW

 

 

VOLUME

 

 

‘FLOW

 

 

VOLUME

 

 

 

 

Figure 4. Sample oscilloscope tracing demonstrating the prema-
ture termination artifact procedural error

second). (These trials are automatically dis-
carded as totally invalid and are not considered
in a set of five.)

Venturi artifact.—The Venturi artifact is
evident when FVC volumes and/or flow rate
values are greater than clinically expected (fig-
ure 6). This phenomenon is caused by trumpet-
ing into the mouthpiece with pursed lips, which
causes room air to be drawn into the spirometer
along with the expired air (figure 7). This situ-
ation occurs because of a vacuum effect from
the high velocity of air movement from the
pursed lips. The typical morphology of a Ven-
turi trial is a rapid rise of flow rate to a high
level that is sustained until residual volume is

Figure 5. Sample oscilloscope tracing demonstrating the inhala-
tion artifact procedural error

attained, followed by a rapid decrease to the
zero line. Such an uncharacteristic trial is readily
identified by a trained technician by review of
the flow-volume display.

Because some members of the population
(such as highly trained athletes) have extra large
lung volumes and flow rates, large values can be
obtained without any artifact; however, caution
is required before accepting these readings.
Again, if reliability criteria are met, which
includes a careful review of the flow and volume
histories, the test is valid.

Low peak flow artifact.—Peak flow rates of
50 percent of predicted value are sought as a
measure of inital expiratory thrust. Lower values
 

WITH VENTURI

FLOW

 

 

 

VOLUME

 

FLOW

WITHOUT VENTURI

— es we 20 [/SBEONA fre cm ce cn cm  — t— — — — —

Normal

 

 

VOLUME

 

Figure 6. Sample oscilloscope tracings, one normal and one demonstrating the Venturi artifact procedural error.

 

  
   
  
 

THE VENTURI

Outside air drawn in

Spirometry tube

 

 

 

Figure 7. A depiction of the mechanics of a spirometric Venturi
artifact

would indicate the possibility of malingering,
not achieving total lung capacity before begin-
ning to blow, trouble understanding the test
instruction, or severe obstructive lung disease
(figure 8). This possible error check is discarded
if applied reliability criteria are met. No compu-
ter check is used for detecting this artifact:
Detection is left to the technician who must

observe both subject effort and peak flow esti-
mates on the monitoring equipment.

Hesitation artifact.—The hesitation artifact
should not occur during the three phases of the
spirogram. If it does occur, the test may be
considered acceptable only if the reliability
criteria have been fulfilled and flow and volume
histories are similar. This artifact (figure 9) is
generally identified by the technician, and com-
puter identification is limited to detection of a
relatively large hesitation only at phases II and
III.

Table A describes the output codes and
criteria used by the computer program to flag
the described procedural violations. :

Reliability error detection.—Acceptable
spirograms result in reproducible curves.l1,12
The technician makes an initial determination of
reliability by using the monitoring equipment to
superimpose one flow-volume curve over the
other or, alternatively, to compare them side by
side. At the conclusion of the fifth trial, the
technician also examines the paper record for
the two best trials. These trials are deemed
reproducible if the estimates of the FVC and
forced expiratory volume at 1 second (FEV, 4)
are within 5 percent, assuming that these vol-
umes exceed 31 or 10 percent for FVC and
FEV, , volumes of less than 3 I. If reproducibil-

 
 

 

FLOW

 

 

VOLUME

 

 

 

Figure 8. Sample oscilloscope tracing demonstrating the low
peak flow artifact procedural error

 

FLOW

 

 

 

VOLUME

 

 

 

Figure 9. Sample oscilloscope tracing demonstrating the hesita-

tion artifact procedural error

Table A. Procedural error codes and their definitions

 

 

Definition

 

 

Code
Oeicnsinmnvinirase No violations occurred.
Yoctionsecrisussnnse
D.ssreinirrie
BB srirasenmmaneres
8. iirrreminies
interval following EOP (inhalation artifact).
8 urrrsnrnierens
Bsc Computed FVC was less than 0.2 | (invalid trial).
TT ,
Bu iisvsmnriries
subject).

Onset of volume curve occurred less than 150 ms after the beginning of the record (short baseline).
End of trial (EOT) was not identified in the 9.19-second record (premature termination by recorder).

A volume increment of less than 4 percent between 0.5 second and 1 second after onset of the curve, or an increment
between 1 and 2 seconds less than 4 percent (midtrial premature termination by subject), occurred.

A negative flow occurred followed by post-EOT positive flows in excess of 50 ml per second over any 0.50-second

Peak flow was greater than 3 standard deviation units above subject's predicted peak flow (Venturi artifact).

Post-peak flow but pre-EOT signal showed a marked decrease (25 percent of peak flow) in flow for a time interval of
0.1 second or more and was followed by a marked increase (25 percent of peak flow) in flow (hesitation artifact).

The 0.50 second of a trial after EOT had a slope in excess of 50 ml/second (premature termination at end of trial by

 

ity cannot be demonstrated within that test set,
the five-trials test sequence is repeated after the
subject has rested.4»2

The need for technician monitoring and
surveillance. —Uniformity of testing procedures
was achieved in the NHANES by the use of
appropriate operational procedures, care in the
selection and training of technicians, and peri-
odic retraining.

Because data collection in NHANES I ex-
tended over a 5-year period, problems of drift in

technique were anticipated.® This drift was
overcome in part by a surveillance program in
which spirometry data obtained by each techni-
cian were periodically reviewed for trends in
procedural and reliability errors. From this
information, corrective actions were taken to
reduce the continued collection of technically
unsatisfactory data. This procedure was accomp-
lished by directly observing the technician as he
or she performed the testing in order to identify
possible errors in technique. One aspect of this
on-site surveillance was to compare the instruc-
tions given to the subjects with the standard
instructions shown in appendix II. Another
aspect was the on-site review of the subjects’
tracings to determine whether the technicians
could make accurate judgments from the record
and were able to correctly observe the flow-
volume loop.

INSTRUMENTATION

The instrumentation used in the NHANES
program to acquire and store the spirometry
signals in a format suitable for computer analysis
comprised an electronic spirometer, a storage
X-Y oscilloscope to display the flow-volume
curve for monitoring purposes, a single-channel
linear strip chart recorder to provide a perma-
nent record of the volume signals, and a data
acquisition unit to encode, convert, and record
on digital tape the spirometry volume signals.

Figure 10 is a schematic representation of the
system.

Spirometer and Support Electronics

Spirometry examinations were performed on
an Ohio Medical Instruments Corporation model
800 electronic spirometer. This spirometer dif-
fers from the more widely used volume displace-
ment “wet” system in that it consists of a dry
metal cylinder containing a plastic-faced piston.
A silastic rolling membrane forms an air-tight
seal between the piston and the cylinder. The
piston connecting rod is attached to a low-
voltage potentiometer, which varies a fixed
voltage signal in a linear manner proportional to
the piston displacement.

Expired air from the forced expiratory
breathing maneuver flows down the connecting
hose, into the spirometer, and displaces the
piston, causing the output of a signal from the
potentiometer. This signal is transferred to the

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Flow
Flow- :
Spirometer volume Storage Strip chart
converter oscilloscope recorder
Volume
T Volume 1 1
I~ _— — dene — — a—— — — — — m— o— —— 7]
A/D
converter
9-track
magnetic
tape
recorder
i Lead
selector
switch
Encoder
(EBCDIC)
12
selector
switches Digicorder |

 

 

 

 

 

Figure 10. Schematic of the NHANES spirometry system

10
flow-volume converter where the signal is fil-
tered, amplified, and is also differentiated to
generate the flow signal. The outputs from the
flow-volume converter are two signals of varying
voltages, one of which is directly proportional to
the amount of piston displacement (volume
signal) and the other directly proportional to the
rate of piston displacement (flow signal).

Calibrators

The spirometer is calibrated by means of an
internal volume pump operated by a small
electric motor that drives a single-lobed cam
through a gear reduction train. When the calibra-
tor yoke is attached to the connecting rod of the
spirometer piston, and when the electric motor
is engaged, the cam rider pushes the piston back
and forth, causing the in-and-out movement of a
known volume of room air at known flow rates.
When the output volume signal is recorded on a
paper tracing, as shown in figure 11, the known
air movement is represented by a graphic sinus-
oidal signal, with the trough-to-peak distance
representing the volume of air displaced. For
example, if the calibrator movement causes
5,000 milliliters (ml) of air to flow in and out of
the spirometer, the trough-to-peak distance on
any paper recording of the volume signal will
represent that 5,000 ml. The volume displace-
ment shown in figure 11 is representative of a
midrange calibration.

The Ohio spirometer electronics are preset
to convert 1 ml of volume to 1 millivolt (mV);
therefore, the trough-to-peak signal will be
recorded on the magnetic tape as a difference of
5,000 mV from the baseline voltage. Any
variation from the 5,000-mV calibration signal
thus indicates either a change in the calibrator or
a change in the volume-to-voltage ratio. For
preliminary data processing purposes, any such
observed change was assumed to be caused by
the latter. For example, if the mean trough-to-
peak difference was found to be 4,950 ml, a
calibration factor of 1.01 was used for the
subsequent spirometric analysis. Likewise, a
mean difference of 5,050 ml would produce a
calibration factor of 0.99.

Finally, a manual check of the program-
computed standard deviation is performed, and

the coefficient of variation is computed. It is
assumed that a coefficient of variation greater
than 3 percent indicates a daily variation great
enough to warrant the use of different calibra-
tion factors for different periods of testing.
Specifically, if the coefficient of variation is
greater than 3 percent (that is, £150 ml on a
5,000-ml calibration), hardware maintenance is
performed and the affected data set is manually
divided into smaller batches until the variation is
less than 3 percent; a different calibration factor
is computed manually for each batch (from the
list of trough-to-peak differences printed out by
the program). Spirometric analysis is performed
separately for each batch. If the coefficient of
variation is within the 3-percent limit, the data
on that tape are considered to be a single batch.
Before conducting a spirometry test on a
subject, the technician electronically calibrates
the spirometer through the use of the signal
generation capability of the flow-volume conver-
ter. This calibration involves switching the
volume-calibration switch from its normal “op-
erate” position to the zero position. The techni-
cian then switches between this position and a
+5,000 mV (d.c.) position. This switching back
and forth between 0 and +5,000 mV activates
transmission of a signal that displays graphically
as a square wave function, with the bottom step
representing 0 mV and the top step representing
+5,000 mV.
The difference between electronic and pneu-
matic calibrations follows. A pneumatic calibra-
tion involves all parts of the spirometry data
collection system—the mechanical action of the
spirometer piston, the mechanical and electronic
action of the potentiometer, the amplifier and
filtering circuits of the flow-volume converter,
the analog-to-digital (A/D) converter and filter-
ing circuits of the data acquisition unit, and the
nine-track tape recording device. Conversely, the
electronic calibration only tests the electronic
portions of the instrumentation system. Thus an
evaluation of the pneumatic calibration signal is
an evaluation of the accuracy of the entire data
collection system, whereas the evaluation of the
electronic calibration signal assists the examiner
in locating the source of any variation. For
example, should the pneumatic variation deviate
beyond certain preset limits on a given magnetic

1
 

10,000 —

 

 

EXPIRATORY VOLUME IN MILLILITERS

(4,500 mv) A

2,000 |—

 

0 1 2 3 a

 

 

ELAPSED TIME IN SECONDS

 

5 6 7 8 9 10

 

 

Figure 11. Spirometer data—calibration sine wave

tape, the troubleshooter would first examine the
electronic calibration output for that tape. If
this check revealed a consistent step-function
difference of 5,000 mV, it could be assumed
that the electronic portion of the system was
functioning normally and that the problem lay
in the pneumatic system (i.e., the spirometer
itself). Conversely, should the electronic calibra-
tion show a significant step-function difference
from 5,000 mV, it could be assumed that the
spirometer was functioning normally and that
the problem emanated from the electronic cir-
cuits somewhere in the line after the spirometer.
The pneumatic calibration procedure used in
NHANES I (1971-75) was not that which was
recommended by the American Thoracic Soci-
ety (ATS) in its Snowbird Standardization Proj-
ect.2 NHANES II (1976-80) practice did, how-
ever, follow that procedure.

12

Data Acquisition System

Spirometry data are recorded on a Beckman
Digicorder Model No. DRS-1000 digital tape
acquisition system. This unit encodes each signal
with a series of pulses entered by thumb
switches that the computer program identifies as
the recording location, subject identification
number, age, sex, race, height, technician code,
barometric pressure in millimeters of mercury,
and temperature in degrees Celsius. The com-
puter uses temperature and pressure to develop a
BTPS correction factor that adjusts volume from
ambient temperature and pressure saturated
with water vapor (ATPS) to body temperature
and pressure saturated with water vapor (BTPS).
A record of the machine identification number
is also encoded. This encoding permits each
spirogram to be traced to the machine it was
recorded on and a code (lead) number that

indicates to the computer whether the signal was :

a calibration or .an FES. Because this 14-lead
data acquisition system is also used to collect a
12-lead electrocardiogram on the same subject,
unique lead numbers are assigned to the spi-
rometry examination. Data are recorded on a
nine-track digital tape after conversion from
analog form via an A/D converter. The tape is
processed directly by a digital computer at a
later date.

SPIROMETRY DATA ANALYSIS:
PROGRAM DESCRIPTION

The analog spirometric signal is converted to
digital data and encoded by the Digicorder and
then recorded on a digital magnetic tape. Each
individual data record (subject trial, calibration,
etc.) consists of 18 digits representing the header
and identification information, followed by
4,599 data points representing voltages (the
spirometer volume curve). All data are recorded
at a rate of 500 samples per second. As
described below, the number of data points is
reduced by computer processing to 100 samples
per second, and each resulting data point has a
signal resolution of approximately 2 ml of
volume.

The automated computation of spirometer
trial parameters is performed in three stages. In
the first stage the Digicorder data tape is
unpacked (reformatted) and the calibration fac-
tor (which corrects voltage-to-volume ratios) to
be applied to the volume data is computed. In
the second stage the flow data are computed
from the volume data, the calibration and BTPS
factors are applied to the volume and flow data,
and the baseline is computed and removed. The
third stage is the computation of spirometric
parameters from the corrected data.

Calibration Factor Computation
(First Stage)

A calibration data record is recognized by
two conditions in the 18-digit header. The
number 14 must be found in the channel lead
indicator (digits 5 and 6) and at least nine 9’s
must be found in digits 7 through 18. The actual

calibration is a sinusoidal wave with the trough-
to-peak voltage difference corresponding to a 5-1
volume. By computing the average trough-to-
peak voltage, a ratio is formed (the calibration
factor), which is later used to scale all the
volume data.

When a calibration data record is found, the
sinusoidal wave data are first reduced from 500
samples per second to 100 samples per second
by a five-point average:

v, =(V + Vr + V2 * Vin+s + Vines W5s
where

m =5(n=-1)+1 and
n =the number of the averaged data point,
1-919.

This averaging reduces the data from 4,599
points per record (trial) to 919 points per
record. Once the data have been averaged, the
first differences (flows) are computed by the
relation

AV, =p * Vp

Because the sinusoidal curve may begin with
a positive or a negative flow, a starting point
must be determined. This point is located by ob-
serving the first negative flow with a voltage of
less than 4.5 volts (V) (i.e., a point from which
to start looking for the first trough). If no such
point is found, the record is ignored and the
next record is read. When the starting point is
located, a search is made for the first positive
flow. From this positive flow, the next 175 vol-
ume data points are retained (1.75 seconds;
because the sine wave period is 3 seconds, this
time will contain a minimum and maximum
voltage). The trough-to-peak difference is com-
puted and the next cycle is checked, beginning
with a positive flow (minimum).

The trough-to-peak differences are com-
puted cycle by cycle and record by record until
the end of the data is reached. If no pneumatic
calibration signals were found on the data tape,
processing is terminated. If one legitimate cali-
bration is found, the calibration factor is com-
puted (cal =5.0 I/D volts, where D is the average
trough-to-peak voltage difference).

13
Figure 11 shows a typical calibration sine
wave. At A, the first negative flow is encoun-
tered; however, the voltage level is above the
4,500-mV threshold. At A’, the first negative
flow with a voltage less than 4,500 mV is found.
The search for the next positive flow proceeds
to B, where the minimum threshold of 3,320
mV is recorded. The search then continues to C,
where the maximum of 8,510 mV is found. The
trough-to-peak difference (B to C) is computed
as 5,190 mV. A like difference is computed be-
tween D and E, and the average is 5,190 mV or
5.190 V. Thus the calibration factor is

5.0 liters
cal =

= e=— =(.9634 lit It.
5.190 volts rensfvo

Figure 12 shows the data taken from an actual
calibration trial where both the trough and peak
(B and C) were examined for stability of the sig-
nal. As shown in the figure, 9 data points were
recorded at 3,330, which preceded the trough,
and 9 similar data points were recorded, which

 

 

\

C
| 8,500 (n = 9)
| {ry
\ PEAK

:
.

4 BT = 16)
:

\ / 8,500 (n= 11) \

3,330 (n=9)

:

: TROUGH
3820—— (n=36)
:

 

 

3330(n=9)

 

 

 

Figure 12. Digital data array from a calibration curve showing
trough (B) and peak (C)

14

followed the trough; moreover, the trough volt-
age of 3,320 mV appeared as a continuous string
of 36 samples. The figure also shows a similar
stability at the peak end.

Volume and Flow-Rate Signal Data
Corrections (Second Stage)

The corrections applied to the spirometric
trial volume and flow rate data consist of a cali-
bration factor, a BTPS factor, a reduction of the
data sample rate from 500 samples per second to
100 samples per second, and the subtraction of
the baseline (from the volume data).

BTPS correction.—The BTPS factor is com-
puted by using the following formula:

(BP- PH,0) _ (310.16)

BTPS = -—
BP- 47.067 (t,)

where

BP = the barometric pressure (obtained
from the header data)
ty =the spirometer temperature in de-
grees Kelvin (derived from the
header data)
PH,0 =a temperature-dependent water va-
por pressure.

A combined correction factor is then com-
puted by multiplying the BTPS and calibration
factors.

Sample rate reduction.—The sample rate for
the 4,599 volume data points (voltages) is re-
duced from 500 to 100 samples per second by
the five-point averaging technique applied to the
calibration data. The volume data are then con-
verted from centivolts (cV) to liters, and the
combined correction factor is applied to the
averaged volume data. Finally, the flow rates are
computed by taking the first differences of the
volume data as was done with the calibration
curve.

Baseline removal. —Because a volume of zero
liters is not generally represented by a zero-
voltage signal from the spirometer, a baseline
must be determined and removed from the vol-
ume data. This baseline is defined as the average
value of the points preceding the estimated be-
ginning of the trial.
A flow threshold of 1 1 per second, plus a
noise tolerance, is used to estimate the beginning
of a trial. (The noise tolerance is defined as 30,
where 0 is the standard deviation of all baseline
data being processed and is determined as a sepa-
rate computation.) The flow threshold is based
on the minimum step size in the spirometer sig-
nal—which is 1 ¢V or 10 mV. With a combined
correction factor of 1, this method would con-
vert 10 ml at 500 samples per second. When the
sample rate reduction is performed (five-point
averaging), the minimum step size would be re-
duced to 2 ml. At a sample rate of 100 samples
per second, a volume change of 2 ml would pro-
duce a flow rate of 0.2 1 per second. The 1-1-per-
second threshold allows for a small deviation of
the baseline above the 0.2-l-per-second minimum
step size. The noise tolerance is used to increase
the size of the flow threshold if the baseline data
are noisy. Therefore, the first flow rate to
exceed the flow threshold marks the end of the
baseline. (This initial estimate of zero time is
refined during the third stage). The baseline
digits are then averaged with the weighted aver-
age technique at the net volume point, Avg,
= (Avg,.;.* Vol, )/2. Once the average baseline
volume has been determined, it is subtracted
from all volume data to remove the recorder
bias. If the baseline is less than 15 points long
(150 milliseconds (ms) worth of data), the trial
is rejected and the data quality code is set to 1,
indicating a short baseline.

Spirometer Trial Parameter
Computation (Third Stage)

Peak flow determination.—After the volume
and flow curves have been corrected, the flow
data are searched and the largest value is re-
corded as the peak flow, with the corresponding
volume. Predicted peaks are computed on the
basis of the following formulas:

Predicted peak flow for males =~-1.0028
+ (0.0474 X age) + (0.2150 X height)

Predicted peak flow for females ==0.5532
+ (~0.0331 X age) + (0.1493 X height),

where peak flow is in liters per second, age in
years, and height in inches. The data quality

code is set to 5 (Venturi artifact) if the observed
peak exceeds the predicted peak by at least 3.10
(where: 0 male = 1.9585, and 0 female =
1.3821).

Zero time.—Once the peak flow has been de-
termined, the zero-time (beginning of trial) esti-
mate can be refined (figure 13). This method for
determining zero time therefore replaces the ini-
tial estimate derived in the second stage. Using a
triangular method for the flow curve, the zero
time is corrected by the following equation:

 

fie V peak
0 peak F peak >
where
ty = zero time
tpeak = time of peak (referenced to first
guess zero time)
Vpeak = volume at peak flow
Fea = the peak flow rate

with the additional constraint that the corrected
to not precede the first guess zero time.
End-of-trial determination and FVC calcula-
tion.— After the beginning of trial is determined,
the end of the trial (EOT) must be found. The
EOT is found by a two-step process. First, the
volume data are searched for a plateau. The

 

Peak flow

    

Volume at peak

FLOW RATE ————»
2 =»
if

 

TIME roms:

 

 

 

Figure 13. Schematic of flow-volume curve showing the rela-
tion of zero time to time of peak flow

15
plateau is said to be reached when, starting with
the zero-time volume and comparing at every
10th point (every 0.1 second), the volume has
not increased from the previous 0.1-second
point. When a plateau is found, the time of the
earliest point is recorded as the first guess of
EOT, and the corresponding volume is recorded
as the first-guess FVC. If a 10-point plateau is
found, a search is made from this first-guess
EOT to the end of the volume data (919 data
points or 9.19 seconds) for the maximum vol-
ume. Current recommendations? are for a mini-
mum signal duration of 10 seconds; however,
design of this system preceded the development
of these recommendations and the current sys-
tem in use conforms to the 10-second duration
of signal. If no volume is found larger than the
first-guess FVC, the previously recorded FVC
and EOT are used, and the data quality code is
set to zero. If a larger volume is found, a search
is made of the flow data between the first-guess
FVC and the maximum volume. If a negative
flow rate is encountered between the first-guess
FVC and the higher maximum volume, the EOT
is defined as the point just prior to the negative
flow and the corresponding value is recorded as
FVC. A negative flow is defined as any 0.01-
second flow rate less than zero when the base-
line o is zero; when 0 is not equal to zero, nega-
tive flow is equal to the noise tolerance (or
minus 30). If no intervening negative flows are
found, the maximum value is defined as the
FVC, and the corresponding time is recorded as
EOT.

A hesitation artifact (procedural error 7
in table A) is reported if a marked decrease in
flow occurs after the peak flow for a 0.1-second
interval, which is followed by a marked increase
in flow. If EOT is found after the 10-point pla-
teau, a check must still be made for premature
termination at the end of trial. This check is
done by examining the average flow rate during
the 0.50-second period preceding EOT. If the
average flow rate exceeds 50 ml per second, the
data quality code is set to 8 (premature termina-
tion at EOT) and no further processing is per-
~ formed on that trial. If the EOT is at 9.19 sec-
onds and the average flow rate exceeds 50 ml
per second, quality control code 2 (premature
termination by recorder) is set and no further
processing of the trial is performed. If no prema-

16

ture termination is found, the entire trial be-
tween zero time and EOT is searched for inhala-
tion artifacts (negative flow rates greater than
noise tolerance). If any are found, the data
quality flag is set to 4 (inhalation artifact), but
the processing continues on that trial.

Calculation of other parameters and quality
control checks.—Once the beginning and ending
of the trial have been defined and the peak flow
and FVC have been determined, the other trial
parameters can be computed. The forced expira-
tory flow rates at 25, 50, and 75 percent of FVC
(FEFy5q , FEF5q , and FEF, ) are computed
from the FVC. The volume data are then
searched for a forced expiratory volume of at
least 0.2 1. If none is found, the trial is declared
invalid (procedural error 6 in table A) and no
further processing is carried out. If a volume ex-
ceeding 0.2 1 is found, the corresponding time is
found by linearly interpolating between that
volume and the previous one. The same pro-
cedure is followed to determine the time for an
FEV of 1.21. The FEF,40.4 200 m1° is then com-
puted as: :

(Viz ~ Voz)

FEF =
200-1,200 ml Cro ~taa)

In a like manner, the times are determined
for FEV’s at 1, 2, 3, 4, 5, and 6 1. Any volumes
that are not reached have their corresponding
times set to 99.99. The flow rates are also re-
corded at the times of the various FEV values.
Finally, any FEV’s that exceed the FVC,4q
have their corresponding flow rates set to 99.99.

The times and flows for 25, 50 and 75 per-
cent of FVC are determined by locating the first
volume that exceeds that value and recording
the corresponding time and flow rates. The
FEF,; 755 measurement (maximum midexpira-
tory flow rate (MMEF)) is then computed as:

FVC,5q J. FVCysq
FEFas75% = tise = 25%

€Forced expiratory flow rate between 200 and
1,200 ml on the volume curve (FEF200.1,200), form-
erly known as the maximum expiratory flow rate
(MEFR).
 

