NATIONAL id CENTER Series 2 For HEALTH Number i] S37. vane] pilot study on | CREE Statistics U. S. DEPARTMENT OF HEALTH, EDUCATION, AND WELFARE Public Health Service Public Health Service Publication No. 1000-Series 2-No. 28 For sale by the Superintendent of Documents, U.S. Government Printing Office Washington, D.C., 20402 - Price 45 cents NATIONAL CENTER| Series 2 For HEALTH STATISTICS | Number 28 VITALand HEALTH STATISTICS DATA EVALUATION AND METHODS RESEARCH pilot study on Patient Charge Statistics A study to develop methodology for collecting information on hospital charges in the Hospital Discharge Survey. Washington, D.C. May 1968 U.S. DEPARTMENT OF HEALTH, EDUCATION, AND WELFARE Public Health Service Wilbur J. Cohen William H. Stewart Acting Secretary Surgeon General for NATIONAL CENTER FOR HEALTH STATISTICS THEODORE D. WOOLSEY, Director PHILIP S. LAWRENCE, Sc.D., Associate Director OSWALD K. SAGEN, PH.D., Assistant Director for Health Statistics Development WALT R. SIMMONS, M.A., Assistant Director for Research and Scientific Development ALICE M. WATERHOUSE, M.D., Medical Consultant JAMES E. KELLY, D.D.S., Dental Advisor LOUIS R. STOLCIS, M.A., Executive Officer OFFICE OF STATISTICAL METHODS MONROE G. SIRKEN, PH.D, Director Public Health Service Publication No. 1000-Series 2-No. 28 Library of Congress Catalog Card Number 67-62373 KAHIT 5/2 710. 2 8-22 PUBLIC HEALTH LIBRARY FOREWORD This report presents the results of a pilot study which was undertaken to investigate the feasibility of collecting information about hos- pital charges from the patient charge ledgers on file in the business offices of hospitals. The report also presents findings and recommenda- tions regarding the methodology for collecting patient charge statistics on a continuing basis for a national sample of patients in the Hospital Discharge Survey, The objective of the Hospital Discharge Sur- vey is to produce national statistics on hospital patients which are representative of the experi- ence of the civilian population in short-term hos- pitals. Since the latter part of 1964, selected items of social and medical information about discharged patients have been abstracted from the patient medical records, It is planned to expand the con- tent of the Hospital Discharge Survey by linking the information about the patient, which is cur- rently being collected from the medical records, to information about his hospital charges, that will be collected from the patient charge ledgers. In July of 1965, the National Center for Health Statistics convened an Ad Hoc Committee on Hos- pital Charges to advise on matters pertaining to the collection of statistics on hospital charges and sources of payment. In particular, the Center sought the advice of this Committee with regard to (1) the kinds of patient charge data that would be most useful, and (2) the problems that could be expected in abstracting these data from the patient charge ledgers, The members of this Com- mittee included: Mrs. Agnes W, Brewster, Divi- sion of Community Health Services, PHS, HEW; Dr. Donald C. Riedel, Blue Cross Association; Mr. Robert M, Sigmond, Hospital Planning Associ- ation of Allegheny County, Pa; Mr, William Spector, American Hospital Association; and Mr, Theodore D. Woolsey, Dr. Monroe G, Sirken, and Mr. Milton C, Rossoff from the Center, In 1966, the Center contracted with Booz, Allen, and Hamilton, Inc., to undertake a pilot study on hospital patient charges. The project was started in July 1966 and the work was com- pleted as planned about 8 months later. During this period, there was very close collaboration between the staffs of Booz, Allen, and Hamilton, Inc., and of the Center, Dr. Robert H, Hamlin of Booz, Allen, and Hamilton was project director and he and his staff, including Michael A, Green, Douglas W, Metz, and John L. McKee, were re- sponsible for conducting the pilot study and pre- paring this report, The Center staff included John Monroe and Milton C, Rossoff, and Monroe G. Sirken, who was project officer, The Demographic Survey Division of the Bureau of the Census par- ticipated in the field work of the study, From the viewpoint of the National Center for Health Statistics, this was an efficiently conducted pilot study that was exceptionally well planned and implemented on a timely basis. The results of the study presented in this report have demon- strated the feasibility of collecting national hos- pital charge statistics classified by types of serv- ices provided and by sources of payment based on information reported in the patient charge ledg- ers. The findings will be valuable in deciding on methods to be used and in preparing the proce- dures for collecting this information inthe Hospi- tal Discharge Survey. Monroe G, Sirken, Director Office of Statistical Methods 116 — CONTENTS Foreword =---m-- comme meee eee meme mm I. 1. III. IV. Summary of the Study--------=-cc-comm mmm Background---=-===c-m-mmmm momen emo memo momo ooo Objectives of the Study--------==-cccmmmommm mmc cme Methodology =-===========-m mmm mmm mmm em moomoo moomoo Major Findings and Conclusions--------=-=c-mommocmoomommmommomomo Study Methodology --=-==-=====-ccecmmmm mmm mmm momo m moomoo Phase I. Development of Interview Schedules, Abstract Forms, and Pro- cedures =------ mmm meme mmm mm mmmm mmm Phase II. Administration of the Field Test---=--c-ecccmemmcmmanoanm- Phase III. Administration of the Final Field Test----------cccemamnu-- Limitations in Pilot Study--==--m-mmmm mmc mmr Major Characteristics of Patient Charge Systems in Hospitals--------- Types of Hospital Patient Charge Systems==-====-co-=ccmccmmmmmanooo The Completed Patient Ledger---------comommmmmmm ccm cee memo Feasibility of Abstracting Patient Charges on a Continuing Basis------- Major Issues in Abstracting Patient Charge Ledgers--------=--==-=--- Interpretation of Findings and Conclusions------=---ccm-cmommomomaamno Appendix A. Glossary of Terms--------commmmommmmm mcm cm mmm mmo mme Appendix B. Contact Letters Used With Hospitals-----==--cc-cccomocomuom- Appendix C. Interview Schedule Used With Selected Hospital Personnel----- Appendix D. Patient Charge Abstract Form Used in Field Test Hospitals---- Appendix E. Hospital Instruction Manual Used in Final Test Hospitals------ Appendix F. Suggested Patient Charge Abstract Form for Future Use------- Appendix G. Alternative Methods for Selecting and Relating Patient Charge Ledgers and Medical Records for Abstracting in a Continuing Abstracting Program=m mmm momo momo oem ee mmm mmmmm mm oommomo- oe N= = = w 27 31 34 43 44 51 52 vi IN THIS REPORT, the results of a pilot study to develop the method- ology for instituting a continuous program for collecting information on patient charges and sources of payment as a part of the Hospital Discharge Survey ave presented. These results relate specifically to the development of methodology for utilizing hospital ledgers of patient charges as the source of information on charges for the national sam- ble of patients in the survey. Curvently, the Hospital Discharge Sur- vey is limited to the collection of specified items of information about patients available from the patient medical records. The study concludes that it is both feasible and practical to institute a continuous program for collecting patient charges and sources of pay- ment dala as part of the Hospital Discharge Survey. The principal find- ings weve that (1) the dollar amounts of charges by selected types of services and sources of payment are available from the ledgers of patient charges ov can be estimated in all hospitals, (2) the agvee- ment in the definition of types of charges and sources of payment is sufficient to permit gross comparisons among hospitals, and (3) hos- bilals ave cooperative and their employees can accurately abstract the information from the ledgers of patient charges. SYMBOLS Data not available----ccceom momo. css Category not applicable-----ceceemaooo___ Quantity Zero=--=c-coocmm eee - Quantity more than 0 but less than 0.05---- 0.0 Figure does not meet standards of reliability or precision-==-==v-veeeeceun.. * PILOT STUDY ON PATIENT CHARGE STATISTICS l. SUMMARY OF THE STUDY This chapter summarizes the background, objectives, methodology, and principal findings and conclusions of the pilot study on hospital patient charge statistics. BACKGROUND The subject of patient charges in hospitals is receiving increased attention by the medical and health care communities, government agen- cies with public health responsibilities, and the general public. The rapid rise inhospital charges and costs, the increasing demand for hospital services, and the rather substantial restructur- ing of mechanisms for financing hospital care (e.g., the growth of third parties as sources of payment, including the Medicare and Medicaid programs), have stimulated both the need and desire for improved data on patient charges in hospitals. Currently, little information on a national basis is available on patient charges which (1) reflects a cross section of U.S, hospitals, en- abling generalization to the total population of patients and hospitals; and (2) presents a trend analysis, on a continuing basis, of variations in charges over time, To meet the general need for additional and improved data on patient charges in hospitals, the National Center for Health Statistics (NCHS) of the U.S. Public Health Service sponsored this pilot study on patient charge statistics. OBJECTIVES OF THE STUDY The purpose of the pilot study was to deter- mine the feasibility and practicality of institut- ing a continuous program for collecting patient charge and source-of-payment information as a part of the Hospital Discharge Survey of NCHS, which currently is limited to the collection of patient medical record information. The major objectives of the study were to determine whether: Information in hospitals is available on the dollar amount of total charges for a single episode of hospitalization, broken down into room and board charges and other charges, and by source of payment (e.g., Blue Cross, Medicare, personal, etc.), as prescribed by the Ad Hoc Committee on Hospital Charges. Consistent definitions exist among hospitals for the requisite patient charge and source- of-payment information, Specified patient charge and source-of-pay- ment data are readily obtainable. Valid patient charge data can be abstracted from hospital records by hospital personnel. Abstracting can be done by hospital personnel at a reasonable cost, Patient charge and source-of-payment data can be related to medical record data ab- stracted for the same patient and admission. METHODOLOGY The methodology of the pilot study involved three major phases. Phase 1.--Development and pretesting of an interview schedule, abstract form, and study pro- cedures in five hospitals not a part of the Hospital Discharge Survey, Phase 2.--Administration of the interview schedule with selected hospital personnel and ab- straction of patient charge ledgers by survey staff in a national sample of 20 field test hospitals not a part of the Hospital Discharge Survey. Phase 3.--Administration of a revised inter- view schedule and abstraction of patient charge ledgers by hospital personnel in five hospitals already included in the Hospital Discharge Sur- vey, with verification of their abstraction ac- curacy by independent review of survey staff, Throughout the report the 20 hospitals in Phase 2 of the study are referred to as the field test hospitals and the 5 hospitals in Phase 3 as the final test hospitals, A glossary of terms used in the report is presented in Appendix A, at the end of this report, Direct interviews were conducted with more than 50 hospital administrators and accounting department personnel in the 25 hospitals sur- veyed. Of the more than 2,000 patient charge ledg- ers identified by random and purposive selec- tion, 1,064 patient ledgers, with complete data, were abstracted. The study team was composed of six indi- viduals, three of whom conducted the field work. Personnel from NCHS and the Bureau of the Census participated with the study team in ab- stracting patient charge records in several hos- pitals, Hospital personnel in the five final field test hospitals abstracted patient charge records in accordance with a training and an instruction manual developed by the survey team and ad- ministered in each hospital by a team member, In the 20-hospital field test, it was found that data from 2 hospitals with computerized patient charge systems were not obtainable be- cause special programming would have been re- quired to secure the information needed for patient charge abstracting. Therefore, such hos- pitals were excluded from the final field test phase of the study. MAJOR FINDINGS AND CONCLUSIONS On the basis of the principal study findings, it was concluded that it is both feasible and prac- tical for NCHS to institute a continuous program for collecting patient charge and source-of-pay- ment data as a part of the Hospital Discharge Survey. The principal findings were: Data on the dollar amounts of room and board charges, other charges, personal nonmedical charges, and sources of payment are avail- able or can be estimated in all hospitals. Sufficient consistency exists in the elements that hospitals include in room and board charges to permit gross comparison of such charges among hospitals. Hospitals vary in the items included in the category of other charges, but comparisons of other charges among hospitals can be made since significant variations are limited to only a few classes of charges, such as anesthesiology services. Hospital employees can accurately abstract valid patient charge and source-of-payment data if appropriate on-site training is pro- vided by the survey staff, Hospital personnel can abstract patient charge ledgers at a reasonable cost ranging, among hospitals, from $12.50 to $50.00 per 100 abstracts, or an average of $25.00 per 100 abstracts, A continuous program for abstracting patient charge and source-of-payment data can be established as a part of the Hospital Dis- charge Survey since it is possible for NCHS to: Use the same patient sample for patient charge and medical data abstracting, ex- cept for well-newborn infants when their charges are combined with the mother's on the same ledger. Match the abstracts of patient charge data with abstracts of medical data for the same patient, even though the two sets of data will not be available at the same time. Obtain the cooperation of the great majority of hospitals in a continuous abstracting pro- gram, provided the approach to the hos- pitals is carefully planned and personal in nature, and the time required of hospital personnel is held to a minimum by careful planning of hospital time requirements and by training of hospital personnel by survey staff, NCHS and U.S. Public Health Service available to NCHS are evaluated in subse- prestige is sufficient to stimulate hospital quent chapters of this report.) cooperation, although hospital association endorsement should be obtained if possible. Linking medical record abstracts and patient charge record abstracts, once completed, Prior to initiation of a program of abstract- for the same patient and same admission ing patient charge information on a continuing since the two abstracts will be prepared at basis, NCHS should develop detailed procedures different points in time, (Alternative methods for: available to NCHS are evaluated in subse- } quent chapters of the report.) Abstracting patient charges in hospitals with automated patient charge systems, Ensuing chapters of this report discuss in greater detail the study methodology, the major Assuring that the same sample of patients is characteristics of hospital patient charge sys- selected for patient medical and patient tems, and the feasibility of abstracting patient charge data abstracting. (Alternative methods charges on a continuing basis, Il. STUDY METHODOLOGY This chapter describes the methodology em- Sources of Payment ployed to conduct the pilot study on hospital patient charge statistics. Basic methodology con- Blue Cross sisted of the following steps: Commercial insurance Development of an interview schedule (Ap- pendix C). Administration of the interview schedule through direct interviews with hospital ad- ministrators and key accounting department personnel in 25 hospitals—20 hospitals not participating and 5 participating in the pres- ent NCHS Hospital Discharge Survey. The interview schedule was administered by three members of the study team, with one mem- ber responsible for conducting 15 of the 25 direct interviews, training the other team members, and closely supervising their work. Development of an abstract form (Appendix D) for eliciting patient charge and source-of- payment data according to the following prescribed categories. Patient Charges Room and board charges Other charges Total charges Personal payments Welfare (Title XIX— Medicaid) Hospital discounts Private charity Medicare (Title XVIII) Workmen's Compensation Federal public service programs Other Use of the abstract form to abstract patient charge and source-of-payment data from patient charge ledgers in the 25 hospitals. Abstracting was accomplished by members of the study team and personnel from NCHS and the Bureau of the Census in the 20-hos- pital field test. In the last phase of the study, the final field test, hospital employees in the five hospitals performed the abstracting in accordance with an instruction manual de- veloped by the study team. This abstracting was verified by independent abstracting of the same patient charge ledgers by the sur- vey team, Exhibit 1 is a chart depicting the work plan for the study. As outlined in the sections which follow, the study design comprised three major phases and a series of steps to accomplish the objectives of each phase, EXHIBIT |. WORK PLAN FOR THE STUDY ESTIMATED AND ACTUAL PILOT STUDY SCHEDULE | su | August [ serremser | ocroser | wovewsex | DECEMBER 7 vl2 3 TelsTe [0 To Is Tow la Talus Tie [ir J Be nn . Frm} TASK I T EE » [un [a]a 33 | 34 3s T ! In JANUARY T FEBRUARY MARCH | | 1 + T 1 | | | | I | | | | | I | | | - RESEARCH FILES AND STAFF = PRELIMINARY HOSPITAL INTERVIEWS Hh - PREPARE PRE-TEST MATERIAL deb elnterview schedules act f pre-test hospitals ct letters || | CONDUCT PRE-TEST (4 HOSPITA yey | | [ E Ls) bbb | | PREPARE TENTATIVE FORMS FOR FIELD TEST shbbhhbhbabbb | [I ln schedule vo | | [] eAb 6. SELECT HOSPITALS FOR FIELD TEST 11 bbb | | | | | | | | | | | | | | I | | | | 7. REVIEW FINDINGS, FORMS, AND SELECTION OF HOSPITALS WITH NCHS. bbl ibd 8. CONDUCT FIELD TEST MAMMAL | | | | | | | | Io | 9. REVISE AND PREPARE FORMS FOR FINAL TEST ddd, | BE | | | | | 10. SELECT FINAL TEST HOSPITALS ——— RRR 11 CONDUCT FINAL FIELD TEST | egg | | 12. DEVELOP FINDINGS AND CONCLUSIONS nt OF FINAL FIELD TEST 13. DEVELOP REPORT OUTLINE | | | A L . | | [ I 14 PREPARE REPORT LY Ly Lo] | ed on July 15, 1966, and was completed on March 1, 1967 1Field study wos co — PLANNED WORK SCHEDULE HII CTUAL WORK SCHEDULE PHASE |. DEVELOPMENT OF the feasibility and practicality of abstracting INTERVIEW SCHEDULES. ABSTRACT charge data: (1) dataavailability andaccess- ibility, (2) consistency and complexity of FORMS, AND PROCEDURES patient charge Systems, Including coding systems used, (3) time within which most The objectives of this phase were to: patient ledgers are completed and ready for abstracting, and (4) time required for col- Obtain a general understanding of patient . g & p lection of data. charge systems in hospitals, Phase I involved accomplishment of the following Determine attitudes of hospital personnel ps concerning the value of the survey and the key factors that would influence a hospital to participate in the pilot study and, later, a continuous program of collecting patient Personnel charge data. Preliminary Interviews With Knowledgeable In this initial step, interviews were con- Ascertain the opinion of the hospital staff on ducted with selected administrators and account- the following key items which would influence ing department personnel in four hospitals. In addition, actual primary patient charge data sources were examined for format, content, and variations and compared with study require- ments to determine the availability of data. Preparation and Pretesting of Survey Material In this step, the following survey materials were developed: Procedures for contacting hospitals, follow- ing up contacts, and expressing appreciation for participation, Appendix B includes mate- rials developed and used during the study by NCHS and the study team, An interview schedule for use in interview- ing hospital administrators and accounting department personnel, A data collection (abstract) form in a format adaptable to a broad range of hospital pri- mary data source (patient charge ledger) characteristics, A list of five hospitals for pretesting of the initial interview schedule and abstract form. The hospitals which were not a part of the present Hospital Discharge Survey were selected on the basis of diversity in type, size, and other pertinent characteristics. The resulting interview schedule and patient charge abstract form were submitted, as re- quired, to the Bureau of the Budget for approval and subsequently pretested in the five hospitals by a member of the survey team in accordance with the following procedures: Contact letters from NCHS were sent to the five selected hospitals to explain the pur- pose of the study and to request cooperaton, The letter was followed by a personal tele- phone call from a study team member to each hospital administrator to verify co- operation, arrange for a personal visit, and outline the procedures to be followed during the personal visit, The interview schedule was administered by direct interview with each hospital adminis- trator, and with accounting managers or des- ignated employees in the accounting depart- ment, if the administrators were not able to answer the detailed technical questions on accounting procedures, Approximately 100 patient charge records were selected in each hospital—75 randomly from the hospital's daily patient discharge listing for the most current 4-month period and 25 purposively for the types of patient charge records which were identified as potential problem areas (e.g., welfare, Work- men's Compensation, Medicare, and automo- bile accident cases) during the interviews with hospital administrators and accounting department personnel, Hospital personnel in the accounting depart- ment of each hospital searched their files to locate the 100 selected ledgers. A study team member completed abstracts for the selected ledgers, asked questions whenever necessary, and made notes regard- ing portions of the abstracting process which would necessitate changes in the study pro- cedures, the abstract form, or the interview schedule, Because some records could not be located, slightly fewer than 100 ledgers were abstracted in each hospital, Letters of appreciation were sent to the par- ticipating hospitals. Modification of Forms and Instructions for Use in the Field Test This step involved preparation and review of findings from the pretest phase of the study, revision of the interview schedule and abstract form based upon the findings of the pretest, and preparation of an interviewer's instruction manual for the additional study staff to be used in con- ducting the field test involving 20 hospitals. PHASE Il. ADMINISTRATION OF THE FIELD TEST The objectives of Phase II of the study were and delineate problems in linking patient charge ledger and medical record data in a continuous program to collect patient charge information. to: Ascertain the likelihood of hospital coopera- Determine the availability and accessibility ton a : a of collecting of hospital patient charge and source-of-pay- patie Charge statistics, ment data in a sample of 20 hospitals not pres- Test and refine the interview schedule, ab- ently participating in the Hospital Discharge stract form, and procedures developed during Survey, the pretest phase of the study, Determine hospital procedures for handling The following steps were accomplished in this patient charge and source-of-payment data phase, Exhibit II. Hospital characteristics prescribed for the field test Number of Type of hospital hospitals in field test Total hospitals==== mmm momo oe ee een 20 Nonfederal Government Large teaching l-- como ommm eee x Large nonteaching-=--== === moomoo mee ee eee eee 1 Not large Urbam=---- o-oo moe 1 Not large rural-=-=--== === meme 1 Nonprofit church Lar gem = =m mm mm mm me ee mmmceeeem 1 Not large urbam--=- === ome ee ee. 1 Not large rural---- oom ee ccimeee em 1 Other nonprofit Large teaching== === moo oo oo eee ee eee ee 1 Large nonteaching=--==== comme L_. 1 Not large. urban== === === eee ee ee cece 1 Not large rural--==== == eee ee eee eee eee eee 1 Proprietary Lar gem mm mmm mm mm me ee ee ee eee 2 Not large uUrbam===- == mm eee 2 Not large rural--==== === ee ee ees 2 Special types 0SteOopathic mmm mm mmm mmo ee ieee 1 Hospital participating in a prepaid group practice plan------=-meccmccccocococoonn 1 Specialty hospital, such as a hospital for crippled children-=---=cmmecmcmeccoooenn 1 ‘A large hospital is one with 500 beds or more. One of each of the following kinds of hospitals was to be included among the 20 listed above. (a) Hospital owned by physicians or by an industrial corporation (b) Hospital with a highly automated record system (c) Hospital with an all-inclusive charge rate (d) Hospital with long-term care, psychiatric, or other specialized units (e) Hospital in which radiologist, pathologist, etc., bill separately for their departments Selection of Hospitals In the first step 20 hospitals, not presently included in the Hospital Discharge Survey, were selected for the field test, The characteristics of the 20 hospitals, as prescribed by NCHS, are pre- sented in Exhibit II, As reflected in Exhibit II, the purpose was to select a broad, representative sample of hospitals in the United States. The hos- pitals therefore represented a range in size, location (regional distribution was as follows: East, 6; Midwest, 4; South, 5; and West, 5), owner- ship, accounting equipment, patient charge sys- tems, and special characteristics. In the 20-hospital field test, great effort was made to select hospitals with characteristics identical to those requested, This was not possible in all cases because most large hospitals (500 beds and over) were already in the Hospital Discharge Survey and, therefore, were excluded from the requirements of the field test. In addition, the number of 500-bed church-affiliated and 500-bed proprietary hospitals is quite limited. The following variations in characteristics of hospitals selected for the field test were approved by NCHS, since it was felt that the variations would not affect the outcome of the study. Nonprofit church—a 472-bed hospital substi- tuted for a 500-bed hospital. Other nonprofit—a large teaching hospital substituted for a large nonteaching hospital. Proprietary—A 174-bed hospital substituted for a 500-bed hospital, ‘Exhibit III, summarizes the specific char- acteristics of the 20 field test and 5 final test hospitals. Administration of Interview Schedule In the next step, the interview schedule used in the 20-hospital field test (see Appendix C) was administered by three members of the sur- vey team. The member who conducted the pre- test administered the direct interviews in 10 hos- pitals and trained and closely supervised the performance of the two other members in the re- maining hospitals. Administration of the interview schedule consisted of an introductory interview with the administrator, a followup interview with the accounting manager or designated person in the accounting department, and completion of the interview schedule by the interviewer, Information about the hospitals and their patient charge procedures was collected from in- terviews with 52 hospital employees, including 2 medical directors, 19 hospital administrators, 20 controllers, 6 business managers, and 5 account- ing supervisors. Abstraction of Data on Patient Charges This step involved the completion of approxi- mately 100 abstracts in each hospital by members of the study team accompanied in some cases by personnel from NCHS and the Bureau of Census. (See Appendix D for the abstract form used.) In the field test (20 hospitals) and final test (5 hospitals) more than 1,000 abstracts of patient charge ledgers were completed, Of 2,084 patient ledgers identified by random and purposive se- lection, 1,064 were abstracted because they were completed, Exhibit IV analyzes the ledgers re- viewed during the study, 23 of the 25 survey hospitals produced ledgers for review, One hospital had a computer- based patient charge system that would have required programming to collect patient charge data and the other hospital was in the process of converting from a manual to a bookkeeping machine system and ledgers were not available, 2,038 ledgers were requested from the hos- pitals for review. 2,084 ledgers were searched for, of which 1,904 were located. (One hospital searched for more ledgers than requested.) Of the 1,904 ledgers that were located, pay- ments had been completed for 1,064, which therefore could be abstracted. Exhibit III. Hospital characteristics observed in the 25 survey hospitals Number Type of | Type of Hospital of Type of posting | charge Geographic Special characteristic beds ownership equipment | system location Total- ces G=- 5 C - 3 A-I- 4 M - 4 Vv -11 T - 1 s&s-21 E - 10 VC - 3 B -19 ... S - 6 P- 6 M- 2... We- 5 Field test hospitals 704 G C A-1 M | Psychiatric unit, teaching 762 G T A-1 E 160 G B S&S 5 75 G B S&S Ww 472 vc C S&S S | Long-term care unit 249 Vv B S&S S 126 VC B S&S M 656 V C 15&S E | Rehabilitation service,teaching 460 Vv B A-1 E | Teaching 348 Vv B S&S S | Psychiatric unit, teaching 141 Vv B S&S M | Progressive patient care 174 P B S&S E | Doctor=-owned 140 P B S&S W | Doctor =-owned 87 P B S&S W | Doctor -owned 46 P M S&S W | Doctor=-owned 44 P B S&S W | Community=-owned 72 P B S&S E | Corporation 42 Vv B S&S E | Osteopathic hospital 180 V B S&S M | Prepaid group plan 112 V B S&S S | Crippled child care Final test hospitals A ae 289 vec B S&S E A 91 Vv B S&S E A EE 145 V M A-1 E 2pm mmm mmm 244 Vv B S&S E 25mm mmm mem 1,298 G B S&S S | Teaching "Welfare patients have an all-inclusive rate Legend: Type of ownership Type of posting equipment Type of charge system Geographic location G = Government C = Computer A-1 = All-inclusive M = Midwest V = Voluntary, = Bookkeeping machine daily rate E = East nonprofit T = Tabulator S&S = Charge according S = South VC = Voluntary, M = Manual to supply or W = West Coast church service received P = Proprietary Modification of Forms and Procedures for Use in Final Field Test This step involved the development of find- ings and conclusions from the 20-hospital field test, and the design of an instruction manual to be adapted to each individual hospital by a survey team member for use by hospital employees in the final field test (Phase III), In addition, this procedure included the revision of the interview schedule and abstract form, based upon the ex- perience of the 20-hospital field test, PHASE Ill. ADMINISTRATION OF THE FINAL FIELD TEST The objectives of the third and final phase of the study were to: Test the applicability of forms and procedures developed during the 20-hospital field test, Exhibit IV. Patient charge ledgers reviewed Total ledgers Total ledgers Total ledgers found requested Hospital Randomly | Purposively Information | Information selected selected Found | Net found complete not complete Total--==-==m=--mmmenn- 2,038 12,084 1,904 1,649 3891 1,904 180 1,064 840 84 6 65 25 6 59 80 20 98 2 32 66 71 23 97 3 61 36 74 25 96 3 43 53 82 18 100 - 42 58 75 25 97 3 55 42 76 24 99 1 32 67 90 10 68 32 40 28 72 31 103 58 45 73 26 82 17 35 47 89 8 66 3) 58 8 74 26 98 2 48 50 15 25 98 2 47 51 80 20 100 = 59 41 72 28 93 7 49 44 75 25 93 7 58 35 75 25 75 25 3 72 79 24 103 - 92 11 53 - 45 8 45 - 50 = 41 9 40 : § 47 - 47 - 46 1 47 - 45 2 41 4 150 - 95 1 74 21 Ipifference between total ledgers requested and total ledgers accounted for by extra ledgers from hospital No. 25. “Computer system, charge data on magnetic tape. In process of conversion to computer system. 450 ledgers were requested in hospital No. 25 but 96 were searched for. including procedures for preparing an in- struction manual for use by hospital employ- ees who would perform the records abstract- ing. Determine the accuracy and validity of the abstracts prepared by hospital employees by having a study team member independently abstract the same patient charge ledgers. Develop findings and conclusions on the feasi- bility and practicality of establishing a con- tinuous program for collecting patient charge Ledgers not available. and source-of-payment data as a part of the Hospital Discharge Survey. Phase III included the following steps. Selection of Five Hospitals In the first step, five hospitals already par- ticipating in the Hospital Discharge Survey were selected for testing the accuracy and validity of patient charge abstracting by hospital personnel. (See Exhibit III for characteristics of these hos- pitals.) Hospitals with computerized accounting systems were excluded as a result of findings from the field test that special programming was required to retrieve the necessary patient charge and payment data from automated sources, and that consequent delays would be encountered, Administration of the Interview Schedule In step 2, a general interview was conducted by a study team member with each hospital administrator, and a followup interview was held with the accounting manager or a designated em- ployee im the accounting department, Preparation of the Instruction Manual In this step, a member of the study team: Modified the basic instruction manual to make it applicable to the individual hospital and to guide the hospital employee involved in se- lecting and abstracting patient charge ledg- ers, Adapted the manual to each hospital, This was made possible by a format which listed alternative instructions from which the per- tinent instruction for the individual hospital could be selected. (See Appendix E for Hos- pital Instruction Manual.) Explained the instruction manual to the se- lected hospital employee and demonstrated abstracting procedures, Supervised the completion of two or three abstracts by the hospital employee, for train- ing purposes, and answered any questions, Provided the hospital with a supply of blank abstracting forms and the names of patients whose patient charge ledgers were to be ab- stracted. These names were selected from the list of patients whose medical records had been previously abstracted for the Hospital Discharge Survey, 10 Abstraction of Records by Hospital Employees An employee in each of the five hospitals utilized the instruction manual and prepared ap- proximately SO abstracts of patient charge ledgers after the study team member had left the hospital, (See Exhibit IV for data on ledgers abstracted.) Audit of Abstracting Performed by Hospital Personnel The last step of the study consisted of a re- turn visit to each hospital by a study team mem- ber for the purpose of: Independently abstracting the same patient charge ledgers that had been abstracted by the hospital employee, Comparing abstracts completed by the hos- pital employee and the team member to de- termine the accuracy and validity of data collected by the hospital employee, Revising forms and instructions based on findings from the audit of hospital employee performance, LIMITATIONS IN PILOT STUDY Principal limitations in the pilot study on hospital patient charge statistics were the fol- lowing. Hospitals with automated patient charge sys- tems were not analyzed in depth in the study because special programming would have been required to retrieve the patient charge data to be abstracted. It was not possible to include the number of large hospitals prescribed for the field test (six) because most large hospitals were al- ready participants in the Hospital Discharge Survey and because the number of large church-affiliated and proprietary hospitals is very limited. The purpose of the survey was to conduct a "one-time" study of short duration, and therefore the survey did not test the extent of hospital cooperation for a continuing pro- gram of abstracting hospital patient charge data. Also, the study did not provide con- ditions for preparing the staffing require- ments, procedures, and forms for anongoing abstracting program as compared with a "one-time'' program. The analysis of the patient charge data col- lected in the small sample of 25 hospitals cannot be considered representative of the total universe of hospitals in the United States, Il. MAJOR CHARACTERISTICS OF PATIENT CHARGE SYSTEMS IN HOSPITALS The purpose of this chapter is to describe patient charge systems as observed in the 25 test hospitals prior to the presentation of findings and conclusions on the feasibility of initiating a pro- gram of abstracting patient charge data on a continuing basis, The chapter provides informa- tion on the major components of patient charge systems relating to the format and content of a completed patient charge ledger, and the pro- cedures for creating, updating, and maintaining patient charge ledgers in hospitals. TYPES OF HOSPITAL PATIENT CHARGE SYSTEMS Two major types of patient charge systems were observed in the surveyed hospitals—the itemized charge system wherein a charge is posted to the ledger for each supply and service provided to the patient, and the all-inclusive per diem rate system whereinall patients are charged the same flat rate per day regardless of the amount and type of supplies and services pro- vided, THE COMPLETED PATIENT LEDGER The patient charge ledger is the only hospital record of the supplies and services provided to a patient, the charges made, and the payments re- ceived. The information on the ledger is used by the hospital for billing (invoicing) responsible parties for payment and for accounting audits. A completed patient charge ledger contains the information on all hospital charges to the patient and the dollar amount of these charges as well as all payments received by the hospital, a description of the source(s) of payment, andthe dollar amount of responsibility of the source of payment (the source of payment may pay less than its responsibility because of discounts, maxi- mum allowances, etc.). These two items constitute all the data required for patient charge and source-of-payment abstracting. In addition to information on patient charges and source of payment, the patient charge ledger also includes the patient's name, admission num- ber, and admission and discharge dates, These three items comprise all the information needed to locate a patient's charge ledger in the file and to verify that the ledger is for the same patient and admission selected for patient charge ab- stracting. A completed ledger frequently contains other information such as admitting diagnosis, attending physician's name, type of accommodation, etc., but this information is not necessary for patient charge and source-of-payment abstracting. Information on patient ledgers was relatively constant among the survey hospitals, but signifi- cant variations existed in the following areas: Ledger Format Some hospitals use single-column ledgers, whereby descriptions of all chargesand payments 11 EXHIBIT V. SAMPLE OF SINGLE COLUMN PATIENT CHARGE LEDGER iLL TO! TELEPHONE are placed in the same column and usually arc posted consecutively by date of occurrence, These postings may use color (red) or symbol codes for denoting debits or credits. Other hospitals use multiple-column formats, whereby charges are posted to specific prelabeled columns (e.g., col- umn may be labeled ''laboratory'), Again, the postings are consecutive by date, In hospitals using multiple-column ledgers, a description of the source of payment is usually spread across several columns unless the source is coded, in which event any column can be used, Exhibit V is a sample single-column ledger, and Exhibit VI is a sample multicolumn ledger, Equipment Used for Posting The design and format of a hospital's patient ledger is usually controlled by the type of ac- counting equipment used by the hospital, Three types of equipment were observed, Bookkeeping machines. — Entries are man- ually typed and automatically calculated, Tabulators.—Entries are manually typed on cards via keypunch and fed into the tabulating equipment, which automatically posts, calcu- lates, and prints charges and sources of pay- ment on the patient charge ledger, Computers.-—-Entries are manually typed on cards via keypunch and converted to mag- netic tape from which calculations are made on charges and sources of payment, Among the surveyed hospitals, bookkeeping ma- chines were by far the most frequently used type of equipment, Number of Postings Made Hospitals with all-inclusive rates often post only one charge on a ledger, whereas hospitals with itemized charge systems post for each in- dividual charge made. Variation among the hos- pitals using itemized systems is significant be- cause room and board charges may be posted daily or only once per patient stay, While personal charges, such as telephone and television service, may be posted daily, once per patient stay, or may not be posted at all since they may be billed separately, included in other charges, or not pro- vided, Extent of Computations Grand totals for room and boardand personal charges may or may not be shown on the ledger, Method of Describing Charges and Sources of Payment Hospitals may describe the type of patient charge or source of payment on the ledger by spelling out in full or abbreviated form or by alphabetical or numerical code. Method of Handling Writeoffs Some sources of payment (e.g., Blue Cross, commercial insurance paying for Workmen's EXHIBIT VI. SAMPLE OF MULTIPLE COLUMN PATIENT CHARGE LEDGER ADDRESS PATIENT oT ee = CLAss | room | RaTE [ Doc Tor ADMITTING NO. aL to | 1 _ = = - = En - a —— [oare erp DISCHARGED UNIT NO. a - { at : - L ~ a — CONTRACT NO. INSURANCE . _ oo I I 1 2 3 operating s ¢ i g MISCELLANEOUS TOTAL CREDITS OLD BALANCE RM. & BD. IVERY RM. ELECTRO-| MED. & AN DRUGS | LAB DELIVER X-RAY | ‘CARDIO- [SURGICAL | BLOOD CHARGES: | fees rE DATE BALANCE NURSING RATORY |, 0unr fooe GRAM [SUPPLIES AMOUNT [oo amount Fooe | | | | | | | | 4 rs sEeet i edb i TOTAL CHARGES COVERED BY INSURANCE |CHARGES TO PATIENT PAYMENTS 1. ANESTHETIC MATERIALS 4. TISSUE CHARGE 7. EMERGENCY ROOM ALL BILLS ARE PAYABLE UPON PRESENTATION CODE FOR 2. RECOVERY ROOM 3. OXYGEN LYy MISCELL ANEOUS 3. LV. SOLUTIONS 6. TELEPHONE OTHERS CHARGES ARRIVING AT CASHIER'S OFFICE AFTER DISCHARGE WILL BE BILLED AT A LATEX DATE vE-% THE DATES INDICATED FOR THE ABOVE CHARGES ARE THE DATES THEY WERE RECORDED AND ARE NOT NECESSARILY THE DATES ON WHICH THE SERVICES WERE RENDERED. Compensation, and welfare) often do not pay the hospital the full amount of their dollar responsi- bility. When posting these discounted payments, some hospitals show both the amount of the pay- ment and the writeoff, Other hospitals post only the amount of the actual payment to the ledger and subtract the full amount of the responsibility. All patient charge ledgers, regardless of the patient charge system, accounting equipment, or ledger format used, provided for the posting of dollar amounts in the form of debit or credit en- tries, Charges for supplies and services provided patients were posted as debits. Payments for such supplies and services provided are posted as credits. All debits are added to the balance due and all credits are subtracted. PROCEDURE FOR CREATING, UPDATING (POSTING), AND FILING PATIENT CHARGE LEDGERS Creating Patient Charge Ledgers in Hospitals The patient charge ledger is usually created by a clerk in the admitting or accounting office at the time the patient is admitted to the hospital, However, in many hospitals, the admitting office is managed by the controller or business man- ager since the major portion of the admitting information is obtained for the hospital's ac- counting department, 13 The patient's name and address, admission number and date, and name and policy number of the responsible source of payment, if other than the patient are recorded on the ledger at the time of creation. (This information may be first recorded on an admitting form and then trans- ferred to the ledger.) The necessary information is usually typed or written in the heading portion of the ledger, In computerized hospitals, this information is ob- tained in the admitting office and later key- punched and entered into the system through a card-to-tape converter, Posting Charges to the Patient Charge Ledger All charges for supplies and services pro- vided, for which the hospital expects to receive payment, are posted to the ledger by a posting clerk in the accounting department, Ona multiple- column ledger, dollar amounts of the chargesare posted to the column prelabeled for that charge or type of charge. On a single-column ledger, the dollar amount for each charge is posted to the single column, In hospitals using itemized charge systems, charges are posted to the patient's charge ledger usually within 24 to 48 hours after the supply or service is provided. Charge slips are prepared for each supply and service by the servicing departments, These slips are then for- warded to a posting clerk in the accounting de- partment, who transfers (posts) the information from the charge slip to the patient's ledger. The following entries constitute a typical posting operation--date of posting, description of the charge, amount of the charge, and amount of the new balance due, This operation is repeated un- til all charges are posted to the patient's charge ledger. In hospitals withall-inclusive rates, charges are usually posted to the ledger only on the day the patient is discharged. The amount to be posted is calculated by determining the number of days the patient remained in the hospital and multiply- ing this figure by the all-inclusive rate, The re- sulting total is then posted to the patient charge ledger. 14 Posting Payments to the Patient Charge Ledger All payments received by the hospital for supplies and services provided are posted to the patient's ledger by a posting clerk in the ac- counting department, Payments are posted as soon as they are received by the hospital, and the procedure is the same regardless of the type of charge system, The patient charge ledger for which a pay- ment is received is removed from the active ledger file and given to a posting clerk with the payment information. Then the dollar amount of the payment is posted as a credit, and the source of payment is spelled out or a code is employed for the entry. The following entries will be made in posting a payment-—date of posting, a description of the party making the payment, the amount of the payment, and the amount of the new balance, This procedure is repeated as each payment is received until all charges are paid. Filing the Patient Charge Ledgers A hospital maintains a series of files for patient charge ledgers, The particular file within which a patient charge ledger is placed depends upon whether all charges have been posted andall payments have been received or whether payments are substantially overdue, While a patient is in the hospital and for a short time after his dis- charge, ledgers are kept in a file for accounts in which all charges have not been posted, After posting is complete, the patient ledger may be found in one of several files, depending on the status of the account, To find a patient ledger for a discharged patient, therefore, several files may need to be searched. Ledgers are filed alphabeti- cally by the patient's last name in each file, Patient charge ledgers are maintained in the following hospital files: The open file, maintained by a clerk responsi- ble for posting, contains files for which all charges have not been posted, The active file, maintained by a clerk re- sponsible for accounts receivable, is com- posed of ledgers for which all charges have been posted but for which payments are still due, The agency file, maintained by an accounts receivable clerk, consists of accounts sub- stantially past due that have been referred to collection agencies. A hospital may not always keep a record of individual ledgers missing from its files that have been trans- ferred to a collection agency, The completed file, maintained by a ledger file clerk, is composed of ledgers for which all charges have been posted and payments made, The size of the hospital determines the num- ber of personnel involved in posting charges and payments and maintaining ledger files, In small hospitals one person may perform all functions whereas in larger hospitals one or more persons may be responsible for each activity, IV. FEASIBILITY OF ABSTRACTING PATIENT CHARGES ON A CONTINUING BASIS The analysis of the data from the 25 hospi- tals included in the pilot study permits findings and conclusions to be made on issues of abstract- ing in a broad spectrum of noncomputerized hos- pitals of different size, location, and type. Such an analysis can assist in deciding the feasibility of abstracting patient charge data on a continuing basis for a larger national sample of hospitals. MAJOR ISSUES IN ABSTRACTING PATIENT CHARGE LEDGERS Issue 1 Can patient chavge ledgers and medical records beabstracted for the same sam- ple of patients and admissions? Findings A patient charge ledger was prepared for each patient and for each admission except new- born infants in the hospitals. Charges for well- newborn infants who were discharged from the hospital at the same time as the mother were posted to the mother's ledger in 22 of the 25 hos- pitals. (Three hospitals prepared separate ledg- ers for newborn infants.) In hospitals using single ledgers for both the mother and child, it is im- possible to distinguish from the ledger whether certain charges are for supplies or services pro- vided to the mother or to the child (e.g., drugs, special nursing, and central supplies). Oo Of the ledgers requested, 8.5 percent were not found. Major reasons for not finding all re- quested ledgers in the 25 hospitals were as follows: Hospital employees locating the ledgers often searched only one hospiral file for a ledger. Ledgers filed in separate facilities were not searched for. (These were usually completed ledgers for which all payments had been received.) In the 20-hospital field test, some ledgers were requested for patients discharged within the previous few days, and the hospital did not wish to remove these ledgers from the open file. Limited time available in the 20-hospital field test did not permit an extensive search of the files. In the five final test hospitals where sufficient time was provided, two hos- pitals with problems in locating ledgers were certain that the ledgers had been transferred to a collection agency. Conclusions It is possible to use the same sample of pa- tients and admissions for patient charge and med- ical record abstracting except for well newborn infants, whose charges are usually combined with and are often indistinguishable from those of the mother on the same ledger. 15 In a "one-time" study, such as the pilot study, missing ledgers are a problem. The ledgers are not lost, but are difficult to locate since they may be kept in files (particularly the closed file for completely paid accounts and the collection agency file for overdue accounts) that are placed in sep- arate facilities or at some distance from the ac- counting department. In a continuing abstracting program, however, the percentage of ledgers not found should not be significant for the following reasons: Hospital employees will have more time and will search more thoroughly for all ledgers requested. Ledgers for recently discharged patients will not be included. Ledgers will be abstracted in the hospital accounting department before being trans- ferred to separate facilities. Ledgers transferred to collection agencies will be abstracted after they are returned to the hospital. Issue 2 Can patient chavge and medical vec- ovds data abstracts, once prepared, be matched for the same patient and ad- mission? percent of the ledgers were complete. In the other hospital, welfare, whose payments are frequently delayed, was the predominant source of payment, resulting in the high percentage of incomplete ledgers. Conclusions Patient charge and medical records data are not available at the same time, but it is pos- sible to match the two data sets. This can be done in three locations—in the hospital by the accounting department, in the hospital by the medical records department, and in Washington, D.C., by NCHS, Since medical records and patient ledgers for the same patient are generally completed at different times, separate lists of patients selected for patient-charge-ledger and medical- record abstracting will be required in a con- tinuing abstracting program. Unless arrange- ments are made to rely upon the coordination of the two surveys within each hospital, both the medical records and account departments will have to prepare separate listings of patients sampled for later cross-matching by NCHS, Issue 3 Can voom and board chavges be identi fied on patient chavge ledgers? Findings Abstracting medical records and patient charge data can only be done when the source record is complete. Medical records and patient ledgers are com- pleted at different times in most hospitals. Medi- cal data originate in the medical records depart- ment and are usually available 2 to 4 weeks after the patient is discharged. Charge and source pay- ment data originate in the accounting department and are usually available in 90 percent of the hospitals within 60 days after discharge for 90 percent of the patients. This was determined by direct questioning of the hospital accounting managers and then verified by the study staff during actual abstracting in the 20 field test hos- pitals. In the five final test hospitals, all ledgers requested were for patients discharged 60 days or more earlier; in four of the five hospitals 90 16 Findings Patient ledgers in hospitals that charge ac- cording to each supply and service provided have room and board charge data identified separately on the ledger. Room and board charge data do not exist on patient ledgers in hospitals that charge an all-inclusive per diem rate. Some of these hospitals have cost accounting breakdowns of the all-inclusive rate which can be used to calculate per diem room and board charges. The all-inclu- sive-rate hospitals that do not have costaccount- ing can provide estimated room and board charge data which they already supply to third-party payors upon request, Four of the 25 hospitals used all-inclusive- rate charge systems. In two of these hospitals room and board charges had been determined by cost accounting, and in the other two such charges could only be estimated. Conclusions Accurate room and board charge data can be identified in all hospitals that charge according to each supply or service provided and in all-inclu- sive-rate hospitals where cost accounting exists. The room and board charge data obtained from all-inclusive-rate hospitals without cost accounting breakdowns will be estimates of actual charges. Issue 4 Ave items included in room and board chavges consistent among all hospitals? Findings In most survey hospitals the room and board charge was generally considered a catchall item. Any item that is not separately charged for is incorporated in the room and board charge. The major portion of the items which is in- cluded in room and board was consistent among the hospitals surveyed. The following supplies and services were included in room and board charges in all 25 survey hospitals—room, food (routine and special diets), routine nursing, laun- dry (linens, etc.), overhead (administration and supervision), and building and building equipment (maintenance of facility, depreciation, heat and light, etc.). These supplies and services, consistently included in the room and board charge of all survey hospitals, constitute the bulk of items and prices in the room and board charge. Lab- oratory services were excluded from the room and board charge in all 25 hospitals. Exhibit VII analyzes variations in selected items included in room and board charges in the 25 hospitals. Conclusions There is sufficient consistency in the ele- ments that make up room and board charges to permit gross comparison of room and board charges among hospitals. In all hospitals, the room and board charge includes the same major items including room, food, routine nursing, laun- dry, administrative overhead, and building and equipment maintenance. Issue 5 Can charges other than room and board (other charges) be identified on patient] ledgers? Findings Although per diem room and board charges are usually similar for all patients in hospitals which charge for each supply and service pro- vided, other charges will vary from patient to patient according to the type and amount of sup- plies and services provided. Other charges can be most easily computed by subtracting room and board charges from total charges less personal charges on the patient ledger. Other charges in all-inclusive-rate hospitals can be computed on an average per diem basis for all patients; in other words, other charges in all-inclusive-rate hospitals will be the same per day for all patients. Conclusions Other charges can be computed in all hospi- tals, but only accurately for each patient in hos- pitals that charge according to each supply and service provided. Therefore, comparison of other charges among patients (e.g., by diagnosis, length of stay, etc.) will be reasonable only in hospitals which charge according to each supply and service provided; such comparisons will not be meaningful among all-inclusive-rate hospitals since per diem other charges will be the same for all patients. Issue 6 Ave items included in othev charges con- sistent among all hospitals? Findings Only supplies and services billed for by the hospital are included in other charges on the patient charge ledgers. Exhibit VIII summarizes the variations of items billed for as other charges in the 25 hospitals. Not all supplies and services provided in the hospital are billed for by the hospital. Supplies and services may be billed for by various other parties, including physicians, private nurses, hospital auxiliary groups, and agencies other than hospitals. Exhibit VII. Variations in supplies and services included in room and board charges in 25 survey hospitals Drugs Special equipment and services Hospital 1 7 i lood 0 Surgica Recovery ntensive | Intravenous | Bloo Xygen Routine | Special Eo i room care feeding setups | setups Lm mm X X X X 2eccmmm mmm ————— x X X X X 3mm m————————— X X X X NP X X X Lo ro wr Xx X X NP X X 15 X X X NP X 16m X xX X X X 5 we 0 X X x X NP X X X BB ime i X NP X X X CF X X X X X TC orc nv a, X X NP X X X TL i i x X X X X X 12-m mmm meee X X X X X X 13--mmmme meme ms X X X X X X 1fmmm mmm meme mee X X X NP X X X 15-mmmmm mmm eee em X X X X X X 16m -mmmmmmm meena X X X X X X me X X X X X X 18--mmmm emma X X X X NP X X X 19mm mmc cece x X X X X D0 reve emt me i mn X X X NP X X X D1, wm mm mmm X x X NP X X X 22m mmmee nme ————— X X X NP X X X 23m mmm mmm emo X X X X 2m mmo mm om mo ee X X X X X X 25m mmm ————————— X X X X X X Percent of hospitals!==-n-nn- 20 92 68 72 57 88 92 92 Percent of hospitals for which each supply or service is not included in room and board. NOTE: The following supplies 25 survey hospitals: room, food etc.), and overhead (administration, supervision facility, depreciation, heat and light, etc.). and board charge in all 25 survey hospitals. Legend: and services were included in the room (routine and special diets), , building and building equipment (maintenance of and board charge in all routine nursing, laundry (linens, Laboratory services were excluded from the room X = Supply or service not included in room and board charge. NP = Supply or service not provided by the hospital. Items for which the hospitals bill under other charges are constantly changing because of new programs (e.g., Medicare), decisions to separate or combine billings with those of the hospital (e.g. pathology and radiology), and changes inaccount- ing standards. Medicare may increase the frequency of bil- ling by parties other that the hospital for supplies and for services provided in the hospital because of its required separation of charges for certain services such as pathology and radiology. Conclusions Not all other charges for supplies and serv- ices received by the patient in a hospital will be 18 included on the hospital's patient charge ledger. Although variations occur among hospitals on the items they include in the category of other charges, gross comparisons of other charges among hos- pitals can be made since significant variations occur only for a few items, particularly for anes- thesia services. Comparisons can be made more meaningful by obtaining periodic information from each hospital on what items they include in other charges. Issue 7 Can chavges for blood be identified on patient ledgers? Findings Charges for blood were posted to the patient's ledger in 23 of the 25 survey hospitals. In the two Exhibit VIII. Method of billing for selected other charges ' Regular Private Hospital Pathology | Radiology Ancfhagi- Blood Sugseon 8 ese duty gy ee visits nursing Field test hospitals —emmmmmmmmmm mmm ————— H H 0 H 0 0 0 et mm H H H H 0 0 0 sm H H H H 0 0 0 eee mmemmm———————————— H H 0 HM 0 0 0 et H H 0 H 0 0 0 mo H H 0 H 0 (1) (1) ce ro H H HM H 0 0 0 meme mm—m———————————— H H 0 H 0 0 H te H H 0 H 0 0 0 meme me em ————————— HM HM HM H 0 0 0 et EO H H 0 0 0 0 0 JE) H H H H 0 0 0 mt a H H H HM 0 0 0 0 0 H H H HM 0 0 0 a 0 H H H H 0 0 0 mm 0 H HM 0 HM 0 0 0 SI —— H H 0 H 0 0 0 ies et eB BB 0 H H 0 H 0 0 0 eee mmm mmm————————————— HM H H H H H 0 NR HM H H H 0 0 0 H H 0 H 0 0 0 H HM 0 H 0 0 0 H H H H 0 0 0 H H 0 H 0 0 0 HM HM H H 0 0 0 Percent of hospitals®=---- 100 100 48 96 4 4 4 ISupply or service is not provided at the hospital. 2Percentage of hospitals included that billed for specified charges. Legend: H HM physician sends a separate bill. o n other hospitals, blood was provided free through the community's blood bank. Patients were never billed independently by the blood bank. The posting of charges for blood is similar to any other charge posting. When blood is pro- vided the charge is posted to the ledger. If blood is paid for by replacement through donors, a credit posting similar to any other credit posting is made. Conclusions Identifying charges for blood will be a prob- lem only if hospital blood banks bill the patient directly and the hospital does not post the charge to the ledger. (This did not occur in any of the Billed as an individual charge by the hospital. Billed as an individual charge by the hospital except for Medicare patients where the Billed separately by the physician, auxiliary group, or agency (other than hospital). 25 survey hospitals.) In the hospitals that provide blood free of charge, the charges to the patient are truly reflected on the patient charge ledger since there is no blood charge or payment. For hospitals which accept or prefer pay- ment for blood through donor replacement (pay- ment in kind), it will be necessary on the patient charge abstract form to record the credits re- placements as personal payments. Issue 8 Can personal (nonmedical) charges be identified on patient ledgers? Findings Personal charges noted on patient charge ledgers included those for television and tele- 19 phone. Personal charges were not hidden in any other hospital charge in the 25 survey hospitals. Telephone service was provided in 23 of the 25 hospitals and the charges for both long dis- tance and local telephone calls could be identified from the patient's ledger in all 23 hospitals, Television was provided in 24 of the 25 hos- pitals. Charges for television, however, could be identified from the patient's ledger in only 11 of the hospitals (44 percent) because an auxiliary group or television service in the other hospitals independently billed the patient for the service. Conclusions Television and telephone service are the only frequent personal charges that will be encountered on patient charge ledgers, and these charges can be identified from the patient charge ledger only when the hospital does the billing. The hospital will usually bill for telephone service, but other parties such as auxiliary groups will often bill independently for television. A continuing program for abstracting patient charge data may require identification of personal (nonmedical) charges since such charges can distort an analysis of patient medical costs. Per- sonal charges should be deducted from total hos- pital charges before any calculation is made of room and board and other charges. The number of personal charge entries varies considerably among hospitals, and will influence abstracting time significantly in some hospitals. Issue 9 Can source of payment of patient charges be easily identified from patient ledgers? Findings Of the 10 sources of payment prescribed for the study, 7 could be identified with consistency from patient charge ledgers in all 25 hospitals and 3 in only some hospitals, They are: Identified in all hospitals Blue Cross Commercial insurance Personal payments Welfare (Including Title XIX-Medicaid) 20 Hospital discounts Private charity Medicare (Title XVIII) Identified in some hospitals Federal public service programs Other service programs Workmen's Compensation insurance Of the seven sources of payment that can always be identified on patient ledgers, five ac- counted for more than 97 percent of total pay- ments made to the hospitals. Major sources of payment In percent Total -=- =m meee 97.4 Personal payments----------_ 43.0 Blue Cross-------ccoooo___ 30.2 Commercial insurance ------- 18.0 Wel AL Ew mime mim mw sisi css sin sim 5.5 Medicare (Title XVIII)-=------ 0.7 Medicare, which had only been initiated at the time of the study, will account for a greater per- centage of payments in the future, The seven sources of payment most frequently identified in the hospitals (except Medicare) pay the hospital directly. Medicare, paid through an intermediary agent, can be identified easily be- cause the Social Security Administration requires that Medicare payments be marked in red on the ledger. Workmen's Compensation insurance cannot always be identified from ledgers because it is often paid by commercial insurance or govern- ment agencies that may be recorded by the hos- pital as the source of payment, Federal public service programs and other service programs also often do not pay bills di- rectly but rather through intermediary agents that are identified as the sources of payment on the patient charge ledger. With special effort Workmen's Compensation often can be identified, but this is not true with Federal public service and other service programs. Writeoffs (discounts) are frequently made by hospitals for some sources of payment. Discrep- ancies, therefore, frequently occur between the amount of patient charges for which the source of payment is responsible and the amount that it actually pays. Personal payments to the hospital by patients often include third-party payments made to the patient as well as his out-of-pocket payments. These third-party payments to patients are rarely recorded on the patient charge ledgers. Conclusions Except for Medicare, sources of payment can be accurately identified only when the source pays the hospital directly. Therefore, Workmen's Com- pensation, Federal public service programs, and other service programs are probably not appro- priate sources of payment to use in obtaining ac- curate information from patient charge ledgers. These three sources, furthermore, accounted for a minor proportion of total payments to the 25 survey hospitals. (See Appendix F for suggested patient charge abstract form for future use.) Information can be readily obtained for more than 98 percent of the total payments to the hos- pitals (five of the seven easily identifiable sources of payment accounted for 97.4 percent of the total payments) if it is recognized that the amount of third-party sources of payment will be understated and that personal payments will be overstated, since third-party payments made directly to the patient will be posted as personal payments. Although personal payments are one of the most frequent sources of payment (43 percent of sources of payment identified on all ledgers), they are often only for personal charges such as television and telephone, which constitute less than one-half of 1 percent of total dollar charges (although they may be significant in some individ- ual cases). To collect meaningful information on personal payments as a source of payment, a subdivision of personal payments into payments for nonmedi- cal charges (e.g., telephone and television) and payments for medical charges may be required. By dividing this source of payment into two divi- sions, it will be possible to analyze patient respon- sibility more objectively and, if desirable, to discount patient responsibility as a source of payment when the payment is for telephone and television only. Also, a provision in the abstract form for a writeoff (discount) for Blue Cross, commercial insurance, and welfare, when the actual payment is less than the source of pay- ment's financial responsibility, may help assure the collection of accurate data on who pays patient charges in hospitals. Issue 10 Can valid patient chavge data be oblaine ifabstracting is done by hospital employ- ees? Findings In the five final test hospitals in which hospi- tal employees and study staff independently ab- stracted the same patient charge ledgers, em- ployee abstracting errors occurred in 6.2 percent of the abstracts, as follows: Six arithmetic errors, one patient charge overlooked, two figures copied incorrectly, and two incorrect sources of payment recorded. Most errors found are correctable since arithmetic errors can be controlled by validity checks, and incorrect sources of payment can be avoided by clarification of definition of source of payment in the instruction manual for hospi- tal employees. Incorrect transposition of figures and over- looked figures cannot be controlled. Uncontrol- lable errors occurred in only 3 out of 176 ab- stracts, or 1.7 percent of abstracts completed. Conclusions Hospital employees can abstract patient charge ledgers with considerable accuracy if properly trained. Proper training will require personal contact with the hospital employee and the use of a carefully developed instruction man- ual for teaching and reference. 21 Issue 11 What will be the cost to hospitals to ab-| stract patient charge data on a contin- uing basis? Findings Average time for a hospital to complete ab- stracts of patient charge ledgers varied consider- ably among the survey hospitals. Exhibits IX and X summarize the time required to abstract patient charge ledgers by study staff in the 20 field test hospitals and by hospital employees in the 5 final test hospitals. Average time required to abstract a ledger ranged from 3.1 minutes to 12 minutes for the following reasons: Accessibility of the files. Allfiles may not be in the same location. Quality of record maintenance. Files not located averaged 8 percent and ranged from 0 percent to 32 percent among hospitals. Number of room and board postings. Hospi- tals may post room and board charges daily or only upon discharge. Extent of computations on the ledger. Some hospitals have already totaled room and board and personal charges and these figures need only be transposed without computation. Number of postings. Low number of postings in hospitals with all-inclusive rates makes abstracting easier. Frequency of posting of personal charges (e.g., telephone and tele- vision) depends upon hospital systems and policies and will vary considerably. Ability and familiarity with the procedures. Quality of staff assigned by the hospital and training of these personnel influence the rate of abstracting. Time required by hospital employees to find and remove patient ledgers from the files aver- aged 32 percent of total abstracting time for the 25 hospitals and more than 50 percent of abstract- ing time in several individual hospitals. The difference in time required by study staff and hospital employees to abstract a patient charge ledger was not great (4.0 minutes for study staff and 4.5 minutes for hospital employees). Hospi- 22 tal employees located ledgers in both parts of the survey, and the average locating time in the field test hospitals (1.9 minutes) was more real- istic for future estimates of time requirements because one of the final test hospitals required excessive ledger locating time. Conclusions Based upon results of the survey and an as- sumed employee cost of $2.50 per hour, the cost to the hospital to prepare 100 abstracts will range from $12.50 to $50.00, or an average of $25.00 for 100 abstracts. Reimbursement to hos- pitals may need to be on a variable rather than a flat rate basis since time for abstracting can vary considerably among hospitals based upon their individual accounting system characteris- tics. Issue 12 When will patient charge ledgers be ready for abstvacting? Findings Since source-of-payment data can only be accurately identified after all payments are re- ceived, patient charge ledgers cannot be ab- stracted until all charges and payments have been posted to the ledger. The time required for this complete posting of payments varied considerably among survey hospitals. The time required after discharge of patients for 90 percent of the ledg- ers to be completed in the 25 survey hospitals was: within 15 days for 1 hospital, within 30 days for 8, within 60 days for 14, and more than 1 year in 1 hospital. This time lag was determined by direct questioning of the hospital accounting manager and then verified by the study team while abstracting. In 23 of the 25 hospitals, 90 percent or more of the patient charge ledgers were ready for abstracting within 60 days. Variations among the study hospitals in the time required to complete ledgers primarily result from the degree of efficiency of the hos- pital's billing procedure and the dominant sources of payment in the hospital (e.g., Blue Cross pays more rapidly than welfare, which often delays payment for months). Exhibit IX. Time required by study team to abstract patient charge ledgers Time to Time to Average Fila test hospital. | hs | he oe (hours) (hours) (minutes) TO Lomb srs rm 0 30 HEB SS 0 0 im 1,631 50.8 109.1 5.9 7 70 ER RT A Re HE SESE 3 wei = woe rope Sires srt ma 65 2.0 4.0 5.5 emi sc eo i i TEE R18 EH 98 1.0 4.0 3.1 3 mimic we ER 97 5.0 6.6 7.2 LR A A i APART msc mo wo wi me Aimar 96 2.0 4.0 3.7 Boon cn mr i 5 oR RRS RR RI i i 100 5.0 6.0 6.6 Bis ws srs we pcr msn ss BAS 97 2.5 7.5 6.2 TD SB et 1 A 99 3.0 9.5 7.5 02 ee ro SS ER 0 tt ei wwe oi, ro sm rm si i bir REA RS 68 4.0 8.0 10.5 simran Ee se were Hii en 103 2.5 8.0 6.1 [Lien rin sR RR EE ER IH rasan Siar ase 82 4.0 4.0 5.9 [Grose msm ss mo 58 A ER RRR 66 2.0 4.0 5.5 Le wm os im tn mn st SE 98 2.0 3.0 3.1 Lo Se ER ER TE RE Rp pases rm se A 98 1.5 4.0 3.4 LS msm smc bs ms 5 NE TBO 8 0 100 2.0 10.5 7.5 TB ces ss: sm eis oi em im tn Em 10 A LE SE ER 0 0 I I pr a RR Er A SR ARR I O's ip insmninmise arms onlo i 93 4.0 6.0 6.5 LB ms SR 58 8 tr 93 2.0 5.0 4.5 Yl mmm mm mm spn mis 3s SAA BAR AA RS Wi SR 75 3.0 6.0 7.2 | Pp EE EE EE EEE La 103 3.3 9.0 7:2 IThe locating time 1s an estimate because this was done by hospital employees while the study staff member was abstracting. 