 

 

 

 

 

 

 

 

 

 

 

SUBJECT NUMBER 12-345 SEX MALE AGE=41.YEARS HEIGHT=70. INCHES WEIGHT=178.LBS. TECH. NO. 2
TRIAL* VOLUME(L) TIME(SEC) FLOW(L/S) * TIME(SEC) VOLUME(L) FLOW(L/S) * TIME(SEC) VOLUME(L) FLOW(L/S)*
* 0. .02 10.31 * 1/4 2.32 6.02 *PEAK  .03 .31 12.46 *
* 1.0 .09 10.74 * 1/2 3.38 3.22 » .10 1.43 8.81 *
* 1.2 11 10.74 * 3/4 3.93 1.29 * .50 3.47 2.58 *
*- 2.0 .21 6.66 * 1.0 4.23 99.99 * 1.0 4.26 99.99. *
* 3.0 .40 3.65 * 2.0 4.75 99.99 i 2.0 4.76 99.99 *
1 * 4.0 .80 99.99 * 3.0 4.97 99.99 * 3.0 4.97 99.99 *
* 5.0 3.18 99.99 * 4.0 5.06 99.99 * 4.0 5.06 99.93 *
* 6.0 99.99 99.99
* .25FVC 11 9.24 - ZERO TIME= 1.54 END TIME= 5.04 TIME OF FVC= 3.50
* .50FVC .29 4.73 - FVC= 5.06 MEFR= 11.48 MMEF= 4.43 RELIABILITY CODE(S)= 8 7
* .75FVC .67 1.50 -
* 0.2 .01 11.82 * 1/4 2.28 5.16 *PEAK  .03 .40 11.82 *
* 1.0 .09 10.31 * 12 3.26 3.22 » .10 1.42 8.38 *
* 1.2 11 9.45 * 3/4 3.88 1.72 * .50 3.35 2.79 *
* 2.0 .21 6.66 * 1.0 4.22 99.99 * 1.0 4.25 99.99 *
* 3.0 .43 3.44 * 2.0 4.75 99.99 * 2.0 4.75 99.99 *
2 * 4.0 .83 99.99 * 3.0 4.95 99.99 * 3.0 4.95 99.99 *
* 5.0 3.32 99.99 * 4.0 5.06 99.99 * 4.0 5.06 99. =
* 6.0 99.99 99.99
* .25FVC 11 8.38 - ZERO TIME= 1.06 END TIME= 4.76 TIME OF FVC= 3.70
* .50FVC 31 4.08 - FVC= 5.06 MEFR= 10.45 MMEF= 4.15 RELIABILITY CODE(S)= 8 7
* .75FVC a1 1.93 -
* 0.2 .01 11.82 * 14 2.21 6.02 *PEAK  .04 .50 12.25 *
* 1.8 .09 9.02 > 1/2 3.27 3.01 * .10 1.46 7.52 *
* 1.2 11 9.02 * 3/4 3.90 2.15 * .50 3.39 2.19 *
* 2.0 22 6.45 * 1.0 4.26 99.99 * 1.0 4.30 99.99 *
* 3.0 .43 3.44 * 2.0 4.81 99.99 > 2.0 4.82 99.99 *
3 * 4.0 .80 99.99 * 3.0 5.03 99.99 * 3.0 5.04 99.99 *
* 5.0 2.84 99.99 * 4.0 5.06 99.99 te 4.0 5.06 99.99 *
* 6.0 99.99 99.99
* .25FVC 12 7.95 - ZERO TIME= .88 END TIME= 3.98 TIME OF FVC= 3.10
* .50FVC 31 4.51 - FVC= 5.06 MEFR= 10.27 MMEF= 4.27 RELIABILITY CODE(S)= 8 7
* L75FVC .70 1.93 -
*. 0.2 .01 12.28 * 1/4 2.28 5.37 *PEAK  .04 .52 12.68 *
* 1.0 .09 9.67 > 4/2 3.30 3.44 * .10 1.51 8.81 *
* ‘1.2 .10 8.81 * 3/4 3.93 1.93 * .50 3.41 2.79 *
* 2.0 .21 6.88 * 1.0 4.31 99.99 x 1.0 4.35 99.99 *
* 3.0 .41 4.08 *. 2.0 4.83 99.99 * 2.0 4.83 99.99 *
4 * 4.0 .79 99.99 * 3.0 5.03 99.99 > 3.0 5.04 99.99 *
* 5.0 2.83 99.99 * 4.0 5.06 99.99 * 4.0 5.06 99.99 *
* 6.0 99.99 99.99 = eceeeccecemeccecceeeeeeceeee
* .25FVC 1 8.38 ZERO TIME= .93 END TIME= 4.03 TIME OF FVC= 3.10
* .50FVC .30 4.73 - FVC= 5.06 MEFR= 10.89 MMEF= 4.27 RELIABILITY CODE(S)= 8 7
* 75FVC .69 1.93 -
* 0.2 .02 11.82 * 1/4 2.14 6.02 *PEAK  .03 .33 12.89 *
* 1.0 .10 8.81 * 172 3.2 3.22 * .10 1.33 7.52 *
* 1.2 .11 8.59 * 3/4 3.88 1.72 * .50 3.32 3.22 *
* 2.0 +23 6.45 * 1.0 4.27 99.99 * 1.0 4.30 99.99 *
* 3.0 .44 3.22 * 2.0 4.81 99.99 * 2.0 4.83 99.99  *
5 * 4.0 .82 99.99 * 3.0 4.96 99.99 * 3.0 4.96 99.99 *
* 50 99.99 99.99 * 4.0 4.96 99.99 * 4. 4.96 99.99 *
*.6.0 99.99 99.99 = mmmeemeeemeeeeeecemeeeeeo —
* ,25FVC 12 8.59 - ZERO TIME= 1.36 END TIME= 4.16 TIME OF FVC= 2.80
*° .50FVC .31 5.37 - FVC= 4.96 MEFR= 10.89 MMEF= 4.44 RELIABILITY CODE(S)= 8 7
* _I5FVC .68 2.36 -

 

Ll

Figure 14. Data output sheet for a normal subject showing a set of 5 spirograms

 
8l

 

 

 

 

 

 

 

 

 

 

 

 

 

SUBJECT NUMBER 12-345 SEX MALE AGE=41. YEARS HEIGHT=70. INCHES WEIGHT=178.LBS. TECH. NO. 2
TRIAL* VOLUME(L) TIME(SEC) FLOW(L/S) * TIME(SEC) VOLUME(L) FLOW(L/S) *  TIME(SEC) VOLUME(L) FLOA(L/S)*
* 0.2 .07 1.72 * 1/4 .55 1.72 *PEAK  .06 .18 2.79 +
* 1.0 .57 1.50 + 12 .92 ‘ * .10 .39 1.72 *
* 1.2 75 .86 * 3/4 1.20 .85 * .50 1.00 86 *
* 2.0 2.41 99.99 * 1.0 1.40 .86 * 1.0 1.44 86 +
* 3.0 99.99 99.99 * 2.0 1.90 99.99 * 2.0" 1.91 99.99 *
1 * 4.0 99.99 99.99 * 3.0 2.12 99.99 * 3.0 2.13 99.99 *
* 5.0 99.99 99.99 * 4.0 2.17 99.99 * 4.0 2.17 99.99 *
* 6.0 99.99 99.99 -- eee me em me mmm en
* .25FVC .25 1.72 - ZERO TIME= 3.47 END TIME= 6.87 TIME OF FVC= 3.40
* .50FVC .64 1.07 - FVC= 2.17 MEFR= 1.46 MMEF= 97 RELIABILITY CODE(S)= 8 7
* .75FVC 1.36 43 -
* 0.2 .08 1.72 * 1/4 51 1.93 *PEAK 05 .14 2.79 *
* 1.0 .62 1.07 + 2 .86 1.50 * 10 .34 1.07 *
* 1.2 .82 .86 * 3/4 1.14 .86 * 50 92 1.50 =
* 2.0 2.53 99.99 * 1.0 1.33 1.97 * 1.0 1.37 86 *
* 3.0 99.99 99.99 * 2.0 1.85 99.99 * 2.0 1.86 99.99 *
2 * 4.0 99.99 99.99 * 3.0 2.10 99.99 * 3.0 2.11 99.99 *
* 5.0 99.99 99.99 * 4.0 2.16 99.99 * 4.0 2.16 99.99 +
* 6.0 99.99 99.99
* L25FVC .26 1.72 - ZERO TIME= 2.01 END TIME= 5.41 TIME OF FVC= 3.40
* .50FVC .69 .86 - FVC= 2.16 MEFR= 1.34 MMEF= 87 RELIABILITY CODE(S)= 8 7
* L75FVC 1.50 .64 -
* 0.2 .07 1.72 * 1/4 .52 1.93 *PEAK  .05 .15 2.79 *
* 1.0 .63 .86 * 1/2 .85 1.50 * .10 .35 1.93 *
* 12 .83 .86 * 3/4 1.13 1.07 * .50 .91 1.50 *
* 2.0 2.38 99.99 * 10 1.34 1.07 * 1.0 1.38 2»
* 3.0 99.99 99.99 * 2.0 1.86 99.99 * 2.0 1.89 99.99 +
3 * a0 99.99 99.99 * 3.0 2.13 99.99 * 3.0 2.14 99.99 *
* 5.0 99.99 99.99 * 4.0 2.18 99.99 * 4.0 2.18 99.99 *
* 6.0 99.99 99.90 momen eee omens
* .25FVC .26 1.07 - ZERO TIME= 1.47 END TIME= 4.77 TIME OF FVC= 3.30
* .50FVC 72 .86 - FVC= 2.18 MEFR= 1.31 MMEF= 89 RELIABILITY CODE(S)= 8 7
* 75FVC 1.48 43 -
* 0.2 .08 1.93 * 1/4 .52 1.72 *PEAK  .02 .08 3.40 *
* 1.0 .61 .86 * 1/2 .87 1.50 * .10 .29 2.15 *
* 1.2 .80 .86 * 3/4 1.16 .21 * .50 .90 86 *
* 2.0 2.35 99.99 * 1.0 1.36 86 * 1.0 1.38 64 *
* 3.0 99.99 99.99 * 2.0 1.88 99.99 * 2.0 1.89 99.99 *
4 + 4.0 99.99 99.99 * 3.0 2.15 99.99 * 3.0 2.15 99.99 *
* 5.0 99.99 99.99 * 4.0 2.28 99.99 * 4.0 2.28 99.99 +
* 6.0 99.99 Es nn
* L25FVC .28 1.50 - ZERO TIME= 1.61 END TIME- 5.81 TIME OF FVC= 4.20
* L50FVC .75 .21 - FVC= 2.31 MEFR= 1.38 MMEF= 84 RELIABILITY CODE(S)= 8 7
x LTSFVC 1.65 .21 -
* 0.2 .08 1.72 * 1/4 72 .86 PEAK .31 79 2.58 *
* 1.0 .51 .86 * 1/2 1.00 .86 * .10 .90 1.50 *
+.1.2 71 86 x 3/4 1.24 .86 * .50 1.29 86 *
* 2.0 2.27 99.99 = 1.0 1.42 .43 * 1.0 1.60 64 *
* 3.0 99.99 99.99 * 2.0 1.92 99.99 * 2.0 2.01 99.99 *
5 * 4.0 99.99 99.99 * 3.0 2.17 99.99 * 3.0 2.18 99.99 *
* 5.0 99.99 99.99 * 4.0 2.18 99.99 * 4.0 2.18 99.99 *
* 6.0 99.99 99.00 meee
* L25FV 13 1.50 - ZERO TIME= 3.06 END TIME= 6.16 TIME OF FVC= 3.10
* 50FVC .60 .86 - FVC= 2.18 MEFR= 1.11 MMEF= 87 RELIABILITY CODE(S)= 8 7
* L75FVC 1.37 .64 -

 

Figure 15. Data output sheet for an abnormal subject showing a set of 5 spirograms

 
The volumes and flow rates are recorded at the
following time points: 0.25, 0.50, 0.75, 1.00,
2.00, 3.00, and 4.00 seconds from zero time.
For the times 1, 2, 3, and 4 seconds, if the cor-
responding volume is less than FVC,zq , the
corresponding flows are set to 99.99. Using the
peak time as the reference time, volumes and
flows are recorded similarly for 0.1, 0.5, 1.0,
2.0, 3.0, and 4.0 seconds after peak flow. Again,
whenever the volume is greater than FVC, 4 ,
the corresponding flow rate is set to 99.99.

A final data quality check is performed using
the FEV’s at 0.5, 1.0, and 2.0 seconds from zero
time (FEV, 5, FEV, , FEV,). If FEV, is not at
least 4 percent larger than FEV, , or if FEV, ,
is not at least 4 percent larger than FEV, 4, the
data quality code is set to 3 (midtrial premature
termination artifact).

Figures 14 and 15 are examples of five trials
of normal and abnormal subject data,
respectively.

SPIROMETRY DATA ANALYSIS:
VALIDITY OF ALTERNATIVE
ALGORITHMS

Methodology

Five separate analyses were performed to
evaluate the accuracy, consistency, and validity
of the logic selected for calculating five spiro-
metric parameters: zero time, EOT, and the
three most commonly used ventilatory para-
meters (FVC, FEV, ,, and FEF, ;54). The
calculation of the latter three parameters de-
pends directly on the determination of the first
two, and in this section the ventilatory param-
eters are used to evaluate the performance of
various algorithms for those determinations,
both in the presence and absence of electronic
or physical noise in the volume signal. The
algorithms that were chosen as best, and the
subsequent calculation of FVC, FEV, ,, and
FEFy5 759, are described in the previous
section entitled ‘“Spirometer Trial Parameter
Computation (Third Stage)”.

The first analysis consisted of comparing
calculations obtained on 19 trials; the compari-

sons for each of the five parameters were based
on three independent measurements:

1. Computer methods as described in the
previous section

2. Manual calculations by technician no. 1

3. Manual calculations by technician no. 2.

Differences between the computer-derived pa-
rameters and each of the manually derived val-
ues were obtained, as well as differences between
the values obtained by the two technicians on
the 19 trials. The manual calculations were
obtained from curves plotted from the same
digital data that were introduced into the
computer program and were adjusted by the
correction factor after subtracting the baseline
and averaging to 100 four-figure (nearest milli-
liter) digits each second. Figure 16 shows 1 1 and
1 second measuring 11.4 and 16.8 millimeters
(mm), respectively. This figure also shows vol-
ume and time-base sensitivities reasonably close
to recommended minimums? of 10 and 20 mm,
respectively.

The second analysis was a comparison of
computer-derived parameters in which several
alternative algorithms were used. As indicated in
table B,f four computer methods were devel-
oped for determining zero time and four were
developed for EOT identification. The initial
methods for determining zero time and EOT
referred to in the preceding section, entitled
“Spirometer Trial Parameter Computation
(Third Stage),” constitute methods 1 in table B;
moreover, the definitive zero time and EOT time
given in that section are methods 2 and 3, re-
spectively. The extrapolation method (method
3) for zero time differs only slightly from the
triangular method (method 2) in that the former
assumes that flow from time zero to the time
when peak flow occurs is equal to the peak flow
rate, whereas the latter method assumes that
flow averages one-half of peak flow during this
short time interval and that the flow increases in

fTables 1-17 showing the results of these analyses
are grouped at the end of this report, preceding appen-
dix I.

19
 

10,000 —

8,000 —

4000 |—

EXPIRATORY VOLUME IN MILLILITERS

2,000 —

 

0 1 2 3 4

 

ELAPSED TIME IN SECONDS

 

 

Figure 16. Example of a normal time-volume spirogram

a linear manner between zero time and time

when peak flow occurs. Both methods are rela-
tively easy to use in graphic analysis of spiro-
grams, as well as in computer analysis. The ex-
trapolation method for zero time has been rec-
ommended previously,13-15 and the volume
threshold (method 4) for determining zero time
also has been considered previously.!? The selec-
tion of 30 ml as the threshold (see table B) was
based on the assumption that one can read
graphic records within 0.5 mm with reasonable
accuracy and that for most spirograms this incre-
ment would be no less than 30 ml when ampli-
tude is converted from millimeters to milliliters.
Thus zero-time comparisons include both algo-
rithms given in the above-mentioned section plus
two similar methods that have been examined

20

before. The methods to determine EOT time
also consisted of two methods described in the
above-mentioned section plus method 2 (slope
threshold method) previously recommended!
for determining EOT time and method 4 (max-
imum volume method), which disregards any
negative or zero-flow events from zero time to
EOT time.

The third analysis was a comparison of these
algorithms when a noise signal at each of two
levels of amplitude was superimposed on the 19
trials. The first noise signal was a sine wave with
an average amplitude of 2 cV (0.02 V) superim-
posed on the original 500-samples-per-second
digitized volume signal. The sine wavelength was
0.017 seconds (60 hertz (Hz)). This superim-
posed noise did not significantly increase the
114

Table B. Methods for zero-time and end-of-test determinations

 

 

 

Method number and name Curve used Critical value(s) Method description
ZERO TIME (zg)
Method 1, flow threshold............ccceeeveeeieerrneeernnenns Flow Flow > (1 | per second +t), | Search flows from beginning of data (baseline) every 0.01
where t is noise tolerance second. Record tg as time when criteria are met: per-
(liters per second) cent of flow > 1 + t in liters per second.
Method 2, triangular (triangular extrapolation) ........ Flow and Flow peak (largest value on | Determine peak flow from flow data every 0.01 second;
volume flow curve) compute tg based on following formula:
Volume at peak
t to =t — 2 V peak | where flow peak is largest value
peak 0 = Ipeak Fisk pe 9
on flow curve, V peak is corresponding volume, and
tpeak is corresponding time to nearest 0.01 second.
Method 3, extrapolation (rectangular Flow and Flow peak Determine peak flow from flow data every 0.01 second,
extrapolation)................ npsesnsssns ARS SSR RRS SSR SR volume Volume peak and compute tq based on following formula:
peak
tQ = tpeak — Y PeaK where flow peak is largest value
F peak
on flow curve, V peak is corresponding volume, and
tpeak is corresponding time to nearest 0.01 second.

Method 4, volume threshold...........cccccveeerirrnneeeennnnee Volume Volume => 30 ml Search volume every 0.01 second for criteria to be met.

Record tg as first volume to equal or exceed 30 ml.
END-OF-TIME (tgoT)

Method 1, 10-point plateau ............cciersrsrsrinsssrssens Volume (Volj+10 - vol}) <0 Starting with tg, compare volumes every 0.1 second
(10th point). Record tgQT as time of vol; when
(volj+10 — volj) <0.

Method 2, slope threshold.........ccccceeeeviererinnennnnnnnnnns Volume (Volj+50 — volj) < 25 mi Starting with tg, compare volumes every 0.01 second
with an interval of 0.5 second (50th point). Record
teQT as time (corresponding to vol;) when
(volj+50 — volj) < 25 ml.

Method 3, negative fIOW..........cceeveeeeereeiiininnnennsennnns Flow Flow (-t) liters per second Starting at tg, compare flows every 0.01 second. Record

where t is noise tolerance tEQT as time corresponding to flow; _ 1, when
(liters per second) flow; < —t.
Method 4, maximum volume...........cccceeueueennnecrnrennes Volume Maximum (volume) Starting at tg, compare volumes every 0.01 second.

 

 

 

Record tgQT as time corresponding to volmax.

 
 

 

EXPIRATORY VOLUME IN MILLILITERS EXPIRATORY VOLUME IN MILLILTERS

EXPIRATORY VOLUME IN MILLILITERS

8,000

6,000

4,000

2,000

8,000

6,000

4,000

2,000

8,000

6,000

4,000

2,000

 

NORMAL SPIROGRAM~NO NOISE

 

 

 

ELAPSED TIME IN SECONDS

NORMAL SPIROGRAM — 2 cV OF NOISE

 

ft

 

 

1 2 3 4 5 6 7 8 9

ELAPSED TIME IN SECONDS

FLOW RATE IN LITERS/SECOND

FLOW RATE IN LITERS/SECOND

NORMAL SPIROGRAM-~4 cV OF RANDOM NOISE

 

 

ELAPSED TIME IN SECONDS

FLOW RATE IN LITERS/SECOND

15

10

 

 

©
N
>

EXPIRATORY VOLUME IN LITERS

 

 

 

0 2 4
EXPIRATORY VOLUME IN LITERS
0 2 4

EXPIRATORY VOLUME IN LITERS

 

22

Figure 17. Normal spirogram (time-volume and flow-volume curves) with noise signals superimposed

 
calculations for either the flow threshold for
zero time (method 1) or EOT time (method 4);
however, the noise tolerance (t) was equal to 11
per second when the second noise signal was
superimposed—a 4-cV amplitude random sine
wave with the same wavelength. Thus the flow
threshold for #9 was defined as 41 per second
with the 4-cV noise and the cutoff for a
significant negative flow was -31 per second
with the 4-cV noise when using method 3 for
tgot. Figure 17 shows the results of two levels
of noise on a spirometry signal displayed as a
time-volume curve and as a flow-volume curve.
Note the increased visual sensitivity of the
flow-volume representation to detect noise.
Because the 19 trials were obtained from
four well-trained subjects with normal ventila-
tory parameters, the second and third analytic
routines were also applied to a set of abnormal
spirograms that were derived from the 19
normal spirograms by computer manipulation of
the time and volume variables. Therefore, the
fourth analysis was a comparison of algorithms
for abnormal spirograms with superimposed
noise signals. Figure 18 shows an example of a
time-volume and flow-volume representation of
an abnormal spirogram with superimposed noise.
The electronic spirometry system was care-
fully adjusted to present as noise-free a signal as

possible. This entire baseline signal was ex-
amined by visually reviewing a computer listing
of all 19 trials; when no noise signals were
observed, it was concluded that the spirometry
signal was noise free. This conclusion that the
signal was relatively noise free should be under-
stood in terms of the necessary 2-ml threshold
for a trigger of one bit evident on the computer
listing of the 500-signals-per-second data. (This
one-bit trigger was described in the section
entitled “Baseline Removal.”) The computer
listing showed a constant 1,050 digital voltage
reading for all 44,000 digits examined at 500
samples per second over a baseline signal totaling
almost 30 seconds. All four persons used in the
tests were symptom free, and each entered into
a brief training session on forced expiratory
maneuver. Each of the four persons performed
the maneuver correctly and no trials were
rejected because of subject error or poor motiva-
tion. With an excellent or relatively noise-free
signal and carefully trained and apparently
healthy subjects, one would expect to find few
quality control flags and computed values simi-
lar to manual values, especially when using
rigorous algorithms for zero time and EOT.
Alternative algorithms might be more suitable
for “abnormal” subjects tested on equipment
with a random-noise problem. For example,

 

 

 

FLOW RATE IN LITERS/SECOND

 

-

 

 

8000 —
w

oc

w

E

3 6000 (—

-

ul

=

2

g

2 4000 (—

4

o

>

>

S

E200

«

g

8

VLA
0 1 2 3 4 5 6 7
ELAPSED TIME IN SECONDS

 

EXPIRATORY VOLUME IN LITERS

 

 

Figure 18. Abnormal spirogram (time-volume and flow-volume curves) with noise signals superimposed

23
when calculating zero time, a 1,000-ml-per-
second flow threshold (method 1) was used
initially. Zero time was defined as the interval
preceding the first nonzero flow by examining
each 0.01-second interval. This determination
implies that a volume difference of 10 ml or
more adjusted for BTPS would be the first
nonzero volume when comparing successive vol-
ume signals (5 volume signals averaged to 100
signals per second). When a noise signal is
superimposed on a noise-free signal, this flow
threshold might be expected to be too rigorous,
and an extrapolation or volume method might
be preferred. Thus four alternatives, including
the triangular method (method 2) described
previously, were compared to determine zero
time and the impact of noise signals. Similarly,
various end-of-trial alternatives were compared
to a rigorous negative flow algorithm (method
3), as indicated in the following section. The
four end-of-trial alternatives included in this
analysis include the two approaches described in
the section entitled “End-of-trial determination
and FVC calculation.”

Zero-Time and FEV, , Calculations

Zero time was initially identified as the first
0.01-second signal achieving or exceeding a
1-l-per-second flow. For manual readings, the
technicians were asked to identify the first
detectable departure from baseline. The techni-
cians then measured the 1-second time and the
volume at this specific elapsed time. Both time
and volume measurements were attempted to
the nearest 0.1 ml, recognizing that the fractions

of a millimeter were only rough estimates. The

computer labeled the zero-time signal as the
0.01 second that was previous to the flow
threshold signal and identified the FEV, 4
volume as the 100th baseline-corrected digital
signal after zero time.

For the first trial (figure 16), the computer
calculated a value of 1.50 seconds for zero time,
indicating that the first significant departure
from baseline occurred with digitized flow signal
151, located on the 100-sample-per-second array
that had been derived by averaging the previous
500-sample-per-second array. Technician no. 1
examined the visual display of the same 100-

24

‘sample-per-second digits and estimated that the

first significant departure from baseline had
occurred at a point measured as 1.52 seconds, or
25.5 mm of baseline from the initiation of a
signal on the plot. Technician no. 2 estimated
this point at 1.56 seconds, or 26.2 mm of
baseline. The paired differences of zero time for
this first trial were calculated as follows:

® Computer minus technician no. 1 = 0.02
seconds

® Computer minus technician no. 2 = 0.06
seconds

® Technician no. 1 minus technician no. 2
= (0.04 seconds

The average paired differences and standard
deviations of the 19 differences for the above
comparisons are given in table 1. The conclusion
was that the two technicians had sets of zero
times the means of which were not statistically
different when compared as paired differences;
however, technician no. 2 identified a mean zero
time at an elapsed time significantly later com-
pared with the computer-calculated mean zero
time.

The FEV, o manual calculation began from
the manually identified zero time, and then the
volume was estimated at 1 second from this
point. If the manual readings were slightly
biased to late zero-time estimates compared with
the computer, the expectation would be that
manual FEV, , values would be slightly greater
than the computer values. The mean computer
FEV, , value for 19 trials was 3.974 1 (standard
deviation = 0.471), whereas technician no. 1 and
technician no. 2 had measured mean values of
4.002 and 3.9771, respectively. These paired
mean differences were not significantly differ-
ent. Although the manual calculations by one
technician disagreed slightly with the zero time
obtained by the computer program, both techni-
cians were in remarkable agreement (with an
average difference of less than 30 ml) with the
computer-calculated FEV, , for 19 trials, and
both technicians agreed on FEV, , within a
mean of 30 ml. Note that 30 ml is considerably
more than one bit (2 ml) when the computer is
used to calculate the FEV, ,, but only about
one-third of a millimeter when volume is meas-
ured from the visual display.

The second test compared the computer-
derived values for FEV, ; when three alterna-
tive methods for determining zero time were
used (see table B for description of methods). As
just indicated for trial no. 1, the zero-time value
for the flow threshold, as defined above, was
1.50 seconds of elapsed time of baseline signal;
if the triangular method had been used, it would
have placed zero time at 1.52 seconds. More-
over, by the extrapolation method, zero time
would have been at 1.54 seconds, and by the
30-ml volume threshold method it would have
been 1.53 seconds. Table 2 presents mean
differences of zero-time values for all 19 trials.
Thus the extrapolation method (method 3)
resulted in a significant shift toward later times,
but no significant differences were found for the
triangular and the volume threshold methods
when compared with the threshold method. It
was predicted that the extrapolation method
would result in excessive mean FEV, , values
for the 19 trials when compared with the other
three methods. Table 3 presents the mean
FEV, , results. The paired mean FEV, , differ-
ences and standard deviations were calculated;
significant differences from the flow threshold
method were obtained for both the volume
threshold (p <0.01) and the extrapolation
(p <0.05) methods, as indicated in table 4.
This first comparison of methods indicates that
(1) zero time and FEV, , are in remarkable
agreement for the flow threshold method and
the triangular method as well as in agreement
with manual calculations, (2) the volume thresh-
old is of intermediate agreement with the flow
threshold method, and (3) the extrapolation
method would appear to yield a sizable bias for
trial estimates of zero-time and FEV, , testing
for subjects with normal spirograms and a
noise-free spirometry system. Figure 19 illus-
trates each of the four zero-time computations,
showing a representative delay in zero time that
results from the extrapolation method. As indi-
cated from data in table 4 this delay results in an
average error in FEV, , in the range of 1-2
percent.

After the four methods were compared by
using the same 19 signals, a third set of
comparisons with a noise signal superimposed on
the test signal was performed. For trial no. 1,
table 5 gives the shifts in FEV, , that occurred
with a 2-cV amplitude sine wave (slight noise),
and with a 4-cV random-amplitude sine wave
(greater noise). No significant shift occurred in
the mean FEV, , for 19 trials with increasing
noise when the flow threshold method was used;
that is, the no-noise FEV; , was 3.974, the
slight noise FEV, , was 3.971, and the greater
noise FEV, , was 3.979. These differences were
considered negligible when the 3.974 value for
the noise-free trial was used as the basis for
comparison; the extrapolation method (method
3) showed significant effects on FEV, ,, as
table 6 indicates. Methods 1, 2, and 4 were
reasonably similar to one another when the
mean-paired differences were used as a basis for
comparison. A close examination of the spirom-
etry curves showed that the noise signal often
increased the magnitude of peak flow and
delayed its occurrence; thus the extrapolation
method would yield increased FEV, values.

The fourth test was an attempt to have the
computer create abnormal curves from the 19
trials by mathematically reducing the volumes at
any one time point by first multiplying the
volume by 0.5 and then extending the time
required for each expired volume by multiplying
the time scale by 3.0, as figure 20 shows. For
example, the FEV, , for the first trial was
reduced as follows:

® Normal—Trial no. 1 computer-calculated
FEV, ( =4.191

® Abnormal—-Trial no. 1
calculated FEV,  =1.331

computer-

The mean of 18 of the 19 values from the
transformed trials was 1.251 1. (By performing
this mathematical transformation, the baseline
was also extended in time; in the process, trial
no. 7 was rejected because the baseline was too
short. Analysis of 18 abnormal trials was under-
taken, however, because the loss of trial no. 7

25
 

to,
Volume
threshold
method

toy

 
  
 

method

to,
Triangular
method

to,
Flow

threshold se—————
method

 

 

   

Extrapolation

  
  

 

 

 

 

 

 

NOTE: The upper curve is a time-volume plot with a superimposed line extrapolating a tangent from the point of inflection to estimate zero time by the extrapolation method (method 3). Also evident in the insert
is the ease of estimating 7g by the triangular method (method 2) by simply doubling the time from peak flow. The upper plot is a volume-time data display; the lower plot is a flow-time display. From the lower plot
one can see the close proximity of fq by the triangular method to that of the flow threshold method, an intermediate position of the volume threshold method, and the delayed zero time of the extrapolation method.

 

 

 

 

 

Figure 19. Manual calculation of to showing a representation of the 4 alternative methods

would have no bearing on our inferences.) The
zero-time and FEV, , calculations used for
these 18 trials again were based on the flow
threshold method, and the outcome was com-
pared with the other methods. Table 7 gives the

26

three mean paired differences and standard
deviation for zero time and FEV, , by using the
results of the noise-free signal as a basis for
comparison. Noteworthy differences, in the
range of an error of 3 percent for FEV, ,, were
 

 

 

 

 

10,000 —
8,000 —
«
«
w
-
oS
3
= 6,000 —
=
Zz
w
2
2
=
Oo
>
>
@
e
< 4,000 —
«
a
x
w
2,000 —
0
0 1 2 3 4 5 [} 7 9

ELAPSED TIME IN SECONDS

 

Figure 20. Example of an abnormal time-volume spirogram

found when the flow threshold results were
compared with those obtained by direct extrap-
olation from peak flow (method 3); however,
the triangular method of extrapolation from
peak flow (method 2) yielded results similar to
the flow threshold results. The volume threshold
method (method 4) was compared with the flow
threshold method and produced paired differ-
ences that were statistically significant but were
not in the error range as noted above for the
extrapolation method.

In the fifth and final test, the introduction
of 2- and 4-cV sine wave random noise had no
consistent effect on zero time compared with
the noise-free flow threshold method. Table 8
presents the results. The shift in zero time was a
consistent delay in the ¢;3 from that obtained on
the noise-free signal when the flow threshold
method was used as the index for these abnor-

mal spirograms. The extrapolation method
(method 3) continued to show the greatest
paired difference when compared with the flow
threshold index; moreover, at both noise levels
the volume threshold method yielded a smaller
mean difference for zero time when compared
with the triangular method of extrapolation
from peak flow. The flow threshold method
showed an interesting zero time shift with noise,
as follows:

® A'—Flow threshold method with noise-
free signal minus flow threshold method
with 2-¢V noise =-0.076.

® A" —Flow threshold method with noise-
free signal minus flow threshold method
with 4-cV noise = +0.046.

27

 
Table 9 shows I .w the shift in zero time
obtained by various alternative methods and
superimposed noise signals combine to yield
errors in FEV, ,, by using again the mean
FEV, o from 18 abnormal spirograms obtained
by the flow threshold method and a noise-free
signal as the index for comparison. The paired
differences are expressed in table 9 both in liters
and as a percentage of the index mean FEV, ,.
The only method that showed an error in excess
of 5 percent was the extrapolation method
(method 3), which overestimated FEV, in the
range of 8-9 percent. The conclusion was that
although normal spirograms analyzed by the
extrapolation method would result in an error of
1-2 percent, much greater errors are found when
abnormal spirograms with or without noise are
analyzed with this method. The only method
that showed an average error in FEV, , less
than 3 percent was the volume threshold
method (method 4). The flow threshold method
applied to abnormal spirograms showed a much
greater mean FEV, , change with noise (-38
and +46 ml for the two noise signals given in
table 9) compared with the data given previously
for normal spirograms and the same two noise
signals (-3 and +15 ml). By comparing the
results in table 6 with those in table 9, the
greater impact of noise on the FEV, , of
abnormal spirograms is evident for all four
methods. In table 6, all except the extrapolation
method yielded errors less than 1 percent of the
mean noise-free FEV, , by the flow threshold
method, although the errors shown in table 9 are
8-9 percent for the extrapolation method com-
pared with approximately 2-4 percent for the
other three methods.

In both the normal and abnormal spirograms
in the presence of 4 cV of noise, the volume
method (4) is affected least since it more closely
approximates the threshold reference measure-
ment (method 1) than methods 2 or 3. A close
examination of the spirometry data reveals that
the additional sensitivity achieved by using the
triangular method is jeopardized by the need for
a clear peak flow signal. This problem is severely
increased in the presence of abnormal data
where flow rates are relatively low and peak
flow rates are not well defined. (Usually the
flow curve demonstrates several points in close

28

proximity to each other in the area around the
peak flow.) In many cases, noise caused flow
points following the peak flow to be elevated to
a greater amplitude than the true peak flow rate,
which caused the time of peak flow rate to
occur later. This development generally resulted
in the zero-time point occurring to the right of
the threshold reference point. Method 4, which
does not rely on any flow rate data to identify
zero time, was relatively unaffected by the high
noise levels. This finding would suggest that in a
relatively noise-free system (<2 cV of noise),
method 2 is superior to all other methods,
although in a system with considerable noise (4
cV or more) method 4 would be the preferred
technique. Computerized spirometry data acqui-
sition systems with noise levels similar to that
simulated in this study generally should not be
used to collect data because measurement accu-
racy would be severely compromised.

End-of-Trial, FVC, and FEF, ;5
Calculations

The same five tests were performed for the
other end of the spirogram and the volume of
FVC. Because fractions of FVC are used as the
first steps to determine FEFy5,54 volume
points, the calculation of FVC is critical to the
logic for FEF 5 450

The mean value for the initial estimate (10-
point method—method 1) of end-of-trial time
and the mean FVC were both considerably dif-
ferent from the mean values for end of trial and
FVC obtained by the more rigorous negative
flow method (method 3)# This difference is
shown in table 10. A significant difference be-
tween these two methods is also evident for

"FEF 35.759 . A smaller FVC due to early termina-

tion would be expected to result in a much
higher FEFy5 755 because of a shift of this slope
toward the left where flows tend to be greater
on normal spirograms (table 10). The two tech-
nicians were asked to identify the FVC plateau
and mark the beginning of this plateau as the
end of test. By using this approach, the two
technicians next calculated FVC and FEF 5 754
for each of the 19 trials. The end-of-trial time

gRefer to table B for a description of methods.
and the FVC values were in remarkable agree-
ment (see table 11); however, technician no. 2
had a mean FEFg5.,55 that was significantly
higher than that of technician no. 1. The data
collected by each technician were initially com-
pared with the 19 values obtained by the more
rigorous negative flow method that was applied
for each of the three measurements. The mean
paired difference results are given in table 12.
Technician no. 1 reported FEF, ;54 measure-
ments that were significantly lower than those
obtained by the computer algorithm consisting
of the negative flow method; otherwise, no note-
worthy differences were observed.

Two alternative methods to determine the
EOT were next compared with the negative flow
and the 10-point plateau methods. The first
alternative was to consider the EOT to be the
end point of the first 0.5-second interval where
the average flow over this 0.5-second interval
was equal to or less than 50 ml per second
(method 2); the second alternative (method 4)
was the same as the first, but extended the EOT
to a larger value whenever the maximal FVC
value. occurred at a later point (irrespective of
any intervening negative flows). Figure 21 illus-
trates each of the four EOT computations.

Data previously presented on the 19 trials
in table 13 show the means and standard devia-
tions for the four methods plus the results ob-
tained by an average of the two manual measure-
ments for each trial. The calculations by the two
technicians agreed with the results from both
the negative flow and the maximal volume meth-
ods, but were very different from the results ob-
tained with the other two methods. It was con-
cluded that a choice between method 3 and
method 4 would be difficult. This conclusion is
supported by comparisons of these two methods
with the average manual measurements. Table
14 provides differences of less than 1 percent for
FVC and approximately 1 percent for
FEF, 759 . This comparison suggested that the
means of the average of manual readings are ap-
proximately midway between the results of the
two computer methods, perhaps somewhat
closer to the maximum volume method. The
mean FVC by the technicians was 15 ml more
than that determined by the negative flow
method and 12 ml less than that from the maxi-

mum volume method; in addition, the mean
FEFy5 59 obtained by the two technicians was
0.054 1 per second higher (+1.4 percent) for the
negative flow method, but only 0.018 I per sec-
ond lower for the maximum volume method.
These differences between method 3 and
method 4 and the average manual readings were
small and not statistically significant. In a sec-
ond analysis, the paired differences from the
two methods and the data from each technician
were compared separately (table 15). The sec-
ond comparison also showed no statistically
significant differences except for FEFg5.75%,
where three of the four differences were signifi-
cant (p <0.05), the only exception being the
0.080-l-per-second difference of the negative
flow method.

The next analysis compared the four end-of-
trial methods as related to the effect of the two
random noise signals that were superimposed on
the normal spirogram. Table 16 gives these re-
sults for EOT, FVC, and FEF; 555 by compar-
ing the mean paired differences for the noise
signal obtained by the negative flow method
with the other three methods. The effect on the
three parameters of the 2-cV noise was to
shorten the signal, reduce the FVC by a prema-
ture EOT, and increase the FEFgg 754 . For the
4-cV random noise signal, the trend was to pro-
long the EOT time, increase the FVC, and de-
crease the FEF, 555 for the negative flow and
maximum volume methods; but the trends were
inconsistent.

For the final analysis, abnormal spirograms
were created and noise signals were superim-
posed, as described previously. Only 15 of the
19 trials could be interpreted at the tail end of
the curve for all noise levels because of the lack
of a plateau at the end of the trial when noise
was superimposed. Table 17 shows the mean
paired differences.

Again, the 2-cV noise signals produced con-
sistent effects of reducing EOT and FVC and of
increasing FEF 754 ; similarly, the 4-cV signal
prominently showed relatively small reductions
in FVC and FEF5 555 when the negative flow
and maximum volume methods were used.
These two methods most closely simulated the
technicians’ measurements; the changes in the
FVC, when noise was added to the signal, were

29
 

 

(A) METHOD 1

 

10-point plateau FVCand EOT
identified here
FVCand EOT
Volume

 

 

 

 

 

 

|

 

     
 

(B) METHOD 2
EOT
see FvC
4430, <25ml
0.5 second

ecovoceo od
@eoe0cevccec oo

 

 

 

 

 

 

 

 

Time
8,000 —
0
@
w
EE 6,000 [—
-
3
g
Hi
Zz ’
w
E 4000 [—
2
>
>
S
=
= 2000 |—
*
uw
; J..d
0 1 | 3
(C) METHOD 3
Eve
Negative flow
Volume
|, }
y ; | Time
EOT
Flow
rate
Ll
[A Volume

 

 

 

NOTES: :
Method 1 shows the FVC and EOT identified as the first point of a 10-point plateau.

Method 2 shows the FVC and EOT as the first point within a 0.5-second window where the volume increase is less than 25 ml.

4
ELAPSED TIME IN SECONDS

Method 3 shows the FVC and EOT as the first point prior to the first negative flow of 0.01 | per second.
Method 4 incorporates methods 2 and 4 criteria but extends EOT and FVC to a larger value whenever the highest point on the volume curve occurred at a later point thereafter, irrespective of any

intervening negative flows.

 

7 8 9
(D) METHOD 4
FVC
ee
tees <25ml
0.5 second

 

NOTE: Method 2 criteria superseded if higher volume found regardless of
presence of any intervening negative flows.

 

 

 

 

30

Figure 21. Manual calculation of end of trial showing the 4 alternative methods

 
all less than 3 percent when compared with the
noise-free computed values; and for FEFy5 59
the changes were under 5 percent.

The two methods appeared to have a reason-
able accuracy when compared with technician-
measured EOT time, FVC, and FEFg5 754 . In
examining the graphs of individual trials where
the results of the negative flow method differed
from those of the maximum volume method,
locations on the spirographic curve were ob-
served where an instantaneous negative flow
could have occurred, and in many of these in-
stances the technician may have noted a termi-
nation hesitation. The recognition of instantane-
ous negative flows is beyond the resolution of
the human eye; however, if a negative flow con-
tinues over a duration of about 0.25 seconds,
such an event would likely be identified as an
end of test. In an effort to test this hypothesis a
fifth method for EOT was developed, which
identified the first point in time followed by an
average negative flow of 0.25-second duration.
The analysis is not given in this report; however,
it showed no appreciable accuracy advantage
with abnormal spirograms in the presence of
noise. Another possible method not yet tested
that may yield a more accurate estimate of
relatively low FVC and FEFg4 554 in the pres-
ence of noise would be a protracted negative
flow of 0.25 second. Therefore, either of two
methods (negative flow or maximum volume)
are recommended with the reservation that
neither are very accurate under conditions of a
high noise-signal ratio. The slope threshold
method recommended in this report is different
from recently published recommendations.! 2
In this report the spirometry signal was 9.19 sec-
onds in duration rather than the recommended
minimum of 10 seconds; however, conclusions
were not affected because all trials of these
“normal” persons were terminated within the
9.19-second signal duration when the slope
threshold was used. One subject had a terminal
flow pattern suggesting that if the signal dura-
tion would have been extended to 10 or more
seconds, methods 3 and 4 may have produced
FVC values even larger than reported herein and
the differences with method 2 would have been
even greater.

SUMMARY AND CONCLUSIONS

Because FEV, ; and FVC are the most im-
portant ventilatory parameters, three important
considerations were: (1) that these two param-
eters be estimated accurately by computer meth-
ods when compared with manual calculations;
(2) that the measurements be reproducible with
the superimposition of noise in the signal; and
(3) that the sequence of computational steps
proceed in a logical way, including the determi-
nation of significant digits used in the adjust-
ment for baseline signals, BTPS conditions, and
calibration information.

With a logical computation of spirometry
signals, FEV, , depends most on the accuracy
and consistency of zero time, and FVC depends
most on the accuracy and consistency of the
EOT time. For zero time, any of three methods
(methods 1, 2, and 4) appeared satisfactory;
however, the extrapolation method (method 3)
led to excessive FEV, , values even without
superimposed noise or when subjects had abnor-
mal or low values. Moreover, when the latter
conditions did pertain, the error of method 3
was shown to be in an 8-9-percent range. A de-
fense of the use of the triangular method
(method 2) as the definitive method used in the
NHANES program can be made because some
subjects have hesitation patterns on forced expi-
ration as do certain spirometers with high iner-
tia; both would seem to require some form of
linear extrapolation from a peak flow. Method 2
was shown to be essentially equivalent to meth-
ods 1 and 4, and the triangular extrapolation can
be as easily adapted to manual spirometry as the
unacceptable peak flow back extrapolation
method. One can calculate method 2 zero time
by doubling the time from peak flow to method
3 zero time as shown in figure 19. Figure 22
provides a summary analysis of the comparative
zero-time algorithms used in determining the
FEV, , in both normal and abnormal spiro-
metric data, with and without noise superim-
posed on the signal. The means and standard
errors of paired differences from the recom-
mended algorithm (triangular, method 2) are
presented as the index in this figure rather than
method 1.

31
 

 

NORMAL DATA-NO NOISE

Flow threshold
Triangle
Extrapolation

Volume threshold

NORMAL DATA-2cV OF NOISE

Flow threshold
Triangle
Extrapolation

Volume threshold

NORMAL DATA—4 cV OF NOISE
Flow threshold

Triangle

Extrapolation

Volume threshold

ABNORMAL DATA-NO NOISE
Flow threshold

Triangle

Extrapolation

Volume threshold

ABNORMAL DATA-2 cV OF NOISE
Flow threshold

Triangle

Extrapolation

Volume threshold

ABNORMAL DATA-4 cV OF NOISE

Flow threshold
Triangle
Extrapolation
Volume threshold

 

 

 

s—

 

 

TT
®

 

-—
<

 

 

 

 

 

—

 

 

 

 

 

2

 

 

 

-0.2 0.0 0.2

MEAN DEVIATION IN LITERS

04 06

 

Figure 22. FEV, ( analysis—means and standard deviations (30) of paired differences from triangular extrapolation (method 2)

32

measurements

 
 

 

NORMAL DATA-NO NOISE

 

P=

10-point plateau ; °
Slope threshold

 

a

Negative flow b

Maximum volume br

NORMAL DATA-2 cV OF NOISE

 

10-point plateau I
Slope threshold |

Negative flow fs
be

Maximum volume

 

NORMAL DATA-4 cV OF NOISE

10-point plateau

 

T
<

Slope threshold lL -

 

Negative flow °

Maximum volume bp

 

ABNORMAL DATA-NO NOISE

10-point plateau F—eo—]
Slope threshold mee]

Negative flow

Maximum volume I

ABNORMAL DATA-2cV OF NOISE

10-point plateau |

Slope threshold me
bs
.

 

Negative flow

Maximum volume

11+

ABNORMAL DATA-4 cV OF NOISE

10-point plateau

 

 

Slope threshold | —
|
1

Negative flow

 

——

 

Maximum volume I

®

 

 

 

 

 

 

 

 

 

-1.6 -1.2 -0.8 -04 0.0 04 08 1.2 16

MEAN DEVIATION IN LITERS

 

Figure 23. FVC analysis—means and standard deviations (30) of paired differences from negative flow (method 3) measurements

 
Similarly, the slope threshold method
(method 2) for determining the end of trial was
shown to produce an unacceptable bias in the
FVC by lowering FVC by more than 50 ml com-
pared with the negative flow method (method 3)
or the maximum volume method (method 4).
When either of the latter methods was examined,
the FVC accuracy was within 3 percent of man-
ual readings, even with a modest superimposed
noise and with abnormal trials with FVC values
in the range of 2 1. Figure 23 provides a sum-
mary analysis of the comparative EOT algo-
rithms for both normal and abnormal data, with
and without noise present on the signal. The
means and standard errors of paired differences
from the recommended algorithm (negative
flow, method 3) are presented.

The procedures used to collect spirometric
data during NHANES 1 represent the state-of-
the-art at that time; further improvements have
been made for NHANES II, reflecting advances
in available instrumentation, such as recording
of flow and volume data on sensitive strip-chart
recorders. In the future, further refinements are
anticipated by use of on-line computerized tech-

niques that judge data quality and reliability and
by use of terminal displays showing data trends
to assist the attending technician with his or her
acceptance decisions.

The computer program criteria described
using the recommended algorithms for determi-
nation of zero time (method 2) and end-of-test
detection (method 4) represent the state-of-the-
art to date, superseding those recommendations
given in the current NHLBI standards.! (These
two algorithms, as described in the section en-
titled “End of trial determination and FVC
calculation,” have been implemented for cur-
rent analyses.) More work is required on the
development of quality control criteria to pre-
clude the acceptance of questionable data and
on development of algorithms to determine the
best trial. Regarding the latter issue, this pro-
gram applies the criteria specified by the ATS
recommendations. However, more sensitive pro-
cedures must be explored, such as those
suggested by Discher and Palmer where a 10-
parameter procedure was used to determine the
optimal total curve.l?

000
REFERENCES

1 Epidemiology standardization project: III. Recom-
mended standardized procedures for pulmonary testing.
Amer. Rev. Resp. Dis. 118:55, 1978.

2Report on Snowbird Work Shop on Standardization
of Spirometry. Amer. Thor. Soc. 3:20, 1977.

3National Center for Health Statistics: Methodologic
problems in children’s spirometry. Vital and Health
Statistics. Series 2-No. 72. DHEW Pub. No. (PHS) 78-
1346. Public Health Service. Washington. U.S. Govern-
ment Printing Office, Nov. 1977.

4Rosner, S. W., Palmer, A., Ward, S. A., Abraham, S.,
and Caceres, C. A.: Clinical spirometry using computer
techniques. Amer. Rev. Resp. Dis. 94:187, 1966.

5Ayers, W. R., Ward, D. A., Weihrer, A. et al:
Description of a computer program of the forced expira-

' tory spirogram. Part II. Validation. Comp. Biomed. Res.
2:220, 1969.

6National Center for Health Statistics: Quality con-
trol and measurement of non-sampling error in the
Health Interview Survey. Vital and Health Statistics.
Series 2-No. 54. DHEW Pub. No. (HSM) 73-1328. Health
Services and Mental Health Administration. Washington.
U.S. Government Printing Office, Mar. 1973.

7 National Tuberculosis and Respiratory Disease Asso-
ciation: Chronic Respiratory Disease Screening Manual.
New York. National Tuberculosis and Respiratory Dis-
ease Association, 1968.

8National Center for Health Statistics: Forced vital
capacity of children 6-11 years, United States. Vital and

Health Statistics. Series 11-No. 164. DHEW Pub. No.
(PHS) 78-1651. Public Health Service. Washington. U.S.
Government Printing Office, Feb. 1978.

9Discher, D. P., Massey, F. J., and Hallett, W. Y.:
Quality evaluation and control methods in computer
assisted screening. Arch. Environ. Health 19:323, 1969.

10 Rosner, S. W., Abraham, S., and Caceres, C. A.: Ob-
server variation in spirometry. Dis. Chest 48:265, 1965.

11National Center for Health Statistics: HANES Ex-
amination Staff Procedures Manual for the Health and
Nutrition Examination Survey, 1971-1973. Part 15a.
Public Health Service. Washington. U.S. Government
Printing Service, June 1972.

12Dayman, H. G.: Flow control during forced expira-
tion. Amer. Rev. Resp. Dis. 95:887, 1967.

13Smith, A. A., and Gaensler, E. A.: Timing of forced
expiratory volume in one second. Amer. Rev. Resp. Dis.
112:882, 1975.

14Kanner, R. E. and Morris, A. H.: Clinical Pulmonary
Function Testing. Salt Lake City. Intermountain Tho-
racic Society, 1975. p. 1.

15Knudson, R. J., Slatin, R. C., Lebowitz, M. D., and
Burrows, B.: The maximal expiratory flow-volume
curve. Normal standards, variability, and effects of age.
Amer. Rev. Resp. Dis. 113:587, 1976.

16Discher, D. P., and Palmer, A.: Development of a
new motivational spirometer—Rationale for hardware
and software. J. Occ. Med. 14:9, Sept. 1972.

35
10.

11

12.

13.

14.

15.

16.

17.