2Computer system, charge data on magnetic tape. In process of conversion to computer system. Ledgers not available. Exhibit X. Time required by hospital employees to abstract patient ledgers in five final test hospitals Time to | Time to Average Zand Ledgers locate abstract |time per TpioYes Final test hospital reviewed | ledgers | ledgers abstract sepsleting (hours) (hours) (minutes) Total-=--====-cecmecccec meme mmm ———— 273 13.25 20.25 8.5 se eo 45 2.00 1.00 4.0 Accountant DD mm an ww we A 41 0.50 4.25 6.9] Clerk 23mmmmmmmmmm mmm em meme mmmm mmm mmm —————— 47 2.00 1.00 5.1| Controller 2 mr ms em ei A 45 1.75 7.00 11.7| Clerk mr i 95 12.00 7.00 12.0| Clerk (insurance department) 23 In all hospitals payments for certain types of cases (e.g., Workmen's Compensation or automobile accident) may not be received for several years, Conclusions More than 90 percent of patient charge ledg- ers will be complete and ready for abstracting in the great majority of hospitals within 60 days after patient discharge. Availability of completed patient charge ledgers will be significantly de- layed in some hospitals (particularly hospitals with a high proportion of welfare patients) and for some cases in all hospitals (e.g., Workmen's Compensation, automobile accident, and welfare). Issue 13 Will hospitals cooperate in a continuing program to abstract patient charges? Findings No hospital approached in the pilot study refused to cooperate, although considerable ex- planation and persuasion were required to achieve cooperation of several hospitals. No hospital raised objections to the releasing of patient charge data because of the issue of confidentiality of hospital records. In the five final test hospitals, certain prob- lems in implementing a continuing abstracting program were identified by hospital personnel, including extensive new manpower and space problems in the accounting department due to Medicare workload and prior commitments to other studies and surveys. Survey staff found that the most important factors that hospitals will consider in deciding whether to cooperate in a program of continuing patient charge abstracting are the time required by hospital personnel to participate in an abstract- ing program and the attitude of top hospital ad- ministrative staff to the value of providing data to sources outside the hospital. Conclusions The great majority of hospitals will cooper - ate in a continuing program to abstract patient charge data, provided the approach is carefully 24 planned and personal, the time required is held to a minimum by careful planning of hospital time requirements, and training of hospital per- sonnel is conducted by survey staff, Other factors of somewhat less importance in obtaining cooperation include making equitable financial reimbursement for required hospital time, stressing the national importance of the information gathered from continued abstracting, and providing an advance indication to the hospital of the amount of time required of hospital staff to abstract patient charge ledgers (such estimates can be based on findings of the pilot study). INTERPRETATION OF FINDINGS AND CONCLUSIONS In summary, the pilot study of hospital patient charge statistics in 25 hospitals throughout the United States supports the following conclusions. Conclusion 1 It is feasible lo collect patient charge and source-of-payment data Patient charge data can be abstracted since information is readily available in noncomputer hospitals, valid data can be collected by hospi- tal employees, and patient charge and source-of- payment data can be collected at a reasonable cost, Conclusion 2 Most of the information collected from an ongoing patient charge ab- stracting program will be accurate and useful; however, caution must be exercised when using some in- formation Some incorrect source-of-payment data will be collected because hospitals cannot determine whether personal payments by patients include third-party reimbursement made to the patients. Special studies will be needed to determine the extent of the third-party reimbursement that is made directly to patients. Patient charge data in all-inclusive-rate hospitals will not reflect the individual patient's treatment requirements because all patients in these hospitals have the same per diem charge. Also, in addition, most all-inclusive-rate hos- pitals do not have cost accounting, and therefore patient charge data cannot be broken down ac- curately between room and board and other charges. Only a limited number of hospitals presently use all-inclusive rates, so this prob- lem is not great. Items included in room and board charges vary among hospitals, although most of the items (i.e., room, food, and nursing) are consistently included in all hospitals. Therefore, valid com- parison of room and board data among hospitals is possible. Although items included in other charges differ somewhat among hospitals because of vari- ations in items for which the hospitals bill, gross comparisons of other charges can be made since significant variations occur with only a few items. Periodic information from the hospitals regarding items included in other charges can assist in making more accurate comparisons of other charges among hospitals. Personal charge practices vary among hos- pitals because of different methods of billing for television. Charges for television may be included on the patient ledger or billed by a separate group or agency. This should not create any major prob- lem because total charges for television account for a very minor percentage of total charges (less than one-half of 1 percent for both tele- vision and telephone). Personal (nonmedical) charges and payments may be eliminated altogether in a continuing abstracting program since they are small in amount. By separately collecting charge and payment data for personal items at the early stages of a continuing program, a decision can be made on whether personal charges should be eliminated in a full-scale abstracting program. Conclusion 3 Some technical problems must be resolved before implementation of a continuing program of patient charge abstracting Development of a Method to Insure That Medical and Patient Charge Abstracts Are Prepared for the Identical Discharge The following four alternative methods are available to ensure that the same sample of patients is selected for medical record and patient charge abstracting. (See Appendix G for a de- tailed description of these four alternatives.) Initial selection of patient sample by the ac- counting department from the daily admis- sions list, with referral of the sample list- ing to the medical records department. Initial selection of patient sample by the medical records department from the daily discharge list, with referral of the sample listing to the accounting department. Independent sample selection by the account- ing department from the daily admissions list and by the medical records department from the daily discharge list. NCHS referral of sample list for patient charge abstracting to the hospital accounting department after NCHS receipt of hospital medical record abstracts. Unless NCHS selects the last alternative, the choice of sample selection method should be left to the individual hospital. Permitting a choice from among the other three alternatives should assist in obtaining hospital cooperation. The selection of the sample by the account- ing department and referral of the sample list- ing to the medical records department is the most economical because only one hospital department has to select the sample and prepare the sample list. In addition, ledgers for later abstracting could be identified at the time of their preparation and the time-consuming and therefore costly process of locating completed patient ledgers would be essentially eliminated. Development of a Method for Matching Medical and Patient Charge Abstracts A decision must be made on the basis of the following choices as to where abstracts of medical record and patient charge data are to be matched. Abstracts can be matched at the hospital by the medical records or accounting departments, or at NCHS, with the two sets of abstracts being sub- mitted independently by the medical records and accounting departments. It would appear to be better for NCHS to match the medical record and patient charge 25 abstracts because hospital time and cost will be reduced, NCHS can match at a lower cost by computer use, and NCHS will need a matching program for control anyway. NCHS matching will also minimize the need for additional contact be- tween the accounting and medical records de- partments, which normally have limited associ- ation. In addition, completed abstracts for medical record data will arrive earlier at NCHS because they can be sent as soon as they are prepared without waiting for the delayed patient charge abstract. The possibility of abstracts being lost within the hospital while waiting to be matched will also be eliminated. Development of Methods for Abstracting Patient Charges on a Continuing Basis From Automated Data Sources It was decided not to abstract patient charge data for automated sources in the pilot study be- cause special computer programming to retrieve patient charge data from automated sources would be costly for a one-time survey, and be- cause time requirements to complete such pro- gramming would have substantially delayed the study. Since automated data system hospitals were deemphasized in the survey, the following ques- tions still need to be answered: Is data on patient charges retrievable from automated sources? Can patient charge abstracting be automated? Will hospital staff and computer time be available to develop programs to retrieve data? Will machine time be obtainable on a periodic basis for abstracting? Can valid patient charge data be collected from automated sources at a reasonable cost? Will different interview schedules, instruc- tion manuals, and abstract forms be required? Until these questions are answered, NCHS cannot begin abstracting patient charges ona con- tinuing basis in hospitals with autoniated patient charge systems. Very few hospitals now have fully automated patient charge systems, and even in these hospitals it will often be possible to abstract patient charge information using the same procedures as in nonautomated hospitals since printed ledgers are available. However, as the number of hospitals with automated patient charge systems increases, programs to abstract mechanically will be needed because many hos- pitals will otherwise be unable or unwilling to provide the data. 000 26 APPENDIX A GLOSSARY OF TERMS Accounting Equipment Bookkeeping machine.—Equipment used by most hospitals for preparing patient charge ledgers and posting patient charge and source-of-payment information to the ledgers. The machine consists of a typewriter keyboard and a calculator. Computer. — Equipment used in some of the larger hospitals for preparing patient charge ledgers and posting patient charge and source-of-payment information to the ledgers. The ledger in these hos- pitals is often on magnetic tape and a program is required to abstract the patient charge and source- of-payment data. Tabulator.— Equipment used in a few hospitals for preparing patient charge ledgers and posting patient charge and source-of-payment information to the ledgers. Its capabilities are midway between those of the bookkeeping machine and the computer. The tabulator can compute automatically but cannot store information on magnetic tape. Accounting Terms Debit.— Accounting term for an entry on the side of an account (ledger) showing indebtedness. All charges to the patient are posted to the debit column of the ledger. Credit.— Accounting term for an entry to an account (ledger) of the reduction of a debit. All pay- ments to the hospital for charges incurred by a patient are posted to the credit column of the ledger. Balance due.— Amount of money owed to the hospital. When balance due is zero, the ledger is com- pleted, or paid. Charges Other chavges.— All charges to the patient, except room and board and personal, which are posted to the patient charge ledger. Personal chavges.—All nonmedical charges to the patient for items such as telephone, television, guest meals, etc., which are posted to the patient charge ledger. Room and board charges.—Hospital charges to the patient, usually for room, food, and nursing plus any other miscellaneous item for which the hospital does not make a separate charge. Total charges.— Total of room and board, personal, and other charges which are posted to the patient charge ledger. Charges Covered For a ledger to be completed (paid), a payment covering each charge on the ledger must be received by the hospital. Usually, the amount of the payment equals the amount of the charge, but in several cases 27 this may not be true. For example, Blue Cross has contracts with hospitals in many States to pay less than the full charge. Commercial insurance companies, when paying for a Workmen's Compensation case, may pay less than the full charge and welfare often pays the hospital less than the full charge. In all these cases, the payment covers less than the full charge, so that the hospital must write off the difference between the charge and the payment received. The pilot study investigated the amount each source of payment is responsible for, not how much it actually pays. Therefore, the abstract form collected data on total amount covered, To avoid errors in abstracting the source-of-payment information in hospitals that post the amount of the payment, provi- sions were made on the abstract form for writeoffs. This alerts the abstracter to a possible reason for total charges not equaling total amount covered. It is not necessary to record the amount of the writeoff if the hospital would prefer just to add this amount to the amount covered and record this as a single figure. Charge Systems All-inclusive rate.— Patient charge system used by some hospitals whereby the patient is charged a flat daily rate no matter what supplies or services are provided by the hospital. All patients, no matter what the severity of their illness, are charged the same rate in these hospitals. Itemized-charge, according to supply or sevvice provided.— Patient charge system whereby the patient is charged for each supply or service provided. Each charge is separately itemized on the patient charge ledger. Files Active file.—File of patient charge ledgers for which all charges are posted but are still unpaid. Sometimes referred to as Accounts Receivable, Awaiting Payment, or Billing Files. Agency file.—File of patient ledgers determined uncollectable by the hospital and turned over to a collection agency for recovery. Sometimes referred to as the Bad Debt File. Completed file.—File of patient charge ledgers for the current fiscal year for which all charges have been posted and all payments received, leaving a balance of zero. Sometimes referred to as the Closed File, Inactive file.—File of completed patient ledgers (all charges posted and all payments made) more than one year old. (Seven years is the usual legal requirement for file retention.) Sometimes referred to as the Dead File. Open file.—File of patient charge ledgers for which all charges have not been posted. Sometimes referred to as the In-House or Posting File, Hospital Departments, Accounting department,—Hospital department responsible for preparing, updating, and maintaining patient charge ledgers which are the source documents for patient charge and source-of-payment ab- stracting, Medical records department.—Hospital department responsible for preparing, updating, and main- taining the hospital's medical records which are the source documents for patient medical data ab- stracting. Hospital Forms Bill.—Document sent by the hospital to the patient or other responsible party describing the charges and requesting payment for the amount shown on the bill. Also called an Invoice. 28 Patient charge ledger.— Accounting department record of all supplies and services provided, charges made, and payments received for all patients admitted to the hospital. Patient Charge A hospital charge to the patient for supplies or services provided. Patient Charge Abstracting Process by which patient charge and source-of-payment information on the patient charge ledger is recorded on a patient charge abstract. Payments Discounts.— Full or partial credit for a patient given by hospitals, usually to employees, physicians, etc. A discount is posted like a payment since full charges are posted to the ledger for patients receiv- ing discounts. Sometimes called an Allowance. Personal. — Payments made to the hospital by the patient. Third party,— Payments made by party other than the patient for supplies and services received in the hospital. Posting Process of recording patient charge and payment information to the patient's ledger. Information usually includes a description of the charge or party making the payment plus the amount of the charge or payment. Separate Billing Billing for a supply or service provided in the hospital but billed for by a physician, an agency, or auxiliary groups (e.g., anesthesiology, pathology, radiology, television). Survey Materials and Forms Hospital instruction manual, —Manual prepared for each individual hospital by the survey team dur- ing the initial visit to the hospital. It includes instructions for: Selecting sample for abstracting Identifying ledgers for abstracting Removing completed ledgers from file Preparing abstracts for completed ledgers Mailing abstracts to NCHS Interview schedule.— Document used to record information collected by direct interview of hospital personnel pertaining to characteristics of the hospital and its patient charge system. Patient charge abstract,—Form used for recording patient charge data for patients included in the study. Training manual.—Document used to train pilot study staff to conduct interviews and abstract patient charge data during the field test. 29 Survey Staff Personnel who were responsible for planning and conducting the pilot study of hospital patient charge statistics, 30 APPENDIX B CONTACT LETTERS USED WITH HOSPITALS INTRODUCTORY LETTER FROM NATIONAL CENTER FOR HEALTH STATISTICS TO HOSPITAL Dear Administrator: The National Center for Health Statistics of the U.S. Public Health Service is planning to under- take a national survey to collect hospital charge statistics for patients treated in short-stay hospitals. This project is a phase of the Hospital Discharge Survey, which is currently collect ing national hospital patient statistics. The objectives and design of the Hospital Discharge Survey are described in the enclosed report. The purpose of this letter is to request your cooperation in a pilot study on patient charges involving a small number of hospitals. Your hospital was recommended by of the areawide Planning Council as one whose information would make an important contribution toward developing sound conclusions in this project. Within the next several days, a representative of Booz, Allen, & Hamilton Inc., acting as an agent of the National Center for Health Statistics, will telephone you to arrange for an appointment to discuss the details of the project more fully. The National Health Survey program, of which the Hospital Discharge Survey is a part, has been approved by the House of Delegates of the American Medical Association, and the Hospital Dis- charge Survey itself has been endorsed by the Board of Trustees of the American Hospital Association. All information collected in this pilot study will be accorded strict confidential treatment as provided by Federal regulations and will be used for statistical purposes only. Without the cooperation of your hospital and others like yours, this important research project cannot succeed. I therefore urge you to participate in this project. Sincerely yours, ~~ Monroe G. Sirken, Ph.D. Chief, Division of Health Records Statistics 31 32 LETTER FROM STUDY STAFF TO HOSPITAL TO EXPRESS APPRECIATION FOR HOSPITAL COOPERATION IN PRETEST Dear Administrator: We would like to express our sincere thanks for your help in the pretest phase of our study for the Public Health Service. The information we received at is now being used to modify our questionnaire and abstract forms for the next phase of the study. Although we cannot promise that there will be a direct way in which we can repay you for your help, you can be assured that if we do come upon any ideas that we think would be beneficial to your hospital, we will certainly pass them along to you. Very truly yours, BOOZ, ALLEN & HAMILTON, Inc. LETTER FROM STUDY STAFF TO EXPRESS APPRECIATION FOR COOPERATION OF FIELD TEST AND FINAL TEST HOSPITALS Dear Mr. We deeply appreciate the time which you have extended to us and the U.S. Public Health Service in the important study we are conducting for the National Center for Health Statistics to de- velop a methodology for abstracting patient charges as a means of securing important national data for health program planning. The information we obtained in your hospital and the excellent results achieved by your people in our field survey were most useful and encouraging to us. Thank you again for your cooperation. Sincerely yours, BOOZ, ALLEN & HAMILTON, Inc. 000 33 34 APPENDIX C INTERVIEW SCHEDULE USED WITH SELECTED HOSPITAL PERSONNEL Inidal Mail Contact Telephone Contact Summary of Personal Interview with Administration Interviewer Form Approved Budget Bureau No. 68-R620.R2 INTERVIEW SCHEDULE FOR FINAL TEST HOSPITALS CONFIDENTIAL. 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 purposes of the survey and will not be dis- _ 001 closed or released to other persons or used for any other purpose (22FR 1687) Revised 10/27/66 Hospital Charge Pilot Study Name and Address (verify Booz, Allen & Hamilton on behalf of and correct, if necessary) National Center for Health Statistics U.S. Public Health Service Department of Health, Education, and Welfare Section A - Summary of Telephone Appointment and First Personal Visit Date Sent Received a. National Center for Health Statistics Letter yes no b. Booz, Allen & Hamilton Letter yes no a. Date of telephone contact b. Appointments made for personal visit a.m. Date Time p.m. Name Title Room and Building c. Comment and important items discussed with administrator on the telephone d. Other telephone contacts Name Title Name Title Section B a. Hospital Cooperation Unqualified cooperation Qualified cooperation (describe) Refused to cooperate (describe) Coordinator's Name: If additional space is needed for comments, use last page of questionnaire Name: Date form completed: INTERVIEW SCHEDULE WITH ADMINISTRATOR The U.S. Public Health Service is interested in collecting information from patient ledgers on charges and sources of payment, and in relating this data to the medical information now being collected by the medical records department for the Hospital Discharge Survey. The medical records department in your hospital is currently supplying the Public Health Service with data from the medical records for about 30 discharges per month. The additional information that the Public Health Service is interested in collecting is: Total Charges Room and Board Charges Other Charges Sources of Payment Dollar Amount of Responsibility by Source of Payment WB WN = This information would be obtained from the patient's ledger and would have to be for the same patient and hospitalization episode now being used by the medical records department for medical data abstracting. Booz, Allen & Hamilton Inc. has been conducting this study for the U.S. Public Health Service. The study is now in its final phase. To date, hospitals in Illinois, Michigan, California, Louisiana, and New Jersey have been studied. As a result of these studies, it has been determined that it is possible to collect this information from patient ledgers. Procedures and forms have been developed which need to be tested in this final study phase. The study will require a maximum of two days and probably less in each hospital. Each hospital will be visited during the week of November 7, 1966, to develop the procedures and again on the week of December 15, 1966, to see how the procedures were followed and to discuss any problems that may have occurred. It is to be emphasized that this study will not continue beyond the above mentioned dates and that all information obtained will be kept strictly confidential. Before talking to your business manager, I would like to ask you some questions about the survey and how a survey of this nature would be received by your hospital IF it were to be conducted on a continuing basis for a small sampling of discharges per month. As we mentioned earlier and as you can see from the abstract form, we are interested in obtaining information from patient ledgers on Total, Room and Board, and Other charges and sources of payment including their dollar amount of responsibility. i, Can you foresee any objections that hospital's management might have to releasing this information? If yes, explain 2. What would be your feelings about participating in this survey if the work were to be done on a continuing basis by your business office staff, with the Public Health Service compensating your hospital for the time spent as they are now doing with the medical records department? 35 36 Ask this question if the administrator feels that the major cause of concern would be availability of time in the business office. 3. Would you think it would be better if staff from the Public Health Service were to do the abstracting instead of your own hospital staff? Yes No 4. From your knowledge of your hospital and others, which factors, if any, do you think would cause the greatest concern about conducting a survey of this nature? INTERVIEW SCHEDULE WITH BUSINESS MANAGER The U. S. Public Health Service is interested in collecting information from patient ledgers on charges and sources of payment and in relating this data to the medical information now being collected by the medical records department for the Hospital Discharge Survey. Owing to many differences in hospital systems, records, and policies, it is necessary to obtain the following information before a procedure can be developed to abstract patient charge information in a hospital. Type of equipment used to post and record charge data . Type and amount of information available from a patient's ledger or equivalent Filing and coding practices used by your hospital for patient ledgers Time required to complete records and receive payments from discharged patients . Make-up of Total and Room and Board charges The purpose of my visit today is to obtain this information about your hospital and, with your help, to develop and test a patient charge abstracting procedure, and to teach one of your employees to use the developed procedure. I have a sample of the Hospital Discharge Survey listing and a sample abstract form. If you could obtain a sample of the hospital's discharge listing and a representative paid patient ledger, I believe we will have all the documents that are needed for this discussion. a, One of the major considerations for developing abstracting procedures is the type of equipment your hospital uses for posting charges and payments to a patient's ledger. Which of the following is used at your hospital? For Charges For Payments Computer Tabulating equipment Bookkeeping machines (NCR, etc.) Manual BW Ne Does your hospital have any plans to change this equipment in the near future? Yes No If yes, explain. If computer--use Computer Hospital Instruction Manual for developing the abstracting, procedures. 2; As we mentioned earlier, the patient charge information needs to be collected for the same patient and hospitalization episode that the medical records department is using for patient medical data abstracting. If the medical records department were to supply your office with the list of patients they have selected, (show sample list), would your office be able to locate the ledgers for the same patients and hospitalization episodes? Yes No 2a. Ifitem 2isno, what additional information would your office require? 2b. If item 2 is yes, skip to item 4. This list (show list) is prepared by the medical records department using the hospital's daily discharge listing. Each month, a sample of discharges is selected (usually through a terminal digit) to be used for medical daca abstracting. 3. Can the additional information required by your department (refer to question 2a.) be obtained from the daily discharge listing? Yes No If item 3 is yes, do you think the medical records department would supply this additional information to your office? Yes No If item 3 is no, does your office have access to the daily discharge listing or any other records which could be used to obtain this additional identifying information? Yes No 4. If patient charge abstracting were to be done on a regular basis in your hospital, which of the following methods would you suggest for coordinating patient charge and medical data abstracting? a. Medical records department will continue to prepare listing but will prepare one extra copy for the business office, which will add additional data as required. b. Business office will select and identify patient ledgers on its own, using the same procedures now being used by the medical records department for selecting the sample. C. Medical records department will prepare one extra copy and will add any additional information to the business office copy, as required. d. Other? Explain. The U. S. Public Health Service is also interested in the type of charge system used at each hospital since it will have a direct effect on the procedures used for abstracting. 5. Does your hospital charge any of its patients on an all-inclusive per diem (daily) rate or do you charge all your patients according to each service or supply received? All-inclusive According to service or supply 37 38 If all-inclusive, is it possible to break down this rate to determine the amount charged for Room and Board? Yes No If no, what figure is supplied to insurance companies that request such a breakdown? $ per day Now let us consider the information we are trying to obtain. As you can see from the abstract form, we are interested in obtaining Total Charges. Since the makeup of Total Charges varies among hospitals, it is necessary to determine what is included in Total Charges in your hospital. For the purposes of this study, Total Charges are defined as: All charges to the patient that are posted to the patient's ledger. I will read off a list of services or supplies that may or may not be included in your hospital's Total Charge figure. For each service or supply, could you tell me if the charge for each is: Billed Included Separately in Your and not Not "Total" Included in Rendered Hospital "Total" at This Not Charge Charge Hospital Charged For Professional Fees for 1. Pathology 2. Radiology 3. Anesthesiology 4. Surgery 5. Regular physician(s) visits Professional Fees for 6. Physical therapy 7. Other professional fees (explain) 8. Private duty nursing Services 9. Television 10. Telephone 11. Local telephone calls 12. Guest meals Supplies 13. Blood 14, Other Can you think of any other supplies or services which may be billed separately at your hospital? (Billed separately means that the charge is not posted to the patient's ledger.) Yes No If yes, list: Again, looking at the abstract form, you can see we are also interested in Room and Board Charges. We would like to see what is included under Room and Board at your hospital. Could you tell me if your hospital includes the following supplies and services under Room and Board or charges separately for them? Not Considered Included Separate in Either Food Routine diets Special diets Drugs Routine drugs (aspirin, etc.) Special drugs Nursing Routine nursing Private duty nursing Building and Building Equipment Maintenance of facility Depreciation Heat and light Air-conditioning Laundry Services Overhead Administration Supervision Special Equipment and Services Intensive care Intravenous feeding Recovery room Surgical dressings Oxygen setups Blood setups Telephone Television Guest meals Laboratory Services Urinalysis Routine blood counts Biopsies Basal metabolism 39 40 9. Are there any other supplies or services not mentioned on the preceding page that your hospital includes in Room and Board? If yes, explain. The next item to consider is the requirements for completing a patient charge and source of payment abstract. Based on our experience to date, it has been determined that patient charge and source of payment abstracting can only be done accurately from a paid or confirmed patient ledger. Assuming this to be true, and this survey were to be conducted on a continuing basis, we would be interested in determining whether a system could be worked out in your office so that the selected patient ledgers would only have to be searched for and located once. 10. Considering this as an ongoing survey, do you think the procedure to be described in the following paragraph, or one similar to it, could and should be used to conduct this survey at your hospital ? Obtain (or develop) a list of patients records to be abstracted. Locate the selected patient ledgers in the file. If the ledger is paid or confirmed, remove from file. If ledger is unpaid or unconfirmed, identify it for abstracting with a TAB or STAMP, etc. This would alert the posting clerk to give this ledger to the abstractor once it is paid or confirmed. Abstract from all paid or confirmed ledgers removed from file or forwarded by the posting clerk. Could this procedure be followed? Yes No Should this procedure be followed? Yes No If no, can you recommend any other procedure that could be used to eliminate the need for locating an unpaid account several times? Yes No If yes, explain. 11. Paid ledgers are usually necessary to obtain accurate source of payment information. We are interested in identifying the following sources of payment and the amount for which each source is responsible. Could you tell us, for each of the following sources of payment, if: (1) Source and amount can be identified from the paid ledger (2) Source and amount cannot be identified from the paid ledger Yes No Explain Can Be Cannot Be (Yes--Where and How) Identified Identified (No--Why) 1. Blue Cross and Blue Shield 2. Patient Responsibility Medical charges Personal charges (TV, phone, etc.) 12. 13. 14, Yes No Explain Can Be Cannot Be (Yes--Where and How) Identified Identified (No--Why) 3. Commercial Insurance 4, Social Security Administration (Title XVII Medicare) 5. Welfare (includes Title XIX Medicare) 6. Workmen's Compensation 7. Federal Public Service Programs (V. A. etc.) 8. Other Service Plans (Kaiser, CHA, etc.) 9. Probate Courts 10. Hospital Discounts and Allowances 11. Private Charity Union Welfare Fund [on tr 13. Other (explain) Since we are interested in abstracting from paid or confirmed ledgers, we would like to get an idea of how long your hospital requires to collect its bills so we can determine when abstracting can be done. For all your patients with third-party coverage discharged today, how many days would your hospital require to receive payment or confirmation from the third party on 90% of them? Days As you know, many patients who pay their own hospital bills are reimbursed by their commercial insurance. Have you any way of distinguishing those who are reimbursed from those who are not? 0] Yes [] No If yes, how and how accurately? Would it be possible to make copies of a completed ledger, and also copies of other hospital records, if any, which contain source of payment information that is not on the ledger? The name of the patient can be blanked out. [] Yes UJ] No This concludes the interview portion of the study. To complete our study, we would like to develop the abstracting procedure to be used at your hospital. To save time, an outline has already been developed. From this outline, the procedure can be prepared by crossing out any statements that do not relate to your hospital's system or records. The statements that remain or are filled in should be the abstracting procedure. We would then like to explain the abstracting procedure to the person selected by you to do the abstracting. A few abstracts would be completed to show him (her) what has to be done. 41 42 We would then leave this list (show list) with you. Within the next couple of weeks (before the week of December 15th), we would appreciate it if you would complete the abstracting for the patients on the list. During the week of December 15th, we will return to your hospital and check on the work and the problems encountered. To save time, it would be very helpful if the ledgers used for abstracting (or copies thereof) could be kept aside with the completed abstracts for our review so that they would not have to be located again when we return. I believe we are now ready to develop the abstracting procedure from the prepared outline. INTERVIEWER'S COMMENTS AND ANALYSIS 1. Availability of source documents and information a. Daily discharge listing Yes ___ No b. Patient ledgers ~~ Ys=s ___No Cx Sources of payment Yes No If no, why was information NOT available? 2. Problems encountered and other comments INTERVIEWER'S CHECK LIST All questions on questionnaire completed Abstracting procedure developed Samples of patient ledgers received Selection list completed with necessary identifying information Hospital employee instructed on abstracting Time summary explained Appointment made for follow-up visit APPENDIX D PATIENT CHARGE ABSTRACT FORM USED IN FINAL TEST HOSPITALS Form Approved Budget Bureau No. 68-R620.R2 CONFIDENTIAL. 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 purposes of the survey, and will not be disclosed or released to other persons or used for any other purpose (22 FR 1687). . PATIENT CHARGE ABSTRACT I. TO BE COMPLETED BY ACCOUNTING DEPARTMENT FROM MEDICAL RECORDS LISTING. Patient's Admission Number Medical Record Number Discharge Date Admission Date HDS Number II. TO BE COMPLETED BY ACCOUNTING DEPARTMENT Total Charges Total Amount Covered Room and Board Charges Other Charges (less Amount Personal Charges) Sources of Payment Covered Personal Charges [Blue Cross and Blue Shield Blue Cross and Blue Shield write-off [] patient Responsibility (Medical Charges) [1 Patient Responsibility (Personal Charges, e.g., TV, Telephone, etc.) O Commercial Insurance Social Security Administration (Title Xvi Medicare) [J Welfare (includes Title XIX Medicare) Welfare write-off O Workmen's Compensation [J Federal Public Service Program (Veteran's Administration, etc.) [C] Other Service Plans (Kaiser, CHA, etc.) [J Hospital Discounts [_] Private Charity UJ Other (explain) Total Sources of Payment (number of Boxes Checked) (Work space for abstractor's calculation and comments) Abstracting Date Abstractor's Initials 43 44 APPENDIX E HOSPITAL INSTRUCTION MANUAL USED IN FIELD TEST HOSPITALS Form Approved Budget Bureau No. 68-R620.R2 HOSPITAL INSTRUCTION MANUAL * for Abstracting Patient Charges and Source-of-Payment Data for the U.S. Public Health Service at (Name of Hospital) Manual Completed on rhe Manual presents alternative methods for handling each step in patient charge abstracting. All inappropriate al- ternatives for the individual hospital are to be crossed out on the Manual by the study staff member explaining the Manual to the hospital. The remaining method(s) for each step in abstracting is the one the hospital is to use for each step in abstracting. PATIENT (ACCOUNT) IDENTIFICATION PROCEDURE Patient charge and source of payment information will be abstracted for the same patient and hospitalization episode that the medical records department uses for abstracting patient medical data for the National Center for Health Statistics of the U.S. Public Health Service. 1. Obtain the list of accounts that have been selected for abstracting. 2. Remove ledgers (identified from medical records list) from the file and complete the following items in Section I of the abstract for each ledger. HDS number Patient number Discharge date Other 3. Abstract from a paid or confirmed ledger only.2 ABSTRACTING PROCEDURE FOR OBTAINING "TOTAL" CHARGES The first item to obtain is total charges (see instruction sheet for location of total charges on abstract). 4a, Total charges on a patient's ledger are for one hospitalization only—proceed to num- ber 4. 4b. Total charges on a patient's ledger may include a balance from a previous admission. If so, subtract any previous balances before completing item 4. Proceed to number 4. 4c. (Add any other procedure of choice for the hospital.) Sa. Enter total charges on abstract from the total shown on the ledger. (See 3b if not crossed off.) 5b. Compute total charges by scanning ledger for all payments (credits) and adding the total of these payments (credits) to the balance due, if any, (see 3b if not crossed off) and enter on the abstract. Sc. Compute total charges by adding all subtotals for supplies and services. (See 3b, if not crossed off.) Then enter on the abstract. 5d. (Add any other procedure of choice for the hospital.) “For the field test, return all abstracts for unpaid and unconfirmed accounts with a statement ‘‘account not paid” instead of completing Section II. Be sure to complete Section I of «/l abstracts 45 46 PROCEDURE FOR OBTAINING "ROOM AND BOARD" CHARGES The second item to obtain is room and board charges. This hospital has an "'all-inclusive-rate' system. 6a. Use $— per day as room and board rate. Then: Multiply figure initem Sa by the number of days the patient stayed in the hospital. Then: Enter result of item 5b under room and board charges on the abstract. This hospital charges according to each supply or service provided the patient. 6b. Enter total for room and board (as shown on ledger) on the abstract. 6c. Compute room and board by totaling all room and board postings in the room and board column of the ledger and enter total on abstract. 6d. Compute room and board by scanning entire ledger for all room and board postings (not in separate column) and total all these postings. Enter this total on abstract. 6e. (Add any other procedure of choice for the hospital.) PROCEDURE FOR OBTAINING "PERSONAL" CHARGES The third item to obtain is personal charges. Enter personal charges on abstract, as follows: 7a. Total all charges for telephone, television, and guest meals and enter total on abstract. and enter on abstract. 7b. Total all charges for codes and ’ ’ 7c. (Add any other procedure of choice for the hospital.) PROCEDURE FOR OBTAINING "OTHER" CHARGES’ The last charge item to obtain is other charges. Calculate other charges as follows: 8a. Total room and board and personal charges and subtract this sum from total charges. Enter result on abstract for other charges. 8b. (Add any other procedure of choice for the hospital.) Are necessary, use work space at bottom of abstract to make this calculation. PROCEDURE FOR ABSTRACTING SOURCE OF PAYMENT INFORMATION 9. Start by identifying the source of payment. Shown below is a list of the sources of pay- ment and how they can be identified from the patient's ledger. (See Definitions of Sources of Payment for what is included in each source of payment. *) Identified on Cannot identify S e Payment ouice of PF ledger by from ledger 1. Blue Cross and Blue Shield Blue Cross and Blue Shield writeoff 2. Patient responsibility - Medical charges - Personal charges (TV, phone, etc.) 3. Commercial insurance 4. Social Security Administration (Title XVIII Medicare) 5. Welfare (includes Title XIX Medicare) Welfare writeoff 6. Workmen's Compensation” 7. Federal Public Service Program (Veterans Administration, etc.) 8. Other service plans (Kaiser, CHA, etc.) 9. Hospital discounts and allowances 10. Private charity 11. Other (explain) 10. Check box (if one exists) on the abstract for each source of payment identified from the ledger. 11. Enter the amount paid by each source on the corresponding line to the right on the ab- stract (column labeled amount covered on abstract). 4 : : ; . The memberof the study team who is working with hospital personnel is to complete column as to how each source of payment can be identified from ledgers in the specific hospital. SWorkmen’s Compensation cases (#6) usually cannot be identified from the payment information on the ledger. For this reason, each ledger heading should be scanned before abstracting to determine if a Workmen’s Compensation case is being abstracted. 47 48 12. Total all amounts in amount covered column and enter this total on the total amount covered line on the abstract, 13. Compare total amount covered to total charges. If these totals are not equal, deter- mine error and correct. 14. Total all boxes checked and place total on total sources of payment line on abstract. PROCEDURE FOR COMPLETING ABSTRACT 15. Enter abstracting date and abstracter's initials at bottom of abstract. Place assigned hospital number on top right-hand corner of the abstract. 16. Place completed abstracts, with ledgers or copies of ledgers in folder. Section I com- pleted for unpaid and/or unconfirmed ledgers and Sections I and II completed for paid and/or confirmed ledgers. Hold this folder until the abstracts are picked up and the work discussed. DEFINITIONS OF SOURCES OF PAYMENT 1. Blue Cross and Blue Shield All charges paid for by any Blue Cross and/or Blue Shield plan except for Title XVIII Medicare payments made by a Blue Cross and/or Blue Shield organization acting as fiscal agent for the Social Security Administration, or for Workmen's Compensation Federal Govern- ment Program, or welfare cases when Blue Cross and/or Blue Shield provides the insurance coverage or acts as fiscal agent. It makes no difference whether Blue Cross or Blue Shield insurance premium is paid directly by the patient, by the patient and his employer, or entirely by the employer, by a union health and welfare plan, or otherwise. Blue Cross and Blue Shield plans vary in their formal name from region to region but any Blue Cross and/or Blue Shield plan which is a member of the Blue Cross Commission or Associations of Blue Shield Plans should be included. It is to be emphasized that in any case where Blue Cross and/or Blue Shield is acting as the fiscal representative for the Social Security Administration for Title XVIII, the charges are to be put under Source of Payment—Social Security Administration (Title XVIII Medicare). la. Blue Cross and Blue Shield writeoff.—All charges written off the Blue Cross and Blue Shield payment responsibility by the hospital based on the terms of the Blue Cross/Blue Shield contract in that region. Such discounts are not required by all plans. 2. Commercial insurance Any charges paid by commercial insurance companies (Continental Casualty, Aetna, Connecticut General, Liberty Mutual, etc.) should be included in this category, except for Workmen's Compensation cases which are to be put under Source of Payment— Workmen's Compensation. All commercial insurance payments are included here whether the premiums are paid for by the individual, by the individual and his employer, or by the employer alone, by a union health and welfare plan, or otherwise. 3. Social Security Administration Included in this category are all charges paid ultimately for Social Security (Medicare) beneficiaries by the Social Security Administration, including those payments made by Blue Cross and/or Blue Shield and commercial insurance companies acting as the fiscal agent for the Social Security Administration. 4, Welfare Included are charges paid for by welfare and/or health departments as part of public wel- fare programs for the aged, aid to families with dependent children, aid to the blind, aid to the permanently and totally disabled, medical assistance to the aged, independent State and local welfare programs and Title XIX Medicaid programs which have replaced any of the foregoing public assistance categories. This category includes any charges paid by Blue Cross and/or Blue Shield in those cases where welfare and/or health departments pay the Blue Cross and/or Blue Shield premiums, provide the funds to public assistance beneficiaries to pay such pre- miums, or where Blue Cross and/or Blue Shield act as the fiscal agent for the welfare and/or health department. This category includes the Title XIX Medicaid programs which exist in some States. 4a. Welfare writeoff.—All charges written off by the hospital due to the limits set by welfare within the region. S. Patient responsibilities Sa. Patient responsibilities—medical.—This includes charges paid directly by the patient or his representative for supplies and services received. 5b. Patient responsibility-—personal.—All charges paid by the patient for telephone, tele- vision, and guest meals. 6. Hospital discounts Includes charges for which the hospital directly assumes responsibility and requires no payment from the patient or his representatives. It includes discounts provided hospital em- ployees, members of the medical profession, members of churches or religious organizations working for the hospital, etc. 7. Federal Public Service programs Any program under which the Federal Government pays for medical services for a Govern- ment employee beneficiary such as the Veterans Administration, military dependents" medical care services, etc. 8. Private charity Any payments for medical services donated or given from any source when no direct re- sponsibility for payment exists, such as individuals paying bills for others, and voluntary health agencies, foundations, and church groups, who pay for medical services with no legal respon- sibility involved. This does not include discounts given by hospitals to their employees, which should be placed under source of payment—hospital discounts. 49 50 9. Workmen's Compensation These are payments made by any local, State, and Federal Workmen's Compensation pro- gram for injuries or illness related to the patient's occupation. It does include Workmen's Compensation medical payments made by commercial insurance, Blue Cross, or other third parties acting as the insuring agent for Workmen's Compensation coverage. 10. Other service programs Payments made by any third-party payment group other than Blue Cross, commercial insurance, welfare, or Government. This category includes primarily prepayment insurance plans such as Kaiser Foundation Health Plans, Community Health Associates, and Group Health Association. 000 APPENDIX F SUGGESTED PATIENT CHARGE ABSTRACT FORM FOR FUTURE USE Form Approved Budget Bureau No. 68-R620.R2 CONFIDENTIAL. 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 purposes of the survey and will not be disclosed or released to other persons or used for any other purpose (22 FR 1687). PATIENT CHARGE ABSTRACT I. TO BE COMPLETED BY THE ACCOUNTING DEPARTMENT FROM LIST OF PATIENTS SELECTED FOR ABSTRACTING Patient's Admission Number Patient's Admission Date Patient's Discharge Date Patient's Hospital Discharge Survey Number Patient's Medical Record Number (if other than Admission Number) Total Charges Total Amount Covered Room and Board Charges Amount Personal (Nonmedical) Covered Charges Sources of Payment Other Charges (Total Charges [] Blue Cross and Blue Shield less Room and Board Blue Cross and Blue Charges and Personal Shield write-off Charges) [] Commercial Insurance [] Welfare (includes Title XIX Medicare) Welfare write-off [] Social Security Administration (Title XVIII Medicare) d Patient Responsibility (Medical Charges) O Other Check Source of Payment Private Charity x. 2. Hospital Discounts 3. Federal Public Service Programs 4. Other Service Programs 5. Workmen's Compensation . Other Total Sources of Payment (Number of Boxes Checked) Comments and Work Space Abstracting Date Abstractor's Initials 51 APPENDIX G ALTERNATIVE METHODS FOR SELECTING AND RELATING PATIENT CHARGE LEDGERS AND MEDICAL RECORDS FOR ABSTRACTING IN A CONTINUING ABSTRACTING PROGRAM ALTERNATE 1: INITIAL SELECTION OF PATIENT SAMPLE BY ACCOUNTING DEPARTMENT WITH REFERRAL TO MEDICAL RECORDS DEPARTMENT Assumptions I. The hospital's admission list is an appropriate instrument for sample selection. 2. The accounting department would be willing to select the patient sample for patient charge and medical data abstracting from the hospital's admission list. Procedure 1. Accounting Department— ledger preparation clerk (1) Obtains instructions from National Center for Health Statistics for selecting the sam- ple. (2) Stamps or tabs each ledger selected for abstracting as it is prepared. (3) Stamps additional pages of ledgers selected for abstracting since each new page be- comes the face page. (4) Prepares in quadruplicate a list of patients selected for abstracting for use as follows: Two copies to medical records department (one for files and one to be sent to NCHS with medical records abstracts) Two copies for accounting department employee designated for abstracting (one for files and one to be sent to NCHS with patient charge abstracts) 2. Accounting Department—file clerk (1) Removes each ledger identified by the ledger preparation clerk by a tab or stamp be- fore filing in the closed file. (2) Gives completed ledgers to the accounting department employee designated for ab- stracting. 52 3. Accounting Department—employee designated for abstracting (1) (2) 3) 4) (5) (6) (7) (8) Receives two copies of list of patients selected for abstracting from the ledger prepa- ration clerk. Receives completed ledgers identified for abstracting from the accounting department file clerk. Completes abstract for each ledger received. Checks off name of patient in both copies of list of patients selected for abstracting as abstract is completed. Indicates on ledger that abstracting was completed and returns ledgers to file clerk for refiling. Sends monthly to NCHS completed abstracts for that month, abstracts for prior months clipped separately, list of patients selected for abstracting, and completed statement on changes in accounting system and participating personnel (transmittal sheet). Audits periodically the list of patients’ ledgers selected and not checked off (abstracted) and attempts to find absent ledgers. Files the completed list of patients for sample when all abstracts are completed and uses it annually to check NCHS reimbursement. Evaluation 1. Advantages (1 (2) 3) (4) Costs hospital less than other alternatives since ledger is identified by accounting de- partment at time of original preparation of ledger and constant searching of files is avoided. Requires NCHS to monitor sample selection of only one hospital department. Minimizes problems of matching patient charge and medical record abstract since the HDS abstract number is assigned by only one hospital department. Eliminates need for using patient's name to find patient charge ledgers for abstracting. 2. Disadvantages 0) (2) 3) Requires training accounting department ledger preparation clerk to select the sample from the hospital's admission list. Creates possibility of ledger tampering since ledger is identified before charges are posted. Necessitates use of hospital admission rather than discharge list which is usual basis for sample selection of hospital patient studies. 53 54 ALTERNATE 2: INITIAL SELECTION OF PATIENT SAMPLE BY MEDICAL RECORDS DEPARTMENT WITH REFERRAL TO ACCOUNTING DEPARTMENT Assumptions 1. The medical records librarian will consent to senda copy of the list of patients selected for abstracting, including the names of patients, to the accounting department. 2. The accounting department will be willing to cooperate with the medical records department in permitting the medical records department to prepare the list of patients whose ledgers will be abstracted. 3. The ledgers selected for abstracting are accessible after they are paid. Procedure 1. Medical records librarian (1) (2) Selects patients for medical record and patient charge abstracting from the hospital's daily discharge listing in accordance with NCHS instructions. Prepares in quadruplicate a list of patients selected for abstracting for use as follows: Two copies to accounting department file clerk with the patient's name since the name is necessary to find the ledgers in the accounting department (one for files and one to be sent to NCHS with patient charge abstracts) Two copies for medical records department employee designated for abstracting (one for files and one to be sent to NCHS with medical record abstracts) 2. Accounting Department—file clerk (1) (2) (3) 4) (S) (6) Receives two copies of the list of patients selected for abstracting with name of each patient from the medical records department. Places list from medical records in a follow-up file since most patient charge ledgers will not have been paid and therefore will not be ready for abstracting. Removes list of patients selected for abstracting from follow-up file after sufficient time has elapsed to allow most of the ledgers to be paid. Uses the list of patients selected for abstracting, after sufficient time has elapsed for the majority of ledgers to be completed: Locates all selected ledgers Removes completed ledgers Stamps or tabs incomplete ledgers Gives completed ledgers to the accounting department employee designated for ab- stracting with two copies of list of patients selected for abstracting. Forwards all stamped or tabbed ledgers to the employee designated for abstracting as they are closed out. 3. Accounting Department—employee designated for abstracting (1) (2) (3) (4) 4) (6) 7) Receives completed ledgers and both copies of list of patients selected for abstracting from accounting department file clerk. Completes abstract for each ledger received. Checks off names of patients as abstracted on both copies of list of patients selected for abstracting. Indicates on ledger that abstracting was completed and returns ledgers to file clerk for refiling, Sends monthly to NCHS completed abstracts for that month, abstracts for prior months clipped separately, list of patients selected for abstracting, and completed statement on changes in accounting system and participating personnel. Audits periodically the list of patient ledgers selected and not checked off (abstracted) and attempts to find absent ledgers. Files the completed list of patients selected for abstracting when all abstracts are completed and uses it annually to check NCHS reimbursement. Evaluation 1. Advantages (1) (2) (3) Maintains most consistency with present Hospital Discharge Survey since medical records department is already designated for selecting the patient sample for abstract- ing. Requires NCHS to monitor sample selection of only one‘hospital department. Minimizes problems of matching patient charge and medical record abstracts since the HDS number is assigned by only one hospital department. 2, Disadvantages (1) (2) Costs more than sample selection by accounting department since each ledger selected for abstracting by medical records department must be located and removed from files in the accounting department. Medical records department must write the patients’ names on the list of patients selected for abstracting before forwarding to the accounting department. 55 56 ALTERNATE 3: INDEPENDENT SAMPLE SELECTIONS BY ACCOUNTING AND MEDICAL RECORDS DEPARTMENTS Assumptions 1. Identical patient numbers will be used on the hospital's daily admission and discharge lists. (Required because accounting department will select sample from the admission list and medical record department from the discharge list.) Procedure I. Accounting Department—ledger preparation clerk D (2) (3) (4) Obtains instructions from National Center for Health Statistics for selecting the sam- ple. Stamps or tabs each ledger selected for abstracting as it is prepared. Stamps additional pages of ledgers selected for abstracting since each new page be- comes the face page. Prepares in duplicate a list of patients selected for abstracting for use by accounting department employee designated for abstracting, who retains one copy for files and one to be sent to NCHS with patient charge abstracts. 2. Accounting Department-—file clerk (1) (2) Removes each ledger identified by the ledger preparation clerk by a tab or stamp be- fore filing in the closed file. Gives completed ledgers to the accounting department employee designated for ab- stracting. 3. Accounting Department--employee designated for abstracting (1) (2) 3) (4) Receives two copies of list of patients selected for abstracting from the ledger prepa- ration clerk, Receives completed ledgers identified for abstracting from the accounting department file clerk. Completes abstract for each ledger received. Checks off name of patient in both copies of list of patients selected for abstracting as abstract is completed, Indicates on ledger that abstracting was completed and returns ledgers to file clerk for refiling. (6) Sends monthly to NCHS completed abstracts for that month, abstracts for prior months clipped separately, list of patients selected for abstracting, and completed statements on changes in accounting system and participating personnel (transmittal sheet). (7) Audits periodically list of patients' ledgers selected and not checked off (abstracted) and attempts to find absent ledgers. (8) Files the completed list of patients for-sample when all abstracts are completed and uses it annually to check NCHS reimbursement. Evaluation 1. Advantages (1) Costs only slightly more than sample selection by accounting department alone (only additional cost is preparation of separate listing of patients selected for abstracting by the medical records department). Low cost because ledger is identified by accounting department at time of original preparation of ledger and constant searching of files is avoided. (2) Requires no contact between medical records and accounting department in hospitals. (3) Eliminates need for using patient's name to find patient charge ledgers for abstracting. 2. Disadvantages (1) Requires NCHS to monitor sample selection of two departments. (2) Increases problems of matching patient charge and medical record abstracts because common HDS numbers will not exist. ALTERNATE 4: REFERRAL BY NCHS TO HOSPITAL ACCOUNTING DEPARTMENT AFTER RECEIPT OF HOSPITAL MEDICAL RECORDS ABSTRACTS Assumptions 1. NCHS will find it advantageous to designate the patients for patient charge abstracting by using the same sample listing prepared by the medical records department for medical data abstracting. The accounting department will be willing to cooperate with NCHS, permitting NCHS to designate the patients for patient charge abstracting. The ledgers selected for patient charge abstracting are accessible after they are paid. 57 Procedure 1. Medical records department (1) Completes medical records abstracts and forwards abstracts with list of patients selected for abstracting to NCHS. 2. NCHS (1) Receives medical records abstracts and list of patients selected for abstracting from the hospital's medical records department. (2) Uses list of patients selected for medical record abstracting to designate patients for patient charge abstracting. (3) Prepares list of patients designated for patient charge abstracting in triplicate for use as follows: Two copies to accounting department file clerk One copy for NCHS file 3. Accounting Department—file clerk (1) Receives two copies of list of patients for patient charge abstracting from NCHS, (2) Obtains patient names from medical records department in cross-reference file and writes name on list. (Some hospitals will require names of patients to locate ledgers.) (3) Using the list of selected patients: Locates all patient ledgers. Removes completed ledgers and forwards to accounting department employee designated for abstracting. Stamps or tabs all selected ledgers that are not complete. (4) Gives both copies of list to employee designated for abstracting. (5) Forwards all stamped or tabbed ledgers to employee designated for abstracting as they are completed. 4. Accounting Department—employee designated for abstracting (1) Receives completed ledgers and both copies of list of patients selected for abstracting. (2) Completes abstract for each ledger received. (3) Checks off name of patient on both copies of list of patients selected for abstracting as abstract is completed. (4) Indicates on ledger that abstracting was completed and returns ledgers to file clerk for refiling. 58 (5) Sends monthly to NCHS completed abstracts for that month, abstracts for prior months clipped separately, list of patients selected for abstracting, and completed statement on changes in accounting system and participating personnel. (6) Audits periodically the list of patients’ ledgers selected and not checked off (abstracted) and attempts to find absent ledgers. (7) Files the completed list of patients for sample when all abstracts are completed and uses it annually to check NCHS reimbursement. Evaluation 1. Advantages (1) Maintains most consistency with present Hospital Discharge Survey since medical records department is already designated for selecting the patient sample for abstract- ing. (2) Requires NCHS to monitor sample selection of only one hospital department. (3) Minimizes problems of matching patient charge and medical record abstracts since the HDS number is assigned by only one hospital department. 2. Disadvantages (1) Costs more than sample selection by accounting department since each ledger selected for abstracting by medical records department must be located and removed from files in the accounting department. (2) Medical records department must write the patients’ names on the list of patients selected for abstracting before forwarding to the accounting department. O O 59 Series 1. Series 2. Series 3. Series 4. Series 10. Series 11. Series 12. Series 13. Series 20. Series 21. Series 22. OUTLINE OF REPORT SERIES FOR VITAL AND HEALTH STATISTICS Public Health Service Publication No. 1000 Programs and collection procedures.—Reports which describe the general programs of the National Center for Health Statistics and its offices and divisions, data collection methods used, 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, 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 Intevview Survey.— Statistics on illness, accidental injuries, disability, use of hospital, medical, dental, and other services, and other health-related topics, based on data collected in a continuing national household interview survey. Data from the Health Examination Survey.—Data from direct examination, testing, and measure- ment of national samples of the 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 Institutional Population Surveys.— Statistics relating to the health characteristics of persons in institutions, and on medical, nursing, and personal care received, based on national samples of establishments providing these services and samples of the residents or patients. Data from the Hospital Discharge Survey.— Statistics relating to discharged patients in short-stay hospitals, based on a sample of patient records in a national sample of hospitals. Data on mortality.—Various statistics on mortality other than as included in annual or monthly reports—special analyses by cause of death, age, and other demographic variables, also geographic and time series analyses. Data on natality, marriage, and divorce. — Various statistics on natality, marriage, and divorce other than as included in annual or monthly reports—special analyses by demographic variables, also geographic and time series analyses, studies of fertility. Data from the National Natality and Mortality Surveys. — Statistics on characteristics of births and deaths not available from the vital records, based on sample surveys stemming from these records, including such topics as mortality by socioeconomic class, medical experience in he last year of life, characteristics of pregnancy, etc. For a list of titles of reports published in these series, write to: Office of Information National Center for Health Statistics U.S. Public Health Service Washington, D.C. 20201 NATIONAL PSA RE; mE 2 For HEALTH Number 29° yr Nae) Bw oe ae of Age on the Death Certificate and Matching Census Record United States-May- August 1960 , U.S. DEPARTMENT OF tin arse EDUCATION, AND WELFARE Public Health Service Public Health Service Publication No. 1000-Series 2-No. 29 For sale by the Superintendent of Documents, U.S. Government Printing Office Washington, D.C., 20402 - Price 45 cents NATIONAL CENTER| Series 2 For HEALTH STATISTICS | Number 29 VITALand HEALTH STATISTICS DATA EVALUATION AND METHODS RESEARCH comparability of Age on the Death Certificate and Matching Census Record United States-May-August 1960 Comparisons of age as stated on the death certificate with age as stated on the matching census record by color, sex, geographic region, and specified causes of death. Based on a sample of death certificates for deaths occurring in the United States during May-August 1960 matched to 1960 cen- sus records. Washington, D.C. June 1968 U.S. DEPARTMENT OF HEALTH, EDUCATION, AND WELFARE Public Health Service Wilbur J. Cohen William H. Stewart Secretary Surgeon General NATIONAL CENTER FOR HEALTH STATISTICS THEODORE D. WOOLSEY, Director PHILIP S. LAWRENCE, Sc.D., Associate Director OSWALD K. SAGEN, PH.D.,, Assistant Director for Health Statistics Development WALT R. SIMMONS, M.A., Assistant Director for Research and Scientific Development ALICE M. WATERHOUSE, M.D., Medical Consultant JAMES E. KELLY, D.D.S., Dental Advisor LOUIS R. STOLCIS, M.A., Executive Officer MARGERY R. CUNNINGHAM, Information Officer OFFICE OF HEALTH STATISTICS ANALYSIS IWAO M. MORIYAMA, Ph.D., Director DEAN E. KRUEGER, Deputy Director Public Health Service Publication No. 1000-Series 2-No. 29 Library of Congress Catalog Card Number 68-60057 PREFACE This report is one of a series of comparability studies on selected items from the death certificate and the matching census record for a sample of deaths which occurred inthe United States during May-August 1960. The data are a byproduct of the study, Social and Economic Dif- ferentials in Mortality, United States, 1960, being carried out by Evelyn M. Kitagawa and Philip M. Hauser at the Population Research and Training Center, University of Chicago, in cooperation with the National Center for Health Statistics and the Bureau of the Census supported by PHS Grant RG-7134 (later changed to CH-00074). The items for which comparability is being studied are residence, age, marital status, race, nativity, and country of origin. As a latecomer to this project, the author gratefully acknowledges the tremendous assistance provided by Lillian Guralnick, now of the Office of Research and Statistics in the Social Security Administration, and Evelyn M. Kitagawa who were both involved in this study from its inception. SYMBOLS Data not available--====e-mcmcmemm ee ——— Category not applicable---=--ueooaoaaao Quantity zZero==-=-eecm comme - Quantity more than 0 but less than 0,05----- 0.0 Figure does not meet standards of reliability or precision-------ccceeoeooo "CONTENTS Page PLOLACE worm 0 sm mm emo 0 0000 0 0 00 0 0 iii IILEOQUCTION == = = mm mo mm mm 0 mm 0 0 0 00 dr 2 2 1 Accuracy of Age Information-------=======-=-=--===-=-=-o-osomoomoos Measuring Age Comparability----=-=-=========-====o-==-osooosomooomooos 2 Age Comparability Results---=-======-===---om=om-soosommsoossoosoooos 3 Sex, Color, and Age----=-==-===============--=-==sso-sossoosossosooss 4 Geographic Region----===============-oo-ommoomsooooooooooommsooooos 9 Cause of Death------mmm-=mmmmm-momommmmmm—mm—mmomsom—msee——o moms 12 Results of Related Studies-----==-==-===--mm==c-=---=-=--o-oooooomossoos 14 Evaluation of Accuracy of Age-Specific Death Rate------=-==-----=-====-=== 18 DISCUSSION ===-==-m=-=m==-=-=m-m--————e ee —e—— sm ——e——eeees= esse 25 SUMMATY=-======m=mmm========-==m=m=o= oso =s=ososossososSssosoooss 26 EEE —————— EE EEE EE inininieieieitieiaiaiea 27 Bibliography =---=--===========s===s=ooso-osomsoossoosoosooosoossonoos 27 Detailed Tableg=rrmrmmm====mmmmem=———=——-——————— === meee 29 Technical AppendiX----===========-=m=-=-=-=o-—=m=msososoooomsososoooos 45 Design of the Study----=-==-==-====-=====--====-=----osoososoosoooooo 45 Definitions--====m======-mmmmm==-m===-ee-m-===——=—=-esss=-=-s somos 45 Unmatched DeathS-e-==-==-cemm=me===mm—-——————————————————— == 45 Sources and Limitations of Data----==-==-====m=-==-om-o====-oo-==-=s 47 Alternative Method for Adjusting Death Rates Using Comparison of Response Results----------mmommommmomoomoomomoommomms moos m mmm mm os 49 vi THIS REPORT discusses differences between age as stated on the death certificate and age as stated on the matching census record and the effect of these differences on the published age-specific death rates. The dala presented inthis report are based ona sample of deaths which occurred in the United States during May-August 1960, Analysis of the data by means of two measures (net difference rates and percent agreement) involved examining age comparability by a number of characteristics of the decedent: age, sex, color, geographic region, and cause of death. Age comparability was relatively higher for white decedenls than for nonwhite, especially for older nonwhite decedents, and slightly higher for male decedents than for female. There were some regional differences in levels of comparability: they were lower in the Northeast Region for white decedents and lower in the South Jor nonwhite decedents than in the other vegions. Age comparability was higher among persons dying from accidents, poisonings, and violence than among persons dying from major cavdiovasculav-venal diseases. The results of this study indicate that if the census record age is used as both the numerator and denominator of the age-specific death rate, the observed excess of mortality for the older white population com- pared with the older nonwhite population, if not reversed, is consider- ably diminished. COMPARABILITY OF AGE ON THE DEATH CERTIFICATE AND MATCHING CENSUS RECORD Thea Zelman Hambright, Office of Health Statistics Analysis INTRODUCTION The major objectives of this report are (1) to evaluate the comparability of age as stated on the death certificate and as stated on the census record for the same individual, and (2) to evalu- ate the accuracy of the age-specific death rates in light of the age comparability results. In the past most of the measures of health status have been provided by mortality statistics. The census record and the death certificate are the basic sources of data for these statistics, and the adequacy of these records in terms of com- pleteness of coverage and accuracy of information is therefore important. As far as death certificate coverage is con- cerned, "although there has never been areliable evaluation made of death registration complete- ness, it has always been assumed that virtually all deaths are reported in the United States. It is unlikely that more than 1 or 2 percent of deaths go unrecorded.’ On the other hand, considerable effort has been expended to determine the com- pleteness of the census enumeration. ?” Depend- ing on the method used, undercoverage rates were estimated to be from 1.9 to 2.3 percent of the enumerated population in 1960.%° However, these figures represent coverage of the total population, and various portions of the total may ‘seriously differ in their coverage rates. For example, the estimated number of individuals thought to have been missed was much higher for nonwhite individuals and for certain age groups than for the population as a whole. An analysis of coverage errors for any subgroups of the popu- lation—such as age, sex, and color groups— necessarily involved consideration of content errors; that is, the number of persons reported in a given age, sex, or color group was affected not only by the degree to which they were missed but also by errors in age, sex, or color reporting and tabulation.’ Insofar as this report deals with the evalua- tion of age statements, the following discussion of the accuracy of the death certificate and census record relates primarily to considerations of accuracy of age statements. The accuracy of age information on both records has been questioned repeatedly, particu- larly with regard to the subject of mortality among the nonwhite population. In contrast with the over- all picture of large excesses in death rates for the nonwhite population compared with the white, death rates for the white group exceeded those for the nonwhite at ages 75 years and over. This reversal at older ages has existed as far back as 1900 and still existed as of 1965toan even greater degree. It was suggested on the basis of 1963 age-specific death rates that the errors in age reported on the census record for the nonwhite population do not correspond to the errors on the death certificate. When a set of ''corrected Negro populations’ developed by Bogue’ was used, the difference noted at older ages was not eliminated. Accuracy of Age Information Is the reported age on either the census record or the death certificate inaccurate? If so, is the age statement on the census record more nearly accurate than that on the death certificate or vice versa? From a two-way comparison of response, such a judgment is difficult unless some a priori reason existed for believing one record to be superior to the other. One acceptable criterion for superiority would be the extent to which one of the two records was more in agreement than the other record with age as derived from the matching birth certificate. However, the birth certificate is sometimes impossible to obtain, especially for older people, and this kind of check was not done. Another criterion would be how closely both of the records compared with a third, independent source. Although such information was available for the 1960 census on a nationwide basis (the CPS-Census Match?), nothing compa- rable was done for the death certificate. On the other hand, while it would be important to know which record was more nearly accurate, the results would not permit correction of the death rates. That is, beyond the purpose of com- paring information on the two records is the more pragmatic purpose of evaluating the accuracy of the death rates on the basis of this comparison. Even if it were known that the age on the death certificate is correct, it cannot be used to adjust the age of the total population which contains a large number of living persons of which only a small proportion die in a short period of time. However, the reverse is possible, and, in fact, is the procedure used here. The evaluation involves adjusting the death certificate information in the numerator of the death rate to the information contained in the census files for those decedents, thereby creating a rate in which the numerator and denominator of the death rate are based on census age reports, Thus no statement is made or implied about which record is the more accurate but, rather, the differences between them and the effect of these differences in moving from the actual age on the death certificate to one which is consistent with what was reported on the census record are the subjects of this report. MEASURING AGE COMPARABILITY Correspondence or lack of correspondence between recorded ages on the census record and the death certificate represents the combined effect of the differences in circumstances between the two records surrounding the collection, re- sponse, and processing of data. For 1960 the 100- percent census enumeration (stage I) forms were generally filled out at home, where the subject, if riot supplying the information himself, could be consulted. Age information was recorded as date of birth; returned forms were processed mechani- cally, and items not completed were assigned responses (1.7 percent of the 1960 census records had age estimated). Personal items on the death certificate were usually filled out by a funeral director to whom information was generally provided by the next of kin; age was recorded in completed years as of the last birthday; completed forms were manually coded and punched; and items not filled out were left incomplete (.04 percent of 1960 death certifi- cates had no age stated). The data used here are a byproduct of the study, Social and Economic Differentials in Mor- tality, United States, 1960, in which a sample of all deaths that occurred in the United States dur- ing the 4 months of May-August 1960 was selected and manually searched in the 1960 census for matching records. Almost 80 percent of these decedents were found in the 100-percent census enumeration (stage I). For the purposes of this study, two important changes were made in the usual census proce- dures. First, the census data used were unedited, manually coded responses with no assignments made for nonresponse. Second, the date of birth was converted to age at death, and a correction was made for those records in which a birthday occurred between the date of the census and the date of death. Thus the unmeasured error pro- duced by coding and card punching may be ex- pected to be about the same in both sets of records. The adjustment of age on the census record to age at date of death simplified the comparison of age information. 