36

LIST OF DETAILED TABLES

. Mean, standard deviation, and significance level for paired differences between zero-time methods on 19 trials for method 1

ant) Manual MBINOU, BY 2 t80IIMICIBNG. ......cconierrissrmsmrerssscrratiss ss tirsas tras siss sSEasss HAT a sRIs SAT 91 2402450 TARAS SERIAL LASERS pms ib amma

Mean, mean paired difference, standard deviation of differences, and significance level for zero-time measurements, by 4
MBLIVOOS OF COMPUTBEION ON 1D TIBIS...coiisiriimmmnssmsinisssmmsansmsssnssvss ssossnnsssrssnseantons ses hss aa vR ATE ARISE AFA APS VERRY RIAH A TEA as RR REARS

Mean forced expiratory volume at 1 second (FEV g) for 19 trials, by 4 zero-time methods.............cceeevueverrereeeerenrennereeenenes

Mean paired difference, standard deviation of differences, and significance level for forced expiratory volume at 1 second
(FEV. 0}, DY ZBIOAIMS MBINOM DEINE... cocverssrssissssniirissisisisinmmmusissssss posse bests sos tases aes a se nosvs sss sits pms Soasans soap esaptsses apes ns sons

Mean forced expiratory volume at 1 second (FEV go), by 3 levels of noise and zero-time method...........cceceereerueccnnuenenerunanns

Mean paired difference for forced expiratory volume at 1 second (FEV g), by 3 levels of noise and zero-time method

 

Mean paired difference, standard deviation of differences, and significance level for zero time and forced expiratory volume
at 1 second (FEV g), by zero-time method pairs on 18 abnormal SPIrOGrams.............e.ceceeeeereeeneesersesrerassessesessessesessesassassessens

Mean paired difference between zero-time methods, by 3 levels of noise for 18 abnormal spirograms............cccceeevevermmmrereeeeennns

Mean paired difference and percent difference for forced expiratory volume at 1 second (FEV g) between zero-time meth-
ods; by 3 1evelsiof noise Tor 18. aDN0NING) SDITOUAINIS. cs irepisserssssiissmmsssssssnmmsssssassssesskaspobsissunsssissbastisvatmmsss saps vas EA URREI SERRATE

Mean and standard deviation for end-of-trial time, forced vital capacity (FVC), and forced expiratory flow rate between 25
and 75 percent of the FVC (FEF 25.759) with mean paired difference, standard deviation of differences, and significance
level, by 2 methods of detecting the end Of tri@l...........ciiiiiieiiiiiiiiiiieteeerccreer ee cree cere eere ease sraeae essere ee es ssaae seen aeaaeassssssanssannnen

Mean and standard deviation for end-of-trial time, forced vital capacity (FVC), and forced expiratory flow rate between 25
and 75 percent of the FVC (FEF 55.759) with mean paired difference, standard deviation of difference, and significance
CE TL OOO

Mean paired difference, standard deviation of differences, and significance level for end-of-trial time, forced vital capacity
(FVC), and forced expiratory flow rate between 25 and 75 percent of the FVC (FEF 25.759) between manual method by 2
technicians and metNOg) 3. OF BNC-OFIIIBI CTILBIIB. cv. crisimisimsiinirssisssarssssssssmssissssimmsssi stb ns ss STAs SR AFOSR SSSA APRS SE EA RIBS oS

Mean and standard deviation for end-of-trial time, forced vital capacity (FVC), and forced expiratory flow rate between 25
and 75 percent of the FVC (FEF 5.75%), by 5 methods of detecting the end Of trial..........ccccceeveeirireneeneneereenieecneeee serene

Percentage differences in end-of-trial time, forced vital capacity (FVC), and forced expiratory flow rate between 25 and 75
percent of the FVC (FEF 25_759,) between 4 methods of detecting the end of trial and the average manual reading.................

Mean paired difference and standard deviation of differences for end-of-trial time, forced vital capacity (FVC), and forced
expiratory flow rate between 25 and 75 percent of the FVC (FEF 25.759) between manual methods by 2 technicians and
yy JE SE OP SI J) Re BU NR

Mean paired difference from index or negative flow method without noise for end-of-trial time, forced vital capacity (FVC),
and forced expiratory flow rate between 25 and 75 percent of the FVC (FEF5.75%), by 3 levels of noise on 18 normal
SPIr00rams ane 4 SNOOT-rIA) MNBUNIOUS ....coccoivniinririminnssisss mis temsiiassirsst esas Hrs ELE ELEAF E124 ER AEII ASE REPSTISEREASIA SR SRR RISI RRR RR HA A ER RSIF SAR

Mean paired difference from index or negative flow method without noise for end-of-trial time, forced vital capacity (FVC),
and forced expiratory flow rate between 25 and 75 percent of the FVC (FEF 5.759), by 3 levels of noise on 15 abnormal
SRIFODrams antl 4 SIVC-Of APIBE IIVBINOES, cc uriiissserrmomsssimrnns stm surmss ss ess oss aE IIHR AE SABE AAA SSAA SR APA 4 CHATTY R ERAEHE SAREE

37

37

37

38

38

38

39

39

40

40

41

aM

41

42

42
Table 1. Mean, standard deviation, and significance level for paired differences between zero-time methods on 19 trials for method 1

and manual method, by 2 technicians

 

 

 

 

 

 

Paired
Pairs differences p
X4 sd
Seconds
Computer and technician #1 ........cccovevverrerirrneencccsenessssssesessnes -0.007 | 0.037 NS
Computer and technician #2 .........ccccoeviernriinsseensssssesnrens -0.020 | 0.030 |<0.05
Technician #1 and technician #2....... ~0.013 | 0.030 NS

 

 

 

 

 

 

NOTES: Xq = mean paired difference; sq = standard deviation of difference; p = significance level; NS = not significant.

Table 2, Mean, mean paired difference, standard deviation of differences, and significance level for zero-time measurements, by 4

methods of computation on 19 trials

[ Zero time in seconds]

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Zero-time Alternative Paired differences
method and zero-time method ~
mean and mean Xd sd p

1 (0.926) 2 (0.917) -0.009 | 0.051 NS

1 (0.926) 3B {0.984) .cocinciiminmvsimanriri i Ee ese een eS eS FR eT TVR OTIS +0.058 | 0.037 | <0.05

1 (0.926) BUOIIBOY , ioreomvinsccmmvensnenpstone is snes SEI EST SARTO T AR SSE AE Hi UTS OATY +0.024 | 0.011 NS

NOTES: X4 = mean paired difference; sq = standard deviation of difference; p = significant level; NS = not significant.

Table 3. Mean forced expiratory volume at 1 second (FEV 1 g) by 19 trials, by 4 zero-time methods
Zero-time method Mean FEV1 go
Liters

1 rh ore rin rushers iri paneer en RR ER RET RTS 3.974
Brrr rR RRS SEER HEARSE pans RS 3.966
B rrsmerrirpreamsrerisrivsisssispreriperrerscsrs 4.034
Bs ieusniisirnmpsiarertanrim es REAPER EEDA EERE eR TA Se epee 4.002

 

 

 

 

37
Table 4. Mean paired difference, standard deviation of differences, and significance level for forced expiratory volume at 1 second
(FEV o), by zero-time method pairs

Paired differences
Zero-time method pairs

 

 

 

 

Xd sd Pp
Liters
1and 2 -0.008 | 0.055 NS
RL BNONB..rrrsiinsssisssssiassainissssoninsessatessaniisstiosnss +0.060 | 0.037 | <0.05
1and 4

 

 

+0.028 | 0.014 | <0.01

 

 

 

 

NOTES: Xq4 = mean paired difference; sq = standard deviation of difference; p = significance level; NS = not significant.

Table 5. Mean forced expiratory volume at 1 second (FEV g), by 3 levels of noise and zero-time method
[1 trial]

 

No SI ight Greater
noise noise
(2cV) | (4cV)

Zero-time method

 

 

 

Mean FEV g (liters)

 

 

 

 

rire hr RISE RIA EER eR Spe RE EE RE EERE SERS REE EERE SERRE EE OS SRSA ELDAR ex R 4.19 4.19 4.21
sit ianin sensors saan RSs AES AERIS ARRAS AAAS SIRES TSAR EAS aa T SISA AAA EAS - | 820 4.19 4.16
3 4.23 4.24 4.22
4 4.22 4.22 4.21

 

 

 

 

 

Table 8. Mean paired difference for forced expiratory volume at 1 second (FEV 1 g), by 3 levels of noise and zero-time method pairs

 

No Slight Greater
noise noise
(2cV) (4 cV)

Zero-time method pairs noise

 

Mean paired difference (liters)

 

 

 

 

 

 

 

 

1and 2 -0.008 | +0.008 -0.022

1and 3 140.060 | 1+0.063 +0.027

1and 4 +0.028 | +0.027 +0.008
15<0.05.

38
Table 7. Mean paired difference, standard deviation of differences, and significance level for zero time and forced expiratory volume at
1 second (FEV g), by zero-time method pairs on 18 abnormal spirograms

 

 

 

 

Zero-time paired FEV o paired
Zero-time method pairs differences differences
Xq Sd Pp X, d Sd Pp
Seconds Liters

0.028 | 0.094 NS | 0.022 | 0.073 NS
0.179 | 0.173 | <0.05 | 0.177 | 0.114 | <0.05
0.042 | 0.042 | <0.05 | 0.033 | 0.037 | <0.05

 

 

 

 

 

 

 

 

NOTES: Xq = mean paired difference; sq = standard deviation of difference; p = significance level; NS = not significant.

Table 8. Mean paired difference between zero-time methods, by 3 levels of noise for 18 abnormal spirograms

 

 

No Slight | Greater
Zero-time method pairs noite noise noise
(2cV) | (4cV)

 

 

 

Mean paired difference

(seconds)
0.028 0.037 0.177
0.179 0.127 0.265
0.042 0.001 0.099

 

 

 

 

 

Table 9. Mean paired difference and percent difference for forced expiratory volume at 1 second (FEV { g) between zero-time methods,
by 3 levels of noise for 18 abnormal spirograms

 

 

 

 

 

 

 

N . Slight noise Greater noise
Noise (2cv) (4 cv)
Zero-time method pairs
x Percent ~ Percent ~ Percent
Xd difference Xd difference *o difference
Liters Liters Liters
0.022 1.76 0.049 3.92 | 0.056 448
0.177 9.35 0.106 8.47 | 0.105 8.39
0.033 2.64 0.029 2.32 | 0.021 1.68
A ins RISER Eee ae SRA SRR HRSA S 4l3e ... | -0.038 3.04 was 0
A? servi . aon sme op ‘ne ... | 0.046 3.68

 

 

 

 

 

 

 

 

NOTES: A’ = Flow threshold method with noise-free signal minus flow threshold method with 2 ¢V of noise; A” = flow threshold
method with noise-free signal minus flow threshold method with 4 cV of noise.

39
Table 10. Mean and standard deviation for end-of-trial time, forced vital capacity (FVC), and forced expiratory flow rate between 25
and 75 percent of the FVC (FEF 25.759) with mean paired difference, standard deviation of differences, and significance level, by

2 methods of detecting the end of trial

 

 

 

 

 

 

 

 

 

 

 

Method 1 Method 3 Paired differences
Parameter — Fr =
X Sx X Sx Xd sd p
Seconds
ENO OT ATI THIN sussnonssiussmsrsssossss moms seis ios a SAA SAO 4.164 | 0.6081 6.024 | 1.689 | -1.860 | 1.830] <0.01
Liters
PVC, ismmnmemdinisniihsinmm ann cermin nrisemscimiasetvvas 4.787 | 0474 | 4.939 | 0.436 | -0.152 | 0.124| <0.01
Liters per second
PEP DBGBOL «cigs reessurensenemssupsrensssbenenrensisentessimersserssrsnsetnibmassisstrost vest channs 4.287 | 0.719 | 3877) 0.905 | +0.310 | 0.269 | <0.01

 

NOTES: X = mean; sx = standard deviation; Xg = mean paired difference; sq = standard deviation of difference; p = significance level;

NS = not significant.

Table 11. Mean and standard deviation for end-of-trial time, forced vital capacity-(FVC), and forced expiratory flow rate between 25
and 75 percent of the FVC (FEF 25.759) with mean paired difference, standard deviation of difference, and significance level, by 2

technicians

 

 

 

 

 

 

 

 

 

 

Technician Technician Paired differents
Parameter no. 1 no. 2
x Sx x Sx Xq sd Pp
Seconds
ERCh OTANI) G00... rsisissmpsromisstiss missions ssiiisnsiasi sins ssitsiissmis 6.268 | 1.627] 6.244 | 1693 | 0.014] 0.219 | NS
Liters
Lh EE ORTOP Se PU 4957 | 0.426] 4.951 | 0.435 | 0.006 | 0.025 NS
Liters per second
PE I. FBO sc vesuisi cs sssumriss infor arses dreams OAR ER ke isi as ATA 08 szeel 0560] 0571 0.987 {020 0.433 [<0.01

 

NOTES: X = mean; sx = standard deviation; Xq = mean paired difference; sq = standard deviation of difference; p = significance level;

NS = not significant.

40
Table 12. Mean paired difference, standard deviation of differences, and significance level for end-of-trial time, forced vital capacity
(FVC), and forced expiratory flow rate between 25 and 75 percent of the FVC (FEF25.759) between manual method by 2

 

 

 

 

 

 

technicians and method 3 of end-of-trial criteria
Paired differences Paired differences
between technician between technician
Parameter no. 1 and method 3 no. 2 and method 3
X, d Sd Pp X d Sd Pp
Seconds Seconds

0.234 | 1.040 NS | 0.220 | 1.034] NS

 

End-of-trial time

 

 

 

 

 

Liters Liters

BEV Cs coriiiimcssmrmmanss sve msbiness loti thesia shia da eashss pis oes ER eT FATA RET 0.018 | 0.070 | Ns | 0.012 | 0.072 NS
Liters per . Liters per
second second

FEF 25.75% 0.189 | 0.321 |<0.01 0.080 | 0.324] NS

 

 

NOTES: Xq = mean paired difference; sq = standard deviation of difference; p = significance level; NS = not significant.

Table 13. Mean and standard deviation for end-of-trial time, forced vital capacity (FVC), and forced expiratory flow rate between 25
and 75 percent of the FVC (FEF 25.759), by 5 methods of detecting the end of trial

 

 

 

Average
Method 1 Method 2 Method 3 Method 4 of manual
Parameter readings
x Sx x Sx x Sx Xx Sx x Sx

 

 

 

 

 

 

 

 

 

 

End-Of-trial timMBu....viesrsirissssssimssssisismsssisssrn

 

FVC

 

PEP OB. 789 «ocr srsimsimssimnsenmessessessrssrosssssesssiven
NOTES: X = mean; Sy = standard deviation.

Seconds
4.164 0.608 | 5.036| 1.077 |. 6.024 | 1.689 | 6.342 | 1.611 | 6.251 | 1.657
“Liters
4.787| 0474 | 4.878 | 0482] 4.939 | 0.436 | 4.966 | 0.444 | 4.954 | 0.430
Liters per second

4.287 | 0.719 | 4.103 | 0.758 | 3977 | 0.905 | 3.905 | 0.888 | 3923 0.900

Table 14. Percentage differences in end-of-trial time, forced vital capacity (FVC), and forced expiratory flow rate between 25 and 75
percent of the FVC (FEF 25.759) between 4 methods of detecting the end of trial and the average manual reading

 

 

Method 1 | Method 2 | Method 3 | Method 4

 

 

 

 

 

 

 

 

 

Parameter
Percent
End-of-trial time . -33.4 -19.4 -3.6 +1.5
BEV. isis vessnnnrisninsoninssninmoessssss it sas uss sshssss bess sss RESTA LTR RA CARR OSV RPE Laas -3.4 -1.5 -0.3 +0.2
FEF QB.7B%, esresrserserssssssesssssssnsisssssssssssssssssssssssssssssssssnsassstssssssssssssssstsesssssssssssssessssssas 19.3 +4.6 +1.4 -0.5

 

 

41
Table 156. Mean paired difference and standard deviation of differences for end-of-trial time, forced vital capacity (FVC), and forced
expiratory flow rate between 25 and 75 percent of the FVC (FEF 25.759) between manual methods by 2 technicians and methods

 

 

 

 

 

 

 

 

 

 

 

 

 

3and 4
Paired differences between Paired differences between
technician no. 1 and: technician no. 2 and:
Pevamgiar Method 3 Method 4 Method 3 Method 4
Xq sd Xq sd Xq sd Xq sd
Seconds
ENG-OF-trial tM .....vvoorvveveesereeessssssesssesssssessssssssssssesssssens 0.234 | 1.040 | 0.100] 0.359] 0.220] 1.034 | -0.086 | 0.325
Liters
PVC co nssimiusssi sis oss i oss issn Ss ss 3 mi ss seis 0.018 | 0.070 | 0.009 | 0.021 | 0.012] 0.072| -0.016| 0.270
Liters per second
LT -0.189 | 0.321 | -0.116 | 0222 | +0.080 | 0.324 | 0.112 0.270

 

NOTES: a = mean paired difference; sq = standard deviation of difference.

Table 16. Mean paired difference from index or negative flow method without noise for end-of-trial time, forced vital capacity (FVC),
and forced expiratory flow rate between 25 and 75 percent of the FVC (FEF 25.75%), by 3 levels of noise on 18 normal spirograms

and 4 end-of-trial methods

 

 

Parameter and method

Slight noise | Greater noise

No noise | ™ ev) (4 cv)

 

 

 

End-of-trial time

Mean paired difference (seconds)

 

 

9 is -1.740 +0.476
+0.320 -0.943 +0.473
-0.943 -0.943 -1.427
-1.818 -1.762 -1.532

 

Mean paired difference (liters)

 

 

 

 

-0.142 -0.030
+0.022 -0.057 +0.001
-0.061 -0.061 -0.115
-0.304 -0.150 -0.163

Mean paired difference
(liters per second)

+0.352 -0.127
-0.073 +0.112 -0.125
+0.137 +0.124 +0.726
-0.310 +0.112 -0.125

 

 

42
Table 17. Mean paired difference from index or negative flow method without noise for end-of-trial time, forced vital capacity (FVC),
and forced expiratory flow rate between 25 and 75 percent of the FVC (FEF25.759), by 3 levels of noise on 15 abnormal

spirograms and 4 end-of-trial methods

 

 

Parameter and method

No noise

Slight noise
(2¢cV)

Greater noise
(4cV)

 

End-of-trial time

 

 

 

 

 

 

 

 

 

 

 

 

Mean paired difference (seconds)

0.000
-2.318
-3.151

-3.050

-2.310
-2.318
-3.064

+0.155

+0.155
-2.446
-3.120

Mean paired difference (liters)

0.000
+0.555
-0.370

 

-0.144

-0.080
-0.083
-0.153

 

-0.050

-0.050
-0.214
-0.765

Mean paired difference

0.000
+0.074
0.000

+0.134

+0.089
+0.093
+0.138

(liters per second)

-0.033

-0.033
+0.079
+0.100

 

43
44

APPENDIXES

~ CONTENTS

I. Glossary

 

II. General Spirometric Test Procedure Used by NHANES

 

 

Instructions for Older Children and Adults
Instructions for Children 6-10 Years of Age

 

APPENDIX FIGURE

I. Three phases in a forced expiratory spirogram

 

45
47

47

45
APPENDIX |
GLOSSARY

Trial. —The term “trial” is used to specify a
single effort to perform the forced expiratory
spirogram (FES) maneuver. A test set comprises
five trials; however, a spirometric examination
may comprise one to three test sets. The FES
always has three parts as shown in figure I and
the accompanying explanation.

® Phase I-effort-dependent flow, entailing
peak flow for the breath.

® Phase II—constant deceleration of vol-
ume exhaled (critical flow).

® Phase III—terminal leakage flow.

 

 

 

 

 

PHASES
| | n | m
I T 1
a Ir Critical gn
Z Pa |
8
EL
: |
9 f= |
0 2 8

VOLUME IN LITERS

 

 

 

Figure |. Three phases in a forced expiratory spirogram

Procedural error.—A violation of the ex-
pected shape (morphology) and completeness of
the spirometric flow-volume signal is known as
a “procedural error.” If artifacts are observed,
the indication is that one or more of the follow-
ing procedural errors has occurred during the
test: an inhalation or hesitation during the per-
formance of the test, a less-than-expected initial
respiratory thrust, an absence of the character-
istic decay portions of the flow and volume due
to a premature termination of the expiratory
effort before reaching residual volume, or flow
or volume values higher than clinically possible,
perhaps due to a Venturi error.

Detection of these artifacts is cause to chal-
lenge the validity of the test effort. For exam-
ple, if an inhalation artifact is observed, the test
has no clinical value and should be discarded. In
contrast, if a Venturi phenomenon is observed
but the flow and volume history meet the re-
producibility criteria on another trial, the cri-
teria for reliable data have been fulfilled and the
test should be declared valid (in this example,
the higher values observed may be because a per-
son is exceptionally fit, e.g., an athlete).

Best trial. —The best trial refers to a particu-
lar trial, within a set of five, that clearly demon-
strates the optimum presence of the two vari-
ables that are of the most interest (i.e., thrust
and sustained expiratory effort). Thrust, as
measured by the flow rate, is judged through
the visual observation of the flow-volume curve
velocities. Sustained expiratory thrust therefore
results in the presence of higher flow rates
throughout the spirogram, in contrast with other
trials in which less thrust was applied.

Variations in thrust can be reflected by over-
all reductions in the peak flow rate (PFR), the

45
forced expiratory flows at 25 percent, 50 per-
cent, and 75 percent of the vital capacity, the
FEFg5 759, or in all of these. In contrast, sus-
tained expiratory effort is judged by observation
of the FVC signal on the volume (horizontal)
axis of the oscilloscope display. Because all flow
rates are volume dependent (except for the
PFR), all flow rates must be judged in the con-
text of the largest FVC. The best trial is the one
that exhibits the highest flow rates in the pres-
ence of the largest FVC (and in the absence of
obvious procedural errors).

Biologic variation often causes some varia-
tion of these values from trial to trial, such as
the presence of the highest observable flow rates
on a trial in which the FVC is slightly lower than
in another trial, or conversely, a trial where a
larger FVC contains slightly lower flow rates.
During an examination, the technicians must use
their best judgment to identify the best trial,
taking into consideration the phenomenon of
biologic variation; when the technician consci-

entiously applies reliability criteria (described
in the following section), this problem will re-
solve itself.

This “eyeball” type of judgment is validated
and supplemented during data processing by a
digital computer at a later time when each flow
and volume parameter in the spirometric curve is
accurately measured and compared with like
measures in other curves in the trial sequence. In
this way, a consistent method of identifying the
best and most reliable data is used, which
reduces the probability of unacceptable data
quality.

Reliability. —Reproducible spirographic trac-
ings are assumed to represent the very best ef-
forts of the examinee. Reliability is determined
by comparing the flow and volume characteris-
tics of the best and second-best spirograms in a
test set of five trials by determining the repro-
ducibility of the two trials. If limits of repro-
ducibility are violated, the best trial is declared
not reliable, and the test sequence is repeated.

000
APPENDIX II
GENERAL SPIROMETRIC TEST PROCEDURE USED BY NHANES

Instructions for Older Chiidren
and Adults?

Examinees follow four simple steps during
the test procedure:

® Inhaling as deeply as possible (examinee
maximally inhales from room air).

® Holding in all the air while placing the
lips tightly around the mouthpiece,
being careful to keep the tongue under
the mouthpiece.

® Blasting out of all the air.

® Maintaining the expiratory effort until
all the air is out.

These directives, that is, a deep inspiration,
proper placement of the mouthpiece, and the
forceful exhalation of air into the tube, are dem-
onstrated to the examinee by the attending tech-
nician. The technician overemphasizes the last
step by doubling up in an effort to squeeze out
air by full exhalation.

Before commencing the first trial, the ex-
aminee should have the mouthpiece firmly
seated in the hose with both hands around the
hose an inch or so from the mouthpiece. The
standard instructions for the test procedure are
as follows:

® “Take in a deep breath of air, as far as
you can go.”

® “Hold in the air and place your lips
tightly around the mouthpiece.” (Move
the hose toward the subject so he or she
will insert mouthpiece; when the mouth-
piece is in place, press the recorder

* button.)

® “Keep blowing!!! Keep blowing!!! Get it
all out!!! Keep blowing!!!” By monitor-
ing the oscilloscope, the technician can
detect when the volume is not increasing
and can judge when 1 or 2 seconds of
effort at full expiration have transpired.

For subsequent trial instructions, the techni-
cian must repeat the last four standard instruc-
tions or use a special instruction for solving a
misinterpretation based on the observations that
he or she has made. A complete test set requires
the recording of five spirograms.

Instructions for Children 6-10 Years
of Age?

In the current NHANES and in previous sur-
veys, children 6-10 years of age were tested. The
following spirometry instructions were applied
to the less behaviorally mature children; this
group included most of the children 6 and 7
years of age (except for a few poised children
with mature behavior), about half of the chil-
dren 8 years of age, and the least mature of the
children 9 and 10 years of age.

Technicians were instructed to proceed
slowly, to be patient with the children, to use
few words, and to demonstrate mostly by vivid,
clear example. A typical instruction situation
was

“We want to see how your lungs work. We
will measure how much air you can take in
(accompanied by a strenuous demonstra-
tion) and how much you can blow out—as
hard and as fast as you can.”

At this point, the technician asked the child if
he or she could blow up a balloon quickly. The

47
response usually gave the technician an indica-
tion of how ready the child was to perform a
satisfactory spirogram. The instruction contin-
ued with “Let me show you”:

“Take in as much air as you possibly can.
Like this.”

“Hold it all in until I tell you to blow.”
“Lips closed around the tube like this.”

“When I say ‘blow,’ blow out all of your air
as fast and hard as you can. Keep blowing
out hard until I tell you to stop.”

The child was then given one or two practice
trials in a relaxed way. If still unsure of the

child’s understanding and ability to perform, the
technician repeated the entire demonstration,
stressing anything that the child did not do per-
fectly (i.e., reemphasizing by use of exaggerated
demonstration rather than words how to per-
form correctly).

A complete spirometry test set also com-
prised five trials. If further instruction between
trials was necessary, it was vivid, pertinent, and
brief. If no satisfactory data were obtained dur-
ing the test, a decision was made either to have
the child come back later during the examina-
tion session if the technician held hope for
better trials, or mark the test as unsatisfactory if
the technician believed that a satisfactory test
could not be obtained because the child was
frightened, too confused, or did not understand.

000

48

#U.S. GOVERNMENT PRINTING OFFICE: 1980 341-161/24 1-3
 

 

 

VITAL AND HEALTH STATISTICS Series

Scites 4; Programs and Collection Procedures.—Reports which describe the general programs of the National

Center for Health Statistics and its offices and divisions and data collection methods used and include
definitions and other material necessary for understanding the data.

Series 8 Data Evaluation and Methods Research.—Studies of new statistical methodology including experi-
mental tests of new survey methods, studies of vital statistics collection methods, new analytical
techniques, objective evaluations of reliability of collected data, and contributions to statistical theory.

Serle 8, Analytical Studies.—Reports presenting analytical or interpretive studies based on vital and health
Statistics, carrying the analysis further than the expository types of reports in the other series.

Sores d, Documents and Committee Reports.—Final reports of major committees concerned with vital and
health statistics and documents such as recommended model vital registration laws and revised birth
and death certificates.

Series 90, Data From the Health Interview Survey.—Statistics on illness, accidental injuries, disability, use of
hospital, medical, dental, and other services, and other health-related topics, all based on data collected
in a continuing national household interview survey.

Serles 9, Data From the Health Examination Survey and the Health and Nutrition Examination Survey.—Data
from direct examination, testing, and measurement of national samples of the civilian noninstitu-
fionalized population provide the basis for two types of reports: (1) estimates of the medically defined
prevalence of specific diseases in the United States and the distributions of the population with respect
fo physical, physiological, and psychological characteristics and (2) analysis of relationships among the
various measurements without reference to an explicit finite universe of persons.

Sorles 88, Data From the Institutionalized Population Surveys.—Discontinued effective 1975. Future reports from
these surveys will be in Series 13.

Soles 48, Data on Health Resources Utilization. —Statistics on the utilization of health manpower and facilities
providing long-term care, ambulatory care, hospital care, and family planning services.

Series 44, Data on Health Resources: Manpower and Facilities. —Statistics on the numbers, geographic distri:
bution, and characteristics of health resources including physicians, dentists, nurses, other health
occupations, hospitals, nursing homes, and outpatient facilities.

Servs 80, Data on Mortality. — Various statistics on mortality other than as included in regular annual or monthly
geports. Special analyses by cause of death, age, and other demographic variables; geographic and time
geries analyses; and statistics on characteristics of deaths not available from the vital records based on
sample surveys of those records.

Sorles 83, Duta on Natality, Marriage, and Divorce.—Various statistics on natality, marriage, and divorce othép
+ than as included in regular annual or monthly reports. Special analyses by demographic variables;
geographic and time series analyses; studies of fertility; and statistics on characteristics of births not

available from the vital records based on sample surveys of those records.

Serles 22. Data From the National Mortality and Natality Surveys. —Discontinued effective 1975. Future reports)
from these sample surveys based on vital records will be included in Series 20 and 21, respectively,

8orles £3, Data From the National Survey of Family Growth. Statistics on fertility, family formation and dis:
solution, family planning, and related maternal and infant health topics derived from a biennial survey)
of a nationwide probability sample of ever-married women 15-44 years of age.

Pov a list of titles of zeposts published in these series, write tog Scientific and Technical Information Branch
a National Center for Health Statistics
Public Health Service
Hyattsville, Md. 20782

 

 
 

 

edn”
© “

Hi
ela

Z,
3 Nex)

8;
La

National Survey of
Family Growth, Cycle II:

Sample Design,
ATE OR OL LE
Variance Estimation

U.S. DEPARTMENT OF

HEALTH AND HUMAN SERVICES

Public Health Service

Office of Health Research, Statistics, and Technology
National Center for Health Statistics

 

 
 

 

 

Library of Congress Cataloging in Publication Data

Grady, William R.
National survey of family growth, cycle II.

(Vital and health statistics : Series 2, Data evaluation and methods research ; no. 87)
(DHHS publication ; no. (PHS) 81-1361)

Includes bibliographical references.

1. Family size—United States—Statistical methods. 2. Fertility, Human—United States.
3. Birth control—United States. I. Title. II. Series: United States. National Center for
Health Statistics. Vital and health statistics : Series 2, Data evaluation and methods re-
search ; no. 87. III. Series: United States. Dept. of Health and Human Services. DHHS
publication ; no. (PHS) 81-1361.
RA409.U45 no. 87 [HQ766.5.U5] 312'.0723s [804.6'3]
ISBN 0-8406-0199-9 ; 80-20552

 

 

For sale by the Superintendent of Documents, U.S. Government Printing Office, Washington, D.C. 20402

 
DATA EVALUATION Series 2
AND METHODS RESEARCH Number 87

National Survey of
Family Growth, Cycle II:

Sample Design,
Estimation Procedures, and
Variance Estimation

 

DHHS Publication No. (PHS) 81-1361

 

U.S. DEPARTMENT OF HEALTH AND HUMAN SERVICES
Public Health Service
Office of Health Research, Statistics, and Technology
National Center for Health Statistics
Hyattsville, Md. February 1981
NATIONAL CENTER FOR HEALTH STATISTICS

DOROTHY P. RICE, Director

ROBERT A. ISRAEL, Deputy Director

JACOB J. FELDMAN, Ph.D., Acting Associate Director for Analysis and Epidemiology

GAIL F. FISHER, Ph.D., Acting Associate Director for the Cooperative Health Statistics System
GARRIE J. LOSEE, Acting Associate Director for Data Processing and Services
ALVAN O. ZARATE, Ph.D., Assistant Director for International Statistics
E. EARL BRYANT, Acting Associate Director for Interview and Examination Statistics
ROBERT C. HUBER, Acting Associate Director for Management
MONROE G. SIRKEN, Ph.D., Acting Associate Director for Research and Methodology
PETER L. HURLEY, Acting Associate Director for Vital and Health Care Statistics

ALICE HAYWOOD, Information Officer

STATISTICAL METHODS STAFF, DATA SYSTEMS

E. EARL BRYANT, Chief

DIVISION OF VITAL STATISTICS

JOHN E. PATTERSON, Director
ALICE M. HETZEL, Deputy Director
WILLIAM F. PRATT, Ph.D., Chief, Family Growth Survey Branch
JOSEPH D. FARRELL, Chief, Programming Branch
MABEL G. SMITH, Chief, Statistical Resources Branch

COOPERATION OF WESTAT, INCORPORATED
In accordance with specifications established by the National Center for Health Statistics,

Westat, Inc. of Rockville, Maryland, under a contractual agreement, participated in the design
and selection of the sample and carried out the data collection.

Vital and Health Statistics-Series 2-No. 87

 

DHHS Publication No. (PHS) 81-1361
Library of Congress Catalog Card Number 80-20552
PREFACE

This report presents a detailed description of the sample design, estimation
procedures, and variance estimation method used in Cycle II of the National Sur-
vey of Family Growth. The survey was designed and conducted by Westat, Inc. of
Rockville, Maryland under a contractual arrangement with the National Center
for Health Statistics (NCHS). The sampling plan was developed under the super-
vision of Joseph Waksberg of Westat, Inc., in consultation with E. Earl Bryant and
William F. Pratt of NCHS.

Much of the report is based on survey specification documents and the final
report prepared by Westat, Inc., and on internal NCHS memoranda. Parts of the
report are also largely based on a previous report, prepared by Dwight K. French,
on the Cycle I survey. Mr. French was the primary resource person for methodo-
logical questions.

In addition to the usual internal review, NCHS policy stipulates that meth-
odological reports are to be given a peer review for technical merit and readability
by one or more persons who are familiar with the subject matter area but who are
not involved in producing the report. George A. Schnack and Peter W. Ries, both
of NCHS, carried out the peer review of this report and provided many construc-
tive suggestions.
 
CONTENTS

Preface

 

Introduction

 

Design Specifications

 

Sample Design

 

Summary

 

 

First-Stage Selection of Primary Sampling Units
Second-Stage Selection of Enumeration Districts

 

Third-Stage Selection of Segments

 

Fourth-Stage Selection of Households

 

 

Fifth-Stage Selection of Sample Persons
Nonresponse Followup

 

 

Characteristics of the Sample
Response Rates

 

Sample Size

 

Estimation

 

Weighting Procedures

 

Estimating Equation

 

Variance Estimation

 

Background

 

 

Summary of Applicable Theory
Application to the National Survey of Family Growth

 

References

 

Appendixes
I. Glossary of Terms

 

II. Household Screener

 

LIST OF TEXT FIGURES

1.  First-stage selection of primary sampling units

 

2. Third-stage selection of sample segments

 

3. Relative standard errors for aggregates of currently married women, by race

4. Relative standard errors for percent of currently married women of all races

LIST OF TEXT TABLES

A. States in conterminous United States, by geographic region

 

B. Sample primary sampling units (PSU's), by type

 

ii

27
29

11
22

24
vi

RTE pos

=

Estimated number of women eligible for inclusion in the sample, expected sample size, and
approximate sampling fraction required to produce the expected sample size, by race .....ccceeeennenes

Total and expected number of dwelling units (DU’s) in the United States, and percent difference,
by stratum

 

 

Example of sampling table on the Household Screener

Percent of sample women, by race and sampling rate

 

Number of dwelling units (DU’s), by nonresponse subsample status and final disposition ......ceceee..
Weighted number of dwelling units (DU’s), by final diSPOSIitiON .cccsseesserisesscssasssasessesansssseasssasensans

Weighted screener, interview, and combined response rates, by Stratim .....cccccseeccceseescccsssncecscsanease

 

Poststratification adjustment factors, by race and age

Example of a half-sample replication pattern

 

Estimates of parameters A and B for relative standard error curves, by type of statistic and race...

 

SYMBOLS

Data not available---------eeeememmme eee.

Category not applicable-------woroeoeeeeeeeee |

 

Quantity zero -
Quantity more than 0 but less than 0.05-—- 0.0

Figure does not meet standards of
reliability or precision---------e-eeseeeeeeemeeen *

 

 

 

13
14
14
15
16
16
17
19

21
NATIONAL SURVEY OF FAMILY GROWTH, CYCLE II:
SAMPLE DESIGN, ESTIMATION PROCEDURES,
AND VARIANCE ESTIMATION

William R. Grady, Division of Vital Statistics

INTRODUCTION

The primary mission of the National Center
for Health Statistics is to collect and publish
data relating to the health of the U.S. popula-
tion. In carrying out this mission the Center
data on vital events registered in the United
States are collected, inventories of health facili-
ties and manpower are conducted, and proba-
bility sample surveys based on household inter-
views, health examinations, and medical records
are conducted. Data collection programs are
supplemented by research projects that investi-
gate new techniques of data collection and
evaluate currently operating programs.

In response to the need for current informa-
tion on fertility and family planning and their
effects on population growth, the National
Survey of Family Growth was established in
1971 as an integral part of the Center’s Division
of Vital Statistics. The purpose of the survey is
to collect data relating to natality and the
process of family formation and dissolution. The
National Survey of Family Growth was designed
as a cyclic survey; that is, data are collected
every few years by means of a sample survey.
The first cycle of the survey was conducted in
19738, the second in 1976; Cycle III is scheduled
to be conducted in 1981.

The target population of Cycles I and II of
the National Survey of Family Growth consisted
of the civilian household population of women
aged 15-44 years, living in the conterminous

United States, who were currently married,
previously married, or never-married mothers
with offspring living in the household at the
time of interview. Data were collected from a
probability sample by means of personal inter-
views lasting about an hour. The interviews
provided information on fertility trends and
differentials, family planning practices, sources
of family planning advice and services, use and
effectiveness of contraceptives, and aspects of
maternal and child health that are closely related
to family planning.

The sample design and data collection for
Cycle I were contracted to the National Opinion
Research Corporation of the University of Chi-
cago. A complete description of that survey can
be found in another report.! The sample design
and data collection for Cycle II were contracted
to Westat, Inc. of Rockville, Maryland. The
sample consisted of 9,470 eligible women, of
whom 8,611 (90.9 percent) were interviewed.
All interviews were conducted between January
and September 1976 and centered on May 12.
The sample design employed to select the
women is described in detail in this report. Also
described are the techniques used to estimate
population parameters and the procedures used
to obtain provisional sampling variances.

DESIGN SPECIFICATIONS

The development of an efficient sample

design mut take into account the primary
survey objectives, the amount of funds available,
logistical problems, time limitations, estimates
of population characteristics and distribution,
and operating costs. These requirements dictated
a stratified multistage probability sample design
for Cycle II, based essentially on the following
set of specifications:

1. The target population was defined to be
the civilian household population of
women living in the conterminous
United States who were 15-44 years of
age and either
(a) currently married,

(b) previously married, or

(c) never-married mothers with one or
more children born to them cur-
rently living in the household.

2. The sample would consist of approxi-
mately 10,000 women, selected from an
initial probability sample of households.
It would include about 4,000 black
women and 6,000 women of other races.
Trained field staff were to conduct a
screening interview with a responsible
member of each sample household to de-
termine if there were any eligible women
(the screener questionnaire is reproduced
in appendix II). When a household con-
tained one eligible woman, she was in-
cluded in the sample. In households with
more than one eligible woman, the staff
member would randomly select one
woman for the sample.

3. Data were to be collected from the
sample women (no proxy respondents
were to be accepted) by means of
personal interviews lasting an average of
1 hour.

All interviewers would be female.

The interviewer would collect informa-
tion on fertility, family planning prac-
tices, sources of family planning services,
and related maternal and child health
practices.

6. The fieldwork would be completed in
approximately 6 months.

7. The target interview completion rate for
the total sample and both major subsam-
ples by race was 90 percent of the ex-
pected number of women from all
sample households (i.e., screener and in-
terview nonresponse combined would
ideally be no more than 10 percent).

8. The contractor, in cooperation with the
National Center for Health Statistics
(NCHS), would design and implement
procedures to measure and control the
quality of data collection and data prep-
aration.

SAMPLE DESIGN

Summary

The sample design for Cycle II of the
National Survey of Family Growth (NSFG) was
a five-stage probability design that incorporated
a supplementary sample of new (post-1969)
housing units. This section provides an overview
of the design, and it is followed by sections
discussing each stage in detail.

The counties and independent cities that
comprise the total land area of the conterminous
United States were combined to form a frame of
primary sampling units (PSU’s). During the first
stage of the sampling process, which involved
extensive stratification, 79 PSU’s were chosen
from this frame. Census enumeration districts
(ED’s) were then identified for each of the
selected primary sampling units; during the
second stage, these enumeration districts were
stratified according to the percent of their
population that was black, and a systematic
sample was drawn. The rate at which enumera-
tion districts were sampled varied by second-
stage strata. These differential sampling rates
were the first step in producing the desired racial
composition of the final sample of women.

The third stage required the identification of
area segments (groups of houses) within sample
enumeration districts and the random selection
of one segment from each district. All sample
dwelling units (DU’s) built prior to 1970 were
selected this way. However, segments of new
construction units (post-1969) were drawn from
a supplementary sample of building permits
selected from building permit offices in the 79
primary sampling units chosen during the first
stage of the sampling process.

The fourth stage resulted in the selection of
households within sample segments. In segments
from enumeration districts with a 10-percent or
greater black population, households of races
other than black were selected at a rate that was
lower than for black households. These different
rates of selection were obtained through a
subsampling process (to be described later in this
report) and ensured that the desired proportions
of black and other women would be included in
the final sample. At each sample household an
interviewer attempted to complete a Household
Screener and identify women eligible for inter-
view. When more than one potential respondent
was found during this fifth-stage operation, all
eligible women in the household were listed and
one was randomly selected.

First-Stage Selection of
Primary Sampling Units

Sampling frame.—The counties and inde-
pendent cities of the conterminous United
States have been grouped into about 1,900
primary sampling units (PSU’s) by the U.S.
Bureau of the Census for its Current Population
Survey. These census PSU’s were used by
Westat, Inc., with minor modifications, as the
sampling frame for the National Survey of
Family Growth. Each PSU consisted of an
individual county or a grouping of contiguous
counties. Where standard metropolitan statistical
areas (SMSA’s) were defined, the counties com-
prising each SMSA were used as a primary
sampling unit.2 Seventy-nine PSU’s were se-
lected in the first stage of the sampling process.

The sample contains 25 self-representing
(included with certainty) primary sampling units

ASMSA definitions used are those of 1970 (see
appendix 1). No attempt was made to have the sample
reflect new SMSA’s created since 1970 or changes in
SMSA definitions since that date. In New England,
where SMSA’s are not defined in terms of counties,
metropolitan state economic areas (MSEA’s), which are
comprised of counties, were used as primary sampling
units instead of SMSA’s. Thus primary sampling units
were in all cases defined in terms of complete counties.

composed of 18 separate SMSA’s. These include
the 14 largest SMSA’s, with 1970 populations of
more than 1,850,000, and 4 slightly smaller ones
that were made self-representing because they
could not easily be placed in strata with other
primary sampling units. Four of the largest
SMSA’s, which contained several counties each,
were subdivided into smaller primary sampling
units, each with an average population of about
2,500,000.

Selection of nonself-representing PSU’s.—
Prior to the selection of the remaining PSU’s, all
nonself-representing (probability of selection
less than unity) PSU’s were grouped into 35
strata of approximately equal size. Nineteen
strata contained SMSA’s and 16 contained non-
SMSA areas.

The more than 200 nonself-representing
SMSA’s were sorted into 19 strata of about
4,000,000 persons each. In defining the appro-
priate stratum, four characteristics were con-
sidered. In order of priority, these characteristics
were the following:

1. Region of the country (see table A).

2. Percent change in population between
1960 and 1970.

3. Percent of the population employed in
manufacturing.

4. A Socioeconomic Index (developed by
Westat, Inc.) that was based on the per-
cent of the population that is white, the
percent of households that either lacks
plumbing or is overcrowded, and the de-
pendency ratio P

The nearly 1,700 remaining PSU’s were then
sorted into an additional 16 strata. The criteria
in order of priority for this stratification were as
follows:

1. Region of the country.

2. Percent of the population living in urban
areas.

PThe dependency ratio is the ratio of the number
of people who are under 18 or over 64 years of age to
the number of people who are between the ages of
18-64, inclusively.
Table A. States in conterminous United States, by geographic region

 

 

Region

 

 

States
Connecticut New Hampshire Pennsylvania
Maine New Jersey Rhode Island
Massachusetts New York Vermont
Illinois Michigan North Dakota
Indiana Minnesota Ohio
lowa Missouri South Dakota
Kansas Nebraska Wisconsin
Alabama Kentucky South Carolina
Arkansas Louisiana Tennessee
Delaware Maryland Texas
District of Columbia Mississippi Virginia
Florida North Carolina West Virginia
Georgia Oklahoma
Arizona Montana Utah
California Nevada Washington
Colorado New Mexico Wyoming
Idaho Oregon

 

3. Percent change in population between
1960 and 1970.

4. The Westat Socioeconomic Index.

This stratification insured the proportionate rep-
resentation of women by region and socioeco-
nomic status, and may also have reduced sam-
pling error.

Following the stratification process, PSU’s
were selected in two stages.© In the first stage, 1
PSU was selected from each of the 35 strata
with a probability proportionate to size. The
second-stage selection of the remaining 19 PSU’s
was then accomplished in 3 steps. First the 35
strata were combined into 19 superstrata. This
recombination was done in such a way as to
produce strata that were, as far as possible,
homogeneous with respect to region, metropol-
itan composition (SMSA vs. non-SMSA), rate of
population change between 1960 and 1970, and
the percent of the population that was black.

¢Two stages were used because a 50-PSU design
existed prior to the decision to change to a 79-PSU de-
sign. It was thus more efficient to go through a second
stage of selecting additional PSU’s than to select a new
independent sample.

Four of the resultant superstrata contained a
single stratum each, 14 contained 2 strata, 1
contained 3 strata. The second step was the
selection, with a probability proportionate to
size, of a single stratum within each super-
stratum. The final step was the selection, also
with a probability proportionate to size, of 1
PSU from each of the 19 strata. Since the
second-stage selection of 19 PSU’s was done
independently of the first-stage selection of 35
PSU’s (i.e., sampling was done with replace-
ment), it was possible for PSU’s to be selected
twice; this was the case for 1 PSU. All selected
PSU’s are shown in table B, and a diagram of the
first stage of the sampling process is shown in
figure 1.

Second-Stage Selection of
Enumeration Districts

Sampling rates.—Because the final sample of
women for the National Survey of Family
Growth was intended to consist of approxi-
mately 4,000 black women and 6,000 women of
other races, different sampling rates were re-
quired for the 2 groups. Given a 100-percent re-
sponse rate, the approximate sampling fraction
needed to produce the required number of black
women was 1 in 950, while the corresponding
Table B. Sample primary sampling units (PSU's), by type

 

 

Self-representing SMSA's

 

Anaheim-Santa Ana-Garden Grove, Calif.
Baltimore, Md.
Boston-Lowell-Lawrence, Mass. (MSEA)
Buffalo, N.Y.

Chicago, Ill. (represented by 2 PSU's)
Cleveland, Ohio

Detroit, Mich. (represented by 2 PSU's)
Houston, Tex.

Kansas City, Mo.-Kans.

Los Angeles-Long Beach, Calif.

Newark, N.J.

New York, N.Y. (represented by 5 PSU's)
Paterson-Clifton-Passaic, N.J.

Philadelphia, Pa.-N.J. (represented by 2 PSU's)
Pittsburgh, Pa.

San Francisco-Oakland, Calif.

St. Louis, Mo.-lll.

Washington, D.C.-Md.-Va.

Nonself-representing SMSA's

Allentown-Bethlehem-Easton, Pa.-N.J.
Birmingham, Ala.
Bridgeport-Stamford-Norwalk, Conn. (MSEA)
Chattanooga, Tenn.-Ga.

Dallas, Tex.

Denver, Colo.

Erie, Pa.

Fayetteville, N.C.

Grand Rapids, Mich.

Hartford-New Britain-Bristol, Conn. (MSEA)
Indianapolis, Ind.

Jacksonville, Fla.

Lansing, Mich.

Las Vegas, Nev.

Little Rock-North Little Rock, Ark.

Madison, Wis.

Miami, Fla.

Milwaukee, Wis.
Minneapolis-St. Paul, Minn.
New Orleans, La.
Oklahoma City, Okla.
Omaha, Nebr.-lowa
Salinas-Monterey, Calif.
San Angelo, Tex.
Seattle-Everett, Wash.
Spokane, Wash.

Syracuse, N.Y,

Tacoma, Wash.

Tampa-St. Petersburg, Fla.
Toledo, Ohio-Mich.

 

Nonself-representing non-SMSA’s

Auglaize Co.-Shelby Co., Ohio

Benton Co.-Carroll Co., Ind.

Calhoun Co.-Clay Co.-Roane Co., W. Va.

Calvert Co.-Charles Co.-St. Mary's Co., Md.

Caroline Co.-Fredericksburg City-King George Co.-Spotsylvania
Co.-Stafford Co., Va.

Chaffee Co., Calif.

Claiborne Co.-Hamblen Co.-Hancock Co.-Hawkins Co., Tenn.

Clarendon Co.-Sumter Co., S.C.

Danville City-Henry Co.-Martinsville City-Pittsylvania Co., Va.

Darlington Co.-Dillon Co.-Marlboro Co., S.C.

Davidson Co.-Rowan Co., N.C.

De Soto Co.-Sarasota Co., Fla.

Franklin Co.-Jackson Co.-Williamson Co., Ill.
Gallatin Co.-Saline Co., Ill.

Kitsap Co., Wash.

Lee Co.-Van Buren Co., lowa

Lincoln Co., Mont.

Middlesex Co., N.J.

Montgomery Co.-Otsego Co., N.Y. (represented by 2 PSU's)
Orangeburg Co., S.C.

Reeves Co., Tex.

Reno Co., Kans.

Sheboygan Co., Wis.

 

fraction for women of other races was about 1
in 4,610 (see table C). The first step in pro-
ducing these disparate rates occurred in the
second stage of the sampling process: The
selection of 1970 census enumeration districts
(ED’s) within sample primary sampling units.
This was accomplished by stratifying ED’s by
the percent of the population that was black,
and by using a higher rate of selection in strata
with a 10-percent or greater black population.

Stratification of enumeration districts.—
Before the ED’s were stratified, a certain
amount of recombination was necessary. Al-
though the block groups (small ED’s used by the
U.S. Bureau of the Census in densely populated
areas) and ED’s averaged about 400 housing
units in size, there was considerable variation
about this mean. Thus the smallest ED’s and
block groups were combined to form ED’s that
contained 10 or more housing units each. For
 

 

SAMPLING FRAME OF PRIMARY
SAMPLING UNITS

 

 

 

 

The 18 largest All SMSA's except Non-SMSA
SMSA's the 18 largest areas
4 SMSA's <afemmmmm= Sorted into Sorted into
that could 19 strata 16 strata
not be
stratified
{ Y
Selection of one PSU from each stratum with a probability
Divided into proportionate to its size
21 PSU's
| 3
Selected Nonselected Selected
Recombination
into 19 super-
strata
Selection of one
stratum from each
superstratum with
a probability pro-
portionate to its
size
Selection of one PSU
from each stratum with a
probability proportionate
to its size
Y Y Y Y
21 SMSA self- 4 SMSA self- 19 SMSA 11 SMSA and 16 non-SMSA
representing representing nonself- 8 non-SMSA nonself-
PSU's PSU's representing nonself- representing
PSU's representing PSU's
PSU's

79 sample PSU's

 

Figure 1. First-stage selection of primary sampling units

 
Table C. Estimated number of women eligible for inclusion in
the sample, expected sample size, and approximate sam-
pling fraction required to produce the expected sample size,

 

 

 

 

 

by race
Number Expected | Approximate
Race of eligible sample sampling
women?! size fraction?
All races..... 31,444,000 10,000 1in 3,140
Blak iversssssirrenens 3,795,000 4,000 1 in 950
Other races............ 27,649,000 6,000 1in4,610

 

 

 

 

INumbers are estimates, produced by the National Center
for Health Statistics, of the number of women 15-44 years of age
who were either ever-married or never-married and who have
children of their own living with them. Estimates are for January
1976,

2Sampling fractions represent the overall proportion of
eligible women in each race group that would have to be selected
in order to produce a sample of the appropriate size and racial
composition. The various sampling rates used in the sample
design were intended to produce these overall rates after such
factors as subsampling in certain strata and nonresponse were
taken into account.

ease of discussion, block groups, ED’s, and re-
combined block groups and ED’s will be referred
to as ED’s.