4 Two measures are used throughout this re- port to evaluate age comparability. The net dif- ference rate measures the difference in the num- ber of individuals in an age group between the death certificate classification and the census record classification relative to the number Yeported on the census record: d-c c; i x 100, where d;=Number of decedents inthe study group classified as age i on thedeathcertifi- cate. c;=Number of decedents in the study group classified as age i on the census record. The net difference rate here is algebraically equivalent to the "Index of Net Shift Relative to CPS Class" used by the U.S. Bureau of the Census in evaluating the 1960 census data in light of the results of the Current Population Survey. * Theo- retically, the range for this measure is from -100.0 percent to positive infinity. It is conceiv- able, but highly unlikely, that there would be no individuals in an age group. However, where age is analyzed by detailed characteristics or where characteristics other than age are analyzed, it is possible for the rate to become infinity. In this study no age category on one record contained more than twice the number of individuals than the other record, and thus, empirically, the rate never approached an absolute value of 100.0 percent. The comparison is made with respect to the marginal totals of the data according to each record, as shown below. Death certificate Census record NUDE Nusbey Total in age age group i group i Total----- d, Number in age group i w=e=r- c; s; c,—s p Number not in age group: -- d,—s; A net difference rate of zero means that both records had the same number of individuals in that age group. However, no indication is givenas to whether the same individuals are in that age group on both records. This measure may be used to evaluate the accuracy of age-specific death rates with the condition that the data to be evaluated have the same age distribution as the data from which the net difference rate was derived. Under these cir- cumstances, the net difference rate indicates the direction and the size of the difference between the actual age-specific death rate where numer- ator and denominator come from different sources and a rate that would result if both numerator and denominator came from the census record. For example, a negative net difference rate of -5.0 percent suggests that the actual age-specific death rate is too low. The number of deaths at that age on the death certificate is less than the number of deaths at that age according to the census records for those decedents. An age- specific death rate computed from age as reported in census records would be 5.0 percent greater than the actual age-specific death rate. The. second measure, percent agreement, is a more direct measure of the correspondence between records. It indicates the extent to which the same individuals are classified in the same age group on both records: * x 100, where s,— Number of decedents in the study group classified by both the census record and the death certificate as age 1 c;= Number of decedents in the study group classified as age i on the census record. In this case since only the agreements are considered (see the above diagram), there is no opportunity for differences between records to cancel each other as is possible in the net dif- ference rate. The numerical value of one meas- ure does not determine the numerical value of the other except when the value of the net differ - ence rate is -100.0 percent; then the percent agreement must be zero. The tables in the text of the report contain summary figures of percent agreement which were obtained by cumulating the number of rec- ords identically classified in the specified age intervals over a number of age categories and dividing by the total number of census records in these age intervals. For example, percent agreement for 5-year age intervals for decedents 45-64 years of age would be calculated as follows: i=60—64 Xs, i=45— 49 x 100. i=60— 64 Se i=45— 49 Comparisons were made between death cer- tificate and matching stage I unedited census rec- ord information for inflated sample data, (See Technical Appendix.) AGE COMPARABILITY RESULTS Sex, Color, and Age Percent agreement, —Slightly more than two- thirds, 69 percent, of the total study group had the same single year of age on both the census record and the death certificate. Considerable variation around this overall figure existed for the component sex and color groups. Substantially less agreement was found for the nonwhite group for all ages (1-99 years) than for the white. Agree- ment levels were somewhat lower for females than for males although differences between males and females were not so pronounced as those be- tween the white and nonwhite groups (table A). Three-fourths of the white males had the same single year of age reported on both records as compared with less than one-half of the nonwhite males; two-thirds of the white females had the same single year of age on both records as com- pared with a little more than one-third of the non- white females. It should be established at this point that the designation of an individual as white or nonwhite is in accordance with the classification used by the Bureau of the Census. Although there were some differences in color assignments on the census record and the death certificate, they were very small: 0.2 percent of the decedents reported as white on the census record and 2.3 percent of 4 those reported as nonwhite were reported differ- ently on the death certificate (preliminary esti- mates from another report in this series). Results on age comparability would not be affected by color discrepancies; for example, if there were lower age agreement among the 0.2 percent of decedents reported as nonwhite on the death cer- tificate but white on the census record, it is un- likely that it would contribute to lowering the age agreement level of the total white group. In addition to large differences in the amount of agreement in same single year of age between the color groups, there were at least three major differences in the patterns of disagreement. For the white group, agreement remained fairly con- stant with increasing age. For the nonwhite group, however, there was less likely to be agreement between the two records as age increased. While percent agreement was at the same level for both nonwhite and white decedents at age 1, itdeclined rapidly thereafter for nonwhite decedents and, consequently, the difference between the color groups was greater at the older ages (fig. 1). Second, age for the white group as reportedon the death certificate was usually within 1 year of that reported on the census record when the two rec- ords did not agree. The difference was usually greater by more than 1 year for the nonwhite group, however, particularly for decedents aged 45 years and over. For example, in this older age group the age on the death certificate was within 1 year of that on the census record for 91 percent of white males but only for 61 percent of nonwhite males. And finally, where age statements between records did not agree, the age reported on the death certificate was more often older than the age given on the census record for white dece- dents of all ages as well as for nonwhite decedents under age 45. In contrast, the age given on the death certificate was more often younger than the age on the census record for nonwhite decedents over age 45. Data for single years of age are subject to errors which may contribute to differences in correspondence between records. One such error is age heaping, or the tendency for reported ages to be concentrated at particular ages or at groups of ages ending in the same digit. Distributions of deaths by single years of age as reported on the census record and onthe death certificate indicate that heaping is found in varying degrees in the four sex-color groups and that the extent and the Table A. Percent distribution of matched census records, by age agreement with death certificates according to color, sex, and age: United States, May-August 1960 White Nonwhite Death certificate age Tosal (relative to census age) Sa y group Male | Female | Male | Female Percent distribution All ages, 1-99 yearS==wmwmmmmmmmnn nun 100.0 100.0 | 100.0 | 100.0 | 100.0 Same single year of age=e==wewnussm~e=cewansn 68.8 74.5 67.9 44,7 36.9 1 FEAT JOUIZEE ww i 6 sei mm mr 3m 2 50 8.1 7.5 8.3 10.5 10.7 1 YORE OLAGT wp «wmmimmmm swim mmm mmm oo thin tw 2 9.7 9.1 10.5 10.3 9.5 Younger or older by more than 1 year---------= 13.4 8.9 13.3 34.5 42,9 1-44 years=----======---====------sssoss 100.0 100.0 | 100.0 | 100,0 | 100.0 Same single year of age----=-=-==-=-------==---- 74.8 79.0 76.8 61.3 57.4 1 year younger==---==-==============-=sss-sooses 6.7 5.6 6.5 9.5 10.6 1 Year Dlderei=weecrescunwnmmmnmm emmy 9.0 8.4 8.5 11.0 11.3 Younger or older by more than 1 year------=-== 9.5 7.0 8.2 18.2 20.2 45-99 years-----==mmm====-ceme-==--e-=- 100.0 100.0 | 100.0 | 100.0 | 100.0 Same single year of age--=-=-------==----=---°" 68.1 74.0 67.1 40.5 31.9 1 year younger----=-==-===-m==========s===o--on 8.2 7.7 8.5 10.8 10.7 1 Year oldererwennmm~ecnmsnmmnmennnnnmss se, 9.8 0.2 10.7 10,1 9.0 Younger or older by more than 1 year=--=---== 13.9 9.1 13,7 38.6 48.4 particular ages at which heaping occurs differ from one record to the other (figs. 2 and 3). Generally speaking, although heaping existed on both records, it was less pronounced on the census record distribution than on the death cer- tificate distribution. This can be partially ex- plained by the fact that date of birth was used for the age item on the census record as opposed to completed years as of last birthday for this item on the death certificate. Another reason may be that when the informant on the death certificate was uncertain about the deceased's exact age, he would tend to report the age in round numbers. This tendency shows up as a preference for ages ending in 0 and 5 and is seen in both distributions although to a more marked degree in the death certificate distribution. In addition there appears to be specific preference for ages 59 and 60 on both records for all sex and color groups, which may reflect a preference for 1900 as the year of birth. On the whole, heaping is greater for fe- males than males and greater for nonwhite in- dividuals than for white. Percent agreement figures for single years of age discussed above are affected by the extent to which heaping occurs on one record more than the other. While age heaping errors are part of the total difference in age information between records, they can be minimized by combining ages into intervals. Since ages are usually tabu- lated by 5- or 10-year intervals for most vital statistics purposes, the effects of age heaping and the lack of agreement in single years of age found here would probably not seriously distort the patterns in broader age groups. When ages were combined into 5-year inter- vals, 86 percent of the total study group were classified in the same age groups on both records. A further but smaller improvement to 90 percent agreement occurred when ages were combined into 10-year intervals. This improvement in agreement occurred for each sex and color group 00 WV TTTITTTIT II TIT TTI TT Try II TT TT777 A TTR TTTTTTITT White male 80 - & s 60 Ww w ox © 2 L —- < S a0l- ox w a 20 |- LLL bn nde et bed erred wucdiauediendiwdidius 0 10 20 30 40 50 60 70 80 90 99 SINGLE YEARS OF AGE ON CENSUS RECORD 100 THY TY IT ATT TY RIT FYVITTIT] i hl ll a a E Ww = w Ww x oO < —- Zz w o ox w a EL taba mn dae biti Lh oh Las | 0 10 20 30 40 50 60 70 80 90 99 SINGLE YEARS OF AGE ON CENSUS RECORD Figure I. Percent agreement in age between the death certificate and the matching census record, by color and sex. 6,500 6,000 5,000 4,000 3,000 NUMBER OF DEATHS 2,000 1,000 6,500 6,000 5,000 4,000 3,000 NUMBER OF DEATHS 2,000 1,000 White male Nonwhite male 20 40 60 80 99 | 20 40 60 80 DEATH CERTIFICATE SINGLE YEARS OF AGE CENSUS RECORD SINGLE YEARS OF AGE 6,500 6,000 5,000 4,000 3,000 2,000 1,000 6,500 6,000 5,000 4,000 3,000 2,000 1,000 29 Figure 2. Distributions of deaths by age as stated on the death ce census record for male decedents, by color. riificate and as stated on the matching 6,500 6,500 6,000 White female 6,000 5,000 5,000 z 4,000 4,000 << Ww o uw Oo x 3,000 3,000 w m = - 2 2,000 2,000 1,000 1,000 0 0 6,500 I 6,500 60 6,000 | Nonwhite female —16,000 5,000 (— — —15,000 60 w E 4000 — fre 80 —1{4,000 < a8 Ee 80 © 3,000 S 5 —{3,000 w m 5 40 2 40 2,000 (— fe — 2,000 fi I — =] sr 000 g 2 20 1,000 ol} 0 | 20 40 60 80 99 | 20 40 60 80 99 DEATH CERTIFICATE SINGLE YEARS OF AGE CENSUS RECORD SINGLE YEARS OF AGE Figure 3. Distributions of deaths by age as stated on the death certificate and as stated on census record for female decedents, by color. the matching along with some narrowing of the gap in color differences (table B). The relationship between age and degree of agreement noted for single years of age was also seen in the 10-year age intervals. While the agreement between records was relatively high and fairly stable from one age interval to another for the white group, for the nonwhite, agreement declined with increasing age. For example, while about 95 percent of the nonwhite group in the census age interval 1-4 years were similarly reported on the death certificate, only 61 percent of those in the census age interval 85-99 years were reported in the same age group onthe death certificate. Net difference rate.—In addition to the fact that reported ages were often different between records for the nonwhite group, particularly at ages 45 and over, age as reported on the death certificate was frequently younger than that re- ported on the census record for the same individ- ual at ages over 45. This finding is implied by the pattern of the net difference rates (fig. 4 and table 4), where the death certificate showed consider - ably more decedents aged 45-64 years than the census record and considerably fewer aged 75-99. The number of nonwhite decedents aged 65-74 years was almost the same on both records. This is because there were more decedents aged 65- 69 years but fewer at ages 70-74 years on the death certificate than on the census record. Thus the differences between records for these two 5-year age intervals were cancelled when the ages were grouped into the 10-year interval 65- 74. (See table 1 for data on 5-year age intervals.) The implications of these findings are clear insofar as these data apply to all nonwhite dece- dents for 1960: the number of deaths for nonwhite individuals aged 75-99 years is substantially understated relative to the number of nonwhite decedents reported in this age range onthe census record. Consequently, the published age-specific death rates for this group are lower than rates based on census ages. The pattern for the com- bined age group 65-99 years is the same as that for ages 75-99 years although the understatement on the death certificate relative to the census record (and hence, the net difference rate) is not as great. In a later section the effect of these Table B. Percent agreement in age between the death certificate and the matching census record, by color, sex, and speci- fied age intervals: United States, May- August 1960 5-year |10-year Single age age years inter- |inter- vals vals Color and sex Percent agreement Total study group-----=-- 68.8 85.9 90.3 White male----- 74.5 89.9 93.0 White female--- 67.9 85.8 90.5 Nonwhite male-- 44,7 68.3 77.1 Nonwhite female-------- 36.9 60.4 71.5 findings on the age-specific death rates is further investigated. No clear pattern was observed for the white group in the net difference rates which oscillated around 0.0 for most of the age intervals. Thus results in age comparability were rather dif- ferent for the two color groups. The important differences were (1) the rapid decline in agree- ment with increasing age for nonwhite decedents in contrast with fairly constant agreement at each age for white decedents; and (2) the large negative net difference rates for older nonwhite decedents arising from the reporting of younger ages onthe death certificate than on the census record for individuals over age 45. Because of these findings, statements about age comparability between rec- ords need to be made separately for the white and nonwhite groups. Geographic Region Even though the trend is toward diminishing geographic variation in age-specific death rates, differences still exist and are of interest.® Some of the variation in age-specific death rates may not reflect geographic differences in mortality risks but rather geographic dissimilarities inthe accuracy of age information. If the latter were the 32 | — White male White female Nonwhite male Nonwhite female NET DIFFERENCE RATE o -32 | | | | 5 An i’ GW | | | | | 15-24 25-34 AGE IN IO-YEAR INTERVALS 35-44 45—54 55-64 65-74 75—84 85-99 Figure 4. Net difference rates for 10-year age intervals between the death certificate and census record, by color and sex. case, then the measures ofage consistency (the net difference rate and percent agreement) would be expected to vary among the geographic regions. Net difference vate.—For the white group, net difference rates are very similar in both size and direction among regions (fig. S and table 2), and some of the regional differences could be ex- pected on the basis of sampling error alone. The overall impression is that since lack of age corre- spondence between records was not concentrated in any one region then regional differences in mortality of the white population at specific ages are probably not a result of regional differences in the accuracy of age information on the records. If age-specific death rates had been calcu- lated for the nonwhite group by region, the accu- 10 the matching racy of the rates for the South for ages 45 years and over would probably be more in question than those for the other three regions; net difference rates, although sizable for each region for older ages, were still larger for the South. Percent agreement.—The amount of age agreement between the census record and the death certificate for the white decedents in this study did not differ much from one geographic region to another (tables C and 2). Differences consisted mainly of slightly lower agreement for the Northeast Region and slightly higher agree- ment for the North Central as compared with the rest of the country, regardless of the age of the decedent. On the other hand, agreement for the nonwhite group varied dramatically among the -12 1 1 | 1 | 1 1 1 | | i 1 1 1 1 I= 5. 10- 15- 20- 25- 30- 35- 40- 45- 50- 55- 60- 65- 70- 75- 80- 85- 90- 95- 4 9 14 19 24 29 34 39 44 49 54 59 64 69 74 79 84 89 94 99 AGE IN 5-YEAR INTERVALS 32 T T T T T T T T I I T T T T T T T T 1 NONWHITE A 4 Ww sm Northeast /\ R24 mmmmm North Central o@ smn SOuth / \ monn West : 7 \ Zz w x w uw uw o - w Zz -32 1 1 | 1 1 1 1 | | ] 1 | 1 | 1 | | | 1 = 5. 10. 15- 20- 25- 30- 35- 40- 45- 50- 55- 60- 65- 70- 75- 80- 85- 90- 95- 4 9 14 19 24 29 34 39 44 49 54 59 64 69 74 T9 84 89 94 99 AGE IN 5-YEAR INTERVALS Figure 5. Net difference rates for 5-year age intervals between the death certificate and the matching census record, by geographic region and color. regions; it was higher for the West and somewhat lower for the South than for the other regions. It should be pointed out that regional differ- ences found in age agreement may be related to differences in population composition. For ex- ample, the lowest agreement found among white decedents was for the Northeast Region, where the proportion of foreign born decedents is higher than for the rest of the country. Racial composi- tion of the nonwhite group varied considerably among regions; in this study 99 percent of the nonwhite decedents in the South Region were Negro as compared with 52 percent in the West. Cause of Death Cause of death may affect the reporting of information on the death certificate. Procedures for completing the death certificate vary some- what by State although the general practice is for the funeral director to complete the personal in- formation on the death certificate from informa- tion provided by a relative of the deceased. There are, however, at least two kinds of circumstances surrounding death which could lead to variations in this procedure. Table C. The first involves deaths which occur by accident, violence, or which were for one reason or another unattended by the individual's physi- cian. In such cases, a medicolegal officer or coroner is usually required to certify the death. Although the certificate would then be sent to a funeral director, the medicolegal officer may take the responsibility for filling out the nonmedical part of the certificate, Under this general cate- gory are deaths from motor vehicle accidents, where the decedent may have carried some iden- tification (such as a driver's license) that pro- vided a statement of age. The second case includes deaths occurring in the hospital. Hospital records containing per - sonal information provided by the individual when he entered the hospital (assuming that the indi- vidual was able to provide such information when he entered) may be used for filling out the death certificate. These cases can be only approximately distinguished on the basis of what is coded as cause of death. However, if age agreement, for example, between the death certificate and the census record varied by cause, this would support a hypothesis of different circumstances surround- Percent agreement for 5-year age intervals between the death certificate and the matching census record, by geographic region,color, and age: United States, May- August 1960 Region Uni ted Solos and age Beates North- North east Central South | West White Percent agreement 1-99 years-=----cemceccmce mecca 88.1 85.4 90.1 88.3 | 89.0 1-44 years =o cco oe oe ee 90,7 88.8 91.9 | 90.9 | 91.1 45-64 yearS=mm-m-ee cece cececeee 89.4 87.8 90.8 | 89.6 | 90.1 65-99 years-=-=-=---mm meee 87.2 84.1 89.7 | 87.3 | 88.1 Nonwhite 1-99 years-====-=eemcmccccccncecccancaan 64,8 70.0 69.5 60.2 79.7 1-44 years=--=mme emcee eee 80.7 81.9 81.6 79.3] 85.6 45-64 years======-memcmeeee eee cemccmcceeaaa 69.8 70.7 73.5| 67.0 | 80.1 65-99 years=-=-c-ccccoe meee 54,6 62.9 61.3 | 48.7 76.3 12 Table D. Percent agreement for 10-year age intervals between the death certificate and the matching census record, May-August 1960 by cause of death, age, color, and sex: United States, Cause of death! Age, color, end sex Cardio- Acci- | All All 1 Malig- causes || vascular | a. Tes dents, | other diseases etc. causes 1-99 years? Percent agreement White male-==-=---=---memmecccecccccnnonn= 93,0 93.0 93.1 93.9 92.8 White female=----==-=cc-cmmmecmconoccconnn 90.5 90.1 91.2 92,1 90.5 Nonwhite male------=-=-ce-eemcemcenocccnnn 77:1 74.0 77.1 86.2 79.2 Nonwhite female-----=--===-mcmeeoeccccnen- 71.6 67.7 74.7 86.1 15.9 1-34 years White male====--==----cmeommmmmooooonnonnn 94,2 83.0 92.8 96.1 94,7 White female=---==---=--cmccmmmmccocnonon 93.4 84,1 90.8 95,7 95.6 Nonwhite male==----=-=ccmmmemcococcccnon—" 89.1 74.4 88.0 92,2 88.8 Nonwhite female-----=-=--omecmccomooccnnnn 89.2 78.8 82.4 95.0 91.4 35-99 years” White male==-=m-==--c---cmcmcemmmmme oom m 93.0 93.1 93.1 92.6 92.6 White female-=--===---cecmcmemmmoccomonon 90.4 90.2 01.3 90.6 90.0 Nonwhite male-------c-cmcmmemmcccnmcononm- 75.4 74.0 76.6 79.2 77.3 Nonwhite female-----==ccccmmmccccmcncnnnnn 69.6 67.4 74.2 77.2 71.6 lComplete category titles and numbers of the Seventh Revision of the International Lists, 1955, are as follows: Major cardiovascular-renal diseases (330-334, 400-468, 592-594) Malignant neoplasms (140-205) Accidents, poisonings, and violence (E800-E999) All other causes (residual) Includes a total of 103 death certificate. records ing the death affecting the accuracy of information reported on the death certificate. The level of age agreement did in fact vary by cause of death. When deaths were divided into the four cause-of- death categories shown in tables D, 3, and 4, more agreement was found for decedents whose deaths were from accidents, poisonings, and violence (International List Numbers E800-E999) than from all causes combined. Because age of the decedent is so closely related not only to cause of death but also to the degree of age agreement between records, espe- cially for the nonwhite group, it should be taken into account in this analysis. Since patterns of both agreement and cause of death seemed to with age reported as 100 years and over on the differ primarily between ages under 35 and those 35 years and over, only these two age groups are considered here. For decedents under age 35, the amount of age reporting agreement varied considerably among cause-of-death categories for each of the four sex and color groups. The amount of agree- ment was always substantially higher for acci- dents, poisonings, and violence and lower for major cardiovascular-renal diseases (330-334, 400-468, 592-594) than for all causes. The lowest age agreement was found for deaths from major cardiovascular-renal diseases which are rather unlikely occurrences at these ages, suggesting that age on the census record for some of these decedents was probably much younger than the true age. In fact, the major part of the disagreement was a result of age on the death certificate being from 10 to 30 years older than that on the census record. These cases of large discrepancies in age may reflect incorrectly matched records or processing errors in the coding and punching of age on the census record. Perhaps even more noteworthy is the negli- gible difference in the amount of age agreement between the white and the nonwhite individuals who died from accidents, poisonings, and violence, Several factors could play a partinexplaining this finding. Accidents, poisonings, and violence are the most frequent cause of death at younger ages. A large number of these deaths resulted from motor vehicle accidents, and some evidence of age would probably have been available for those decedents. The bulk of deaths in this general category were probably certified by a medicolegal officer or a coroner. Any one or a combination of these factors or ones in addition to these could be involved, but their contribution cannot be determined directly from these data. Curiously, however, the results for ages 35 or over did not show the same strong relationship between cause of death and degree of agreement as that found for younger ages. For older ages, there were only small differences in age agree- ment by cause. Since agreement for the white group was at approximately the same level, re- gardless of age or cause with the exception of major cardiovascular-renal diseases at the younger ages, the small differences at older ages are not remarkable. For the nonwhite group the generally lower age agreement that existed for older decedents was reflected to practically the same extent in each cause category although agreement was somewhat higher for accidents, poisonings, and violence. The effect of these differences in age corre- spondence among causes on the age-specific death rates by cause will be examined in a later section. RESULTS OF RELATED STUDIES There have been a number of other studies comparing responses on vital and census records. Procedures and target populations differ among these studies so that results are not quite com- parable. However they are of value as indicators of the variation in the circumstances of these studies that may account for the differences among the results. The study most analogous in scope and design to the one discussed in this report was that done by the General Register Office of the United Kingdom to evaluate the results of the 1951 Cen- sus of England and Wales.” For this British study, all death certificates filed during the week of May 1-7, 1951, were selected and searched for matching census records for April 8, 1951. Their match rate was 87 percent—7 percenthigher than the rate for the present study. In addition to higher match rates, agreement in single years of age for matched records was also higher for the British study: 80 percent were in complete agreement, with only 4 percent of cases differing by more than 1 year; whereas 69 percent of the present study were in complete agreement, with almost 14 percént differing by more than 1 year (table E). The similarities in results between the two studies are interesting. Generally speaking, there was more agreement in age among males than among females in both countries, Agreement tended to decline with in- creasing age and, where there was disagreement, the age on the death certificate was more often older than the age on the census record for the same individual. For the British study this last pattern was partly a result of the natural aging process between date of the census and date of death, which was not corrected as it was in the present study. The conclusion of the British study? was that: Since population estimates are based initially on the census enumeration these discrep- ancies tend to distortdeath rates in thedirec- tion of exaggerating longevity, but. . . itis clear that any exaggeration is of trivial mag- nitude for normal vital statistics purposes. On the whole it is fair to say that age state- ments at death registration have a high order of reliability, This conclusion agrees with the findings for the white group in the present study—that, although differences in age between records existed, they would probably not appreciably affect the actual age-specific death rates. However, if results for the white group were compared with those for the group in the British study, agreement would still be higher for the British study ( table E). Recently, data from an evaluation of the 1961 Census of England and Wales were made avail- able.!’ There were some procedural differences between the earlier study and the later one, but the two sets of results were very similar: for 1961, 79 percent of matched records were in agreement in single years of age, and 4 percent of the records differed by more than 1 year of age. Reproduction of almost identical results, sepa- rated in time by a decade, supports the reliability of the figures in the 1951 British study and sug- gests that the differences found between their Table E. according to sex and age: Matched Record Study results and those of the present study are prob- ably real. This raises the question as to whether either the death-registration system or the enu- meration process or both in England and Wales produce more accurate data than those in the United States. Or, on the other hand, is the popu- lation in England and Wales different on the average from that in the United States with re- spect to those characteristics which contribute to lower age agreement between records? Another study, "The Comparability of Reports on Occupation From Vital Records and the 1950 Census,"!! was done in the United States in 1950 to compare occupation information on vital and census records. Part of the study involved com- paring age information on the death certificate Percent distribution of matched census records, by age agreement with death certificate 1951 British Study and 1960 United States Census-Death Certificate 1951 British Study! 1960 U.S. Census-Death Certificate Matched Record Study Sex and death certificate age (relative to Total Whit census age) All 0-14 | 15-34 | 35-64 | years BLLe, ages || years | years | years) and |.gg | 1-14 | 15-34 | 35-64 | 65-99 | years years || years | years | years | years Percent distribution Male------- 100.01] 100.0 | 100.0 | 100.0 | 100.0 | 100.0 || 100.0 | 100.0 | 100.0 | 100.0 | 100.0 Same single year of age---------- 81.7 92.2 | 82.4 | 80.8 | 8l.6 71.8 83.3| 75.6 73.9 | 69.8 74.5 1 year younger--- 4.6 - 4.4 5.2 4.6 7.7 5.1 6.7 7.3 8.2 7.5 1 year older----- 10.4 6.9 9.9 | 10.5 v3 9.2 8.1 8.9 8.8 9.5 9.1 Younger or older by more than 1 year-----===-=---- 3.3 0.9 3.3 3.5 3.3| 11.3 3.5 8.8 10.0 | 12.5 8.9 Female----- 100.0 || 100.0 | 100.0 | 100.0 | 100.0 | 100.0 || 100.0 | 100.0 | 100.0 | 100.0 | 100.0 Same single year of age---------- 78.9 98.8 | 83.3 | 76.7 | 78.9] 64.8 82.9| 72.9 | 66.3| 63.3 67.9 1 year younger--- 5.1 - 1.0 5.4 5.4 8.6 5.2 7.8 8.5 8.7 8.3 1 year older----- 11.3 1.2 | 14.6 11.9 .3| 10.4 8.3 8.8 9.9 | 10.7 10.5 Younger or older by more than 1 year----==-====== 4.7 % 1.0 6.1 4.5| 16.2 3.6] 10.31 15.3] 17.3 13.3 1Gt. Brit. General Register Office, 1951 Census of England and Wales, General Report, H.M. Sta- tionery Office, London, 1958. 15 with that on the matching census record. For this comparison, the sample was confined to death certificates for white males aged 45-64 years who died during May-August 1950. These death certificates were selected in such a way as to in- clude a certain number of each of the major occupation categories. Of this sample, 79 percent were found in the 1950 census compared with a match rate of 81 percent for the white male decedents aged 45-64 in the present study. Age agreement between the two records for ages classified into 5-year groups was consist- ently lower for each age group in the occupation study (table F). The differences between the re- sults of the two studies are probably due to dif- ferences in procedures rather than to improve- ments in information on records from 1950 to 1960 although some difference may be attributed to the use of self-enumeration in the 1960 census and the use of year of birth in 1960 instead of age as was used in 1950. One procedural difference was that ages on the census record were not corrected for the time that elapsed between Table F. Percent agreement for 5-year age intervals between the death certif- icate and the matching census record for white male decedents aged 45-64 years: United States, 1950 Occupation Study and 1960 Census-Death Certificate Matched Record Study 1960 Death 1950 Census-Death certificate Occupation | Certificate agel Study? Matched Record Study Percent agreement 45-49 years---- 88.1 92,7 50-54 years---- 87.5 91.5 55-59 years---- 84.3 91.2 60-64 years---- 85.8 90.1 Death certificate used as base for percent agreement. 2National Office of Vital Statistics: The comparability of reports on occupa- tion from vital records and the 1950 cen- sus, by D.L. Kaplan, E. Parkhurst, and P. K. Whelpton, Vital Statistics—Special Re- orts, Vol. 53,No. 1. Public Health Serv- i Washington, D.C., June 1961. 16 enumeration and date of death in the occupation study. It was estimated that 4 or 5 percent of the decedents should be reported in different 5-year age groups. This correction would certainly raise the levels of agreement to ones approaching those in the present study. A second factor that might explain some of the, differences between studies was the sampling procedure used in the occupation study. Since the sample for the occupation study was drawn to represent the distributions of major occupation groups in 1950, the sample in the present study from 1960 deaths might be slightly different in occupation distribution. To the extent that occu- pation is related to age agreement, differences in the sample composition of occupations would affect the results, A third study, '"Matched Record Comparison of Birth Certificate and Census Information," 2 was done as part of the 1950 Birth Registration Test in the United States. The relevant data con- cern age of mother as reported on the birth cer- tificate and the matching infant card for infants born during the month of March 1950. The infant cards were filled out by enumerators during the 1950 census for infants born during January- March of that year. Both for single years of age (fig. 6) and for ages grouped into 5-year intervals, there was higher agreement in age for mothers in the birth study than for all female decedents aged 15-44 years in the present study (table G). In both studies agreement was higher for the white group than for the nonwhite, but an interest- ing difference appeared in the relative positions of the color groups when single years of age were combined into 5-year intervals. That is, agree- ment in single years of age was almost always higher for white female decedents than for non- white mothers. However, for 5-year age groups, agreement was higher for nonwhite mothers than for white female decedents up to age 35. The apparent reason for this is that 22 percent of the differences for nonwhite mothers were within 1 year of the same single year of age compared with 16 percent for the white female decedents. This relatively high agreement in the birth study existed even though no corrections were made for a possible birthday between the date the birth occurred and the date of the census. One explanation for the higher age agreement in the 100 [~ — White mothers mmmm=m White female decedents we mmm mem NOnwhite mothers 90 [— wan Nonwhite female decedents 80 — PERCENT AGREEMENT PA ~ Oo < T a Oo I 40 - 30 i 0 Inn 15 20 25 30 35 40 44 SINGLE YEARS OF AGE Figure 6. Percent agreement in age between the vital record and the matching census record for mothers (1950 Birth Study) and for female de- cedents (1960 Census-Death Certificate Matched Record Study) aged |5-Ul years, by color. birth study is the great likelihood that the respond- ent was the same (i.e., the mother herself) for both the infant card and the birth certificate. Another reason could be that if the mother's age were reported far outside the range of child- bearing ages, 15-44 years, it probably would have been detected. On the other hand, the age reported on the census record or the death cer- tificate would not fall into any such narrow range of possibilities. Finally, there are other studies carried out by the Bureau of the Census to evaluate census results, The Post-Enumeration Survey (PES)? 1950 was an intensive reinterview program car- ried out for a sample of persons to evaluate the coverage of the censuses and the accuracy of the responses obtained. The agreement in response for ages classified into 5-year groups between the census and PES was quite high—94 percent for the total group compared with 89 poreent for the present study. In the CPS- Census? match study done in 1960 to evaluate the accuracy of the data in the 1960 census, slightly higher age agree- ment was found than for the PES—O95 percent agreement for the total group classified into 5- year age groups by the census and the CPS. This may reflect improved techniques in the CPS over the PES or improvements in the 1960 census, including the use of self-enumeration and year of birth. The same sex, color, and age patterns of age agreement observed in the present study were found in both of these studies, although the per- cent agreements were much higher in the census studies. Agreement was higher for males than females, higher for white individuals than for nonwhite, and higher for younger ages than for older ages among nonwhite individuals in all three studies (table H). In view of the basic simi- larities in the collection and proeessing of data for the census, PES, and CPS, and the likelihood that the respondent was the same for both records being compared, it is not surprising that age agreement was higher for these studies than for the present study. Moreover, the census studies refer to samples of the living population while the present study refers to a sample of deaths. As an independent investigation of the 1950 census, several record check studies were done. One of these involved a sample of census records matched to birth certificates.!® Of those matched, 96 percent were in the same 5-year age group according to the census record and the birthcer- tificate. Unfortunately these results are based largely on findings for younger individuals since birth certificates were found for less than one- third of persons aged 45 years and over. In the 1951 British study discussed above, a similar record check was done, and 99 percent of the sample were in the same 5-year age group OR both the 1951 census and the birth certificate. Birth certificates were found in the British study for more than 85 percent of individuals 65 years and over in the sample. The following hypotheses are suggested by the results of these diverse studies: 1. Age reported in the registration system and the enumeration process inthe United States is somewhat less consistent than 17 Table G. matching census record for mothers color: United States, Record Study Percent agreement for 5-year age intervals between the vital record and the and for female decedents aged 15-44 years, by 1950 Birth Study and 1960 Census-Death Certificate Matched 1960 Census-Death 1950 Birth Study? Certificate Matched (mothers) Record Study (female Age! decedents) Total | White | No®= | rotall|| white| Nom- white white Percent agreement 15-19 years------ccccmmmmcmemeeeeeeeeooo 93.3 93.4 93.0 92.2 92.4 01.6 20-24 yearS=----=--cccccccmmmmemeeeeeoo 95.1 95.6 92.1 88.8 90.6 82.4 25-29 years=--=---ccmcmmmmmmmeemeeeecee 95.3 95.9 90.1 87.5 89.3 81.9 30-34 yearS=-------cmmcmccccmcemecm——eeen 94,7 95.4 88.6 84.2 87.1 75.1 35-39 years--------cemmmmmmeeeeeeee em 92.9 93.7 87.3 86.3 89.5 75.5 40-44 years------cmcccmm mm meeeeeeeeees 92.8 93.7 85.6 83.7 86.9 71.0 lage in the Birth Study refers to age on the birth record which was used as base for percent agreement; age on census record is base Matched Record Study. for 1960 Census-Death Certificate ®National Vital Statistics Division: Matched record comparison of birth certificate and census information: similar information reported in England and Wales, 2. Age information is less reliable for cer- tain portions of the population thanothers no matter which record is being consid- ered. These are generally females, non- white individuals, and persons of older ages. 3. Age information is more likely to be con- sistent from one record to another if the same individual is the respondent on these records, EVALUATION OF ACCURACY OF AGE-SPECIFIC DEATH RATE The age-specific death rate as an estimate of the probability of dying at a given age has had worldwide significance as a measure of the health status of the Nation and as an identifier of sub- groups within the population that are at high risk of death. Its importance cannot be overestimated, and for this reason its accuracy is of considerable interest. 18 United States, 1950, No. 12, Public Health Service, Washington, D.C., Mar. Vital Statistics—Special Reports, Vol. 47, 1962. For the purposes of this report accuracy is taken to imply consistency: the age reported on one record is consistent with that on the other record for the same individual. Specifically, the question posed is "If the figures for the numer - ator and the denominator of the age-specific death rate come from the same source (the census record) how would such an age-specific death rate (herewith called the adjusted age-specific death rate) compare with the published (herewith called the actual) figures?" To answer this question figures from census records for all decedents in a given year should be used. For the year in which the data for this study were collected, 1960, considerably less than all such records are available so that only an approximation to an adjusted rate is possible. The net difference rate, used in the previous sections of this report, can provide an estimate of the percentage difference between the actual and the adjusted age-specific death rates. However, the net difference rates used in this section are based on different figures from those used in the previous sections. They represent the compari- Table H. Percent agreement for States, three census studies and 1960 Census 5-year age intervals between matching records: United -Death Certificate Matched Record Study [Census record was used as base for percent agreement; 5-year age intervals were used through age 74; ages 75 years and over were treated as one interval] 1950 1960 1950 1960 census and | 02% 4 | census ang | Gengus-Death Color, sex, and age on census Birth Certificate record Certificate POSE Corpons Matched Record Enumeration | Population record Check! Survey Survey? Study? Total Percent agreement All ages-----==-=====-=----= 96.3 94.0 94.8 89.3 Under 45 years------===--==-------= 96.8 95.1 95.9 88.9 45 years and over----------=-----= 91.9 91.3 92.1 89.4 White All ageS=------=-==--c------ --- 94.7 95.4 91.4 Under 45 years-------==--=-=-==----= --- 95.7 96.5 90.7 45 years and over=---------------= --- 92.4 92.8 91.5 Nonwhite All ageS------==-=-===------ --- 88.3 89.8 69.4 Under 45 years----------=-------=-= --- 90.8 91.5 80.7 45 years and over-----------------= --- 79.3 84.7 66.6 Male All ages------===--=====----= --- 94,3 94.9 90.3 Under 45 years-----=---=--=-=-----== --- 95.5 95.7 89.5 45 years and over-------------=--- -——— 91.3 92.9 90.4 Female All ages-------=-====-------~ -— 93.8 94.7 88.0 Under 45 years=-------=-=-==--=--=== -——— 94.7 96.1 87.8 45 years and over----------------- --- 91.4 91.5 88.0 ly.S. Bureau of the Census Post-Enumeration Survey, 1950, Results 24, Age Statistics-""Record Check' Studies, Jan. 1954. (unpublished) 2y.S. Bureau of the Census: The Post-Enumeration Survey: sus, Technical Paper No. 4, Washington, D.C., 1960. Memorandum No. 1950, Bureau of the Cen- 3Uy.S. Bureau of the Census: Evaluation and research program of the U.S. Censuses of Population and Housing, by CPS-Census Match, Series ER60, No. 1964. iRefers to ages 1-99 years. 5, Washington, 1960: accuracy of data on population characteristics as measured U.S. Government Printing Office, 19 sons found between the age information reported on the death certificate and that reported for the same individual in the edited 25-percent sample of the 1960 census (stage II) (table 5) as opposed to the unedited 100-percent census enumeration (stage I) used in the previous sections. There are several advantages in using stage II data rather than stage I for evaluating age-specific death rates. The most important one is that stage II data were processed by FOSDIC (Film Optical Sensing Device for Input to Computer), whereas stage 1 data were manually coded for this study, and errors resulting from the manual process necessitated the exclusion of certain groups of records. A more detailed discussion of the dif- ferences between stage I and stage II and the reasons for preferring one to the other for the different sections of this report are contained in the Technical Appendix. This analysis is confined to death rates rather than numbers of deaths to facilitate com- parisons of results between the various sub- groups. Estimations were made of adjusted num- bers of deaths which are presented in table 8 but which will not be discussed here. The death rates considered are specific for sex, color, and 10- year age intervals and for sex, color, cause of death, and 10-year age intervals. The adjusted rate was calculated in the following way: Adjusted = Actual + (1 + Net Differcnce Rate) This method is one of several possible proce- dures for correcting age-specific death rates from the data. Another method has been suggested by Dr. Kitagawa which would yield slightly dif- ferent results. See the Technical Appendix for an outline of this method. The net difference rates used to calculate adjusted rates are shown in table 6. Figure 7 and table 7 show that the adjusted rates for white males and females at every age are remarkably close to the actual rates. For nonwhite males and females, the adjusted rate is sometimes quite different from the actual. The adjusted rates for the nonwhite group are lower at ages 45-74 (except for females 65-74) and higher at ages 75 and over than the actual. Both of these changes and the small decrease in the adjusted rate for the white group at ages 75 and over contribute to slightly smaller differences 20 between white and nonwhite individuals in ad- justed rates for ages 45 and over. The most interesting finding relates to the oldest ages 75 and over: actual rates for the white population are higher than those for the nonwhite at ages 75 and over, but at all other ages rates are higher for the nonwhite population. This reversal is diminished but not eliminated in the adjusted rates. The adjusted rates for the non- white group are as high as those for the white at ages 75-84 years. The ''crossing of the curve" occurs only at ages 85 and over, where although the adjusted rates for the nonwhite population are still lower than those for the white group, the gap is smaller than was the case for the actual rates, For the combined age group 65 years and over, the adjusted death rate is higher for the non- white group than for the white, reversing the color differentials found in the actual death rates (table 7). There is some reasonto believe that the net difference rates used to calculate these ad- justed rates are a conservative estimate of the "true" difference in age statements since these rates are based on records only for decedents who were matched with the census. In any event, the results lend support to the hypothesis that the excess in mortality for the white population over the nonwhite at ages 75 and over, in contrast with the overall pattern of excesses for the nonwhite population, is an artifact of the data rather than a real phenomenon, The adjusted rates for three major cause- of-death categories (figs. 8-10) were similar to the pattern for all causes. The adjusted rates for the white group in each cause category showed smaller changes from the actual rates than those for the nonwhite group in the same category. The adjusted rates for the nonwhite group at ages 75 and over were higher than the actual, tending to approach the rate for white individuals. (For accidents, poisoning, and violence little or nothing can be said about the rates for nonwhite persons 85 years and over because the study group had frequencies of less than 10 nonwhite individuals.) The largest differences between the actual and adjusted rates were found for major cardio- vascular-renal diseases and the smallest for accidents, poisonings, and violence, correspond- ing to the findings in the earlier section of lower and higher than average age agreement for these two categories, respectively. 800 600 400 200 MALE 800 600 400 200 FEMALE T I T T T 0 1 1 1 1 0 2 1-4 5-14 15-24 25-34 35-44 z 1-4 5-14 15-24 25-34 35-44 = = gS AGE IN 10-YEAR INTERVALS =< AGE IN I0-YEAR INTERVALS 2 2 $ P o 24.000 —— : T T T 3 24,000 T T T T — o S S o ! o ' o sms White male, actual o me White female, actual @ smmm: White male, adjusted @ rmmms White female, adjusted & Nonwhite male, actual fu memsmw= Nonwhite female, actual w 20,000 = Nonwhite male, adjusted Tow 20,000 = uuu Nonwhite female, adjusted = 5 g « x r x S g w w © 16,000 — © 16,000 — 12,000 wd 12,000 wl 8,000 — 8,000 -1 4,000 - 4,000 — 0 1 1 1 1 1 0 1 1 1 45-54 55-64 65-74 75-84 85+ 45-54 55-64 65-74 75-84 85+ AGE IN 10-YEAR INTERVALS AGE IN 10-YEAR INTERVALS Figure 7. Actual and adjusted age-specific death rates for 10-year age intervals, by color and sex. v4 MALE 4 FEMALE 0 T T T T i 240 T p= = T T mmm White male, actual i ——— White female, actual i lL smmmm White male, adjusted i - | mmmma White female, adjusted / | mmm Nonwhite male, actual m—w=s Nonwhite female, actual E wine Nonwhite mote. ogivered 4 wnnnn Nonwhite female, adjusted fF 160 - = / — 80 . 80 | J - 0 0 | = 1-4 5-14 15-24 25-34 35-44 Zz 1-4 5-14 15-24 25-34 35-44 ° o = = dl AGE IN IO-YEAR INTERVALS « AGE IN IO-YEAR INTERVALS g z o a 18,000 I [ [ I I a 18,000 T I T— I | 2 o o 8 oc oc Qe 1 e = 7 @ x b a a w 14,000 - 14,000 E < - o T x E g < - = —) W a 10,000 — 10,000 wl 6,000 =~ 6,000 - 2,000 — 2,000 -) 0 1 | | ] 0 1 45-54 55-64 65-74 75-84 85 + 45-54 55-64 65-74 75-84 85+ AGE IN IO-YEAR INTERVALS AGE IN I0-YEAR INTERVALS Figure 8. Actual and adjusted age-specific death rates for Major Cardiovascular-Renal Diseases for 10- year age intervals, by color and sex. 22 100 80 60 40 20 DEATH RATE PER 100,000 POPULATION nN ° o Oo 1,600 1,200 800 400 MALE sm White male, actual smmwm: White male, adjusted mmmmw Nonwhite male, actual aman Nonwhite male, adjusted I. 1 1 1 1 1-4 5-14 15-24 25-34 35-44 | 1 1 1 1 45-54 55-64 65-74 75-84 85+ AGE IN I0-YEAR INTERVALS DEATH RATE PER 100,000 POPULATION 100 80 60 40 20 2,000 1,600 1,200 800 400 FEMALE White female, actual White female, adjusted Nonwhite female, actual Nonwhite female, adjusted 1-4 5-14 15-24 25-34 35-44 AGE IN 10-YEAR INTERVALS 1 5 1 45-54 55-64, 65-74 75-84 85+ AGE IN 10-YEAR INTERVALS Figure 9. Actual and adjusted age-specific death rates for Malignant Neoplasms for 10-year are intervals, by color and sex. 23 MALE 300 T T T T T ] 300 200 200 100 + = 100 0 1 1 1 1 1 0 1-4 5-14 15-24 25-34 35-44 AGE IN I0-YEAR INTERVALS 2 700 T T T T T 700 z o o < & = ] 5 5 a a © o a a o 600 Bn 600 o Qo o o Q a % S 8 x x w w a a 500 - 500 Fe ke < @ x x x = = < < w w © 400 +4 © 400 300 = 300 200 ‘a, = 200 =m White male, actual 100 | smmms White male, adjusted — 100 wm wm Nonwhite male, actual mun Nonwhite male, adjusted 0 1 1 1 1 45-54 55-64 65-74 75-84 85+ AGE IN I0-YEAR INTERVALS FEMALE T T T T T sm White female, actual Coe smmms White female, adjusted 1 smmmm Nonwhite female, actual wine Nonwhite female, adjusted Nn, ITT pups Re 1 1 L 1 1-4 5-14 15-24 25-34 35-44 AGE IN I0-YEAR INTERVALS 0 1 1 1 1 45-54 55-64 65-74 75-84 85+ AGE IN I0-YEAR INTERVALS Figure 10. 24 Actual and adjusted age-specific death rates for Accidents, Poisonings, and Violence, color and sex. by Adjusted rates are not intended to be closer representations of reality than the published rates but rather to reflect the change in rates brought about by removing some of the age statement differences between the census record and the death certificate. Also these adjustments do not allow for estimates of undercoverage in the cen- sus record. The analysis is limited to only those age-specific death rates published for the year 1960, and no attempt is made to generalize these results to other years. It is further restricted in generality because of the number of limitations inherent in the data which are discussed in the Technical Appendix. These limitations include the fact that the adjusted rates are based on data that refer only to deaths sampled during May- August and matched with the 25-percent sample (stage II) census records. Moreover, this evalua- tion is specific for the characteristics considered here: sex, color, age, and cause of death. Insofar as the value of the net difference rates between ages on the census record and those on the death certificate is related to other characteristics, age-specific death rates by such characteristics should be evaluated on the basis of net difference rates for each characteristic. For example, an age-specific death rate for white males by marital status should be evaluated from net difference rates calculated for white males by age and marital status. In spite of the specificity of this evaluation, some indication of the general quality of the actual age-specific death rates for various portions of the population in 1960 is provided. DISCUSSION The results of this report confirm the fre- quently stated argument that there are inconsist- encies in the age information between the census record and the death certificate. Moreover, it has been shown that these inconsistencies can affect the level of the age-specific death rates. The question arises at this point as to what can be expected in the future in terms of improved age information. It is difficult to identify the factor or factors responsible for differences in age statements. Measures of age correspondence between records reflect the combined effect of the various differences between records, including the questions designed to elicit age response, the respondent providing the information, the care taken by the person completing the forms, and the processing of the responses. Because of these differing circumstances, a certain amount of difference in age information between the records might be expected. However, since an average of 93 percent of the white male decedents were classified in the same 10-year age interval on both records, it is clear in view of the results for white males in this study that at least this level of agreement should be possible for all subgroups of the study population. That subgroups vary radically from this average for white males is evidence that more is involved than differences in procedures between records. While the entire amount of disagreement between records cannot be attributed to any one factor, the disagreements in excess of what was found for white males can probably be ascribed to differences in characteristics of the subgroups. Then the question becomes what reasons can ac- count for those portions of the population provid- ing unreliable information and what possibilities exist for future improvements. It seems reasonable to suppose that consist- ent age reporting on records implies the existence of two prerequisites on the partofthe respondent: (1) That he knows either his age or the age of the individual for whom he is responding, and (2) that he is willing to provide this information ac- curately. The existence of these prerequisites is related to the individual's social and economic environment. The low age agreement in the older nonwhite group probably reflects the fact that they do not have birth certificates and do not know their age. For the majority of white males the environment is largely urban and industrial where records of age—including birth certifi- cates, social security enrollment forms, and insurance policies—are widely used. To the extent that the trend has been for nonwhite individuals to move from rural farm areas to urban indus- trial ones, differences in age information as a result of differences in the environment should diminish. The relatively greater age consistency for younger than older nonwhite decedents, which approached the levels found for the white group, may be taken as evidence of this movement. At any given moment in time, individuals within the two color groups may be quite similar with re- spect to age agreement. For example, if age comparability had been analyzed by socioeconomic 25 status, age agreement might have been the same for white and nonwhite individuals within each status group. In any event, as more and more use is made of records and as record keeping becomes increasingly accurate, knowledge of one's age should increase. Although sex differences were not as great as color differences, the lower agreement for females than males deserves some attention. The emphasis here, particularly for white females, should probably be on willingness to report their accurate age rather than on knowledge of it. Where disagreement existed between records, it was often a result of the age on the death certifi- cate being reported as older than that on the census record for white females. If it is assumed that women are quite likely to respond for them- selves on the census record, then the finding would support the contention that women tend to give a younger age. There is some hope for improvements in this area with the more wide- spread use of self-enumeration in the census. On the other hand, some of the age inconsistency for females might arise from a difference in respond- ents on the two records. That is, information for males might be more likely to come from the wife or other household member on both the census record and the death certificate, whereas infor- mation for females might be more likely to come from themselves on the census record and hence from a different respondent on the death certifi- cate, Much could be gained from further research into this subject of comparison of age statements in an effort to isolate the causes of disagreement and thus to provide a basis for making improve- ments. There are many characteristics of indi- viduals that need to be studied which, separately or in combination, might affect age agreement— e.g., education, urban-rural residence, marital status, nativity, and the relationship of the de- ceased to the informant on the death certificate. In addition to such studies, one involving a three- way comparison of age on the census record, death certificate, and social security form or the birth certificate would be of considerable im- portance, SUMMARY The outstanding finding in this evaluation of age response comparability on the death certifi- 26 cate and the census record for the same individ- ual is that age agreement is generally quite high for the white group and relatively low for the nonwhite. Agreement levels appeared to be closely related to age for nonwhite decedents: age agree- ment although somewhat lower for nonwhite indi- viduals than for white under age 45 was sub- stantially lower at ages 45 and over. Large discrepancies for nonwhite individuals between the ages reported on the death certificate and those on the census record have their impact on adjusted age-specific death rates based on age of decedent as reported on the census record. The adjusted rates differed from the actual rates for ages 45 and over, casting some doubt on the accuracy of the actual rate as a measure of non- white mortality risks at these ages. Other characteristics of the decedent ex- amined for their bearing on the problem were related to the amount of age agreement but to a lesser degree than color and age. Sex differences were small although there was usually less age agreement for females than for males. Geographic region of residence and the factors which vary concomitantly with geographic area—race and nativity—seemed to be related to amount of age agreement. White decedents of the Northeast Region and nonwhite decedents of the South Region had lower agreement than did decedents of other regions. Cause of death appeared to affect age agreement: higher than average agreement was found for decedents whose deaths were from acci- dents, poisonings, and violence while lower than average agreement was found for decedents whose deaths were from the major cardiovascular-renal diseases. The extent of these differences is indi- cated approximately in the comparison between actual and adjusted rates. Actual rates were dif- ferent from adjusted rates to a slightly larger degree for females than males. For accidents, poisonings, and violence the actual rates did not noticeably differ from the adjusted rates although they did for major cardiovascular-renal diseases. The results from related studies indicated similar patterns of age, sex, and color differences in age agreement between two sources although the size of the differences varied from study to study. REFERENCES National Center for Health Statistics: The change in mortality trend in the United States. Vital and Health Sta- Yt. Brit. General Register Office: 1951 Census of Eng- land and Wales, General Report. London, H. M. Stationery tistics. PHS Pub. No. 1000-Series 3-No. 1. Public Health Service. Washington. U.S. Government Printing Office, Mar. 1964. p. 37. Marks, E. S., and Waksberg, J.: Evaluation of Coverage in the 1960 Census of Population: Through Case-By-Case Checking. Paper prepared for presentation atthe annual meet- ing of the American Statistical Association, Los Angeles, Calif., Aug. 15-18, 1966. 3Siegel, J. S., and Zelnik, M.: Evaluation of Coverage in the 1960 Census of Population by Techniques of Demo- graphic Analysis and by Composite Methods. From the 1966 Proceedings of the Social Statistics Section of the American Statistical Association. fy.s. Bureau of the Census: Evaluation and Research Program of the U.S. Censuses of Population and Housing, 1960, Accuracy of Data on Population Characteristics as Measured by CPS-Census Match. Series ER60, No. 5. Wash- ington. U.S. Government Printing Office, 1964. Sy.s. Bureau of the Census: The Post-Enumeration Sur- vey, 1950. Bureau of the Census, Technical Paper No. 4. Washington, D.C., 1960. 8 raeuber, C., and Hansen, M. H.: A Preliminary Evalua- tion of the 1960 Censuses of Population and Housing. Paper for presentation and discussion atthe meeting of the American Statistical Association, Sept. 5, 1963, Cleveland, Ohio. "Bogue, D. J., Misra, B.D., and Dandekar, D. P.: A new estimate of the Negro population and Negro vital rates in the United States, 1930-60. Demography 1(1):339-358, 1964. Metropolitan Life Insurance Company: Geographic dif- ferences in longevity diminishing. Statis.Bull.Metrop.Life Insur.Co., Vol. 46. New York. July 1965. pp. 1-4. Office, 1958. pp. 35-43. 10G¢. Brit. General Register Office: 1961 Census of Eng- land and Wales. To be published National Office of Vital Statistics: The comparability of reports on occupation from vital records and the 1950 cen- sus, by D. L. Kaplan, E. Parkhurst, and P. K. Whelpton. Vital Statistics—Special Reports, Vol. 53, No. 1. Public Health Service. Washington, D.C., June 1961. I2National Vital Statistics Division: Matched record com- parison of birth certificate and census information, United States, 1950. Vital Statistics—Special Reports, Vol. 47, No. 12. Public Health Service. Washington, D.C., Mar. 1962. 13y.S. Bureau of the Census: Post-Enumeration Survey, 1950, Results Memorandum No. 24, Age Statistics-‘‘Record Check’’ Studies, Jan. 1954. Unpublished manuscript. Hagan, E. M., and Hauser, P. M.: Methods used ina current study of social and economic differentials in mortality. Emerging Techniques in Population Research, Proceedings of the 1962 Annual Conference of the Milbank Memorial Fund. Lgitagawa, E. M., and Hauser, P. M.: Education and In- come Differentials in Mortality, United States, 1960. Paper presented at the annual meeting of the Population Associa- tion of America, New York, Apr. 29-30, 1966. Revised, Mar. 1967. 16k tagawa, E. M., and Hauser, P. M.: Education differ- entials in mortality, by cause of death, United States, 1960. Demography, Vol. 5, No. 1, 1968. BIBLIOGRAPHY U.S. Department of Health, Education, and Welfare: White- nonwhite differentials in the United States. Health, Education, and Welfare Indicators. Washington, D.C., June 1965. Myers, R. J.: Accuracy of age reporting in the 1950 United States Census. J.Am.Statist.A. 49:826-831, Dec. 1954. National Center for Health Statistics: Annual summary for the United States, 1965. Monthly Vital Statistics Reports, Vol. 14, No. 13. Public Health Service. Washington, D.C., July 1966. Sauer, H. I.: Adequacy of Age Data for Age-Specific Death Rate, Reliability and Validity of Age as Entered on Death Certificates, Charleston County, South Carolina, 1961-63. Paper presented atannual meeting of the Population Associa- tion of America, New York, Apr. 29-30, 1966. 000 27 Table 1. DETAILED TABLES Cross classification of 5-year age intervals as stated on the death certificate and on the matching 100-percent enumeration census record (stage I), by color and sex: United States, May-August 1960-------------------eceeooooccoocconncon=- Net difference rates and percent agreement for 5-year age intervals between the death certificate and the matching 100-percent enumeration census record (stage 1), by geographic region, color, and age: United States, May-August 1960-------- Number of matching 100-percent enumeration census records (stage I) for 10-year age intervals, by cause of death, color, sex, and age: United States, May-August 1960===-m===m--eceememememeemeemeemem memes eeeemmmmmmemeeemseSssssssss-oooooo-c-- Net difference rates and percent agreement for 10-year age intervals between the death certificate and the matching 100-percent enumeration census record (stage I), by cause of death, color, sex, and age: United States, May-August 1960------ Cross classification of 5-year age intervals as stated on the death certificate and on the matching 25-percent sample census record (stage II), by color and sex: United States, May-August 1960---=--==------c-c-occmoccoomomomomooooooooo ooo m=s Number of matching 25-percent sample census records (stage II) and net differ- ence rates for 10-year age intervals, by cause of death, color, sex, and age: United States, May =BUSUSE LOB0 mmm mm mies mm mm mms unm us ohio ob om 0 ml so Sah kf Actual and adjusted age-specific death rates for 10-year age intervals, by speci- fied causes of death, color, sex, and age: United States, 1960-----------==-----= Estimated corrected number of deaths for 5-year age intervals, by color, sex, and age: United States, 1960------=-=--=-==-c-=-c-co-=cooo-oo-oooooooooosoooosooooomsos Page 30 34 35 36 38 42 43 Dh 29 Table 1. Cross classification of 5-year age intervals as stated percent enumeration census record (stage I), by color on the death certificate and on the matching 100- and sex: United States, May-August 1960 © oN OoOV PWN ~ I I i eS SS pS HF OoOwomNOOUNWNHO N N 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 30 Age on death certificate Color, sex, and age on census record Total, 1-99 1-4 5-9 | 10-14 | 15-19 | 20-24 | 25-29 | 30-34 | 35-39 | 40-44 years years | years | years | years | years | years | years | years | years White male Total, 1-99 years-------c---ao- 199,642 || 1,697 | 1,243 | 1,255 | 1,951 | 1,540 | 1,482 2,082 3,177 5,261 1-4 years====---c-occmmmmmeeee oo 1,708 1,649 36 1 - 1 2 - - 1 5-9 years======c--emeeccmcmmemeeaoo 1,243 33 (1,179 22 1 x - x - - 10-14 years=------cmommmme eo 1,251 4 10| 1,177 41 4 - 1 - 2 15-19 years------cocmmemo cmon 1,938 - 3 28 | 1,837 40 4 - 2 4 20-24 yearS=--m=c- mmm 1,546 3 - 2|° 281,416 51 14 5 4 25-29 years------mm-mmmmmeeoemeoaoo 1,536 - 2 1 4 46 | 1,365 45 10 4 30-34 years=----m-mmmmmmmm 2,178 3 4 1 2 3 40 | 1,931 84 13 35-39 years------ecmmmmmeo_ 3,280 4 3 6 3 2 8 63] 2,977 108 40-44 yearS=---mccmmmmme oo __ 5,516 - - 8 11 4 - 10 65| 4,868 45-49 years------ommme oo. 8,733 - 1 1 13 7 4 3 9 182 50-54 years=--=---cmmmmmee__ 12,626 - - 1 4 5 3 4 3 37 16,503 - - 1 4 4 2 3 8 8 21,688 - 1 - - 1 1 1 6 6 65-69 years=-----oocmomooooo. 27,097 - - 1 1 6 1 4 4 70-74 years=-----ecmmmmmmeee o_o 29,843 - 1 4 1 - 1 - 2 10 75-79 yearSweceemmmsermmnmnmmnnean= 26,917 - 2 1 - # - ~ x 80-84 years=------mmmmmmeeo_ an 19,817 - 1 - - - 1 2 1 85-89 years------ecocmmmmmeeo_ 11,487 i - x - - - - - 90-94 years------mmommmeeeeoooo_o 3,946 - - - - - - - - 1 95-99 years--mmeceememccec em ————— 789 - - - - - - - - - Not stated or not valid------------ 2,091 8 4 4 8 10 16 19 43 51 White female Total, 1-99 years-----=-o----- 149,902 || 1,274 768 634 730 649 767 | 1,265 2,160 | 3,220 1,285 1,249 27 2 1 - - 1 - - 773 17 | _726 10 2 1 - - 1 1 651 4 7| _604 18 1 1 - - 2 747 - 2 13 690 17 1 - 5 657 1 2 - 22 595 15 5 2 801 1 - 1 2 23 713 33 10 2 1,336 1 - 1 - 2 23 | 1,164 79 14 2,226 - - 2 - 2 8 45) 1,992 99 3,376 - 1 1 - 1 - 3 49] 2,935 4,786 - - - 5 2 1 i 12 122 6,411 = 1 7 = 2 - 1 2 21 8,393 - 1 - - 1 - 2 5 7 12,150 1 - - - 1 - - 1 4 16,617 - 1 - - - 1 1 - 2 22,199 - - - - - - - 1 3 23,238 - - - - - - 1 1 1 21,372 - - - - - - 2 1 - 14,699 - - - - 1 2 1 1 = 6,516 - 1 - - - - - - - 1,669 - - - - - - - - - 1,842 5 3 2 5 6 7 11 19 24 Table 1. Cross classification of 5-year a percent enumeration census record 2 intervals as stated on the death certificate and on the matching 100- stage I), by color and sex: United States, May-August 1960 —Con. - Age on death certificate a, 45-49 50-54 55-59 60-64 65-69 70-74 75-79 80-84 85-89 90-94 | 95-99 not years years years years years years years years years years | years |valid, or 100+ years 8,560 12,55) 16,602 21,647 | 27,109| 29,650 | 26,871| 20,431| 11,661| 4,144 728 37 - - 3 1 4 6 - - 5 = - - - 2 3 x - - - - - - - - 5 1 2 1 - 2 - 1 - - - - 1 4 1 - 2 2 - 5 5 - - - 6 8 3 2 4 - - - - - - - 7 9 18 9 6 - 10 - - - - - 7 17 14 15 16 18 10 - - - - - 30 13 15 11 20 12 - 5 - - - - 263 147 20 24 26 24 5 25 11 5 - - 7,938 317 118 26 26 22 20 35 6 5 - - 223 11,479 586 129 51 26 30 25 15 5 - 1 62 446 15,142 557 131 44 46 20 20 - 5 2 7 73 488 19,512 865 426 91 150 35 20 5 10 3 18 145 955 | 24,490 955 370 50 75 10 - 2 2 6 26 333 1,041 | 26,338 1,356 546 151 25 - 1 4 6 8 30 360 1,454 | 23,868 935 217 30 - 1 2 4 4 27 33 244 838| 17,818 703 136 - 3 - x 4 6 22 55 192 735| 10,195 265 10 1 - - 1 6 8 22 20 76 198 3,558 56 1 - - & 2 4 - 15 5 25 85 _652 15 98 148 206 264 312 292 225 206 112 60 5 5 4,731 6,325 8,378 12,046 | 16,657 | 21,074 | 23,118| 22,585| 14,987 6,886| 1,648 50 - 1 3 1 - - - - - - - 6 - 2 2 - 6 - - - 5 - - - 5 1 - 1 2 - - 5 - - - - 1 1 3 1 2 4 - 5 - - - - 3 6 2 3 2 2 - - - - - - 3 2 4 1 4 - - - - - - - 10 7 8 2 4 6 5 5 - 5 - - 21 10 12 6 6 3 5 5 5 - 5 - 215 71 17 17 15 16 20 15 - - - 1 4,251 223 87 22 24 16 15 5 - - - - 148 5,627 393 102 18 26 25 20 25 - - - 59 293 7,345 419 159 42 20 15 20 5 - 1 3 55 338 10,358 813 351 90 80 25 30 - 1 4 13 137 799 | 14,155 893 370 96 105 30 10 1 1 4 16 246 1,006 { 17,973 1,741 840 208 150 10 3 4 4 7 36 351 1.3151 19,495 1,533 425 35 30 10 2 4 3 22 50 317 1,002 | 18,783 1,015 151 20 - 1 1 1 5 20 85 295 1,023 | 12,731 496 36 5 - - - 4 6 23 25 120 373 5,819 145 10 - - - 1 14 2 10 35 50 165| 1,392 10 52 93 124 136 206 262 320 257 185 115 10 5 — OO ON On WN I I I TE rT = HH OW ®ONOGOWLPPHWNHEO N N 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 31 Table 1. Cross classification of 5-year age intervals as stated percent enumeration census record (stage I), by color and on the death certificate and on the matching 100- sex: United States, May-August 1960-—Con. 