The stratification of ED’s was accomplished
on the basis of information included in the 1970
Census of Population and Housing. These data
were used to determine the proportion of the
population of each ED that was black, and thus
to which stratum the ED should be assigned.
The strata were defined as follows:

Stratum 1: ED’s with a black population of
59.5 percent or more.

Stratum 2: ED’s with a black population of
29.5 10 59.499 percent.

Stratum 3: ED’s with a black population of
9.5 to 29.499 percent.

Stratum 4: ED’s with a black population of
0.0 to 9.499 percent.

Determination of size.—After this initial
stratification by race, each ED was assigned a
measure of size, which was the number of area
segments (groups of year-round housing units)
that it contained. The size of each area segment,
in turn, was a function of the ED stratum. The

average number of housing units per segment
was 15 in strata 1 and 4, 30 in stratum 2, and 60
in stratum 3. The measure of size for ED’s in
strata 1 and 4, for example, was thus the num-
ber of year-round housing units they contained
divided by 15. This ratio was then rounded to
the nearest integer for sampling purposes.

If an ED assigned to stratum 2 or 3 did not
have a number of housing units equal to at least
two-thirds of the average segment size for that
stratum, a second combining operation was
performed. The newly created ED’s were then
placed in the proper stratum, and a new measure
of size was computed. In both the first and
second combining operations two restrictions
applied: No more than four ED’s were to be
grouped together, and combined ED’s must all
be from the same county.

Sequencing of enumeration districts.—Fol-
lowing the assignment of a measure of size, all
enumeration districts were sorted by: (1) pri-
mary sampling unit number, (2) stratum code,
and (3) the Westat Economic Indexd The sort-
ing by primary sampling unit number had the
effect of grouping enumeration districts by re-
gion and by type of primary sampling unit
(self-representing SMSA’s, nonself-representing
SMSA’s, and non-SMSA’s) within a region. The
Westat Economic Index was sequenced within
strata in ascending order in the first and third
strata, and in descending order in the second and
fourth strata. The sequencing operation, com-
bined with the systematic sampling of ED’s that
was carried over from one primary sampling unit
to the next and from one stratum to the next,
produced the effective stratification of ED’s by
region, residence, and income. This stratification
was intended to achieve a small reduction in
sample variance, and it ensured a representative
sample of women by these variables.

dThe census ED file contains only information col-
lected on a 100-percent basis and thus does not include
income data. Westat, Inc. therefore imputed a median
income for each ED by creating a regression of income
on such items as value of home and rent paid, and apply-
ing the regression model to each ED. These imputed
values form the Westat Economic Index.
Selection of enumeration districts.—Prior to
the sampling operation, the measure of size for
each ED (number of segments) was inflated by
the inverse of the probability of selection for the
primary sampling unit that it was in. This
weighting has two effects: Households from all
PSU’s have exactly the same overall probability
of being included in the final sample, and a
representative sample can be achieved while
using the same sampling fraction across all
PSU’s.

A systematic sample of ED’s was drawn
using a random start and a sampling interval of
760 segments in strata 1 through 3, and 4,256
segments in stratum 4. Thus the segment corre-
sponding to the random start and every 760th or
4,256th segment thereafter was identified, and
the ED’s in which they fell were selected. A few
ED’s with an inflated measure of size that was
greater than the sampling interval were selected
more than once.

Only one random start was needed. The
systematic sampling was carried over from one
stratum to the next within primary sampling
units and from the last ED selected in one
primary sampling unit to the first ED in the
next.

Third-Stage Selection of Segments

Initial subdivision of segments.—Each enum-
eration district was divided conceptually into
the number of segments that had been deter-
mined during the second stage of the sampling
process. When the ED was from an area for
which census block statistics were available,
block maps and census counts were used in the
first step of this subdivision. Each block (or
occasionally, groups of contiguous blocks) was
then assigned one or more segments based on
the number of dwelling units it contained and
on segment size (determined by the ED
stratum).

When block statistics were not available,
census county and place maps were used. If the
physical features shown contained enough infor-
mation, the maps were used to accomplish the
segmentation. Where these maps were inad-
equate, one of two alternative procedures was
used. Wherever U.S. Department of Agriculture
aerial photographs (taken about 1970) were
available, a cartographer was employed to do the

segmentation. In other areas field interviewers
were sent to the sites to prepare maps for this
purpose. However, the physical features of these
rural areas did not always allow the ED’s to be
divided into segments of the desired size. Thus,
as with block areas, these initial subdivisions
sometimes included more than one segment.

Selection, field listing, and subsampling of
area segments.—Regardless of the method used
to divide each enumeration district into land
areas, one area was randomly selected from each
district with a probability proportionate to the
number of segments it contained. If the area
contained only one segment (had a measure of
size of 1) it became a sample segment. However
some areas had a measure of size that was
greater than 1; that is, they included more
dwelling units (DU’s) than was appropriate for a
single segment. In these cases the subsampling of
dwelling units was employed to produce sample
segments of the proper size.

The first step in this subsampling process
was the field listing of area segments. This ad-
dress listing operation was not limited to seg-
ments that were too large, however, since all
segments in strata 1-3 were later to be subsam-
pled by race, and also because the listing pro-
vided a more accurate count of dwelling units
than did census data.

As interviewers were recruited for the sur-
vey, they received training as listers and were
assigned three segments each. For each segment
the lister received a ‘segment folder” that
included a census map, two sketch maps, and
address listing sheets. She then proceeded to the
segment, verified the segment boundaries, cor-
rected the maps when necessary, and listed the
address of each dwelling unit in the segment (or
recorded a detailed description of the unit where
the house or apartment number was not evi-
dent). The listing was also verified through the
independent relisting of one out of three seg-
ments (and another relisting if large discrepan-
cies were found), and through checks for missed
dwelling units during the interview stage of the
survey.

For exceptionally large segments, however,
an additional operation was introduced prior to
the listing of addresses. The maps for these seg-
ments were used to prepare sketch maps con-
taining further subdivisions of the areas. The
resultant subsegments were then assigned a
measure of size based on approximate DU
counts, and the maps were returned to Westat,
Inc. for the selection of one subsegment for list-
ing and inclusion in the final sample.

After the listing of addresses, segments that
contained more than the required number of
DU’s were subjected to a systematic sampling
procedure. Thus a segment that included twice
the number of DU’s dictated by the stratum
from which it was drawn, for example, had every
second DU on the address listing sheets desig-
nated for interview.

New construction segments.—Approximately
90 percent of the respondents included in the
final sample lived in area segments that were
selected this way; however, about 10 percent
lived in new construction segments. Although
area segments would have provided an unbiased
sample of new construction units (dwelling units
for which building permits were issued during
the period September 1969-July 1975), it was
thought that such a procedure would signifi-
cantly increase sample variances. This would
occur since the measure of size used to define
area segments was based on 1970 census data
and thus would not reflect the true number of
DU’s in segments containing units built after
1969. Since new construction tends to occur in
large clusters, and thus in a small proportion of
~ area segments, a small, highly clustered subset of
new DU’s would be obtained through area
segment selection. Consequently, new construc-
tion segments were selected in a separate sam-
pling step whenever possible. However, 15-20
percent of new construction (according to cen-
sus figures) occurred in areas where no easily
accessible file of building permits (the bases of
new construction segments) existed, and for
these localities new construction was picked up
in the area segments along with older units.

The new construction segments were defined
to be 15 DU’s in size (the same size as area
segments from enumeration districts in strata 1
and 4) and were systematically sampled at the
rate of 1 in 4,256 (the same as stratum 4 area
segments). The sample selection procedures were
slightly different for DU’s with permits issued
during the part-years September-December
1969, and January-July 1975, than for those
with permits issued during the period 1970-74.

The U.S. Bureau of the Census produced
magnetic data tapes that contain annual totals of
the number of housing units authorized for
1970-74 for each permit-issuing place. These
tapes were used to draw the sample of new con-
struction DU’s with permits issued during
that period. The file was first sorted by primary
sampling unit (the same 79 PSU’s selected in
the first stage), by permit-issuing place within
each PSU, and by year of permit within each
place. The total authorized number of units for
each place was then calculated for each year,
and this number was divided by 15 (the average
segment size) and rounded to the nearest in-
teger. The result was used as the measure of size
for each “place-year” on the tapes and is
analogous to the measure of size assigned to
enumeration districts during their second-stage
selection. This measure was next weighted by
the inverse of the probability of selection for the
PSU in which the place was located, and a
systematic sample of place-years was drawn at
the rate of 1 in 4,256 segments. Just as during
the selection of enumeration districts, this sam-
pling was accomplished using a single random
start and continuous sampling across all PSU’s.

After the sample of place-years was drawn, a
measure of size was assigned to every month
within each place-year. This measure of size
(number of segments of 15 building permits
each) was determined by published census data
for larger permit-issuing places,® and through
contact with building permit personnel for the
remainder. One month, or in some cases a group
of adjacent months, was then randomly selected
from each place-year with a probability propor-
tionate to size, an operation that is analogous to
the selection of area segments from enumeration
districts.

For 1969 and 1975, census data tapes could
not be used for the selection of place-years since
only part of each year was included. Monthly
census reports? and reports from permit-issuing
offices, used in the subsampling of place-years
for 1970-74, were thus used to directly sample
place-months. To this end, a measure of size
(number of segments) was assigned to each
month for each place, and the measure was
weighted by the inverse of the probability of
selection for the PSU in which the place was
located. The months were then sequenced by
chronological occurrence, PSU, and place, in
that order, and a systematic sample of place-
months was drawn at the rate of 1 in 4,256
segments.

The new construction segments formed by
selected place-months (whether selected directly
or from place-years) were next subjected to
listing and subsampling operations in order to
produce sample segments of the proper size.

Listing and subsampling of new construction
segments.—The listing of new construction seg-
ments (hereafter referred to as stratum 5 seg-
ments) was done in a manner substantially
different from that for area segments from strata
1-4. For stratum 5 segments, the listing was
accomplished using the file of building permits
in each selected building permit office. The lister
was required to obtain access to the file of
permits for the selected month or months,
determine the number of dwelling units repre-
sented by each permit, and list all dwelling units
on an address listing form similar to that used
for area segments. The DU’s were then concep-
tually grouped into segments so that the first 15
encountered formed one segment, the next 15
another, and so on until all DU’s had been
exhausted. This procedure was used so that the
segments thus created formed geographically
compact clusters and resulted in reduced inter-
view costs. Following segmentation a random
sample of segments was drawn at the rate
necessary to produce the final self-weighting
sample of 1 in 4,256.

The sample design for new construction
segments had a number of desirable characteris-
tics. The most important of these is that all new
construction units in the United States had the
same overall probability of being included in the
final sample. Further, that probability was 1 in
4,256, the same probability established for DU’s
from stratum 4. The method also yielded seg-
ments of 15 dwelling units in size (identical to
stratum 4) and produced a maximum spread
among places within the 79 sample PSU’s.
Figure 2 illustrates the selection process for both
new construction and area segments.

Integration of area and new construction
segments.—The listing and subsampling of area
and new construction segments produced ap-
proximately 1,700 segments, consisting of about
32,000 households. In order to provide an

10

unbiased sample of all households, it was then
necessary to integrate households from the new
construction segments with the rest of the
sample in such a way that all households had a
known, non-zero probability of selection. To
this end, reports published by the Bureau of the
Census® were used to determine which area
segments were in localities that did not require
building permits. In these segments interviewers
were instructed to include all households, both
new and old construction, in the sample. Thus
the probability of selection for households in
any segment in a non-permit locality was not
conditional on the number of new dwelling units
the segment contained.

In the remaining areas, where new construc-
tion segments were sampled separately, all new
units were excluded from the area segment
sample. This operation precluded the possibility
of new units being sampled twice and ensured
that the probabilities of selection for old and
new units were independent. The exclusion of
new units was accomplished by including a
question on when the housing unit was built in
the Household Screener questionnaire (see ap-
pendix II).

Missed dwelling units.—During the interview
phase of the survey, interviewers were required
to check for dwelling units not discovered
during the listing operation. These dwelling units
may have been missed as the result of the lister
having overlooked either entire structures or
individual DU’s within listed structures. The
procedures for locating and including missed
DU’s in the sample were designed to give them
the same probability of selection as other DU’s
from segments in the same stratum. The proce-
dure was as follows:

1. At each address that had been listed as a
single-family home, the interviewer
checked to ensure that all persons living
at that address were included in the
screening. Any additional dwelling units
located at that address were also in-
cluded in the sample.

2. In multiunit structures, the interviewer
looked for missed dwelling units within
the structure only if the first-listed unit
in the building was included in the sam-
ple. The check included questioning re-
 

 

NEW CONSTRUCTION SEGMENTS AREA SEGMENTS

New construction building permits Sample enumeration
issued within the 79 sample PSU's districts (ED's) N

to nearest integer) assigned Selection of one area from each
to each year within each ED with a probability propor-
place (place-year) tionate to the number of segments
| (groups of dwelling units) it
contained

Permits for complete Permits for part-years Urban ED’s Rural ED’s
years 1970-74 1969 and 1975 |
Sorted by PSU, permit- Measure of size (number Subdivision of Mapping
issuing place within PSU, of segments = number of ED’s using block of ED’s
and year of permit permits/15 and rounded to statistics
within place nearest integer) assigned to
each month within each
place (place-month) Subdivision
Measure of size (number USING Maps
of segments = number of |
permits/15 and rounded |

Measure of size weighted
by inverse of the prob-

ability of selection for the
PSU in which the place was
located Subdivision of very large
| areas and the selection of
| one subarea for listing

Systematic sample of place- Place-months sequenced
years at the rate of 1 in by chronological occurrence,
4,256 segments PSU, and place
Field listing of
areas

Measure of size assigned to
months within place-years

Selection of one place-month Systematic sample of
from each place-year with place-months at the rate
probability proportionate of 1 in 4,256 segments

to size

Subsampling of areas that
contained more than one
segment

Listing and subsampling
of place-months to produce
segments of the appropriate

size

SAMPLE SEGMENTS
(groups of dwelling units)

 

 

Figure 2. Third-stage selection of sample segments

11
spondents about possible missed dwell-
ing units and looking at mailboxes or
doors that might give indications of a
unit not listed. Any additional dwelling
units located in this manner were in-
cluded in the sample.

3. In segments, missed structures were
searched for only if the first-listed struc-
ture in that segment was included in the
sample. When missed structure checks
were initiated, the procedure included
canvassing the segment by foot or by
car, using the segment map and listing
sheets to look for additional structures,
and querying respondents about possible
missed structures in the area. Any dwell-
ing units located in missed structures
were included in the sample.

The procedure of having the first-listed unit
determine whether a missed DU or missed
structure check was made ensured that the
self-weighting feature of the sample was re-
tained. This was the case since the check was
made for all segments and structures in which all
the DU’s were included in the sample, for half of
the segments and structures in which the sub-
sampling rate had been 1 in 2, for one-third of
the segments and structures in which the sub-
sampling rate had been 1 in 3, and so on. The
only deviation from this self-weighting principle
occurred when five or more missed DU’s were
located in a single missed structure or at a DU
listed as a single unit. In these cases the
interviewer contacted the sampling department
at Westat, Inc., which then selected four of the
DU’s for interview and inclusion in the sample.
About 700 DU’s were included in the sample as
a result of the missed DU checks.

Fourth-Stage Selection of Households

Subsampling for race.—During the second
stage of the sampling process, a higher sampling
rate was used for the selection of enumeration
districts (ED’s) in strata 1-3 than for those in
strata 4 and 5. The higher rates in strata 1-3
(ED’s where 9.5 percent or more of the popula-
tion was black) were intended to produce a
higher probability of selection for black women
than for women of other races and thus yield

12

the approximately 4,000 sample black women
(out of a total sample of about 10,000) required
by the National Survey of Family Growth design
specifications. However, since about 45 percent
of eligible women in strata 1-3 were expected to
be of races other than black, this procedure
alone would also have produced an unnecessarily
large sample of these women. Consequently, the
fourth-stage selection of sample households in-
volved a subsampling operation designed to
reduce the overall sampling rate for these
women and produce a sample of the desired size.

This subsampling operation included the use
of a computer to designate a portion of sample
dwelling units in those strata as subsample units.
All households containing eligible black women
became sample households, but households con-
taining eligible women of other races were
included in the sample only if they were found
in designated subsample units. The interviewers
received instructions on the Household Screen-
ers as to which households were to include any
eligible woman and in which ones only black
women were to be included.

The initial sample design called for the
subsampling of households in strata 1-3 at the
rate of 1 in 5. This rate would have produced
approximately equal sampling rates for women
of races other than black across all 5 strata: 1 in
3,800 in strata 1-3, and 1 in 4,256 in strata 4
and 5. However, after the fieldwork was well
underway and the 1 in 5 subsample already
selected, a decision was made to increase the
subsampling rate to 13 in 20. This decision was
necessary because significantly fewer of these
women than expected were being included in
the final sample. This undercoverage resulted
primarily from two factors: fewer-than-expected
dwelling units in strata 1-4 (see table D), and a
population shift that produced fewer-than-
expected eligible women of races other than
black in strata 1-3.

Since the fieldwork was more than half
completed under the first regime (a 1 in 5 sub-
sample), the increased subsampling rate was
achieved through a supplementary sample se-
lected from households with eligible women that
were rejected in the initial subsample. The way
in which this supplementary sample was drawn
was dictated by the fact that the entire National
Survey of Family Growth sample had previously
Table D. Total and expected number of dwelling units (DU’s) in
the United States, and percent difference, by stratum

 

 

 

 

 

Total Expected Percent
Stratum DU’s in DU’s in ditterence
thousands! | thousands
All strata.......... 72,509 74,725 -3.0
Stratum 1 5 4,612 5,370 -14.1
Stratum 2 so 3,096 3,420 -9.5
Stratum 3 4,030 4,940 -18.6
Stratum 4 52,306 52,995 -1.3
Stratum 5 8,465 8,000 +5.5

 

 

 

 

 

1Estimates are based on the total number of DU’s assigned
for screening (including those from missed DU checks) and are
the product of the number assigned and the rate at which the
DU’s were sampled.

been divided into four random subsets, which
were assigned for interviewing in sequence. At
the time the decision was made to increase the
subsampling rate the third-period interviewing
was almost completed, and the fourth-period
interviewing was well underway. Thus a com-
plete list of households with eligible women of
races other than black that had been rejected in
the initial subsampling during the first three
periods was soon to be available. The supplemen-
tary sample of households was easily obtained
and involved only the removal of the message:
“Conduct extended interview if black only”
from the label on the screeners for 4 out of 5 of
strata 1-3 households. If the household had
already been screened and rejected because the
respondent was not a black woman, the case was
returned to the field. If the case had been
returned as a nonresponse, it was included in the
followup effort.

T'o supplement the fourth period would have
required either changing methods during on-
going fieldwork or waiting until the fourth-
period interviewing was completed under the old
regime and then instituting the same procedures
used for periods 1-3. Neither alternative was
desirable because each involved too much poten-
tial for error or delay. Thus the entire supple-
mental sample effort was restricted to the first
three periods of interviewing. Since households
had been randomly assigned to each of the four
periods, the probability of selection for women
of races other than black was constant and not
dependent on interview period. Because about

three-quarters of the strata 1-3 households was
subjected to a 4 in 5 subsample, and about
one-quarter of - these households was subjected
to a 1 in 5 subsample, the overall subsampling
rate achieved was the following:

13
20

2
—_— 4
4 5

|
ot | =

Fifth-Stage Selection of Sample
Persons

Subsampling when there was more than one
eligible respondent in the household.—To avoid
the high correlation of information from eligible
women within the same dwelling unit, the
sample design stipulated that no more than one
eligible woman from any sample DU would be
interviewed. During completion of the House-
hold Screener, the interviewer listed all residents
of the DU on one page of the form and relisted
the eligible females in order of age in item S-14
on another page (see appendix II). Item S-14
provides space for listing up to six persons
because six was considered to be a reasonable
limit for the number of eligible females to be
expected from a single DU. When the inter-
viewer listed more than one eligible woman, she
referred to the sampling table at the beginning
of the questionnaire to determine which person
she was to interview. The sampling table consists
of five numbers that designate which person to
interview when the number of eligible women is
two, three, four, five, or six or more (see table
E). Westat, Inc. home office personnel filled in
the table on every Household Screener by
systematically assigning 12 versions of the table
to the screeners. This method of assigning
interviews gave each eligible woman in a given
household the same probability of being
selected.

Nonresponse Followup

Subsampling of nonresponse cases.—An elite
corps of traveling interviewers and assistant
supervisors was assigned the task of obtaining
interviews at sample households included in the
followup effort. A household was assigned to
this operation after all efforts to complete an

13
Table E. Example of sampling table on the Household Screener

 

 

 

 

 

 

IF NUMBER OF THEN INTERVIEW
ELIGIBLE FEMALES PERSON .LISTED
LISTED IN SUMMARY ON SUMMARY BOX
BOX IS: J LINE !
Two 1
Three 3
Four 1
Five 4
Six or - 3
More

 

 

 

 

interview (either a screener or an extended
interview) were exhausted by local interviewers.
Nonresponse cases that appeared to be non-
convertible refusals were excluded, however.

Prior to the followup interview effort, a
50-percent subsample of these cases was selected
using systematic cluster sampling procedures.
This subsampling, designed to reduce interview
costs, was accomplished in two ways. In large-
city PSU’s, where large numbers of nonresponse
cases were expected, the households were
grouped by segment, the segments were se-
quenced in descending order by the number of
followup cases they contained, and a systematic
sample of half of the segments was drawn. For
the remaining PSU’s, nonresponse households
were grouped by PSU, the PSU’s were se-
quenced in descending order by the number of
followup cases they contained, and a 50-percent
sample of PSU’s was selected systematically.
Approximately 4.5 percent of the respondents
who were included in the final sample came in as
a result of this followup effort. They represent
about 9 percent of the population of respond-
ents (since they were subsampled at the rate of 1
in 2) and thus are weighted double relative to
other women in the sample (see table F for the
percents of women selected at various sampling
rates).

14

Table F. Percent of sample women, by race and sampling rate

 

 

 

 

 

Race
Sampling rate!
Black Other
races
Percent
Tin TBO csciscsivinnivervsnmmsinssmsssiunivessorsesnses raisins 89.5
1i 4.3 are 4
1i sist 25.1
1i Ge 0.8
1i 5.9 70.5
1i 03 3.6

 

 

 

 

1The sampling rates exclude the effects of subsampling
multieligible households.

CHARACTERISTICS OF THE SAMPLE

Response Rates

The first four stages of the sample design
(plus the missed dwelling unit procedure) re-
sulted in the identification of 32,683 sample
DU’s. However, during the fifth-stage screening
operation and prior to the subsampling of
nonresponse cases (see preceding discussion), in-
terviewers discovered that 5,306 of these were
either vacant or not a DU as defined by the
National Survey of Family Growth. Screeners
were completed for 23,734 of the remaining
27,377 occupied DU’s, and 8,226 extended
interviews were completed among the 9,064
DU’s found during screening to contain an
eligible woman (see table G).

Prior to the intensive followup effort to
convert the nonresponse DU’s, a 50-percent
subsample of these DU’s (excluding non-
convertible refusals) was drawn. A total of 3,946
DU’s was held for subsampling and 1,958 were
selected. The final disposition of these cases is
shown in table G.

After the followup effort was completed,
30,695 DU’s remained in the sample (the orig-
inal 32,683 minus the 1,988 held for subsam-
pling but not selected). Screener interviews were
completed at 24,188 of the 25,297 DU’s found
to be legitimately occupied DU’s, producing a
fifth-stage sample of 9,470 women eligible for
an extended interview. Extended interviews
were completed for the 8,611 women compris-
Table G. Number of dwelling units (DU’s), by nonresponse subsample status and final disposition

 

 

 

 

 

 

 

 

 

 

 

Subsample status
Final disposition Before in Lffian g
PIRsIOnse NONIESPONsE no ¥ Sen;
subsample subsample subsample
drawn drawn
All DU's
TOR... cscinsnnisssisrsssrmimmsmis irs SE EE HAASE A ER ETAT TSR T STATA S TARAS AEE 32,683 1,958 30,695
Vacant or nota DU............. 5,306 92 5,398
Legitimately occupied DU 27,377 1,866 25,297
Disposition of legitimately occupied DU’s

COMPILE SETORNBIINIEIVIOW ..couurssucrrssisisismsmmisseriieninsissssnssmmssvesessnsisnsssssarssnnnssnsunnsissunsives 23,734 1,292 24,188
Eligible for extended interview 9,064 732 9,470
Completed extended interview 8,226 385 8,611

 

 

 

ing the final National Survey of Family Growth
sample (see table G).

When calculating screener, interview, and
combined response rates, the nonresponse sub-
sampling procedure must be taken into account.
For example, to ignore the subsampling proce-
dure entirely would mean that the 1,988 DU’s
held for subsampling but not selected would
enter the denominators of the rates even though
they did not receive the complete effort to
convert them to response DU’s (which enter the
numerators of the rates). Thus the procedure
would negatively bias the rates. On the other
hand, to simply delete the 1,988 nonselected
cases from the calculations would positively bias
the rates. That is, if the unselected DU’s had
been subjected to the same followup efforts as
the selected DU’s, and if both groups were
included in the calculation of response rates, the
rates would probably be lower than if the
unselected DU’s were simply excluded from the
calculations. Response rates calculated from
table G provide supporting evidence for this
position. For example, the screener response
rate (the ratio of completed screeners to oc-
cupied DU’s) for the nonresponse subsample is
69.2 percent while the comparable rate for the
full sample before the subsample was drawn
(and which thus includes all the nonresponse
cases) was 86.7 percent.

In order to overcome these problems the
response rates presented in this report are

weighted response rates. That is, they are based
on numbers of DU’s in which subsampled DU’s
are weighted double relative to others. This
weighting allows the subsampled DU’s to repre-
sent not only themselves but also the DU’s held
for subsampling but not selected. Thus the
weighted rates are those that would have oc-
curred if there had been no nonresponse subsam-
pling and the DU’s held for subsampling but not
selected had been converted to response DU’s at
the same rate as subsampled DU’s.

The weighted numbers of DU’s from which
the response rates are calculated are found in
table H. The weighted response rates are found
in table J, which also contains weighted response
rates for each of the five first-stage selection
strata. The screener response rate is the percent
of occupied legitimate DU’s for which a screener
interview was completed. The interview response
rate is the percent of eligible women for whom
completed interviews were obtained. The com-
bined response rate is the product of the
screener and interview completion rates divided

by 100.

Sample Size

The 8,611 women included in the final
sample are nearly 14 percent less than the
desired sample size of 10,000. Further, this
undersampling is greater among black women
than among women of other races; the number

15
Table H. Weighted number of dwelling units (DU’s), by final

 

 

 

 

 

disposition

Weighted

Final disposition number

of DU's!

All DU’s
TR OHAE, Lois tovtsviescrsinesssssietnrensnttne ssn sies vilunsndbenins 32,653
Vacant or not a DU .......ccmmmmniiininmsunmssmrises 5,490
Legitimately occupied DU 27,163
Disposition of legitimately occupied DU's

Completed screener interview. .........ueuueeeeeereereeeenanns 25,480
Eligible for extended interview........ccccccccceiinieneennens 10,202
Completed extended iNterview.........ccceeeeerernrreennnnns 8,996

 

 

1pU’s included in the nonresponse subsample are weighted
double.

Table J. Weighted screener, interview, and combined response
rates, by stratum

 

 

 

 

 

 

 

 

 

 

Screener | Interview | Combined
Stratum response | response response
rate rate rate
Percent

AN STata .cocrrerrines 93.8 88.2 82.7
Satu 1..couisissasisvsssennss 92.3 87.8 81.0
Stratum 2..... 95.0 87.5 83.1
Stratum 3..... 94.2 85.2 80.3
Stratum 4..... 94.4 89.4 84.4
Stratum B....... 90.8 89.0 80.8

 

 

1The results in this table are weighted for the subsampling
of nonresponse cases.

of sample black women was about 25 percent
below the expected total compared to an under-
sample of about 7 percent for women of other
races. However, much of this racial differential is
due to the decision, made after data collection
was well underway, to increase the subsampling
rate for women of other races in strata 1-3, thus
increasing the overall sampling rate for these
women and reducing the extent to which they
were underrepresented.

The unexpectedly small sample is due to
several factors, the most important of which is
the subsampling of nonresponse followup cases.
Since 50-percent subsampling was used for these

16

cases, each woman brought into the sample as a
result of the followup effort actually represents
two nonresponse cases. Allowing these women
to assume a double weight, the number of
women in the sample becomes 8,996 (nearly 90
percent of the expected total), of which 3,161
are black women (about 79 percent of the
expected total) and 5,835 are women of other
races (about 97 percent of the expected total).

Other important factors are the following:

1. Higher-than-expected nonresponse rates.
The expected combined response rate
was 90 percent compared to an actual
rate of 82.7 percent.

2. Fewer-than-expected dwelling units in
some strata. This factor may have been
the result of higher-than-expected rates
of abandonment and demolition - of
structures in some areas (see table D).

3. Missed eligible women within some
dwelling units.

4. Population movement across strata. This
problem is especially critical for black
women since they were sampled at a
much different rate in strata 1-3 than in
strata 4 and 5. It is estimated that about
670,000 black women had moved from
areas where they would have been sam-
pled at the rate of 1 in 760 to areas
where they were sampled at the rate of 1
in 4,256. This factor alone could have
reduced the sample of black women by
about 725.

ESTIMATION
Weighting Procedures

Overview.—Since the National Survey of
Family Growth (NSFG) is designed to produce
unbiased estimates for the entire population of
eligible women in the conterminous United
States, the sample data must be inflated to the
level of the population from which the sample
was drawn. The inflation factor, or weight, for
each woman is the product of several adjust-
ments, including one or more for each stage of
sampling. Three types of adjustments are in-
volved: inflation by the reciprocal of the proba-
bilities of selection, nonresponse adjustment,
and poststratification adjustment.

Inflation by the reciprocal of the probabil-
ities of selection.—The weight for each woman is
the product of the reciprocals of the probabil-
ities of selecting (1) the primary sampling unit,
(2) the enumeration district, (3) the segment,
(4) the household, and (5) the eligible sample
person.

Nonresponse adjustment. —Each sample
weight is adjusted for nonresponse to the House-
hold Screener (screener nonresponse) and non-
response to the detailed NSFG questionnaire
(interview nonresponse). These adjustments are
necessary because the phenomenon of nonre-
sponse introduces bias into any probability
sample; women with certain characteristics are
more likely to be nonrespondents than other
women; thus the former are likely to be under-
represented in the sample. Nonresponse adjust-
ments minimize the impact of this bias. Adjust-
ment for screener nonresponse was achieved by
imputing to women in nonresponding house-
holds the characteristics of women in responding
households in the same primary sampling unit
and stratum. Adjustment for interview nonre-
sponse was achieved by imputing to nonrespond-
ing women the characteristics of responding
women in the same age-race class and primary
sampling unit.

Poststratification by age and race.—The
weight for each ever-married respondent is multi-
plied by a poststratification adjustment factor
that is determined by the woman’s age and race.
The 12 adjustment factors shown in table K
make NSFG estimates of ever-married women in
each age-race class equal to independent control
totals for April 1976, the approximate midpoint
of data collection. The control totals are based
on data from the Current Population Survey
(CPS) conducted by the U.S. Bureau of the
Census. No poststratification adjustment is ap-
plied to the weights for never-married mothers
because reliable control totals are not available.
Poststratification achieves much of the improve-
ment in precision that would have been attained
if the sample had been drawn from a population
stratified by age and race. A discussion of the

Table K. Poststratification adjustment factors, by race and age

 

 

 

 

 

 

 

Race
Age

Black | Oo

races
15-19 years 1.457 0.937
20-24 years 1.078 1.134
25-29 years 0.882 1.079
30-34 years 1.078 1.212
35-39 years 1.000 1.119
40-44 years 1.156 1.106

 

1Ratio of May 1976 population control totals, based on
Current Population Survey data, to National Survey of Family
Growth weighted estimates.

method used to compute the CPS control totals
and a comparison of simple inflation and post-
stratification estimators can be found in another
report.!

Estimating Equation

The Cycle II estimator of an aggregate
parameter Y for all women in the NSFG target
population is

Y=Y, +7,

where Y] is the estimated total for ever-married
women and Y} is the estimated total for never-
married mothers.

 

Ba
3 .
Y; T 3 Yet x’
o=1 al

where

Y,, =the weighted, nonresponse-adjusted
estimator of Y for ever-married
women in age-race class a. The 12 «
classes are defined by 2 race groups—
black and other races—crossed with 6
age groups: 15-19, 20-24, 25-29,
30-34, 35-39, and 40-44.

X,,, =the weighted, nonresponse-adjusted
estimator of the number of ever-
married women in class a.

= an independent estimate of the num-
ber of ever-married women in class &
based on data from the Current Popu-
lation Survey.

Xx

al

17
12
Zl ses ’
Y, = Y Yoo
o=1

is the sum of the corresponding weighted,
nonresponse-adjusted estimates for never-
married women.

The weighted, nonresponse-adjusted esti-
mator for ever-married women in age-race class
ais given by

5 79
Yi =4® ba ts, Sy (er

g=1 h=1
Len Hop;
: 2 Wogni * Wagni * Rani * > D, * Fy
= 7=1
“Lowy Yeahs bon) )
where

S., § Household Screener.

4 _ adjustments for nonresponse to the

I, § personal interview.

W,, = first-stage weight = reciprocal of the
probability of selecting primary sam-

pling unit A.

B, b= adjustments for nonresponse to the
h

Wogni = second-stage weight = reciprocal of
the probability of selecting second-
stage unit 7 within stratum g and pri-
mary sampling unit A. Strata 1-4 are
the race strata, stratum 5 is the new
(post-1969) housing stratum.

Wags; = third-stage weight = the reciprocal of

* the probability of selecting the sam-
ple segment from second-stage unit i.

the reciprocal of the large-segment
subsampling rate:
Rggp; = { 2 for certain segments with large
numbers of households;
1 for all other segments.
the reciprocal of the household-
race subsampling rate:
Do 20/13 if g=1, 2, or 3 and the re-
& spondent in the household is not
black;
1 otherwise.

 

18

the reciprocal of the nonresponse

subsampling rate:

2 if household j, or the sample

women therein, was in the nonre-

sponse sample;

1 otherwise.

Tygon; = the number of eligible women in
household ; of the segment in
second-stage unit z.

Y,; = the value of characteristic Y for the
sample women from household j.

1 iI’ the respondent has ever been

5 married and belongs to age-race

class a;

0 otherwise.

L,, = the number of second-stage units in
stratum g and primary sampling
unit A.

Hy; = the number of completed interviews
from households in the sample seg-
ment from second-stage unit z.

Fos ™

Y,o is exactly the same as YJ, except that

8 oni = 1 for never-married women in class «.
VARIANCE ESTIMATION
Background

The balanced half-sample replication tech-
nique described in detail in other National Cen-
ter for Health Statistics reports*:5 is used to
estimate National Survey of Family Growth
(NSFG) variances. An empirical study by Bean®
gives evidence that the half-sample technique
produces highly reliable, essentially unbiased
variance estimates.

Three important practical reasons why half-
sample replication is being used are as follows:

1. Programming difficulties are reduced be-
cause half-sample variances are computed
by taking a simple average of squared
deviations of half-sample estimates from
the estimate based on the full sample.
Instead of having to program an exceed-
ingly difficult variance formula, the pro-
grammer must simply adjust the estima-
tion formula to compute estimates from
appropriately chosen half samples.
2.. The complete algebraic formula for
NSFG variances is unknown because of
the complexity of the design. Although
algebraic expressions can be derived for
particular subprocedures—such as the
individual stages of sampling and the
poststratification and nonresponse ad-
justments—a single, exact variance equa-
tion has not been developed.

3. As stated by McCarthy# : “Variance esti-
mates based upon the replicated esti-
mates will mirror the effects of all
aspects of sampling and estimation that
are permitted to vary randomly from
replicate to replicate.” Also, replicated
half-sample variances include some of
the variability due to nonsampling
(measurement) error, as well as sampling
variability.

Summary of Applicable Theory

The population of interest is classified into
L strata, and two sample primary sampling units
(PSU’s) are drawn from each stratum. Selection
of exactly two sample PSU’s reflects an essential
element of the theory. This requirement may be
met for noncertainty PSU’s by collapsing two
strata having one PSU each, or for certainty
PSU’s by creating two artificial, or pseudo,
PSU’s by random methods from a single PSU.
The collapsing method produces somewhat posi-
tively biased (overstated) variance estimates by
introducing a between-stratum component of
variance that does not exist.’

Let the parameter of interest be denoted by
Y, for which an estimate Y’ has been obtained
from the complete sample. If Y' is a linear com-
bination of the sample observations, it can be
shown that Y' is an unbiased estimate of Y.
However, several empirical investigations indi-
cate that the bias of half-sample variance esti-
mates for certain ratio estimators and correla-
tion statistics is negligible, if detectable at
all 4,5,8,9

A half-sample replicate is defined as a collec-
tion of L primary sampling units obtained by
selecting one of the paired sample PSU’s from
each stratum. If the PSU’s within each stratum
are designated by the subscript ¢ =1 or 2 and

there are K half samples, where K 2 L, the pat-
tern may be summarized as in table L. The “+”
indicates that a PSU falls into a particular half
sample, and the “-” indicates that it does not.

Analogs of Y' corresponding to each half
sample are then computed. That is, for the kth
half sample, Y” is given by .

L
+2 ¥Y 7,
h=1

where 7 = either 1 or 2 depending on which PSU
of the stratum is the half-sample &, and Yj}; is,
in this example, a total. The estimator Y' is

L
® > (Yar + Yao)
r=

and its variance is estimated by

Because it is impractical to compute the Y},
for the entire set of 2L possible half samples
when L is large, a subset of half samples is se-
lected to produce the estimates. A set of side

Table L. Example of a half-sample replication pattern

 

 

 

 

 

 

Stratum
Half-
saniple 1 2 3 L
replica. psu | psu | Psu |...| psu
tion
1 2.41 2 2 1 2
1 +|-1-1+]-]+ + -
2 -l+]-1+] +] - - +
3 - | ++] -} -1 + ~ +
K “| -{+1-f +1 - + -

 

 

 

 

 

 

 

 

 

 

19
conditions relating to the selection of PSU’s for
the half samples has been developed by
McCarthy,*5 based on work by Plackett and
Burman!® and Gurney.!! These side conditions
greatly increase the stability of 3 by elimina-
ting a between-strata component of variance
that is otherwise present. The value of £3
obtained from a subset of half samples that is
chosen according to the McCarthy criteria is
equal to the value that would be obtained using
all 2L half samples. A set of half samples that
satisfies the McCarthy criteria is called a “bal-
anced set,” and the procedure is referred to as
“balanced half-sample replication.”

Application to the National Survey
of Family Growth

As a first step in applying the balanced half-
sample replication technique, the National Cen-
ter for Health Statistics grouped the 79 Cycle II
primary sampling units into 37 replicate strata.
Eighteen of the strata were self-representing;
each consisted of the PSU or PSU’s associated
with a single self-representing SMSA (see table
B). Within each of these strata two pseudo-PSU’s
were created by: (1) listing the PSU’s in numeri-
cal order within each stratum (for multi-PSU
strata); (2) listing the sample segments in numer-
ical order within each PSU; and (3) systemati-
cally dividing the segments into two groups,
with the first segment and every second-listed
segment thereafter assigned to the first pseudo-
PSU, and the remaining segments assigned to the
second pseudo-PSU.

The remaining 19 strata included the noncer-
tainty PSU’s. Their composition was dictated by
the way in which these PSU’s were originally
selected for inclusion in the sample. As described
previously, this selection was accomplished in
two stages: First, the PSU’s were stratified and
35 PSU’s were systematically selected; second,
all PSU’s (both selected and nonselected) were
recombined into 19 superstrata and 1 supple-
mental PSU was selected from each. This design
produced two independently chosen observa-
tions for each superstratum (the original selec-
tion or selections and the supplemental selec-
tion) and made the superstrata ideal replicate
strata. By pairing the original selections from
each strata with the supplemental selection, two

20

pseudo-PSU’s were created for each. There was
no need for a “collapsing” of population strata
to form replicate strata; the common problem
of extraneous between-stratum variance in the
half-sample variance estimates was avoided.

Within each of the 37 replicate strata, a
value of 1 was assigned to one of the pseudo-
PSU’s, and a value of 2 was assigned to the
other. Forty half samples were then created by
selecting one of the two alternative values within
each strata for each half sample; the values
chosen were determined by the elements of an
orthogonal 40 X 40 matrix of 1’s and 2’s
adapted from Plackett and Burman.10

In order to estimate the variance of an aggre-
gate statistic Y', 40 half-sample analogs of Y’
were computed. The formula for the kth half-
sample estimate is

ha !
Y. =¥tYy,

where Y}, is the half- sample estimate for ever-
married women and Yj, is the half-sample esti-
mate for never-married mothers. The half-sample
estimates Yj, and You correspond to the full-
sample estimates Y] and Yj, respectively, and
are computed in the same manner (see the pre-
ceding section on estimation), but with case
weights adjusted to compensate for the half-
sample procedure.® The variance of Y’ was then
estimated by

Types of aggregate statistics produced from
the National Survey of Family Growth include
the number of currently married women, num-
ber of ever-married women, number of ever-
married women and never-married women with
offspring, and number of children ever born to

€For replicate strata where one of the pseudo-PSU’s
was composed of more than one sample PSU, case
weights were adjusted so that each pseudo-PSU repre-
sented half of the stratum population. In addition, case
weights were inflated approximately double to com-
pensate for the fact that only one-half of the sample was
used to obtain the estimate.
ever-married women. Half-sample variances were
not computed for all aggregate statistics, because
to do so would have required a prohibitive
amount of time and money. In addition, data
reports would be cumbersome if a variance esti-
mate was published for each statistic. Thus vari-
ances for each type of statistic were computed
only for selected population subgroups, which
were chosen to represent a wide variety of de-
mographic characteristics and a wide variation in
the size of the estimates. Curves were then fitted
to the relative standard error (RSE) estimates
for each type of statistic according to the model

2, B

(vy? F

Il

oa
+
|

 

A and B are parameters whose estimates deter-
mine the shape of the curve. The rationale for
the model and the iterative method that was
used to estimate 4 and B are explained else-
where. 12

Table M shows estimates of 4 and B for rela-
tive standard error curves by type of statistic
and race. For each type of statistic, separate esti-
mates were produced for black women, for
women of other races, and for women of all
races combined, because black women were sam-
pled at a higher rate than other women. Thus an
estimate of a given number of black women has
a smaller relative standard error than an estimate
of the same number of women of other races.
For example, an estimate of 500,000 currently
married black women has a relative standard
error of 7 percent while an estimate of 500,000
women of other races has a relative standard
error of 12 percent. This relationship is also re-
flected in figure 3, which shows the relative
standard error curves, by race, for currently mar-
ried women.

The relative variances of the aggregate statis-
tics are used to derive the relative variances of
percents, which are ratios of two aggregates with
the numerator being a subclass of the denomina-
tor. The relative standard error (RSE) of a per-
cent estimate

Table M. Estimates of parameters A and B for relative standard error curves, by type of statistic and race

 

 

Type of statistic and race

Parameter Parameter
A B

 

Number of currently married women

 

 

Number of ever-married women and never-married women with offspring

 

-.0001858989 6751.0619

 

-.0006310400 2798.6440
-.0002056235 7021.1665

.0001700390 6486.5185

 

 

-.0004520643 2848.2362
.0000422037 7111.5185

 

ANE FBOBE esi criusiienminamsmmm sims sat AtEm ARTE TART RT RR ean

Fr Re Wi FRE, mf LL AL el -.0001926913 6494.6569

 

 

crimes TSR Avs eR sR RE TTS -.0004813358 2698.6043
EE WL A -.0002362857 6892.2852

 

ES RR RT RRR SARE Ea 0 -.0001015087 | 18450.3253

 

 

rere er AE An Ak dn gies 2 -.0003690139 9111.1866
TEA STEER ERR -.0001792083 | 19967.9466

 

 

 

21
ze

RELATIVE STANDARD ERROR (PERCENT)

Figure 3. Relative standard errors for aggregates of currently married women, by race

100
9

70
60

50
40

30

20

A UO NOOO

 

A 2 3 4 se789) 2 3 4 567894 2 3 4 56789} 2 3 4 567894 2 3 4 56789
1 10 100 1,000 10,000

ESTIMATED NUMBER OF CURRENTLY MARRIED WOMEN IN THOUSANDS

Example of use of chart: An aggregate of 1 million black women (on the scale at the bottom of the chart) has a relative stand-
ard error of 4.7 percent, or a standard error of 47,000 (4.7 percent of 1 million).
is given by the expression!2

RSE (P') = V/[RSE? (Y')- RSE? (Z')]
Jas3- (13)
ditt— — (A+ —
y Zz
. BZ’ BY’
“NTr7
- [BZ - BY (PY)
% Y'z' (PY)
i [B(100 - P')
x PZ

where B is the least squares estimate from the
relative error curve for Y' and Z' (table M).

Notice that the relative standard error of P'
is a function of the values of both P' and Z'.
This relationship is demonstrated in figure 4,
which shows separate relative standard error
curves for percents based on different numbers
of currently married women of all races com-
bined. Each curve satisfies the equation

6751.0619 (100 - P')

where P' is the estimated percent and Z’ is the
denominator of P'.

An estimate of the standard error of the dif-

ference between any two aggregates or percents
is given by

: sri? 2
S(ri-v3) Nh toy,

=V(Y})? RSE2 (¥}) + (Y})? RSE? (Y})

 

This expression provides a good estimate of
the standard error for uncorrelated statistics, but
it can only be considered a rough approximation
otherwise. Because estimates from Cycle II of
the National Survey of Family Growth are based
on a large sample of women, the distributions of
Y| and Yj, (and, therefore, Y] - Y3) are approx-
imately normal. Frankel!® shows empirically
that, using balanced half-sample replication esti-
mates of variance, the test statistic

is, - Yh
(ry - 13)
approximates the student’s ¢ distribution under
the null hypothesis of no difference between the
parameters estimated by Hg and 4 against a
two-sided alternative. The number of replicates
in the replication design (40 for Cycle II) can
reasonably be used as the number of degrees of
freedom for the ¢ statistic, although the exact
value for the degrees of freedom remains
unknown. Therefore, individual two-tailed sig-
nificance tests of differences between statistics
from Cycle II data can be performed with an
approximate significance level of a by com-
puting ¢ and comparing it to the two-tailed 1-«
critical value for the t distribution with 40 de-
grees of freedom.

Example: In 1976, 29.0 percent of
24,795,000 currently married white women
had been surgically sterilized, compared to
21.6 percent of 2,169,000 currently married
black women. To test this racial difference
at the a = .05 level of significance, compute

% 29.0 - 21.6
. 2 2 Z
V(29.0)% RSEZ2(29.0) + (21.6)2 + RSEZ(21.6)

from table M

 

 

7021.1665 (100 - 29.0)

 

 

RSE(29.0) =
(299) (29.0)(24,795,000)
= .026
and
2798.6440 (100 - 21.6)
RSE(16)= [Zo A 20d
(21.6)(2,169,000)
= .068
thus
4 29.0 - 21.6
V(29.0)2(.026)% + (21.6)%(.068)?
=448

The two-tailed .95 critical value (1 - «) for a
t statistic with 40 degrees of freedom is
2.02. Therefore, the difference is significant
at the .05 level.

23
Figure 4. Relative standard errors for percent of currently married women of all races

100
90

80 Base of percent in thousands
70

60
50

100

40

30

20

Ad 0 OO NowWO

RELATIVE STANDARD ERROR (PERCENT)

vo No

>

 

A 2 3 4 56789) 2 3 4 56789) 2 3 4 56789)
1 10 100 1,000
ESTIMATED PERCENT

Example of use of chart: An estimate of 10 percent (from the scale at the bottom of the chart) of
a population of 1 million women (fourth curve from top) has a relative standard error of 24.6 percent,
or a standard error of 2.5 percent (24.6 percent of 10 percent).

000

24
REFERENCES

National Center for Health Statistics: National Sur-
vey of Family Growth, Cycle I: Sample design, estima-
tion procedures, and variance estimation, by D. K.
French. Vital and Health Statistics. Series 2-No. 76.
DHEW Pub. No. (PHS) 78-1350. Public Health Service.
Washington. U.S. Government Printing Office, Jan.
1978.

2U.S. Bureau of the Census: Census of Population
and Housing, 1970, Census Tracts, Final Report PHC(1).
Washington. U.S. Government Printing Office, 1972.
pp. 1-238. N

3U.S. Bureau of the Census: Housing Authorized by
Building Permits and Public Contracts, Monthly Sum-
mary Report C40, Sept. 1969-July 1975. Washington.
U.S. Government Printing Office.

4National Center for Health Statistics: Replication:
An approach to the analysis of data from complex sur-
veys, by P. J. McCarthy. Vital and Health Statistics. PHS
Pub. No. 1000-Series 2-No. 14. Public Health Service.
Washington. U.S. Government Printing Office, Apr.
1966. p. 12.

5National Center for Health Statistics: Pseudoreplica-
tion: Further evaluation and application of the balanced
half-sample technique, by P. J. McCarthy. Vital and
Health Statistics. PHS Pub. No. 1000-Series 2-No. 31.
Public Health Service. Washington. U.S. Government
Printing Office, Jan. 1969.

6National Center for Health Statistics: Distribution
and properties of variance estimators for complex multi-

stage probability samples: An empirical distribution, by
J. A. Bean. Vital and Health Statistics. Series 2-No. 65.
DHEW Pub. No. (HRA) 75-1339. Health Resources Ad-
ministration. Washington. U.S. Government Printing
Office, Mar. 1975.

"Kish, L.: Survey Sampling. New York, John Wiley &
Sons, 1965.

8Simmons, W. R., and Baird, J.: Use of pseudorepli-
cation in the NCHS Health Examination Survey, in Pro-
ceedings of the Social Statistics Section of the ASA.
American Statistical Association, 1968.

9Kish, L., and Frankel, M. R.: Balanced repeated
replications. J. Am. Stat. Assoc. 65(331): 1071-1094,
Sept. 1970.

10Plackett, R. L., and Burman, J. P.: The design of
optimum multifactorial experiments. Biometrika. 33(2):
305-325, Aug. 1946.

11 Gurney, M.: The variance of the replication method
for estimating variances for the CPS sample design. U.S.
Bureau of the Census, 1962. Unpublished document.