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 Age on death certificate Color, sex, and age on census record Toes), 1-4 | 5-9 | 10-14 | 15-19 | 20-24 | 25-29 | 30-34 | 35-39 | 40-44 years years | years | years | years | years | years | years | years | years fe Nonwhite male Total, 1-99 years-=-=-e-ma= -=-| 20,142 470 244 271 343 296 378 443 662 900 1-4 years 476 450 20 1 - - - - - - 5-9 years 233 7 214 5 1 - 3 - - - 10-14 280 4 5 252 9 1 2 - x 1 15-19 343 1 1 10 312 8 1 - 1 2 20-24 315 1 - 1 4 256 22 3 1 4 25-29 384 2 1 - 3 21 306 24 4 4 30-34 464 - 1 - - 1 31 347 51 12 35-39 695 1 1 - - 2 6 56 532 54 40-44 897 2 - - 2 3 2 6 45 663 45-49 1,220 1 - - 2 - 1 i 12 124 50-54 1,578 1 1 2 - 1 2 4 5 19 55-59 1,878 - - - - - 1 1 5 6 60-64 2,158 - - - - 2 - - 4 7 65-69 2,525 - - - - 1 - - 1 2 70-74 2,389 - - - - - - 1 - - 75-79 1,999 - - - - - § - - 2 80-84 1,150 - - - - - - - - - 85-89 790 - - - - - - - - - 90-94 266 - - - - - - - - - 95-99 105 - - - - - - - - - Not stated or not valid-==e=-eeceeca- 391 3 1 1 2 3 2 13 25 20 Nonwhite female Total, 1-99 years=---eeecceecan- 16,752 350 176 121 146 174 264 413 643 854 356 343 9 - - - - - 1 - 170 6 159 2 - - 1 - - - 134 ¥ 5 117 3 2 1 1 » - 154 - 1 2 141 6 - 1 - - 182 - - - 1| 150 13 4 3 2 260 - - - - 12| 213 17 10 1 434 - - - - 3 29| _326 51 10 657 - 1 - - 1 4 46 | 496 63 856 - - - - - 2 11 59 608 1,044 - - - - - - 2 15 112 1,174 - - - - - - 2 3 29 1,454 - - - - - - 2 4 17 1,610 - T - - - - - 1 4 2,035 - - - - - - - - 2 2,043 - - - - - - - - 3 1,700 - - - x - - - - 1 1,182 - - - - - 1 - - - 803 - - - - - - 1 - 1 364 - - - - - - - - - 140 - - - - - - - - 1 Not stated or not valide=-e-=cec-caa- 349 3 1 - - 1 6 7 13 13 32 Table 1. Cross classification of 5-year age intervals as stated on the death certificat percent enumeration census record (stage I), by color and sex: United States, May e and on the matching 100- -August 1960--Con. Age on death certificate Not stated, 45-49 50-54 55-59 60-64 65-69 70-74 75-79 80-84 85-89 90-94 | 95-99 not years years years years years years years years years years | years va BD 1,251 1,721 2,125 2,456 2,727 2,291 1,662 1,012 592 215 83 45 1 2 1 1 - - - - - - - 1 - 1 = - - 2 - - - - - » 1 - - - 3 1 - - - - - - - 2 1 2 - 1 - 1 - - - 1 1 6 2 1 - 2 1 - - - - - 5 5 2 3 3 - 1 - - - - - 5 6 3 1 - 1 1 3 - - 1 - 15 7 9 1 5 1 2 2 1 - - 1 99 32 15 9 7 6 2 2 - 2 - - 923 89 37 10 8 3 6 1 2 - - 2 139 1,191 150 34 17 10 2 - - - - 3 39 256 1,376 124 45 14 7 2 2 - - 1 13 68 281 1,494 187 60 16 18 4 1 - 1 8 34 126 410 1,718 156 48 10 7 4 - 3 1 12 66 198 386 1,418 203 72 25 5 2 7 - 3 37 104 224 364 1,100 117 40 3 4 4 1 4 10 42 66 147 166 _635 65 11 3 10 - 1 8 17 39 73 78 124 _400 40 10 4 - 2 - 2 14 27 24 15 37 130 15 3 - - 1 3 5 5 5 10 9 19 _48 4 30 39 56 52 34 37 29 19 15 7 3 2 1,097 1,417 1,703 2,085 2,053 1,843 1,425 962 631 286 109 48 - - 2 - 1 - - - - - - 3 - 1 1 - - - - - - - - - 1 1 - - 1 1 » - - - - - 1 2 -" - “ = - - - - - - 2 2 - 5 - - - - - - - - 1 i 2 2 - 1 - - - - - - 5 Zz 2 1 1 3 1 - - - - - 16 5 10 6 3 2 z 1 1 - - - 94 38 17 12 6 3 4 2 - - - 2 8 92 33 19 5 8 6 - 1 - - - 135 834 116 | ~ 32 9 3 5 4 - - 2 - 58 262 916 123 39 18 7 7 - 1 - 1 18 84 295 971 143 55 22 10 4 2 - 1 13 48 181 446 1,057 190 66 19 8 4 1 5 1 27 61 267 404 998 169 74 30 8 1 2 1 12 38 108 228 301 835 120 40 12 3 7 = 3 20 57 93 158 159 377 93 12 9 4 - 2 5 26 43 83 103 101 _390 44 4 6 - - 2 5 12 17 33 3 53 182 29 5 - 1 2 5 8 2 13 16 11 21 12 25 40 36 42 44 31 39 27 11 8 2 1 — © ONO UL WN EEE ETE EE aE A = — OWN WAH WNFEO ~N N 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 33 Table certificate and the matching 100 gion, color, and age: United Sta 2, Net difference rates and percent agreement for 5-year age intervals between the death -percent enumeration census record (stage I), by geographic re- tes, May-August 1960 Region Region United United Color and age 5 States North- North States North- North South | West east Central east Central oy es White Net difference rate Percent agreement Total, 1-99 years- Saw con ces "ve 88.1 85.4 90.1 88.3 89.0 1-4 years==-==w--- -0.7 1.2 -2.2 -0.5 96.8 97.5 96.1 96.9 97.2 5-9 years------e-- -0.2 -3.1 2.0 -0.2 94,5 01.6 96.7 94,2 94,8 10-14 years------- -0.7 - -0.3 -1.9 93.6 93.6 94.8 92.3 93.9 15-19 years------- -0.1 -0.3 -0.1 0.3 9.1 93.6 95,1 93.4 94,3 20-24 years-=--=--- -0.6 -0.6 -0.6 0.7 91.3 89.2 92.6 91.5 91.4 25-29 years------- -3.8 -6.9 -3.8 -3.2 89.0 85.9 89.7 89.6 91.2 30-34 years------- 4.8 -5.7 -4,1 -3.8 88.1 85.4 89.9 88.1 88.5 35-39 years-=---== 3.1 -5.7 -2.9 -0.6 90,2 87.9 91.2 91.6 90.3 40-44 years------- 4,6 -5.9 4,2 ~4.4 87.8 86.3 89.0 88.0 87.8 45-49 years------- -1.7 -3.0 -1.9 -1.0 90.2 88.4 91.4 90.3 91.0 50-54 years=------ -0.8 -0.5 -0.3 -1.1 89.9 88.7 91.2 89.7 89.7 55-59 years------- 0.3 -0.2 0.4 0.7 90.3 88.4 91.5 90.4 92.0 60-64 years------- -0.4 0.3 -0.9 -0.2 88.3 86.6 89.8 88.5 88.5 65-69 years------= 0.1 1.7 -0.2 -1,2 88.4 86.1 90.3 88.4 89.6 70-74 years------= -2.5 -2.7 -2.9 -2.2 85.1 82.3 87.2 85.4 86.5 75-79 years=------ -0.3 -0,2 -0.1 -0.8 86.5 82.8 89.4 86.3 87.7 80-84 years------= 4,4 5.1 3.4 4,8 88.9 85.6 91.2 89.4 89.1 85-89 years------- 1.8 -1.6 3.5 2.4 87.6 83.0 91.3 87.6 87.6 90-94 years=------ 5.4 6.1 5.4 6.6 89.6 87.9 91.5| 89.7 | 88.9 95-99 years------= -3.3 4,5 -6.8 -9.9 83.2 79.6 85.3 79.8 88.3 Nonwhite Total, 1-99 years- be FE oe ‘on 64,8 70.0 69.5 60,2 79.7 1-4 years--------- -1.4 0.8 1.4 -3.1 95,3 96.2 95. 9 94,8 95.7 5-9 years--------- 4,2 6.6 - 5.8 92.6 95.1 93.2 92.4 88.9 10-14 years------- -5.3 -3.6 -12,2 =5.1 89.1 87:5 87.8 88.7 97.1 15-19 years------- -1.6 2.3 - 1.4 91.1 93.0 90.5 91.4 88.6 20-24 years------= =5.4 5.4 10.0 -7.9 81.7 85.1 88.3 78.4 87.9 25-29 years------- -0.3 1.6 - - 80.6 84.4 82.6 17.9 85.5 30-34 years-=----- 4,7 -10.6 -6.0 -1.3 74.9 73.3 79.3 73.4 78.1 35-39 years------- w3eS -6.8 -0.4 -3.3 76.0 80.0 78.3 72.9 82.0 40-44 years------- 0.1 0.7 -6.5 2.3 72,5 74.3 72.6 70.7 80.4 45-49 years------- 3.7 - 2+1 5.9 73.9 712.7 77.1 72.4 79.5 50-54 years------= 14.0 11.4 10.4 16.9 73.6 74.1 27+0 71.4 81.5 55-59 years------- 14.9 6.4 10.3 20.6 68.8 70.4 72.3 65.6 79.2 60-64 years------- 20.6 6.8 16.3 28.1 65.5 67.4 69.9 61.7 | 80.4 65-69 years------- 4,8 3.4 1.8 6.5 60.9 67.3 64.6 56.9 76.4 70-74 years------= -6.7 -9.,2 -9.1 -6.0 54,5 61.1 60.2 49,2 74.4 75-79 years------- -16.5 -4.8 -10,0 | -22,3 52.3 61.3 60.0 45,8 77:5 80-84 years------- -15.4 =4.4 -9.5| -21.9 52.0 56.6 63.7 44,0 80.3 85-89 years------- -23.2 -15.9 -12.4 | -29.7 49.6 64.1 59.8 40,8 75.2 90-94 years------- -20.5 -10.7 -16.0 | -25.0 49.5 67.9 47.9 45,5 75.0 95-99 years------=- -21.6 =5.6 -26.8 44,1 72.2 51.2 39.5 42,9 34 Table 9. Number of matching 100-percent enumeration census records (stage I) for 10-year age intervals, by cause of death, color, sex, and age: United States, May-August 1960 Cause of death? L Color, sex, and age wy Cardio- — Accis All a vascular 2 Z dents, | other cause diseases | M@1¢ gs etc. causes White male TOLal, TFT VEAL Grew mmm mim ann sia 8m 0 hm tm 199,667 118,044 34,147 | 17,084 30,392 1-4 yedr§=-====-smm==-mmsm=sssmee-ee===SsssmessseTEoToTEmmTmmn 1,708 53 247 661 747 5-14 years 2,494 110 361 1,425 598 15-24 years 3,484 210 285 2,528 461 25-34 years 3,714 677 474 1,912 651 Bll FOAL me mmm mm mo 8,796 3,673 1,369 2,162 1,590 45-54 years----=mmmmmmmm==mmmmmmm=m=-e-=====sssssessmmmm TTT 21,359 11,403 3,998 2,426 3,532 55-64 years--------===s--===S--ssso-SSSSSSoooooo-SSSoooSTonTT 38,201 22,045 8,195 2,216 5,745 55274 EI Grrmrrishis sm mmm rset ot SR EU SE om ro 56,940 35,556 10,858 1,925 8,601 75-84 YeArS----smmemmmmmr mmm emma Ww mmm TT 46,734 32,096 6,823 1,382 6,433 85-99 years----------=-m-====-=-=-=-s--ssssesssooosoomomsosEEsS 16,237 12,219 1,537 447 2,034 149,942 90,164 30,030 7,363 22,385 1,290 41 178 437 634 1,424 88 267 527 542 1,404 178 178 609 439 2,137 408 489 542 698 5,602 1,362 2,045 751 1,444 11,197 3,726 4,478 817 2,176 20,543 9,959 6,453 850 3,281 65-74 years---------mm---s=ssms==-—mmessse=sms=-msssssssTeComTT 38,821 24,098 8,279 961 5,483 75-84 years--------=======o-------s-so-o--ooosoSoSSSoSmooomTT 44,615 32,424 5,963 1,120 5,108 85-99 YEArS----==--memmmsmm mmm 22,909 17,880 1,700 749 2,580 Nonwhite male Total, 1-99 years--==-mmm======m-=--s--e-==-=-=sss-oo---os 20,157 10,289 3,031 2,583 4,254 FE et 477 22 31 153 271 SL Tr st mb wm msm i RE 0 mt i 513 32 30 319 132 15-24" years-==-=======-m==m-memmm------oo-osssoosoSoSsSoo-osSmTT 658 52 28 477 101 25-34 848 160 44 435 209 35-44 1,592 626 190 396 380 45-54 2,798 1,361 460 339 638 55-64 4,033 2,151 801 231 850 65-74 4,915 3,028 856 135 896 75-84 3,154 2,040 486 72 556 85-99 1,169 817 105 26 221 TOLALs L0G YORE Grom mm mom msm ios Saf eo 10m a to = 0 16,775 9,604 2,734 872 3,565 1-4 years----====mm==m-=sssmomssmosseiemo-o-SSsSmoosomoTomoTo 358 17 18 116 207 5-14 years=---=-====---======-=-=------o-o-- 304 33 31 122 118 15-24 years 336 64 28 89 155 25-34 years-------------ss-ssssSSosoosesososo-SooSoSSSoooTTo 694 198 82 111 303 35-4l years--------=-m=----oSSSSSSSssssoooooSSSSoSSSSoooTooTT 1,513 581 326 117 489 45-54 years=---==--=====-=--ssSSsssooooooo-o-ssoososoooooooooo 2,218 1,094 557 94 473 55-64 YEAr§--=======memmmm-sSmm=—m—-e-ees=—Sssesessemesnes oT 3,064 1,837 622 72 533 65-74 years---------==m==-=-==-oSSSSSSssssoso--oSooooosoooooooo 4,079 24723 649 65 642 75-84 years------------m=mm-ooosSSSSoSooooooososoSSssssooooooo 2,884 2,068 323 61 432 85-99 years-----mm-sm-mmsmmessmmsm—ssssmmssssssmsssmsemm mess 15325 989 98 25 213 Includes 103 records with age reported as 100 years and over on the death certificate. Complete category titles and numbers of the Seventh Revision of the International Lists, follows: Major cardiovascular-renal diseases (330-334, 400-468, 592-594) Malignant neoplasms (140-205) Accidents, poisonings, and violence (E800-E999) All other causes (residual) 1955, are as 35 Table 4. Net difference rates and percent agreement for 10-year age intervals between the death certificate and the matching 100-percent enumeration census record (stage I), by cause of death, color, sex, and age: United States, May-August 1960 Cause of death? Cause of death? Color, sex, and age! 8 All Carin Malig- pi ALL All Caran. Malig- Ace ALL causes vascular nancies ents, | other causes vascular nancies ents, | other diseases etc. causes diseases etc, causes White male Net difference rate Percent agreement Total, 1-99 years- _— “on cos cee oe] 93.0 93.0 93.1 93.9 92.8 1-4 years====e=eecanna -0.6 -22.6 2.4 0.6 0.4 96.5 75.5 96.0 96.2 98.5 5-14 years-=------cee-- 0.2 -6.4 -0.3 0.1 1.7 95.7 88.2 96.1 95,9 96.5 15-24 years=-=--eee--- 0.2 -13,3 -3.2 2.0 -1.5 95.3 81.9 91.9 97.0 94.1 25-34 years-------=-u- 4.0 -11.2 -6.3 0.1 7.1 91.0 83.2 89.2 95.0 88.9 35-44 years----=---ean =4.1 -5.7 -5.8 2.0 -1.6 91,2 90.1 89.4 93.4 91.9 45-54 years-----eeeuna wleZ -1.4 -0.9 -0.3 led 93.4 93.4 93.8 93.7 93.0 55-64 0.2 0.1 0.1 -0.1 0.5 93.5 93.3 93.9 93.9 93.3 65-74 -0.3 -0.4 0.8 2,5 -1.1 92.8 92.8 93.5 90.2 92:1 75-84 1.2 1.1 1.1 2.1 15 93.0 93.3 92,2 90.9 92,6 85-99 1,9 2.0 -1,0 3.1 3.9 92,7 93.1 90.2 91,9 92.5 White female Total, 1-99 years- en i me tie “hw ‘em 90.5 90.1 91.2 92.1 90.5 -0.9 ~17.1 -1.1 -2.5 0.6 97.2 82.9 94.9 96.6 98.4 =-1.5 =10,2 -7.1 1.9 -0,7 94.6 84,1 90.6 97.0 95,9 -1.8 -12.4 -6,2 1.8 -0.7 93,6 83.1 88.8 96.9 95.2 -4.,9 -10.8 5,7 -3.0 -2.4 90.5 84.6 90,2 92.4 92.8 =4.,0 =5.3 =4.4 «1,1 -3.7 90.6 88.4 90,9 92.8 91.1 -1.3 -3.2 -0.5 -1.3 0.6 91.5 89.4 92.9 92.0 92.3 55-64 years-------c-u- -0.6 1.7 1.3 -0.6 -1.1 89.9 88.2 92.3 92.0 89.4 65-74 yeargs------e---- -2.8 -2.8 -2.6 -6.5 -2.6 87.7 87.4 89.5 84.0 86.7 75-84 years-----=-c--- 2+5 1.7 4,1 7.2 4,1 91.5 91.4 92.0 92.5 91,0 85-99 years-------e--- 2,8 3.2 2.8 1.5 1.2 92.7 93.0 89.5 90.8 92.8 Nonwhite male Total, 1-99 years- .e — TI Eon te 77.1 74.0 77.1 86.2 79,2 1-4 years-=---e-cece-- “1.3 -18.2 - -4.6 145 94,5 81.8 96.8 92.8 95.9 5-14 yearse-----eme-u-- 0.4 -18.8 3.3 1.9 0.8 92.8 71.9 100.0 94.4 92.4 15-24 years===-------- -2.9 -23.1 -21.4 1:5 -7.9 89.7 69.2 75.0 93.9 84.2 25-34 years=---===-==-o= “32 -11.3 4,5 2.5 -8.6 83.5 75.6 81.8 88.5 79.4 35-44 -1.9 -0.8 -10,5 -0.5 -0.8 81.3 79.7 79.5 83.8 82.1 45-54 6,2 7.1 1.6 2.9 5.2 83.7 82.4 84,3 83.8 85.9 55-64 13.6 18.0 9.1 5.6 8.8 81,2 80.8 81.5 79.2 82.6 65-74 2.1 2.6 3.6 -15.6 1.7 74.8 74.8 77.1 65,2 74,1 75-84 -15.1 -15.9 -17.3 -27.8 -9.4 64,1 63.3 64,0 56.9 67.4 85-99 «23.3 -23.5 -26.7 11.5 -21.3 61.0 61.4 54,3 84.6 61,1 Nonwhite female Total, 1-99 years- ev wid FF — «wie 71.6 67.7 74.7 86.1 75.9 1-4 years=----=-eeecan -1.7 -23.5 -11.1 -2.6 0.5 96.3 76.5 88.9 95.17 98.1 3-14 years—--m=mewmnmm -2,3 =6.1 3.2 1.6 -5.9 93,1 87.9 93.5 96.7 90.7 15-24 years==-----ea--- 4.8 -18.8 -14.3 4,5 -2.6 88.7 75.0 75.0 96.6 92,3 25-34 years--------u-- =2.4 -10.6 3.7 uid 1:3 84.3 78.8 79.3 91.0 86.8 35-44 years------u-u-- -1.1 -0.7 -1.8 -3.4 -0.4 81.0 77.8 81.9 86.3 83.0 45-54 years=----=-mea= 13.3 16.7 12.2 11.7 7.2 81.7 80.8 85.6 85,1 78.4 55-64 years-----=----- 23.6 26.6 15.3 13.9 24,6 7542 74.0 11.5 70.8 71.5 65-74 years-------e--- 4,5 -1.5 -13.7 “7 sd -7.3 65.0 65.2 65,3 64.6 63.6 75-84 years-----=----- 1742 -18.0 -13.0 -19.7 -16.4 58.7 57.5 63,5 67.2 59.0 85-99 years--------ea- 21.5 -21.5 =24,5 - -18.3 60.7 60.6 56.1 80.0 64.8 Includes 103 records with age reported as 100 years and over on the death certificate. 2 LjComplete category titles and numbers of the Seventh Revision of the International Lists, 1955, are as ollows: Major cardiovascular-renal diseases (330-334, 400-468, 592-594) Malignant neoplasms (140-205) Accidents, poisonings, and violence (E800-E999) All other causes (residual) 36 Table S. Cross classification of S-year age intervals as stated on the death certificate and o cent sample census record (stage II), by color and sex: United States, [Numbers include allocations of ages not reported on the census record; number of allocations is in superscript] n the matching 25-per- May -August 1960 38 Age on death certificate Color, sex, and HS en-gensus eoas Totaly 1-4 | 5-9 |10-14 | 15-19 |20-24 | 25.29 | 30-34 | 35-39 | 40-44 | 45.49 a years years years years years years years years years years White male Total, 1 year and over-- |47,972%*2 || 4022 | 309% | 330% | 447 | 379% | 390% | s285| 8145 1,370 | 2,07818 1-4 yearsammmmmemmn= ——— 406% 387° 121 i= - 1 1 2 ” = a 5-9 JEArSmmmmmmmmm mmm 301% 9 | 281 pt i - - |, ® = - 10.14 Jorrmeennemn nme 329 3 4 302 16 1 - 1 @ - L AS 19 FEET Hmmm 4311 ~ 2 S 408 5 1 - 1 1 2 P02 JEEP Burris 4042% - w- 1 4 | 349 14 BR 4 - 1 25-29 1 3 - 2 132 | 356 71 6 1 1 30-34 1 1 1 1 4 483? 181 gt 32 35.39 - 1 3 1 1 3 20 762 27° iz? 40-44 2 1 3 3 1 - 3 il [1.252 67% ii dn 45-49 - 2 1 5 « 4 - 2 49 1,886 50-54 " 1 2 3 2 1 2 32 3 53 55.59 = = - 1 2 4 1 3 22 32 60-64 2 - i” - 1 @ 1 1 2 62 65-69 = = = 1 1 1 - 2 gt 59 70-74 - 1 1 1 1 1 - # 41 &* 75-79 w = - - " “ 1 2 “ 3* 80-84 - - 1» = = a 1? = el ~ 85-89 yearsmmm- - " = ® = - - - 12 - 90-94 yearsmmm- 8852 w = = - = - - - - - 95-99 yearsemmm-mn- 165 ” " = - R 2 - w @ 12 100 years and over 41 = = - ~ - - - - = - White female Total, 1 year and over-- |35,624%°7 || 3242 | 181% | 161% | 195 | 1431 | 186 | 305% | 5555 | 8078 1,124 led YEATSmmmmmmmmmmccee man 555° 3161 5 1 - ” 1 1 - 2 = Sul VERT mmm macnn stim 1861 E 174 4? = = = - - = 1 10.12 JOarmcinummmmn oo 1701 a 2 1511 5 2 - - 1 2 we 15-19 yearsmmmmmmmmm—————— 199° 21 % 1 183 - - - oa 1 = 20-24 yearSmmmmmmmmmcee—an 1541 - - - S| 132 - 7% 5 w 1 25-29 2258° ~ = = gt] 175 9 3 22 - 30-34 35225 1 # 1 “ & 8 | 280%] 212 6° g* 35-39 58313 = - wv - - - 6 501 241 7+ 40-44 88224 ® 5 2 2 " = 1 18 719% 468 45-49 years-mmmemmemmemann 1,16322 - - - - 1 = Ll 33 1,008 50-54 yearsSmmmmmmnmnea———— 1,589°¢ - wn - w - - - - 131 452 55.59 5 = = - - = w 3 2 28 60-64 = ~ ~ = - - ” 2 3 65-69 - p- od - - a - 1 4? 70-74 - = ~ - wo = 1 1 - 1 75-79 yearsmmmmmmmmmmmman=n 5,44942 w = - - - - - - % 2 80-84 yearsm—-mmm—aem-- -—-| 4,749%° = - - - - - = = - - 85-89 years—--eammne- ——---| 3,289%° - - - - - - - - - - 90-94 years—--a-- ————————— 1,433 - - - - - - - - - - 95-99 yearsmmmmmmnm a —-— 3544 - " w “ w - = - - = 100 years and OVeTrammmm ama 88% - - ® - - - - » wi 1 Table 5. Cross classification of 5-year age intervals as stated on the death certificate and on the matching 25-per- cent sample census record (stage II), by color and sex: United States, May-August 1960—Con. [Numbers include allocations of ages not reported on the census record; number of allocations is in superscript] \ . Age on death certificate 100 Not 50-54 55-59 60-64 65-69 70-74 75-79 80-84 85-89 90-94 95-99 years | stated years years years years years years years years years years and or not over valid 3,088%¢ | 4,06727 | 5,2715° 6,5357%|7,1807® | 6,3177°| 4,596%°|2,686%%| 1,000°| 155 30° 8| 1 = = 2 = - - - = - - - il 2 # - - = ” - - - - = - -1 3 1 - - = = wo - - - - - =] = 2 4 - ~ w - - - - - =) 3 x 2 2 44 22 1538 = - - - - -| 6 - 4 62 22 we 5 - - = ot a ~y 7 32 4 3 oe 4 55 - - ~ - = -| 8 4 1t gt 62 6 5 - = 5 - ~ -| 9 59° fat 94 6% 144 - - Po " 55 -| 10 735 57 72 102 8 15° - - = = - =i 3 2,757° 138° 7314 12% 168 - 10 10 - - = 212 4 6 8 6 5 C1 3,656 115 88 24 10 10 10 - 5 - -l13 485 140° 4,691% 24728| 15518 2515 10 55 10 - 5 114 74 334 238° 5,805° 232° 1752° 301° 10 - ® - 21s 1% 3 87° 25312 |6 2472 2511° 145 155 5 w ® 1/16 3 32 133 83% 365° 5,596 202 55 201° 5 v 117 2 2 114 9 794 185 4,019 150 40 - ow 118 1 1 1 4 248 20 155 |2,366%° 90 5 =) -| 19 # 1 - 42 w 5 39) 810 5 * | 20 - - - 4 5 5 5 10 135 - -l21 - 1 = - - 5 5 = 10 - 20 -| 22 1,574%2 | 2,055! | 2,971%° 4,073%6|5,0599¢ | 5,43381 | 5,161°5|3,199°° | 1,620*%| 400° | 70 -|23 1 1 3 22 » - = - - - - -| 24 . - = 2 " - - - 0 = -|2s w 2 - - - - 5 - - - - -| 26 » - 1 22 22 - » ” - 5 = «| 27 # 1 - 2 - 5 ” " - - -|28 43 31 - 42 22 - 55 55 - 55 - -|29 43 2 - 4? wo wi 10S - = = | 30 78 3 1 4 42 15° - 10 = - - - | 32 40% 52 4 10% 0° 5 15° 5 - - - «| 32 51% 192 6 4 42 255 - 55 “ - = -|33 1,3352 928 548 14° 16° 5 5 - 55 5 w -| 34 798 1,774* 1067 60° 10 151° w- - - 5 % = 1135 393 1017 2,4961° 20916| 1222° 15% 4515 5 55 £ - -| 36 62 382 2032 3,564° 222’? 901° 501° 10° 10% 52 - - 37 1 2 7 5 4 7 67 2801° (4,172 425 260 31 40 5 - -| 38 32 4? at 822 358° 4,5221° 311° 120 201° 10 - -| 39 1 2 gt 198 9412 225 4,130 1901S 60 205 - - | 40 - - 5 10% 242 85% 275%5(2,726 1561° 10 - - 21 i - 1 3 3 - 40 82 1,279 25 - -|42 - 1 - = 22 6 - 10 Rh 295 - - 143 - - - - vd w 5 - 5 5 70 - laa 39. Table 5. Cross classification of S-year age intervals as stated on the death certificate and on the matching 25-per- cent sample census record (stage II), by color and sex: United States, May-August 1960—Con. [Numbers include allocations of ages not reported on the census record; number of allocations is in superscript) Color, sex, and Age on death certificate #46 on census Fecond Dovel, 14 | 5-9 | 10-14 {15.19 | 20-24 | 25-29 | 30-34 | 35-39 | 40-44 | 45.49 Mol years |years | years years | years | years years | years years years Nonwhite male Total, 1 year and over-- | 4,90072 120 65 70 | 98'| 69% | 84% | 1172 1542 2125 309% 1-4 years 123 117 5 - - - - - 1 - - 5-9 years--- eat 1 59 # - 1 1? 7 - - - 10-14 752 1 - 66 4 2? - " ” 2 15-19 962 - - 4 | gat 2 1 - - = 1 20-24 732 ® ® & 2 56 5 2% " 2h 1 25-29 932 - 1 - 3 5 68 5 i 1 2 30-34 120° 1 - - - 1 6 | 91 9 4 - 35-39 169° - - - 1 1 zx 14 | 2127 Hn? 4 40-44 2157 - w - - = g* 8 160% 192 45.49 2923 = = w - - - 1 2 27 2221 50-54 3997 - - = - ” - 1 1 3 36 55-59 456° - - w ” 1 ” - 2 = 14 60-64 497° - - w w i” - 1 w - 6 65-69 617° - > - - - - 2 # 2 70-74 5851S - - - - 13 - - = - - 75-78 JEOTBmmmmmamwiesms - 464% - w - wo 5 1 ® - * - 80-84 YyEArS-mmmmmmmmmmn—n- 281° - - - - - - - od - - 85-89 yearsmmmmmmemmmeenn - 184 - - - - - - - - - - 90-94 yearsSmmmemmemmmcmean 62 - - - - - - - - - - F5uTY FOUL G mmr wns mmm Ze ” - - 2 - - - " = - 100 years and OVeremmmm= - 14 - - - - - - - = - - Nonwhite female Total, 1 year and over-- | 4,098%% got 41 30 | 43 30 51 | 105 | 175° 211% 306° got 78% 1 “ ~ - - - * 42? - 38 1 % - 1 w at " 1 32 1 1 29 - 1 - - = = - 47° ® 1 -| 40 1 & 2 * - - 35% - - - -] 2 2 1 - 2 2 sit i - - @ 1 = 4 31 1 1 119° = = = - - 7 82 19, 22 2 169 w - 1 = 1 1 | 1m: 18 1 208° - - - » - = 4 12? 142 28! 279° - - wo 1 “ = x 6 28% 202 283° - - - - Rl - - q~ 12k 38 3326 - - - - - 1 - 1 3 18% 419% = a * w - w- ® 1 3 8 E589 YOUTSwmuwmenmmnnnnm - 483° - - - - - hic - - - 3 TOTAL FEBLB mmmwwsimmnmm a 4947 - w » = - - - - " - 75-79 YearSmmmmmmmmmmmmen= 425° - - - 1 - ” - x = - 80-84 27? - - - w - ® = - - - 85-89 1891 - - - wo " 1 - - 1 5 90-94 79% - - - = = = = = - = 95.99 YEArBmuwwwmnmmnmn -— 39% - - “ w - s = = - = 100 years and OVeremmemmemm= 16 - - - - - - - - - - 40 Tabde 5. Cross classification of 5-year age intervals as stated on the death certificate and on the matching 25-per- cent sample census record (stage II), by color and sex: United States, May-August 1960—~Con. [Numbers include allocations of ages not reported on the census record; number of allocations is in superscript] Age on death certificate 100 Not 50-54 55-59 60-64 65-69 70-74 75-79 80-84 85-89 90-94 95-99 |years | stated years years years years years years years years years years and or not over valid 4325 5047 5868 67112 538° 4018 231% 1442 542 24 > 6 [1 = - - - - - . - - N -ifd 1 - 1 - = - - - ~ = - -|3 - - ~; - - - - " - - - - 14 a " w w " - = » » 5 = = |. 2 1 1 i] - - A - - - ~y8 2 1 w 3 ~ - » - ©» = A -~ 17 3 1 1 12 2 - " “ = = * “|8 3 1 1 2 E 2 = - " - = |g 142 3 4 2 i 2 » ~ = - S -|10 18 121 4 2 i 2% = > ie] wo - WE 2953 391 18° 4 4 1? “ » w = > wi {12 54 320% 32! 19 7r 3 i} - = = a «1s 23 64 3441 422 92 4 2 2 - w - = [iad 12 35 1041 405° 371 15 - 5 1 wo w 1 kas 3 11° 39 1082 341° 51% 192 51 1 “ - 2 |16 - 1 24 442 85 256 29 10 4? 2 1 2 az 2 2 112 242 34 41 142 17 4 i) 2 118 1 1 4 9 10 19 33 97 8 2 = 119 . - 3 4 7 4 3 5 32 5 1 - | 20 “ - 1 = 2 = 1 1 = 3 = [12% - - - - - 2 1 1 i 1 10 ~ | 22 3585 3953 53813 489° 415° 3667 233° 1261 622 28t 16 5 |23 : - - ~ s - = 1 - = = J2a s - w - “ = " = - - -|2s - ~ - - - - a - = - 3 - |26 - - 2 a - 11 22 - w - “ - 27 1A x 1 = - w o ~ = w - 128 ig - 1 32 - - » wo - - |29 S 1 2% - 1 i - - 1 w- 3 - [30 Ja gl - 1 - # i 1 = vt vo w [B 1g 5 3 1 = x we - “ - = 1{32 26 6 72 - i 1 = w - ” ~ 133 190 26 6 6 2 g “ “ " - - 13a 72 190? 29 gt 5 22 - - = 2 " -|35 33 8l 2422 27 15? 4 4 a = - [38 9 47 119? 247 352 ei i 3 2 = Ps “ = 137 7 15 792 104 220 352 20 9 4 1? = 38 3 10 262 572 76% 214 30 5 1 1 i 2 |39 2 7 3 mR 37 46 135 20 2 " - 1 |40 1 1 10 11 19 30 26 18* 8 3 4 1 {41 - 1 2 4 1 7 7 10 37 7 2 - | 42 oy oh - 3 2% 2 5 2 6 1s 3 = |4% = 2 = 31 1 2 - 2 2 2 5 - |4a 41 Table 6. Number of matching 25-percent sample census records (stage II) and net difference rates for 10-year age intervals, by cause of death, color, sex, and age: United States, May-August 1960 Cause of death! Cause of death! Gola, Sez, #nd age di Acci All Cardi Acci All Cardio- cci- ardio- cci- in vascular Malige dents, | other mil vascular elise dents, | other diseases ete. causes diseases etc. causes White male Number of matching census records Net difference rate Total, 1 year and over------- 47,972 28,273 8,275 4,203 7,221 — wey vie ces ces 1-4 years------ccouc-o- 406 11 51 142 202 -1.0 -18,2 2,0 -2.1 - 5-14 years------------ 630 31 87 361 151 1.4 -3.2 -2.3 2.5 2.0 15-24 years----------- 835 59 15 591 110 -1.1 -23,7 -20.0 3.9 -2.7 25-34 years------=-un- 949 176 129 490 154 “343 «9,7 -8.5 -1.0 1.3 35-44 years-----=-oua-n 2,324 1,031 315 553 425 -6.0 -9.7 -7.0 -2.0 -1.6 45-54 years--------oo- 5,219 2,821 950 614 834 -1.0 -0.2 -0.7 -3.4 -2.4 55-64 years----------n- 9,438 5,526 1,943 544 1,425 -1.1 -1.8 -0.4 0.6 0.1 65-74 years----------- 13,564 8,458 2,531 481 2,094 1.1 1.7 3.1 -4.0 -2.6 75-84 years------==--- 10,848 14355 1,780 315 1,398 0.6 -0.1 -0.4 6.7 3.9 85 years and over----- 3,759 2,805 414 112 428 3.0 3.4 -2.2 2.7 5.1 White female Total, 1 year and over------- 35,624 21,336 7,074 1,729 5,485 Fa wes win ruin eee 1-4 years--------coooo 333 } 15 43 114 161 -2.7 -26.7 -7.0 1.8 -2.5 5-14 years- 356 26 77 124 129 -3.9 -26.9 -1.3 -5.6 0.8 15-24 years 353 62 43 130 118 -4,2 -25.8 -7.0 4.6 -1.7 25-34 years 577 135 128 135 179 -14.9 -31.9 «13.3 -6.7 -9.5 35-44 years 1,465 376 536 195 358 -7.0 -12,2 -9.5 -0.5 -1.4 45-54 years 2,752 961 1,069 208 514 -0.9 -7.4 1.6 1.4 4.9 55-64 years 5,129 2,493 1,59 220 822 -2.0 -3.4 1.4 -2.3 4.4 65-74 years 9,297 5,731 1,964 227 1,375 -1.8 -1.0 -2.6 -10.1 -2.3 75-84 years 10,198 7,484 1,275 250 1,189 3.9 2.8 71.5 8.0 6.2 85 years and over----- 5,164 4,053 345 126 640 2.4 3:1 -2.6 11.1 -0.6 Nonwhite male Total, 1 year and over------- 4,900 2,490 719 627 1,064 “vin vow "wa “ee cee 1-4 years------ccoeua- 123 2 9 40 72 -2.4 - - -7.5 - 5-14 years------------ 139 4 8 84 43 -2.9 -25.0 12.5 -3.6 -2.3 15-24 years----------- 169 10 12 126 21 -1.2 -50.0 -33.3 7.1 9.5 25-34 years----=--=-o- 213 45 12 108 48 -5.6 -17.8 - -0.9 -6.3 35-44 years----------- 384 158 47 84 95 -4.7 -4.4 -14.9 - -4.2 45-54 years------=---- 691 328 111 87 165 7.2 10.1 9.0 -5.7 7.3 55-64 years-- 953 522 196 47 188 14.4 13.2 9.7 19,1 21.3 65-74 years----------- 1,202 137 200 29 236 0.6 3.8 - 6.9 -9.7 75-84 years=---------a 745 475 105 18 147 -15.2 =15,6 -15.2 -61.1 -8.2 85 years and over----- 281 209 19 4 49 ~14.9 -16.7 -15.8 75.0] -14.3 Nonwhite female Total, 1 year and over------- 4,098 2,369 661 218 850 wiv o cor ve cee cee 1-4 years=-------co--- 80 5 3 31 41 - =20,0 - - 2.4 5-14 years=-------u--- 74 10 4 29 31 4,1 - - 6.9 -16.1 15-24 years--------=-- 82 19 6 24 33 -11.0 -21.1 -16.7 -4,2 -9.1 25-34 years-----=--a-- 170 54 19 27 70 -8.2 -22,2 -10.5 3.7 -1l.4 35-44 years----------- 377 143 79 31 124 2.4 2.8 -5.1 - Zs3 45-54 years------=-u-- 562 281 138 27 116 18.1 26.0 11.6 7.4 9.5 55-64 years----------- 751 468 147 14 122 24,2 25.4 17.0 7.1 30.3 65-74 years--=----=---- 977 656 158 14 149 75 -5.5 -10.8 -14.3 -12,1 75-84 years---------=-- 702 491 83 15 113 -14.7 -13.6 -12.0 -33.3 -18.6 85 years and over----- 323 242 24 6 51 -28.2 -31.4 -29.2 33.3] -19.6 £ lComplete category titles and numbers of the Seventh Revision of the International Lists, 1955, are as ollows: Major cardiovascular-renal diseases (330-334, 400-468, 592-594) Malignant neoplasms (140-205) Accidents, poisonings, and violence (E800-E999) All other causes (residual) 42 Table 7. Actual and adjusted age-specific death rates for 10-year age intervals, by specified causes of death, color, sex, and age: United States, 1960 All causes Cause of death? Cardiovascular Color, sex, and age i diseases Malignancies Accidents, etc. Ad- Actual 1 jugeed Actual Ad- , | Actual Ad- | | Actual Ad- justed justed c justed! White male Rate per 100,000 population 1-4 years=-=-==--==-===== 104.9 106.0 2.2 2,7 13.1 12.8 32.4 33.1 5-14 years-----========== 52.7 52,0 2.5 2.6 8.0 8.2 25.7 25.1 15-24 years 143.7 145.3 8.0 10.5 10.3 12.9 105.2 101.3 25-34 years 163.2 168.8 25.9 28.7 18.8 20.5 88.8 89. 35-44 years 332.6 353.8 128.4 142.2 46.3 49,8 88.2 90.0 45-54 years 932.2 941.6 478.5 479.5 164.1 165.3 111.0 114.9 55-64 years---=--====ceceeee-- 2,225.2 2,249.9 1,263.5] 1,286.7 450.9 452.7 128.7 127.9 65-74 years---=-=-=-==--==c--== 4,848.4 | 4,795.6 2,998.0 | 2,947.9 887.3 860.6 163.1 169.9 75-84 years-------=====iccme=- 10,299.6 | 10,238,.2 7,029.2 7,036.2 | 1,413.7 | 1,419.4 297.1 278.4 85 years and over-----===-===== 21,750.0 | 21,116.5]|| 15,977.8 | 15,452.4 | 1,791.4 | 1,831.7 664.5 647.0 65 years and over------------= 7,137.3 | 7,052.7 4,713.7 4,653.2(1,073.4 1,059.6 223.5 222.2 White female 1-4 years=---=r=mememesecccun 85.2 87.6 1.6 2.2 9.7 10.4 23.5 23.1 5-14 years-----=-==-=-c--=--=- 34,7 36.1 2.1 2.9 6.2 643 10.8 11.4 15-24 years-=--=============e= 54.9 57.3 6.5 8.8 6.5 7.0 23.5 22.5 25-34 years------------mm====- 85.0 99,9 15.3 22.5 18.8 21.7 21.8 23.4 35-44 years-=------mmm=m=-aa-=- 191.1 205.5 45.3 51.6 66.6 73.6 26.0 26.1 45-54 years=---=--==--=c-oo=== 458.8 463.0 150.9 163.0 175.7 172.9 34.2 33.7 55-64 years=-----r==mm===a===- 1,078.9 | 1,100.9 520.4 538.7 329.0 324.5 42.0 43.0 65-74 years-=--==s=m=mmreccecea 2.11931 2,830.2 1,728.9 1,746.4 562.1 577.1 69.2 77.0 75-84 years------r===m-m=ecee- 7,696.6 7,407.7 5,556.3 | 5,405.0 | 939.3 873.8 211.4 195. 85 years and over----------=-- 19,477.7 | 19,021.2 || 14,998.9 | 14,547.9 | 1,304.9 | 1,339.7 683.5 615.2 65 years and over----=------=-- 5,256.7 5,184.1 3,674.4 | 3,616.5 718.4 711.3 149.1 146.5 Nonwhite male 1-4 years=-----====--=-=-=-=-- 207.3 212.4 6.8 6.8 8.2 8.2 61.6 66.6 5-14 years------====-==-=-==-- 15.2 77.4 4.9 6.5 449 4.4 40.1 41.6 15-24 years------=---==--=---= 213.8 216.4 16.6 33.2 9.2 13.8 147.8 138.0 25-34 years=--=--=====-==ce-o=-= 386.4 409.3 65.8 80.0 18.2 18.2 203.7 205.5 35-44 years-------=-m=m-m-ae-=- 129.2 765.2 248.2 259.6 11.7 84.3 196.2 196.2 45-54 years=-====r==mnmu=ucewe 1,551.0 1,446.8 719.1 653.1 233.6 214.3 192.1 203.7 55-64 years------====-===ce===- 3,151.5; 2,754.8 1,738.5 1,535.8 549.8 501.2 175.1 147.0 65-74 years-----====--e-m=e=== 5,664.0 | 5,630.2 3,400.7 | 3,276.2 927.5 927.5 174.0 162.8 75-84 yearS==---==--ce-ccaooa- 8,662.6 | 10,215,3 5,571.1 | 6,600.8 | 1,086.1 | 1,280.8 234.8 603.6 85 years and over-------=-==== 15,238.7 | 17,906.8 || 10,356.8 | 12,433.1| 1,211.7 | 1,439.1 414.5 236.9 65 years and over-----=====-=-= 6,923.9 | 7,413.2 4,314.0 | 4,574.8 982.4 | 1,044.0 202.0 229.0 Nonwhite female 1-4 years=--------m--===a=o=== 174.4 174.4 4.9 6.1 7.0 7.0 53.2 53.2 5-14 years==--=-===-=-=c=----== 53.4 55.7 5.2 542 4.8 4.8 20.9 19.6 15-24 years=-=----============-= 106.1 119.2 17.4 22.1 6.7 8.0 32.4 33.8 25-34 years=-=--===--=-------= 260.0 283.2 65.2 83.8 29.6 33.1 49.5 47.7 35-44 years---------=--==-=--- 547.3 534.5 209.3 203.6 98.3 103.6 52.6 52.6 45-54 years-----mmmemmm==-———— 1,144.9 969.4 572.4 454.3 249.3 223.4 50.3 46.8 55-64 years--------=======-==-= 2,409.7 1,940.2 1,460.5 1,164.7 427.8 365.6 61.0 57.0 65-74 years-------=-=-------== 3,981.4 | 4,304.2 2,681.1 2,837.1 537.6 602.7 82.7 96.5 75-84 years---------mm==-m=a-- 6,708.4 | 7,864.5 4,756.5 | 5,505.2 702.3 798.1 153.4 230.0 85 years and over-----=-====== 12,871.2 | 17,926.5 9,330.6 | 13,601.5 727. 1,027.5 347.7 260.8 65 years and over>--=========w 5,215.2 6,015.2 3,613.0 | 4,148.1 591.0 677.8 116.7 136.2 Ipdjusted age-specific death rate = actual age-specific death rate + (1 + net difference rate). The net difference rates come from stage II data. 2Complete category titles and numbers of the follows: Seventh Revision of the Major cardiovascular-renal diseases (330-334, 400-468, 592-594) Malignant neoplasms (140-205) Accidents, poisonings, and violence (E800-E999) All other causes (residual) International Lists, 1955, are as 43 Table 8. Estimated corrected number of deaths for 5-year age intervals, by color, sex, and age: United States, 1960 White Nonwhite Age Male Female Male Female Number of deaths! Total, 1 year and Over=-=-=--ccmemccccmcccccccccccmmean 812,527 | 609,892 | 98,748 | 79,222 7,486 5,957 2,486 2,027 4,294 3,106 924 806 3,837 2,340 898 499 7,050 2,963 1,289 709 8,270 | 3,142| 1.815| 1,103 7,499 | 4,187] 2,314 | 1,462 9,283 | 6,028] 2.751 | 2.531 14,847 | 8,828] 4.146 | 3.031 22,585 | 13,807] 4,949 | 3.990 34,922 | 18,451] 6.193 | 4.686 50,937 | 24,944| 7.912 | 5.152 55-59 years=m === mcm meee eam 67,032 | 32,797] 8,987 | 6,578 60-64 years===-= == omen eee 87,139 | 47,953| 9,854 7,082 65-69 years===--=-- commerce meme 108,871 | 64,794] 11,542 | 9,238 70-74 years==- == cm come emeeecemcmccmcemmem 116,627 | 89,014 11,734 | 10,120 75-79 yearS===---m cme 109,779 ( 96,5121 9,161 7,715 80-84 years===--== commen 82,107 | 84,505| 6,110 | 5,339 85-89 years==- com-coo meee 49,174 | 66,875] 3,835 | 4,218 90-94 years 16,158 26,048 1,346 1,818 95-99 years 4,016 6,286 344 796 100 years and over 614 1,355 158 322 Estimated corrected number of deaths in age group i = published number of deaths in age group i z total number of published deaths all ages (1+ net difference rate for age group ;) total number of estimated deaths all ages where the net difference rates were derived from a comparison of age statements between the death certificate and the matching 25-percent sample census record (stage II) including the allocations for ages not reported on the census record, and the published number of deaths came from Vital Statistics of the United States, 1960, Vol. II, Part A, page 5-182, table 5-11. 44 TECHNICAL APPENDIX Design of the Study The data used in this report are a byproduct of the study, Social and Economic Differentials in Mortality, United States, 1960, which has as its primary objective the provision of nationwide statistics on mortality differentials by various social and economic characteristics collected in the 1960 census. To this end, for those death certificates selected for the 4-month period May- August 1960, a manual search was made for matching census records in the 1960 files. Of the approximately 535,000 deaths which occurred during the 4-month period, about 340,000 were selected for the search. All nonwhite dece- dents were selected as well as all white decedents under age 65, one-half of the white decedents 65-74 years old, and one-fifth of the white dece- dents 75 years and older. Thus the data presented in this report have been inflated by the following factors: 1 x Sample number of all nonwhite decedents and white decedents under 65 years ofage 2 x Sample number of white decedents 65- 74 years of age 5 x Sample number of white decedents 75 years of age and over If the decedent was found in the 100-percent census enumeration (stage I) and it was indicated there that a 25-percent census sample record (stage II) existed for him, his stage II record was searched since most social and economic characteristics in the 1960 census were collected in stage II. Definitions The demographic characteristics of the dece- dent—sex, color, and geographic region of resi- dence—used in this report are census rather than death certificate designations. Cause of death was taken from the death certificate and grouped into the four following major cause categories: Major International Lists cause -of -death category Numbers Major cardiovascular- renal diseas€Semamenma= 330.334 ,400-468,592-594 Malignant neoplasms emm===- 140-205 Accidents, poisonings, and violenCeemeemmmmmmn= EB00-E999 All other causeSemmmm=m=- Residual codes Unmatched Deaths Approximately 23 percent of the total sampled death certificates were not matched with the 1960 stage 1 records. This rate is slightly reduced to 21 percent when decedents under 1 year of age on the death certificate are eliminated—about 35,000 such records. Many of the infant decedents were born after the April 1960 enumeration, and their death certificates were matched with the mother's census record. Although over 65 percent of the death certificates for infants were matched with the census record, many of these records would be for the mother only, prior to birth of the infant, so that exclusion from this analysis of all decedents under 1 year of age on the death certificate seemed justified. There was considerable variation in the non- match rate for subgroups of the total sample. For example, rates of about 35 percent existed for white decedents aged 15-34 years as com- pared with 20 percent or less for those aged 55 years or over (table I). Other subgroups with 45 Table I. Percent of death certificates matched with census records, by color, sex, and age on death certificate: United States, May-August 1960 White Nonwhite Age Male | Female | Male | Female Total, 1+ years-| 80,7 81.3] 69.6 72.7 1-4 years=---| 76.4 75.41 66.9 66.1 5-14 years--| 80.8 81.7] 71.9 76.6 15-24 years-| 64,8 69.6 | 59.7 60.1 25-34 years-| 66,1 74.9 | 52.0 60.6 35-44 years-| 75.0 80.3] 56.9 69.0 45-54 years-| 79,4 82,7 66.7 71.7 55-64 years-| 81.3 83.2.7 73.0 75.0 65-74 years-| 82,9 82.8 | 75.4 74.5 75+ years---| 81,9 80.5] 75.5 75.2 higher than average nonmatch rates were the nonwhite decedents, decedents of the South and West Regions, and decedents whose deaths were from accidents, poisonings, and violence and from other causes of death (table II). Many factors may be responsible for the large number of unmatched records. One may be residential mobility between the time of enumera- tion and the date of the individual's death. Other factors may be census underenumeration of various subgroups of the population, processing and/or coding errors on either set of records, and reporting errors on either set of records on crucial, identifying pieces of information. Potentially "match bias'' (i.e., the bias intro- duced into the data as a result of using only matched records to analyze age comparability) could be a serious problem, and it is one which cannot be easily quantified. From the compari- Table II. Percent distribution of total, matched, and unmatched decedents, by selected death cer- tificate characteristics: United States, May-August 1960 Death certificate Un- Death certificate Un- characteristic Total Matched matched characteristic Total Matched matched Color Percent distribution Region Percent distribution Total--=------- 100.0 100.0 100.0 Total-===---- 100.0 100.0 100.0 White=====cccecaaaan 89.0 90.3 84.2 Northeast--=------ 28.0 28.9 24.2 Nonwhite-=-===------ 11.0 9.7 15.8 North Central----- 29,7 30.6 26.2 South-----ccceaaa- 28.2 26.7 34.1 Sex West-==-cecccaaaa- 14.1 13.8 15.4 Total--==-=--- 100.0 100.0 100.0 Cause of death! Hig wenn einem 56.8 56.6 57.9 Total--=--=--- 100.0 100.0 100.0 Female=----cccceeu-- 43.2 43.4 42,1 Major cardiovascu- Age lar-renal dis- - eases----330-334, Total===------ 100.0 100.0 100.0 400-468, 592-594 57.8 58.9 53.1 Malignant neo- 1.1 1.0 1.4 plasms----140-205 17.8 18.1 16.7 1.2 1.2 1.2 Accidents, poison- 1.9 1.5 3.2 ings, and vio- 2.2 1.8 38 lence---E800-E999 8.2 7.2 11.8 427 ih] 6.2 All other 9.9 9.7 10.6 causes---residual 16.2 15.7 18.3 55-64 years--------= 17.1 17.3 16.3 65-74 years--------- 25.9 26.6 23.0 75 years and over--- 35.9 36.3 34.2 Not reported-------- 0.0 0.0 0.1 INumbers after tional Lists, 1955. 46 causes of death are category numbers of the Seventh Revision of the Interna- sons made between the distributions of the matched and unmatched decedents by their re- spective death certificate characteristics, it is not possible to infer whether the age correspond- ence for the unmatched decedents, had their census records been found, would have been worse, the same, or better than that for the matched. On the other hand, there are several reasons for assuming that the findings of this study based on matched records probably indicate more age agreement than is actually the case for all records. One major reason for this assumption comes from the match process itself. Age may have been used in the match operation in order to determine if a census record with similar butnot identical spellings of name and/or street consti- tuted a match. Also it is conceivable that two records would be classified as not matched be- cause of large age discrepancies. Moreover, results from an independent follow-back survey indicated some relationship between levels of education and income and the proportion of decedents not found in the census. As part of the study, Social and Economic Dif- ferentials in Mortality, United States, 1960, a sample of about 10,000 death certificates was selected from the original 340,000 certificates, and questionnaires similar in content to the stage II census forms were mailed to the informants on the death certificates. One purpose of this follow-back survey was to provide survey infor- mation for decedents not found in the census.'* According to survey responses for white dece- dents, a somewhat larger proportion of white fe- males (but not white males) with low education and white males (but not white females) with low income levels were not found in the census than the proportion not found with high education and high income.!® If these characteristics of educa- tion and income are positively related to age agreement between records, this is a further indication that the unmatched record would prob- ably have less age agreement than the matched. A rough indication of how different results would have been had census records been found for the unmatched decedents can be arrived at by assuming that these records would have had either perfect or no age agreement with the death certificate. The measures calculated under these assumptions suggest that if there were no agreement, the results presented here are ex- treme overstatements of the level of age agree- ment that actually exists, whereas if there were perfect agreement, these results only slightly understate the amount of agreement. Since the data used to make these two assertions are hypothetical and might confuse the issue, they are not shown. Sources and Limitations of Data Two series of comparisons of age informa- tion are available: (1) deathcertificate age infor - mation cross classified by census age information from the 100-percent enumeration (stage I), and (2) death certificate age information cross clas- sified by census age information from the 25- percent sample (stage II). Series 1—Death Certificate and Stage I Census Record Comparisons of Age Information Stage I data were selected for evaluating age comparability for tworeasons: (1) theyrepresent to a greater extent than stage II data the actual, unedited age information reported on the census record, (2) the estimates derived from them are subject to less sampling error than those from stage Il—e.g., the number of records available for detailed subcategories such as age, sex, color, and cause of death become so small in stage II that sometimes fewer than 10 records are in- volved. For this study stage I data were manually coded, not processed by FOSDIC as was the pro- cedure for the official census publications. The date of birth reported on the stage I census was coded into quarter of year, century, decade, and year and later converted to an "updated age" based on the month of death and the quarter of year of birth to correct for a birthday that might have occurred in the interval between the date of the census and the date of death. A combined code was used for quarter of year and century of birth which resulted in some obvious errors whencon- verted to age because of the confusing construc- tion of the code: a number of records turned up with negative ages or ages implausibly over 100 years, Ultimately, all records with negative ages 47 calculated from the stage I census and all records with ages of 100 and over on either the census or the death record were eliminated from the comparison of response analysis—835 such rec- ords were omitted. In addition to these, other records were excluded—3,033 records where age was not stated on one or the other source since these are essentially out of the realm of measuring comparability in age statement (the great bulk of them had no age information on the census rec- ord); 34 records with processing errors resulting in impossible age codes ("not valid" ages); and an estimated 2,687 records where sex and/or color were not stated on the census record (table III). More record losses were incurred in the data by single years of age where a couple of thousand records did not have year but did have century and decade of birth (e.g., 194-) on the census record. These records were allocated to year 5 and were included with the reported ages inthe data by 5- and 10-year age intervals. Measures of age comparability based on these data are of somewhat limited generality. The smallest contribution to the limitations comes from the sampling error. For most statistics shown in this report, the sampling error is either zero or of negligible size. Except for statistics based on the stage II census match, the sampling error is zero for all measures for nonwhite dece- dents and for all white decedents under age 65. Many of the measures for white decedents over 65 have sampling errors of less than 2 percent of the measure itself, A rough approximation to the sampling error of a proportion or rate, P, shown in the report may be computed from the formula: 0, = v2e where n is the sample size and @=1-P n Stage I— White persons 65 years and over (1) Ages 65-74: Divide the recorded esti- mate in table 1 by 2. (2) Ages 75 and over: Divide the recorded estimate in table 1 by 5. However, more serious limitations in these measures of age comparability do arise from other sources. The largest is probably the poten- tial "match bias" discussed previously. Another, 48 Table III, Actual and inflated number of records ultimately used in Census-Death Certificate Matched Record Study and enumeration of excluded records: United States, May-August 1960 Actual Inflated Status of records count of | number of records records Searched in stage I census- | 340,033 533,743 Ultimately used in study==---=-----=- 232,752 386,438 Exe ludedwmmmimnes 107,281 147,305 Not found in stage I census----- 77,067 112,656 Rejected—im- possible codes----- 483 798 Decedents under 1 year on death certificate---==--- 23,176 23,176 Sex and/or color not stated on CensuS===mmmmeecae= 12,687 4,371 Ages negative or 100 years and over- 835 1,557 Ages not reported or "mot valid"----- 3,033 4,747 1By subtraction from total searched. but perhaps not as serious, source would be the potential ''seasonal bias" introduced by the selec- tion of deaths occurring only in the summer months May- August. Series 2—Death Certificate and Stage II Census Record Comparisons of Age Response The denominator of an actual or published age-specific death rate is based on edited census data. One part of the editing process involves allocation of an age to persons with no age reported on the census record, taking into ac- count other related information for those persons such as sex, color or race, marital status, and relationship to head of household. To evaluate the accuracy of the death rate by the method discussed in the text, estimates of the percentage error in the age-specific death rate come from the stage IT data which were treated in the usual census manner just described, as opposed to the rather unusual process that stage I data were subject to in this study. Stage II data, therefore, do not contain the kind of record losses as stage I data. Records of ages under 1 were omitted but ages 100 and over are included, and there were no cases of not stated age, sex, or color since they were allocated. Of the 64,675 records which were designated on the stage I census as being included in the stage II sample, 62,487 were found. Beyond the initial loss of records not found in stage I, a further loss of less than 4 percent of the records was incurred in the stage IT match operation. As an interesting consequence of examining stage II data, it was found that the census alloca- tion of not stated ages agreed only 20 percent of the time with the 10-year age group containing the age stated on the death certificate. It may be that the allocation procedure is more satisfactory on a group than on an individual basis. Or it may be satisfactory for the population as a whole but not suitable for the population of decedents in this study, particularly for the age groups 25-34 and 35-44 years. On the other hand, it may be that individuals with nonresponses constitute a special group and that information on either record for them is subject to question. Table IV contains figures for the two measures used (per- cent agreement and net difference rates) based on the three types of data available—stage I un- edited data, stage II data with allocations, and stage II data excluding allocations. Although the source of data does affect the degree of age com- parability within each age, sex, and color cate- gory, the relative differences among age, sex, and color groups are not disturbed to any appre- ciable extent. When net difference rates based on stage II data without allocations were used to adjust age- specific death rates, these adjusted rates were not seriously different from those based on the data with allocations. For purposes of com- parison, adjusted age-specific death rates based on the three types of data are presented in table V. For almost all categories of age, sex, and color the choice of data makes little difference in the adjusted rates. The exception to this occurs in the age group 85 and over for nonwhite males and females where adjusted rates differ consid- erably. But while the data from stage Il with allo- cations give the lowest adjusted rate for the non- white males, the same data give ahigher adjusted rate for nonwhite females so that the choice of data does not appear to affect the results in any one direction. The data for Series 2 comparisons are sub- ject to the same limitations in generality de- scribed in stage 1 with the additional disadvantage of being subject to larger sampling errors. That is, beyond the sampling done to select decedents to match with census records, the stage II infor- mation was collected for only 25 percent of the population. The sampling errors of estimates based on stage II census reports may be quite large for certain groups where proportions and frequencies are small. This is true for both white and non- white decedents and particularly for the younger ages. For stage II estimates most of the required sample sizes may be obtained directly from table 8, except for white persons 65 and over. For estimates of white persons 65 and over based on stage II and for estimates not based on stage II census records, the appropriate sample sizes may be obtained as follows: Stage II— White persons 65 years and over (1) Ages 65-74—Divide the recorded esti- mate in table 8 by 2. (2) Ages 75 and over—Divide the recorded estimate in table 8 by 5. Alternative Method for Adjusting Death Rates Using Comparison of Response Results In the text of this report estimates were provided of adjusted age-specific death rates where the net difference rates calculated from the comparison of age response data were applied directly to the published age-specific death rates. In addition, table 8 presented the estimated cor- rected number of deaths in 1960 where the same net difference rates were applied to the published number of deaths. One consequence of using these net difference rates is that the total number of corrected deaths over all ages differed from the total number of deaths reported in official publica- tions and, therefore, the individual figures had to 49 Table IV. Comparison of net difference rates and percent agreements based on three types of data, by color, sex, and 10-year age intervals: United States, May-August 1960 White Nonwhite Stage 1 Stage II census Stage I Stage II census Sex and age er. ages = reported: census ages not reported: ages not ages not reported: reported: excluded | Allocated| Excluded | excluded | Allocated | Excluded Male Net difference rate 1-4 years=em===mmemeececcececacaa= -0.6 -1.0 -0.5 -1.3 =2.4 -2.4 5-14 years====s=m=smcmm-ccmece-enn= 0.2 1.4 0.8 0.4 -2.9 -1.5 15-24 years======-c--mmmmeceoeoaoa= 0.2 ct 1.9 -2.9 «1,2 -1,2 25-34 years--=-===mm-mcemeccocaan -4,0 -3.3 -1.4 -3.2 -5,6 -5.8 35-44 years=-=-==-====-mme-eecccooa- -4,1 -6.0 -5.2 -1.9 -4,7 -4,0 45-54 years-------=----cmmmconno= “leZ -1.0 -0.9 8:2 7.2 7.5 55-64 years=---------cc-emmcoona= 0,2 =1l.1 -0.7 13.6 14.4 14.1 65-74 years-=-----=----cmcmcemeennn -0.3 1.1 0.8 2.1 0.6 0.8 75-84 years----------c-cmmennono- 1.2 0.6 0.2 -15.1 -15.2 -15.5 85 years and over=-=------=-=---- 11.9 3.0 2.5 1.23.3 -14.9 -16.4 Female : 1-4 years--====m==m===mumunanmen -0.9 -2.7 -2.1 -1.7 - = IE LE et «13 -3.9 -4.5 -2.3 =4.1 -2.7 15-24 years----==rm=mmmmmmennnnnew -1.8 -4,2 -2.6 -4,8 -11.0 -6.4 25-34 years-=--==--m=m-ccecmcooaoan -4,9 -14.9 -7.4 =2.4 -8.2 -3.7 35-44 years--=-===-c-mmmcccmeem—n- 4.0 -7.0 =5.5 -1.1 2.4 1.6 45-54 years=-----=-cecemccnaccnnn -1.3 -0.9 -0.5 13.3 18.1 18.4 55-64 years=-=-=----mcccececmooeaon -0.6 -2.0 -1.2 23.6 24,2 23.6 . 65-74 years--=-=----c-cececnccnna- -2.8 -1.8 -1.7 =4.5 -7.5 -7.7 75-84 years-------c--c-ommocnnan- 2.5 3.9 31 -17.2 -14.7 -14.9 85 years and over--------------=- 15.3 2.4 1.3 -21.5 -28.2 -29.1 Male Percent agreement 1-4 years-s==seemsme-e-ceececceea-- 96.5 95.3 95.8 94,5 95.1 95.1 5-14 years=======cee-ccececmemnn=- 95.7 94,6 94.6 92.8 89.9 91.2 15-24 years=---------ccccmcenaann 95.3 91.7 94.6 89.7 87.0 88.5 25-34 years=------ec-cecemecccoonn 91.0 89.6 91.7 83.5 79.8 81.3 35-44 years==-----cemcccncnncann= 91,2 88.4 89.8 81.3 79.7 81.1 45-54 years=-----=--cececcemnnccnn 93.4 91.4 92.2 83.7 82.6 83.4 55-64 years====-------cccencceanonn 93.5 91.1 92.0 81.2 79.7 80.4 65-74 years=--------eeceeennenan- 92,8 92.4 92.9 74.8 74.1 74.9 75-84 years=-=-------ceemcccanonn 93.0 92.2 92.5 64.1 62.8 63.6 85 years and over=--------======= 192.7 93.5 93.8 161.0 64.4 64.3 Female 1-4 years---==m-mmescmecmeeeecne——- 97.2 94,9 95,7 96.3 97.5 97.5 5-14 years-===-=m=mecemccmecaeeana- 94,6 93.0 92.9 93.1 93.2 94.5 15-24 years----=-=-cececcecnnnana 93.6 90.7 92.5 88.7 82.9 87.2 25-34 years=---------emcccccenanan 90.5 81.8 89.2 84.3 716.5 80.7 35-44 years---------mcccmemmenn—- 90,6 86.1 88.2 81.0 80.4 81.1 45-54 years----====--cemccccnncan 91.5 88.6 90.2 S1l.7 81.1 82.3 55-64 years-=---==---cmecmceceeeen- 89.9 87.3 88.7 75.2 72.2 72.6 65-74 years-----=----mmmccccccoan 87.7 86.5 87.6 65.0 62.0 62,6 75-84 years=-----=--eceemcceeecnnn 91.5 90.1 90.7 58.7 60.5 61.3 85 years and over=--------====-== 192,7 91.1 91.6 160, 7 55,7 55.9 50 lRefers to ages 85-99 years. Table V. Comparison of adjusted age-specific death rates for 10-year age intervals based on three types of data used for adjustments, by color, sex, and age: United States, May-August 1960 White Nonwhite Adjusted age-specific death Adjusted age-specific death rate rate Sex and age Apa acoal specific Stage 1 Stage II census specific Stage I Stage II census “death census ages not reported death census ages not reported ratel ages not rate! ages not reported: reported: excluded | Allocated | Excluded excluded Allocated | Excluded Male Rate per 1,000 population 1-4 years=e=====-= 1.0 1.0 1:0 1.0 2,1 2.1 2.2 2.2 5-14 years======-= 0.53 0.5 0.5 0.5 0.8 0.8 0.8 0.8 15-24 1.4 1.4 l.4 1.4 2+1 2.2 2:1 2% 25-34 1.6 1.7 Yd 1.6 3:9 4.0 4.1 4,1 35-44 | 3.4 3.5 3.5 7.3 7.4 7.7 1.6 45-54 9.3 9.4 9.4 9.4 15.5 14.6 14.5 14.4 55-64 22.3 22.3 22.5 22.5 31.5 27.7 27.5 27.6 65-74 48.5 48.6 48.0 48.1 56.6 55.4 56,3 56.2 75-84 103.0 101.8 102.4 102.8 86.6 102.0 102.1 102.5 85+ years 217.3 2213.4 211.2 212,2 152.4 2198.7 179.1 182.3 Female 1-4 years===ww=sr= 0.9 0.9 0.9 0.9 1.7 1.7 1.7 1.7 5-14 years======- 0.3 0.3 0.3 0.3 0.5 0.5 0.5 0.5 15-24 years------ 0.5 0,5 0.5 0.5 1.1 1:2 1.2 1.2 25-34 years----=- 0.9 0.9 1.1 1.0 2.6 247 2.8 2:7 35-44 years------ 1.9 2,0 2.0 2.0 5.9 5.6 5.4 5.4 45-54 years----== 4.6 4,7 4.6 4.6 11.4 10.1 9.7 9.6 55-64 years----=-- 10.8 10.9 11.0 10.9 24,1 19.5 19.4 19.5 65-74 years=----== 27.8 28.6 28.3 28.3 39,8 41.7 43.0 43.1 75-84 years------ 77.0 75.1 74.1 74.7 67.1 a 81.0 78.7 78.8 85+ years------== 194.8 2189.5 190.2 192.3 128.7 163.9 179.2 181.5 INational Center for Health Statistics: Vital Statistics of the United States, 1960, Vol. II, Part A, Public Health Service, Washington, U.S. Government Printing Office, 1963. 2Refers to ages 85-99 years. be adjusted to add to the total. It was mentioned in the text that the adjustment procedure described is highly specific: if deaths to be corrected are by sex, color, age, and cause of death or marital status or some other variable, the net difference rates used to correct these deaths should come from comparison of age response data by age, sex, color, and cause of death or marital status. Since such comparisons of age response are not available by detailed cause of death or by marital status, some provision should be made to adjust those deaths with the data available. Insofar as the deaths to be corrected are similar in age distribution to the deaths in this study group, although age reporting may differ within detailed cause of death or within marital status categories, it is possible to use the method described in the text for approximate adjustment purposes. How- ever, when the age distribution of the deaths to be corrected is very different from that for the deaths in this study group, an alternative method for adjustment may be preferred. One alternative procedure proposed by Dr. Kitagawa'® has the additional advantage (besides being applicable regardless of age distributions) of forcing the total estimated corrected number of deaths to be the same as the official totals. The primary methodological difference in the two adjustment procedures is that whereas the one used in this report relies on the marginal dis- tributions of the comparison of age response results, the alternative procedure uses the entire cross classification of age response data. In both cases the analysis is specific for age, color, and sex. The basic data involved are shown below. 51 Comparison of Response Results Death certificate age Census record age Total 1-4 5-9 “ww k a lily n 4 me A =d = Ix =d Ix =d Total ty 0 1 2%, q, i Xk k n= 1-4 years---=--===-mmmmmmmeoo. x, - 5 » T= 5 5-9 years NINN Sw 2 2 21 X50 * Xok Xon rS=mmm—memmecmcmemmccan- = k yea 2x, c, oy %.. we Xe Xn Il yearS=-==-cmmemeccccmccacan. 2x, =C, mil Xn “en I. cee x, Actual Deaths, United States, 1960 D=(D, D,...D,...D) Estimated Corrected Number of Deaths, United States, 1960 By the method used in this study A d,— ¢ D D =D + [ae (B20 I nis [ir {5203] § - (2) 2 kik)" 3 52 where £2 is the adjustment factor needed to D make the adjusted age-specific deaths add to the published total number of deaths By the alternative method B-: [%, : ou] J j As an example, consider the comparison of age response results shown in table 8 for non- white males presented here in broad age cate- gories. Comparison of Response Results Death certificate age Census rec- ord age 1.14 |15.44 |45.64 | 65+ Total years | years |years | years Totale-n- | 4,900 255 | 734 |1,831 | 2,080 1-14 years--- 262 249 8 4 = 15-44 years-- 766 6 677 65 18 45-64 years--| 1,644 - 42 | 1,498 104 65+ years----| 2,228 - 6 264 | 1,958 Actual Deaths, United States, 1960 Age Total l.14 15-44 45 -64 65+ years | years | years | years Number--- [98,748 |[4,218 16,521 | 36,806 | 41,203 Estimated Corrected Number of Deaths, United States, 1960 Method used in this report Altera Age native Not method corrected orpacien to total 2 Totalememcmcaaa 98,756 98,748 | 98,748 Leld JOULE mum mmmmmn 4,332 14,332 | 24,402 15 hl FEAT Bmummmmmm 17,248 17,247 | 17,000 45.64 years-mmmmmn- 33,048 33,045 | 33,118 65+ YeArSmmmmmmmnn= 44,128 44,124 | 44,228 Method used in this report: 262 98,748 4,332 = 4,21 == i Gln ’ s218 x B38 = HEE 2Alternative method: 4,218 16,521 36,806 = 9 epee p! 4,402 = 249 x —f== + 9 x —4— +4 x TE 41,203 mea +0 x 57560 000 53 s d= elm = I Series 1. Series 2. Series 3. Series 4. Series 10. Series 11. Series 12. evies 13. eries 20. eries 21. eries 22. OUTLINE OF REPORT SERIES FOR VITAL AND HEALTH STATISTICS Public Health Service Publication No. 1000 Programs and collection procedures.—Reports which describe the general programs of the National Center for Health Statistics and its offices and divisions, data collection methods used, 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, 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, based on data collected in a continuing national household interview survey. Data from the Health Examination Survey.—Data from direct examination, testing, and measure- ment of national samples of the 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 Institutional Population Surveys.—Statistics relating to the health characteristics of persons in institutions, and on medical, nursing, and personal care received, based on national samples of establishments providing these services and samples of the residents or patients. Data from the Hospital Discharge Survey.— Statistics relating to discharged patients in short-stay hospitals, based on a sample of patient records in a national sample of hospitals. Data on mortality.— Various statistics on mortality other than as included in annual or monthly reports—special analyses by cause of death, age, and other demographic variables, also geographic and time series analyses. Data on natality, marriage, and divorce. — Various statistics on natality, marriage, and divorce other than as included in annual or monthly reports—special analyses by demographic variables, also geographic and time series analyses, studies of fertility. Data from the National Natality and Mortality Surveys. —Statistics on characteristics of births and deaths not available from the vital records, based on sample surveys stemming from these records, including such topics as mortality by socioeconomic class, medical experience in the last year of life, characteristics of pregnancy, etc. or a list of titles of reports published in these series, write to: Office of Information National Center for Health Statistics U.S. Public Health Service Washington, D.C. 20201 VITALand HEALTH STATISTICS | ~ DATA EVALUATION AND METHODS RESEARCH NATIONAL . . : CENTER Series 2 For HEALTH Number 30 STATISTICS ——— comparison of the Classification of Place of Residence on Death Certificates e 1 [ELH PE ETE CTH United States - May - August 1960 U.S. DEPARTMENT OF Zea) HEALTH, EDUCATION, AND WELFARE [5( (730 \ Public Health Service \%\u\ = a Health Services and Mental Health Administration 3 Zs Public Health Service Publication No. 1000-Series 2-No. 30 For sale by the Superintendent of Documents, U.S. Government Printing Office Washington, D.C., 20402 - Price 60 cents NATIONAL CENTER| Series 2 For HEALTH STATISTICS | Number 30 VITALand HEALTH STATISTICS DATA EVALUATION AND METHODS RESEARCH comparison of the Classification of Place of Residence on Death Certificates and Matching Census Records United States - May - August 1960 Comparison of the classification of residence stated on the death certificate with residence as stated on the matching census rec- ord for selected geographic areas by age, race, and sex. Based ona sample of death certificates for deaths occurring in the United States during May-August 1960, matched to 1960 census records. U.S. DEPARTMENT OF HEALTH, EDUCATION, AND WELFARE Public Health Service Health Services and Mental Health Administration Washington, D.C. January 1969 NATIONAL CENTER FOR HEALTH STATISTICS THEODORE D. WOOLSEY, Director PHILIP S. LAWRENCE, Sc.D., Associate Director OSWALD K. SAGEN, Ph.D., Assistant Director for Health Statistics Development WALT R. SIMMONS, M.A., Assistant Director for Research and Scientific Development ALICE M. WATERHOUSE, M.D., Medical Consultant JAMES E. KELLY, D.D.S., Dental Advisor EDWARD E. MINTY, Executive Officer MARGERY R. CUNNINGHAM, Information Officer DIVISION OF VITAL STATISTICS ROBERT D. GROVE, Ph.D., Director ROBERT A. ISRAEL, M.S., Deputy Director ROBERT A. ISRAEL, M.S., Acting Chief, Mortality Statistics Branch JOHN E. PATTERSON, Chief, Natality Statistics Branch LOREN E. CHANCELLOR, Chief, Registration Methods Branch ALICE M. HETZEL, Chief, Marriage and Divorce Statistics Branch ARNE B. NELSON, M.A., Chief, Vital Records Survey Branch MICHAEL J. ZUGZDA, Acting Chief, Statistical Resources Branch Public Health Service Publication No. 1000-Series 2-No. 30 Library of Congress Catalog Card Number 68-62238 CONTENTS Page Introduction ==-- === mmm mm eee mmm 1 Major Findings ==-=-==-==cm mmm mmm meee 2 Urban Places -===c comm mmm mee meee m em 2 Population-Size GroupS--=---= === =m moomoo ome meme mmm momo mmo 4 Underenumeration in 1960 Census of Population----=-=-------oocoeouou 5 Followback SUrvey-------=--cmmmmm meme meme meme 5 Metropolitan and Nonmetropolitan Counties for Urban and Rural Areas------ 6 Standard Metropolitan Statistical Areas----==-==--c--momommmommmcmmeo Age by Geographic Division----------n-mmmmmmm meme cme 9 Race by Geographic Region-------=---omommmm mmm meee meee 10 References ==---=mmmmmcmmmmmc meme eee mm mmm meme 11 Selected Bibliography -=-=-==-=c=cmmm mmm meme mmm moo 12 Detailed TablesS--=-mmm mmo m meme meee mmm meme 13 Technical AppendiX---=-===- =m mmo mmm meme mmm mmm mmo 56 Design of Study--=-===-====--==c-mm mmm mmm mmmmm moomoo 56 Criteria for Matching-------=----cmmmmmm mmm meen mm mee 57 VAariQnCes ===e= mcm mmm mmm me meme mmm mmm mm mem 57 AGE m mm mmm mmm mem meee meme m mmm ommm mmo 57 Race and Color -=--==o moomoo 58 SEX == mm mm mmm mm eee mmm mmm 58 Urban and Rural Areas----=--c-mcmmmm mmm meme mmm meee 58 Standard Metropolitan Statistical Areas----------=-=-=---=c-m-me------ 59 Metropolitan and Nonmetropolitan Counties--------------o-mommmmmoomm- 59 Geographic Divisions and Regions----------=-m-omoocmooommmmmmm mmo 59 Formula for Adjusting Annual Death Rate-----------=---moccccomoom- 59 Net and Gross Difference RateS---=--mmmmmmmm mmm meme meme mmm 60 iv IN THIS REPORT a comparison of the classification of residence stated onthe death certificate with residence as stated on the matching census record for selected geographic regions by demographic variables is presented, The classification of vesidence information on the death certificates corresponds closely to the residence on the census records for the de- cedents whose records weve matched. An inverse relationship exists between the size of the geographic avea and the degree of difference be- lween census assignments and those by the National Center for Health Statistics (NCHS) with differences approaching zero as the size of the area increases. NCHS somewhat overstates the numbers of deaths for individual urban places and understates those for rural aveas. Towns, townships, boroughs in Connecticut, New Jersey, and Pennsylvania, territorial annexations between 1950 and 1960, and urban fringe areas presented problems to NCHS in properly allocating the residence of the decedents, The poorest age agreement between the census vecords and the death certificates was for the nonwhite group at ages 85-99 years in the geographic divisions comprising the South Region. The average pro- portion of records unmatched was 50 percent higher for the nonwhite group than for the white. Both the differences by race between census and NCHS records and the match status weve most favorable for the white population and the Japanese when compared with other racial and ethnic groups. Differences by sex weve negligible, SYMBOLS Category not applicable-----cceommmceeaao Quantity zero---------cmmmmcmmm meee em Net difference rate not computed if census frequency is the same as NCHS frequency- Net difference rate not computed if either census frequency or NCHS frequency or both are zero---=-=--eemrccmm ee meem ee Quantity more than 0 but less than 0,05----- 0.0 COMPARISON OF THE CLASSIFICATION OF PLACE OF RESIDENCE ON DEATH CERTIFICATES AND MATCHING CENSUS RECORDS Mary A. McCarthy, Division of Vital Statistics INTRODUCTION The chief purpose of the 1960 Comparison Study is to measure differences for selected characteristics as reported on the death certifi- cates and on the questionnaires for those dece- dents enumerated in the Eighteenth Decennial Census of the United States. A secondary objec- tive is to assess the accuracy of the annual death rate on the basis of the results of this study. This comparison study involves a sample of deaths occurring during the 4-month period May- August, 1960. The design of the sample, number of records, procedure for matching, and related information are noted in the Technical Appendix. The present study is a byproduct of the Uni- versity of Chicago Mortality Study, Social and Economic Differentials in Mortality: United States, 1960, which is now providing nationwide statistics on mortality differentials collected in the 1960 census by social and economic charac- teristics.!2 This report evaluates the comparability of classification of residence as reported on the death certificate and on the 1960 census sched- ules. The report primarily evaluates coding dif- ferences between the Bureau of the Census and the National Center for Health Statistics (NCHS). The standard certificate of death (1956 Revision) contained the following item for eliciting resi- dence information: 2. USUAL RESIDENCE (Where deceased lived. If institution: Residence before admission) a. STATE b. COUNTY c. CITY, TOWN, OR LOCATION d. STREET ADDRESS e. IS RESIDENCE INSIDE CITY LIMITS? ves(J nod J. IS RESIDENCE ON A FARM? ves(J wo In order to develop uniform national statistics, NCHS issues standard certificates which serve as models for the States. With respect tothe death certificate, the usual residence is the place where the deceased resided at the time of death. How- ever, in the case of persons residing in long- term institutions, assignment is made to the place where the deceased lived prior to admission, if it is given. In the 1960 Census of Population each indi- vidual enumerated was counted as an inhabitant of his usual place of abode which, for the most part, was the place where the individual lived most of the time. Differences exist between cen- sus and vital statistics in determining the usual residence for certain numerically small groups such as inmates of long-term institutions and residents of foreign countries temporarily living in the United States. The residence data shown in this report were grouped into six major categories: (1) Urban places of 25,000 population or more in 1960 (2) Population-size groups of areas (3) Metropolitan and nonmetropolitan counties, urban and rural areas (4) Standard metropolitan statistical areas (5) Geographic divisions (6) Geographic regions (For definitions of these geographic areas, see the Technical Appendix.) MAJOR FINDINGS The classification of residence information on the death certificates corresponds closely to the place of residence as stated on the census records for the decedents whose records were matched. The proportion of unmatched records was high—23 percent. Considerable improvement in the quality of the residence data has taken place since 1950. This observation is based on a comparison of the results of this study with those for a matched record study involving births for 1950. An inverse relationship exists between the size of the geographic area and the degree of dif- ference found between census assignments and those by NCHS. Differences diminish as the pop- ulation size of the area increases. Compared with census assignments, NCHS somewhat overstates the numbers of deaths for individual urban places and understates those for rural areas. A com- parison of NCHS annual data with figures tabulated by selected State offices of vital statistics shows the same pattern. Towns, townships, and boroughs in Connecti- cut, New Jersey, and Pennsylvania; urban fringe areas; and territorial annexations made between 1950 and 1960 presented the greatest problems to NCHS in properly allocating the residence of decedents. The poorest age agreement between the census records and the death certificates was for the nonwhite group at ages 85-99 years in the geographic divisions comprising the South Region, For three-fifths of the 729 urban places of 25,000 population or more in 1960, either no matched deaths for nonwhite persons were re- ported or fewer than 10 deaths were reported, thereby obscuring the comparison of differences by color. The average proportion of records unmatched was 50 percent higher for nonwhite than for white persons. Both the net difference rates by race and the match status were more favorable for the white and Japanese groups than for other racial and ethnic groups. Differences by sex were negligible, A followback survey involving almost 10,000 dzcedents showed that the distribution of matched and unmatched records by nine population-size groups of geographic areas was not different. However, this general finding cannot be extended to the more detailed residence data used in this report. The findings of this study on the annual death rate are limited by the fact that a sample of deaths is taken only for the summer months, and data are for a single year. Other limiting factors are covered in the Technical Appendix. The re- sults of this study generally indicate that the annual death rate for an individual urban place is slightly overstated. URBAN PLACES A frequency distribution of net difference rates for 729 urban arcas with a population of 25,000 or more in 1960 is shown in table 1. Net difference rates were computed by subtracting the number of deaths matched at the usual place of residence according to census assignments from those by NCHS assignments, dividing by the census figure, and multiplying by 100. A negative rate indicates more assignments by census than by NCHS. The group "matched at usual residence" represents those decedents who died at the place where they had usually resided and had been enumerated in the 1960 Census of Population, Net difference rates based on 100 frequencies or more (according to census designations) are shown separately from those based on fewer than 100 deaths, The reason for this separation is that the variance would be greater for rates based on small frequencies, This latter group has a wider range of rates-—from 0 to 94.9-—compared with the range of 0 to 49.9 for the group in which the rates are based on 100 deaths or more, The median net difference rate for the 729 areas fallsat 3.0 percent, Based on 100 deaths or more, the median is 2.9, and for small frequen- cies, 3.2. The size of the difference varies in- versely with the size of the urban area. For areas of under 100,000 population, differences of understatement by NCHS exceed those of over- statement, As indicated in table 1, most of the differ- ences are positive, This finding is consistent with that of other independent sources used for comparison, In a previous matched-record study involving births for 1950, the results were the same as those of this study--an overstatement of events in the urban areas by NCHS compared with census records.’ Net difference rates for deaths in 1960 and births in 1950 for 221 selected urban areas were reviewed, The basis for the selection of the 221 areas was the number of births included in the 1950 study, namely, 100 events according to either the census record or the birth record. For 66.5 percent of the areas the difference was positive for births and deaths (more assignments by NCHS than by census). The direction of the differences for births and deaths for the balance of the areas was as follows: Percent of Direction of difference total Both negative -----=--===c=--c----- 7.2 One positive and one negative------- 16.3 Zero net difference for either births or deaths ---=-cmmmmcmcmmmme mmm 10.0 The negative differences were not concentrated in any one State but were scattered among 32 of the 42 States and the District of Columbia, shown in table 2. This pattern was also typical of the other types of differences shown above, The magnitude of the difference was substan- tially less for deaths than it was for births (table 2). Two factors contributed to this difference. First, the addition of the residence checkbox item on whether the place of residence was inside or outside the city limits on the standard certifi- cates in 1956 aided in properly allocating the residence of persons living near cities, but out- side the corporate limits, Certificates of births and deaths for the majority of States contain this checkbox item. The second factor involves the mobility of persons using hospitals. There is more likelihood of movement for the utilization of hospitals for births than for deaths. A comparison was made between the annual number of deaths by place of residence as tabu- lated by NCHS and those tabulated by the various State offices of vital statistics. For most regis- tration areas, NCHS figures were larger than the State tabulated data, Three-fifths of the States or registration areas had State assigned residence codes on the microfilm copies of the death certificates re- ceived by NCHS for 1960. It is not known if these codes were assigned on the basis of census tracts, street maps, or if the codes were as- signed only on the basis of the information entered in the residence section of the certificate, NCHS used the State codes in determining its own geo- graphic codes for only four States and for selected local areas and counties in 11 other States. The findings of the 1960 Comparison Study show the same pattern of difference when com- paring NCHS figures with either those of the State or the Bureau of the Census. The primary reasons for the differences between census and NCHS figures are to be found in the enumeration process and vital registra- tion-—the former giving greater support to the accuracy of the census figures, Inthe 1960 Census of Population, geographic locations can be fixed with a high degree of precision by the use of street maps and similiar aids. The situation is very different with respect to the vital record, however. In assigning residence codes, NCHS is, for the most part, dependent on the information entered in the section of the vital record per- taining to usual residence. Due to the great number of vital records which NCHS must code, it is not feasible to verify all addresses by means of a census tract or street guide. Few extreme differences between census and NCHS assignments were observed. Only 14 of the 729 areas had net difference rates of 50 percent or more. With the exception of Boise City, Idaho, the remaining 13 areas had fewer than 100 deaths matched at the usual place of residence as re- ported by either census or NCHS designations, The average number of such records for the combined 14 areas was below 50. Not all of the excess differences can be ex- plained. Generally, the areas with these extreme differences in code assignments between census and NCHS fell into three groups. First, towns, townships, and boroughs were difficult to identify and code for vital statistics purposes. Almost three-fourths of the areas with large net dif- ference rates were towns or townships in Con- necticut, New Jersey, and Pennsylvania, In each of these areas, the numbers of assignments were understated by NCHS when compared with census numbers, For example, the first area listed in Pennsylvania was Bristol Township-—an area of 59,298 population which was classified as ur- ban by special rule. In the same county-- Bucks County--there was also the incorporated area of Bristol borough (population of 12,364) which NCHS also classified as utban, Duplication of place names within a county and double entries of place names on certificates resulted in prob- lems in correctly classifying geographic areas, The second factor relating to extreme dif- ferences may be the number of annexations be- tween the two decennial census periods, For example, Hayward, California, had over a 400- percent increase in population, primarily due to annexations between 1950 and 1960. Frequently, the place names of the annexedareasare reported as the decedent's residence and thus may be im- properly classified. Several of the areas with substantial net differences had large annexations between 1950 and 1960. Finally, unincorporated areas which have mailing addresses and/or place names similar to that of an adjacent urban place present a coding problem. One such example is Boise City, Idaho, which is surrounded by unincorporated areas with place names similar to Boise City such as South Boise, The numbers of deaths matched at the usual place of residence were tabulated by color for the 729 urban places having 25,000 population or more in 1960. Comparisons of net difference rates for the white and nonwhite decedents were limited by the small frequencies for the nonwhite group. Almost one-third of the 729 areas had no matched deaths assigned according to either the census record or death record for the nonwhite group (table 3). Another 4 percent of the areas had no frequency according to either census as- signments or those by NCHS and with only one or two deaths in the alternative cell, Consequently, no net difference rate could be computed. An additional one-third of the areas had frequencies of fewer than 10. Net difference rates for white individuals were lower than those for the nonwhite group for most areas. As the number of matched deaths increased, the net difference rates for the nonwhite group approached those for the white. The urban places with at least 50 nonwhite deaths or more assigned by both census and NCHS were concentrated in the South Region (table 4), In this region the match status for both color groups was least favorable. About one-fourth of the records in the South were not matched and thus obscured the analysis of white-nonwhite dif- ferences, The proportions of records which were un- matched for the nonwhite group were higher than those for the white in 324 of the 729 areas, How- ever, for three-fourths of the 324 areas there were fewer than 20 unmatched deaths, The un- matched proportion was higher for white than nonwhite persons in only 92 of the 729 areas. The balance (43 percent) had no matched and/or un- matched deaths. It can be concluded that no judgment can be made about the white-nonwhite differences for the unmatched group for urban areas of 25,000 popu- lation or more throughout the country because of the small frequencies for nonwhite deaths except in the South Region. POPULATION-SIZE GROUPS Table 5 shows the numbers of deaths matched at the usual place of residence for urban places grouped by population size as of 1960. As indi- cated previously, for the less definitive popula- tion-size groups in table 1, the magnitude of the net difference rates varies inversely with the size of the area. This inverse relationship is more evident from data shown in table 1, than from data shown in table 5. For all urban areas of 100,000 population or over and for rural areas, the net difference rates are positive, indicating more assignments according to the death record than by the census record. For all other urban places, the direction of the difference is negative. The group, urban areas of 2,500-10,000 population, had the largest net difference rate (-9.0) compared with other population-size groups. There are slightly over 3,000 such areas and the problem of properly identifying the resi- dence of decedents of such areas has increased so extensively since 1960 that starting with data- year 1964 NCHS no longer separately identifies such places. The reason for this change was primarily due to the fact that mailing addresses rather than the actual residence of the decedents were often entered on the death record. The proportions of records which were not matched at the usual place of residence were similar for the various population-size groups with the exception of that for rural areas. The proportion for this latter group was 26 percent and the range for the other eight groups was from 20 to 22 percent, Lack of a well defined place of residence on the vital record undoubtedly contributed to the high proportion of unmatched records for rural areas. Data for white and nonwhite individuals are also included in table 5. The net difference rates by population-size groups follow the same pattern for each color group as for the total population. For most of the population-size groups in table 5, the net difference rates for the white group were higher than those for the nonwhite, How- ever, the proportion of the records which were not matched at the usual place of residence was about S50 percent greater for nonwhite than for white persons. For urban places of a combined population of 25,000 or more, the unmatched proportion was 19 percent for the white and 30 for the nonwhite group. Underenumeration in 1960 Census of Population There appears to be an association between the net undercount in the 1960 Census of Popu- lation for geographic areas grouped by population size and the proportion of records unmatched in the 1960 Comparison Study. Various methods were used in estimating the coverage errors in the 1960 census. One method was an estimate of housing unit coverage errors noted in Waksberg's paper.? The data relate only to field enumeration errors and not to errors resulting from FOSDIC Processing (Film Optical Sensing Device for In- put to Computer), The highest net undercounts were reported for urban places of 500,000 popu- lation or more and for areas of under 10,000 population, Likewise the proportion of records unmatched in the 1960 Comparison Study showed somewhat the same distribution--low proportions of unmatched records for groups of urban places of 10,000 to 250,000 population and high propor- tions for areas above and below these population groups. Followback Survey One method of evaluating the character of the unmatched group was to followback by an in- dependent survey on a sample of matched and unmatched decedents. This was done for 9,475 decedents of the 340,033 included in the compar- ison study. A survey questionnaire was mailed to the 9,475 informants whose names were entered on the death certificates, On the basis of the questionnaires returned and after personal inter- view followups, there was over 90-percent re- sponse, The only residence data included in both the survey and the 1960 Comparison Study were those for population-size groups of areas, Table 6 shows the number and percent distribution of deaths by population-size groups of areas included in both the survey and the comparison study. The distributions were almost identical, thereby indi- cating that the survey was representative of the major study. The number of records matched in both the survey and the comparison study are shown in table 7. With the exception of the cate- gory 25,000 to 50,000, the distributions of matched events followed closely those of total events shown in table 6. The large difference for this category was the result of processing errors. It is probable that an additional 300 records in the survey in the group ''25,000 or more, but not further defined" should be properly classified to the size of areas, 25,000-50,000. Table 8 shows data for the unmatched group. The number of decedents for which there was a response in the followback survey is shown separately from that for which a response was not received, Since this latter group included only 221 records, it is not feasible to make a judgment about the differences between this distribution and that representing a response in the survey. The percentage distribution by population- size group of the unmatched decedents for which there was a response in the survey differed only slightly from the matched groups (table 7). There- fore, it is probable that the matched and unmatched decedents were from the same 'population.' To be conclusive, it would be necessary to have a distribution of deaths matched in the survey only and coded according to the responses in the survey. This was not done, however, METROPOLITAN AND NONMETROPOLITAN COUNTIES FOR URBAN AND RURAL AREAS Net difference rates and the proportion of records unmatched for the combined groups of metropolitan and nonmetropolitan counties are presented in table 9, The two groups of counties are further classified by urban and rural areas for color-sex groups. Metropolitan counties had smaller net dif- ference rates and lower proportions of unmatched records than nonmetropolitan counties, For the latter group of counties there were more assign- ments according to the census record, whereas for metropolitan counties there were more assign- ments according to the death record. Consistent with the findings noted in the previous section, NCHS classified more deaths to urban areas than to rural. The rural areas of both types of counties have higher proportions of unmatched records than urban areas, This may be related to the fact that rural areas generally have less definitive mailing addresses than urban areas, Conse- quently, matching the census and death records for individuals who had resided in these areas was not as successful as in urban areas. The difference between census assignments and those by NCHS was greatest for rural areas in metropolitan counties (table 10). Likewise, the proportion of records not matched--27.4--was considerably higher for this geographic compo- nent than for the others shown in the table, The rural section of metropolitan counties is unique for several reasons. First, the rural component of metropolitan counties as defined for vital sta- tistics includes the "urban fringe," that is, places surrounding large metropolitan areas. This defi- nition differs from that used by the Eighteenth Decennial Census insofar as the Bureau of the Census classifies these fringe areas as urban, This difference is necessitated by the fact that census can set up one-time boundaries for un- incorporated areas for its decennial enumerations, whereas NCHS must depend upon established political delineations for a 10-year period. A rapid growth of new fringe areas took place be- tween the two census periods--—1950 and 1960. Such areas frequently had mailing addresses of the adjacent incorporated urban places, and these addresses were reported as the usual place of residence, Consequently, there were difficulties in matching, The second factor to adversely affect the match status was migration, As used here, mi- gration pertains to those persons 5 years of age and over who moved from one county to another between April 1, 1955, and April 1, 1960. The proportion of persons who moved to different counties in this quinquennial period was 21.1 percent for the urban fringe of metropolitan areas and only 14,0 percent for the central cities of metropolitan counties,’ The social and economic characteristics of the rural parts of metropolitan areas are very different from those for rural areas in nonmet- ropolitan counties, From one point of view, the rural section of metropolitan areas is more ur- ban in character than the rural section of non- metropolitan areas, In regard to the net difference rates and the proportions unmatched, the differential between the white and the nonwhite groups is pronounced-— both measures are higher for nonwhite than for white persons. The contrast between the direction of the net difference rates is noteworthy, These rates for rural areas are negative for the white group and positive for the nonwhite, The individual geographic areas which comprise the rural total are not separately identified in coding geographic information on the death certificates, This detail would be necessary to attempt to explain the dif- ferences in the rates for the white and nonwhite groups. Rural data were reviewed for the four geographic regions and the differences observed in the North Central Region contributed most to the high positive net difference rate for the non- white group. Nonwhite Region metropolitan-rural net difference vate Totaleemmemm-amu- 10.3 Northeast-----------= 25.9 North Central -------- 57.6 South----mmmmemmmea = 8.0 -13.6 (Death record is base of rate and minus sign indicates more assignments by the death record than by the census record.) STANDARD METROPOLITAN STATISTICAL AREAS Table 11 shows a comparison of net differ- ence rates for 25 standard metropolitan statistical areas (SMSA's) which were randomly selected, and the urban components of these same SMSA's, Here the urban components are limited to those places of 25,000 population or more in 1960 since deaths for urban places below 25,000 popu- lation were not separately tabulated, All subse- quent references to these components will be limited to urban areas of 25,000 population or more, Data presented in this table show that the net difference rates are, for the most part, lower for the SMSA than for selected urban areas whichare a part of that SMSA. The SMSA is a county or group of contiguous counties and therefore in- cludes urban and rural areas, As noted in table 9, NCHS data generally show an overstatement of events for urban places and an understate- ment for rural areas, As a result, the net difference rate is low-—the result of compensat- ing differences. The SMSA's with the highest proportion of unmatched records were located in the South Region with the West having the next highest proportion, This is not only true for the 25 SMSA's in table 11 but for all 201 SMSA's (MSEA's in New England) and for other residence data in this report, Both the South and the West Regions have a higher proportion of nonwhite persons than either the Northeast or North Central Regions. For this color group the unmatched proportion is less favorable than that for the white group. Another factor is that migration is higher in the West and South; therefore it is likely that the unmatched proportion would be higher in these two regions than in the Northeast or the North Central Regions. The largest of the SMSA's in table 11-- Chicago, Illinois—had 22 urban components of 25,000 population or more. The net difference rates for these areas ranged from O for several components to 91.4 for Oak Lawn, Illinois, but the average was only 5.9 percent and the net dif- ference rate for the SMSA, 0.2 percent. The number of matched deaths for SMSA's was tabulated separately for two groups: Matched at usual place of residence (UPOR) and matched but not at the usual place of residence. This latter group was numerically small and repre- sents those matched at the place of death (POD), The number and percentage breakdown of the number of records by match status are as follows: Number Percent Tolalswwwwemmmme 532,948 100.0 Matched ~--~--==-===eu=- 420,292 78.9 UPOR---=-mcecmem mm 388,754 72,9 POD trimeric sassesamassnsoi 31,538 5.9 Unmatched-----=-=-=o=- 112,656 21.1 The group matched at the place of death (POD) is different insofar as any comparison of residence coding by census and NCHS is not applicable, One factor which accounts for a match at the place of death rather than at the usual place of residence is the difference between the enumeration process and the vital registration system, Persons resid- ing in long-term institutions were enumerated in the census at these institutions prior to their death, but the death record asks for the usual place of residence prior to admission to the in- stitution. Therefore, the two geographic codes assigned are not the same, The net difference rate for the total matched group was 0.3 and for the two subgroups, UPOR and POD, the rates were 0.0 and 4.1 (table 12). The highest rates for the total group were ob- served for the three SMSA's of Columbia, South Carolina; Pueblo, Colorado; and Tuscaloosa, Alabama, The high difference rates for the total group were the result of matching at the place of death. The distribution of net difference rates by type of match is as follows: Total UPOR POD Columbia, S.C ------ -22.3 -1.7 -77.5 Pueblo, Colo-------- -30.3 -2.1 -85.7 Tuscaloosa, Ala----- -37.4 - -92.5 In comparing the net difference rates by color only 57 SMSA's were considered, The areas were those to which 100 or more nonwhite deaths were assigned (table 13), Although the net differ- ence rates for the white group were higher for slightly over half of the 57 areas, the proportion unmatched was considerably higher for the non- white group (table 14). The frequency distribution of the proportion unmatched (using census as base) for the white and nonwhite groups is presented below: Proportion unmatched Number of SMISA's White Nonwhite Total-=meeceeeee 57 57 10.0-19.9 ————————————— 20 1 20.0-29.9 «nnmeemmmmmm 28 26 30.0-39.9 me —————— 9 28 40.0~40 GF mo or wesw in gsi = 2 Those areas for which the net difference rates for the nonwhite group did exceed those for the white were located for the most part in the South Region, where the match status was least favor- able, One purpose of the 1960 Comparison Study was to evaluate the effect of the findings of this study on the annual death rate, The annual death rates for SMSA's in 1960 are published in Vital Statistics of the United States, 1960, Vol, 11, Part B. The method involves the following assumptions, First, the assignment of deaths to any SMSA made by census is presumably the ''correct' figure rather than the NCHS figure for the May- August study period, As indicated previously, the difference between the enumeration process and vital registration gives greater support to the accuracy of the coding of geographic locations by the Bureau of the Census. The second assumption is that the population enumerated as of April 1, 1960, for an SMSA is ''correct."" Measures of misstatement of residence or underenumeration for individual SMSA's as of April 1 have not been published by the Bureau of the Census, The annual rate could be adjusted on the basis of the findings of this study, An upper and lower range of the annual rate could be computed as indicated below, The upper range is derived as follows: (1) Increase or decrease the annual number of deaths by the percent difference between cen- sus and NCHS assignments for the study period for the 'SMSA, If the census figure is lower than the NCHS figure, the annual num- ber of deaths would be reduced; if the NCHS figure is lower, the annual figure would be increased, (2) Compute the upper range of the annual ad- justed rate using the frequency derived in step 1. In computing the upper range of the annual ad- justed rate, all unmatched deaths are assumed to be correctly assigned to the SMSA. In computing the lower range of the annual adjusted rate, all unmatched deaths are assumed to be incorrectly assigned to the SMSA, Therefore, the range of the annual adjusted rate is determined by the un- matched deaths, The lower range of the annual adjusted rate is computed as follows: (1) Follow step 1 for the upper range, (2) Reduce the annual number of deaths by the proportion that unmatched deaths are of total deaths for the study period. The figure for total deaths is the sum of the matched and unmatched deaths, (3) Compute the lower range of the annual ad- justed rate using the frequency derived in step 2, It is probable that the true adjusted rate is closer to the upper range since it is most likely that the majority of unmatched deaths are correctly as- signed. The formulas for computing the adjusted rates and their application for one SMSA are shown in the Technical Appendix. AGE BY GEOGRAPHIC DIVISION Net difference rates vary slightly among the nine geographic divisions for the total population. In five of the divisions, the highest net difference rates—all negative—were reported for the age group 25-34 years. (Negative indicates more as- signments by census than by NCHS to an age group.) For the three divisions comprising the South Region, maximum positive net differences were observed for the interval 55-64 years. The highest net difference rate (-5.5) in the Middle Atlantic division was reportedatages 35-44 years (table 15), Division data discussed here and re- gion data discussed in the following section are according to census designations. These sections differ from those appearing previously insofar as they are a comparison of age and race, re- spectively, rather thana comparison of residence. A detailed analysis of age for the United States has been covered in a report in Vital and Health Statistics, Series 2, No. 29.° There was only a slight difference by sex, but a pronounced difference by color when data were compared by age for the divisions. Net dif- ference rates for nonwhite males and females were considerably higher than those for white males and females. The highest rates were re- ported for nonwhite individuals below 25 years of age. However, the rates are based on extremely small frequencies——generally about 10 deaths— for all divisions except those in the South Region. In addition to net difference rates, the per- cent agreement between census and NCHS assign- ments was also computed, This measure was derived by dividing the number of matched records classified by a given age group by both census and NCHS (the common cell) by the number of matched records classified to that same age group by census. As was noted for net difference rates, extensive variability was observed for the percent agreements for the nine divisions by 10-year age groups. Generally, the poorest agreement oOC- curred at 85-99 years—the age group for which the annual age-specific death rate is maximal, The New England Division was an exception tothe pattern; the percent agreement was poorest at the younger ages rather than the older ages. The observations made for the net difference rates by color and sex are applicable tothe percentagree- ments. Nonwhite differences between census and NCHS assignments were more extreme than those for white decedents and these differences were greatest in the divisions comprising the South Region, With few exceptions, the proportions of rec- ords unmatched for males and females and for white and nonwhite persons were highest for the age groups 15-24 and 25-34 years in the nine geographic divisions. For the most part, the exceptions to this pattern were for the interval 1-4 years of age and the proportions were based on small frequencies. With respect to the characteristics of age and race, there were two phases in the 1960 Comparison Study. The difference between the two phases of the study relates to the two-stage enumeration in the 1960 Census of Population. All matched records in stage 1 represent dece- dents who were in the 100-percent enumeration. Stage II includes those decedents who were matched in stage I and who also were included in the 25-percent sample of households. With re- spect to this study and the impact on vital rates, the important distinction is that in stage II, not stated and not valid codes for the majority of items were allocated. The method of allocation is described in the various publications relating to the 1960 Census of Population found in the bibli- ography. It is data from stage Il which are used in computing selected vital rates. All data by age previously discussed have been derived from stage I. Table 16 shows four series of net difference rates for white males for the South Atlantic Division--stage 1 with and without the "not stated and not valid" codes and stage II with and without the allocations of the "mot stated and not valid" codes, Stage I illus- trates better than stage II the '"true' comparison between data derived from the census record and those from the death record. The fact that the effect of the allocations on most age groups was to increase the disparity between age reporting on the census record and the death record has an impact on the analysis of mortality data, RACE BY GEOGRAPHIC REGION Detailed data for race for the United States and four regions are presented in tables 17 and 18. The net difference rates for each of the re- gions were less than 1 percent for the white group. The direction of the difference wasalways positive, that is, more assignments were made to the white group by NCHS than by census (census record used as base). Two factors may contribute slightly to the direction of these differences. First, the number of "race not stated and not valid" assignments according to census tabula- tions was 3,214, compared with 22 ''race not valid" assignments according to NCHS tabulations. Therefore, the figures for the white and for the nonwhite category by census designations are be- low those by NCHS designations for the United States and each region. Second, the death records for which the race was not stated were assigned to "white" in the initial coding operation in NCHS, The number of "not stated'' entries is not known, but presumably the figure is small, Not only are the net difference rates gener- ally lowest for the white group but the proportions of unmatched deaths are lowest, The latter factor gives greater validity to the net difference rates for the white population. The net difference rates for the Negro race were higher than those for the white and were generally positive; the rates were less than 2 percent for the United States and each region. However, the proportion of records unmatched was approximately 30 percent for the Negro race — almost 50 percent above that for the white. The net difference rates for the Indian group were negative for each region and lowest in the West where the majority of the Indian reserva- tions are located, Self-enumeration of race, which was introduced in the 1960 Census of Population, may contribute to the excess of census records over death records which were coded as Indian. The annual death rate for the Indian group was 8.6 per 1,000 population in the United States for 1960 compared with 13.0 in 1950. The self- enumeration procedure in the 1960 census and 10 changes in classification by census of persons of mixed stock of Indian and other races contrib- uted to a sharp increase in the Indian population between the two census years, This artifact ob- served in the anual death rates is consistent with the findings of this study, With the exception of those for the West, the regional net difference rates for Chinese, Japa- nese, and Filipino are based on such small fre- quencies as to render the rates statistically unreliable, The net difference rate for the Japa- nese in the West was -0.6 which was the lowest of any net difference rate for any racial category for the United States or any of the four regions. The match status for this racial group——18 per- cent-—is superior to that of the other racial categories in the West Region or in the United States as a whole, The net difference rate for the Chinese is 0.9--the same as that for the white group in this region. The proportion unmatched is about the same for the two categories-—22 for the white group and 23 for the Chinese. The net difference rates for the ''other nonwhite" group showed the greatest divergence between census assignments and those by NCHS. Likewise the proportion unmatched is highest—approxi- mately 50 percent (census record used as a base). If the death record is used as a base, the equiva- lent proportion is below 30 percent because the denominator of the proportion would be substan- tially larger. It is difficult to explain the large difference between census assignments and those by NCHS for this group because itis an "all other" category. Differences may exist in interpretation between census and NCHS as to what should be included in this "all other" group. Regional differences for each sex group were examined only for the white and nonwhite cate- gories since the frequencies for the racial groups were very small, Net difference rates for females were lower than those for males for each region except the West, In stage II of the 1960 Comparison Study there were 87,905 matched records. Of this number 1,145 or 1.3 percent had no entry for race (''mot valid" included). After the 1,145 rec- ords were assigned a code for race, the assign- ments were compared with the codes for race on the death records for those 1,145 decedents, There was only 20 percent agreement, The effect on the comparison study was to bring into closer agree- ment the assignments of codes by census and NCHS to the white category. The reverse effect occurred with respect to most of the nonwhite racial groups. A comparison of the net differ- ence rates for stages I and II by race for the United States is shown below. (Census record is used as base and minus sign indicates more as- signments by census than by NCHS.) Race Net difference rates for stages: I In White--------- 0.8 -0.4 Nonwhite------ 1.1 -1.2 Negro ------ 1.1 0.2 Indian ------ -9.4 -11.2 Japanese --- -0.9 7.5 Chinese ---- -4.6 -25.0 Filipino ---- -23.2 -33.3 Other ------ -125.4 21.9 Not stated and not valid---- -99.3 eee If the annual death rates for the two groups, white and nonwhite, are evaluated in terms of the findings noted above, the effects are negligible. However, rates for specified nonwhite racial groups would be affected. For example, the net difference rate for the Japanese according to stage 1 is -0.9 and according to stage II, +7.5. The latter net difference rate (+7.5) represents fewer allocations of not stated" to the Japanese than the findings of this study would warrant. The effect would be to overstate the published annual death rate for the Japanese from 5.1 per 1,000 population to 5.5. REFERENCES Kitagawa, E. M., and Hauser, P. M.: Methods used in a current study of social and economic differentials in mor- tality. Emerging Techniques of Population Research, Pro- ceeding of the 1962 Annual Conference of the Milbank Mem- morial Fund. New York. Milbank Memorial Fund, Mar. 1967. 2Kitagawa, E. M., and Hauser, P. M.: Education and In- come Differentials in Mortality, United States, 1960. Paper presented at the annual meeting of the Fopulation Associa- tion of America, New York, Apr. 29-30, 1966. Revised Mar. 1967. 3National Vital Statistics Division: Matched record comparison of birth certificate and census information, United States, 1950. Vital Statistics—Special Reports, Vol. 47, No. 12. Public Health Service. Washington, D.C. Mar. 1962. tWaksberg, J.: Housing unit coverage errors by type of geographic areas, 1960 census. Memorandum No. 2. Bureau of the Census. Unpublished data. 5U.S. Bureau of the Census: U.S. Census of Popula- tion, 1960. General Social and Economic Characteristics. United States Summary. Final Report PC (1)-1C. Washington. U.S. Government Printing Office, 1962. 6National Center for Health Statistics: Comparability of age on the death certificate and matching census record; United States, May-August 1960. Vital and Health Statis- tics. PHS Pub. No. 1000-Series 2-No. 29. Public Health Service. Washington. U.S. Government Printing Office, June 1968. "National Office of Vital Statistics: The comparability of reports on occupation from vital records and the 1950 census, by D. L. Kaplan, E. Parkhurst, and P. K. Whelpton. Vital Statistics—Special Reports. Vol. 53, No. 1. Public Health Service. Washington, D.C. June 1961. 8U.S. Bureau of the Census: U.S. Census of Popula- tion, 1960, Number of Inhabitants, United States Summary. Final Report PC (1)-1A. Washington. U.S. Government Print- ing Cffice, 1961. 9U.S. Bureau of the Budget: Standard Metropolitan Sta- tistical Areas. Washington. U.S. Government Printing Office, 1961. 10y.s. Bureau of the Census: U.S. Census of Popula- tion, 1960, Selected Area Reports, State Economic Areas. Final Report PC (3)-1A. Washington. U.S. Government Print- ing Office, 1963. 1 SELECTED BIBLIOGRAPHY Arriaga, E.: Index for the Detection of the Underregistra- tion of Rural Deaths Due to the Inaccessibility of the Regis- tration Office. A paper presented at the April, 1966, annual meeting of the Population Association of America. Fellegi, I. P.: Response variance and its estimation. J. Am. Statis. Assoc. 59(308):1016-1041, Dec. 1964. Kitagawa, E. M., and Hauser, P. M.: Educational differen- tials in mortality by cause of death, United States, 1960. Demography , Vol. 5, No. 1, 1968. Marks, E., and Waksberg, J.: Evaluation of coverage in the 1960 Census of Population through case-by-case checking. Proceedings of the Social Statistics Section. American Sta- tistical Association, Los Angeles, Calif., Aug. 15-18, 1966. National Office of Vital Statistics: Coding and punching geographic and personal particulars; births, deaths, and fetal deaths occurring in 1960. Vital Statistics-Instruction Manual, Part II, Section B and Supplements. Public Health Service. Washington. U.S. Government Printing Office, Oct. 1960. National Office of Vital Statistics: Interviewer’s Manual; National Mortality Study, 1960. Washington. U.S. Government Printing Office, Oct. 1960. National Vital Statistics Division: Vital Statistics of the United States, 1960, Vol. 1. Public Health Service. Washing- ton. U.S. Government Printing Office, 1962. National Vital Statistics Division: Vital Statistics of the United States, 1960, Vol. II, Parts A and B. Public Health Service. Washington. U.S. Government Printing Office, 1963. National Center for Health Statistics: Vital Statistics of the United States, 1964, Vol. II, Part A, Section 6. Public Health Service. Washington. U.S. Government Printing Office, 1966. Neter, J., Maynes, E. S., and Ramanathan, R.: The effect of mismatching on the measurement of response errors. J. 4m. Statis. Assoc. 60(312):1005-1027, Dec. 1965. Siegel, J. S., and Zelnik, M.: An evaluation of coverage in the 1960 Census of Population by techniques of demographic analysis and by composite methods. Proceedings of the Social Statistics Section, American Statistical Association, 1966. Taeuber, C., and hansen, M. ii.: Self-enumeration as a census method. Demography 3(1):289-295, 1966. Zelnik, M.: Errors in the 1960 census enumeration of na- tive whites. J.Am.Statis Assoc. 59(306):437-459, June, 1964. U.S., Bureau of the Census: County and City Data Book, 1962, A Statistical Abstract Supplement. Washington. U.S. Government Printing Office, 1962. U.S. Bureau of the Census: Evaluation and Research Pro- gram of the U.S. Censuses of Population and Housing, 1960; Record Check Studies of Population Coverage. Series ER 60, No. 2. Washington, D.C., 1964. U.S. Bureau of the Census: Evaluation and Research Pro- gram of the U.S. Censuses of Population and Housing, 1960; the Employer Record Check. Series ER 60, No. 6. Washington. U.S. Government Printing Office, 1965. U.S. Bureau of the Census: U.S. Censuses of Population and Housing, 1960; Principal Data-Collection Forms and Pro- cedures. Washington, D.C., 1961. U.S. Bureau of the Census: The Post-Enumeration Survey, 1950. Bureau of the Census Technical Paper No. 4. Washing- ton, D.C., 1960. U.S. Bureau of the Census: U.S. Census of Population, 1960, Detailed Characteristics, United States Summary. Final Report PC(1)-1D. Washington. U.S. Government Printing Office, 1963. U.S. Bureau of the ‘Census: U.S. Census of Population, 1960, General Population Characteristics, United States Sum- mary. Final Report PC(1)-1B. Washington. U.S. Government Printing Office, 1961. U.S. Bureau of the Census: U.S. Census of Population, 1960, Selected Area Reports, Standard Metropolitan Statisti- cal Areas. Final Report PC(3)-1D. Washington. U.S. Govern- ment Printing Office, 1963. U.S. Bureau of the Census: U.S. Census of Population, 1960, Subject Reports, Inmates of Institutions. Final Report PC(2)-8A. Washington. U.S. Government Printing Office, 1963. U.S. Bureau of the Census: U.S. Census of Population, 1960, Subject Reports, Mobility for States and State Economic Areas. Final Report PC(2)-2R. Washington. U.S. Government Printing Office, 1963. U.S. Bureau of the Census: U.S. Census of Population, 1960, Subject Reports, Mobility for Metropolitan Areas. Final Report PC(2)-2C. Washington. U.S. Government Printing Office, 1963. O00O0 12 Table 1. 10. 11, 12. DETAILED TABLES Frequency distribution of size of net difference rates based on deaths matched at usual place of residence, for urban areas of 25,000 population or more: United States, May-August 1960 ===== = ccm mmm eee mmmmmmmme em Net difference rates based on events matched at usual place of residence for 221 selected urban areas of 25,000 population or more:United States, selected months, 1950 and 1960=== === om comm mee eee em memmmem oes Distribution of nonwhite deaths matched at usual place of residence for 729 ur- ban areas of 25,000 population or more, by frequency groupings and type of dif- ference: United States, May-AugusSt 1960=- === comme emmmmmmeas Number of deaths matched at usual place of residence and net difference rates, deaths unmatched, and proportion unmatched for 89 selected urban areas of 25,000 population or more, by color: United States, May-August 1960=-==---e-comemoaoooao Number of deaths matched at usual place of residence and net difference rates, by population-size group of area: United States, May-August 1960==-------o-coooao Number and percent distribution of deaths in the 1960 Comparison Study and the followback survey, by population-size group of area according to death record designations: United States, May-August 1960===-==-m-mmmommmmmmmmmmmmemmaoas Number and percent distribution of deaths matched in the 1960 Comparison Study and the followback survey, by population-size group of area according to census record designations: United States, May-AugusSt 1960====-=----ccmommoooomoommaaooo Number and percent distribution of unmatched deaths, for which there were re- sponse and no response in the followback survey, and unmatched deaths in the 1960 Comparison Study, by population-size group of area according to death record designations: United States, May-August 1960===-=- co mmmom mmm Net difference rates and proportion of death records unmatched for urban and rural areas in metropolitan and nonmetropolitan counties, by color and sex: United States, May-AuguSt 1960 -===== mm mmm meee meee ee eee en Number of deaths matched at usual place of residence and not at usual place of residence, according to census record and death record for urban and rural areas in metropolitan and nonmetropolitan counties, by color and sex: United States, May -August 1960 ==== === == mm eee ee eee Comparison of net difference rates and proportion unmatched, based on deaths matched at usual place of residence for 25 selected standard metropolitan sta- tistical areas (SMSA's) and urban area components of those SMSA's: United States, May-August 1960==-= === == cme eee emma Number of matched deaths reporting same standard metropolitan statistical area (SMSA) on both census and death records, on census record only, and on death record only; and net difference rates for total matched, matched at usual place of residence, and not at usual place of residence: United States, May-August Page 15 16 19 20 23 24 24 25 25 26 29 31 Table 13. 14 14. 15, 16. 17. 18. DETAILED TABLES—Con. Number of matched deaths reporting same standard metropolitan statistical area (SMSA) on both census and death records, census record only, and death record only; and net difference rates, by color: United States, May-August 1960-------- Number of unmatched deaths for standard metropolitan statistical areas (SMSA's), by color: United States, May-AuguSt 1960===== m-mec como ccmeeecmme em Net difference rates, by color, sex, and age: United States and each geographic division, May-August 1960-----memccccccec ee eee ccccceeeeneccmm mm meee eee ————— Net difference rates for white males for stage I and stage II, by age: South Atlantic Division, May=-August 1960 ====== === mmm eee Number of matched deaths reporting same race on both census and death records, census record only, and death record only; and net and gross difference rates: United States and each region, May-August 1960=-====-=- coco mm moomoo Number of matched and unmatched deaths and proportion unmatched, by race: United States and each region, May-August 1960--ceceecmmco emcee meee Page 48 52 53 54 55 Table 1. Frequency distribution of size of net difference rates based on deaths matched at usual place of residence, for urban areas of 25,000 population or more: United States, May-August 1960 [Census recor used as a base. Negative indicates more assignments by census than by NCHS] Positive difference Negative difference Areas of: Areas of: Net difference rate Total Total 100,000 50,000 25,000 100,000 50,000 25,000 OF MOTE | 100,000| 50,000 OF MOTE | 100,000 | 50,000 Based on 100 deaths or more Total-=-=---=cmmomcommn 289 I 110 | 110 | 69 83 I 21 45 17 0-0.04=-mm-mmmmmmmm mmm eee em 47 5 22 20 vee vee veo ry w 0.1-4,9=--bmmmmmm meee ee eee 172 91 535 26 68 18 38 12 5.0-9.9-=----mmmmmmm eee 50 12 23 15 10 2 5 3 10.0-14.9-------mmmmmmmm mo 14 2 6 6 1 1 - - 15.0-19,9---=--=-mcmmmmemem oo 3 - 1 2 1 - 1 - 20.0-24,9==-==mmmmmmm eee em 2 - 2 - - - - - 25.0-29.9---mmmmmmm meee ——— - - - - 1 - 1 - 30.0-34,9-=-=--mmmmmeme eee - - - - - - - - 35.0-39,9-===-mmmmmmme em 1 - 1 - 1 - - 1 40.0-44,9-=-=-cmmmmmmm meee - - - - - - - - 45,0-49,9------mmrmmmm mee em - - - - 1 - - 1 All other--------ceececmccea-- - - - - - - - - Based on less than 100 deaths Total-=-=-==c-ccmccnano- 243 - 21 222 114 - 16 98 0-0.04=---mm-mmmmm meme meee 84 - 7 77 . —_ cos ‘ee 0.1-4,9=mmmmmmmmmmm meee 89 - 8 81 61 - 10 51 5.0-9,9---rmmmemmne meee 32 - 3 29 21 - 1 20 10.0-14.9-=-=--cmmmmmmem eee 12 - 1 11 12 - 3 9 15.0-19.9---==-mmmmmmmmm meee 2 - - 3 3 - - 3 20.0-24,9==-c-emmmmmmmmeee 12 - 1 11 3 - - 3 25.0-29,9-==-==-mmmmmmmem ene 4 - - 4 3 - - 3 30,0-34,9=---cccmmmmcm meee 2 - - 2 1 - - 1 35.0-39,9----=--memmmmmnmm no - - - - - - - - 40,0-44,9=-=-mmmemmmm meee em - - - - - - - - 45,0-49,9-=---ccmcmmmmeeee eo 1 - - 1 - - - - 50.0-54,9===---cmmecmnmceee - - - - 2 - - 2 55.0-59,9-=-=-mmmmmmmmm eee 1 - 1 - - - - - 60.0-64,9=-=-==-mcmommemem oo 1 - - 1 2 - - 2 65.0-69,9~-=---=mmmememeeme 1 - - 1 1 - - 1 70.0-74,9=-----mmmmm meee - - - - - - - - 75.0-79,9-===mmmmmrcmce ee - - - - 1 - - 1 80.0-84,9==-=--mcmcommmene - - - - 3 - 2 1 85.0-89,9----=-mmcmmmmmmeeen - - - - 1 - - 1 90.0-94,9===--mmmceme meee 1 - - 1 - - - - 95.0-99.9----mmmmmmmm meee - - - - - - - - 100.0 or more----=====-cece== - - - - - - - - Table 2. Net difference rates based on events matched at usual place of residence for 221 se- lected urban areas of 25,000 population or more: United States, selected months, 1950 and 1960 [census record used as base. Minus (-) sign indicates more assignments by census than by NCHS. Includes only those areas for which at least 100 births or more were assigned according to either the census record or the birth record in 1950) Net difference rate Net difference rate Area Area Deaths, Births, Deaths, Births, 1960 1950 1960 1950 Alabama: Georgia: Birmingham-----==mmemmeu- 2d 5.0 Atlanta---------------~ 2.7 4.0 GAAS AT mm mmm mim mim mm mm mm mon -2.7 -11.7 Augusta---------=-=---- 8.2 34.7 MOL. Li mmm mim mmr sm me -0.6 -4.9 Columbus -===-=========-= - 34.5 MODE GOEL om i im mm 4.2 -7.4 Macon-=----=======--==n- 17.5 53.3 Savannah--=----=-=---o-- 0.9 9.1 Arizona: Phoenix=--=--=-=omnoummuun -5.5 22.9 Illinois: Tucson=--=-=-===mmmcccmme——e 4.6 271.0 Chicago-----=-==-=-==---= 0.5 0.8 Decatur--=----=---=====- 9.3 4.3 Arkansas: East St. Louis--------- 5.6 8.1 Little Rock----=--=-----= 8.1 10.3 Joliet--=-=--n-mmceeo- 4.4 71.1 Peoria--------=--c-mn-nmn 0.7 1.8 California: Rockford------=-==----=- 2.5 2.3 Alameda-------===-c-eocumm- 0.7 1.2 Springfield------==---- 8.0 - Bakersfield-----==---uou-- 14.4 41.1 Berkeley---==---ce-ceueonn 1.4 3.2 Indiana: Burbank-=--=---cememaaaann 0.6 0.8 East Chicago-------=---- - - Compton-=-=-=-======-=ee==- 12.9 15.1 Evansville-------=-===- -0.3 10.6 Fresno-------=--=--=------- 4.1 18.9 Fort Wayne-------=-===-- 1.5 9.6 Glendale-~-=mmmmmmmmm———— - -1.7 Gary----=-=====m-m--—-- 4.3 20.2 Long Beach---=--veoeceaa- 0.5 1.6 Hammond ----=-===-====-=- -1.1 -15.0 Los Angeles-=====------=-- -0.1 3.1 Indianapolis------==--- 247 9.0 Oakland=------=---=ceeuu-- 0.9 3.7 Muncie---------=-c-m--= 13.3 15.9 Pasadena----------------- 1.1 -1.4 South Bend=-------=-=--- 4.9 12.1 Richmond-----===ececeoan- - 9.1 Terre Haute-------=-=---- 1.6 25.7 Riverside-=----=-ccceeaan- 11.0 13.3 Sacramento-=-=-==========-= 8.4 3.9 Iowa: San Bernardino----------- 1.6 38.8 Cedar Rapids----------- 1.1 4.1 San Diego------=---------- -1.2 2.5 Davenport---------===-=-- -3.2 -4.9 San Francisco------------ - 1.8 Des Moines----==-=====-- 2.2 17.4 San Jose----~---=-------= 6.8 35.17 Dubuque--=--=--=-==-===== - 4.4 Santa Monica 2.6 0.9 Sioux City----=======--- -0.5 -5.9 Stockton=-----=--=-=---m-=-- 5.3 11.4 Waterloo------===---==- 3.4 6.6 Colorado: Kansas: Colorado Springs----=----- 7.6 28,2 Kansas City----===-=--- 11:5 10.3 Denver-------=----ceeemu- -0.2 0.7 Topeka---=--=-==-==-=-=-= 2.8 17.1 Pueblo-----=---=-=--=-omoun-nn Z.0 7.9 Wichita-------======-un= 6.8 23.3 Connecticut: Kentucky: Bridgeport--------------= -1.0 1.6 Covington--=--=-=======- -1.0 6.1 Hartford---------oeeeeana- -5.9 1.4 Louisville-------=====-= 12.1 12.7 New Britain--------==-=--- 1.4 0.9 New Haven-----=--=------= -0.5 -0.5 Louisiana: Stamford-----=-===c------- -1.0 -6.8 Baton Rouge------------ 4.0 0.4 Waterbury------=-=-=----- 0.3 4.1 New Orleans----=-=-==-=---- 0.7 1.9 Shreveport----=----===-=-=- 0.7 11.2 Delaware: Wilmington-====mmmmmm———- 3.3 7.4 || Maine: Portland--------=---=--== -0.4 -3.9 District of Columbia: Washington-------==-====--= -0.7 7.4 || Maryland: Baltimore------======-= 0.6 1.1 Florida: Jacksonville-==--=====cuu- 7.4 25.7 || Massachusetts: Miami----==--c=cccceoonn- 6.0 15.2 Boston----=--=---==-==== 0.1 1.1 Pensacola----=------=u=-=-- -3.5 41.1 Brockton-------==---==-- -1.7 -5.7 Tampa---=--===-=========--= 3.4 63.1 Cambridge-----===-=-=---- 0.7 1.1 Table 2. Net difference rates based on events matched at usual place of residence for 221 se- lected urban areas of 25,000 population or more: United States, selected months, 1950 and 1960—Con. [census record used as base. Minus (-) sign indicates more assignments by census than by NCHS. Includes only those areas for which at least 100 births or more were assigned according to either the census record or the birth record in 1950 Net difference rate Net difference rate Area Aren Deaths, Births, Deaths, Births, 1960 1950 1960 1950 Massachusetts—Con. New Mexico: Fall River---==-=======-- - 3.6 Albuquerque-----=------- 6.0 19.0 Lawrence-----==-========== -0.5 2.3 Lowell-=-=-ccmeommmmmmmm— - 3.0 || New York: Lynn---=====mm=ceeeaao-—— -0.7 -0.5 Albany-----=-------====- 1.9 6.9 Malden-------==-==-----== 0.6 2.0 Binghamton------------- 1.2 -3.6 New Bedford-------------- 1.3 2.5 Buffalo------=-=-=-=------ 0.5 -0.2 Newton----=-=-=-=======c-u-- -6.5 4.9 Mount Vernon----------- -1.1 -2.0 Quincy=-=-==========c-u--= -0.9 2:1 New York City---------- 0.2 0.5 Somerville-----------=--- -0.5 -2.2 Niagara Falls---------- 1.2 3.0 Springfield-------=------ - 4.4 Rochester----=--==-=------ 1.4 2.4 Worcester-------=---=---- 0,2 2.2 Schenectady----=-=-=------ 5.8 10.8 Syracuse----==-=-=-=------- 1.1 0.3 Michigan: Troy------============= 1.0 14.8 Bay City=-=-=========---=-- 3.6 44.2 Utica--=-=---==========- -2.4 - Dearborn-----==-====c-=-- 9.4 24.0 Yonkers-----=====-=----- -2.6 -5.2 Detroit----=---====-=--m= 0.4 1.5 Flint--------=--emcemcomn 0.9 6.1 North Carolina: Grand Rapids=-----------= 3.0 11.4 Charlotte---=--===-===---- -0.6 6.5 Jackson------==-=-==------ 7.5 35.4 Durham-----==--==-===---- 0.6 21.7 Kalamazoo----=--=-======-- 3-5 13.7 Fayetteville-----=------ 20.9 9.6 Lansing------==========-- 2.0 8.0 Greensboro-----=---=----- 1.1 32.7 Muskegon=--=--=======-=--= 3.3 26.9 Raleigh-------==------- - 21.2 Pontiac----=--=======m=--- 0.6 6.9 Winston-Salem=-=-=-===---- 3.4 15.5 Royal Oak=-=-==========-=-- - 22.2 Saginaw---=-=-==-===------ 1.9 1.0 Ohio: Akron-----===-===------- 17 3.5 Minnesota: Canton-----===========- 3+3 25.7 Duluth--==-=cceemcmemem m= 2.1 - Cincinnati-------====== 0.4 0.4 Minneapolis=-------=----- 1.1 2:2 Cleveland-----========- 1.7 l.1l St. Paul-e-wmemmmeennn——- 0.2 3.6 Columbus--============- -0.1 -2.5 Dayton-----======m====- 2:1 46.1 Mississippi: Lakewood--=======-=-=---- -0.6 -3.9 Jackson------====c=---mm- 2.5 - Lima-----=-=======-=--- 3.3 32.0 Lorain--=--========-=--- -0.8 3:0 Missouri: Springfield------------ 5.1 14.1 Kansas City----=--===-==--= 0.2 -1.7 Toledo-----============ 0.3 16,2 Springfield-----==-==-umuo- 1.3 2.9 Warren----------------- 5.1 13.8 St. Joseph-=-==-=mcceco-- -0.5 3.1 Youngstown=---=--=--=-=-- 0.2 15.0 St. Louis----=--====---m= 0.6 1.5 Oklahoma: Nebraska: Oklahoma City-----=-=---- 5.2 3.8 Lincoln=--====--=c-=-o--= 3.0 5.6 TU LEG wm me mim mm rio mi mn fe 8.7 7.3 Omaha--------=-=ccececoo--= 4.1 5.1 Oregon: New Hampshire: Eugene-----==-=====------ -8.6 62.7 Manchester------==------= - 3.0 Portland--------=------- 5.0 30.5 New Jersey: Pennsylvania: Bayonne------==-=======-=- - 0.8 Allentown----==-=-=-=-=--- 1.5 1.9 Camden-----=-=-=====--===== 1.6 - Altoona-=-=-=-=-====-==------ 2.4 5.3 Clifton----=---==---=m---= 0.6 -4.8 Bethlehem-------=------ 4.0 6.7 East Orange-------------= -1.3 -0.9 Chester----===========~ 5.1 17.2 Elizabeth-====--oo-oueuon- 1.1 2.4 Eriemmmmmmemmm ieee 1.0 8.7 Jersey City-------------- -0.8 2.6 Harrisburg-----=--=------ 12.0 13.0 NEAT wei mm sm i i 0 0.9 -0.2 Johnstown-----=-==-=-===- 36.9 42.6 Paterson-----=-===--=--=--- 0.8 -0.5 Lancaster------==------- 21.0 13.2 Trenton------====-====-===-- 3.3 4.3 Philadelphia----=-=-=--- 0.3 1.8 17 Table 2. Net difference rates based on events matched at usual place of residence for 221 se- lected urban areas of 25,000 population or more:United States, selected months, 1950 and 1960~Con. [census record used as base. Minus (-) sign indicates more assignments by census than by (WCHu. Includes only those areas for which at least 100 births or more were assigned according to either the census record or the birth record in 1950 Net difference rate Net difference rate Area Area Deaths, Births, Deaths, Births, 1960 1950 1960 1950 Pennsylvania—Con. Texas=—Con. Pittsburgh---------ccoo-- 5.0 30.5 Houston=-==-cceccomaaaa- -12,3 14.2 Reading--------ccccocomoo 1.0 10.3 Laredo------=-coocuooo- » 3+9 Scranton----=-==--c--oooo- 1.9 1.6 Lubbock===occmaeoo = 4.7 0.6 Upper Darby-------------- -3.4 -16.9 Port Arthur------------ 0.7 9.7 Wilkes-Barre------------- - 23.7 San Antonio----------_- 1.2 8.7 YOrk----mmmmm meee 4.0 26.7 San Angelo-----c---ooo- 2.0 1.8 Waco-----ommmee eee -1.7 2:5 Rhode Island: Wichita Falls--------o- 2.7 10.5 Pawtucket--=--=-ec-ueooooo -0.8 7.3 | Utah: Providence--=--=---------- -0.7 1.1 Ogden-----ccocmoooooo 5:3 5.5 Salt Lake City------=--- 3.7 6.6 South Carolina: Charleston=-----cccaeaaao 2.0 11.9 Virginia: Columbia-=-----cccmmcaaa -1.0 45.2 Alexandria----------_-- 1.5 13.2 Greenville---=-=---couono 4.2 7.7 Arlington County------- - 0.3 Norfolk--=---ccccoaooo 1.6 5.1 South Dakota: Portsmouth--------o-o-- 0.5 2.0 Sioux Falls--------c-eceoo- 6.7 6.9 Richmond----ceceeeo-_ 0.3 6.9 Roanoke----=--ccouooo- 3.9 0.6 Tennessee: Washington: Chattanooga------=----=-- 3:2 27.7 Seattle----=-cceceono__ 3.7 43.9 Knoxville--=-----cmcoooan 0.4 5.0 Spokane--------ccooo-_ 1.6 9.7 Memphig-=--=--ceuoooeooo oo 1.8 =7.9 Tacoma-----=---occoooo- 1.2 10.3 Nashville-------oucuooooo 3.0 58.1 Yakima--------ccooooo- 17.6 33.0 . West Virginia: Texas: ADEE LL Lp = mn wm wn -0.5 7.4 sm Semen s 2.6 38.0 untington-------ceo--- 3.5 9.7 Austin----=--coccoomoooo -l.4 7.8 Whee 1ingm m= mm comme m 36 4.9 Beaumont-=--=-------oooo_- 1.6 1.9 & - TT : : Corpus Christi--------=-= 0.4 1.4 || Wisconsin: Dallas==-===ccocccmmmanao 0.8 5.2 Green Bay--------oc-o-o- 6.8 El Paso---=----ccecmcoooao 1.7 6.0 Madison-------cocooooo- i 9 -0.5 Fort Worth------ceceeono- 3.4 14.6 Milwaukee------------_- 0.5 1.1 Galveston==-=------coccanon 0.5 23,2 Racine------ccmoooo- 0.5 11.5 NOTE: May-August 1960 used for deaths and January-March 1950 used for births. Source for births: National Vital Statistics Division: tificate and census information: No. 12. Public Health Service. 18 United States, 1950. Vital Statistics— Washington, D.C., Mar. 1962. Matched record comparison of birth cer- Special Reports, Vol. 47, Table 3. Distribution of nonwhite deaths matched at usual place of residence for 729 urban areas of 25,000 population or more, by frequency groupings and type of difference: United States, May- August 1960 Number Direction of difference of urban areas All AreaS=--=--==ceem meme meme eee eee meme meee oo-o—so—----——-—- 729 No deaths assigned according to census record or death record--------------------------o- 201 Frequency of 1 or 2 according to either census record or death record and zero frequency in other----=---eceecccmccccccccmee cee cme meee meee e eee ee nme em em memes mee—e—o————-—- 26 Frequency of less than 10 deaths according to census record-------------=--=-=------------- 239 Net difference rate is Zero=----------c-cmmemmmmmmm cme me ones seosmmoooo—oo--oooo- 150 Net difference rate for white is higher than for nonwhite----------------=co-oooooooooo 5 Net difference rate for white is lower than for nonwhite----------------c-cmcmooononono 84 Frequency of 10 deaths or more according to census records-------=-=--=--=----------------=- 1263 Net difference rate is zZero=--=----------mcmcemcmmmcmmcmeeo nooo s-——oeeom——ooooo- 58 Net difference rate for white is higher than for nonwhite--------------c----eoooooooooo 79 Net difference rate for white is lower than for nonwhite-----------=---------ooooooonon 125 includes one area for which the net difference rate was the same for white and nonwhite. Table 4. Number of deaths May-~-August 1960 matched at usual place of residence and net difference rates, deaths unmatched, and proportion unmatched for 89 selected urban areas of 25,000 population or more, by color: United States, [Census record used as a base for net difference ratc anc proportion uiiuatched. Minus (-) sign inaicates more assignments by census than by NCHS. Includes only those areas for which at least 50 nonwhite deaths were assigned according to either the census record or death record - —— Deaths matched at usual residence ccording tot Net Unmatched Proportion difference rate deaths unmatched Area Census record Death record White | Nonwhite | White | Nonwhite White | Nonwhite | White | Nonwhite | White | Nonwhite Alabama: Bessemer-------=-w-n 23 42 29 51 26.1 21.4 12 26 34,3 38.2 Birmingham---=------- 450 370 466 380 3.6 2.7 169 136 27.43 26.9 Mobile--=---=cccooon 222 120 227 117 2.3 -2.5 72 45 24.5 27.3 Montgomery=-=-------=-- 110 127 123 127 11,8 - 63 55 36.4 30.2 Arkansas: Little Rock----=----- 171 74 189 77 10.5 4.1 51 37 23.0 33.3 California: Los Angeles--------- 4,975 627 | 5,021 618 0.9 -1.4 | 1,362 305 21.5 32.7 Oakland------------- 858 150 877 146 2.2 2.7 257 65 23.0 30.2 San Diego------=-=-=--- 884 56 886 54 0.2 -3.6 323 24 26,8 30.0 San Francisco------- 2,012 17Z | 2,035 169 Axl -1.7 514 69 20.3 28.6 Colorado: Denver-------------- 1,183 56 | 1,193 53 0.2 ~0.8 230 24 16.3 30.0 Delaware: Wilmington---------- 254 51 262 54 3.1 5.9 51 33 16,7 39.3 District of Columbia: Washington---------- 1,070 854 | 1,064 863 -0.6 1.1 260 358 19.5 29.5 Florida: Jacksonville-------- 251 205 278 214 10.8 4.4 100 83 28.5 28.8 Miami---=--=-cccoeeo- 639 103 687 108 74:3 4.9 293 64 31.4 38.3 St. Petersburg------ 720 66 751 65 4.3 -1.5 135 25 15.8 27.5 Tampa--=-=-=---==-=-=-=- 584 138 614 140 5.1 1.4 152 56 20.7 28.9 Georgia: Albany---------couoo 56 50 58 54 3.6 8.0 16 22 22,2 30.6 Atlanta---- 691 428 724 438 4,8 2.3 227 174 24.7 28.9 Augusta---- 67 84 81 90 20.9 7.1 41 34 38.0 28.8 Columbus --- 147 59 146 60 -0.7 1.7 73 26 33.2 30.6 111 88 132 103 18.9 17.0 26 33 19.0 27.3 169 152 163 163 -3.6 7.2 47 41 21.8 21.2 17 43 19 56 11.8 30.2 3 29 15.0 40,3 103 231 106 294 2.9 27.3 44 72 29.9 23.8 Illinois: Chicago---===-=-=---- 8,081 1,643 | 8,199 1,682 1.5 2.4 [2,061 849 20.3 34,1 Fast St. Louis------ 133 92 149 95 12.0 3.3 41 36 23.6 28.1 Indiana: Gary--=====cecooononoo 262 132 276 137 5.3 3.8 46 56 14.9 29.8 Indianapolis-------- 970 239 | 1,017 235 4.8 -1.7 259 98 21.1 29.1 Kansas: Kansas City--------- 237 78 272 86 14.8 10.3 67 22 22.0 22.0 Kentucky: Lexington----------- 142 54 182 61 28.2 13.0 36 24 20.2 30.8 Louisville--=----=-- 806 225 933 237 15.8 5.3 281 111 25.9 33.0 Louisiana: Alexandria---------- 46 52 51 51 10.9 -1.9 31 25 40.3 32.5 Baton Rouge--------- 169 108 1727 111 4.7 2.8 45 38 21.0 26.0 Monroe---==-=------- 46 69 48 75 4.3 8.7 16 28 25.8 28.9 New Orleans--------- 1,111 609 | 1,128 618 1.5 1.5 231 213 17.2 25.9 Shreveport---------- 234 171 231 179 -1.3 4.7 83 72 26.2 29.6 Maryland: Baltimore----------- 1,744 679 11,773 687 1.7 1.2 469 329 21.2 32.6 20 Table 4. Number of deaths matched at usual place of residence and net difference rates, deaths unmatched, and proportion unmatched for 89 selected urban areas of 25,000 population or more by color: United States, May-August 1960--Con. [Census record used as a base for net difference rate and proportion unmatched. Minus (-) sign indicates more assignments by census than t.y NCHS. Incluces only those areas for which at least 50 nonwhite deaths were assigned according to either the census record or death recorc Deaths matched at usual residence according to: Net Unmatched Proportion difference rate deaths unmatched Area Census record Death record White | Nonwhite | White | Nonwhite White | Nonwhite | White | Nonwhite | White | Nonwhite Massachusetts: Boston---=========== 1,927 134 | 1,946 129 1.0 =3,7 457 71 19.2 34.6 Michigan: Detroit----==-======-- 3,344 885 3,383 898 1.2 j EP 663 341 16.5 27.8 Flint-------=======- 380 73 387 71 1.8 -2.7 68 16 15.2 18.0 Mississippi: Greenville---==----- 32 62 34 60 6.3 -3.2 12 13 27.3 17.3 Jackson--==========-= 134 105 139 106 3.7 1,0 54 32 28.7 23.4 Vicksburg----=------- 31 67 31 69 - 3.0 12 14) 27.9 17.3 Missouri: , Kansas City--------- 1,139 213 | 1,146 216 0.6 1.4 242 85 17.5 28.5 St. Louis--=====-=-- 1,742 5141 1,771 513 Lyd -0.2 471 242 21.3 32.0 Nebraska: Omaha--====--=-=-==--- 584 65 613 66 5.0 1.5 151 33 20.5 33.7 New Jersey: Atlantic City--==--~ 165 53 160 58 -3. 9.4 | 51 24 23.6 31.2 Jersey City----=-=---- 767 69 761 74 -0.8 7:2 153 44 16.6 38.9 Newark------=-==-==== 740 266 761 264 2.8 -0.8 239 160 24.4 37.6 New York: New York City------- 18,703 2,074 (18,813 2,132 0.6 2.8] 4,350 1,252 18.9 37.6 Buffalo-----==-==-=--=-- 1,520 129} 1,559 116 2.6 -10.1 225 36 12.9 21.8 North Carolina: Charlotte----===-=-=-=-- 208 139 206 139 -1.0 - 73 40 26.0 22.3 Durham----- 120 55 123 54 2.5 -1.8 28 31 18.9 36.0 Greensboro- 114 63 117 62 2.6 -1.6 45 24 28.3 27.6 Raleigh---- 93 60 94 59 1:1 -1.7 29 11 23.8 15.3 Wilmington---=--=---- 1 60 73 65 2.8 8.3 12 13 14.5 17.8 Winston-Salem------- 122 112 128 115 4.9 2.7 25 46 17.0 29.1 Ohio: Akron-------=-=--===-= 602 58 616 56 2.3 -3.4 127 29 17.4 33.3 Cincinnati-----=-=---- 1,082 247 1,091 254 0.8 2.8 232 101 17.7 29.0 Cleveland-----=-=----- 1,934 4921 1,995 489 3:2 -0.6 339 184 14.9 27.2 Columbus---=-=-====-=-~ 836 167 835 173 -0.1 3.6 179 46 17.6 21.6 Dayton----=-=-=====-=== 529 127 549 126 3.8 -0.8 133 44 20.1 25.7 Toledo---=-=======-=-= 783 81 791 83 1.0 2.5 186 22 19.2 21.4 Youngstown---=-==-=--- 416 53 418 52 0.5 -1.9 55 18 11.7 25.4 Oklahoma: Oklahoma City=------- 546 81 581 83 6.4 2.5 201 44 26.9 35.2 Tulsa---============ 447 65 493 70 10. 2:7 131 31 22.7 32.3 Pennsylvania: Philadelphia-------- 4,464 1,174] 4,529 1,162 1.5 -1.0 +235 518 21.7 30.6 Pittsburgh-------=--- 1,551 264 | 1,639 281 5.7 6.4 410 88 20.9 25,0 South Carolina: Charleston------==-=-- 79 71 84 70 6.3 -1.4 26 20 24.8 22.0 Columbia----==-==--- 132 59 132 62 - 5.1 20 40 13.2 40.4 Greenville---------- 65 52 74 50 13,8 -3.8 36 26 35.6 33.3 Tennessee: Chattanooga---=--=---=- 200 114 212 113 6.0 -0.9 85 65 29.8 36.3 Knoxville--=-=-=====-= 215 57 221 53 2.8 -7.0 95 22 30.6 27.8 Memphis--======----- 594 437 617 443 3.9 1.4 135 173 18.5 28.4 Nashville------==--- 279 190 294 171 5.4 0.5 78 65 21.8 25.5 Texas: Beaumont -=--=-=====-=- 163 75 172 76 5.5 1.3 45 21 21.6 21.9 Dallas--=-==-==--- 928 251 945 247 1.58 -1.6 32) 137 25,7 35.3 Fort Worth-- 569 130 598 131 5.1 0.8 211 58 27.1 30.9 ‘Galveston--- 144 55 147 56 2.1 1.8 34 20 19.1 26.7 Houston=----- 1,152 3741 1,017 334 | -11.7 -10.7 348 161 23.2 30.1 San Antonio- 925 98 940 99 1.6 1.0 286 42 23.6 30.0 Waco-----==========- 170 59 177 54 4,1 -8.5 62 18 26,7 23.4 21 Table 4. Number of deaths May-August 1960--Con. matched at usual place of residence and proportion unmatched for 89 selected urban areas of 25,000 and net difference rates, deaths unmatched, population or more by color: United States, [census record used as a base for net difference rate and proportion unmatched. Minus (-) sign indicates more assignments by census than by NCHS. Includes only those areas for which at least 50 nonwhite deaths were assigned according to either the census record or death record) Deaths matched at usual residence according to: Net Unmatched Pioportion difference rate deaths unmatched Area Census record Death record White | Nonwhite | White | Nonwhite White | Nonwhite |White | Nonwhite | White | Nonwhite Virginia: Newport News---=---= 108 71 109 72 0.9 1.4 31 31 22.3 30.4 Norfolk--------cuom_ 377 190 383 195 1.6 2.6 107 81 22.1 29.9 Petersburg---------- 56 49 56 51 - 4.1 13 14 18.8 22,2 Portsmouth---------- 140 78 143 79 2.1 1.3 40 36 22.2 31.6 Richmond-----=uceuoo- 432 185 440 185 1.9 - 85 92 16.4 33.2 Washington: Seattle--------=----o 1,205 591,260 60 4.6 hy 468 25 28.0 29.8 Wisconsin: Milwaukee----=-=-=-- 1,750 92] 1,766 91 0.9 «1.1 326 34 15.7 27.0 22 Table 5. Number of deaths matched at usual place of residence and net difference rates, by popu- lation-size group of area: United States, May-August 1960 [census record used as base. Minus (-) sign indicates more assignments by census than by NCHS] Deaths matched at usual residence according to: Net Population-size group of area differ- ence Census Death rate record record Total BLL ACCES wwe mne mmm mm oo J 0 0 tw 388,754 388,754 Pr Total areas Of 25,000 OF TOKE www mimmmmme mmm mmm mm mi 189,363 189,307 -0.0 1,000,000 Or MOYE==-=========--memee—m o-oo ———————————— 46,317 46,452 0.3 500,000 £0 1,000, 000 =m mmm mmm mom mm mi sim om mo om om 27,760 27,924 0.6 250,000 £0 S00, 00m wren mm sm mmm sm mmm si oof 0 0 0 25,671 26,259 2.3 100,000 £0 250, 000 w www wm wm mmm mm om mm mm mm mm mr mr te et 0 0 00 27,203 27,530 1.2 50,000 to 100,000==========-=-=---------------o-emmomoomoo——— 30,768 30,543 -0.7 25,000 to 50,000--=-=====---mmmmcmme—eeeemmemeomeommmomo mm 31,644 30,599 -3.3 10,000 TO: 25, QOD wre umm mimmim mim me mm rm om me 0 a 36,159 34,464 -4,7 2,500 to 10,000-----=ceecremmrmmm——— meee meme ———————— 41,510 37,786 -9.0 RUral=--=--mccmeece meme meme ommeeoooooos-me—=- 121,722 127,197 4,5 White All areas------=----m--mmemmemeeeeemeeeee—eo-e--------o-= 344,495 347,247 0.8 Total areas of 25,000 or more--~----memmmeeeceeeeecee cece ————— 163,749 164,963 0.7 1,000,000 OY MOTE===-=========m--m=-————m————-—o------o-=--=- 39,567 39,957 1.0 500,000 to 1,000,000=-=====m==mmee-ecomeeememmm— mmm 22,529 22,399 1.6 250,000 to 500,000----==c=rmmmemeeme—eceeee mmm mee ————————— 21,255 22,025 3.6 100,000 CO 250, 000 = tom we mm eso mm mm mm mm mmm or 1 2 0 23,423 23,899 2,0 50,000 CO 100,000 = = www mm mmm mmm mm mm mm me 0 0 0 2 28,003 27,982 -0,1 25,000 to 50,000---=-===-=-cmmmeceem—eeeeeeeemmmem——— mmm 28,972 28,201 -2,7 10,000 to 25,000-----===-==em-eemeemeeeeeeememmemme— momo 33,294 31,854 -4,3 2,500 to 10,000--=====----=-m-memmememmemeeeeeeo—-me—e—e——————— 38,142 34,879 -8.6 RUTr@le----mmemcmcmem meme ceeem meme mmmememmmmemeeo-cooomosooo- 109,310 115,551 5.7 Nonwhite All areasS=-=-======mmmmmceme meee eeememmemm—————------ 41,111 41,486 0.9 Total areas of 25,000 or more----=--=--------------c---oo-------- 24,084 24,335 1.0 1,000,000 Or MOYE@--===mmmrmmm;e—ee———————eeeeees——e—-————————— 6,401 6,492 1.4 500,000 £0 1,000,000 «wwwm mmm mmm mm mmm mm wm oom mmm en 0 5,026 5,024 -0,0 250,000 to 500,000 4,140 4,233 2,2 100,000 to 250,000 3,593 3,628 1,0 50,000 to 100,000-=-=======---- 2,509 2,561 2.1 25,000 to 50,000-c==rmemmmmm=mmmem— meee e————— 2,415 2,397 -0.7 10,000 LO 25, 000 = = mee me mmm mm mmm mmm mm mo 2 2,645 2,607 -1.4 2,500 £0 10, DOU «mw wm mm mmm mmm sm mmm mr to 5m 0 0 0 0 2,995 2,903 -3.1 RUT@l--=--memmmmmm moe ccce emcee emeemmme—ee—e-—ooo--—ees-- E1387 11,641 2.2 NOTE: For explanation of why white and nonwhite do not Appendix. add to the total, see Technical 23 Table 6. Number and percent distribution of deaths in the 1960 Comparison Study and the follow- back survey, by population-size group of area according to death record designations: United States, May-August 1960 1960 Comparison Followback Study survey Population-size group of area Percent Percent Number distri- | Number | distri- bution bution Total mm mmm mmm em mee ee ee 501,410 100.0 9,475 100.0 1,000,000 OF MOTE@= === === mmm eee oom 59,390 11.8 1,134 12,0 500,000 to 1,000,000== === === =m eo 35,800 7.1 717 7.6 250,000 to 500,000-======m mmm meee eee —————e 33,830 6.7 651 6.9 100,000 to 250,000 === === =m ome 34,600 6.9 653 6,9 50,000 to 100,000== === === == meee emo 37,991 7.6 708 7.5 25,000 to 50,000= === == === meee 38,000 7.6 718 7.6 10,000 to 25,000== === === ome eee ememmmmem 43,425 8.7 825 8.7 2,500 £0 10,000== == === ooo m mm ee eee moon 47,434 9.5 843 8.9 All Other=- == comme eos 170,940 34,1 3,226 34,0 Table 7. Number and percent distribution of deaths matched in the 1 960 Comparison Study and the followback survey, by population-size group of area according to census record designations: United States, May-August 1960 1960 Comparison Followback Study survey Population-size group of area Percent Percent Number distri- | Number | distri- bution bution Total =m mm mmm mmm meee ee ee eee 388,754 100.0] 7,393 100.0 1,000,000 OF MOTE==== === mmm ee eee ememmmmm 46,317 11.9 889 12,0 500,000 to 1,000,000======== =m eee emmmoooo 27,760 7.1 539 7.3 250,000 to 500,000== === === =m meee 25,671 6.6 495 6.7 100,000 to 250,000= === === mmo 27,203 7.0 474 6.4 50,000 to 100,000 === === comme momo 30,768 7.9 564 7.6 25,000 £0 50,000 == === == meee eee oem 31,644 8.1 292 3.9 10,000 £0 25,000== === === mmm ee eee eee 36,159 92.3 700 9.5 2,500 to 10,000= === === oom omen 41,510 20.7 761 10.3 25,000 OF TTR! cxvmmmmmammsmmmmsinss msm m—————————— en ve. 302 |— 4,1 All Other=--mm ooo oem lLo os 121.722 31.3 2,377 32,2 'Processing errors by census category represents areas of 25,000 population or more, of area could not be further defined. 24 but size Table 8. Number and percent distribution of unmatched deaths, for which there were response and no response in the followback survey, and unmatched deaths in the 1960 Comparison Study, by population-size group of area according to death record designations: United States, May-August 1960 Followback survey Viadtched deethe in 1960 Comparison Response No response Study Population-size group of area Percent Percent Percent Number | distri- | Number | distri- | Number distri- bution bution bution Total--===m=m=m--c-eeemcememm—eo———m-emo———— 1,861 100.0 221 100.0 | 112,656 100.0 1,000,000 OF MOYE@=--==mm==mm============--=-eceee=—==———— 189 10,2 30 13.6 12,938 11.5 500,000 to 1,000,000-- 127 6.8 34 15.4 7,876 7.0 250,000 to 500,000---- 121 6.5 16 142 15571 6,7 100,000 to 250,000---- 140 745 11 5.0 7,070 6.3 50,000 to 100,000--==============------=-------=----o--o 126 6.8 24 10.9 7,448 6.6 25,000 to 50,000===========-=-c----------m-o-oom--soooo 127 6.8 17 7.7 7,401 6.6 10,000 £0 25, Q00 = = w= wommom mmm mm mm mie mo on a a 0 0 0m i 1355 8.3 18 8.1 8,961 8.0 2,500 TO: 1 D0 wwe mmm wm mm 558 5 A ES A mr mr 164 8.8 3 2,3 9,648 8.6 All other---==-===mm-e-ecmceememmmmmm——ommmmmomm mmo 712 38.3 66 29,9 | 43,743 38.8 Table 9. Net difference rates and proportion of death records unmatched for urban and rural areas in metro- politan and nonmetropolitan counties, by color and sex: United States, May-August 1960 [ceath record used as a base. Minus (-) sign indicates more assignments by census than by NCHS. For definition of areas, see Technical Appendix] Propor- Propor- na. tion of nk, tion of Area, color, and sex deaths Area, color, and sex deaths ence ar ence Ta Rote matched rete matched on 21.1 White female—Con. -0.3 20.8 0.5 21.5 2.4 19.7 0.3 19.4 . * -3.6 27.4 -0.9 22.9 0.5 21.6 sw 34.6 2.0 20.0 -0.7 35.4 -0.5 22,7 -2.2 33.8 ww» 21.0 15.0 48.1 -0.3 20.8 1:1 33.2 0,3 19,2 0.3 31.8 -3.4 27.4 2.0 34.1 0.5 21.,2 31.5 2:3 19.8 -0.3 31.0 -0.7 22,0 -1.2 29.4 NI 20.4 | Rural------------e-eocmmmem—————— 10.0 45,1 -0.3 19.8 Nonmetropolitan counties----- 0.4 32.4 0.6 18.0 [[ Urban----------===--m-mmmmm————- -0.2 28.8 -5.0 27.8 || Rural----=--===-===------c-omo-oo- 1.0 34.9 NOTE: Figures with color and/or sex not stated are included in the total, but are not shown separately. 25 Table 10. Number of deaths matched at usual place of residence and not at usual place of resi- dence, according to census record and death record for urban and rural areas in metropolitan and nonmetropolitan counties, by color and sex: United States, May-August 1960 [F or definition of areas, see Technical Appendix] Area according to death record Area according to census record Total Metropolitan county Nonmetropolitan county Total Urban Rural Total Urban Rural USUAL PLACE OF RESIDENCE Total All counties-===-ceeomaaoo_ 388,754 241,075 || 201,366 | 39,709 147,679 || 60,191 | 87,488 All metropolitan counties---- 241,030 240,717 (|201,154 | 39,563 313 118 195 Urban----ceocmm meee 205,660 || 205,469 (|197,373 8,096 191 100 91 Rural-=--e-coo momo 35,370 35,248 3,781 | 31,467 122 18 104 All nonmetropolitan counties- 147,724 358 212 146 147,366 || 60,073 87,293 Urban=-=-=-comm mee. 61,372 120 86 34 61,252 || 57,565 3,687 Rural-=---cooommm eo 86,352 238 126 112 86,114 || 2,508 | 83,606 White Male--==ccommcmm ee 190,119 115,442 || 94,749 | 20,693 74,677 (| 29,139 | 45,538 All metropolitan counties---- 115,438 115,263 || 94,624 | 20,639 175 60 115 96,833 96,733 [| 92,632 | 4,101 100 56 44 18,605 18,530 1,992 | 16,538 75 4 71 74,681 179 125 54 74,502 11 29,079 | 45,423 29,867 57 47 10 29,810 || 27,870 1,940 44,814 122 78 44 44,692 1,209 | 43,483 Female-----eccccmmaaaao 138,013 87,880 73,672 | 14,208 50,133 || 21,899 28,234 All metropolitan counties---- 87,854 87,761 73,614 | 14,147 93 41 52 Urban=-=---cmcmmmmm ee _. 75,443 75,381 72,430 2,951 62 30 32 Rural=-=--omccmcmee eo 12,411 12,380 1,184 | 11,196 31 11 20 All nonmetropolitan counties- 50,159 119 58 61 50,040 || 21,858 28,182 Urban=-===cocoocm mmm 22,380 44 28 16 22,336 || 21,071 1,265 Rural --=---mocm momo 27,779 75 30 45 27,704 787 26,917 Nonwhite Male-==-=-ccmcmmmme 19,886 12,113 11,076 1,037 73273 2,998 4,775 All metropolitan counties---- 12,108 12,089 11,061 1,028 19 6 13 Urban-====ccoom me 11,141 11,129 10,924 205 12 5 7 Rural 967 960 137 823 7 1 6 7,778 24 15 9 7,754 || 2,992 4,762 2,996 9 7 2 2,987 || 2,851 136 4,782 15 8 7 4,767 141 4,626 26 Table 10. Number of deaths matched at usual place of residence and not at usual place of resi- dence, according to census record and death record for and nonmetropolitan counties, by color and sex: United States, May-August 1960-—~Con. [For definition of areas, see Technical Appendix] urban and rural areas in metropolitan Area according to death record Area according to census record Total Metropolitan county Nonmetropolitan county Total Urban | Rural Total Urban Rural USUAL PLACE OF RESIDENCE--Con. Nonwhite—Con. Female----=-vm-mmmecencne—a- 16,814 10,520 9,666 854 6,294 || 2,703 3.591 All metropolitan counties-=--- 10,523 10,507 9,660 847 16 6 10 Urban=---====c-eecmemccmm cme m 9,732 9,722 9,566 156 10 4 6 Rural-=----c-ccecmmcmm emcee eee 791 785 94 691 6 2 4 All nonmetropolitan counties- 6,291 13 6 7 6,278 || 2,697 3,581 Urban----=====c-ececeeme cece eee 2,703 4 1 3 2,699 | 2,591 108 Rural---==-c-ccmmommcc cece e eee 3,588 9 5 4 3,579 106 3,473 NOT AT USUAL PLACE OF RESIDENCE Total All counties---=---=-=--m-- 31,538 19,992 16,936 | 3,056 11,546 5,038 6,508 All metropolitan counties---- 19,210 17,041 14,445 | 2,596 2,169 932 1,237 Urban-----=--c-cccmmeee emcee 13,368 11,886 10,5811 1,305 1,482 632 850 Rural-----=-ceccccccmcmncnceeeu= 5,842 5,155 3,864 | 1,291 687 300 387 All nonmetropolitan counties- 12,328 2,951 2,491 460 9,377 4,106 5,271 Urban---------csmocmmmnnc cena ae 5,165 1,075 890 185 4,090 || 2,170 1,920 Rural----=---eeccmmmm mmm cece ee 7,163 1,876 1,601 275 5.287 1,936 3,351 White Male---------ccccmcmmmee em 13,450 8,325 7,030 | 1,295 5,125 || 2,183 2,942 All metropolitan counties---- 7,920 6,891 5,807 | 1,084 1,029 424 605 Urban------===--cccmmcoooanonannn 5,274 4,550 4,049 501 724 309 415 Rural---------cccccccmmem ccm eeee 2,646 2,341 1,758 583 305 115 190 All nonmetropolitan counties- 5.530 1,434 1,223 211 4,096 1,759 2,337 UEDA m= wm mm mim re mm em ee rm em 2,223 543 437 106 1,680 849 831 Rural-----=----cmemcmmc cee cceeem 3,307 891 786 105 2,416 910 1,506 Female--------=-cc-ceocuan- 15,493 9,902 8,322 1,580 5,591 || 2,552 3,039 All metropolitan counties---- 9,634 8,689 7.323 1,366 945 430 315 Urban----------c-ccccmcmmmcnennan 7,049 6,372 5,632 740 677 291 386 RIEL orion i sis wo on SR oo ml i 2,585 2,317 1,691 626 268 139 129 All nonmetropolitan counties- 5,859 1,213 999 214 4,646 2,122 2,524 Urban----=ccemmmmmm meee meee 2,662 439 373 66 2,223 || 1,221 1,002 Rural------cececccmceccccccccea= 3,197 774 626 148 2,423 901 1,522 27 Table 10. Number of deaths matched at usual place of residence and not at usual place of resi- dence, according to census record and death record for urban and rural areas in metropolitan and nonmetropolitan counties, by color and sex: United States, May-August 1960—Con. [F or definition of areas, see Technical Appendix] Area according to death record Area according to census record Total Metropolitan county Nonmetropolitan county Total Urban | Rural Total Urban Rural NOT AT USUAL PLACE OF RESIDENCE—Con. Nonwhite Male====---cmcoommmcm ceo 1,273 886 793 93 387 148 239 All metropolitan counties---- 800 696 625 71 104 39 65 Urban=------mceccmc meee meme eee 467 428 405 23 39 15 24 Rural-------omcccmmm meee meme - 333 268 220 48 65 24 41 All nonmetropolitan counties- 473 190 168 22 283 109 174 Urban=---=---cccccmmm mmc cee 139 61 52 9 78 42 36 Rural=--=--coocomm ieee me eee 334 129 116 13 205 67 138 Female-=-------cocomocomn 903 617 565 52 286 115 171 All metropolitan counties---- 585 518 476 42 67 26 41 Urban=-------cccmmmmmmcee eee 379 350 330 20 29 12 17 Rural------cooommcmcmcm emcee mmo 206 168 146 22 38 14 24 All nonmetropolitan counties- 318 99 89 10 219 89 130 Urban---==-cecmcmcmm ccm eeeeemo oo 108 27 25 2 81 48 33 Rural----=-ccommmm meme mee mm 210 72 64 8 138 41 97 NOTE: Figures with color and/or sex not stated are included in the total but are not shown separately. 28 Table 11. Comparison of net difference rates and proportion unmatched, based on deaths matched at usual place of residence for 25 selected standard metropolitan statistical areas (SMSA's) and urban area components of those SMSA's: United States, May-August 1960 [urban area components are only those of 25,000 population or more. For further definition of areas, see Technical Appendix] Proportion Net unmatched prea fr Census Death 1 vas record | record base base ALDUGUETTGUE , Ny MH wri mom mmm mmm mo 0 0 000 9 tt ct 0 0 -0.3 31.4 31.4 AlDUGUETQUE, IN, IEE em rm mow mm me ff 0 20 0 6.0 30.6 29.3 Augusta, Ga.-S.C--=-----=-==------------e-----e---—o-o-----o-sosososomes 0.3 36.1 36.1 Augusta, Ga---------=====-mm----- e-em eoo---—eoeo-ooooooo-o--o-o-sss 8.2 32.2 30.5 BRULLIIEE , MOTVE m0 mien rw aw mm mo m0 5 0 900 0 ft - 31.1 31.1 Billings, MONG === mmm mom momma 0 0 0 0, 2 0 8.5 29.3 27.7 Chicago, Ill----==---m=meeccec eee eee — eee ees e-em em m= ——————— 0.2 21.3 21.3 Arlington Helghts, TLL mmmmmm mim ime imi mi oo hon mon on me wm oe cw 4,2 32.4 31L.5 BUTOTR , TL Lm moo escent mm me em ee 0 00 0 a mt 8.3 20.0 18.8 Berwyn, Ill--------c--mmmmmmmomoooooooooooooom-o-ooooo-oooooooosses 1.1 2.6 2.7 Calumet City, Ill----=--------m---oo-coooomoo momo oooooo—oo-oooo== 4.5 21.4 20.7 Chicago, Ill--=----cccmmocommmommmooooooooooooooomoo—moooooooooooes 0.5 22.3 22.8 Chicago Helghts, Ill cw mmmmmmmm min amo im mim imine shee mah or or 1:2 16.0 15.8 Cicero, Ill-==--mm-m-mccmemocmmmmmmmeme—oo— meme m-m-———---—-——m———--o- -0.9 6.6 6.6 Des Plaines, Ill==— === cmc em mmm mm mmm mo 0 10 1000 0 0 0 00 00 0 mm 15.8 33.3 30.2 Elgin, Ill--------=--mmommmmmm oo ooooooo ooo oo--oooooooooooooooooss «3. 17.7 18.2 Elmhurst, Ill----------m----mmmmmmo moomoo ooo om ————oooooooo- o-oo 4.7 14.7 14.1 Evanston, Ill------------cmmemmommccoecm meme mm mmm —m momo m—— mmm mmm -0.5 10.9 11.0 HArvey, TLLi-mimmism comm mmm mmm mim mm mo coo on ao oo - 25.6 25.6 Highland Park, Ill-----====----------------m-omoooooooooooo=om=omm= - 15.1 15:1 Joliet, Ill=--m--mmemmmmmmmcmmommmmmommoomoooooo--o-------ooooo-ooos 4.4 13.2 12.7 Maywood, Ille--=-=ememmm mm nn -8.5 1.7 14.8 Oak Lawn, Ill--=-----------mceommmmmmmmmo——o moomoo —-o---—-o-o-oom-oo 91.4 18.6 10.7 Oak Park, Ill------=-=----emmmccmemmmeo mmm m mmm mmm mmm mmm mmm 0.6 20.4 20.3 PATK TOT@EE, LLL =m mir mission mse amos am mom mm mim mo mr om 0 0 0 00 0 oo - 8.0 8.0 Park Ridge, Ill=------------m--mmmmmooomooooo-ooo—---o--o-o-oooooo- 2.) 9.4 9.3 Skokie, Tlle=mm-========m=m mmm mmmmmm mmm meme —--ooe -1.1 10.7 10.8 Waukegan, Ill=-=-=====--=mmmmmmo= mmm mmo emo mmm memes 5.6 19.1 18.3 Wilmette, Ill-------------meeccce cee meme esses meee ————————— - 4.3 4.3 Davenport-Rock Island-Moline, Iowa-Ill-----------=-=-====-=--------------- -1.4 19.1 19.4 DEVENDOLE ; LOM iim swim im ce wm a mo mm 5 0 0 2 -3.2 20.3 20.8 Moline, Ill----------eemmmmcmmmmmmcmmcecmmmem meme m mmm mm mmm mmm mmm 6.5 23.9 22.8 Rock Island, Ill-----------=------m-mommcoooooo-—ooo-—-----o-o-----o= -1.5 9.6 9.0? El Paso, TeX=-=-=--=--==-=m- meee emo —e———-——-o--—so—------es 3.3 28.9 28.2 El Paso, TeX==--====-==----- mee eemme cee em em -—----————o-------= 1.7 28.7 28.3 Fort Wayne, Ind---------------mm----oooococooo-o-oooooooooooooooooooss 0.2 12.5 12.5 Fort Wayne, Ind---------=cc---mmm-omooooooomcooo moo —o—o—-ooooooooo os 1.5 12.5 12.3 Jersey City, N.J=-=------emmmecece meee emer meme mm meme cee mmm =m — 0.2 17.9 17.9 BAYONNE, Ny Jimmi oo or am mc mm mm te mm 20 00 0 0 0 30 0 gr - 20.0 20.0 HOBOKEN, NJ mi mimo oo or am om on am mm mm mt 0200000 0 0 0 Se -0.8 25.6 25.7 Jersey City, N.J-------omcommmmmmomcm moomoo momo moomoo moomoo -0.8 18.9 19.1 Kearny, N.J=--=--cc-cmmmcommm mmm mm momo mo mmmmoo oo ooo -----ooooo-oooooo -0.9 11.9 12.0 North Bergen, N.J----------mcmcommmmoceocooooooo—o————-—--o--—--o= - 12.3 12.3 Union City, N.J---=-=-=-scmememecmeceee meee memmmeme mo mm =m mom m === 0.6 9.9 9.8 West New York, N.J--=---c-cmcmcmmmmm mmm mcmm mmm mmm mm mmmm mmm mmo me -1.1 20.7 20.9 Laredo, TeX=-====-=--=-=--- eee ooo-----oo---oooo--o-ss 0.8 25,3 25,2 Laredo, TeX-----=-=-=---=--eeeeece eee eee memes see eee eee ————— - 25.7 25.7 Lorain-Elyria, Ohio=---==--====-=-=------c--ooommomomoooo moomoo moomoo -0.3 25.0 25.1 Elyria, Ohio----=--------m-momocoooommc ooo mmmmmoo moo omoo moomoo -3.0 17.3 17.7 Lorain, Ohio--==-=---=c--c--commmmmmemo oo ocoocoommommooooooomooomses -0.8 19.6 19.7 29 Table 11. Comparison of net difference rates and proportion unmatched, based on deaths matched at usual place of residence for 25 selected standard metropolitan statistical areas (SMSA's) and urban area components of those SMSA's: United States, May-August 1960 —Con. [urban area components are only those of 25,000 population or more. For further definition of areas, see Technical Appendix] Proportion Net unmatched res il 1 Census | Death rate record | record base base Milwaukee, Wis == == cm cm momo eee 0.1 16.6 16.6 Milwaukee, WisS=-=== como mm mmo ooo eee eeeee eo 0.5 16.3 16,2 Waukesha, Wis=- == common oom oeeeo aoe -7.0 26.0 24.7 Wawatosa, Wis === mm moomoo ooo eee eam 6.3 13.0 13.8 West Allis, WiS==mmm momo mm mmo ome ee eee em 5.7 746 7:3 Muncie, Ind - 25.2 25.2 Muncie, Ind 13.3 19.5 17.6 Nashville, Tenn=-===cc comm mmm eee -0.3 25,5 25.5 Nashville, Tenmn=--- === como meee ee emeeeeee eee 3.0 2343 22.8 Odessa, TeX=-=======-c ceca meme mmm mmm meme mm - 29.9 29.9 Odessa, TeX===-=-==-c comme meme 1.2 33.1 32.8 Omaha, Nebr.-Towa==--c o-oo emcee meee 2.1 20.1 19.8 Council Bluffs, ITowa==-====momc comme emma - 3.6 3.6 Omaha, Nebr-- o-oo ee eee eee 4.1 22.0 21.3 Racine, WiS=-=-m com mm moo me ere - 15.4 15.4 Racine, WiS=-=--mmmcmo oo mmm eee 0.5 12.0 12.0 Rockford, Tll=-=---co oom mmm om meee m -0.2 18.3 18.3 Rockford, Ill=-----mcomm mmm eee eee ema m 2.5 14.2 13.9 Salt Lake City, Utah-=----ccmomm mmm meee eee - 23.7 23.7 Salt Lake City, Utah=-=--cccmcm mmm eee eee 3.7 20.9 20.3 Shreveport, La=--------ccc mmm mmm -0.4 31:1 31.2 Bossier City, La=---== =o omc eee - 28.9 28.9 Shreveport, La=-=----c-c--commm oe reece mem 0.7 27.6 27.4 Steubenville-Weirton, Ohio-W.Va===== cece momo eee eee om 0.8 24.8 24.7 Steubenville, Ohio=====cc comme rere ema - 19.9 19.9 Weirton, W.Va=-- =o moomoo momo ee eee 2.9 32.0 31.4 Tacoma, Wash=- coco eee eee eee eee -0.3 28.4 28.5 Tacoma, Wash=-- cc cocm mm me eee meee = 1.2 17.9 17.7 Tuscaloosa, Ala==---- =m om moomoo rmrccmcee meme een - 48.1 48.1 Tuscaloosa, Ala=--=-=- mmc eee 13.3 46.5 43.4 Wichita Falls, TeX==-===-cc comme cecmccmmm meme 1.9 24.0 23.6 Wichita Falls, TeX=-===--- com omom mooie em 2.7 20.0 19.6 Wilkes -Barre-~Hazleton, Pa=--== o-oo mmm - 15.1 15.1 Hazelton, Pa=--==-=-o ooo ee eee - 13.7 13.7 Wilkes-Barre, Pa=---===---cmcomm momen 21.3 21.3 Youngstown-Warren, Ohio=--=-=c--ccommm mmm emma -0.1 19.9 19.9 Warren, Ohio=------cmmmm meme meee e 5.L 18.0 17.2 Youngstown, Ohio=---- ccc mmm oo ee eee 0.2 13.5 13.4 ICensus record used as base. Minus (-) sign indicates more assignments 30 by census than by NCHS. Table 12, Number of matched deaths reporting same standard metropolitan statistical area (SMSA) on both census and death records, on census record only, and on death record only; and net difference rates for total matched, matched at usual place of residence, and not at usual place of residence: United States, May-August 1960 [census record used as a base. Minus (-) sign indicates more assignments by census than by NUHS. For definition of areas, see Technical Appendix] Number reporting Net differ- Ares Same Census | Death ence on record | record | rate both only only Total All areaS-=---mmmrmrrmmcee seem mm ——————————— 412,197 8,095 | 8,095 ow Total SMSA'S=-=-cmmoccmmmm meme 255,454 4,786 | 5,613 0.3 Abilene, TeXx~--=======rreeecc-ccc; ce; ; ccc; cee; — ce ——————————— 179 4 9 2.7 Akron, Ohio=-=---ccmeeemm ccc 1,042 24 46 2.1 Albany, Ga--====-=---ccmcem meme eo 126 - 1 0.8 Albany-Schenectady-Troy, N.Y=---e-emcmcmcmm ccc cece meme 1,833 19 76 3.1 Albuquerque, N.MeX--=-=---meeeeeeee ccc ccc ccc ecm ————— 364 3 3 - Allentown-Bethlehem-Easton, Pa.-N.J-===--ccceememcecccmcananooo 1,398 14 16 0.1 Altoona, Pa==-===-=--memccmccmmemee cece e mmm emcee mmm mn 404 21 4 -4,0 Amarillo, TexXe=wwememmmrrececccr cree —— ee —————————————— 210 15 7 -3.6 Ann Arbor, Mich------c-cemmcmmm cme eee emer em 288 70 9 -17.0 Asheville, N.C-------mmmmcmcmmm meme mc eecc meme e meme 264 22 6 -5.6 Atlanta, Ga---=-=-=----memceccecmeme meee cece ee ee —————————— 1,869 24 62 2.0 Atlantic City, N.J-==-eercrermemr ccc cc ccc mmm mm mmm 517 6 17 2.1 Augusta, Ga.-S.C--=------rerercmecm cence m ec —————— 350 14 12 0.5 Austin, TeX---==-=----ecmccrcmm cece cce cece eeeece— eam 329 55 6 -12.8 Bakersfield, Calif-----------cccccmmm ccc 467 - 16 3.4 Baltimore, Md-====vrrrerremre cree; cc; ec; ————————————————————— 3,855 64 55 -0.2 Baton Rouge, La---=----mmcmoomm mmc o 356 7 12 1.4 Bay City, Mich------ccccmm mmm meee 224 1 8 3.1 Beaumont -Port Arthur, Tex--------c-emcmmeomm mmm cm meee mmm 546 8 18 1.8 Billings, Mont--=--------eeecce cece eee e ee em ee emm—e——————— a 134 - 1 0.7 Binghamton, N,Y==-----cmcmmmmmm mmm e meee meee eee 522 33 14 -3.4 Birmingham, Ala-----------cerrcecemc cme m mmm eee 1,200 8 37 2.4 Boston-Lowell-Lawrence, Mass! --=cmmomomommooommceeeeo 8,713 71 129 0.7 Bridgeport -Stamford-Norwalk, Conn! -=--evemmeme eee eens 1,532 28 34 0.4 Brockton, Mass! - == momo mmm 692 20 28 1:1 Brownsville-Harlingen-San Benito, Tex-----===-m-ecccccenecaa. 223 5 7 0.9 Buffalo, N.Y=--c--ommommmmm meme eee meee 3,468 57 53 -0.1 Canton, Ohio==-=-----cemecmem emcee ccc meme 764 42 9 -4,1 Cedar Rapids, Iowa--=-=-cc-cemomemcm emcee enema 241 - 3 1.2 Champaign-Urbana, Ill--------eeeecmccmm meme ccc ccc 192 2 13 5.7 Charleston, S.C-=---=---eeceee;cecccccccc cc; ee; cece e——————— 295 6 12 2.0 Charleston, W.Vas-=--eeeccomm mmc cece cee mmc ccm mmo o 436 3 9 1.4 Charlotte, N.C===---cecmccmme mec ccccccc cece mcccccme momen 405 4 2 -0.5 Chattanooga, Tenn.-Ga---====-s-e-e;cceccee cece e cee ——————— 512 6 10 0.8 Chicago, Ill--=cccomcommcmcmmmm mmm mmm meme m 15,504 91 377 1.8 Cincinnati, Ohio-Ky---==-=-cccmcmmmme mmc ce meme e 2,694 32 29 -0.1 Cleveland, Ohio=--==----cmocmmmm emcee meee emcee mmm 4,406 12 94 1.9 Colorado Springs, Colo-=--=-=--ccecccmcmme cece mmm m meme 245 13 19 2.3 Columbia, S.Ce==--s-cmomcm meme eee 432 124 - -22,3 Columbus, Ga.-Ala-=--=====----ccmemcccee cece e— em mma 309 6 19 4,1 Columbus, Ohio--=------e-eemeccccccc cece cece ce ce—ce ema e 1,490 35 24 -0.7 Corpus Christi, Tex=--===--ccmccmcmmmce cece cmcmmmm mo 312 7 11 1.3 Dallas, TeX=------=--c-mmmemee eee eee mccem meee m mmm 1,880 43 54 0.6 Davenport-Rock Island-Moline, Iowa-Ill------cccmemeoceenananxn 633 32 10 -3.3 Dayton, Ohio=-=-----cccocccm meme 1,458 37 10 -1.8 31 Table 12. Number of matched deaths reporting same standard metropolitan statistical area (SMSA) on both census and death records, on census record only, and on death record only; and net difference rates for total matched, matched at usual place of residence, and not at usual place of residence: United States, May-August 1960—Con. [census record used as a base. Minus (-) sign indicates more assignments by census than by NCHS. For definition of areas, see Technical Appendix] Number reporting Net differ- Aves Same Census | Death ence on record | record rate both only only Decatur, Ill------oooo mmm 237 5 5 - Denver, Colo=-----=------ 1,942 22 66 2.2 Des Moines, Iowa 622 3 19 2,6 Detroit, Miche=--cmoom mmm eee 7,848 26 141 1.5 Dubuque, Iowa-=-=-=-=cccm comme 186 23 4 -9.1 Duluth-Superior, Minn. -WiS-=-----ccmm mmm cece 741 16 37 2.8 Durham, N.C=---mo mmm moomoo eee ee 222 1 11 4:3 El Paso, TeX==-=-----em mmm eee 439 7 22 3.4 Erie, Pas----ccmoommo moon 649 3 21 2.8 Eugene, Oreg=-=-----mmmo mmm meee eee eo 238 - 9 3.8 Evansville, Ind. -Ky---=----c-mmmmm mcm em 440 19 2 -3.7 Fall River-New Bedford, Mass! mmmmm mmm eee 998 13 16 0.3 Fargo-Moorhead, N.Dak.-Minn--=---=--ccccocmmmmm cameo 170 7 14 4,0 BUABE ; TEL Clim scmmmems mmm mon mmo msm isnt so wer os wit 720 9 14 0.7 Fort Lauderdale-Hollywood, Fla---=----cccemmmmmmm cm eecmem eo 702 26 5 -2.9 OTE SINLEIL, AE im scion owt ins 00 0 127 5 3 -1.5 Fort Wayne, Ind-=------cmmmmmm mmm mmm ee mem 521 25 27 0.4 Fort Worth, TeX=---=----mmmmmm meee meme ee em om 1,025 14 28 1.3 Fresno, Calif---cccommmmm mmm meme eee 715 6 13 1.0 Gadsden, Ala--=-=--c-cmmmm meme eee e em 150 10 3 -4.4 Galveston-Texas City, TeX===---scomcomom mmc ema n 278 3 14 3.9 Gary-Hammond-East Chicago, .Ind----==---cccmmmmmmmmn ane 1,166 1 35 2.9 Grand Rapids, Mich=-----ccmmmmm meme e 808 34 20 wl,? Great Falls, Mont=-------cocmmmmm meme 180 6 3 -1.6 Green Bay, WiS-=----cmmomm mmm meee eee em 253 8 12 1.5 Greensboro-High Point, N.,Ce=-=--occmcmmmm mmm meme em moo 392 11 9 «0,5 Greenville, S.Ce=-=----om mmm - 269 - 12 4,5 Hamilton-Middletown, Ohio--===-=-cocmmmm mmm eeeeeee eo 400 - 7 p Harrisburg, Pa==-----cmcmmcmm meme e 885 39 29 -1.1 Hartford-New Britain-Bristol, Conn! =-=----mmmooooommmmaaaoon 1,551 16 41 1.6 Honolulu, Hawaii-=--==--ccmcmcm mmm meme emma eo 602 6 2 -0.7 Houston, TeX=-=-=-=s- meme me meme eee 1,934 18 31 0.7 Huntington-Ashland, W.Va.-Ky.-Ohio=-==-=-cccmmmmmmmmnnea 529 23 13 -1.4 Huntsville, Ala---=---ccmm comme eee em 144 6 4 -1.3 Indianapolis, Ind---------cccmmmm mmm emcmeee eee - 1,673 19 49 1.8 Jackson, Mich=-=-eooomo oom meee eee e 345 5 15 2,9 Jackson, Mig =====m-- momo ome emma - 346 22 17 -1.4 Jacksonville, Fla-------cccommmm imme eemem een 839 13 22 Lal Jersey City, N.J-----mmmmo mmm meme meme - 1,865 9 60 247 Johnstown, Pas-------omoo momma 813 21 20 -0.1 Kalamazoo, Mich=----mm commen mmm mmmeeemee eee em 322 29 2 «Ty? Kansas City, Mo. -Kan§--=-==--cccmommmmm mmo mmmemmemom 2,330 26 94 2.9 Kenosha, Wis=---=---cmomm ome 192 5 10 2.53 Knoxville, Tenm=--==-=-=o- mmm mmmemmmem mem 622 36 8 -4,3 Lake Charles, La--------cmmmomom mmm meee meee 185 - 7 3.8 Lancaster, Pas------co common meee ee 656 64 24 -5.6 Lansing, Mich=----ocmmmmm mmm eee 568 12 16 0.7 Laredo, TeX====== =m mmm momo eee memmeme moo 118 - 8 6.8 Las Vegas, Nev=--- oom omm mmo ome oe oe emma 199 - 9 4.5 Lawton, OKla---=---c moomoo mmo meme eee eee 99 1 6 5.0 32 Table 12. Number of matched deaths reporting same standard metropolitan statistical area (SMSA) on both census and death records, on census record only, and on death record only; and net difference rates for total matched, matched at usual place of residence, and not at usual place of residence: United States, May-August 1960-—Con. [census record used as a base. Minus (-) sign indicates more assignments by census than by NCHS. For definition of areas, see Technical Appendix] Number reporting Net differ- Area Same Census | Death ence on record | record rate both only only Lexington, Ky------===cscseccmmmmmeme eee meme eee ——————— 292 40 6 -10.2 Lima, Ohio--=---==---==--m-mmmmmoooocooooom ooo oomomooom momo 228 14 21 2.9 Lincoln, Nebr-----------eeeecocccooocm omen mm mmm mmm mmm mm 355 21 3 -4,8 Little Rock-North Little Rock, Ark--------------------------o 410 32 4 -6.3 Lorain-Elyria, Ohio------------mcomcmcmom momen mmm mmm 394 17 7 -2.4 Los Angeles ~Long Beach, Calif mmm mm mmm. mmio im mmmom mmm mew 14,583 90 203 0.8 Louisville, Ky.-Ind------=--=-mememeceececemem meme me mmm mmm 1,643 37 16 -1.3 Lubbock, TeX=====-=--==-=-c--mm-mmomome——ooooo-ooco--o—-—-o 179 - 24 13.4 Lynchburg, Va--------------m-mmceoocoomommmmmmm mmm mmm mmm mn 215 22 12 -4,2 Macon, Ga==-===-======-=c---mmmmmee moe -------s-o------o-e 336 16 11 -1.4 Madison, Wis---s--sceccncurunmmmneee emer eccn cnn nm —————— 426 23 14 -2.0 Manchester, N.H!o-coomommmm moomoo meee mem 450 2 10 1.8 Memphis, Tenn------=-=====------ccoo-coocoooomoomoo—oo——o—o 1,297 22 18 -0.3 Miami, Fla----==----cecmommmmmmmmo momo mmmmmmmmomommmm mmm mmo 1,984 3 16 0.7 Midland, TeX------=-recmrreeeeeee eee eee eee cece ——————————— 61 - 6 9.8 Milwaukee, Wis---=--=---c---mcmmmcomcccoccommmmmmmm mmm mmm 2,762 23 33 0.4 Minneapolis-St. Paul, Minn-----=--=------------c-oommmmmomonoo 3,207 65 90 0.8 Mobile, Ala=======-=----c-mmmmmmmmmmm—emoeoem-m-oo-o—----- 520 35 21 “2v5 Monroe, La==--=--c-cmccomommmmmmm moms -oem——--———m eo 195 3 9 3:0 VONUGOMBLY , AL mw wre mem mmm mm ms i oo 280 16 10 -2.0 Muncie, Ind-------=-c----mmmemmocccooooooooomsoommmommm ooo 242 - 6 2.5 Muskegon-Muskegon Heights, Mich------------ccoomeomommoonommo 328 5 10 1.5 Nashville, Tenn-----====----=---mm-cce-e-eeeceemee——— moo 819 49 4 -5.2 Newark, N,J-----=-c-ce-cmmcmmmmcccccccememm meme mmm mmm mm 4,163 116 115 -0.0 New Haven-Waterbury, Conn! ----eeecoooooom omen 1,672 18 60 2.5 New Orleans, La-=------=c--mcmmm-e-comooooooooooooooomm—o 2,169 24 56 1.5 Newport News-Hampton, Va--=--=--===--=--c-----c-coo-mooommoooono- 366 34 8 -6.5 New York, N.Y-----o-o-oomemmmmmmm moomoo mmmmommmmmmomm mmm mmm 28,001 93 333 0.9 Norfolk-Portsmouth, Va---------ccmomcmmommec mmm m mmm mm oe 974 3 34 3:2 OA@S8A, TER m= wren wre wm mm mmm mm mm mm oo 0 mm 101 - 2 2.0 ORE, TE aie immo mm som 1mm mim mmm mm mmm ms ie 0 0 0 0 0 179 1 4 1.7 Oklahoma City, Okla-----====-=--mecmecocccccmemmemm mmm mmm 915 55 15 =4,1 Omaha, Nebr, -Iowa----====--===-=-c-c-coco-c-omomomoooooooo 968 29 48 1.9 Orlando, Fla------==-----cemceomeccmmoemommm mmm mmm mmm mmm mm 555 5 8 0.5 Paterson-Clifton-Passaic, N.J--=--=-=-mmeecocmemcmemnnnen mane 2,619 42 100 2.2 Pensacola, Fla------==------emeemcecccee meen mem mmm mmm 306 4 10 1.9 Peoria, Ille---momcmoommmoomcmmmmmmmmmmmmommmmmmmmmmm mom 644 52 8 -6.3 Philadelphia, Pa.,-N,J-----====-------eccoocoooomoooomm mmm mmo 10,931 94 134 0.4 Phoenix, Ariz--------------mmomeoocccooeommmom momo mmm mm 994 42 22 -1.9 Pittsburgh, Pa-----------c--mmmemccccmcccmememmmm mmm mmm 6,061 57 54 -0.0 Pittsfield, Massl---mmmmmoocom mmm 418 6 10 0.9 Portland, Mainel----=--c-cmmmoomo momo mmmmmmemo momo mo 465 11 23 2.3 Portland, Oreg.-Wash=--=--=--=--c-c--cooomoomonom mmm mmm 2,130 29 67 1.8 Providence, R.I!occmmmooommm mmm o mmm 2,012 21 28 0.3 Provo-Orem, Utah-----------cmeemoccmmecme meee mm meen mmm 180 31 - -14.,7 Pueblo, Colo-=-======----o-mememmmmmmmm—eememmemmo mmm 245 108 1 -30.3 Racine, Wis=====c-=-omcccmocoommmmmmmmmmommmomommmmommommmm 320 12 7 «1,5 Raleigh, N.C---e=--o-m-mmommmmcccccccem omen mm mmm mm mmm mmo 267 40 13 -8.8 Reading, Pa------===----m-emecccecccemm mmm m mmm mmm mmm me 852 29 5 =2.7 Reno, Nev=--w-=-rccccmcununmmnmnn meee eee e mca 158 21 1 -11.2 33 Table 12, Number of matched deaths reporting same standard metropolitan statistical area (SMSA) on both census and death records, on census record only, and on death record only; and net difference rates for total matched, matched at usual place of residence, and not at usual place of residence: United States, May-August, 1960—Con. [census record used as a base. Minus (-) sign indicates more assignments by census than by NCHS. For definition of areas, see Technical Appendix] Number reporting Net area differ- Same Census | Death ence on record | record rate both only only Richmond, Va=-=-- omc Loo 830 24 34 1.2 Roanoke, Va ==== mmm oo oe eee ee eee 316 41 17 -6.7 Rochester, N., Yee coco mmm mee 1,631 33 33 - Rockford, Ille=--omoo momo eeeeeee am 467 4 12 1.7 Sacramento, Calife----ccccmmm mmo __ 918 7 56 5.3 Saginaw, Mich==e ooo mmo ooo. 436 3 11 1.8 St. Joseph, MOo=-=-m mmm mm eee 263 39 2 -12.3 BE, LOULS, MO, wh LL wr imm sm meme mimmres ss otros se sm oo om emesis om 4,881 42 47 0.1 Salt Lake City, Utah 652 15 20 0.7 San Angelo, TeX===== == mmm ememme em 114 10 13 2.4 San Antonio, TeX===-c=- commen 1,153 64 ie) -4,3 San Bernardino-Riverside-Ontario, Calif--=---coocomooaooooo 1,570 58 37 ~1.3 San Diego, Calif---=c-oomo mmm 1,694 26 33 0.4 San Francisco-Oakland, Calif---=---cmemommmm oom 6,234 65 276 3.3 San Jose, Calife-ccco cm omm mmo 1,080 83 7 -6.5 Santa Barbara, Calif----c-eemommoomo ooo 331 1 10 2,7 Savannah, Ga==---==-- comm eao- 399 1 13 3.0 Scranton, Pa=---e-ooo moon 835 11 22 1.3 Seattle, Washe-- oom moomoo eo 2,384 19 88 2.9 Shreveport, La--====cc comme eee memeeeee 562 8 3 -0.9 Sioux City, Iowa--=--=--===-- mmm meee o 299 5 13 2,6 Sioux Falls, S.Dak=====c-cmmo ooo eeme eo 174 = 3 1.7 South Bend, Ind-=--=--c- mmm eee a 537 3 11 1.5 Spokane, Wash== === coon ooo 701 42 15 -3.6 Springfield, Ill=--mm comme cme el 440 9 24 3.2 Springfield, MoO-=--=--oc mmm 290 7 6 -0.3 Springfield, Ohio=-=-=-ccomoo momma 322 18 7 -3.2 Springfield-Holyoke, Mass! === omooo momo. 1,357 33 22 -0.8 Steubenville-Weirton, Ohio=-W.Va=====coemmoommmm 391 7 19 3.0 Stockton, Calife-e-co comm momma 528 33 13 -3.6 Syracuse, N.Y --- como mmo eee 1,450 20 63 2.9 Tacoma, Wash=== == cocoon momma 658 80 7 -9.9 Tampa-St. Petersburg, Fla-------ccommmmomommommmeeooo 2,440 40 37 -0.1 Terre Haute, Ind-==-----ommmm momma 329 2 8 1.8 Texarkana, Tex. =ArK==-- o-oo mmo 161 3 14 6.7 Toledo, Ohio=====ccmmm mmm eee 1,141 23 24 0.1 Topeka, Kans -=----m como momo ol 300 29 7 -6,7 Trenton, N.J--c coool 664 27 15 wl, 7 Tucson, Arize------coc comme eo 445 12 4 -1.8 Tulsa, Okla-=== como mmm moe emcee 840 10 37 3.2 Tuscaloosa, Ala 143 87 1 -37.4 Tyler, TeX=====- ooo eee eee eemecmeem eee 143 23 1 -13.3 Utica-Rome, N.Y 872 122 8 “11,3 Waco, TeX=mm=mm meme me ee ee eee meen 322 22 10 -3.5 Washington, D.C.-Md.-Va====mm moomoo emmmeea 3,612 65 86 0.6 Waterloo, Iowa---------cccmmmcoccaaano- 264 8 3 -1.1 West Palm Beach, Fla 563 8 11 0.5 Wheeling, W.Va. -Ohio 500 4 57 10,5 Wichita, Kans ==--=-oo mmm eee Lo 602 7 16 1.5 Wichita Falls, TeX==--=---=cc mmm cememcmmmccen 215 55 7 -17.8 34 Table 12. Number of matched deaths reporting same standard metropolitan statistical area (SMSA) on both census and death records, on census record only, and on death record only; and net difference rates for total matched, matched at usual place of residence, and not at usual place of residence: United States, May-August 1960—Con. Number reporting Net differ- Area Same Census | Death ence on record | record rate both only only Wilkes-Barre—Hazleton, Pa=-----=--ceceoeoccccacoccomoonnnon 1,168 18 19 0.1 Wilmington, Del.-N.J----=-=-meeee-mmemccm emcee ee — ccm ———— 756 25 32 0.9 Winston-Salem, N.C-------ec-emmmccmcceccooocmcm mmm mmm mmm 325 5 5 - Worcester, Massl--=--ec--comccocmmmmmmmm mmm memoem ome 1,652 92 24 -3.9 York, Pa-------cc-ememmmomeocoeoocoooomommooomoomoooomoo 545 8 33 4.5 Youngstown-Warren, Ohio------==-==---c-c-cc-cmooonmnmnmoononn- 1,116 7 31 2.1 All other areas=--=======me-ceccecememcemomomoococooooo-oo-o- 156,743 3,309 2,482 -0.5 Usual place of residence All areas==----=-=-=e-mmmmeccccco----soesoo-oo--------- 387,763 991 991 Total SMSA'S=-----------ecmommmmmmcmmmomoomcoocoommom mmo 240,397 633 678 0.0 Abilene, TeX=--=----=-=----m--cmmcomcooo-oo-o-oo-oo-moooo-ooo 167 - - - BIEEOrr; OIE 0 mmm mm mm mmm rm smc mo 5. wi 1,010 3 - -0.3 Albany, Ga---------=--s-e-mem-mmeecememeee-ece-se—s-—e—me————= 124 - 1 0.8 Albany-Schenectady-Troy, N,Y=---------c--cocomommmmomoooooon— 1573) - 12 0.7 Albuquerque, N.MeX-----=-=---c--co-ecmoccmmmomoooonmoooo———— 360 1 - -0.3 Allentown-Bethlehem-Easton, Pa.-N.J-===----=-----ocomcmoonnn~ 1,330 6 - -0.4 Altoona, Pa----------cccmommmommmcooo memos mm mm mmmmmmm mmo 378 2 2 - Amarillo, TeX-------=-c--m-emmmemoococcooo-oocoo-o--o--oo-oooeo 208 1 2 0.5 Ann Arbor, Mich------cc-mccmomceoccom cmc m mmm mmm mmm 273 8 3 -1.8 Asheville, N.Ce---c-c-cccomommmmmmmmmmmomomomoooomommmmmo mo 257 1 1 - Atlanta, Ga----=c-cemcmmmmmmemm— memes cess e————————————— 1,802 6 3 -0.2 Atlantic City, N.J---------memmmmmccmccccccccm mmm m mmm mmm 462 1 - -0.2 Augusta, Ga.-S5.C-----=-==----------scecmemmosmomoooooooooo-oo- 327 - 1 0.3 Austin, TeX=-------------mmmmmmoccmcoo-ooo-coseo-o-oomooo-oooo 301 4 - -1.3 Bakersfield, Calif-------c--momccoocooomommmm mmm mmm mmm 460 - 2 0.4 BALELTIOTE, Mt wie we mo mw mm mmm mm ro oe 0 0 a 3,581 2 - -0.1 Baton Rouge, La-- 350 - 1 0.3 Bay City, Mich--=-=--e-mmoeccccocccmnonm mmm m mmm mmm 209 1 - -0.5 Beaumont -Port Arthur, Tex 538 6 12 1.1 Billings, Mont------======-c---e------o-ceoo-m-oo-o-o-sooooo- 133 - - - Binghamton, N.Y---------eemeemeeoememee meen eee em mea 494 2 - -0.6 Birmingham, Ala------====--ecememeccmcoooocoomommmmmooo moe 1,175 6 1 -0.4 Boston-Lowell-Lawrence, Mass! ---=---mmmmmmmooomoommm emo 7,844 26 29 0.0 Bridgeport-Stamford-Norwalk, Conn! ==--mmmmm mmm meee 1,456 2 2 - Brockton, Massl==mmmmmmoooomoooeee mmm mmem moomoo 619 7 8 0.2 Brownsville-Harlingen-San Benito, Tex----=-=-=-==------------ 215 - - - Buffalo, N.Y=-=-----o--m-mmmmmmomooeocooomomomomm mm mmmm mmo oe 3,230 14 6 -0.2 Canton, Ohio--------=-sceeemememeeeee—e eee ccccee ne —————————— 714 2 5 0.4 Cedar Rapids, Iowa---=--===-===-cc-ce-co-o-o-cooooooooooooooo- 227 - 1 0.4 Champaign-Urbana, Ill--=-=-=--=-------cce--o-cooooonooooomono- 180 1 1 - Charleston, S5.C------reeeemmmmemeee—eeeee cece ceen en ——————— 289 6 1 -1.7 Charleston, W.Va-=--=-c--cemomcoeoocommnoo- 422 1 1 - Charlotte, N.C--a-c---ememcmomccoonmonmomn 394 2 - -0.5 Chattanooga, Tenn.-Ga-=--=========-=--------coceo-—moooooooooo- 501 1 2 0.2 Chicago, Ill=-=---------eeemmmmoommcocoooooooomommmommmoo 14,683 18 46 0.2 35 Table 12, Number of matched deaths reporting same standard metropolitan statistical area (SMSA) on both census and death records, on census record only, and on ~ of residence: United States, May-August 1960—Con. .th record only; difference rates for total matched, matched at usual place of residence, and net and not at usual place [Census record useu as a base. Minus (-) sign indicates more assignments by census than by NCHS. F or definition of areas, see Technical Appendix) Number reporting Net differ- Area Same Census | Death ence on record | record rate both only only Cincinnati, Ohio=Ky==--=-c-m comme. 2,438 9 3 -0.2 Cleveland, Ohio-==-== momo mmm 4,241 3 12 0.2 Colorado Springs, COlO===-mmmmm mmo e 245 - - - Columbia, S.Ce-=-cmm mmm eee ooo 398 7 - -1.9 Columbus, Ga.=-Ala=-=====m cme eee eee eemmm—mme 305 1 3 0.7 Columbus, Ohio==- mmo oom o meee 1,377 5 1 -0,3 Corpus Christi, TeX----===-mcmmmm meme cece ccccccccceee em 312 2 - -0.6 Dallas, TeX==== === mo oe ee ee ee eo 1,822 19 v 3 -0.9 Davenport-Rock Island-Moline, Iowa-Ill====-ccooomooooao noo 612 9 - -1.4 Dayton, Ohio==-=cc moomoo eee 1,351 1 1 - Decatur, Ill----eccmm ome eee cecmmmmmmm 229 - 1 0.4 Denver, Colo==-=-c== moomoo emeee eo 1,863 2 - -0.1 Des Moines, IOWA =-======= ccc emeeemmaeoo 577 2 1 -0.2 Detroit, Mich=== mm ommm moomoo. 7,451 3 11 0.1 Dubuque, Towa -======= mmm meee meme 171 - 2 1.2 Duluth-Superior, Minn. -WiS=--m eco mo moomoo 718 8 3 -0.7 DUEIGUL, NO cio ion msi 5 0 5 0 5 Bri ee 222 - - - El Paso, Tex=--------- 427 2 16 3.3 Erie, Pa---=--ceee--- 624 - - - Eugene, Oreg----==--- comme eee ememmcee 231 - 2 0.9 Evansville, Ind.-Ky 429 - - - Fall River-New Bedford, Mass! -==-emommomm ome 896 1 6 0.6 Fargo-Moorhead, N.Dak.-Minn-------eccommme me meeeeeeoo 165 1 1 - Flint, Mich-----om momo meee me 711 2 - -0.3 Fort Lauderdale-Hollywood, Fla=====-=mecccmommooo eo eeeeemeeno 668 4 1 -0.4 POLE BHLLH, Bll mmm mms om om im sm 5 0 0 127 = 1 0.8 Fort Wayne, Ind--=-----ccommm mmm eeee cece 508 2 3 0.2 Fort Worth, TeX=-----cccmmmm meee ee eee 1,000 - 2 0.2 Fresno, Calif=----ocmmmmm cee oo 697 5 3 -0.3 Gadsden, Ala-=--==--- comme emmmemmemooo 148 3 - -2,0 Galveston-Texas City, TeX===-==--mmmemeemcmcceceemcceceeemeon 277 - 2 0.7 Gary-Hammond-East Chicago, Ind-------cecmmmmmccoeeeeeeeeeeoo 1,133 1 10 0.8 Grand Rapids, Mich=-=---mc comme ee oe 749 - 3 0.4 Great Falls, MOnt-------- common emo eee e eee 159 1 - -0.6 Green Bay, WisS-=-coooommmm mmm eee 253 3 1 -0.8 Greensboro-High Point, N.C 371 - 1 0.3 Greenville, S.C------c-ccmmmmmccceaaa 250 - 5 2.0 Hamilton-Middletown, Ohio=-==--c-ccmommm meme C 366 - - - Harrisburg, Pa=-------cccemm momma - 848 1 1) 1.2 Hartford-New Britain-Bristol, Conn! ---e-oeeocoomooooooo_. 1,487 5 2 -0,2 Honolulu, Hawaii---==-oocmommm mmm 577 1 2 0.2 Houston, TeX=-====--ccc comme mmm meee em 1,896 5 2 -0,2 Huntington-Ashland, W.Va. -Ky.-Ohio 515 2 5 0.6 Huntsville, Ala--------ccccmommaanann 144 1 - -0.7 Indianapolis, Ind=-=----omcommmm meee 1,539 1 22 1.4 Jackson, Mich=---commmo mmm cme 331 - 2 0.6 Jackson, Miss =--oo oom momma em 335 - 1 0.3 Jacksonville, Fla-=-=--cmmmcmm meme eee eee 784 i 2 0.1 Jersey City, N.J=--cmcommm mmm 1,769 4 7 0.2 Johnstown, Pa=--e=emcmm oom om eee 773 1 2 0.1 36 Table 12. Number of matched deaths reporting same standard metropolitan statistical area (SMSA) on both census and death records, on census record only, and on death record only; and net difference rates for total matched, matched at usual place of residence, and not at usual place of residence: United States, May-August 1960—Con. [census record used as a base. Minus (-) sign indicates more assignments by census than by NCHS. + or aefinition of areas, see Technical Appendix) Number reporting Net differ- Aven Same Census | Death ence on record | record rate both only only Kalamazoo, Mich-----=--=--=-eccecmmmmmmmomm mmm mmmm mmm mmm mmm 310 1 1 - Kansas City, Mo.-Kans----=-===-----------c--=-o-oocooooomomoo- 2,238 4 7 0.1 Kenosha, Wis----=-===-memmeeeemee mm mem mee mm em mm meme mmm mmm 192 5 5 - Knoxville, Tenn------=-===-==-m--------------eoooooom==mo===== 573 1 1 - Lake Charles, La----===e-=meme--e--emeeece cme e me ————————- 185 - - - Lancaster, Pa------=------m---------o--omomsmoooooomoomomsses 606 11 3 -1.3 Lansing, Mich-=----=-=-m-omomooomomoomom mmm m mmm mmm mm mm mmm 523 - - - Laredo, TeX===-=====-===m=--eme-m-------e-e—-—-o-oeoooooso--ssos 118 - 1 0.8 13S Vegas, Nav—— «ecm wm mmm mmm mm cm oe a tm om mm a on 198 - - - Lawton, Okla---==-======se-eeeeeem——e—eeeceese-————————————— 99 1 5 4.0 Lexington, Ky------=-======-e=-cc-mocoommmomooooooommm monn 279 4 - -1.4 Lima, Ohio=--=--=-=---=-m-mmemoceoooooommmomomomommmm mmm 215 - 1 0.5 Lincoln, Nebr--------=-meecemcmemmmoem mmm meme m mmm mmm mmm 340 13 2 -3.1 Little Rock-North Little Rock, Ark-----------ce-eomooocco--m- 391 3 2 -0.3 Lorain-Elyria, Ohio=-=—==emwmmmmmm nn on = nm 378 2 1 -0.3 Los Angeles-Long Beach, Calif----------------------oomooomom- 13,321 10 16 0.0 Louisville, Ky.-Ind--=--------===-----cocmnoommmmmmmmm mm mmm 1,554 - 4 0.3 Lubbock, TeX-===----emeccmmmmm mm es cese em ——————————————— 179 - 7 3.9 Lynchburg, Va---=----====-c---momcommmmmomomoo mmm m mmm mmm 202 3 - -1.5 Macon, Ga=-==============--m-m--------o-—---ooo-soosos-osossss 332 1 3 0.6 Madison, WiS=====-==m---c-emccmomomomomommoom—mmmmoooommoo 391 1 5 1.0 Manchester, N,H!---==----omommmmmmmmmom mm cmmo momo m mmm 443 - 1 0.2 Memphis, Tenm-------====-=----c-------------oooooooooo-soss-os 1,232 - 3 0.2 Miami, Flamm===========m==-=-==e=====—-==-seem-—e-—e——ese=- 1,916 2 1 -0.1 Midland, TeX=-==-=====----===-mmmm-----co---------o----o-o-o- 61 - - - Milwaukee, WiS==-===-c--=----------o-ooooooomoooooooooommomms 2,635 4 6 0.1 Minneapolis-St. Paul, Minn------~--==-==-==--e=ooe===-e=—==-=—- 2,998 5 3 -0.1 Mobile, Ala-==========-------mm=mme--—-em—e--—--—o--mosomes- 509 1 2 0.2 Monroe, Las==m=====-sceccsmsemee—— meses esem ee. —————— 194 3 2 -0.5 Montgomery, Ala------====-==--=---------------ooo-ooo-oossosss 274 1 3 0.7 Muncie, Ind=-------==------------c----omooooooosmomoosmsmmsses 238 - - - Muskegon-Muskegon Heights, Mich-------=--=--c-c-con-coommmmmon= 320 - - - Nashville, Tenn--------==-====-====--------=----=---o-oo-o---o=- 736 4 2 -0.3 Newark, N,J mmm mmm oe om mmo mow m mmo om om 0 0 0 0 a = 3,920 9 13 0.1 New Haven-Waterbury, Conn!-----cocoemmomoooommmomcoommm mo 1,610 3 11 0.5 New Orleans, La=--------==-=-==-=-----c---=-=---=o--ooo-ooo===s 2,104 4 2 -0.1 Newport News-Hampton, Va-------==-=-=-====-==--=c----------=--- 335 - 1 0.3 New York, N.Y==-------m-m-mmmmceeomommomm momo momo mm mmm = 26,017 20 17 -0.0 Norfolk-Portsmouth, Va---=-=-memmemeemmeee meee meee mmm mmm 949 - - - Odessa, TeX=--=--=====-m--m--=----=------osooooooossosososossss 101 - - Ogden, Utah----crmecemmmmme mee —————— = = “--- 169 - - - OR1LANOME CLEY, ORLA «ww mmm mm mim icin cm sn n/m 2m mm ie 1 mm 2 871 7 7 = Omaha, Nebr, ~lowa coc cmmmm mmm mo iat uo 0 mm i 10 2 wm oe 922 2 21 2.1 Orlando, Fla======--==---m=--m=-m-c-o--s-oso----ooooooooooos-os 523 2 3 0.2 Paterson-Clifton-Passaic, N,J=--===-===---------o=mmmoooooooo- 2,529 13 5 -0.3 Pensacola, Fla==--==----c----omm-ooomocoooomomoomoooommoommo 306 3 2 -0.3 Peoria, Ill--===-=====-----eememe———mem mmm mme—— mmm 599 1 - -0.2 Philadelphia, Pa.-N,J------====--==--------------o-o-oooooo-os 10,302 25 23 -0.0 Phoenix, Ariz-------=-=-------------------omoooo-oo-ooosmsmsos 955 7 7 - Pittsburgh, Pa=-------==--m----------om-m-moomoooooomoomomss 5,645 13 9 -0.1 37 Table 12, Number of matched deaths reporting same standard metropolitan statistical area (SMSA) on both census and death records, on census record only, and on death record only; and net difference rates for total matched, matched at usual place of residence, and not at usual place of residence: United States, May-August 1960—Con. [Census record used as a base. Minus (-) sign indicates more assignments by census than by NCHS. F or definition of areas, see Technical Appendix] Number reporting Net differ- Lo Same Census | Death ence on record | record rate both only only PEELRTLOUA, MAGE wie mem oom sro toi i i 78 5785 mm me ir rar 403 ® g - Portland, Maine! -=--e-ommomm ooo _____ 459 2 - -0.4 Portland, Oreg.-Wash 1,968 - - - Providence, R, IT mi mmmmm me mmm me mm mmm mmm mm mo oe om in mm mo 50 1,809 6 i -0.3 Provo-0rem, Utah==--= o-oo mee. 168 1 - -0.6 Pueblo, C0lo== === mmm mmm ele 228 6 1 -2.1 Racine, Wis == -=- comm. 283 2 2 - Raleigh, N.C=m=mm momo... 266 - = * Reading, Pa=--m==cooo ooo o_ 804 1 - -0.1 Reno, Nev---- oe ooo eee eeee 150 - - - Richmond, Va----cocom mmo 814 5 3 -0.2 Roanoke, Va === coo mmm ooo. 309 - 2 0.6 Rochester, N,Y--oooooo mm cccmoo 1,447 8 1 -0.5 ROCKEOTrd, Ill=== comm moomoo emeeceee 442 1 - -0.2 Sacramento, Calif---- comm omm oe. 879 1 11 1.1 Saginaw, Mich-- === comm mmo. 422 - 1 0.2 St. Joseph, Mo=-==c comme ello 233 - - - St. Louis, MO. =T1l===c comme emcees 4,579 11 3 -0.2 Salt Lake City, Utah===-mememcccoe meee eeeeeeeececmmmmmm 613 2 2 - San Angelo, TeX === =m mmm emmmecmceeo 109 - 1 0.9 San Antonio, TeX == === =m mommies 1,097 1 - -0.1 San Bernardino-Riverside-Ontario, Calif--------cmemomocooonoo 1,495 10 4 -0.4 San Diego, Calif=-=--comomom mmm... 1,571 10 1 -0.6 San Francisco-Oakland, Calif---=-e-eoooooomoo ooo __. 5,892 10 14 0.1 San Jose, Calif-=--coocmmmm eos 1,015 7 3 -0.4 Santa Barbara, Calif=------ocmmmmm oe. 323 1 2 0.3 Savannah, Ga===- == =o cmon. 390 1 - -0.3 Scranton, Pa=-- =o coon. 756 2 5 0.4 Seattle, Wash=-m= coco om ooo aos 2,160 10 3 -0.3 Shreveport, La=====--mcmmo meee eeemmmmmmema 544 5 1 -0.4 Sioux City, LOoWa====== =m mmmmmee 269 - 2 0.7 Sioux Falls, S.Dak 167 = - - South Bend, Ind-==---cemm ome ao 508 3 3 - Spokane, Wash=-- =o comom eo... 612 4 5 0.2 Springfield, Il1l-----oom mmm eeeooe 434 - 9 2.1 Springfield, MO== === ccm mmm ceceoa 270 - 1 0.4 Springfield, Ohio--=--=-mo mmm _o_- 298 - 2 0.7 Springfield-Holyoke, Massle meccecmcmrecennrmmnun meee enn 1,300 3 5 0.2 Steubenville-Weirton, Ohio=W,Va====-meeooomoooomooo oo. 383 2 5 0.8 Stockton, Calif==-=-co comme. 496 1 2 0.2 Syracuse, N.,Y--sooo oo oom loooos 1,333 3 3 - Tacoma, Wash==-= oom moe ao 594 3 1 -0.3 Tampa-St. Petersburg, Fla---=--eeoemomoomo ooo 2,303 7 12 0,2 Terre Haute, Ind=-------cm mmm moe 328 - - - Texarkana, Tex. =-Ark--- == ccm mm moe aoaooo 154 3 - -1.9 Toledo, Ohio 1,074 - 5 0.5 Topeka, Kans 274 - - - Trenton, N.J 627 - 1 0,2 Tucson, Ariz 433 1 3 0.5 Tulsa, Okla-====--c ooo meee 823 3 4 0.1 38 Table 12. Number of matched deaths reporting same standard metropolitan statistical area (SMSA) on both census and death records, on census record only, and on death record only; and net difference rates for total matched, matched at usual place of residence, and not at usual place of residence: United States, May-August 1960-~Con. [census record used as a base. Minus (-) sign indicates more assignments by census than by NCHS. For definition of areas, see Technical Appendix] Number reporting Net differ- Ares Same Census | Death ence on record | record rate both only only Tuscaloosa, Ala--=--------cm--ommmoooomom mmm momo mmm mmm 137 - - - Tyler, TeX=---==--mm-mmmmeem ecm eeemm mmo —-————--—--——————-—- 131 - - - Utica-Rome, N,Y-----c--commmmmmmmmmm com mmmmmmmmmmm mmm mmm mmm 778 5 3 -0.3 WAGE, TEM mom oc er mm a om mm mm om 0, 0 303 5 - -1.6 Washington, D.C.-Md.-Va===-===-------moomoomommmmm mmm mmm 3,397 16 3 -0.4 Waterloo, Iowa-----=-cecmcrremmem meee — ee —————————— 259 2 5 1.1 West Palm Beach, Flaw--c-ecemmmmmmmee meen c eee mm mmm oe 542 2 - -0.4 Wheeling, W.Va.-Ohio==-==-=-=c=c-c-coocommoomo mmm mmm monn mmm 470 2 14 2.5 Wichita, Kans=-----==-c-m-mcooommcmm mmm mmmm mmm mmm mmm mmm 601 5 2 -0.5 Wichita Falls, TeX=---==----mmmeeeee—ee— meee eee —e— eee ——————— 204 2 6 1.9 Wilkes-Barre-~Hazleton, Pa-===wrmmm===mmmmmmemmmecccn nm ———— 1,137 7 7 - Wilmington, Del.=N.J=---------c---m-ooocmoomomooommm mmm mm mmm 718 - 1 0.1 Winston-Salem, N.C----=--ccm--ommmomooomommmommmmo monn mmm mmo 324 - - - Worcester, Massl--------ommcmmommommmmmmom momo mmmmmm mmm mmo 1,485 9 5 -0.3 York, Pam--=-=-=----mmmmmmmmmmmmmmmemmme me memmmmm— mmm 520 4 - -0.8 Youngstown-Warren, Ohio-=------=-===------o-ooocommomommomono- 1,083 2 1 -0.1 All other areas==-=-===-=-c------m--mmmmemm—mo—o--oo--------- 147,366 358 313 -0.0 Not at usual place of residence All areas==-=-=====mc----e-mmmmmmmeemmmo——-o------o- 24,434 7,104 7,104 . Total SMSA'S=m-mmmmmmmm==mm--eeemmmmeee-mmmmmmemm——————nn 15,057 4,153 | 4,935 4,1 Abilene, TeX==-===--------cmoommmmmmmom—ooo-o----------o- 12 4 9 31.3 Akron, Ohio=----=----ceomommmmmm moomoo mmooomomoooommmmmmm oo 32 21 46 47.2 ALBANY, IG mmm eee erm warms ce wacom mm mm mr oe 3 2 - - v Albany ~Schenectady Troy, N.Ymmmmmmmmm mmm mm on 0 10 0 wm 0 im 102 19 64 37.2 Albuquerque, N.MeX--===-=-----------co-ooooooooooo----o-oo-o- 4 2 3 16.7 Allentown-Bethlehem-Easton, Pa.-N.J===--c------ooocnomoommmnn 68 8 16 10.5 Altoona, Pa----------ereeccmmmcc meme mmm mmm ———— 26 19 2 -37.8 AMBTLLLO, TEM mi mim rc ai se om wm mm mm mt eo 0 0 0 2 14 5 -56.3 Ann Arbor, Mich=-------cmommmocmm moomoo mmm momo 15 62 6 -72.7 Asheville, N,Ce-=--ccocmcmmmmmmm momo mmm momo momo mmmmo oe 7 21 5 57.1 Atlanta, Ga=—=-—=-ccec ccm mm 67 18 59 48,2 Atlantic City, N.J-----=--ceemecmemeeec eee m meme em mm meme ——— 55 5 17 20.0 Augusta, Ga.=-S.C-------=-=-----------ooooo---o-----------o- 23 14 11 -8.1 Austin, TeX=-=-=-==c-=--c----mmmmm—eoeoo--o----o----------e 28 51 6 -57.0 Bakersfield, Calif-----------oommmmmcmcmmee meme mmm mo 7 - 14 200.0 Baltimore, Md------==--------m-mmmmeo—eom—o-oo-o------------ 274 62 55 -2.1 Baton Rouge, La-------=------m--m-m-ooooooooooooooo-o-mmoooo 6 7 11 30.8 Bay City, Mich--------c-errerceeememe remem c cece ——————— 15 - 8 53.3 Beaumont -Port Arthur, Tex=-------=--------------------o--o-=o- 8 2 6 40,0 Billings, Mont=----=-=-=-----mme-mmeme——m---e--o-----——--o- 1 - x 100.0 Binghamton, N,Y=-----c--moommmmommmmmommmmmoommomom—mmmm moe 28 30 14 -27.6 Birmingham, Ala-----==-----------mc-oomooommmoo-o-o-o—oo- 25 2 36 125.9 Boston-Lowell-Lawrence, Mass! —=----ccmcmmommmomommm cme 869 45 100 6.0 Bridgeport-Stamford-Norwalk, Conn! ==cmcmmmmmm meee 76 26 32 5.9 Brockton, Massl-----mommmmmoo moomoo eee mmm 73 13 20 8.1 39 Table 12, Number of matched deaths reporting same standard metropolitan statistical area (SMsA) on both census and death records, on census record only, and on death record only; and net difference rates for total matched, matched at usual place of residence, and not at usual place of residence: United States, May-August 1960—Con. [Census record used as a base. Minus (-) sign indicates more assignments by census than by NCHS. F or definition of areas, see Technical Appendix] Number reporting Net differ- Ares Same Census | Death ence on record | record rate both only only Brownsville-Harlingen-San Benito, TeX-==--=-mmemeeeeeecmccaanna 8 5 7 15.4 Buffalo, N.Y--m-ooom mmo ee eememce een 238 43 47 1.4 Canton, Ohio====-o moomoo... 50 40 4 -40.0 Cedar Rapids, IoWa=-== =m mmm om o_o. 14 - 2 14.3 Champaign-Urbana, T1l-----cemmmmmm oo ________ 12 1 12 84.6 Charleston, S.C==-=== =m m meee 6 - 11 133.3 Charleston, W.Va=---=oom oom... 14 2 8 37.5 Charlotte, N.C---o-m ooo. 11 2 2 - Chattanooga, Tenn. -Ga=--=======c comme ______ 11 5 8 18.8 Chicago, Ill----mmm meee eme eee 821 73 331 28.9 Cincinnati, Ohio-Ky-==-mmummeccamcm cece mmm nena n—nn————— 256 23 26 1.1 Cleveland, Ohio=======-eceomomoo mmm emma - 165 9 82 42,0 Colorado Springs, COLlO=======m moomoo. - 13 19 46,2 Columbia, S.C==--oommmm meee. 34 117 - «77.53 Columbus, Ga.=Ala==== == mmm cee 4 5 16 122,2 Columbus, Ohio====== mmm... 113 30 23 -4,9 Corpus Christi, TeX=-===-meomom ome iiemoae - 5 11 120.0 Dallas, TeX====== === oo ee eee 58 24 51 32.9 Davenport-Rock Island-Moline, Iowa-Ill--==--=cmcmmmomcmcaunonon 21 23 10 -29.5 Dayton, Ohio===== ooo oom eos 107 36 9 -18.9 Decatur, Tll=-==-oo como ome eo. 8 5 4 “le? Denver, Colo== == === me ooo 79 20 66 46,5 Des Moines, IOWA ===== === mmo. 45 1 18 37.0 Detroit, Mich=--o oom ooo. 397 23 130 25:45 Dubuque, IoWa==-= === ===. 15 23 2 «55.3 Duluth-Superior, Minn. -Wis==-===mmooomomoo oo ___ 23 8 34 83.9 Durham, N.Ce=-m mmm mmm meee eee eee eieeee ee = 1 11 1,000.0 El Paso, TeX====-mm mmm mmo oe eee eee 12 5 6 5.9 Erie, Pa----cm coco momma 25 3 21 64.3 Eugene, Oreg=----c-oommmm meio. 7 - 7 100.0 Evansville, Ind. -Ky-=====mm-eoeeoo om meeeeecmcccccecmme 11 19 2 -56.7 Fall River-New Bedford, Mass! 102 12 10 -1.8 Fargo-Moorhead, N.Dak.-Minn------=cceecmoooaoo_o 5 6 13 63.6 Flint, Mich=e==m momo oem 9 7 14 43.8 Fort Lauderdale-Hollywood, Fla======-=meeeoccommcmmcemeemmee 34 22 4 -32.1 Fort Smith, Ark-----ceoomm memo. - 5 2 -60.0 Fort Wayne, Ind-------ocmmmomimcmeoo. 13 23 24 2.8 Fort Worth, TeX----=c-mmm mmo ccmcmme 25 14 26 30.8 Fresno, Calife-eoeooom om mmm eo ._ 18 1 10 47.4 Gadsden, Ala=-=--=- mmm 2 7 3 -44 4 Galveston-Texas City, TeX=--=-==-e--cecmeeccocecmeceemcm————— 1 3 12 225.0 Gary-Hammond-East Chicago, Ind=---=--=ccoommommoooo o_o. 33 - 25 75.8 Grand Rapids, Mich=====c-- mmo... 59 34 17 -18.3 Great FALLS, MOTE === =m mm we me te oe mmm mm mm em 21 5 3 wy? Green Bay, Wis === moomoo mmm cmeao. - 5 11 120.0 Greensboro-High Point, N.,C=-=m-memmmom moomoo. 21 2 8 -9.4 Greenville, S.C-====m moomoo ao. 19 - 7 36.8 Hamilton-Middletown, Ohio======-mmom mmm emeeeeaooos 34 - 7 20,6 Harrisburg, Pa--=--c-cemomm me _.. 37 38 18 -26.7 Hartford-New Britain-Bristol, Conn!=--=-e-eooooooooooooooo__ 64 11 39 37.3 40 Table 12. Number of matched deaths reporting same standard metropolitan statistical area (SMSA) on both census and death records, on census record only, and on death difference rates for total matched, matched at usual place of residence, of residence: United States, May-August 1960-—~Con. record only; and net and not at usual place [Census record used as a base. Minus (-) sign indicates more assignments by census than by NCHS. For definition of areas, see Technical Appendix] Number reporting 2 Net iffer- Area Same Census | Death ence on record | record rate both only only HONONUIN, AWA wm mm mm mm sm me se = 25 5 - -16.7 Houston, TeX=--=-=-===-=====-c---—====--=--so-oooo--oooososoooos 38 13 29 31.4 Huntington-Ashland, W.Va. -Ky.-Ohio-=-===---==----==-=====--==== 14 21 10 -31.4 Huntsville, Ala------==-======--==-===--=----------s-oosossoooos - 5 4 -20.0 Indianapolis, Ind 134 18 27 5.9 Jackson, Mich=-==----mm-mmooomooommmmmomoooom mom mmm mmm mmm 14 5 13 42.1 Jackson, Mis§=-==--=-==--==-------------ms-ooooooosossossooos 11 22 16 -18.2 Jacksonville, Fla 55 12 20 11.9 JErSEY CLEY, F.Jiwwmomm mm mm mio sm mcm em nl ho mo 96 5 53 47.5 Johnstown, Pa---=-==--=--meee-eemmm--—memee—eo mmm mmm 40 20 18 -3.3 Kalamazoo, Mich----==-===-sememmeeen-een———- 12 28 1 -67.5 Kansas City, Mo. -Kans 92 22 87 57.0 Kenosha, Wis-----=--=me=meommmmmmmmmm meen mmm mmm mmm mmm mm mm mm - - 5 - Knoxville, Tenn--------========-=-=----==-=--=--c--oo-oo=-o-ososs 49 35 7 -33.3 Lake Charles, Laws mmm mmm mmo oo of ini om mon imo ni on a - - 7 - Lancaster, Pa=====-=-------m==------------oossoooooooomssoossos 50 53 21 -31.1 Lansing, Mich------ 45 12 16 7.0 Laredo, Tex--=------ - - 7 - Las Vegas, Nev 1 - 9 900.0 Lawton, Okla----=-==========-c---=-=--=-==-----o-o---soosoooooos - - 1 - Lexington, Ky-------=======------===-===-oooooo-oooosooosooos 13 36 6 -61,2 Lima, Ohio=======-=-===--==-------------oooooooo-ommoommmmoos 13 14 20 22.2 Lincoln, Nebr=-----=---------cmo-omm-ooosooosomoomms mmm mmo ms 15 8 1 -30.4 Little Rock-North Little Rock, Ark------=---=-===-=-=----==-=-===-- 19 29 2 -56.3 Lotain-~Elyris, Ohio me wmmmmm mm om mm immo mm fom i me mm 16 15 6 -29.0 Los Angeles-Long Beach, Calif 1,262 80 187 8.0 Louisville, Ky.-Ind--=m====esenenne—- 8 37 12 -19.8 Lubbock, TeX======--========-===----=--=---=s-oooooooomsossooooos - - 17 - LITICHDUEE, Viais =e mmm mm sm mm oc i 0 5 13 19 12 -21.9 Macon, Ga---============meeem=—m-eee——mee———esse———=ssss ese 4 15 8 -36.8 Madison, Wis==-=--==-=-==-=-------o-m=-o-ocooooomsomsoooooooo 35 22 9 -22.8 Manchester, N.Hle--oomoommmcommmmmmmoom mmo m momo mmm mm mm 7 2 9 77.8 Memphis, Tenfm=mmm=———=nennn 65 22 15 -8.0 Miami, Fla------=---==----- 68 1 15 20.3 MUALATIA, Tim mmm mm cm mm mm mm mo 0 0 0 1 0 a 2m = - - 6 - Milwaukee, Wis=---==---=-==--------omm=mmoomsoooooooomoosmoos 127 19 27 5.5 Minneapolis-St. Paul, Minn 209 60 87 10.0 Mobile, Ala==--===-=======-=-=-----==----=-ooo-moooomoosos 11 34 19 -33.3 Monroe, La--============-m-m=-------m--eoooooosooosssossooos 1 - 7 700.0 Montgomery, Ala----========-==-=------sssssso-oosooooosooooooos 6 15 7 -38.1 Muncie, Ind--==-===-========--------==------ooooo-ooo-sssoooos 4 - 6 150.0 Muskegon-Muskegon Heights, Mich-----=----------=--------==-="" 8 5 10 38.5 Nashville, Tenn----=--=========-==---=====--==o-o=--o--==-o-=-o= 83 45 2 -33.6 Newark, N,J===-------c---mm-me-oo---—m---—oosooo-oosossosooos 243 107 102 -1.4 New Haven-Waterbury, Conn! =--mmmmmm mmm mmm mmm 62 15 49 44,2 New Orleans, La-===-===-=-=====--=---=-o=—---==-oo-oooso==--s-c- 65 20 54 40,0 Newport News-Hampton, Va--=-=---===---========----=-=-=-===----=s 31 34 7 41.5 New York, N.Y-------so---mmmomoomommm—=oomomom mmm mmm momo 1,984 73 316 11.8 Norfolk-Portsmouth, Va---------====-==----o-=-=-ooo-omoooommm os 25 3 34 110.7 Odessa, TeX-======-==--========----o-----ossooooososoosssSoosos - - 2 - Table 12. Number of matched deaths reporting same standard metropolitan statistical on both census and death records, on census record only, and on death difference rates for total matched, matched at usual place of residence of residence: United States, May-August, 1960—Con. area (SMSA) record only; and net , and not at usual place [Census record used as a base. Minus (-) sign indicates more assignments by census than by NCHS. F or definition of areas, see Technical Appendix] Number reporting Net . differ- sven Same Census | Death ence on record | record rate both only only Ogden, Utah=---=oommmmm______. 10 x 4 27.3 Oklahoma City, Okla=======-ooooooomo oo ___________ 44 48 8 -43,5 Omaha, Nebr. -Towa--====-omm oo _________ 46 27 27 - Orlando, Fla=-=--=-oommmme oo _______ 32 3 3 5.7 Paterson-Clifton-Passaic, N.J 90 29 95 55.5 Pensacola, - 1 8 700.0 Peoria, T1lem--eeooommmm TT 45 51 8 -44, 8 Philadelphia, Pa.-N.,J==em-ooooooom oo __________. 629 69 111 6.0 Phoenix, Ariz------eeeeoooo oo _____ 39 35 15 -27.0 Pittsburgh, Pa------e-m oo ______ 416 44 45 0..2 Pittsfield, Mass! —~-ooommom oo ________ 15 6 10 19.0 Portland, Maine! ~==== mmm TT 6 9 23 93.3 Portland, Oreg.-Wash=-----eeoooo oo _______ 162 29 67 19.9 Providence, R.Il=-ooommmmmo ooo. 203 15 27 5:3 Provo-Orem, Utah-=-----oemmmoo oo ______ 12 30 - =71.4 Pueblo, Colo=====mmmm meee. 17 102 - “85.7 Racine, Wis=-=---ooomemmmmm oo ____ 37 10 5 -10,6 Raleigh, N.C--mooommemmm LL il 40 13 -65.9 Reading, Pa=-=-=----oemmmmm o_o __ 48 28 5 -30.3 Reno, Nev === oom 8 21 1 -69.0 Richmond, Va=--oooommom ee _.. 16 19 31 34,3 Roanoke, Va===-mem moo _____ 7 41 15 ~54,2 Rochester, N.Y----eoeemmoo oo. 184 25 32 3:3 Rockford, T1l----emmmmmm eo ___ 25 3 12 32.1 Sacramento, Calif=-==ommommmmm ee ______ 39 6 45 86.7 Saginaw, Mich=-=-m mmm. 14 3 10 41,2 St. Joseph, Mo==--oommmm em _____ 30 39 2 -53.6 St. Louis, MO. =Tll-=-omoommm mm ______ 302 31 44 3.9 Salt Lake City, Utah=---=--ommm oo _______ 39 13 18 9.6 San Angelo, TeX===-=== == mmo. 5 10 12 13.3 San Antonio, TeR=--===mmescccmm mc cmnee cee ————————————— 56 63 12 -42.9 San Bernardino-Riverside-Ontario, Califemcmcmm meee 75 48 33 -12,2 San Diego, Calif------memmmm oo ________ 123 16 32 11.5 San Francisco-Oakland, Calif 342 55 262 52.1 San Jose, Calif=----oommmmmm oo ________ 65 76 4 -51,1 Santa Barbara, Calif---memececeeccemmmeeeeeccenm—————————— 8 - 8 100.0 Savannah, Ga==-====-coceoooo______ 9 - x3 144.4 Scranton, Pa=-=---=c-ceeeoo_____ 79 9 17 9.1 Seattle, Wash=---ooomemmmmo oo _____ 224 9 85 32.6 Shreveport, La=-==-=--em ooo. 18 5 2 -13.0 Sioux City, IOWA =====m mmo ___ 30 5 11 17.1 Sioux Falls, S.DaK====mmm mmm ooo 7 - 3 42,9 South Bend, Ind=-=---=---mooo oo ________ 29 - 8 27.6 Spokane, Wash===- meme _______ 89 38 10 -22.0 Springfield, T1l-----meoommoo o_o ___ 6 9 15 40.0 Springfield, Mo------ommmmmm 20 7 5 -7.4 Springfield, Ohio-===mmmmomemom om ooo_ 24 18 5 -31.0 Springfield-Holyoke, Mass! =---oooooo oo _______________ 57 30 17 -14.9 Steubenville-Weirton, Ohio=W.Va====mmmeeooooooooo ooo oo_____ 8 5 14 69.2 Stockton, Calif-=----mmooo om ________ 32 32 11 -32.8 42 Table 12. Number of matched deaths reporting same standard metropolitan statistical area (SMSA) on both census and death records, on census record only, and on death record only; and net difference rates for total matched, matched at usual place of residence, and not at usual place of residence: United States, May-August 1960—Con. [census record used as a base. Minus (-) sign indicates more assignments by census than by NCHS. For definition of areas, see Technical Appendix] Number reporting Net differ- Aved Same Census | Death ence on record | record rate both only only BYTACUSE, NM, Yom mmm mim mio ce io 10 1 he 0 mm 0 0 0 mm 117 17 60 32.1 Tacoma, Wash=-==-----m-mcmcooomomomoomooooooomoomomm mm mmm mmo 64 77 6 -50.4 Tampa-St. Petersburg, Fla------=--====-----------==-ooo=--oooo- 137 33 25 4.7 Terre Haute, Ind----=-==----=-----------=-o-o-ocooooso=ooosoooos 1 2 8 200.0 Texarkana, Tex.-Ark------==--=--------==----ooooooooooo=-osmsss 7 - 14 200.0 TOLEAD, ORLO= === mmm mmm mm om 0 tm dm mm mm mm a mm = = 67 23 19 4.4 Topeka, Kans------========--------==---=-oo-ooooossooosoosoos 26 29 7 -40.0 Trenton, N,J--------=-===-----------=---osoosooooooossossooomos 37 27 14 -20.3 Tucson, Ariz----------==m==---------=--momosoooooooommemoooss 12 11 1 -43,5 Tulsa, Okla----==---======-===---=---===---o===-o--o==ooosooooos 17 7 33 108.3 Tuscaloosa, Ala 6 87 1 -92.5 Tyler, TeX---===================- 12 23 1 -62.9 Utica-Rome, N.Y 94 117 5 -53.1 Waco, TeXm=m=======mc====m=-m——--——-=-=-—osssoooooo-mssssoos 19 17 10 -19.4 Washington, D.C.-Md, -Va=mem==-c-=-eeecnmmmme—————e enn -————— 215 49 83 12.9 Waterloo, Iowa------==========------===-=-=--oo-oooo-sosoosoooos 5 6 - -54.,5 West Palm Beach, Fla-------===---------------=--oo--oo-ooooos 21 6 1) 18.5 Wheeling, W.Va.-Ohio=========-====--==--=-----oooo-mosmmosooe 30 2 43 128.1 Wichita, Kans===--====-======-==--=----==-=------ooooooo-msosoos 1 2 14 400.0 Wichita Falls, TexX=-----====-======-----====-==-----c-=o-s=-sooocs 11 53 1 -81.3 Wilkes-Barre—Hazleton, Pa-----=--==c-=----------m--oo=o-o== 231 11 12 2.4 Wilmington, Del.-N.J=-=-=--===--==--=---m-=-=-m-oooooo-==-oooooo - 38 25 31 9.5 Winston-Salem, N.C-------=-===-=----c-=--=—=-----oo-cooo=moooooos 1 5 5 - Worcester, Mass! 167 83 19 -25,6 York, Pas=---=m======-----mmmm---mmo—------mo-sooms-oossmssoos 25 4 33 100.0 Youngstown-Warren, Ohio 33 5 30 65.8 All other areas----=---==-=--=-====--=c---=---=--c------- 9,377 2,951 2,169 -6.3 IMetropoli tan State economic areas. 43 Table 13. Number of matched deaths reporting same standard metropolitan census and death records, census record only, and death record only; United States, May-August 1960 Census record used as a base. Minus (-) sign indicates more assignments by census than by NCHS. For definition of areas, see Technical Appendix] statistical and net difference rates, by color: area (SMSA) on both White Nonwhite Number reporting Number reporting Area Net Net differ- differ- Same Census | Death ence Same Census | Death ence on record | record rate on record | record rate both only only both only only All areas-=-=----cmccmmmme 366,216 7,308 7,308 eee | 42,579 723 723 we Total SMSA'S===-===omcmmmme 226,877 | 4,341| 5,056 0.3] 26,521 395 516 0.4 Abilene, Tex 168 4 9 2.9 10 - - - Akron, Ohio--------- 981 24 45 2.1 60 - 1 1.7 Albany, Ga--=-=---commeme 63 - 1 1.6 62 - - = Albany-Schenectady-Troy, N.Y--=====--cu-oc 1,791 19 75 3.1 37 # 1 2.7 Albuquerque, N, Mex---===--memooomuaoooooo 342 3 2 -0.3 20 5 1 S50 Allentown-Bethlehem-Easton, Pa.-N,J------ 1,375 13 16 0.2 7 1 3 =12.5 Altoona, Pa=--==---eommmmm 400 21 4 -4.,0 4 ¥ - - Amarillo, Tex=----=--ocemmmmmmm 195 14 7 ve 14 1 = -6.7 Ann Arbor, Mich--- 275 63 8 -16.3 13 5 1 "22.2 Asheville, N.C---- 227 16 5 -4,5 37 6 1 -11.,6 ALLONER, Qin mmm ims mmo mmm mmm sms simi 1,331 21 52 2.3 514 3 10 1.4 Atlantic City, N.J----c--mmmoommeoo 443 5 15 2.2 71 1 2 1.4 Augusta, Ga.-S.C 218 12 7 -2,2 124 2 4 1.6 Austin, Tex------------- 280 40 5 -10.9 49 15 1 21:9 Bakersfield, calif 417 - 14 3.4 46 - 2 4.3 Baltimore, Md---=--=ccmmmamaoao 3,002 38 51 0.4 823 26 4 -2.6 Baton Rouge, La=====-ccemmmmmmeeo_ 204 2 6 1.9 152 5 6 0.6 Bay City, Mich-----ccccmmmao 222 1 8 ad 1 " - - Beaumont-Port Arthur, 403 8 12 1.0 136 - 6 4.4 Billings, Mont=--==-mcmmmmomo 132 - 1 0.8 2 - - - Binghamton, N,Y--===ccommmeo oo. 508 33 14 35 13 - - - Birmingham, Ala=----=-cocmmmmam_ 678 4 25 3:1 510 4 12 1.6 Boston-Lowell-Lawrence, Mass!-===ccouoo_- 8,438 70 123 0.6 186 1 6 2.7 Bridgeport-Stamford-Norwalk, Connl------- 1,462 25 33 0.5 51 2 1 ~1.9 Brockton, Mass!-=-ememommmmmemee 674 18 27 1.3 13 2 1 -6.7 Brownsville-Harlingen-San Benito, Tex---- 219 5 7 0.9 3 * # - Buffalo, N.Y 3,272 56 51 -0,2 172 1 2 0.6 Canton, Ohio 718 34 8 -3.5 36 3 1 -5.1 Cedar Rapids, Iowa 240 - 3 13 1 - = 2 Champaign-Urbana, 185 2 13 5.9 6 - - - Charleston, S,C---mmemmcmemmcemmceeeeemm 176 1 7 3.4 117 4 5 0.8 Charleston, W. Va-=--==ccmmomooaaoo_____ 391 3 8 13 40 - 1 2:5 Charlotte, N.,C----=-muuaa- 254 4 - -1.6 151 - 2 1.3 Chattanooga, Tenn.-Ga 376 6 10 1.0 132 - - = Chicago, Tll-=----cooommmao_ 13,489 82 338 1.9 1,853 4 39 1.9 Cincinnati, Ohio=Ky-=====oocoomoooo___ 2,338 32 25 -0.3 332 - 4 1,2 Cleveland, Ohio----=-=coommm___ 3,854 11 83 1.9 527 1 11 1.9 Colorado Springs, Colo 235 12 19 2.8 7 = = - Columbia, S,C---=--=mcmmmmme 304 73 - =19.4 123 51 - ~29.3 Columbus, Ga.-Ala-----------= 195 5 13 4.0 114 1 6 4,3 Columbus, Ohio-======mmomommo___ 1,287 30 23 -0.5 191 5 1 2.0 Corpus Christi, Tex-----==--ceemeoooo___ 288 7 11 1.4 22 - = - Dallas, TeX===-==-occmmmmmmeee__. 1,559 38 42 0.3 307 4 12 2.6 Davenport-Rock Island-Moline, Iowa-I1l--- 617 32 9 oo 14 & 1 7.1 Dayton, Ohio=-=----cmmmmm 1,285 34 9 =-1+9 165 3 1 -12 Decatur, I1l-=--eee imme 224 5 3 -0,9 13 - 2 15.4 Denver, Colo----=----u_-_ 1,866 21 56 1.9 60 1 6 8.2 Des Moines, Iowa-- 593 3 18 25 23 - 1 he3 Detroit, Mich-----=ccuceao__ 6,734 24 118 1.4 1,049 2 21 1.8 Dubuque, Iowa====-==-coommm 185 23 4 -9.1 - - - - Duluth-Superior, Minn.-Wis 735 16 37 2.8 4 - - - See footnote at end of table. 44 Table 13. Number of matched deaths reporting same standard metropolitan census and death records, census record only, and death record only; United States, May-August 1960—Con. [census record used as a base. Minus (-) sign indicates more assignments by census than by NCHS. For definition of areas, see Technical Appendix] statistical and net difference rates, area (SMSA) on both by color: White Nonwhite Number reporting Number reporting Area Net differ- Same Census | Death Same Census | Death ence on record | record on record | record rate both only only both only only 159 - 10 62 1 1 - 425 7 22 11 - - - 630 3 20 18 - 1 5.6 BAGONG, ORG Rr =mimrimrimsmimm ot ms cot 0 cm et 8 235 - 9 2 - # - Evansville, Ind.-Ky---==-rermecemeen————— 398 19 2 36 - - - Fall River-New Bedford, Mass!---------=-- 968 13 16 22 - - - Fargo-Moorhead, N. Dak.-Minn------------- 169 7 14 - - - - Flint, Mich=-====mmmcemmemammmnn- -- 641 8 12 75 1 2 1.3 Fort Lauderdale-Hollywood, Fla-- -- 614 23 2 84 3 3 - Fort Smith, Ark-----=----cecmmmmmeneean—- 113 5 2 13 - 1 7.7 Fort Wayne, Ind----=-mm-recmcecec ce —————— 501 25 27 14 - - - Fort Worth, Tex------m=-ememeeceeee————— 875 13 23 141 1 5 2.8 Fresno, CALE Eww aos sms mm mr me mim 665 5 11 47 1 - -2.1 Gadsden, Alar ===rmmmmm mm 126 9 - 22 1 3 8.7 Galveston-Texas City, Tex---===mmmemee=-- 207 3 11 68 - 3 4.4 Gary-Hammond-East Chicago, Ind----------- 983 1 32 3.2 175 - 2 1.1 Grand Rapids, Mich--=--=--ceeeececceccaa" 769 33 20 -1.6 31 1 - -3.1 Great Falls, Mont--====rrreeecem eee —————— 176 6 3 -1.6 3 - - - Green Bay, WiS--===---moomcommmce cena 251 7 10 1.2 2 1 - -33.3 Greensboro-High Point, N.C---======c=-=-- 291 11 6 -1.7 100 - 3 3.0 Greenville, §.Cr=wemwmmmmnmmben ssi 194 ® 7 3.6 73 - 5 6.8 Hamilton-Middletown, Ohio=--=-==--=-==-=-- 376 - 7 1.9 19 - - - Harrisburg, Paw mmmm mms mmm meme mm 825 39 28 -1.3 55 - 1 1.8 Hartford-New Britain-Bristol, Conn!------ 1,483 15 39 1.6 53 1 2 1.92 Honolulu, Hawaii----=-==-m-ceceomceaaa—-— 159 2 - l,2 342 2 2 - HOUSTON, T@Xrm mmm mom om mam ow mw mm mm im tot tr ret om om 1,495 13 28 1.0 424 5 3 -0.5 Huntington-Ashland, W. Va.-Ky.-Ohio------ 486 22 15 -1l.4 34 1 - -2.9 Huntsville, Ala-==---=-=-e-eececee———————— 116 6 2 -3.3 27 - 2 1.4 Indianapolis, Ind--=---=-=--cccmoocaaonn- 1,400 19 48 2.0 263 - 1 0.4 Jackson, Mich=---==-cecmmmccmcc meee ccccam 33 3 15 3:6 11 2 - -15.4 Jackson, MisS==-=-=-m-mmememme ec ——————— 184 18 15 -1.5 157 4 2 -1.2 Jacksonville, Fla---=-==-=-e-ccmcmenmenna- 584 12 16 0.7 252 1 6 2.0 Jersey CLEY, NJ mmm mmm mmm mm mm mm mmm mm on 1,771 8 58 2.8 78 - 2 2.6 Johnstown, Pa-----=mmeeoceemcec cece —————— 801 19 19 - 9 2 1 -9,1 Kalamazoo, Mich--===m-memeemcc emcee meme 309 28 2 wZ.e7 7 A - -12.5 Kansas City, Mo,-Kang-==memmme mmm —-—————— 1,99) 18 89 5 314 3 3 - Kenosha, Wis--=-=-=--cccmmmcmc cc ccc ccm em 185 5 10 6 6 - - - Knoxville, Tenn-=-=--==m==ececcececeaaan-" 539 36 5 4 80 - 2 2.5 Lake Charles, La---------eeecceeeeeeee——— 130 - 5 8 55 - 2 3.6 Lancaster, Pa---=-=-=-ccmeemccccne cee eenm 646 64 24 6 1 - - - Lansing, Mich---==-mc--ccommmonme nme 555 12 16 0.7 13 - - - Laredo, Tex-~======~~- 113 = 8 7.1 - - - - Las Vegas, Nev 181 - 8 4.4 18 - 1 5.6 Lawton, Okla=--==-=s--cemmcecc emcee eee 90 1 6 5.5 9 - - - Lexington, Ky---====-=-cememmemcecccccana 225 35 6 11.2 67 5 - -6.9 Lima, Ohio=-=-====-=-cmmomm cece meee 217 13 21 3.5 10 1 - -9.1 Lincoln, Nebr------=----emeeeceeceeeeaa—- 343 19 3 -b4.4 12 2 - 14,3 Little Rock-North Little Rock, Ark------- 282 26 1 -8.1 125 6 3 -2.3 Lorain-Elyria, ONio=m=mmmm mmm mmm m—— 369 15 6 -2.3 22 2 1 4,2 Los Angeles-Long Beach, Calif------------ 13,584 84 187 0.8 872 6 8 0.2 Louisville, Ky.-Ind---===-===-cc---c----- 1,366 37 13 263 = 3 1:1 Lubbock, Tex-=====-==m-memceemcc ca ——————— 164 = 22 15 = 2 13.3 Lynchburg, Va=--=mesem ee em. 165 21 10 50 1 2 2.0 Macon, Ga=-====mmmemem-eeeme em ee ———————— 201 15 5 134 1 6 3.7 Madison, Wis--=-==-m-c-omommcm cc mcemceae 414 21 14 5 2 - -28.6 See footnote at end of table. » wn Table 13. Number of matched deaths reporting same standard metropolitan census and death records, census record only, and death record only; United States, May-August 1960—Con. [census record used as a base. Minus (-) sign indicates more assigninents by census than by NCHS. For definition of areas, see Technical Appendix] statistical area and net difference rates, by color: (SMSA) on both White Nonwhite Number reporting Number reporting Area Net Net differ- differ- Same Census | Death ence Same Census | Death ence on record | record | rate on record | record rate both only only both only only Manchester, N,H======cocccmmmcmonmmoe 450 1 10 2.0 - 1 - -100.0 Memphis, Tenn---- 762 19 15 -0.5 523 3 1 -0.4 Miami, Fla------- 1,751 2 10 0.5 207 1 6 A Midland, Tex-------- 57 - 6 10.5 4 - - - Milwaukee, WiS==-==c-coomcmmmmoa oo 2,659 22 33 0.4 96 1 - -1.0 Minneapolis-St. 3,113 65 90 0.8 78 - - - Mobile, Ala=-===---ccccm cece 327 1 18 5,2 187 34 3 -14.,0 Monroe, La---------- 96 3 4 1.0 93 - 5 5.4 Montgomery, Ala 127 11 4 ~5:1 150 5 6 0.6 Muncie, Ind---==--=---mcmmmmm emcee 223 - 6 2.7 17 - - - Muskegon-Muskegon Heights, Mich====------ 293 5 9 1.3 32 - 1 3.1 Nashville, Tenn-=--=--cc-ecomeomeaaaaao- 589 39 2 5.9 216 10 2 -3.5 Newark, N,J------=--c-ce-- 3,674 109 107 -0.1 462 7 8 0.2 New Haven-Waterbury, Conn! 1,604 18 58 2.5 53 - 1 1.9 New Orleans, La=-==-=--=cocomeeoomomoooo 1,468 13 34 1.4 685 11 22 1.6 Newport News-Hampton, Va---====-c-e-e-o-- 244 25 3 -8.2 119 9 5 -3.1 New York, N.Y--=----ccoocm mmc cmee eee 25,332 90 307 0.9] 2,461 2 26 1.0 Nor folk-Portsmouth, Va---- ———- 658 3 19 2.4 310 - 15 4.8 Odessa, Tex-==------=-= -——— 93 - 2 2,2 8 - - - Ogden, Utah====--ccmoccmm comme 170 1 4 1.8 7 - = - Oklahoma City, Okla=-----=c-=cccoooannoon 809 54 12 -4.9 101 - 3 3.0 Omaha, Nebr.-Iowa--------- ———- 893 24 45 2.3 72 - 3 4,2 Orlando, Fla---==-=ceccmeecaannx ——— 463 5 6 0.2 89 - 2 2.2 Paterson-Clifton-Passaic, N.,J--- —— 2,316 42 95 2.1 78 - 4 S.1 Pensacola, Fla=-=-=---cmcmmmmmmmcee eno 229 3 8 2.2 76 1 2 1.3 Peoria, Tll-------oomcmmmmmmmmmemmeme 612 51 8 -6.5 14 1 - -6.,7 Philadelphia, Pa.-N.J 9,336 88 117 0.3 1,512 6 15 0.6 Phoenix, Ariz------=--cmccecccccceeeceeen 924 40 19 2,2 67 2 2 - Pittsburgh, Pa=-=-==--c-commocmcmanaa oo 5,571 53 53 - 453 4 1 -0,7 Pittsfield, Mass'=m=mmmmmmoom eae 409 6 10 1.0 6 - - - Portland, Maine'------c--commmm 457 11 23 2.6 - - - - Portland, Oreg.-Wash---=---=ccecmooaaaoan 2,063 28 67 1.9 47 1 - -2.1 Providence, R.I!'==-coommomm 1,969 20 28 0.4 21 - - - Provo-Orem, Utah-==--c--cccmmommmae 179 30 - -14.4 - 1 - -100.0 Pueblo, Colo=====ccoommmmm imme 239 100 1 -29.2 6 5 - -45.5 Racine, Wis-==---moommmo mcm 312 11 7 -1.2 8 1 - -11.1 Raleigh, N.C-=----comommmm mcm 165 39 9 -14.7 97 1 4 3.1 Reading, Pa 835 29 5 -2,8 14 - - - Reno, Nev------ccmomcmmo ooo 148 21 1 -11.8 8 - - - Richmond, Va 624 19 20 0.2 199 5 14 4.4 Roanoke, Va 265 36 15 -7.0 50 5 2 “5.5 Rochester, N,Y-----ccocmmmoom meee 1,588 33 32 -0.1 36 - 1 2.8 Rockford, Ill-------cmcmomm meee o 451 3 10 1.5 15 1 - -6.3 Sacramento, Calif--==--cccocmmcmmmcaaan- 848 4 53 5.8 52 1 2 1.9 Saginaw, Mich==--=cceoemmmm cee - 406 3 i 2.0 30 - - - St. Joseph, MOo=-=-==ccmcoommm eee 249 38 2 -12.5 13 1 - -7.1 St. Louis, MO.-Ill---c-occcmomm momen 4,125 39 41 0.0 724 3 6 0.4 Salt Lake City, Utah----=---=-cceceeacaaa- 640 15 20 0.8 11 - - - San Angelo, Tex---=--=----eccommocmonnonn 104 10 12 1.8 10 - 1 10.0 San Antonio, Tex---=--==-ccccmcceonnannn" 1,050 62 7 4.9 99 2 5 3.0 San Bernardino-Riverside-Ontario, Calif-- 1,509 50 37 -0.8 56 3 - =5.1 San Diego, Calif-==---cccccmmmccmeaeeae 1,601 26 32 0.4 79 - 1 1.3 San Francisco-Oakland, Calif 5,703 64 266 3:3 477 1 10 1.9 San Jose, Calif---=--------- 1,052 81 7 -6.5 24 1 - -4,0 Santa Barbara, Calif---==-----cecceccooanao 326 1 10 2.8 5 - - - See footnote at end of table. 46 Table 13. census and death records, United States, May-August 1960—Con. Number of matched deaths reporting same standard metropolitan census record only, and death record only; statistical area (SMSA) on both and net difference rates, by color: [census record used as a base. Minus (-) sign indicates more assignments by census than by NCHS. For definition of areas, see Technical Appendix] White Nonwhite Number reporting Number reporting Area Net Net differ- differ- Same Census | Death ence Same Census | Death ence on record | record rate on record | record rate both only only both only only Savannah, Ga---==---===c-cc--mmmmm mmm 216 - 7 3.2 182 - 6 3.3 Scranton, Pa------------=--n-- 827 11 22 1.3 5 - - - Seattle, Wash------------cmu- 2,296 19 85 2.9 77 - 3 3.9 Shreveport, La 320 6 - -1.8 237 2 3 0.4 Sioux City, Iowar===-mewmemeeemcemee——————— 292 5 13 2.7 3 - - - Sioux Falls, S. Dak-=-==-----eememmemen- 174 - 3 1a? - - - - South Bend, Ind 507 3 10 1.4 24 - 1 4,2 Spokane, Wash-------------=-- 683 39 15 -3.3 14 3 - -17.6 Springfield, Ill 426 9 24 3.4 12 - - - SPTANGTLalal, Wom wm se memos mrsd ses wn um eens 278 5 6 0.4 12 2 - ~14.3 Springfield, Ohlor=-wrurseommmmusinmims 272 16 4 -4.2 43 - 3 7.0 Springfield-Holyoke, Mass ----- iia 1,316 33 17 a 33 - - - Steubenville-Weirton, Ohio-W. Va --- 368 7 18 2.9 20 - 1 5.0 Stockton, Calif----------ccmmommmcememm 484 27 10 -3.3 40 4 3 -2.3 Syracuse, N Y= =rmre mmm mm mom mm inti on oe 1,421 20 60 2.8 21 - 1 4.8 Tacoma, Wash-------ommmmmmm cme mme om 637 78 7 -9.9 19 2 - -9.5 Tampa-St. Petersburg, Fla---- 2,190 39 32 -0.3 233 1 5 1.7 Terre Haute, Ind-----------=---- 312 2 8 1.9 17 - - - Texarkana, Tex-Ark-----==-c----- 117 2 9 5v9 44 1 5 8.9 Toledo, Ohio----=-=--ccmcmcmcmneaam 1,049 23 24 0.1 84 - - - Topeka, Kans 270 28 6 -7.4 29 1 - 33 Trenton, N,J 594 21 13 -1.3 68 6 2 -5.4 Tucson, Ariz 417 11 3 -1.9 27 1 1 - Tulsa, 733 9 34 3.4 93 1 3 2.1 Tuscaloosa, Ala-=rr-remmenmmmm 84 82 - -49.4 59 3 1 -3.2 TYler,; Temwmm mmm mu sm ew mii wi ww io - 107 18 - 14.4 36 5 1 -9.8 Utica-Rome, N,Y---meommmce ence me 855 122 8 -11.7 11 - - - Waco, Tex----====---=c--meme meme meme 247 18 6 4.5 69 4 4 - Washington, D.C.-Md.-Va---===---=m-ceuun-- 2,531 63 67 0,2 1,039 2 19 1.6 Waterloo, Towa---------ccccmmmmmmm meena 255 8 5 -1.1 - - - West Palm Beach, Fla-----------occoucooonn 467 4 10 1.3 91 4 1 -3.2 Wheeling, W. Va.-Ohio-----=-==--- 488 4 55 10.4 10 = 2 20.0 Wichita, Kans--------------=--=- 562 7 14 1.2 32 - 2 6.3 Wichita Falls, Tex---------===--- 195 54 7 -18.9 19 1 - -5.0 Wilkes-Barre —Hazleton, Pa 1,160 18 19 0.1 2 - - - Wilmington, Del.-N,J--=-=mmmeemmceecee——— 649 19 26 1.0 104 6 6 - Winston-Salem, N.C 205 5 5 - 119 - - - Worcester, Massl--------ccmeuonn 1,636 88 24 -3.7 6 4 - -40.0 York, Pa--=--------emmemmmm mmm meee mm 533 8 33 4.6 10 - - - Youngstown-Warren, Ohio---=--=----uccouauon 1,024 7 27 1.9 87 - 4 4.6 All other areas-----------===----cooou--- 139,339 2,967 | 2,252 -0,5]| 16,058 328 207 -0,7 "Metropolitan State economic areas. 47 Table 14. Number of unmatched deaths for standard metropolitan statistical areas (SMSA's), by color: United States, May-August 1960 [F or definition of areas, see Technical Appendix] Area Total White Nonwhite All areas-=---===-=mmmemmemeeeeeeeeemeeeeemeemmemee—m—em———- "112,656 | 93,319 | 19,336 Total SMSA'S == mmm mone ee eee em 68,731 || 56,613 12,117 Abilene, TeX==== === momo eee 60 57 3 Akron, Ohio===-= comme mm oe ee meee 268 236 32 Albany, Ga=-==== === meee ee 48 23 25 Albany-Schenectady-Troy, N.Y=-=-cc coco eee meee em 381 367 14 Albuquerque, N.MeX== === ooo ee eee meme 165 158 7 Allentown-Bethlehem-Easton, Pa. 214 207 7 Altoona, Pa---=-ccccmmmm eee o 93 92 1 Amarillo, TeX mm mmm mmm mm mm meee ee meee mem 70 64 6 Ann Arbor, Mich=-==-m comm oem ee a 66 57 9 Asheville, N,C=--ooomm mmm emmmmcm meen 183 158 25 Atlanta, Ga=======- comme eee 677 458 219 Atlantic City, N.J 132 94 38 Augusta, Ga.=S,C-===m common eee ee 185 106 79 Austin, TeX====----mmmmmm meme mmcmcmeeeee oo 137 115 22 Bakersfield, Calif 174 154 20 Baltimore, Md=--==- momo mmm eeeeeemmmme eee eo 1,274 876 398 Baton Rouge, La-======= commoner em 131 69 62 Bay City, Mich=--eooomcm ome eo 57 57 - Beaumont -Port Arthur, TeX=-=--=-eccmo mmo ceeemmemmee mo 177 133 44 Billings, Mont =--=== mmm momo mmemmmmmmmmmmme 60 60 - Binghampton, N,Y======sc common eee eee 114 113 1 Birmingham, Ala------cccmmmm cmos 622 351 271 Boston-Lowell-Lawrence, Mass? 1,724 1,642 82 Bridgeport- -Stamford-Norwalk, ConnZumnncnnn mma mmm —————— 344 308 36 Brockton, Mass? === mmm meee eee een 158 151 7 Brownsville-Harlingen=-San Benito, TeX=-=-=--emcommmmmmo om ceemeeee 116 114 2 Buffalo, N,Y--mooom moomoo ee eee ee em 620 567 53 Canton, Ohio=======-=cc comme meemmmm mmm mem 217 193 24 Cedar Rapids, Iowa=-==-=====cc comme mmmemcmm mmm mmm 92 92 - Champaign-Urbana, Ill-=---c--commem mmm eee meee = 55 53 2 Charleston, S.C====-=-o moomoo eee eee 205 89 116 Charleston, W.Va =--=-- ome m meee mmmmemmemmmee eo 146 130 16 Charlotte, N.Ce=-----mm momo eee mmcmmecmmmcmmmmee em 201 133 68 Chattanooga, Tenn.-Ga===--===== =m comme memcmmcmcmmcmmmmmco 281 204 17 Chicago, Ill=-=--cem mmm emmm meee m 3,979 3,077 902 Cincinnati, Ohio=Ky===-==- common oem - 635 516 119 Cleveland, Ohio===-=== o-oo mmm eee meemeem mmm 869 681 188 Colorado Springs, CoOlOo========m mcm eee eam 5 74 1 Columbia, S.C=-=-mmm mmm emmmmemmmmee eo 146 71 15 Columbus, Ga.-Ala===== cco mmoeeeemmemmmee— 169 113 56 Columbus, Ohio=-====== comme meee meme eo 321 273 48 Corpus Christi, TeX==---c--cm comme eee 113 105 8 Dallas, TeX=======c-- momo mmcmmmmm mmm mmm 663 488 175 Davenport-Rock Island-Moline, Iowa=Ill=--=-ecmcmommco emcee 147 145 2 Dayton, Ohio 349 298 51 Decatur, Ill 75 73 2 Denver, Colo 425 400 25 Des Moines, 125 115 10 Detroit, Mich 1,545 1,163 382 Dubuque, Iowa 16 16 - See footnotes at end of table. 48 Table 14, Number of unmatched deaths for standard metropolitan statistical areas color: United States, May-August 1960—Con. [F or definition of areas, see Technical Appendix] (SMSA's), by Area Total White Nonwhite Duluth-Superior, Minn, -WiS=-=-c-oemcmmm mcm mme meme eee mee 167 165 2 Durham, N,C=-==----cmm cmc m mmm mmo 96 55 41 El Paso, TeX=-==-=-mmmeome mmm emcee memmme meme ee 174 171 3 Erie, Pam----cemccmm meme eee a= 159 156 3 Eugene, Orege--=-----c-emmmm ccm 117 116 1. Evansville, Ind.-Ky=-=----ececmeee cece e - 115 97 18 Fall River-New Bedford, NEE Semis rnin imme moms mis etme isos im sicos ist oo 120 117 3 Fargo-Moorhead, N.Dak,-Minn==--=--ccmommm meme mmmmmmmmme meee 64 64 - Flint, Mich-o---mommc mmm mee eee 152 135 17 Fort Lauderdale-Hollywood, Fla-----=-cccsccmmmm mmm meme meee memo = 226 156 70 Fort Smith, Arke---ce-ccmcm mmc = 67 61 6 Fort Wayne, Ind--==---ceoccmmcm emcee meme emcee mmm mm meme emo 73 68 S Fort Worth, TeX====-sm---ocmee mmm ecm ce emcee memmm meme ne 400 337 63 Fresno, Calife--cmmcmmmm mre emer - 228 205 23 Gadsden, Ala=-=-=-=-mcom om eee ee 109 95 14 Galveston-Texas City, TeX=--==--cmmom mmm meme mmm 133 96 37 Gary-Hammond-East Chicago, 262 197 65 Grand Rapids, Mich=--------cecc-u--- 155 149 6 Great Falls, Mont===---cmcm mmm momma - 49 45 4 Green Bay, WiS==--ccooooomm mmm eee 69 67 2 Greensboro-High Point, N.C--=--=-c-cmomo ome mmm meme eee ee 143 102 41 Greenville, S.C=--=--mommmm meee eee 185 140 45 Hamilton-Middletown, OhiOo===-==--cmmmmmmm mmm meee 78 74 4 Harrisburg, Pa---------ccececeuoo-- ; 178 157 21 Hartford-New Britain-Bristol, 369 345 24 Honolulu, Hawaii--==--cccmmmcmm mmm ecmeceece meee eee mmm em 211 78 133 Houston, TeX==-=--memcm cece mm meme mmmm mmm 810 594 216 Huntington-Ashland, W.Va.-Ky.-Ohio======--ceccmmmm mcm cm meme eee = 173 167 6 Huntsville, Ala=-=---cmcommm mmm emm emcee mmm mem 106 71 35 Indianapolis, Ind=-==---mccommmmm meee eee 433 334 99 Jackson, Mich 43 39 4 Jackson, Miss 125 70 55 Jacksonville, 307 194 113 Jersey City, N.J=---cccmcmomc mmm ecm m meme mmm eee mm 386 336 50 Johnstown, Pas------ccmmmcm meme meme - 122 115 7 Kalamazoo, Mich=-=-=mmmmmmm mmm meee = 76 76 - Kansas City, Mo.,-Kans=-=-cmccomommmc mmm m meee enema 656 544 112 Kenosha, WiS-=----cccmmmm mmm c emer m meee mmm eee em 71 71 - Knoxville, Tenn==---==- oom mmm moe m mao 319 284 35 Lake Charles, La=--=-c-c-commmmm mom memmeeeemeee mem 94 74 20 Lancaster, Pa==---ccccmmm meme 135 134 1 Lansing, Mich-==----- momo mmm meee meee eee em 159 153 6 Laredo, TeX=-=-====-- ooo mmm mmm m meee 40 40 - Las Vegas, Neve--mcmeooom mmc meme cme emma 116 100 16 Lawton, Okla==-cc-ocoo mmm 41 34 7 Lexington, Ky======-=-ccmmmocm emcee meee eee 78 52 26 Lima, OhiO=====mcmco momma 30 26 4 Lincoln, Nebre-e--emcoccm mcm emer cere eem 89 89 - Little Rock-North Little Rock, Ark--------mommmmmmmm mmm meee 226 130 96 Lorain-Elyria, Ohio=-=--=-cccccmm mmr e 127 119 8 Los Angeles-Long Beach, Calif----c-cmcomccmmm mice meme eeeee eee 3,559 3,189 370 Louisville, Ky.=Ind-==-==c oom comme mmmmcmemmmmmme meee 517 401 116 Lubbock, TeX====-m-me moomoo eee m enema 70 65 5 Lynchburg, Va--=---cm momo mmm e meee meme = 88 62 26 Macon, Ga======== =m moomoo eee ememmmmmeee em 128 74 54 See footnotes at end of table. 49 Table 14, Number of unmatched deaths for standard metropolitan statistical areas (SMSA's), by color: United States, May-August 1960—Con. [- or uefinition of areas, see Technical Aopendix] Area Total White Nonwhite MHALBON, WiiGim mmr mom mmm mmm mm mmm 700 90 0A 00 00 0 1010 0 71 71 - Manchester, N.H -= omc com mmm meee meme 132 132 - Memphis, Tenn =---cm ecm mmmeooeeeeeee eee 387 170 217 Miami, Fla----c-o momo mmm mmm mmm ee eee eee 724 601 123 Midland, TeX===-===cm cocoon meme eee 38 32 6 Milwaukee, WiS==-mo como om momma 526 490 36 Minneapolis-St. Paul, Minn-=---=c- common o oo 796 777 19 Mobile, Ala=----- momo m mmm eee 223 134 89 Monroe, La--------mom mmm 109 61 48 Montgomery, Ala 155 79 76 Muncie, Ind=-=--cocom mmm momo eee 80 78 2 Muskegon-Muskegon Heights, Mich-=----cmcmcmm mmm 35 29 6 Nashville, Tenn----------- common eee mmm m m 253 173 80 Newark, N,Je=--c-ooo mmm 987 773 214 New Haven-Waterbury, CONN mmm mm mm eee 336 311 25 New Orleans, La=-------occcom moo cceeccmem een 557 315 242 Newport News-Hampton, Vas==-----c-commmm mcm eee 123 62 61 A 17,119 5,746 1,372 Norfolk-Portsmouth, Va=------ccoem mmm e em 389 220 169 Odessa, TeXm=------o mmm mmmmeee mem 43 40 3 Ogden, Utah=----ccccmcm mmc remem em 44 43 1 Oklahoma City, Okla=-==-= ccm meee meee 339 280 59 Omaha, Nebr, -Iowa-=-=---ccccmm mmm 233 199 34 Orlando, Fla-------mommmmmm mmm mee mmmcmmeeeee ee 209 161 48 Paterson-Clifton-Passaic, N.J 513 483 30 Pensacola, Fla=---c-ccmmm mmm eee 140 94 46 Peoria, Ille---ccocm mmm 175 169 6 Philadelphia, Pa.-N,Je===reeem meme mm mm mm mm mm mm 2,946 2,301 645 Phoenix, Arize=-----eecomcc mecca 545 493 52 Pittsburgh, Pa=---==----oom mmm o 1,554 1,414 140 Pittsfield, RR 68 68 - POTLLANG, MAREE «comm ceo mm rm cr mm se sm mm mm mm mm 143 141 2 Portland, Oreg.-Wash------ccooomm mmm eee 582 565 17 Providence, R.IZ eee eee eo mmm mmm mmo mim i mm mm mm mi im mm im 391 380 11 Provo-0Orem, Utah=----cemomcm mmc e- 30 30 - Pueblo, COlOo=-=-== moomoo momo eee meme eee m 82 78 4 Racine, Wis=---oomcmm mmm cece rere ee 52 51 1 Raleigh, N.,Co-c--mommmm moomoo meee mmm em emma 107 66 41 Reading, Pas-=--c-coccm omc 110 105 S Reno, Nev=-=--=-c ccm merece mmc mmm mmm 96 93 3 Richmond, Vase ccmmm mmo eee 325 210 115 Roanoke, Vas == -coo momo mem eee 101 84 17 Rochester, N,Y=-c-ccoomm mmc 224 208 16 Rockford, Il1l----commmm mm moomoo eee em 99 95 4 Sacramento, Calife-----omcom meee 350 320 30 Saginaw, Mich----ccccmcmm mcm cee rere 81 68 13 St. Joseph, MO=====-mmmm cece eee e meme ee 59 56 3 St. Louis, Mo, =Ill=----momm mmm m em 1,432 1,109 323 Salt Lake City, Utah=------mccmmcmm mmo 191 185 6 San Angelo, TeX=-==-=--ccmmmcccme ccc cmcmemce mmm 60 56 4 San Antonio, TeX=--==---cccem mmm cecmm meme mmm em 389 346 43 San Bernardino-Riverside-Ontario, Calif-------cc-cccmmccmmnnanno 557 522 35 San Diego, Calif=----cccccm mmm reer ee 620 589 31 San Francisco-Oakland, Calif=-------ccccommmmmm mcm eo 1,561 1,379 182 San Jose, Calif----cccmcmmm occa em 314 308 6 See footnotes at end of table. 50 Table 14, Number of unmatched deaths for standard metropolitan statistical areas color: United States, May-August 1960—~Con. [F or definition of areas, see Technical Appendix] (SMSA's), by Area Total White Nonwhite Santa Barbara, Calif-----=----eo-eccccccmcc cence meee cen meme eee 117 114 3 Savannah, Ga-----=======-e-eeeceeeemcccceceeccccecememesmmmmm—mee 12] 70 51 Scranton, Pa-----cecccwccccnnnmmmrmmr mmm mmm mmm ————————————— 119 115 4 Seattle, Wash=--=---ccomcmmmmm ccm cmm mcm ccce meen meme mmm 803 764 39 Shreveport, La=---------eecmeeccm cece ccc m emcee meme mmm essen mmm 247 126 121 SLOUR CLUY, TOWER = www wre rm mmm mmm mm mim mom om om om mw 0m mm et i oe om 0 82 82 - Sioux Falls, S.Dak=--==--e-ecmmemmemcccccccccccmccmcememmmmmmmm mmm 38 37 1 South Bend, Ind-------=-=-ceceemeecccccccccccccceemmem mmm mmm mmm 118 102 16 Spokane, Washe- e-em common coe meme 177 166 11 SpLInG Tied, TY dh wimmummbm mmission iim oi ari iim fe 5 00 0 00 0 0 0 113 107 6 Springfield, Mo----==---eomemmmem mmm mmemmeoomommmmmomm oo 116 114 2 Springfield, Ohio==----=--=--c-comcccommcm mmm mmm meme mmm 82 73 9 Springfield-Holyoke, Mass? -=====ommoo ammo o mmo meeeeee 187 177 10 Steubenville-Weirton, Ohio-W.Va---=-=-ccccmmmommmm meme mmm 127 116 11 STOCKTON, CALLE ww co mim io so on anim om mm sm mf, 1 0 0 225 196 29 BYTACUSE, N,V ww mmm mm mmm mmm om mw mm mm mm om mm mm sn 0 233 221 12 Tacoma, Wash--------e-ocemcmmmcc ccc ccccccme meee meee meen mmm mmm 237 225 12 Tampa-St. Petersburg, Fla=--------c-c-ceconcom cnn mmm m mm mm me = 598 501 97 Terre Haute, Ind------ceemmmmmcocc ccc ccccc mcm mmmm meme mm mmm mmm mmm 98 93 5 Texarkana, Tex.-Ark---==---rmeececeeeeceeccccecoe meee mee ee eee 81 59 22 Toledo, Ohio 303 277 26 Topeka, Kans 91 85 6 Trenton, N.J 173 141 32 Tucson, Ariz 202 187 15 Tulsa, Okla-------===-e-em-emeemeeececccccececeecm—em mee ———————— 251 207 44 Tuscaloosa, Ala------=-ccc-mmcmmmmm momo mmmm oom em-o—o—omo- 127 69 58 Tyler, TeX---=--=-=-cmmmm-e eee me eee omooomo-o—e—-——————— 94 57 37 Utica-Rome, N,Y-------ccmcommmmemm meme mmemmom mmm mmmom mmm om mmm mm 223 219 4 Waco, TeX-=---=-=-memceceecsccee cee ——e eee e-em eee mee se—e————————— 118 95 23 Washington, D.C.-Md.-Va~-=eceeemmrmmcee mee meme meme cee —————— 1,180 728 452 Waterloo, Iowa---===---=---emmcmommo emo o-o-om--ooo—o—oo 62 60 2 West Palm Beach, Fla-----------cmmmemommmommcccemm momo mmmm mmm mmm 198 121 77 Wheeling, W.Va.-Ohio=--------commmmmm mmc ccmmmmmomcmm momo m mmm 164 158 6 Wichita, Kans----=-=--=-e-remeeececcecc ccm cmm em emmme mmm mem 164 151 13 Wichita Falls, TeX=---=--=-----ccemmmmmmem meme emmm——oooooo—-—omo=o= 65 56 9 Wilkes-Barre--Hazleton, Pa------------cmcmmmmemmmcemcmmmme em mmmm moo 203 201 2 Wilmington, Del, -N.J----=---m-memcmmmocmeccecccomemm moomoo mmm mmm 232 179 53 Winston-Salem, N,C=====c--c--c--ccommmmmm em mmmmemmemooooooooomooooooo 146 94 52 Worchester, Mass? -=-=-=------cococmmm meme mmmmemmmom—ooeoomomoooo- 280 280 - York, Pa=-===-c---omeemm meen eee mememcmeoeooooo-osoo—-—m-e 139 133 6 Youngstown-Warren, Ohio==-===-===--c--ccooooooomomom mmm mmm m mmm m mmo 270 240 30 All other Areas==-=-===m=mcmemmoecm ecm ceemememmmmemme—————oo- 43,925 || 36,706 7-219 "Includes one machine error. 2Metropolitan State economic areas. NOTE: For explanation of why white and nonwhite do not add to the total, see Technical Appendix. 51 United States and each geographic division, May=- August 1960 Net difference rates, by color, sex,and age: Table 15. [census record used as a base. Minus (-) sign indicates more assignments by census than by NCHS) Pacific SOON NNITNO NOW DEI OHI rrr ra no ONO ONNS om MNMN—HOO~O To Rw OOOO ONMNOO— FONT —HOoONNI Crore 10OTMOOMNT IT NG reels NN ON $e 1 ANN OOM 00 ®O SEER Te NONITNAHIFTOONT 1 1 —~ 1 1 Mountain NNN FTN N00 — NH OITMNOOO—M 1 tora ra 11 FOOT NO A oONMOoOO—~AaNAN PL TOIT oNnNInNA NMON~OOOWN Err NINO —~moo O FNC NNI~D ~~ 1 3 -16.7 22 West South FTNooONOITONN BOON NmiF Lote ra LI] CONMOO FST NANOS HOS ro Fea NOOOIFNHNOT NONI NOOMNMN ~ 1 ra "a om 1 hk 5 Aer gad - Central | Central East South NANOOAOOTNN IO ~MNO ANN NNNNO~——~O 1 ro 1 1 COO M—HWON—~ FNNOASTOT IO ro 4 DOV OPI © NoFIN in -Ln rrr iy -18.0 1 IE ; Mo 1 South Atlantic voONNOO—A—ON SooNNOoOOoOOON LEQ VOMNANNIIMNMNND CQOHINHOM—AHN 13000 110 SONI OW—~H—N FOMONINO DON 1 [I N 1 Ne 1 West North Central = AFF OoOON~HM 1 ror — reo mesg) oa nro NA oon 1 1 a 1NANNMOOO NN OOMONAN ON ' rN 1 East North Central NO ANTNOTOT NC waa ipN ' | I —o DR OOO 0 1 1 NNO ONANOOT NNT HO AN 1 Aye ONAN ~AINNOOMNO IBS Een ' ~~~ FA AO On I~ ro $4 —o—-— ro Middle Atlantic 1D CNN PS I gir) So~InSSS~S roa | ONO ONNA~O0O ee CA SD rl (SD ed ed roa NONHOIFTAN MO SOHNONOmiNG ror 5) oyoia) Slane OFTMNN—O FO NnNANNNANGO~~SNM [a . (i | ~~ | New England OOM FN WNOTO ONONNOO~AN~ 1 — a MO —~O MOO — HON ON ~~ 1 VOoOFTI~—HIF IN ONANAINO =I NOM — LIE La 1 1 ed roo ~~ Sono @ on hi Total —HMNORNOMONNO NOOMMO~H—~O~ ror 1 OWANNOAN~AMNNO OCOOITIF—HOO~N i ror 1 NNO ONO MO 0-T HESTON NN Err rr a sex, Color, and age Total 55-64 years----- 65-74 years----- 75-84 years----- 85-99 years----- 45-54 years----- 25-34 years----- 35-44 years----- 1-4 years------- 5-14 years------ 15-24 years----- White male 25-34 years----- 35-44 years----- 45-54 years----- 55-64 years----- 65-74 years----- 75-84 years----- 85-99 years----- 1 1 1 1 1 0 ~ [] [9 > aN ~ ny — 1-4 years------- 5-14 years------ White female 1-4 years------- 5-14 years------ 15-24 years----- 25-34 years----- 35-44 years----- 45-54 years----- 55-64 years----- 65-74 years----- 75-84 years----- 85-99 years----- Nonwhite male 85-99 years----- 65-74 years----- 75-84 years----- 25-34 years----- 35-44 years----- 45-54 years----- 55-64 years----- Nonwhite female 1-4 years------- 5-14 years------ 15-24 years----- 25-34 years----- 35-44 years----- 45-54 years----- 55-64 years----- 65-74 years----- 75-84 years----- 85-99 years----- NOTE: Figures with color and/or sex not stated are included in the total, but are not shown separately. 52 Table 16. Net difference rates for Division, May-August 1960 white males for stage I and stage II, by age: South Atlantic [Census record used as base. Minus (=) sign indicates more assignments by census than NCHS. For explanation of stage | and stage Il, see Technical Appendix] Stage 1 Stage II Age Age not Age not Age not Age not stated and | stated and | stated and | stated and not valid | not valid | not valid | not valid included excluded | allocated | unallocated 1-4 years-===-m-mm mmm eee eee 1.1 0.6 -6.3 -6.3 5-14 years==-=-m-mmmmmm meme n - -0.6 1.3 - 15-24 years=====--- mom mmmeeeeemeememeeeem em 0.5 -0.2 -3.8 -2.9 25-34 yearS==== =m mmm mmm eomeoo -1.0 we 0.8 2.3 35-44 years==-== mm mmm meme eee -1.7 -2.6 “5.7 -4.5 45-54 yearS-=- mmm mmm mmm 0.2 -0.6 -3.8 -2.7 55-64 years--=----=-mm meee 0.9 -0.1 -0.5 -0.8 65-74 yearS==--mm mmm mmm 0.6 0. 