12National Center for Health Statistics: Estimation
and sampling variance in the Health Interview Survey, by
J. A. Bean. Vital and Health Statistics. PHS Pub. No.
1000-Series 2-No. 38. Public Health Service. Washington.
U.S. Government Printing Office, June 1970.

13 Frankel, M. R.: Inference from Survey Samples: An
Empirical Investigation. Ann Arbor. Institute for Social
Research, University of Michigan, 1971.

25
26

1.

IL.

Glossary of Terms

APPENDIXES

CONTENTS

 

Household Screener

 

27
29
APPENDIX |

GLOSSARY OF TERMS

Conterminous United States.—The land area
consisting of the District of Columbia and all
States except Alaska and Hawaii.

Dwelling unit (DU).—A single room, or
group of rooms, that is intended for separate liv-
ing quarters. The people who live there must
live and eat separately from everyone else in the
building (or apartment) and the room or group
of rooms must have either

1. A separate entrance directly from the
outside of the building or through a
common hall, or

2. Complete kitchen facilities for the use
of this household only. Complete Kkit-
chen facilities include all of the follow-
ing:

a. a range or cooking stove, and
b. a sink with piped water, and
c. a mechanical refrigerator.

Education.—The highest grade of school
completed.

Geographic region.—For the purpose of clas-
sifying the population by geographic area, the
U.S. Bureau of the Census has grouped the 50
States and the District of Columbia into four
regions, as follows:

Region States included

Northeast......... Maine, New Hampshire, Ver-
mont, Massachusetts, Rhode
Island, Connecticut, New
York, New Jersey, Pennsyl-
vania

North Central... Michigan, Ohio, Indiana, Illi-
nois, Wisconsin, Minnesota,
Iowa, Missouri, North Da-
kota, South Dakota, Kansas,
Nebraska

Delaware, Maryland, District
of Columbia, Virginia, West
Virginia, North Carolina,
South Carolina, Georgia,
Florida, Kentucky, Texas,
Tennessee, Alabama, Missis-
sippi, Arkansas, Louisiana,
Oklahoma

Montana, Idaho, Wyoming,
Colorado, New Mexico, Ari-
zona, Utah, Nevada, Washing-
ton, Alaska, Oregon, Cali-
fornia, Hawaii

Alaska and Hawaii are not included in the NSFG
sample design.

Household.—A family living together, or five
or fewer unrelated individuals living together in
a DU.

Parity.—The number of live births a woman
has had.

Screener interview.—A preliminary interview
at the household to collect information about
the DU and to determine whether or not the
household includes one or more women who are
eligible for the detailed interview.

Standard metropolitan statistical area
(SMSA).—A county or group of contiguous
counties (except in New England) which con-
tains at least one central city of 50,000 people
or more, or “twin cities’”’ with a combined popu-
lation of at least 50,000. In addition, other con-

27
tiguous counties are included in an SMSA if,
according to certain criteria, they are socially
and economically integrated with the central
city.

Urban area.—As defined by the U.S. Bureau
of the Census, the urban areas of the United
States include all cities or “twin cities’ with at
least 50,000 population in 1970 together with
the surrounding closely settled area and all
other incorporated or unincorporated popula-
tion centers with 2,500 inhabitants or more.

Metropolitan state economic area (MSEA).—
In New England, a county with more than half
its population in one or more standard metro-
politan statistical areas was classified as a metro-
politan state economic area if the county or a
combination of counties containing the standard
metropolitan statistical area or areas had
100,000 inhabitants or more. For further discus-
sion see: U.S. Bureau of the Census: State Eco-
nomic Areas, Washington, D.C., U.S. Govern-
ment Printing Office, 1951.

000

28
APPENDIX 11

HOUSEHOLD SCREENER

 

OMB No.
Expires:

OFFICE USE ONLY 68-574071

Dec. 31, 1976

 

Date Received A
Recheck:

 

YeS..v.uaaal NOutasaessa2

BEGIN DECK 25 Collected for the National
Center for Health Statistics
by
Westat, Inc.

Preliminary Disposition:

Final Disposition:

= [TTT] [7
1 2 3 4 5 10

HOUSEHOLD SCREENER

 

 

 

 

ASSURANCE OF

In accordance with Section 308 (d) of the Public Health Service Act (42 USC
CONFIDENTIAL:

242m) and the Privacy Act of 1974 (5 USC 552a), the National Center for
Health Statistics assures each respondent that all information which would
permit identification of any individual or family will be held in strict
confidence, will be used only by persons engaged in, and for purposes of,
this study and will not be disclosed to others for any purposes.

ITY:

ASSIGNMENT BOX:

SAMPLING TABLE
15.1 15 1637 396 Jo. 20 2}

 

 

 

 

 

IF NUMBER OF THEN INTERVIEW
ELIGIBLE FEMALES PERSON LISTED
LISTED IN SUMMARY ON SUMMARY BOX
BOX 181 | LINE J

Two 1
Three 3
Four x
Five 4

Six or 3
More

 

 

 

 

 

 

 

INTRODUCTION

Hello, I'm from Westat Research, Inc. (SHOW ID BADGE) A letter was sent to you
recently explaining the study we are conducting for the U.S. Public Health Service. As you ma

y
recall, from the letter, the study is being conducted all over the countdy and is about family size.

V

CHECK IF R VOLUNTEERS: Did not receive letter or does not remember letter. . . O 22

Received but did not read . + + « + + + 0 + wis x 0 +» O

 

ASK AT SAMPLED HOUSEHOLD AND RECORD VERBATIM. IF A SINGLE YEAR OR RANGE IS GIVEN THAT DEFINITELY

FITS ONE OF THE STATED CATEGORIES, CHECK AND FOLLOW INSTRUCTIONS; OTHERWISE CHECK D.K. AND USE
PROBE.

The sample of households we visit is scientifically selected to represent all households
in our country. In order to be certain our sample is correct, I need to ask

When was this structure originally built? 2

{Year/Range)

1970 or later. Before 1970. D.K., No idea, etc.
TERMINATE: Thank you very CONTINUE

much for your help. My INTERVIEW. Was it built before 1970?
instructions are to inter-
view at homes (apartment
houses) built before 1970.
Those built in 1970 or

later are sampled separately.
(COMPLETE NIR)

 

 

ASK:

(Probably) Yes OO CONTINUE.

(Probably) No QO TERMINATE.

D.K., No idea, O CONTINUE.

etc.
24 25 26 27 20

{Interviewer No.

{Interviewer No.)2s

 

ASSIGNED TO:

 

Interviewer Signature

REASSIGNED TO: (1)

{Interviewer Signature) 30 31 32 9

(2)

{Interviewer Signature) {Interviewer No.)

29
30

CONDUCT ONLY WITH A HOUSEHOLD MEMBER

AGE 15 OR OLDER

HOUSEHOLD ENUMERATION

CONTINUE DECK 25

S-1. To start, how many people live in this household?

 

 

 

 

TIME AM
BEGAN PM
34 35

NUMBER

 

S-2. What is the name of the head of this household?

(ENTER NAME ON LINE 01 BELOW.)

 

S-3. And the other members of this household -- what are their names?
everyone related to (HEAD).

NAMES IN TABLE BELOW.)

Let's begin with
(BE SURE PERSON INCLUDES [HIMSELF/HERSELF])

(ENTER

 

S-4, Are there other people living here who are not related
to (HEAD)? (IF YES, ENTER NAMES IN TABLE BELOW)

S-5. I have listed (READ NAMES IN ORDER).

living here now, such as friends, relatives or roomers?
(IF YES, ENTER NAME BELOW)

Is there anyone else

Ores
Ores

Owe
Ore

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

$-6. 5-7. S-8. | S-9. IF 15 YEARS OR OLDER ASK:
AFTER LISTING HOUSEHOLD, What is CODE SEX| How old Is (PERSON) now married, widowed,
ASK S-6 THROUGH S-9 FOR (PERSON's) |(ASK IF was divorced, separated, or has (he/she)
EACH PERSON AS relation- NOT (HEAD/ never been married? (NEVER MARRIED
APPROPRIATE ship to OBVIOUS)| PERSON) BUT REPORTED LIVING TOGETHER, CODE
(HEAD OF on (his/ IF NEVER MARRIED AND NOT LIVING
HOUSEHOLD)? her) last| TOGETHER AND HAS QWN CHILDREN IN
birthday? | HOUSEHOLD, CODE ”
Sing.
with
Infor own
person First Last mally Div. |Separ |chil |Never
NUMBER Name Name ur Mar. |[Mar. |Wid.|Ann.| ated|dren | Mar.
01 HEAD 1 2 1 2 3 4 5 6 7
02 ¥.2 1 2 3 4 5 6 7
03 2 1 2 3 4 5 6 7
04 1:2 1 2 3 4 5 6 7
05 1: 2 1 2 3 4 5 6 7
06 2 1 2 3 4 5 6 7
07 1-2 1 2 3 4 5 6 7
08 1:2 1 2 3 4 5 6 1
09 1.2 1 2 3 4 5 6 7
10 1:2 1 2 3 4 5 6 7
11 32 1 2 3 4 5 6 7
12 2 1 2 3 4 5 6 7

 

 

 

 

 

 

 

(IF MORE THAN 12 HOUSEHOLD MEMBERS, GO TO CONTINUATION BOOKLET, PAGE 2)

 

(IN S-10, CONTINUE QUESTION WITH

IDENTIFIED.)

S-10, Is there anyone now away from home who usually lives here (PROBE: such
as the mother of [CHILD])?

(IF HOUSEHOLD MEMBER, ENTER NAME ABOVE.)

oh

"PROBE" IF CHILD IS LISTED IN HOUSEHOLD BUT PARENT ([s] ARE NOT

 

S-11, Do any of the people in this household have a home anywhere else?

(IF YES, PROBE FOR USUAL RESIDENCE.

IF NOT

HOUSEHOLD MEMBER, DRAW LINE THROUGH NAME ABOVE.)

O Yes

 

S-12. Are any of the persons in this household now on full-time active
duty with the Armed Forces of the United States?

HOUSEHOLD MEMBER, DRAW LINE THROUGH NAME ABOVE)

(IF NOT

C) Yes

 

36 37 38

40 1 42 43

bu 4S 46

47 48 49 50 51

5

 

el lr TITEL

7
SELECTION OF RESPONDENT

 

CONTINUE DEGK 23

 

S-13, NO ELIGIBLE RESPONDENT by reason of: (CHECK CIRCLE AND SKIP TO S-15)
1. Group Quarters 0 2. Race 0 3. Sex O 4. Age/Marital
(REQUIRES NIR FORM) Status
S-14, ELIGIBLE RESPONDENT between 15 and 44 years old. (CHECK ONLY ONE CIRCLE)
ONE ELIGIBLE WOMAN:

 

1. Currently or informally married;
- Circle R's Person No. on page 2;
- Skip to S-15; USE

CURRENTLY MARRIED QUEX

Oo

2. Post married or SWOC;
- Circle R's Person No. on page 2;
- Skip to S-15; USE

POST MARRIED QUEX

O

 

MORE THAN ONE ELIGIBLE WOMAN (FOLLOW STEPS 1-4):

 

Step 1: List names of eligible women
in Summary Box, in order of
age, beginning with the oldest
on Line #1.

Step 2: Use Sampling Table on Page 1
to.determine which eligible
woman to interview.

Step 3: Circle R's line number in
Summary Box and write selected
R's name here:

Respondent's Name

Step 4: Selected Respondent is:

3. Currently or informally married
- Circle R's Person No. on page 2;
- Skip to S-15; USE

O

 

 

4.

Post married or SWOC;
- Circle R's Person No. on page 2;
- Skip to S~15; USE

POST MARRIED QUEX

Oo

 

 

 

ASK EVERYONE

 

 

 

 

  
  

 

 

 

 

 

 

 

 

 

S-15, Is there a phone. number where you can be reached (in case my office wants to verify this
interview)?
Telephone No. (Area Code) (5-16) (28
HO PHONG. ov a nis a's a #1» #.% » a & ¢ = * 200 ai on os =» + of {9=l7]
Befused. + + os se 4 isl vw oa . i. vw Wee . 7 (5-17)
S-16, Is this phone in your home In household 'v « « + wij 2 © nw + Nl
or elsewhere? In home of neighbor. + ov + +. + s +2
Other (SPECIFY) 3
AM 56 57
S-17, TIME SCREENER ENDED
PM
5-18 Does Assignment Box require: - AND ELIGIBLE "R" AVAILABLE NOW,
: COMPLETE REQUIRED PROCEDURES AFTER
the Missed DU Procedure? (ONo (0) ves EXTENDED INTERVIEW.
the Missed Structure - AND NO ELIGIBLE "R" OR ELIGIBLE "R"
Procedure? Ove (ves NOT AVAILABLE, COMPLETE MISSED DU
PROCEDURE AND FORM - PAGE 4, AND/OR
IF NO TO BOTH OF THE ABOVE, CONTINUE. MISSED STRUCTURE PROCEDURE AS
IF YES TO EITHER OF THE ABOVE QUTLITD ON MISSED STRUCTURE FORM,
INTERVIEWER: FILL OUT S-19 THROUGH $-23 BELOW IMMEDIATELY AFTER YOU LEAVE THE HOUSEHOLD.
S-19, Code Type of Structure: S-21, Date of Screener Interview
Detached single family house. . . . . 1
Trailer”. «'. oe aw dl winte wie ad
2-4 Family house/apartment building . 3 MONTH DAY YEAR
Row house (3 or more attached units). 4 60
Apartment house (5 or more units;
free access to housing units). . . . 5 ss
Apartment house (5 or more units;
locked entry, or guarded by
doornan, or BOth). « + s o.+ sins = 6 S-22, With whom did you conduct the
Other (SPECIFY) screener?
Person No.
7 from pg. 2
66 67 68 69
S-20, Race of Household (BY OBSERVATION): [1]
BLABR/NOGLD s Tan iv te iw, 4 ow iw ww A =
White, «ify ly « vow vie ew bl wd S-23, This screener interview was
Other (SPECIFY) 59 conducted in:
3 English. + ileus se ini io wtid
SPANIBh ov w ss wie ies se 2 70
Not able to observe .. . +. . . . + v . 4 Other (SPECIFY) 3

 

31
|
CONTINUE DECK 25

“ET

Before leaving household, say: "We want to be sure that every household in this area has
been given a chance to participate in this important survey.
At this address we listed households (in your struc-
ture/on this floor). Are there any other living quarters
here that we may have missed?"

MISSED D.U. PROCEDURE

Also, check in the lobby and around the outside of this (house/building) for additional
units or entrances in this structure.

Record discovered D.U.'s on form below. If no additional D.U.'s, check circle in the box in
the upper left-hand corner of the form.

If 1 to 4 Missed D.U.'s are discovered, fill out an assignment box on a blank Screener for
each (instructions for how to do this are in the Interviewer's Manual on page 10-54) and
conduct screener interview. Add the discovered D.U.'s to the Listing Sheet, Selected DU
List, and all copies of the Interviewer Weekly Status Report.

If 5 or more Missed D.U.'s are discovered, call supervisor for instructions before you do
any additional screener interviews. Add all of the discovered D.U.'s to the Listing Sheet
and the selected sample D.U.'s to the Selected DU List and all copies of the Interviewer
Weekly Status Report. Then fill out an assignment box on a blank Screener for each selected
sample D.U. and conduct screener interview.

MISSED D.U. FORM

 

CHECK (v) IF NO MISSED D.U. PSU # SEG #
AT SAMPLED*STRUCTURE O

 

 

 

D.U.

# ASSIGNED ADDRESS OF DISCOVERED D.U.

 

 

 

 

 

 

 

 

 

 

 

D.U.

ber

 

Number dis-
covered D.U.'s TOTAL ADDITIONAL D.U.'s
sequentially

within segments
beginning with

as instructed
on Selected DU
List. Each num-

assigned only
once within a NOTE: Be sure to thank respondent and complete page 3,
segment. S-19 through S$-23.

 

 

 

 

 

800 or 801

must be

 

 

 

 

 

 
Non-INTERVIEW REPORT FORM BEGIN DECK 26

 

N-1. Why were you unable to complete (screener/extended) interview?

Vacant or Not a Dwelling Unit . 1 (N-2)
Household Screener NIR. . . . . 2 (N-4)
Extended Interview NIR. . . . . 3 (WN-5)

 

N-2. Why is the listed address not an occupied dwelling unit

 

 

 

 

for our sample? Condemned + 5x ots + 5 sv ie 401
Demolished. :. +:+ + » +» » » 2 02
Place of Business . . . . . . .03
No such address/No such DU. . .04
Group quarters. . . . . . . . .05
Vacation Cabin. . . . . . . . .06
Not usable as
permanent residence. . . . . .07 (N-3)
Transient use (less
than 1 month). cece. + «+ + 908
Not a DU for other reason . . .09
Still under construction. . . .10
Post 1970 structure . . . . . 11
Improperly listed,
out of segment . . . . . . . .12
Vasant, wh. Juin Wine ow ee eo dX3 (8426)
N-3, Is there any additional information regarding this unit?
i
=
N-4, Whom did you contact in the household?
No one. . . . ow wend
Head or spouse of head. oa te ead
Relative of HH. . . . o & wie 3
Non-related adult in HH . . . . 4 (N-6) 13
Child under 15 years of age . . 5
Other (SPECIFY)
6
N-5. Unable to complete extended interview with respondent who is:
Currently married . . «... wv « 1
Post married or SWOC. . . . . . 2

 

N-6., What was the problem in obtaining information?

Unable to enter structure . . . 22 (¥-8)
No respondent after 4 calls . . 23 (N-8)
Language Problem (SPECIFY)

 

 

 

24 (N-10)

Unavailable during
£ield period . .» +. +s +» vs 4.2.25 (N-8)

POO FIL vv + ¢ = m3 3 + a. = 2608-8)
Breakoff. . . . «is + +» ow 270-7)
Refusal . . wei Cw ww wn B8 REE)
Other (SPECIFY)

29 (N-7)

N-7. What was the reason you could not complete this (screener/extended)

interview? (RECORD ANY EXPLANATIONS "R" GAVE AND YOUR OWN IMPRESSIONS.
THEN CODE THE REASON YOU BELIEVE IS THE MOST IMPORTANT.)

Did not want to answer questions, did
not believe in surveys. . . ids 1
Did not have time, didn't want $0
be bothered . . ode ein edlieinel abd
Afraid to let interviewer in, afraid to
answer, told not to answer questions. . 3
4
5

 

 

 

Objected to this particular survey . .
Claimed this survey did not apply to HH.
Other (SPECIFY)

 

 

 

 

~o

Could not determine any reason . . . . .

 
CONTINUE DECK 26

Name and phone number of sample household, if available.

NAME Phone ( ) =

 

N-9.

Race of household?

Black uo we we Whe
White + « oo vv 8 wmv * % »
Other (SPECIFY)

Could not find-ofE, ~ + ~» = +» =

WN
=

 

N-10.

What information could you find out as to the best time and/or
circumstances at which the (screener/extended) interview could
be obtained?

 

[x EXTENDED INTERVIEW NIR, SKIP TO N-13.

 

 

-11,

Have you learned anything about the age and marital status of the
women in the assigned household? (EXPLAIN)

 

N-12.

Based on the information you have obtained,
do you think the assigned household is:

Definitely eligible . .
Probably eligible . . .
Probably not eligible .
Definitely not eligible
Don't know. . . + + 4

Bs WN

 

N-13.

Code the type of structure?

Detached single family house. .
Trailer or trailer space. . . .
2-4 family house/apartment
Puilding™, . oi alie wee cE
Row house (3 or more

attached units). . . . . . . .
Apartment house (5 or more
units; free access to

housing unit). . «sv + «» 506 + 5
Apartment house (5 or more

units; locked entry or

guarded by doorman or both). . 6
Other (SPECIFY)

No

Sow

 

Date of NIR

17 18 19 20 21 22

 

NAME OF SUPERVISOR WHO APPROVED NIR:

DATE:

 

34
N. 1. R.

 

 

 

SUPERVISOR ONLY: SUMMARIZE PROBLEMS ENCOUNTERED, ACTIONS TAKEN AND YOUR
RECOMMENDATIONS FOR MAIN OFFICE FOLLOWUP:

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

NAME :

 

I.D. Number L L L

MONTH, DAY YEAR

 

 

 

MAIN OFFICE USE ONLY:

ACTION (S) : Telephone Call . . . +. . .
RECORD OF

ENTER ON
HH ACTION

Letter .« « & wriefe-3 wv + i»

Returned to field for
additional followup . . . .

Determined to be Final NIR .

OO OO

NAME :

 

DATE:

 

 
9€

 

00

 

RECORD OF HouseHOLD ACTION

 

Date

Day

Week

NON-ELIGIBLE
TYPE RESULT INPORMANT

CONTACT

 

Extended | g
Screener | Interview || §

(USE CODES) (CHECK BOX

Personal
Telephone

Mail

 

REMARKS

FIELD

Intv
No.
ain
Office
NAME

Is this con-
tact a call?

Yes No

(CHECK BOX)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

ResuLt Copes

 

 

 

 

 

 

 

 

 

IF VACANT OR NOT A D.U.. COMPLETE NIR

 

 

 

 

SCREENER

Vacant/Not DU . . .
Screener completed.
Unable to cnter structure
Eligible screener R not home. . . . .
Language problems . . . . . . . . . 4
Unavailable during field period . .
BB HL cnet ns Be eee i
BrRakole. ois luis. sin ws sow sw kin
ReEOaaY 5 ii dd. de oe
Other (Specify under Remarks) . . .

   

belabhawnro

EXTENDED INTERVIEW (IF SCREENER COMPLETE
USE CODES, OTHERWISE LEAVE BLANK)

  

No eligible R. . . . "8
Interview completed. ie
Unable to enter structure . 2
Eligible R not hore. . . + » +» + + + + 3
Language problems. . . . . . . . . . . 4
Unavailable during field period. . . . 5
Tow THE, lulatty vu 5 0% 0 Shiney 6
Broke ate 0 Sil ie erie a aw ele 7
Retesare ©. LoL Ll Lil anil, 8
Other (Specify under Remarks). . . . . 9

 

 

 

 
pia

Ea
pg #
fk

Bi

 
 

 
 

Series 1.

Series 2.

Series 3.

Series 4.

Series 10.

Series 11.

Series 12.

Series 13.

Series 14,

Series 20.

Series 21.

Series 22.

Series 23.

VITAL AND HEALTH STATISTICS Series

Programs and Collection Procedures.—Reports which describe the general programs of the National
Center for Health Statistics and its offices and divisions and data collection methods used and include
definitions and other material necessary for understanding the data.

Data Evaluation and Methods Research.—Studies of new statistical methodology including experi-
mental tests of new survey methods, studies of vital statistics collection methods, new analytical
techniques, objective evaluations of reliability of collected data, and contributions to statistical theory.

Analytical Studies.—Reports presenting analytical or interpretive studies based on vital and health
statistics, carrying the analysis further than the expository types of reports in the other series.

Documents and Committee Reports.—Final reports of major committees concerned with vital and
health statistics and documents such as recommended model vital registration laws and revised birth
and death certificates.

Data From the Health Interview Survey.—Statistics on illness, accidental injuries, disability, use of
hospital, medical, dental, and other services, and other health-related topics, all based on data collected
in a continuing national household interview survey.

Data From the Health Examination Survey and the Health and Nutrition Examination Survey.—Data
from direct examination, testing, and measurement of national samples of the civilian noninstitu-
tionalized population provide the basis for two types of reports: (I) estimates of the medically defined
prevalence of specific diseases in the United States and the distributions of the population with respect
to physical, physiological, and psychological characteristics and (2) analysis of relationships among the
various measurements without reference to an explicit finite universe of persons.

Data From the Institutionalized Population Surveys.—Discontinued effective 1975. Future reports from
these surveys will be in Series 13.

Data on Health Resources Utilization. —Statistics on the utilization of health manpower and facilities
providing long-term care, ambulatory care, hospital care, and family planning services.

Data on Health Resources: Manpower and Facilities. —Statistics on the numbers, geographic distri-
bution, and characteristics of health resources including physicians, dentists, nurses, other health
occupations, hospitals, nursing homes, and outpatient facilities.

Data on Mortality. —Various statistics on mortality other than as included in regular annual or monthly
reports. Special analyses by cause of death, age, and other demographic variables; geographic and time
series analyses; and statistics on characteristics of deaths not available from the vital records based on
sample surveys of those records.

Data on Natality, Marriage, and Divorce.—Various statistics on natality, marriage, and divorce other
than as included in regular annual or monthly reports. Special analyses by demographic variables;
geographic and time series analyses; studies of fertility; and statistics on characteristics of births not
available from the vital records based on sample surveys of those records.

Data From the National Mortality and Natality Surveys.—Discontinued effective 1975. Future reports
from these sample surveys based on vital records will be included in Series 20 and 21, respectively.

Data From the National Survey of Family Growth.—Statistics on fertility, family formation and dis-
solution, family planning, and related maternal and infant health topics derived from a biennial survey
of a nationwide probability sample of ever-married women 15-44 years of age.

For a list of titles of reports published in these series, write to: Scientific and Technical Information Branch

National Center for Health Statistics
Public Health Service
Hyattsville, Md. 20782
 

DHHS Publication No. (PHS) 81-1361
Series 2-No. 87

 

NCHS

U.S. DEPARTMENT OF HEALTH AND HUMAN SERVICES POSTAGE AND FEES PAID
Public Health Service U.S. DEPARTMENT OF H.H.S.
Office of Health Research, Statistics, and Technology HHS 396

National Center for Health Statistics

3700 East-West Highway THIR D C LASS

Hyattsville, Maryland 20782

OFFICIAL BUSINESS
PENALTY FOR PRIVATE USE, $300

For publications in the Vital and 5 L
Health Statistics Series call 301 a 4 1 4 >

436-NCH

 

 
U.C. BERKELEY LIBRARIES

C02120613)
 
      

WL 7 EERE | EER
Fa le 3 &

nl
Wr dew
»

Lar