0.9 0. 75-84 year S=mm- =m mmm meme eee 1.4 «0.6 1.8 1:3 85-99 years=---=--mmmm mmm 3:1 2.2 85 years and over--------ceocm mma om 542 5.5 Not stated====-==--mooom oom -100.0 Not valid=-=--cmmc meme -42.9 NOTE: Stage I, but not stage II, excludes decedents whose age was stated to be 100 years or over on the census record. 53 Table 17. Number of matched deaths reporting same race on both census and death records, census record only, and death record only; and net and gross difference rates: United States and each region, May-August 1960 Region and race Number reporting Census record as base Death record as base Net difference! Gross difference Net difference’ Gross difference Census | Death $ims on record | record only only Number Rate Number Rate Number Rate Number Rate All regions White--=-=-------- 353,612 847 3,621 2,774 0.8 4,468 1.3 -2,774 -0.8 4,468 1.3 Nonwhite-------- 37,110 881 1,299 418 1. 2,180 5.7 -418 -1.1 2,180 5.7 Negro=--=------- 35,309 685 1,092 407 1.1 1,777 4.9 -407 -1.1 1,777 4.9 Indian-------- 602 154 83 -71 -9.4 237 31,3 71 10.4 237 34,6 Japanese------ 562 19 14 -5 -0.9 33 5.27 5 0.9 33 5.7 Chinese------- 315 33 17 -16 -4,6 50 14.4 16 4,8 50 15,1 Pn ————— 138 56 11 -45 23,2 67 34.5 45 30,2 67 45,0 Other non- white-------- 69 49 197 148 125.4 246 208.5 -148 -55.6 246 92,5 Not stated and not valid------ . 3,214 22 -3,192 =99,3 3,236 100.7 3,192] 14,509.1 3,236 | 14,709.1 Northeast White--====auun- 107,551 275 1,007 732 0.7 1,282 1.2 -732 -0.7 1,282 1.2 Nonwhite-------- 5,467 245 332 87 1:5 577 10.1 -87 -1.5 577 2.9 Negro-----=---- 5,313 229 335 106 1.9 564 10,2 -106 -1.9 564 10.0 Indian-------- 24 21 10 -11 =24.,4 31 68.9 11 32.4 31 91,2 Japanese 8 - 1 1 12.5 1 12.5 -1 -11.1 1 11.1 Chinese------- 74 13 3 -10 -11.5 16 18.4 10 13.0 16 20.8 Pilipinas 3 15 1 -14 -77.8 16 88.9 14 350.0 16 400.0 Other non- htiess —————— 1 11 26 15 125.0 37 308.3 -15 -55.6 B87 137.0 Not stated and not valid------ 824 5 -819 -99.4 829 100.6 819| 16,380.0 829 | 16,580.0 North Central White~----- 113,116 196 1,117 921 0.8 1,313 1.2 -921 -0,8 1,313 1.1 Nonwhite-- 6,587 204 263 59 0.9 467 6,9 -59 -0.9 467 6.8 Negro--- 6,401 153 257 104 1.6 410 6.3 -104 -1.6 410 6,2 Indian-- 110 51 14 -37 -23.0 65 40,4 37 29.8 65 52.4 Japanese 28 3 1 ~2 -6.5 4 12.9 2 6.9 4 13.8 Chinese------- 21 8 2 -6 -20.7 10 34.5 6 26,1 10 43.5 Filipino------ 5 3 2 -1 «12,5 5 62.5 1 14.3 5 71.4 Other non- white-------- - 8 9 1 12.5 17 212.5 -1 -11.1 17 188.9 Not stated and not valid------ 989 9 -980 -99.1 998 100.9 980| 10,888.9 998, 11,088.9 South White--=-==ooonn- 81,809 274 926 652 0.8 1,200 L.5 -652 -0.8 1,200 1.5 Nonwhite-------- 22,369 262 448 186 0.8 710 3.1 -186 -0.8 710 3.1 Negro=--------- 22,188 220 433 213 1.0 653 2.9 -213 -0.9 653 2.9 Indian-- 146 38 20 -18 -9,8 58 31.5 18 10.8 58 34.9 Japanese 5 3 2 -1 -12.5 5 62,5 1 14.3 5 71.4 Chinese------- 15 3 1 -2 -11.1 4 22.2 2 12.5 4 25.0 Filipias —————— - 7 2 -5 -71.4 9 128.6 5 250.0 9 450.0 ther non- white-------- - 6 5 -1 -16.7 11 183.3 1 20.0 11 220.0 Not stated and not valid------ 842 4 -838 -99.5 846 100, 5 838 | 20,950.0 846 | 21,150.0 West White----------- 51,136 102 571 469 0.9 673 1.3 -469 -0.9 673 1.3 Nonwhite-------- 2,687 170 256 86 3.0 426 14.9 -86 -2.9 426 14.5 Negro=--------- 1,407 83 67 -16 -1.1 150 10.1 16 1.1 150 10.2 Indian-------- 322 44 39 -5 -1.4 83 22.7 5 1.4 83 23.0 Japanese------ 52) 13 10 -3 -0.6 23 4.3 3 0.6 23 4,3 Chinese------- 205 9 11 2 0.9 20 9,3 -2 -0.9 20 2.3 Filipino------ 130 31 6 -25 -15.5 37 23.0 25 18.4 37 27.2 Other non- x IL erm 68 24 157 133 144.6 181 196.7 -133 -59.1 181 80.4 ot stated an not valid------ . 559 4 -555 =-99,3 563 100.7 555] 13,875.0 563 | 14,075.0 IMinus sign indicates more assignments 2Minus sign indicates more assignments NOTE: 54 by census than by NCHS. by NCHS than by census. For explanation of why white and nonwhite do not add to the total, see Technical Appendix. Table 18, Number of matched and unmatched deaths and proportion unmatched, by race: United States and each region, May-August 1960 Matched deaths according to: Proportion unmatched $ Unmatched Region and race Census Death deaths Census Death record record record record All regions All races=-===---=-=-------- 395,664 395,664 100,412 20,2 20.2 White=-==--=-=-cocccmmmcm cm cmm a 354,459 357,233 84,560 19.3 19.1 Nonwhite--==-==-=-cccocmmmnncennanan 37,991 38,409 15,851 29.4 29.2 Negro---====-smmememoem cee e me 35,994 36,401 14,913 29.3 29.1 Indian=====-==--cecmcmmmmm em 756 685 432 36.4 38.7 Japanese-- 581 576 131 18.4 18.5 Chinese--- 348 332 153 30.5 31.5 Filipino=---===--cocommcmmmmmeeo 194 149 112 36.6 42.9 Other nonwhite---------=ccc---- 118 266 110 48,2 29.3 Not stated and not valid--------- 3,214 22 i 0.0 4,3 Northeast All races---==--=cememcooan 114,362 114,362 24,32) 17.5 17.45 White---=-=-ommmmmmcmmme meme 107,826 108,558 21,821 16.8 16.7 Nonwhite------==cocmmmmmccnnann 5,712 5,799 2,499 30.4 30.1 Negro--====-==-eecememe emcee 5,542 5,648 2,410 30.3 29.9 Indian-----===-----c-c-oooo-- 45 34 16 26.2 32.0 Japanese--------=---c-c-comoonon 8 9 4 33.3 30.8 Chinese---=-==---cmcoocmmnn- 87 77 58 40,0 43,0 Filipino=-------=-cmccmmccnnoo- 18 4 2 10.0 33.3 Other nonwhite------=----=----- 12 27 9 42.9 25,0 Not stated and not valid--------- 824 5 1 0.1 16.7 North Central All races=-------=---------- 121,092 121,092 26,318 17.9 17.9 TEL ois 8 113,312 114,233 23,792 17.4 17.2 Nonwhite---=---cccmmmmmce eee aa 6,791 6,850 2,526 27.1 26.9 Negro=----==-=---commmmmoae emo 6,554 6,658 2,386 26.7 26.4 Indian---====--==-cccmoemeoo- 161 124 105 39.5 45,9 Japanese-----------cc-mooooooo 31 29 7 18.4 19.4 Chinese==-=-==---=-cc-cccu- 29 23 15 34,1 39.5 Filipino-----=-----ccommomonn 8 7 3 27.3 30.0 Other nonwhite---------=---c-o-- 8 9 10 55.6 52.6 Not stated and not valid--------- 989 9 - - - South All races=------=---------- 105,556 105,556 34,279 24,5 24,5 White--==-meommmmmm mmm mm mmm mmo 82,083 82,735 24,681 23.1 23.0 Nonwhite---=-==--cmcmcoonncnmao 22,631 22,817 9,598 29.8 29.6 Negro-------==----mcomcmmanaon 22,408 22,621 9,487 29.7 29.5 Indian-==-=-=c-ccmomommmanemoo 184 166 83 31.1 33.3 Japanese=------==--c-cmecconnn 8 7 3 27.3 30.0 Chinese------mmmmeececee ec ————— 18 16 13 45.5 48.4 Filipino-------=---cmmmmmmcoomn 7 2 4 36.4 66,7 Other nonwhite------=---c---cc-- 6 5 6 50,0 54.5 Not stated and not valid--------- 842 4 - - = West All raceS=--=---==-=---c---- 54,654 54,654 15,494 22.1 22,1 White-=-======cccmcmmmcmccmm mem 51,238 51,707 14,266 21.8 21.6 Nonwhite==-=====-ccocmcmmoaannnnon 2,857 2,943 1,228 30.1 29.4 Negro=---------cocommcon mean 1,490 1,474 630 29.7 29.9 Indian-=---====-=ccccccemoeoon 366 361 228 38.4 38.7 Japanese---------c--mmmmmoooon 534 531 117 18.0 18.1 Chinese=======-=c-c-ecomomconn 214 216 65 230 23.1 Filipino-----------cccmmoooanm- 161 136 103 39.0 43,1 Other nonwhite-=-----=---=------ 92 225 85 48,0 27.4 Not stated and not valid--------- 559 4 - - - 55 TECHNICAL APPENDIX Design of Study The study includes a sample of registered deaths which occurred during the 4-month period, May-August, 1960. The following age-color groups were included: 1. All nonwhite decedents 2. All white decedents under 65 years of age 3. One-half of white decedents 65-74 years of age 4. One-fifth of white decedents 75 years and over The records for white decedents were systemati- cally selected from the regular mortality punched- card file using the randomly assigned terminal digits of the certificate numbers. Sample groups three and four of both the matched and unmatched records were inflated by two and five, respec- tively, during the processing operation. As indicated in table I, the actual count of death certificates included in the match operation was 340,033, Of this number, 495 records were excluded because of '"impossible'' codes-—483 matched and 12 records unmatched. Also included in table I are the inflated counts of records by match status. A seasonal bias may be introduced since the deaths were limited to May-August. Butan exami- nation of 'monmatch' rates by month, particularly April and September, for a previous study involv- ing the occupation item on the census and death records,” resulted in a decision to limit this study to the 4-month period. Table I. Number and percent distribution of records included in the 1960 Comparison Study: United States, May-August 1960 Actual count of Inflated count of records records Result of match operation Percent Percent Number distribu- | Number distribu- tion tion Total-----==--ecmme reece cee eee 340,033 100.0 | 533,747 100.0 Matched=--=--cccmmm mmm mmm eee eee 262,483 77.2 { 420,292 78.7 Unmatched -=--cccmcmm mmm meme emma 77,055 22,7 | 112,656 21.1 Rejects! mmm mmm emma 2495 0.1 3799 0.1 Includes impossible codes not included in the study. 2Includes 483 matched records and 12 unmatched. Includes 782 matched records and 17 unmatched. 56 Criteria for Matching Copies of the 340,033 death certificates in- cluded in this study were coded for the stage I Census Enumeration District (ED). A record was considered matched if the exact street address of residence and the name of the dece- dent were located in the ED book. Other infor- mation such as date of birth and marital status were used if the above criteria did not result in a match. If infants under 1 year of age were not located in the ED book, an attempt was made to match via the mother. The infants represented a high nonmatch risk group since many of them were born after the 1960 Census of Population and therefore were not enumerated. Almost 6 percent of the records were matched at the place of death, rather than the usual place of residence. Those persons matched at the place of death were for the most part those who died in institutions. Decedents matched in stage I and included in the 25-percent sample of the 1960 Census of Popu- lation, referred to as stage 11, were then searched in the stage II books. There were 64,675 such decedents and all but 2,188 were located in the stage II census records. The names of these 2,188 decedents could not be located on the appropriate pages of the stage II schedules. A host of reasons exists for failure to match the death certificate to the census record: (1) The decedent may have moved after the date of enumeration on April 1. (2) The decedent may not have been enu- merated. (3) Clerical and machine errors occurred at all stages of the study. (4) Misstatements of information were en- tered on the census record and/or the death certificate. Variances The sample design has been discussed in a previous section, "Design of Study.' To obtain the sampling error of an estimate of the net difference rate for a given geographic area, it would be necessary to know the number of death records for white decedents 65-74 years of age and 75 years and over during May-August 1960 for geographic areas both before and after sam- pling. The death records were sampled but the color designations for most geographic areas were based on the color on the census record. Although all nonwhite decedents were included in the study period, a decedent stated to be white on the death record may be nonwhite on the census record and vice versa. Therefore, it is not possible to state that the sampling error is zero for the nonwhite group shown in the tables. Age The age reported on the death record is generally the age of the decedent at his last birthday. If the reported age did not agree with the computed age (the difference between the date of death and the date of birth on the certificate), the stated age was accepted unless the difference was 5 years or more. If the difference was 5 years or more, entries for other items on the vital record such as cause of death, occupation, marital status, and social security number were considered in determining the age. Most of the differences involved infants. Therefore, the afore- mentioned characteristics would give some in- dication of the ''correct' age. If a discrepancy occurred such as 25 years for the computed age and 35 for the stated age, the stated age was accepted. The age classification used in the 1960 Census of Population is based on the quarter of the year of birth as of April 1, 1960. For this study, the age was updated to adjust for a birth- day that might have occurred between the date of the census and the time of death. As indicated in the tables showing data by age, the number of matched records for dece- dents under 1 year of age and 100 years and over, according to the census record, are excluded. The number of records with negative ages are also excluded. Infants were excluded because many were born after the date of enumeration on April 1, 1960. There were 23,176 records 57 for infants. Decedents 100 years and over and those with negative ages were omitted because there was an excessive number of errors in the century-of-birth codes on the census record. For example, if the year of birth was actually 1861 and the century code for 1900 was assigned, the result was a negative age. There were 1,452 such records, that is, for persons 100 years and over and for negative ages. Such exclusions were made to maintain comparability in order to better assess differences in age reported on the two records. Additional technical problems relating to age reported on the vital record are covered in another report,b Race and Color Deaths in the United States in 1960 are classified for vital statistics purposes into white, Negro, American and Alaskan Indian, Chinese, Japanese, Aleut, Eskimo, Filipino, Hawaiian, part-Hawaiian, and other nonwhite. The U.S. Bureau of the Census uses the same classifi- cation, Tabulations in this study by race were grouped as follows: White Nonwhite: Negro Indian (American and Alaskan) Japanese Chinese Filipino Other nonwhite (Aleut, Eskimo, Hawaiian, part-Hawaiian, and racial mixtures ex- cept as noted below) Not stated and not valid The category ''white'" includes, in addition to per- sons reported as ''white,' those reported as Mexi- can or Puerto Rican. If raceisgivenas a mixture of Negro with any other race except Hawaiian, the race is classified as Negro. Mixtures of Hawaiian and any other race are classified as "other non- white." In stage I the number of census records for which the race item was not stated is combined with "not valid" codes (processing errors) and shown as a single group in the race tables. I'he 58 relative number of such records is small—less than 1 percent of the matched records. If the race was not entered on the death record, the code for "white" was assigned in the initial code-punch operation in NCHS. Consequently, the category 'not stated and not valid'' accord- ing to death certificate designations includes only invalid codes. The number of such records is small. This factor is noted, not because it seri- ously invalidates the net difference rates, but because the category 'mot stated and not valid" was excluded from most of the residence tables. Therefore, figures by white and nonwhite do not add to the total, With the exception of table 18 which shows a comparison of detailed groupings of race, the tables show the color item according to census designations. The purpose was to compare a single characteristic such as data for an SMSA according to census and to death record assign- ments. Therefore, all other characteristics such as age, sex, and color in that table were by census designations so as not to distort the com- parison of the variable being compared. Sex In a table showing a characteristic such as age or race classified by sex, the sex of the matched decedent is that stated on the census records in order not to invalidate the character- istic being compared. The small number of deaths for which the sex was not stated—0.5 percent of the total matched records—is included in the total in those tables containing sex, but the category is not shown separately. Urban and Rural Areas The urban-rural definitions used by NCHS differ from those used for data published by the U.S. Bureau of the Census. For this study, the Bureau of the Census assigned the residence codes on the census records of the decedents included in this study according to the rules used for vital statistics. Therefore, the figures for urban and rural are comparable with respect to the application of rules, and the differences between the NCHS figures and those tabulated by census result from the entries on the two records or processing errors. As noted earlier, differ- ences in the enumeration process and the regis- tration system give greater weight tothe accuracy of the census figures. For the decennial census, geographic locations can be fixed with a high de- gree of accuracy by the Bureau of the Census through the use of street maps, census tracts, and the like. NCHS is dependent almost entirely on the information entered in the residence item on the vital record. For the definitions of urban and rural as used for vital statistics purposes, see Vital Statistics of the United States, 1960, Volume II, Part A. Standard Metropolitan Statistical Areas The standard metropolitan statistical areas (SMSA's) used in this report are the sameas those established by the U.S. Bureau of the Budgetas of 1960 (except in the New England States) and used by the U.S. Bureau of the Census. Except in the New England States, an SMSA is a county or a group of contiguous counties which contains at least one city of 50,000 inhabitants or more or "twin cities" with a combined population of at least 50,000 in the 1960 census. Contiguous counties are included in an SMSA if, according to specified criteria, they are (1) essentially metro- politan in character and (2) socially and economi- cally intergrated with the central city or cities. In New England the Bureau of the Budget uses towns and cities rather than counties as geo- graphic components of SMSA's., NCHS cannot use the SMSA classification in New England because its data are not coded to identify all towns. Instead the metropolitan State economic area (MSEA) established by the Bureau of the Census, which is made up of county units, is used. (For a more complete discussion of SMSA's and MSEA's, see references 8-10.) Metropolitan and Nonmetropolitan Counties Counties which are included in SMSA's, or in New England MSEA's, are called metropolitan counties; all other counties are classified as nonmetropolitan, Geographic Divisions and Regions The nine divisions and four regions referred to in this report correspond to those used by the U.S. Bureau of the Census. For States included in each division and region, see reference 8. Formula for Adjusting Annual Death Rate dm, — dm D-D (7) dm, = R, = Upper range of P rate dm —dm_ du D-D ( ) -D ( ) dm, dm, + d, P Symbols D =annual number of deaths, 1960 P =population enumerated as of April 1, 1960 dm_=matched deaths in study period according to census record dm, =matched deaths in study period according to death record du =unmatched deaths in study period R =annual death rate per 1,000 population R, =upper range of adjusted annual death rate per 1,000 population 1 R, =lower range of adjusted annual death rate per 1,000 population Deaths, 1960, by vesidence—SMSA of Abilene, Texas— White Formula A—upper range 794 — 794 (=== - =) 172 = Ry 114,508 771 114,508 x 1,000 = 6.7=R, = upper range 59 Formula B—lowey range 794 _ [104 (-17)] _ 794 (2)- R, 172 172 + 57 114,508 SB x 1,000 = R, = 50 = lower range 114, 508 Rates R = 6.9 published annual rate R =6.7 adjusted rates R,=5.0 Net and Gross Difference Rates Net difference rates for matched deaths were computed as follows: (1) Subtract the number of records accord- ing to census assignments only from those according to NCHS assignments only (2) Divide the number obtained in step 1 by the number of records for which thereis agreement between census and NCHS plus census assignments only (3) Multiply by 100 Most tables contain net difference rates with the census record as base; thus a negative rate indicates more assignments by census than by NCHS. If the death record is the base, a nega- tive rate signifies more assignments by NCHS. In most instances the base is inconsequential. If the differences are relatively large, the rate may vary greatly depending on the base. Gross difference rates were computed by summing the differences rather than subtracting the differences as for netdifference rates. Other- wise the computation was the same. 60 77 U. S. GOVERNMENT PRINTING OFFICE: 1969— 342045/33 Due to a programming error the first digit of the gross difference rates was not printed out on the tabulations in many instances. Rather than manually correct all of these errors, only a few were recomputed for inclusion in this report. Symbols s, =Number of assignments are the same according to both the census record and the death record c, =Number of assignments according to the census record only d = Number of assignments according to the death record only ndr, = Net difference rate gdr, = Gross difference rate Formulas Net difference rate: Census record is base, d -c, ! ! x 100 = ndr, Si + C; Death record is base. c —d iced % YOO = ndr, 5, + d Gross difference rate: Census record is base. d +c x 100 = gdr, Ss. + C Death record is base. c + d x 100 = &dr, s; +d, Series 1. Series 2. Series 3. Series 4. Series 10. Series 11. Series 12. Series 13. Series 14. Series 20. Series 21. Series 22. OUTLINE OF REPORT SERIES FOR VITAL AND HEALTH STATISTICS Public Health Service Publication No. 1000 Programs and collection procedures.—Reports which describe the general programs of the National Center for Health Statistics and its offices and divisions, data collection methods used, 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, 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 veports.— 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, based on data collected in a continuing national household interview survey. Data from the Health Examination Survey.—Data from direct examination, testing, and measure- ment of national samples of the 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 Institutional Population Surveys.—Statistics relating to the health characteristics of persons in institutions, and on medical, nursing, and personal care received, based on national samples of establishments providing these services and samples of the residents or patients. Data from the Hospital Discharge Survey.— Statistics relating to discharged patients in short-stay hospitals, based on a sample of patient records in a national sample of hospitals. 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 manpower occupations, hospitals, nursing homes, and outpatient and other inpatient facilities. Data on mortality.—Various statistics on mortality other than as included in annual or monthly reports—special analyses by cause of death, age, and other demographic variables, also geographic and time series analyses. Data on natality, marriage, and divorce. — Various statistics onnatality, marriage, and divorce other than as included in annual or monthly reports—special analyses by demographic variables, also geographic and time series analyses, studies of fertility. Data from the National Natality and Mortality Surveys. —Statistics on characteristics of births and deaths not available from the vital records, based on sample surveys stemming from these records, including such topics as mortality by socioeconomic class, medical experience in the last year of life, characteristics of pregnancy. etc. For a listof titles of reports published in these series, write to: Office of Information National Center for Health Statistics U.S. Public Health Service Washington, D.C. 20201 VITALand HEALTH STATISTICE ~ DATA EVALUATION AND METHODS RESEARCH NATIONAL ol Pl A RE; NL For HEALTH Number31 STATISTICS Pseudoreplication Further Evaluation and EL i CREE EDT ~ Half-Sample Technique 5 U. S. DEPARTMENT OF i HEALTH, EDUCATION, AND WELFARE 3 Public Health Service ORE Health Services and Mental Health Administration Public Health Service Publication No. 1000-Series 2-No. 31 For sale by the Superintendent of Documents, U.S. Government Printing Office Washington, D.C. 20402 - Price 35 cents NATIONAL CENTER| Series 2 For HEALTH STATISTICS | Number 31 VITALand HEALTH STATISTICS DATA EVALUATION AND METHODS RESEARCH Pseudoreplication Further Evaluation and Application of the Balanced Half- Sample Technique Uses of the method in analysis of data from complex surveys. U.S. DEPARTMENT OF HEALTH, EDUCATION, AND WELFARE Public Health Service Health Services and Mental Health Administration Washington, D.C. January 1969 NATIONAL CENTER FOR HEALTH STATISTICS THEODORE D. WOOLSEY, Director PHILIP S. LAWRENCE, Sc.D., Associate Director OSWALD K. SAGEN, PH.D., Assistant Director for Health Statistics Development WALT R. SIMMONS, M.A., Assistant Director for Research and Scientific Development ALICE M. WATERHOUSE, M.D., Medical Consultant JAMES E. KELLY, D.D.S., Dental Advisor EDWARD E. MINTY, Executive Officer MARGERY R. CUNNINGHAM, Information Officer Public Health Service Publication No. 1000-Series 2-No. 31 Library of Congress Catalog Card Number 68-62232 PREFACE In an earlier publication of the National Center for Health Statistics (Series 2, No, 14) the method of balanced half-sample pseudoreplication was introduced as a device for estimating precision of estimates which come from surveys of modern complex design. The technique was de- veloped by Philip J. McCarthy of Cornell University. The report and McCarthy's work are facets of a series of efforts by the Center to find appropriate methods for analyzing data from complex social and health surveys, in which the assumptions of standard classical techniques rarely are satisfied. The present study describes a number of internal evaluation tech- niques which aid in understanding the behavior of balanced half-sample replication as it relates not only to the variance of linear estimators but also to precision of ratios. And itextends the method to applications of a more analytic character, including adaptations to the sign test and to contingency tables. Replication and pseudoreplication are concepts that have deep roots in statistical theory and that have influenced the thinking of many people. Acknowledgment of all who have contributed is not feasible, but the list of those who have had special influence on the research presented here includes W. Edwards Deming, John Tukey, M.H. Quenouille, Leslie Kish, and several members of the staff of the U.S. Bureau of the Cen- sus. Professor McCarthy carried out the research and wrote the report. Walt R. Simmons has coordinated Center activities in the area of pseudoreplication and, as Project Officer has maintained close contact with the present undertaking. SYMBOLS Data not available----cccemommmmomo Category not applicable-------cccmemoannann QUADLILY ZOE mm mo im mm sm mm mm Quantity more than 0 but less than 0.05---- Figure does not meet standards of reliability or precision--------m-ceeaeano 0.0 Introduction --- Background for Description and CONTENTS Evaluation Studies-----======--=====-----mmm------------- Application of Evaluation Techniques------------=--------- Average of Half-Sample Estimates-----------=--=---------=----------o- Bias of a Ratio Estimator------=-=---=----=====--------------o------ooo Estimating the Variance of r mmm meme meme mmm Estimating the Variance of R ---====-=----------------------ooooooooo Design Effect Approximate Test of Independence Based on a Set of Balanced Half-Sample Replicates --- Order Statistics of a Set of Balanced Half-Sample Estimates--------------- References ---- Detailed Tables ~1 ON Wb 10 12 15 18 19 vi A REPORT of the National Center for Health Statistics (Series 2, No. 14) introduced a pseudoveplication procedure, called balanced half- sample replication, as a basis for statistical analysis when dealing with data derived from complex sample designs and estimation processes. It appeared particularly suited for vaviance estimation in connection with the commonly used multistage design of two primary units per stratum, with the individual observations being subjected to various kinds of ratio and poststratification adjustments. The exact behavior of balanced half-sample replication as a variance- estimating procedure is known only for a simple linear situation where it has no real utility. The first portions of this report review this theory and describe a number of evaluation techniques which can provide evi- dence concerning the behavior of balanced half-sample replication for a particular set of data. They ave illustrated on data devived from the Center's Health Examination Survey (Series 11, No. 1) relating to a variety of body measurements for a sample of approximately 3,000 U.S. adult males and involving estimates of population means, multiple re- gression coefficients, and multiple correlation coefficients. These tech- niques are principally concerned with providing presumptive evidence relating to (1) the extent to which the average of a set of balanced half- sample estimates 'exhausts'' the information in the entive set of 2* half-samples, (2) the existence of bias in the entive-sample estimate, and (3) the adequacy of the variance estimates produced by balanced half- sample replication. Applications to the Health Examination Survey data seem to indicate that balanced half-sample replication is performing well as a variance estimating procedure, Also, as might be expected on the basis of the sample size, there is no evidence of bias in the esti- mates. The variables and subpopulations ave less extensive than one might desive for an explovatory investigation, but this limitation was imposed by the fact that these tabulations were produced as a byproduct of another study. The last two sections of the report suggest several applications of bal- anced half-sample replication to problems ofa move analytic chavacter. Thus an investigation of its use to test the hypothesis of independence in a contingency table is described, and the velationship of balanced half-sample replication to the sign test is explored, PSEUDOREPLICATION FURTHER EVALUATION AND APPLICATIONS OF THE BALANCED HALF-SAMPLE TECHNIQUE Philip J. McCarthy, New York State School of Industrial and Labor Relations, Cornell University INTRODUCTION The extensive literature on theory of sample surveys well documents the fact that commonly used complex sample designs and estimation procedures lead to approximate and extremely complicated expressions for variance estimation. Furthermore, the survey data often do not satisfy the conditions required for the application of even elementary statistical techniques of analysis. In the absence of simple methods for handling these problems, the ordinary researcher will ignore the complexities of design and merely treat the data as though they had been obtained by simple random sampling; the more sophisticated researcher will be frustrated in not being able to account for all the effects of design in his analysis. These points have been discussed in some detail in a publi- cation of the National Center for Health Statistics (Series 2, No. 14). Durbin (1967) has developed methods for designing multistage surveys in such a manner that error computations will be relatively simple, where the emphasis is on the estimation of popu- lation means, percentages, or totals. However, it is unlikely that these methods can ever be ex- tended to handle more complicated problems of statistical analysis, such as the use of multiple regression techniques, Another possible avenue of approach to these problems is through the use of independent replications of the sample design, variously referredtoas interpenetrating samples, duplicated samples, or random groups. Deming, for example, has been a consistent advocate of replicated sampling. He first wrote of it as the Tukey plan (1950), and his recent book (1960) presents references and descriptions of the appli- cations of replicated sampling to many different situations and contains a wide variety of ingenious devices that he has developed for solving partic- ular problems. A major disadvantage of replicated sampling arises from the difficulty of obtaining a sufficient number of independent replicates to provide adequate sampling stability for estimating var- iances and for other purposes. Thus the commonly used design of two primary units per stratum (frequently obtained by collapsing strata from each of which only a single unit has been drawn) gives only two independent replicates. To over- come this problem of having only one degree of freedom available for variance estimation, the U.S. Bureau of the Census originated a pseudo- replication scheme called half-sample replica- tion. The scheme was adapted and modified by the staff of the National Center for Health Statistics (NCHS) and has been used in their Health Exami- nation Survey reliability measurements. A brief description of this approach is given in a report of the U.S. Bureau of the Census (Technical Paper No. 7, 1963, p. 57), and a reference to the Census method of half-sample replication was made by Kish (1957, p. 164). The basic characteristics of this procedure have been described by Gurney (1962) and in the NCHS publication (Series 2, No. 14). A refinement of this pseudoreplication tech- nique, called balanced half-sample veplication, that increases the precision of variance estimates was also introduced in NCHS (Series 2, No. 14) and was applied to the analysis of data derived from the Health Examination Survey (NCHS, Series 11, No. 1), The present report carries further the evaluation of the balanced half-sample technique, with illustrative data drawn from the Health Examination Survey. In addition, several applications to problems of a more analytic character are suggested. Thus an investigation of the use of balanced half-sample replication to test the hypothesis of independence in a contingency table is described, and the relation- ship of balanced half-sample replication to the sign test is investigated. BACKGROUND FOR EVALUATION STUDIES In simplest terms, the idea of half-sample replication is as follows, Consider a stratified sampling procedure where two independent se- lections are made from within each of L strata, Denote the strata weights by W,h=12,...1L and the observationsby y,,, h=1, 2, . . ., L and i= 1,2, Under these circumstances, a half- sample replicate is obtained by choosing one of yy; and Yi,» One of y, and y,,,..., and one of y,, and ¥.,- A half-sample estimate of the population mean is Tu ™ p> W vy,» where i is either 1 or 2 for each h. There are 2! possible half samples, If the estimate made from the entire sample is denoted by Ver = z W, (y,, + Yo) 2, then it can be demonstrated that if k half samples are independently selected from the entire set of 2" possible half samples, with means denoted by Vos, 10 ¥is,2 Se, is, ’ then E[ ,, = 70%] is equal to the or- dinary variance estimate of Ys¢» Where the ex- pectation is taken over the entire set of 2" half samples. As is explained in NCHS (Series 2, No. 14), a balanced set of half samples is obtained by appropriately choosing the half samples so as to eliminate a between-strata contribution which influences the stability of this estimate of vari- ance, The basic characteristics of balanced half- sample replication, as outlined in the preceding paragraph, have been investigated in relation to a linear situation where the method obviously has no real utility, at least as far as ordinary variance estimation is concerned. Under these circumstances the method merely reproduces results that can be obtained by direct and simple methods of analysis. If, however, more com- plicated methods of sampling and estimation are employed, then direct methods of analysis may not be available or may require a prohibitive amount of computation in comparison with half- sample replication, For example, suppose that a multiple regression coefficient is estimated from a complex sampling and estimation operation, No known theory leads to an estimate of its sampling variability. Nevertheless, one may compute a value of the regression coefficient from each balanced half sample and use the comparisons among the computed regression coefficients to estimate the variability of the overall coefficient, Troublesome "domain of study" problems, as described by Hartley (1959) or Kish (1965), may also be handled in the same way, The simplicity of the approach is most appealing in that one has only to apply the appropriate estimation procedure to each half sample, followed by a simple vari- ance computation on the separate half-sample estimates, Since balanced half-sample replication does permit the "easy' computation of variance esti- mates which intuitively seem to mirror most of the complexities in survey sampling, estimation, and analysis, one would like to argue that it can "correctly" account for all of these features, This is clearly not the case. For example: 1. If the ¥,; represent values of avariable associated with individual population elements, if the two elements per stratum are selected with equal probabilities and without replacement, and if the sampling fractions are the same in all strata, then a finite population correction can easily be inserted at the end of the variance estimation process. The effect of differing sampling fractions in the separate strata could be taken into account by working with W's where W| is equal to W, multiplied by the square root of the appropriate finite population cor- rection, but even this simple change effectively divorces the estimation procedure from the variance estimation process. The situation be- comes more involved if the two units are selected with unequal probabilities and without replace- ment, 2. If the y,, representestimatesofa cluster characteristic, as in two-stage sampling, then one must also worry about correctly accounting for the within-cluster contribution to the total variance. As is evident in Cochran (1963, ch. 10), the problem is not substantial if the first-stage units are selected with equal probabilities and with replacement and the first-stage sampling frac- tions are small, or if the first-stage units are selected with unequal probabilities and with re- placement, The situation in which two units are selected with unequal probabilities and without replacement is the variance estimating problem to which Durbin (1967) addressed himself, 3. Most estimation problems are not linear but involve the use of ratios in one form or another. Under these circumstances, it will not even be necessarily true that the estimate made from the entire sample will be equal to the average of the balanced half-sample estimates and thus new dimensions of complexity will be involved. As implied by the foregoing examples, the exact characteristics of estimates of variance obtained from balanced half-sample replication are, for the most part, unknown, These char- acteristics may be investigated in a variety of ways, including the following: (a) exact analytic, in which one assumes the functional form of a distribution (or a joint distribution) and obtains straightforward, although possibly not simple, answers; (b) approximate analytic, in which one uses Taylor series approximations, obtains bounds by analytic methods, and the like; (c) em- pirical studies, in which one employs data obtained from surveys to investigate the behavior of sta- tistical procedures, frequently using such an approach to check on results obtained in (b); and (d) Monte Carlo sampling from synthetic popula- tions, again usually to check on results obtained in (b). In the area of present concern, there appears to be little hope of obtaining exact analytic solu- tions to problems. In particular, exact solutions require quite restrictive models as evidenced by a portion of the work of Durbin (1959) and the work of Rao and Webster (1965) on the Quenouille version of a ratio estimate, Their developments start with the assumption that y, = a + 8 x; + u,, Eu;|x) =0, Vu |x) = né where § is a constant for each n such that né is bounded, and the variates x /n have the gamma distribution with parameter h. Such investigations are ex- tremely valuable in providing clues concerning the behavior of a particular method, but it is not possible to devise specific models for the many different types of data studied by large survey organizations, Taylor series approximations have also been used extensively to investigate the properties of a wide variety of ratio estimators, both with and without models, One of the more recent accounts and summaries of this type of approach is that of Tin (1965). In general, results are ob- tained on bias and variance estimation which one wishes to check against data obtained either from actual studies or from Monte Carlo experi- ments, particularly when the samples are small or moderate in size. Thus Kish, Namboodiri, and Pillai (1962) use actual survey data while Tin (1965) and Lauh and Williams (1963) use Monte Carlo techniques. In this report, the general view is adopted that the behavior of estimates and of variance estimates should be investigated, if possible, on the basis of actual survey results. This enables one to consider a wide variety of types of data that are of current interest and avoids the problem of having to construct "representative" synthetic populations and having the results refer only to these populations. Even the con- struction of appropriate synthetic populations, to which one could apply complex sample de- signs and estimation procedures, would be a difficult undertaking. DESCRIPTION AND APPLICATION OF EVALUATION TECHNIQUES This section of the report will describe a number of internal evaluation techniques which can provide evidence concerning the behavior of balanced half-sample replication. These tech- niques will be illustrated on data derived from the Health Examination Survey (NCHS, Series 11, No. 1, and the appendix to NCHS, Series 2, No. 14) relating to a variety of body measurements for a sample of approximately 3,000 U.S. adult males, (The total sample also contained approxi- mately 4,000 U.S. adult females.) Two different types of estimates will be considered: (1) Esti- mates of population means and (2) Estimates of multiple regression coefficients and multiple correlation coefficients, The original tabulations upon which these results are based were pro- vided by the Survey Research Center of the University of Michigan as a portion of the work performed under a contract between the Survey Research Center and the National Center for Health Statistics. The variables and subpopula- tions are less extensive than one might desire for an exploratory investigation, but this limita- tion was imposed by the fact that these tabu- lations were produced as a byproduct of another study. The Health Examination Survey isan example of the combination of a highly complex sample design and estimation procedure, In brief sum- mary, the roughly 1,900 primary sampling units (PSU's) of the Current Population Survey were grouped into 42 strata, some of which contained only a single PSU that was then included in the sample with certainty, From each of the strata containing more than one PSU, a single PSU was selected and then subsampled in accordance with customary practice to obtain a sample of about 160 persons (75 males). The estimation procedure included four principal operations: (1) Inflation by the reciprocal of the probability of selection, (2) a first-stage ratio adjustment to 1960 popula- tion for eight geographic and population concen- tration classes, (3) anadjustment for nonresponse carried out in 294 age-sex-PSU cells, and (4) a poststratification by 12 age-sex cells. These observations were forced into a ''two observa- tions from each of 27 strata" format by pairing "adjacent" noncertainty strata and by defining "similar" subsamples from within the certainty PSU's, Average of Half-Sample Estimates As mentioned earlier, the formal consider- ation of balanced half samples has been restricted to the ordinary linear situation, One can move away from the linear case in a variety of ways, but only the ordinary combined ratio estimate will be discussed here, The points raised, however, apply to any nonlinear situation and will be illustrated in Health Examination Survey data, In the case of stratified simple random sampling, with two units selected from within each of L strata, there are a totality of 2 half samples, from each of which one may compute a combined ratio estimate, say r,. For the linear case, it has been shown that analyses carried out on a set of balanced half samples give the same re- sults as would be obtained by carrying out the same half-sample analyses on the entire set of 2" elements. We shall now investigate the man- ner in which this carries over to the case of half-sample ratio estimates. In the linear case, denote the means of a balanced set of half samples by Ys 1 Fs, gt «ven Vrs, and the mean of the entire sample by ¥si- Denote the means of the complementary set of half samples by Fos1* Vns,2: © Yhe "* where y! s,i 1s the mean of the half sample ob- tained using the L elements left out of the ith original half sample. If k is greater than L— so that each observation appears in half of the samples--certain relationships hold among these means, In particular, we have that Fs Th v. RE: - i lk ET Ik = Vist - In a sense, we can say that the set of k balanced half sam- ples "exhausts" the information in the totality of 2" half samples, as far as estimation is con- cerned. The question can then be raised of whether the half-sample combined ratio estimators ys» rp, ..., rn, separately from or in com- bination with their complementary estimators By Boy wy r,, also "exhaust" the totality of all 2" half-sample ratio estimators, Notice that we are here asking about the relationship be- tween 2,1 /k 2, ri/k, and the average of the r;'s for all 2 half samples and not about the relationship between these quantities and’ R, the combined ratio estimator for the entire sample. The latter is a separate question which will be discussed shortly. It appears that this problem can be attacked empirically by the following argument, Consider by way of illustration, the three-strata example discussed in the NCHS report (Series 2, No. 14, p. 17), where the observations are denoted by pir Yds h=1,2,3 i=1,2 and W, = W, = W, = 3. Translated into combined ratio estimate terms, for a set of four balanced half samples, this is Yio t Yaz t V3 + X ynt Yat Ya : n= X.. + X + Xx X,, + X 11 21 31 12 22 32 Yip t Ya t Ya yo Nt Ya tra r,= Xp; +t Xp + Xz Xpp + Xp + Xg Yip t Yao t Ya Lon tt ya tt Ya r3= Xj, + Xp + X3 Xt Xp * Xp 12 FR: Yat Ya hm 21 + Vy + Ya - do X;p + Xp + Xz xX; + Xp + X5 In the r, set, note that each observation (x,,» yn) occurs in two of the four samples. Furthermore, if we look at any two of the three strata, say strata h and k, then each of the four possible pairs of observations (Xp; » Yor) and Xpq Vit ( yy) and & ) Xn’ k2’ Yk2 ( ) and (x ) Xh2r Yho x1’ Yi (X55 Yn! and Xr Y2) appears in one of the four samples. It is this "balancing of pairs of observations’ which char- acterizes any set of balanced half samples and, in the linear case, makes them so suitable as the basis for variance-estimation, The set of r,'s is not, however, balanced on triplets of ob- servations, Thus the triplet (X00 Yip) (pps Yop?» (x5, , ¥3,),- obtained by taking one observation from each of the three strata, does not appear in the r, set. Similar observations hold for the rf set. The combined set of r's and r!'s in this case does account for all possible half-sample ratio estimates and is therefore balanced on triplets. If, ina particular example involving three strata, we found that 7 and 7', the averages of the r's and r|'s, respectively, were in close agreement, this would indicate that the effect of the balancing of triplets was negligible, Of course, the average of r and r'is in this in- stance the actual value that one wishes to esti- mate, Even for the artificial and extreme example given in NCHS report (Series 2, No. 14, p. 27), E(7) = 1.1985 and E(r") = 1.1968, thusindicat- ing that the effect of triplet balancing is ex- tremely small, In general, a set of k half samples, where k is a multiple of four, will be balanced on the number of strata corresponding to the highest power of two that is an exact divisor of k. Thus if a set of eight balanced half samples is used for five strata, then the r, and r; sets will each be balanced on singles, pairs, and triplets of observations while the combined set will also be balanced on quadruplets, Under these circum- stances, if ¥ and T' agree, then one can argue that quadruplet balancing is unnecessary, al- though this can be achieved by combining the two sets, When k is not exactly equal to two raised to an integer power, and the Plackett-Burman (1943-46) method of constructing an orthogonal matrix is used, the balancing situation is not as clear cut, For example, the case of k =12 has been completely investigated, Thus the r, and r! sets are both balanced for pairs of obser- vations since 12 is divisible by four, Furthermore, the combined r, and r; sets, which contain 24 half samples, are balanced on triplets of obser- vations. On the other hand, for the case of k = 24 which is given in NCHS report (Series 2, No. 14, p. 19), the r, and r; sets are each balanced on triplets of observations, since 24 is divisible by eight, but the combined set of 48 half samples is not balanced on quadruplets of observations even though 48 is divisible by 16. Nevertheless, a small investigation! has indicated that the combined sets will be better balanced on quadruplets than will either set separately. In ordinary practice one would perform all computations on a single set of k balanced half samples and would not be concerned with the complementary set. The two sets can, however, be compared if there is any doubt concerning the adequacy of a single set, This comparison has been made, for methodological purposes, with data derived from the Health Examination Survey which, as noted earlier, was based on approximately 3,000 adult males, The analyses were performed on a set of 28 balanced half samples, the sample being treated as if two independent selections had been made within each of 27 strata, together with the complementary set of 28 half samples, Two different types of estimates were considered: 1. Ratio estimates of population means were made for 15 physical body characteristics (e.g., chest girth, waist girth, and knee height). 2. Each of eight body characteristics were separately regressed onthe independent variables of age, height, and weight, Estimates were made of the partial regression coefficients and of the multiple correlation coefficient, In each case-—15 means, 24 partial regression coefficients, and eight multiple correlation co- efficients—the average of the estimates made from the original set of 28 balanced half samples can be compared with the average of the esti- mates made from the complementary set, These comparisons are given in tables 1 and 2, In only seven of the 47 comparisons do the two means differ by more than .03 percent of their average and these instances could well have arisen from rounding error. Hence we conclude that for this situation half-sample estimates, based on a set of 28 half samples, effectively give the same re- pen groups of four columns were selected randomly from the population of 21Cs possible groups of four columns. Per- fect balance was LT by combining the r, and ri sets in five of the 10. Improved balance was achieved in four of the 10; and in only one of the 10 did no imnrovement occur. sults as would be obtained by working with the entire set of 227 half samples. Bias of a Ratio Estimator The estimate that is customarily used for the type of sample design to which the present discussion refers is the combined ratio esti- mator which, when based on the total sample, will be designated by R. This estimator may be biased, especially for small samples, and this topic has been extensively researched in a wide variety of ways, e.g., Kish, Namboodiri, and Pillai (1962) and Tin (1965). When balanced half samples are used for variance estimation, it is possible to obtain, as a byproduct, an empirical check on the existence and magnitude of bias. The Quangities r, T', and their average, de- noted by r* are of exactly the same form as R, but are ’ passed on half samples. Since the bias of a ratio estimator decreases with in- creasing sample size, we can expect 7 r* to be subject to greater bias than R. If, for a given set of data, 7, 7', ¥* and R are essentially the same, this is presumptive evidence that the dif- ferential bias is close to zero and it therefore follows that the absolute bias of either estimator is essentially zero. For the artificial example used in NCHS report (Series 2, No. 14, p. 27), this is clearly not the case since E(F) = 1.1985, E(r') = 1.1968, ER) = 1. 1666, and R = 1.1548. It is true that the bias in R may be small even though there is differential bias between 7 and R. On the basis of previously cited empirical investigations and on the basis of the size of the sample upon which the illustrative data of this paper are based, one would expect little, if any, evidence of bias, This expectation is borne out by the summary results presented intable A, which is based on the individual values given in column 7 of table 1 and column 8 of table 2, As can be observed from the table, the differential bias of the 15 ratio estimates of population means is negligible, although one does observe that the sign of the difference is negative in 12 of the 15 cases, possibly indicating some small residual bias in the half-sample estimators. The results for the estimates of partial regression coeffi- cients and multiple correlation coefficients lead to the same general conclusion, although the Table A. Differential bias of 7* and R, as a fraction of the estimated standard error of 1 Estimates of partial regression Ratio coeffi- ‘ A 3 estimates cients an - R/sR of popula- | multiple tion means correla- tion coeffi- cients -.30 to -.25-==-- - 1 -.25 to -,20--=--- - 1 -.20 to =,15=-=--- - - -.15 to -.,10-=--=-- - 2 -.10 to -.,05=----- - 4 -.05 to ,00---=--- 13 11 .00 to +.05--=---- 2 5 do 05 to +.10-=-=-- - 7 +.10 to +.15----- - 1 Total=--=--=-======= 15 32 Mean-----=-=-==-==--= -.016 -.018 Median----=-===--= -.020 -.011 Isee the text for a description of the data upon which these estimates of dif- ferential bias are used. Individual val- ues from which these distributions are derived are given in column 7 of table 1 and column 8 of table 2, values of the differential bias (relative to the standard error) tend to be somewhat larger than in the former case. The three largest (in ab- solute value) of these arise from the independent variable, height, but no specific reason for this could be found. In general, one must conclude that, for this type of data and for this size of sample, there is no need to be concerned about bias in either the half-sample estimators or, more particularly, in the entire-sample esti- mator. If the possibility of ''serious' bias appears to exist for any or all of 7, ¥', 7* and R, as evidenced by large differences among these quan- tities in a specific situation, one can consider the alternative of moving to a Quenouille-type (or Jackknife-type) transformation. There is some evidence, primarily based on Taylor series investigations of the ordinary ratio estimate, that such transformations may reduce bias with- out having deleterious effects on the variance of the estimate, (See, for example, Brillinger (1964), Durbin (1959), NCHS (Series 2, No. 14), Quenouille (1956), Rao (1965), and Tin (1965). There have been no theoretical or empirical investigations of complex design and estimation situations similar to the one being discussed in this paper. Estimating the Variance of 7 Under ordinary circumstances, one has a set of k balanced half samples and the accompanying half-sample ratioestimates, ry, r,, . . ., I. Carrying over the argument outlined earlier for the linear case, the variance of R (ratio estimate computed from the entire sample) would be esti- mated as |Z, (r, — R)%/k. To keep the argument in simplest terms, we shall first restrict atten- tion to the problem of estimating the variance of r instead of R and hence use the variance estimate 2 (r, — F y2/k, here denoted by Vous (7). If the data for the complementary set of half samples are alsoprocessed, as is the case for the illustrative example being used in this paper, then a second estimate of the same form can be computed from ry, hus RE. 2s namely Vous (¥". In the linear case, these two estimates are identical. Another estimate of variance with which to compare Vg, (7) and V V,us(T) may be obtained by the following argument, Viewing a particular half sample and its complement as two independ- ent samples, the variance of a half-sample ratio can be estimated with a single degree of free- dom as +r A : | (r, - AA-7 + Walliy = (5) ~ 2? The average of r, and r!, say rf, makes use of all the sample information and has estimated variance Gat AP If this estimate of variance is obtained from a single half sample and its complement, then it is the simplest form of a replicated estimate of variance and has often been suggested as a crude first approximation to such variances. In the linear case, the value of 7* would be equal to the whole-sample estimator for any : and hence we shall regard (Ya) {r,- rH? as an approxi- mation to the variance of the whole-sample esti- mator, whether it be computed as 7, ¥', or as Th. If the foregoing estimate of variance could be computed for each of the 2'~! complementary pairs of half samples in the entire set of 2' half samples, then the average of these estimates, each of which has the correct expected value, would appear to provide an ''excellent'' estimate of the variance r*. Since this is impossible, we shall substitute its computation over a set of k Table B, Difference 8 of A between LE 2 CBHS balanced half samples, together of course with the complementary set. Thus A — 1 X 1y2 Yours” )= Ca) Z (ri- rk . It seems likely, on the basis of the argument previously set forth on the balancing of singles, pairs, and triplets, and of computations carried out on simple examples, that these estimates are extremely close to those that would be obtained by working with the entire set of 2% half sam- ples. Although no way has been found to put the foregoing considerations on a formal basis, it would appear that Vegus (79 is a "better" esti- mate of variance than either Vous!) or VV (FY In practice, of course, one would have avail- able only one of the latter two estimates, In the A A = _ ; (ops 7 ), and Viegus (T"), as a fraction Fractional difference Ratio estimates of population means Estimates of partial regression coefficients and multiple correlation coefficients BHS A Vv (7) Lr HFFENDNDNONT NI HI oN PNONFERFENT PL I FEFFRFULWWNT Tr 11 NFEPPNNERNUONDNHERF 0 15 -.0l1 -.018 15 +.019 +.017 * o = Ww oN +.012 +.001 -.001 present investigation, all three estimates have been computed for the 15 ratio estimates of population means and for the previously de- scribed regression coefficents and multiple cor- relation coefficients. Agreement of these three variance estimators will be taken as presumptive evidence that the balanced half-sample variance estimation procedure is appropriate for a given set of data. A summary of the results of this comparison is given in table B, while the actual values on which these distributions are basedare presented in tables 3 and 4. It will be observed that the distributions of fractional differences show the three estimates to be in relatively close agreement, Several further observations con- cerning these distributions are: A 1. For any given estimate AL tends to be between V., 7? and Vo(F . 2. For the “ratio estimates of population means (physical body characteristics of U.S. adult males), the estimates Vi, £7) tend to be on the same side of V oF "Tn other words, the values of V. Vos) are correlated with one another. This effect is not as pronounced for the estimates of partial regression coefficients and multiple correlation coefficients. 3. The effects of (1) and (2) show up in the two distributions on the left of table B, one distribution being located somewhat below zero and the other somewhat above zero. 4, If the comparison were made on the basis of estimated standard errors, the distributions would show less dispersion, This follows im- mediately from the fact that (Va — vB)AM/b=(a/ /By—1 is always smaller in absolute value than is (a — b)/b = (a/b) — 1. The foregoing results can be presented in a slightly different manner by introducing the con- cept of correlated variables. Walsh (1947) sum- marizes the following results, If x;, x,,. . ., x, are observations where E(x) = u, V(x) = a? Zr and E(x, joM) (x -w) = pa? then VT) [1+ - 1p) and Elz, -0%/(n-D]= 021g. Furthermore, if the x's are normal, the quanti- ty & — wl =p/a+ (n= Dp] 72 T(x -%)%/n(n — 1] 72 1 has a Student's distribution with (n — 1) degrees of freedom. For present purposes, we replace the x's by the balanced half- sample ratio esti- mators Fys Igy t= na Ger The 1, 's do have a common variance, and they are correlated since the half samples within a balanced set do have elements in common. For the moment, the com- plementary set is ignored. The ordinary practice for balanced half- sample replication has been totake I(r, - RVk as an estimate of the variance of R. As has been done earlier in this discussion, let us consider r rather than Rand see under what circum- stances we can treat Fu, Zr; — 7)? k as having a Student's t distribution with (k —1) degrees of freedom, at least as far as variance considerations are concerned. If the aforementioned equality is tohold, then it is necessary that 1+ (k—Dp Ao = k= which means that k-2 P= 2k-1 . This condition can be shown to hold, through the use of a common-element argument set forth in NCHS (Series 2, No. 14, pp. 20-21), if 1. We are dealing with a lpesy situation in which W,=W,=...=W_ and S =s?=...=5% 2, The number of strata, L is %ne legs than a multiple of four and the set of k=L +1 bal- anced half samples is obtained by deleting the column whose sign entries are all the same from a k x k orthogonal matrix, In the present instance, and in more complicated ones, these conditions cannot be expected to hold exactly, The strata weights and variances will not be equal; the half samples may not have exactly the same number of elements in common, which was required for the "common element" argument leading to a theoretical value of p; and a ratio estimator may behave in a different fashion from a linear estimator. Nevertheless, one would expect these results to hold in some average sense, especially when there isa reason- able number of strata. It is possible to investi- gate this matter empirically. Suppose we consider a set of balanced half samples and their complements, together with the half-sample ratio estimators r rand zd, Ef Then take 12 Bay wo sy ws A k wey — gl 2 Weans T= (V2) 2 tr, = tH%k as an estimate of ¢2 the variance of a single half-sample estimator. Furthermore, take as estimators of ¢2 (1- p). The correlation p can then be estimated as X 2 2 (r=) (k-1 I= w= 1 k da) Zr -h¥k i=1 or as 3 (rl =TH/(k = 1) Pe1oi=1 7 Ls, 2 (72) Zz, (r, —rH7k a and these values, or their average, can be checked against the theoretical value of (k — 2)/2(k - 1, The foregoing estimates of p have been com- puted for the illustrative data used in this paper. Their distributions are shown in table C and the individual values are given in tables 3 and 4. The left portion of table C presents the results for ratio estimates of mean body characteristics of U.S. adult males, while the right portion gives the same data for regression coefficients and mul- tiple correlation coefficients, The theoretical value of p with which these can be compared is 10 (k—2)/2(k — 1) = 26/54 = .4815. The general agreement of the empirical results with the theoretical value again leads to the conclusion that, for the illustrative data being used in this paper, one cantake Z (r, — %k as an approxi- mation to i=) x 2 I, 0-0 [1+ &k=1p] kk -1 a -p) in estimating the variance of ¥. The general ob- servations made concerning the relationship among Gistiibotions in table B also apply here since the 7's are simple functions of Vous'Th A TT * Yous 71 and Vo (7 Estimating the Variance of R Thus far the discussion of variance esti- mation has focusedon 7, 7', and r* rather than R, for the simple reason that we are actually dealing with half-sample ratio estimators. The ordinary procedure in either half-sample or balanced half- - Sample replication has been to use the quantity 2 cr, —RYk as an estimator of the variance ‘of R. This estimate of variance is larger than dhe estimate 2 (r, -™%k by the amount (7 — R)2, Since we have already ob- served that ¥ and r do not differ by an appreci- able amount from R for the illustrative data of this paper, it makes little difference which esti- mate is used. In general, it would seem that - Rk would overestimate even the i= square error of R since it contains a contribution for the differential bias of 7 and R, and this differential bias may well be larger than the absolute bias in R. It is possible to obtain 2 formal relation- ship connecting V(r) and V(R).Consider the ex- pression EF — R)2. It follows that _ A A A EF-R)"'=V@ + V(R) — 2 cov(F, R) +[E® - ER)? A / [A =V(D + VR = 2 p_ AVVO VVR) + [Em — ER)? Table C. Estimated value of the intraclass correlation coefficient for a set of 28 balanced half-sample estimators Estimates of partial Ratio estimates regression coefficients of population means and multiple correlation Estimated value of p coefficients b p b p A433 to Jb4bemmmmmmm meme mm - - 1 - Ab to J45----mmmmmmmmmm meme - - 2 2 JA45 to J4bem--mmmmmmmmmmm mmm m mmm - 1 3 2 46 to L4T----mmmmmmmmmm mmm mmm - 6 5 4 47 to 48----memmmmmmem meme 3 4 8 6 48 to ,49-----mmmmmmmmmmmm mmm 5 4 6 10 49 to ,50==----m-mmmmmmmmmm mmm 6 - 4 5 ,50 to ,5l=-------mmmmmmmm mmm 1 - 3 1 51 to ,52==-----mmmmmmmmmmm mmm - - - 1 .52 to ,53----==-mmmmmmmmmm mmm - - - 1 Total=-=====-==-==c--m-ooooo-o=—- 15 15 32 32 Mean---=======-===-==---o----=-== 4876 4734 4762 4811 Median---=-====-=--=e=m--=-c----=- .4890 4724 4754 .4838 Regrouping of terms leads to VF) and V(R) on the basis of a single sample, A / / NA VF) + V(R) = 2p_A VV) VV(R) A A -EF-R?- [Em - ER)’ Now we observe that p;g <1. Actually, it ap- pears that this IT will be very close to one under almost any circumstances, and this conjecture has been confirmed by direct com- putation in several "small," synthetic examples. Thus N / fA V(r) + VIR) — 2 VV) V(R) Ei - RY? _ [E® — ER) < A RY? oF SEF -R) 2 /[ / A A | vr) — V(R) < E(f-R). Since, as shown earlier, we can estimate VF) from a single sample, and since (FT — R»? is an approximation for E(F — R)? this last ex- pression provides a crude way of comparing The bounds obtained will not be particularly good if the neglected term [E() — ER) is at all appreciable in relation to E(r — R)2, andthis has been the case with various Synietic examples that have been investigated. On the other hand, this may provide quite reasonable bounds in practical situations where the neglected term may be expected to be small, It should be observed that the development of the final expression includes sampling without replacement from finite populations, Furthermore, equality will hold throughout if (r; + r{)/2 =R for each i, as it does in the linear case, and we then have the re- sult that V(F) = V(R). For the illustrative data of this paper, as summarized in table A, all indica- tions are that we are obtaining ''good' estimates of V(R) as well as of V(P. Design Effect As a final point, we note that column 2 in table 3 and column 3 in table 4 contain esti- mates of variance, labeled Van) , computed as if the observations had arisen from simple ran- n dom sampling. Since the estimate Vos) takes into account most of the survey features (strati- fication, clustering, poststratification estimation, etc,), the ratio of Vepnst? 10 Vi) is an estimate of the design effect. These ratios are given in column 8 of table 3 and column 9 in table 4, In table 3, which refers to estimates of mean body characteristics based on approximately 3,000 adult males, these ratios range from 1.158 to 5.729 with an average value of 3.223, In table 4, which refers to estimates of regression co- efficients and multiple correlation coefficients, the ratios range from .570 to 3.675 with an average value of 1,807. The fact that design effect is smaller for the regression coefficients than for the mean body characteristics is con- sistent with other studies. APPROXIMATE TEST OF INDEPENDENCE BASED ON A SET OF BALANCED HALF-SAMPLE REPLICATES The statistical analysis of data that have been produced by the use of a complex sample design and then processed through an involved estimation procedure poses many difficult prob- lems, Although it appears that the application of some form of pseudoreplication will provide a reasonable solution to the problem of esti- mating the sampling variability of a statistic, it is not immediately apparent that one can use these same techniques to solve certain other commonly occurring problems of statistical analysis. This section gives a summary of the results of an investigation carried out by Chap- man (1966) concerning the use of balanced half- sample replications to test the hypothesis of independence in a contingency table, The contingency-table test of independence is ordinarily phrased in the following manner, Observations are classified jointly on each of two qualitative variables, the categories of variable one being denoted by 1, i= 1, 2,.. .r and those of variable two by j;, j =1,2,. . .,c. If p;; denotes the probability of an observation falling in the ith category of variable one and in the jth category of variable two, then the hypothesis of independence states that Pp; = p, 1. pj where p, refers to the marginal distribution 12 of the first variable and p refers to the mar- ginal distribution of the second variable, As- suming that a "large" sample of n independent observations is available, the statistic is approximately distributed as Chi-square with (r — D(c - 1) degrees of freedom when the hypothesis is true, and the hypothesis is re- jected for '"large' values of X2. Data derived from a complex survey operation certainly do not conform to this model, particularly in view of the dependencies that are introduced by the use of stratification and clustering techniques. Furthermore, the effects of commonly used estimation procedures would add to these com- plications. One might attempt to find a solution to this problem in the following manner, If large samples are used, the distribution of the sample esti- mates of (rc — 1) of the p's, which will be denoted by By. can be approximated by a non- degenerate, normal, multivariate distribution. The exponent (ignoring the - 1/2) inthis distribu- tion is a quadratic form in the p's and the coefficients of this form are the elements in the inverse of the variance-covariance matrix of the 3, 's. The variance-covariance matrix of the by; 's can be estimated by a pseudoreplication method, such as balanced half samples, and these estimates would mirror the effects of the design and estimation, It should be observed, however, that the volume of computation would be large since even for r=3 and c¢ = 3 there would be eight variances and 28 covariances, It would then be necessary to invert this matrix and evaluate the quadratic form to take the hypothesis p, = p; p, into account, The resulting statistic would be treated as a Chi-square variable with some appropriate number of degrees of freedom. This approach has not been pursued since we shall now describe a "simpler" alternative. The basic idea upon which the proposed test procedure rests is the following. Suppose that an estimate By; is obtained from a sample and that a product estimate, b, J is obtained from an independent sample, Form the difference of these two estimates p,, = p, Pp, ), and consider the sign of the difference. Define the variable S, to be one when the sign is plus and zero when it is minus. If the hypothesis of independence does hold, then this is the difference of two independent estimates of the parameter, py, and one would intuitively expect the sign of this difference to be plus or minus with equal probability. Actually, even with two independent random samples, each of size n, it can be shown that this result does not generally hold. In particular, 1. If nis "small," then there may be an appreciable probability that the difference is exactly equal to zero. 2. The distributions of py and p Pp, will not be identical and therefore the distribution of the difference may not have a median at zero. For example, it can be demonstrated that VB) = (mp, (=p) vp |B) = W/m) p, A= p> ~[n-v/m?)p a-p0-p). On the othe hand, if we can assume that both p, and p! BY are approximately normally distributed, then their difference will also be approximately normally distributed and the mean of this distribution will be zero when the hypothe- sis of independence is true, Hereafter, we shall make this assumption of approximate normality. Now suppose that it were possible to repli- cate this situation k times and thus obtain k independent observations on the variable Si The sum of the k S's xX, , is a binomial vari- able with EX) =(k|2) VX) = (Yark when the null hypothesis is true and, if k is "large," the quantity k 2 X _- 5 fark is distributed approximately as Chi-square with one degree of freedom. In general it would be impossible to con- sider using this form of analysis since it re- quires 2k independent replicates, each of which is large enough to ensure that py; and p! 3 are approximately normally distributed. Further- more, it is also desirable that k be sufficiently large so that the test will have adequate power. However, if one is using a sample design for which a balanced half-sample replication method of estimating variances is appropriate, itappears that these circumstances can be realized in an approximate manner. Consider, as an example, a situation in which there are two independent selections from within each of seven strata and that one is therefore using a set of eight bal- anced half samples. Denoting the first element selected in each stratum by a + and the second by a —, the following table shows one set of eight balanced half samples and their comple- ments: Half Strata Comple- Strata samples | 1 2 3 4 567 ments 12%45867 1 + + + + + + + l] [| === ==- 2 - + = + = + = 2 EE BD I 3 - t+ + = = + + 4 -- t+ + = =F 4 Ft = =k + = 5 I 5 --- + + + + 6 - + = =F =F 6 EE 7 bom = == + 7 - + + + + = = 8 --t = + + = 8 + + = = = + It will be observed that a half sample and its complement have no elements in common, that a half sample and any other half sample in the same set have three elements in common, and that a half sample and any half sample from the complementary set, other than the complement itself, have four elements in common, In general, if there are L strata and one can use a set of (L + 1) = k balanced half samples, where k is a multiple of four, these numbers of common elements will be zero, (k/2)- 1, and (k/2), respectively. 13 Now consider the following simplified ver- sion of the stated problem: . 1. By is the estimate of p, made from half-sample one and 5! \P is the corre- sponding product estimate obtained from the complementary half sample. These are independ- ent estimators, 2. ,p, is the estimate of p; made from half sample two and ,5, eo is the corre- sponding product estimate obtained from the complementary half sample. 3. Assume that Bi and 2, are ordinary linear estimates based on L Stearn of equal weight and that the within-strata estimates have common variance s2 It then follows, using the argument set forth in NCHS (Series 2, No. 14, p. 21), that {lw _ n/2]} foe?) g2 A A Cov GP; 2P;)) (L-1) 2 er S, - 4. Assume that the product estimates | p 1B behave in exactly the same manner as the pb, Ss. As a result, we have A A A A (L-1) 2 Cov,p| Pj, oP] P= ETE and At A Ay Ary _ Ay A Cov (,B};, ,B| 2P)= Cov(,p,., \b PY -l = {uw +v2 1/12 s, (L - ek 52 2L 5. Using the foregoing results and assump- tions, we find that A Ai A Var(,p;, — ,P; P= Var (, B,, 2Pi. oP. and that the correlation between the two differ- ences is equal to —1/L. For any reasonable num- 14 ber of strata this correlation will be very close to zero, Since the various estimates are assumed to be approximately normally distributed, it fol- lows that the k differences B, - pl p' ), as computed from a set of balanced half samples, can be treated as independent observations. Some evidence concerning the reasonableness of this type of argument is provided by the data on the intraclass correlation coefficient given in table C. It was shown that the estimated values agreed quite well with the theoretical value predicted on the basis of the "common element' approach. The application of the foregoing procedure to a set of k balanced half samples leads to an ob- served value of X, for each of the rc cells inthe contingency table, where EX) = k/2 VIX) = (ark when the hypothesis of independence is true, We would, of course, like to combine these rc values into one overall test, However, such a combination cannot be carried out in a straightforward man- ner since the X, are not independent of one another, Chapman (1966) presents an extensive analysis of this problem which culminates in the suggestion of an approximate test statistic that appears to have reasonable properties and which is simple to use in practice, The basic steps in the development of the statistic are: 1. The rc variables X;; are assumed to have a nondegenerate, normal, multivariate dis- tribution, Although there are constraints on the X's. namely that they cannot all be simul- taneously equal to zero or one, these constraints are not of a form that reduces the rank of the quadratic form, 2. Approximations to the covariance and the correlation of any two Xx, ;'s are obtained in the following manner: a, If two normal variables, with zero means and unit variances, Y, and Y, have a normal bivariate dis- tribution with correlation p(Y,, Y), then the correlation p (S, »Sp between their signs—i.e., S,=1 if VY, >0, and S, =0 if Y, <0,and similarly for Y, —is given by 2 ; p(S;,S) = — arc sin p(Y,Y). This result is given in Cramér (1946, p. 290). b. Assuming that B, and p' p' are based on random samples, each of size n, from a population in which p= (fre), p, = Jr), p= (1/c), it can be shown that the average correlation between ph, _ Bb! A and (p,, - p, P'), where either or both of i +h and ; +t are true, is equal to — 1/trc — 1). c. Thus we have the result that 2 y PX Xp) = 5 arc sin [- 1/tre - D] where either or both of 7+ h and j # t hold. Chapman (1966) inves- tigates some aspects of this approxi- mation procedure, It appears to be a reasonable one except under ex- treme circumstances, e.g., where one or more of the p; are ex- tremely small and other p,; are ex- tremely large. 3. Walsh (1947) proves that if one has n normal variables, X;, X,,...,X,, each with mean pu and variance ¢2 and with a common cor- relation coefficient p then 1 n = T= ——— I (X -X°? ogc (1-— p) i=1 | has a x?distribution with (n — 1) degrees of freedom and ro n(X — w? ?1+n-1p] has an independent x? distribution with one de- gree of freedom, 4, Combining the preceding results, we ob- tain the test statistic Gok - 2 are sin- —L_)] rc—1 which, when the hypothesis of independence is true, has approximately a x? distribution with (rc — 1) degrees of freedom. The hypothesis of independence is rejected for 'significantly" large valuesof T. Since arc sin x=x when x is small, this can usually be simplified to Hark [1+ mA fre— 10]. 5. The use of the independent Chi-square with one degree of freedom — k 2 rc(X — 2 ) Te )] Mark [1 + (rc=1D 2 arc sin (— rc—1 is not recommended since E(X) will tend to be close to k/2 whether the hypothesis is true or not, Under these circumstances, the inclusion of T' in the test statistic will weaken the power of the test. ORDER STATISTICS OF A SET OF BALANCED HALF-SAMPLE ESTIMATES A discussion of the relationship between balanced half-sample replication and the sign test is presented in NCHS (Series 2, No. 14, pp. 20-23). The problem can briefly be stated as follows. Suppose that each half-sample replica- tion provides an estimate, say dpi of a popu- lation difference, and that one wishes to test the hypothesis that the dg ;'s were drawn from a population whose median is equal to zero. If the dy, ,'s were independent of one another, then one solution to this problem would be ob- tained by a straightforward application of the ordinary sign test. The half-sample estimates are not, however, independent of one another, as is observed in the earlier sections of this paper, and the question can be raised of whether it is possible to devise some suitable modification of the sign test, As soon as one introduces dependence among the members of a set of observations, it is no longer possible in general to devise nonpara- metric or distribution-free procedures. If, how- ever, it is reasonable to assume that the dy, si 's have a normal multivariate distribution, then éx- isting tables can, under certain circumstances, be applied to test the stated hypothesis. Thus, Gupta (1963, p. 817) provides a table which gives the "probability that N standard normal random variables with common correlation p are simul- taneously less than or equal to H." In particular, if H is taken to be zero, the appropriate prob- abilities form the basis for obtaining the levels of significance for an "'extreme'' sign test, i.e., all signs plus or all signs minus. For any reasonably large number of bal- anced half-sample estimates, the Gupta tables are not especially helpful, To obtain meaningful levels of significance, one must be able to move in from the extremes on the distribution of the number of plus signs. Since such tables of the normal multivariate distribution do not presently exist, we shall here present a procedure for approximating these probabilities. The degree of approximation appears to be quite acceptable for ordinary applications of the sign test, Consider n independent an normally dob uted varjables XX, , X, with EX) = and EX? y=1. Ie he Collection {x} is ree so that XO 2 x? 2 x , then certain moments of the xV have been extensively tabulated. AP porticglar, Jeiswoew (1956) has tabulated Ex") and EXP XY) for n= 1 20 and Harter (1960) has tabulated EX") for n= 2 I 100, plus values for a number of additional n < 400. Furthermore, Owen and Steck (1962) have shown that the moments of the order statistics of a sample from the equicorrelated, multivariate, normal distribution can be readily obtained from the corresponding moments for n independent variables. Thus, if Z,, Z,,..., Z, are jointly distributed random variables with EZ)=0, EZ} =1, and E(Z, Z)=p for i +j, EZ. a- 2 Ex™) and o : i 2 E[Zz- EZ) = » + a-p EXP - EX] Our present concern is with the probability that U out of the n values Z,Z,,. . ., Z, are positive, where U takes on the values 0,1,2,... . For example, for the firstthree values, itis clear that Pri == PriZzV gs » PreU=0 or D=Prz®< om 3) Pr(U =0 or 1 or 2) =Pr(Z"7° <0). Under ordinary circumstances the calculation of the probabilities on the right side of these expressions would be extremely difficult since they Just Xe obtained from the distributions of zw, z® ., and this is the extreme-value problem that has been considered in some detail by Gumbel (1958). In the present problem, how- ever, moderate values of p, i.e., in the neighbor- hood of 1/2, appear to ensure that the distribu- tions of the 2 can be approximated reasonably well by a normal distribution. Some third and fourth moments of the extreme order statistic, for various values of p, are exhibited by Owen and Steck (1962), and these values confirm this observation, Owen and Steck note that this be- havior is to be expected as p increases since Zz" has a unit normal distribution when »p = 1. As an illustration of this approach consider the case where n=20 and p = .50. From Teichroew's table EXD) = 1.86748 2x Py= 27568, therefore Ez) = 1.32051 02zM)= 63784 . Table D. common mean (less than their common mean) Probability that U or fewer, out of n normal correlated observations, will be greater than their Number 16 20 24 48 of obser- vations 40 .50 40 .50 40 .50 40 .50 U O---=---=-=- .0361 .0603 .0272 . 0492 .0235 +0433 .0096 «+0227 1(.0335) | 1(.0588) | 1(.0251) | 1(.0476) | 1(.0197) | 1(.0440) - - l---eemmme- .0806 ,1178 .0607 .0956 . 0487 .0809 .0193 . 0415 2ecmmme mee .1346 .1746 .1015 .1430 .0808 .1204 .0321 .0615 Gr - - - - - - .0467 .0818 A. - - - - - - .0628 .1022 Ble ismviiomine swell s - - - - - .0805 .1226 IThe values in parentheses are exact and are from Gupta's tables. Assuming a normal distribution for zW ; obtain we Prz® <0)= .0492. The correct value for this probability, from Gup- ta's tables, is .0476. This appears to be quite acceptable accuracy for ordinary purposes. Fur- thermore, the approximation should be better for Zz , z{ 3. .. since the distributions of these variables cannot be as skewed as is the distribution of z*). For n > 20, it was also necessary to approximate 022"). This was done by converting EX), from Harter's table, to a cumulative percentage and using the asymp- totic variance formula for the percentage point of a distribution (see Wilks, 1962, p. 273). The values of these probabilities for n = 16, 20, 24, and 48, and for p = .40 and .50 have been computed and are presented in table D. Where possible, the exact values from Gupta's table are given for purposes of comparison. Not only do these values lead to a sign test, but they also lead to confidence intervals for the population mean (or median), Thus with 16 half-sample estimates, and with a p of .50, the largest and smallest estimates provide a 1 —(2x.06)= 88 percent confidence interval; similarly, the second smallest and second largest estimates provide a 1-(2x .1178) = 76 percent confidence interval. In order to apply the foregoing theory it is, of course, necessary to have a value of p with which to enter the tables, The data given in table C on the intraclass correlation coefficient sug- gest that a reasonable approach is to assume that p has the value suggested by the "common ele- ment'' argument, That is, if a set of k balanced half samples is used for (k — 1) strata, p would be assigned the value (k — 2)/2(k — 1). Finally, we observe that if one is willing to use the set of complementary half samples, then still another approach to this problem is sug- gested by some of the earlier arguments of this report. Suppose one wishes to compare the means of two variables X and Y. Let X, be the esti- mate of X made from the ith member of a set of k balanced half samples and Y,' be the estimate of ¥ made from the complementary half sample. The development on p. 15 suggests that the cor- relation between any two of the k differences Xx; - Y) will be very small and that it would not be unreasonable to treat them as k independent observations. Thus the ordinary sign-test tables could be applied to the signs of the k differences. 17 REFERENCES Brillinger, D. R.: The asymptotic behavior of Tukey’s general method of setting confidence limits (the Jacknife) when applied to maximum likelihood estimates. Review of In- ternational Statistical Institute 32: 202-206, 1964. Chapman, D. W.: An Approximate Test of Independence Based on Replications of a Complex Sample Survey Design. M.S. thesis, New York State School of Industrial and Labor Relations, Cornell University, 1966. Cochran," W. G.: Sampling Techniques, ed. 2. New York. John Wiley & Sons, Inc., 1963. Cramér, H.: Mathematical Methods of Statistics. Princeton. Princeton University Press, 1946. Deming, W. E.: Some Theory of Sampling. New York. John Wiley & Sons, Inc., 1950. Deming, W. E.: Sampling Design in Business Research. New York. John Wiley & Sons, Inc., 1960. Durbin, J.: A note on the application of Quenouille’s method of bias reduction tothe estimation of ratios. Biomet- © rika 46: 447-480, 1959. Durbin, J.: Design of multi-stage surveys for the estima- tion of sampling errors. Applied Statistics 16: 152-164, 1967. Gumbel, E. J.: Statistics of Extremes. New York. Colum- bia University Press, 1958. Gupta, S. S.: Probability integrals of multivariate normal and multivariate ¢. Ann.math.Statist. 34: 792-8928. 1963. Gurney, M.: The Variance of the Replication Method for Estimating Variances for the CPS Sample Design. Unpublished memorandum, U.S. Bureau of the Census, 1962. Harter, H. L.: Expected Values of Normal Order Statistics. Aeronautical Research Lab. Technical Report 60-292. Wright- Patterson Air Force Base, Ohio, 1960. Hartley, H. O.: Analytic Studies of Survey Data. Instituto di Statistica, Rome, Volume in onora di Corrado Gini. Ames, Iowa. Statistical Laboratory, Iowa State University of Science and Technology. Reprint Series 63. 1959. Kish, L.: Confidence intervals for clustered samples. American Sociological Review 22(2): 154-165, Apr. 1957. Kish, L.: Survey Sampling. New York. John Wiley & Sons, Inc., 1965. 18 Kish, L., Namboodiri, N. K., and Pillai, P. K.: The ratio bias in surveys. J.Am.Statist.A. 56(300): 863-876, Dec. 1962. Lauh, E. and Williams, W. H.: Some small sample results for the variance of a ratio. Proceedings of the Social Statis- tics Section of the American Statistical Association, 1963. pp. 273-283. National Center for Health Statistics: Cycle I of the Health Examination Survey, sample and response, United States, 1960-62. Vital and Health Statistics. PHS Pub. No. 1000- Series 11-No. 1. Public Health Service. Washington. U.S. Government Printing Office, Apr. 1964. National Center for Health Statistics: Replication, an ap- proach to the analysis of data from complex surveys. Vital and Health Statistics. PHS Pub. No. 1000-Series 2-No. 14. Public Health Service. Washington. U.S. Government Printing Office, Apr. 1966. Owen, D. B., and Steck, G. P.: Moments of order statistics from the equicorrelated multivariate normal distribution. Ann.math Statist. 33: 1286-1291, 1962. Plackett, R. L.: and Burman, J. P.: The design of optimum multifactorial experiments. Biometrika 33: 305-825, 1943-46. Quenouille, M. H.: Notes on bias in estimation. Biometrika 43: 353-360, 1956. Rao, J. N. K.: A note on estimation of ratios by Que- nouille’s method. Biometrika 52: 647-649, 1965. Rao, J. N. K.: and Webster, J. T.: On Two Methods of Bias Reduction in the Estimation of Ratios. Unpublished pa- per, 1965. Teichroew, D.: Tables of expected values of order statis- tics for samples of size twenty and less from the normal dis- tribution. Ann.math.Statist. 27: 410-426, 1956. Tin, M.: Comparison of some ratio estimators. J.Am.Sta- tist.A. 60(309): 294-307, 1965. U.S. Bureauof the Census: The Current Population Sur- vey, A Report on Methodology. Technical Paper No. 7. Wash- ington. U.S. Government Printing Office, 1963. Walsh, J. E.: Concerning the effect of intraclass correla- tion on certain significance tests. Ann.math.Statist. 18(1): 88-96, 1947. Wilks, S. S.: Mathematical Statistics. New York. John Wiley & Sons, Inc., 1962. O00 Table 1s DETAILED TABLES Estimates of mean body characteristics for population of U.S. adult males based on Health Examination Survey data------==---=====--=-=---==soososssoonsssomosnonnns Estimatesof regression coefficients and the multiple correlation coefficient for population of U.S. adult males based on Health Examination Survey data---------- Estimated sampling variability of mean body characteristics for population of U.S. adult males basedon Health Examination Survey data, and estimated intraclass correlation coefficients among pseudoreplicate samples-----mmmmm==mmmee————————— Estimated sampling variability of regression coefficients and the multiple cor- relation coefficient for population of U.S. adult males based on Health Examina- tion Survey data,and estimated intraclass correlation coefficients among pseudo- replicate Sample§=-=-=-==-=========-------sSSSSoSoSSoSSoSSoSmSTooTIoESTmTETITTIT Page 20 21 22 23 Table 1. Estimates of mean body characteristics for population of U.S. adult males based on Health Examination Survey data Variable R F 7 7* FoR |e hy Sod (1) (2) (3) D (5) (6) (7) Lom mmm 39.76991 | 39.76854 | 39.76836 | 39.76845 | -.00146 -.020 2 mm 30.82899 | 30.82636 | 30.82650 | 30.82643 | -.00256 -.025 3mm mmm 99.57059 | 99.56900 | 99.56929 | 99.56914 | -.00145 -.007 fmm meme eee am 89.00548 | 89.00882 | 89.00932 | 89.00907 .00359 .016 5mm mmm mm eee 1.26982 | 1.26907 | 1.26896 | 1.26902 | -.00080 -.026 mmm mmm eee 1.50193 | 1.50136| 1.50143| 1.50140 | -.00053 -.028 Jom mm me meee 86.63670 | 86.63346 | 86.63346 | 86.63346 | -.00324 -.026 8mm mmm mmm 90.60693 | 90.60443 | 90.60411 | 90.60427 | -.00266 -.022 9mm mmm 54.32630 | 54.32346 | 54.32339 | 54.32342 | -.00288 -.028 10mm mmm mmm meee 44.02929 | 44.02836 | 44.02843 | 44.02840 | -.00089 -.013 Llmmm mmm meee 14.51358 | 14.51357 | 14.51361 | 14.51359 .00001 .000 L2m mm mmm eee 59.22556 | 59.22364 | 59.22371| 59.22368 | -.00188 -.022 LB tees ms im i i i 49.45498 | 49.45439 | 49.45421 | 49.45430 | -.00068 -.007 Lhe mmm mee eee 35.62331 --- --- --- --- --- L5mmm mmm me meee 42.22215| 42.22332| 42.22350 | 42.22341 .00126 .012 16mm mmm me meee 24.26486 | 24.26357 | 24.26364 | 24.26360 | -.00126 -.012 Column 1 Variable 1 Biacromial diameter 9. Knee height 2 Right arm girth 10. Popliteal height 3 Chest girth 11. Thigh clearance height 4. Waist girth 12. Buttock-knee length 5 Right arm skinfold 13. Buttock-popliteal length 6 Infrascapular skinfold 14. Seat breadth 7 Sitting height (normal) 15. Elbow-elbow breadth 8 Sitting height (erect) 16. Elbow rest height A Column 2 Ris the combined ratio estimate based on the entire sample, including adjustments for nonresponse and poststratification. Column 3 The estimate R is obtained for each member of a set of 28 balanced half samples. These estimates are denoted by r,1=1,2,...,28,and Fis their average. Column 4 If r, is the estimate from a half sample, then rl is the corresponding estimate made from the complementary half sample. F' is the average of these 28 complementary estimates. Column 5 T7"is obtained by averaging columns 3 and 4. Column 6 Column 5-column 2. Column 7 The difference F*- R expressed as a fraction of the estimated standard error of R. VR) is computed as Gar E r, — r%/28. The actual values of the estimate of variance used in the computations are given in col- umn 7 of table 3. In the text discussion, these estimates are denoted by Vigus(™® 20 Table 2. Estimates of regression coefficients and the multiple correlation coefficient for population of U.S. adult males based on Health Examination Survey data Dependent Independent a _ _ —. _ x variable variable 8 b b' b (b* - a) JV ih (1) (2) (3) “) (5) (6) op (8) Age-------==m= -.02551 =, 02555 -.02555 -.02555 -.00004 -.016 1 Height--==----- .05260 .05228 .05227 .05228 -.00032 -.056 Weight---===-- «02999 02999 .02998 .02998 -.00001 -.011 Multiple R---- «55033 .55024 .55028 .55026 -.00007 -.006 Age----=--=mn- -.02724 -.02733 -.02734 -.02734 -.00010 -.038 2 Height-- -.12620 -.12689 -.12688 -.12688 -.00068 =o127 Weight .11140 11142 .11142 .11142 .00002 .016 Multiple R---- . 88439 .88481 .88479 . 88480 .00041 +103 Age--mrmmmm——— .03889 .03885 .03884 .03884 -.00005 -.007 3 Height-------- -.20126 -.20248 -.20247 -.20248 -.00122 -.075 Weight---=---= .28900 .28908 .28908 .28908 .00008 .026 Multiple R---- .90431 . 90470 .90468 .90469 .00038 +092 Age--mmmmmmm—— .20804 .20781 «20779 .20780 -.00024 -.028 4 Height-------- -.38187 -+ 38291 -.38293 -.38292 -.00105 -.068 Weight-====--= .38092 .38117 .38119 .38118 .00026 +075 i .91499 »913529 .91529 + 91529 .00030 .069 -.00279 -.00275 -.00276 -.00276 .00003 .037 5 -.02281 -.02283 -.02284 -.00284 -.00003 -.011 .01766 .01764 «01763 .01764 -.00002 -.037 .59132 .59274 .59282 «59278 .00146 L065 .00023 .00025 .00024 .00024 .00001 L011 6 -.03641 -.03650 -.03650 -.03650 -.00009 -.042 «02313 .02311 .02311 .02311 -.00002 -.035 .77307 «77313 .77312 .77312 .00005 .015 .00037 .00010 .00009 .00010 -.00027 -.084 7 .34698 . 34455 . 34455 . 34455 -.00243 232 .02407 .02423 .02422 .02422 .00015 .058 .73569 «73549 . 73544 .73546 -.00023 -.019 Age---==mmmmum -.01787 -.01824 -.01824 -.01824 -.00037 -.101 8 Height=---=--- .37403 .37154 + 37152 «37153 -.00250 -.289 Weight------== .01639 .01654 .01653 .01654 .00015 L067 Multiple R---- .78113 .78183 .78180 .78182 .00069 .079 Column 1 Dependent variable 1. Biacromial diameter 2. Right arm girth 3. Chest girth 4, Waist girth 5. Right arm skinfold 6. Infrascapular skinfold 7. Sitting height (normal) 8. Sitting height (erect) Column 2 Independent variables as identified. Note that the last entry in each set is the multiple corre- lation coefficient. Column 3 These are the regression coefficients and the multiple correlation coefficient as estimated from the entire sample, including adjustments for nonresponse and poststratification. Column 4 The estimate 8 is obtained for each member of a set of 28 balanced half samples. These estimates are denoted by b,i=1,2, ,28, and b is their average. Column 5 If b, is the estimate from a half sample, then b' is the corresponding estimate made from the comple- mentary half sample. b'is the average of these 28 complementary estimates. Column 6 Average of columns 4 and 5. Column 7 Column 6-column 3. — A A Column 8 The difference (b*— 8) expressed as a fraction of the estimated standard error of f. Vir is com- 28 puted as (Va) I (b —bD?Y/28. of the estimate of variance used in the these estimates are denoted by computations are given in column 8 of * cans P ). The actual values table 4. In the text discussion, 21 Table 3. Estimated sampling variability of mean body characteristics for population of U.S. adult males based on Health Examination Survey data, and estimated intraclass correlation coefficients among pseudo=- replicate samples Intraclass correlations Variable VR VR Vo® po | Vv an |v an ofa RAN BHS BHs © BHS Bus (7 cans 77 A 5 5! so" (1) (2) (3) (4) (5) (6) (7) (8) 9) (10) (11) .00143 .00551 .00551 .00564 .00564 .00556 3.888 | .48651 | .47392 .48021 .00342 .01034 .01033 .01086 .01085 .01057 3.090 | .49290| .46736 .48013 .02265 .03740 .03740 .03943 .03943 .03833 1.692 | .49413| ,46661 .48037 .04148 .04623 .04622 .05007 .05006 .04802 1.158 | .50094 | .45950 .48022 .00019 .00096 .00096 .00100 .00100 .00098 5.158 [ .49490| .46939 L48214 .00019 .00039 .00039 .00038 .00038 .00038 2.000 | .47368| .48684 .48026 . 00449 «OL591 .01590 .01599 .01598 .01592 3.546 | .48210( .47959 .48084 . 00444 .01480 .01479 .01528 .01527 .01501 3.380 | .48901 | .47235 .48068 .00282 .01012 .01011 .01083 .01082 .01045 3.706 | .49809| .46316 . 48062 .00241 .00509 .00509 .00503 .00503 .00505 2.096 | .47723] .48317 .48020 .00096 .00556 .00556 .00546 .00546 .00550 5.729 .47545] .48545 .48045 12-mmmmmmmm en .00281 .00713 .00713 .00727 «00727 .00718 2.555 | .48538| .47493 .48015 13-mmmmcmmemm .00310 .01061 .01061 .01111 L01111 .01084 3.496 | .49262| .46863 .48062 14=-mmmmmmm mm .00243 .01126 --- .01030 -——— .01075 4,424 i -—— -— 15-==mmmomeam .00707 .01109 .01109 .01156 .01156 .01129 1.596 | .49070| .46900 .47985 16--memeenen- .00292 .01187 .01187 .01184 .01184 .01184 4.054 | L48015]| .48142 .48078 Column 1 Variable 1. Biacromial diameter 9. Knee height 2. Right arm girth 10. Popliteal height 3. Chest girth 11. Thigh clearance height 4, Waist girth 12. Buttock-knee length 5. Right arm skinfold 13. Buttock-popliteal length 6. Infrascapular skinfold 14, Seat breadth 7. Sitting height (normal) 15. Elbow-elbow breadth 8. Sitting height (erect) 16. Elbow rest height A A Column 2 Vga (R) is computed as if the sample had been drawn as a simple random sample of approximately 3,000 adult males from the total U.S. population. The effects of clustering, stratification, and estima- tion have been ignored. Column 3 Vans (R) is the variance estimate that is ordinarily computed from a set of 28 balanced half samples, If the individual half-sample estimates are denoted by rp, i= 1,2, ...,28 then A A 28 A 2 ons) = Zr — RY’/28 Column 4 Vis (F) is similar to Vs (RO), except that deviations are taken about the mean of the r's, namely FT. A 28 Vous) = Tr, - DY/28 Column 5 Vows (R) corresponds to Vans (R), except that it is computed from the complementary half-sample esti- mates. Denoting these by z, i=1,2,. ..,28 A A 28 A Vous ® = Tr! — R128 Column 6 A 28 2 Vous = 2, — 7/28 Column 7 Vegus(F® is an estimate of variance based on the comparison of a half-sample estimate r, with its complementary estimate ri. This estimate is not available in the ordinary application of balanced half-sample replication for variance estimation. 28 Veons™ = M0 T(r, — 1/28 Column 8 The ratio of column 7 tocolumn 2 estimates the effect on sampling variability of sample design and estimation in comparison with simple random sampling. Column 9 4 is an estimate of the intraclass correlation among the r's. It was computed as 2 Veans FW — (28/2) Vp (TP) = . —~ 2 VegustT ") Column 10 p'is an estimate of the intraclass correlation among the rs, It was computed as pa =1 . 2 Vio oF" — (28/27) Vy (Fh BY pt —— ————— 2 Vegus Column 11 p*is the average of p and p', 22 Table 4. Estimated sampling variability of regression coefficients a lation of U.S. adult males basedon Health Examination Survey data, among pseudoreplicate samples nd the multiple correlation coefficient for popu- and estimated intraclass correlation coefficients Intraclass correlations Dependent | Independent a i \ . _ \ r= i = variable| variable VoanlB? Vans (8) Vagus (P) V0 Vous (b" Vegns P ®)/(3) . . " Pp Pp Pp (1) (2) (3) (4) (5) (6) (7) (8) 9) (10) (11) (12) FR 43432 | 59823 | .59807 |.66892 66876 63165 | 1.454 | .5091 | .4510| .4800 Hulghitn weet 263967 | 327107 | .32608T |.330457 | .32036% |.326037 |1.235| .4814| .4762| 4788 1 Folge ses 15524 | .08954 | .08954 |.08896 .08895 .08854 570 | .4756 | .4791| 4774 Multiple R- 1s748TT | 16116 Tt | 161151 [L15374 TF | .15374TT | L1seeoft | L995 | L667 | .4913| 4790 Agpnnnnnns 32395 | .65564 | .65483 |.63545 .63445 64223 | 1.982 | .4713 | .4878| .4796 Helghgeesne 10689 | .31062t | .30586t |.28883T | .28421% | .20365T [1.491 .4599 | .4982| .4790 2 Velghp=s= 11579 | .17278 | .17274 | .15962 .15958 16537 | 1.428 | .4584 | .4997| .4790 multiple R- 1538 | Lorsia™t | Lo1496Tt |.01605Tt | .o1679TT | .01575TF [1.024 | .s075 | .4a73| 4774 A mcmminl 17950" | 309027 |.300007 |.3s1247 | .3s122" |.373277 [2.078 .a458| .s121| 4790 Height----- c1001stt | 280121 | L28763Tt | .2as61 Tt | L26a15Tt | 2641177 | 2.420 | .4353 | .s5207| .4780 3 Velghitn=nss 64194 | 93129 | .93065 |.94957 .94893 93292 | 1.453 | .4828 | .4726| .4777 Multiple R- 010761 | Lo1762tt | 017471 | 017021" | .o16881T |.01710TT | 1.589 | .4703 | .4882| .4792 T f i f ¥ i Agpmnnnnnn= 29386 | 71384 |.71331 |.73237 .73174 72068 | 2.452 .4868 | .4735| .4802 Helght=--=-= 178601 | 2510171 | .25083"t |.23508"T | .23306 TT |.24006TF | 1.349 | .4603 | .4066| .u784 4 Weight----- a10504T | 11596 |L11534 | 129387 128657 | .12097T | 1.152| .5056 | .4486| .4771 Multiple R- 0085011 | Lo1s70™t | .o1s61tT |.019941T | .01085 1" |.o1014™F | 2.228 | .a959 | .4623| 4791 Hanan 05323 | .09286 |.09270 |.08917 .08908 09038 | 1.698 | .4682 | .4890| .4786 Height----- 03235T | Losao7t |.0s407f |.05507t | .05596T | .0sue1t | 1.688 | .4866 | .4687| .4776 5 Wolggiitn mnie 01903 | .04491 | .04487 |.04726 04717 04576 | 2.405| .4916 | .4655| .4786 Multiple R- 13705™ | Ls2015™ | sa713tt | .a0128'T | 489031" |.s036s't | 3.675 | .4573 | .4966| 4770 gpm mn mmm 03276 | 04977 |.04973 |.05018 .05017 04982 | 1.521| .4824 | .4779| .4802 HG LGHEn mmm o1901’ | Losers’ |.04e06" |.0s438T | .04430" | .0us03' |2.262| .4696 | .4899| .4798 6 Welght==r=r 01171 | .03260 |.03256 |.03196 .03192 03216 | 2.746 | .4751 | .4854| .4802 waleiple Re 052661 | Lor3satT | .o13satt |.o1a16’T | .o1a16't | .013007t | 2.650 | .4837 | .4718| 4778 ALR mmnts 089671 | .11s0sT |.11a32t [L103337 | .ot2sst |.1077a" [1.202] .4s98 | .s065| 4782 Heights 05450 TH L11689TT | 1100071 |. 1157771 | .100877T |. 1000111 | 2.017 | 4764 | .4817| .47% 7 Wolghigeesnb 32052 | .73399 |.73143 |.71576 .71351 71920 | 2.264 | 4727 | 4856 | .4792 Multiple R- 06320 Tt] 1208017 |. 12976 TH |. 1303311 | .13027%1 |.13325%7 | 2.106 | .4970 | .a601| .4786 Agpnnnnanns 7538” | Lusso’ |.13713t [L13so0t | .13aset |.13s00 |1.791| 4733 | .4820| 4776 Height---=-- ass12t | 82611 T | .76411t | .800677 737071 | L7se12T | 1.629 4690 | .4874 4782 8 Welights=-=== 26943 | .48075 |.47850 |.47475 47279 47351 | 1.757 .4760 | .4823| .4792 Multiple R- ouo2s VT] 0733877 | 07280 t |.07896 TT |.07851TT |.07524"1 [1.528 | .a077 | .as90| 4784 See footnotes on next page. 23 Column Column tn dagger Column columns 3-8, the absence of a dagger means that the entry is to be multiplied by 10~5, the presence of a single signifies multiplication by 10-4, and the presence of a double dagger signifies multiplication by 10-3, 1 Dependent variable 1. Biacromial diameter 2, Right arm girth Chest girth Waist girth Right arm skinfold Infrascapular skinfold Sitting height (normal) Sitting height (erect) 2 Independent variables as identified. Note that the last entry in each set is the multiple correlation coefficient. A 3 Vean(B) is computed as if the sample had been drawn as a simple random sample of approximately 3,000 adult males from the total U.S. population. The effects of clustering, stratification, and estimation have been ignored. oN UVL DW Column 4 Vans (Bris the variance estimate that is ordinarily computed from a set of 28 balanced half samples. If the individual half-sample estimates are denoted by b. 1=1,2,...,28 then A A 28 A 2 Vous (8) = Ib - 8/28 A — A A - Column 5 Veus(P) is similar to V, (8) except that deviations are taken about the mean of the b's, namely b, BHS A 28 —2 Vans BD = 2 b — B) /28 A A Column 6 Veus (8) corresponds to Vans (A), except that it is computed from the complementary half-sample estimates. Denoting these by b|, si =1,2,. . 28 A, A 28 v A 2 8 Vous(8) = 2, - BY? A 28 | =12 Column 7 Vous 0 = I (bj — B28 Column 8 Vigus(h*) is an estimate of variance based on the comparison of a half-sample estimate b, with its complementary Column estimate b/. This estimate is not available in the ordinary application of balanced half-sample replication for variance estimation. 5 5° 1, 2 BH Vegus (B™ = (Va) ZB - bh%28 9 The ratio of column 8 to column 3 estimates the effect on sampling variability of sample design and esti- mation in comparison with simple random sampling. Column 10 pis an estimate of the intraclass correlation among the b's. It was computed as A _ Pl . se 2 Vos (BY — (28/2D Vins (B) 2 Vi g,s (B® Column 11 3' is an estimate of the intraclass correlation among the b's. It was computed as Column 12 j*is the average of p and 5'. 24 A Tous A Th eo 2 Vigus (B® — (28/27) Vans (BY 2 Vens PY —_— 00 77 U. S. GOVERNMENT PRINTING OFFICE: 1969 342045/34 Series 1. Series 2. Series 3. Series 4. Series 10. Series 11. Series 12. Series 13. Series 14. Series 20. Series 21. Series 22. OUTLINE OF REPORT SERIES FOR VITAL AND HEALTH STATISTICS Public Health Service Publication No. 1000 Programs and collection procedures.—Reports which describe the general programs of the National Center for Health Statistics and its offices and divisions, data collection methods used, 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, 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, based on data collected in a continuing national household interview survey. Data from the Health Examination Survey.—Data from direct examination, testing, and measure- ment of national samples of the 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 Institutional Population Surveys.—Statistics relating to the health characteristics of persons in institutions, and on medical, nursing, and personal care received, based on national samples of establishments providing these services and samples of the residents or patients. Data from the Hospital Discharge Survey.— Statistics relating to discharged patients in short-stay hospitals, based on a sample of patient records in a national sample of hospitals. 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 manpower occupations, hospitals, nursing homes, and outpatient and other inpatient facilities. Data on mortality.—Various statistics on mortality other than as included in annual or monthly reports —special analyses by cause of death, age, and other demographic variables, also geographic and time series analyses. Data on natality, marriage, and divorce. —Varidus statistics on natality, marriage, and divorce other than as included in annual or monthly reports—special analyses by demographic variables, also geographic and time series analyses, studies of fertility. Data from the National Natality and Mortality Surveys. —Statistics on characteristics of births and deaths not available from the vital records, based on sample surveys stemming from these records, including such topics as mortality by socioeconomic class, medical experience in the last year of life, characteristics of pregnancy, etc, For a listoftitles of reports published in these series, write to: Office of Information National Center for Health Statistics U.S. Public Health Service Washington, D.C. 20201 PUBLIC SITAR N12: PUBLICATION NC 1000 SERIES yi NO 31 VITALand HEALTH STATISTICS : DATA EVALUATION AND METHODS RESEARCH sonal a CENTER Series 2 ‘For HEALTH Number 3 ) NEE er NG A a OTERO RT] Population Change A Systems Analysis Summary S U.S. DEPARTMENT OF /(/ ) HEALTH, EDUCATION, AND WELFARE [i | HIRE CSIR a CE AL a Le IR =r 4 Health Services and Mental Health Administration Public Health Service Publication No. 1000-Series 2-No. 32 For sale by the Superintendent of Documents, U.S. Government Printing Office Washington, D.C., 20402 - Price 3() cents NATIONAL CENTER| Series 2 For HEALTH STATISTICS | Number 32 VITALand HEALTH STATISTICS DATA EVALUATION AND METHODS RESEARCH Methods for Measuring Population Change A Systems Analysis Summary An introductory study of varying requirements of data for the measurement of population change, and of the alternative methods which are available to obtain these data, using a systems analysis approach. Data requirements are grouped into four categories, and seven major statistical methods are explored and rated in terms of their usefulness in furnishing each type of data. U.S. DEPARTMENT OF HEALTH, EDUCATION, AND WELFARE Public Health Service Health Services and Mental Health Administration Washington, D. C. March 1969 NATIONAL CENTER FOR HEALTH STATISTICS THEODORE D. WOOLSEY, Director PHILIP S. LAWRENCE, Sc.D., Associate Director OSWALD K. SAGEN, PH.D.. Assistant Director for Health Statistics Development WALT R. SIMMONS, M.A., Assistant Director for Research and Scientific Development ALICE M. WATERHOUSE, M.D., Medical Consultant JAMES E. KELLY, D.D.S., Dental Advisor EDWARD E. MINTY, Executive Officer MARGERY R. CUNNINGHAM, Information Officer Public Health Service Publication No. 1000-Series 2-No. 32 Library of Congress Catalog Card Number 71-600267 5. CONTENTS . INtrodUCHION mmm mm === === mmm mmm mmm momen mmm mmm mee The Systems Analysis Approach --------=-==-==--om-mmmoomommmommoo-m- Matrix of Required Measurement Factors ---------===-=-=--=-===-=-=-= Data Content Requirements -----=========-===-=----oo-o-o------oo- Population Composition Requirements ------=-=======--======-=-=-=-= Geographic Area REQUITEINENLS =v mem mmm mmm mm mmm map wm mo mn Time Requirements -------=-======mm=--m=m==--om-—=— m= ———=—- The Armamentarium of Statistical Methods-----------=-==--==----=-=-- Conventional Population Census ------======---====-===-==---==-=--= Conventional Vital Registration System ----=------==---=------=-==== Continuous Population Register ------====---cmom-mmo-mmmommomoomo- Sample Registration System ----==-----==--m------omooommmmoooooo Sample Interview SUrvey =-------==----=-===--=---ooo--o--o--oosos Panel Longitudinal Studies---=---==-======-mo-mco-ommooommommmo ooo Demographic Measurement Laboratories -----=-=-==-----=-====--=-- Ratings of Measurement Methods ------=-=mm-=====mmmemmme ne —————— References =---=mm-mmmmmm mmm mmmmmm mm emmm mmm mmm m—mm———————eee Appendix. Detailed Notes on Ratings inTable A ---------no-mo--noomommoooo oN nn 1 Us O 00 NN NN With increasing concern about vapid growth of the population, in this country as well as in many less developed countries, and with the wide- spread adoption of family planning programs, theve is a growing need Jor methods of measuring population change. Considevable resources have beenand will be expended in the collection of data to measure popu- lation change. A systems analysis approach is described which, by exam- ining diffevent methods of measurement in relation to the varying types of data requived, wouldlay the foundation for decisions on where addi- tional theoretical methodological veseavch should be concentrated to improve the adequacy and precision of existing methods and to identify methodological gaps for which new methods are required. The four broad classes of population data requivements are summarized under four headings: (1) data contentrvequivements, (2) population com- position requivements, (3) geographic area vequivements, and (4) time requivements, The armamentarium of statistical methods available to meet these ve- quirements is then exploved. The statistical methods ave considered under seven majov headings: (1) the conventional population census, (2) the conventional vital registration system, (3) continuous population registers, (4) sample registration systems, (5) sample interview sur- veys, (6) panel longitudinal studies or repetitive cvoss-section surveys, and (7) a combination of methods in "demographic measurement lab- oratories," After a brief descriptionof each method, the methods are then examined, taking eachone separately, as to theiv suitability with regard to the four groups of data vequivements. Each method is vated for each of the data requivemenls ona judgmental scale of strong, medium, and weak. These evaluations ave summarized in table A (p. 10). METHODS FOR MEASURING POPULATION CHANGE A SYSTEMS ANALYSIS SUMMARY Forrest E. Linder, Ph.D., University of North Carolina® 1. INTRODUCTION The world is now in a period of extremely rapid population growth. This is a continuing process of growth that started in the far past with a slower rate and smaller absolute incre- ments but is now growing at a faster rate with rapidly increasing absolute increments. Examining official and unofficial estimates of past and future world population totals,b the following first order differences for succes- sive 50-year intervals can be seen: Between 1750-1800, increased 187,000,000 Between 1800-1850, increased 284,000,000 Between 1850-1900, increased 388,000,000 Between 1900-1950, increased 865,000,000 Between 1950-2000, will increase 3,615,000,000. AForrest E. Linder is Professor of Biostatistics at the School of Public Health of the University of North Carolina, Associate Director of the Carolina Population Center, and consultant to the National Center for Health Statistics inter- national statistical program. This study was supported en- tirely by PHS Research Grant No. HD03441 from the National Institute of Child Health and Human Development to the Uni- versity of North Carolina. bMany differing population estimates are available, but within narrow limits determined by different assumptions and methods of calculation, the same broad interpretation emerges. The above figures are derived from the U.N. World Population Prospects 1 and Durand 2. The impact of these rapid increases does not fall equally on every country of the world. The increases are generally proportionally larger in the less developed countries of the world and many of these have adopted national policies and instituted specific programs designed to slow the rate of national population growth. It is estimated that more than half the people of the less developed nations now live in more than 25 countries with policies and programs to reduce birth rates by family planning. 3 The population problem is not limited to countries in the early stages of economic devel- opment. It is true that for many Americans the population question is generally viewed as a world problem, and one removed from day-to-day living in the United States. But the United States also has a domestic population problem although it is of a different type than the global problem. On a global basis and in the less developed countries, there are immediate warnings of ex- tensive starvation with continuing social and political unrest. In the United States there is no foreseeable danger of wide-spread malnutrition due to an overall shortage of food. On a global basis there can be valid concern in many areas that the population is rapidly approaching a magnitude where the land-space is inadequate to permit an acceptable mode of life considering the probability of slow economic and social development. The United States, at least in the near future, does not face this desperate space problem. However, the American people are used to an extremely high level of living in terms of open space and the domestic population pressure may therefore be rapidly approaching unacceptability. There is considerable doubt that Americans will be willing to gracefully accept drastic changes in their accustomed mode of living, and yet U.S. population growth is forcing such changes. These considerations, together with the fact that in the United States the deleterious effects of population growth are concentrated in the less affluent sectors of the population, have led to a positive U.S. population policy and programs to make family planning information available to all who desire it, *5 Appropriations specifically earmarked for population programs by the U.S. Congress for research, domestic action pro- grams, and for foreign assistance have rapidly increased for the fiscal years 1967, 1968, and 1969. It is easy to recognize that a population problem does exist. The crudest type of existing statistical data make this evident, But to measure more subtle population changes; to evaluate the social, economic, and cultural factors that facili- tate population change or create barriers to it, requires more sophisticated measurement instru- ments. Programs for the modulation of popula- tion growth will be extremely costly, will extend over long periods of time, and may produce re- sults that will be obscure in the initial phases. Many false starts will be made, many programs will fail to penetrate to the population sectors most concerned, and some approaches may have a negative rather than the expected positive effect. The timely evaluation of these programs will require effective, sensitive, and relatively inexpensive measurement tools capable not only of giving an early indication of changes in vital rates but of revealing significant factors related to such changes. The statistician concerned with such measurement problems has at his disposal a considerable array of measurement methods or tools, but each of these methods has certain intrinsic limitations as well as varying techni- cal and administrative difficulties or defects. A review of these tools, first in relation to their suitability to meet the broad needs for population change measurement instruments, and then, in more depth, with regard to the special technical power of each, will be of value in select- ing operating methods to be used in the field as well as in identifying basic and operations re- search which may be needed to improve the existing methodology. 2. THE SYSTEMS ANALYSIS APPROACH® An organized analysis of the attributes of statistical methods in relation to the measure- ment of population change is justified because of the extensive resources that are being and will continue to be put into data-collection ac- tivities. It cannot be assumed that all of these methods will be successful and there is sub- stantial evidence, using the criteria of the amount of valid data produced on changing vital rates, that there is substantial waste effort. The conventional census procedure produces no direct data on change of vital rates, and the validity of survey methods at their present level of technical devel- opment has serious defects.d In spite of such defects the needs for data are so great that survey activity for collecting new data on fertility topics will continue to increase. A committee of the International Union for the Scientific Study of Population has published an extensive list of recommended topics for fertility studies.14 The Population Council has published a valuable collection of questionnaires on KAP (knowledge, attitude, and practice) surveys!’ andis preparing a standard operating manual. The U.S. Bureau of the Census is recommending to countries a comprehensive package of survey proposals including suggested methods for measuring vital “The definitions and conceptof “systems analysis’’ differ greatly but generally ‘‘the systems approach involves identi- fying a problem, defining the objectives which mustbe achiev- ed to solve it, considering alternative methods for meeting these objectives, and choosing the mostattractive alternative by rigid cost-effectiveness analysis, by intuition and judg- ment, or by something in between,’’ 6,7 dSom and others have written on nonsampling errors and biases in retrospective demographic inquiries. 8-13 The Na- tional Center for Health Statistics has published a number of studies on survey response error in Vital and Health Statistics, Series 2. rates. 16 There is a substantial possibility that these types of recommendations will encourage expanded data-collection activities without a criti- cal appraisal of the validity of the data collected. The likelihood of waste effort and invalid results are particularly great in reference to ad hoc fertility surveys. The conventional census and vital statistics systems have deficiencies but at least these methods have the benefit of over 100 years of international study and the cooperative formulation of recommendations, so that the sub- ject coverage has been reduced to the most manageable and important topics and the nature of the deficiencies is well known.® Although ex- perience is now being rapidly accumulated, this "shakedown" process has barely started for fer- tility surveys. A systems analysis of existing methods in relation to present requirements is only an initial step in developing a theory of measurement of population change. And even a detailed systems analysis cannot be definitive since a systems analysis is an iterative process leading to dif- ferent conclusions as requirements and methods change and as more precise information on the short-comings and strengths of the different methods is known, The first step in a systems analysis of popu- lation change measurement methods would be an examination of the broad requirements or uses that any system of measurement methods must satisfy. These are more complex and inter- related than may seem at first glance and it is certain that no one statistical procedure will meet all of the requirements. The next step is to identify and clearly define the armamentarium of statistical tools or methods that can be used to deal with the measurement of population change. An examination of the attributes of the various methods in relation to the broad re- quirements would then lay the foundation for deci- sions on where additional theoretical methodologi- cal research should be concentrated to improve the adequacy and precision of existing methods and to identify methodological gaps for which new methods are required. €For the latest formulation of these international recom- mendations see United Nations Recommendations. 17,18 In examining the extent to which statistical methods satisfy the stated requirements it will be difficult to separate characteristics of each method which are to some degree intrinsic to the method from characteristics which are merely a consequence of the available resources or of good or bad administrative or operational practice. The different methods do require different types of field organization, different types of specialized personnel, and different levels of financial expenditure. If these factors are sub- stantially changed, then a method as defined here becomes almost a different method. It is conceivable that almost any deficiency of any method could be reduced or eliminated by an adequate infusion of operating resources and by superior administrative conditions. Inversely, of course, even the best technical methods can be vitiated by insufficient resources or operational incompetence. In spite of this, con- sidering the pattern within which the different methods ave in fact usually used, the. methods have intrinsic limitations and advantages which may be considered in relation to different meas- urement requirements. This discussion is con- cerned primarily with these intrinsic limitations and advantages rather than with the character- istics most susceptible to administrative change. The broad analysis suggested abcve should then be followed by a detailed study of each defined method. These following studies should consider each method in detail, carefully iden- tifying the technical character of each method, the administrative requirements in terms of organization, personnel and budget, synthesizing practical experience that has accumulated, and revealing important technical problems which require new basic or operational research. It is the purpose of this paper to attempt only the initial broad analysis of requirements and methods. A detailed examination of the attributes of each statistical tool can then follow with the benefit of some overall reference frame- work within which each method can be oriented. For either the broad analysis or the detailed examination of an individual method it is impor- tant to have a practical judgment of how much precision is really required to serve the pur- poses for which the data are collected. Per- fect data are seldom if ever necessary. Even approximate data may be sufficient to indicate a general course of action and rough data may be sufficient to permit an administrative judgment as to whether that course of action needs to be changed. With regard to the soundness of their statistical bases, many public health programs have proceeded on a philosophy of 'as if" and with regard to population action programs many different types of considerations other than those based on statistical fact may determine their direction, thrust, and velocity. Nevertheless, it can scarcely be denied that accurate data can be used to avoid tremendous waste in planning and evaluating administrative operations and in general, more precision makes possible more rational program administration. It is also necessary to define the boundaries within which the systems analysis is to be under- taken. The concept of a systems analysis is broad in scope and, without definite boundaries, the analysis could tend to range in a too amor- phous manner over the entire field of statistical method. There are numerous requirements and subrequirements for data and numerous minute features of the different methods. These are interrelated to form a complex too broad for a single treatment. There is, of course, a vast literature dealing with the mathematical or analytical adjustment, graduation, subtabulation and computational treatment of defective data, and with extracting estimates of vital rates from partial information. Some of these methods and new techniques are summarized in various ses- sions of professional conferences.!® The United Nations has also published a detailed manual on these methods. 2’ This article is not intended to deal with this class of method but rather to be directed to the problems of obtaining more ad- equate data in the first place. Even in the area of original data collection there are certain types of problems which will not be considered. In general, consideration will be limited to those methods most directly appli- cable to the measurement of the net effects of programs intended to achieve some reduction in the rate of population growth. Within these terms of reference there are some statistical problems and methods which are not studied. Some of these are methods relating to administrative evaluation of operating programs, the collection of service statistics,2! and the use of microsimulation schemes to estimate vital rate changes ,2 3. MATRIX OF REQUIRED MEASUREMENT FACTORS The required factors or attributes that any measurement system should have depends upon the broad purpose that the system is designed to serve.23 Some statistical systems are de- signed to serve a number of different purposes, and the methods used evolve into a compromise procedure which is thought appropriate to balance its suitability to these various purposes. The intent here, however, is not to examine the major statistical data-collection methods from the point of view of their adequacy in meeting several general needs, but to examine in a more limited way their adequacy in meeting the specialized purposes as measuring tools to detect the end results of programs directed at the reduction of the rate of population growth—that is, the measurement of changes in vital rates (birth and death rates) or the measurement of the net results of such changes. The variety and pioneering character of the population modulation programs that are being initiated in various countries of the world is evidence that even within the limits of the pur- pose stated above, measurement methods must be designed that will satisfy a number of distinct requirements, It is obvious thatone single method or procedure will not suffice. In brief, the re- quirements that must be met can be grouped into four broad classes: (1) data content require- ments, (2) population composition requirements, (3) geographic area requirements, and (4) time requirements. The optimum satisfaction of only one of each of these requirements might result in the selection of a statistical method that would be entirely inadequate in relation to the other groups of requirements. But since these groups of needs are all interrelated and must all be satisfied to some extent, they constitute a matrix of requirements. In effect this matrix becomes a multidimensional mold which par- tially defines the desired measurement tools. It is clear that the matrix described for dis- cussion is not a complete matrix for evaluation purposes, since the choice of a measurement method to be used may depend upon factors not considered in detail here such as cost require- ments, needs for specialized personnel, specifi- cations for minimum sampling errors, and so forth. It is hoped that factors such as these will be considered in more detail in subsequent analy- sis of the methods. Data Content Requirements If the primary purpose under consideration here is the measurement of changes in vital rates, the main requirement with regard to data content is that the method include information on the number of births and deaths together with any additional information which is necessary to convert such gross numbers to rates. While this is the obvious major requirement, it is by no means sufficient, since the appraisal of the effects and causes of population control programs requires knowledge of many factors related to the manner and rapidity of population change as well as the range of variables which may affect, in a positive or negative way, the design and operation of the population programs themselves. The International Union for the Scientific Study of Population!# as well as other agencies has made recommendations for items for such re- lated variables. No doubt some of the many items suggested have little or marginal value and in time a more concise list of essential and tested items will emerge. However, there is no doubt that any com- pletely satisfactory statistical procedure mustbe capable of coping with the collection, processing, and analysis of a wide variety of substantive items. It does not follow, of course, that any single statistical tool must encompass the whole range of substantive topics, but the available armamentarium of tools must meet the needs for many different types of data. Population Composition Requirements The substantive items referred to above are in many cases of little value unless they can be ex- pressed as rates by reference to an appropriate population base. This obviously requires that the total population serve as a base but it is also es- sential that the various strata or elements that makeup the composition of the total population serve as bases. The population of any nation is not usually spread evenly over all population composi- tion elements and population action programs must be designed to differentially relate to these elements. For example, in some countries the population problem may be more or less acute among the different economic classes. In some instances, the more affluent economic classes may already have easy access to family planning information with the resultant birth rates re- flecting their desired family-size norms. In the United States this is the case and accordingly the thrust of the official programs is to bring this same knowledge to the less affluent sectors of the population. In other countries the population problem may be primarily the problem of the predominantly rural areas. For such reasons, the whole necessary analy- sis of the changes of population may depend upon a differential study of numerous sectors of the population—by social and economic variables, by education, by demographic variables such as age, by urban-rural sectors of the populations, and by various religious, ethnic, and nationality groups. The point is not only whether these data can be collected in relation to the data-content items discussed above but whether a population base for rate computation can be also compiled and classified according to these variables. Some sur- vey designs and other data-collection systems may produce the numerator information for rate calculation but fail to produce the denominator population base classified by the required cate- gories. Geographic Area Requirements The geographic area requirements for ade- quate measurement of population change areper- haps the most difficult to meet. For certain important purposes, such as national economic planning, a nationwide measure of change in vital rates is most essential, since these rates determine the gross rate of population growth and this in turn defines an important aspect of a national plan for economic and social progress. However, action population programs are not necessarily administered in a uniform way from a national center and, even if they were, the problems and results vary by geographic areas as well as by population composition factors. In large federated countries, program admin- istration may be a responsibility of state or provincial administrations with great variation in the program design, available resources, and efficiency of administration. This creates a demand for population growth data not only for the national totals, but for subnational geographic or political divisions such as states, cities, and minor civil divisions. Since family planning pro- grams are still in an experimental stage of development there may also be a need for precise data for large program developmental regions or areas and even for small experimental units established to evaluate different program designs. From a theoretical point of view it is con- ceivable that an ideal statistical method could be designed and successfully applied to the smallest geographic area required and that this method could then be expanded to larger and larger geographic units until the total national area was covered. However the resource and logistic demands of this approach for most if not all of the methods considered here would seem to rule it out as a practical approach for the near future. Time Requirements Family planning programs are difficult to organize and are extremely expensive interms of budget and manpower. Such programs are designed in various ways based on differing premises and differing strategies of program operation. To prevent enormous waste effort it is essential that some method of appraising the effective re- sults of such programs be used which can detect changes in population growth rates over rela- tively short time periods. For more classical types of demographic analysis or for historical demography such long-interval statistical pro- cedures such as the traditional census may be adequate. It was from the analysis of decennial census data that the world has been brought to an appreciation of the urgency and magnitude of the world population problem. However, for pro- gram evaluation a decade-interval is much too long as a statistical measure of population change. In addition to the needs for evaluation of popu- lation action programs, the study of short-term changes in vital rates, such as those related to economic fluctuations, war activities, and poor agricultural seasons, require at least a year-to- year measure of changes in vital rates, and for some purposes an even more current index. 4. THE ARMAMENTARIUM OF STATISTICAL METHODS It will be clear that to meet the specifications of the matrix of required measurement factors a variety of statistical methods may have to be employed. Broadly speaking the statistical meth- ods which are available as data-collection pro- cedures may be considered under seven major headings: (1) the conventional population census, (2) the conventional vital registration system, (3) continuous population registers, (4) sample registration system, (5) sample interview sur- veys, (6) panel longitudinal studies or repetitive cross-section surveys, and (7) a combination of methods which might be called "demographic measurement laboratories." The total statisti- cal program in any country directed at measuring population changes and carried on by official and unofficial data-collecting and research agen- cies will probably comprise some combination of all of these methods. Yet, a careful examination of each major method in relation to the matrix of requirements will permit the development of an eclectic compilation and research program of maximum utility. Among the seven major groups of statis- tical methods listed, the conventional population census and the conventional vital registration system have rather clear internationally accepted definitions. 17:18,2425 The other methods require some additional degree of definition and descrip- tion before they can be evaluated in terms of the general requirements that have been set up. Even for the census and the vital registration system it may be worthwhile to reemphasize some of their essential attributes in the same manner as for the other methods, so that the different methods as defined and discussed in this article are clearly differentiated. Conventional Population Census The concept of the conventional population census is generally well understood, but the term is sometimes used also to refer to a variety of data collection procedures which are not ''popula- tion censuses' in the more precise meaning of the term. The examination of the census in this paper refers to the method defined by the United Nations” in these terms: "A census of population may be defined as the total process of collecting, compiling and publishing demo- graphic, economic and social data pertaining at a specified time or times, to all persons in a country or delimited area." The United Nations source lists some of the essential features of a national census as: (1) must be officially sponsored or carried out by a national govern- ment, (2) must relate to a precisely defined area, (3) must be universal in the sense of covering all persons within the defined scope of the census, (4) must be a total enumeration referring to a well-defined point in time, (5) must be a collection of separate data for each individual and not merely some process of aggregating group totals, and (6) must be carried through to the compilation and publication of data by geographic areas and by basic demographic variables. Conventional Vital Registration System Some form of legal provision and permanent organization for -the continuous registration of vital events exists in practically every nation of the world. The United Nations Handbook of Vital Statistics Methods>* defines this process as follows: '. . . a vital statistics system can be defined as including the legal registration, statistical recording and reporting of the occur- rence of, and the collection, compilation, analy- sis, presentation, and distribution of statistics pertaining to 'vital events', which in turn include live births, deaths, fetal deaths, marriages, di- vorces, adoptions, legitimations, recognitions, annulments, and legal separations." The regis- tration method is defined as ". . . the continuous and permanent, compulsory recording of the occurrence and the characteristics of vital events primarily for their value as legal documents as provided by law and secondarily for their use- fulness as a source of statistics...." Obviously the vital statistics systems of most countries fall short of this comprehensive definition in one way or another, either by not covering all of the different vital events specified or by operational deficiencies in covering com- pletely the types of events specified by the national legal basis of the system. In this paper the primary concern is with the system as the source of statistics on births and deaths although the other vital events are also of interest as factors indirectly related to population growth. Continuous Population Register The continuous population register starts with a base of census-type information to which con- tinuous corrections and additions are made with an input of data derived from a vital registration system and other administrative reports. Because of the necessity to make change-of-residence corrections the record systems are typically maintained at a local level. The method of ex- tracting statistical information from a large number of local registers and consolidating it for national totals varies from system to system. Population register systems are theoretically attractive because the same system produces data on the number of births and deaths as well as the corresponding population base. However, the sys- tems are highly sophisticated in that they require very precise input data, a disciplined and literate population to report regularly changes in demo- graphic status, and accurate procedures for matching records and compiling data from them. The concept of a continuous population regis- ter is not uniquely defined. Many variations in the design and the purposes of the systems exist. 26 The best known systems are those in several north European countries and the Japanese Koseki sys- tem. In most instances the systems have not been established for the purpose of creating statistics but for other administrative objectives and in general the statistical output from the systems has been disappointing, although increasing atten- tion has been given to this statistical byproduct of the systems.’ Sample Registration System The conventional registration system fails to serve its purposes in many situations because the technical manpower and financial resources are inadequate to produce vital statistics at a useful level of quality. The concept of a sample vital statistics system is that available resources could be concentrated on the administration of vital registration in a limited group of geographic areas selected as an adequate sample of the whole country or some subnational major civil division and in this way produce an acceptably accurate estimate of birth and death rates for that politi- cal geographic unit. Earlier descriptions of the sample registration concept 28 visualized the sample registration units as reenforced units of the already existing national registration sys- tem so that the number of sample units could be gradually increased resulting ultimately in an improved nationwide conventional vital regis- tration system. Since a vital registration system does not in itself produce the population base figures necessary for the computation of vital rates, most experimental efforts to establish a sample vital registration system have included an overlay of sample interview surveys partly to estimate the population of the sample area and partly to yield an independent report of biiths and deaths based on retrospective responses of the interviewees. A rather wide variety of designs have been tried, some in which the registration element of the system is quite independent of the official registration organization, some in which elaborate matching between the individual records of the registration and survey is attempted, and some in which the method is essentially a sequence of surveys with less emphasis on the continuous registration process. Some of these variations in method are described in a number of papers.29-34 Sample Interview Survey In some respects the sample interview survey method resembles the census technique except that with the advance in the theory and practice of population sampling these surveys can be con- ducted on a smaller number of respondents —a fact which permits many changes in data collecting techniques. Sample surveys for demographic pur - poses may be conducted with sample units con- sisting of individuals, households, or defined geographic areas. The items of data collected may be items of current fact (e.g., present marital status) or retrospective items (e.g., total number of children ever born). Surveys can be designed to produce estimates of a wide variety of items and at the same time population esti- mates in the required detail to convert the gross estimates of the variables to demographic ratios or rates. Sample surveys have almost unlimited flexibility as to the scope of their subject matter and geographic coverage, but their wide potential and relative economy tends to encourage their use without due attention to the many technical prob- lems of survey design and administration and without a critical evaluation of the validity of the data collected for newer and untested topics. Nevertheless the power and the adaptability of the sample interview survey assure it a role of increasing prominence as a method of collecting data on factors related to population change. Panel Longitudinal Studies The concept of the panel longitudinal followup method of studying population change is simply that of selecting an initial panel of persons, according to criteria determined by the purposes of the study, and making periodic observations on this group by repeated interview surveys and the collection of other records so as to create a complete history of the changing facts which are pertinent to the study for each person inthe panel. Unfortunately the needs and requirements for this method of study have not been clearly analyzed and they are generally difficult from an admin- istrative as well as an analytical point of view,3? The term 'longitudinal' study is applied to a number of quite different study designs, and for some of these designs, even if data are collected longitudinally, the analysis is basically little more than a series of cross-section studies. Longitu- dinal studies are essential where a variable must in some sense be integrated over a time period for each individual or where the outcome of a particular act is being studied. While important for medical studies, longitudinal types of investi- gation would seem to have fewer advantages for general studies of population change. Studies of this type require perhaps more sophistication and data collecting precision than population register systems, but are more feasible because the num- ber of persons in a panel is usually quite small. Demographic Measurement Laboratories The concept of a demographic measurement laboratory is a relatively new concept for demo- graphic studies, but is a long established method for the study of health conditions of a population and for developing improved methods for measur - ing such attributes. The Eastern Health District of Baltimore established in the 1930's by Johns Hopkins School of Hygiene, the Hagerstown area administered by the U.S. Public Health Service, the Tecumseh area in Michigan designed by the Michigan School of Public Health, the Arsenal Health district in Pennsylvania established by the Pittsburgh School of Public Health, and the Cali- fornia Population Laboratory under the direction of the California State Health Department are varying examples of this general concept as applied to problems of population health and medi- cal care. These different study areas have pro- duced a great wealth of published data on health and health measurement methodology, but unfor- tunately there seems to be no general review article which examines the general concept as a statistical method. The concept as applied to population’ is simply that of defining a fixed geographic area and using all available continuous and ad hoc statistical procedures for collecting the maxi- mum amount of information for that area that pertains to the topics under study. Since many different types of data can be collected and all available methods can be used, the total pool of data which becomes available has much greater analytical potential than if the same data collection efforts were dispersed over different population areas, Further, since a greatdeal becomes known about the population of the area, it forms an ideal situation in which to test the validity of new measurement methods. The demographic measurement laboratory concept permits a variable design, but ideally the laboratory should include a specific geo- fAs stated above, the laboratory concept has not yet been developed in detail for the study of population or population measurement methods. However, the studies carried on in the Gandigram area of Madras State, India and intensive area studies such as those of Singur, West Bengal, India are ex- amples of early applications of this approach. graphic area, for which maps and a house numbering system are maintained, a periodic population census taken, an upgraded sample or complete vital registration system established, and a permanent central office where data from these continuing processes as well as from ad hoc surveys and studies can be collated, linked, and analyzed. The data obtained from a limited geo- graphic area of this type may have validity for that area, but the extent to which they can be given a broader generality depends upon the extent to which the laboratory population can be con- sidered to represent a broader group. 5. RATINGS OF MEASUREMENT METHODS Table A gives the results of an examination of each method defined above in terms of the degree to which the method seems to satisfy the broad measurement requirements with which this particular analysis is concerned. The ratings or evaluations are on a rough judgmental scale of strong, medium, and weak. It is recognized that these summary ratings are judgmental incharac- ter but they do depend on an examination of essential attributes of the method itself modified by knowledge of the experience with these methods as they have been actually applied and adminis- tered in various countries throughout the world. Because of the qualitative nature of the evaluations or ratings, it is expected that other investi- gators might give somewhat differing weights to various attributes of each method andwouldhave differing appraisals of the accumulated experience on their use. The ratings assigned by the author are supported to some extent by the detailed notes given in the appendix. For convenience in refer- ence the paragraph numbers in the notes corre- spond to the numbers of the entries in table A so that ready reference can be made from the table to the corresponding comment. The com- ments or notes given in the appendix for each rating in table A, by no means, constitute a com- plete justification or detense of the ratings, but they do indicate some of the factors and the line of thought that lead to the evaluation. It would be differences of opinion in the selection and evaluation of these factors that would result in different ratings. Table A. VARIOUS METHODS OF MEASURING CHANGE RATED ACCORDING TO THE DEGREE TO WHICH THEY SATISFY BROAD MEASUREMENT REQUIREMENTS [The numbers in each cell of the table correspond to paragraph numbers in the appendix, which is a justification of the ratings given] STATISTICAL METHODS FOR MEASURING POPULATION CHANGE REQUIRED MEASUREMENT FACTORS Subject Geographic | Population Content Area Composition Time Conventional population census========- 1. weak 2. strong 3. strong | 4. weak Conventional vital registration system- 5. weak 6. strong 7. weak 8. medium Continuous population register--------- 9. medium | 10. strong | 11, weak 12, strong Sample registration system=------=-==-- 13. medium | 14, medium | 15. weak 16. strong Sample interview system=-------cceccceaa 17. strong 18. medium | 19, strong |20., strong Panel longitudinal follow-up---=-------= 21. medium | 22, weak 23. weak 24, medium Demographic laboratory=-=-=-----c-cea-a 25. strong | 26. weak 27. strong |28., strong Such differences in ratings would not, how- ever, destroy the value of table A since even allowing for variation some observations can emerge and these will be a useful guide to the selection of methods that have the greatest po- tential of satisfying the general requirements for measuring current population change. It is im- mediately obvious from table A that no one method alone will meet all of the basic needs of a country for essential population data and the ratings in table A can serve as a guide in designing the overall measurement system, with each method playing a proportionate part so that the total effective result is maximized. This summary systems analysis of the broad characteristics of population measurement meth- ods is also intended to focus attention on the soft or weak features of each to assist in the focusing of future methodological research. For example, the conventional population census is the most fundamental and universally used method of collecting data on population and related charac- teristics. Yet the procedure is weak as regards the "time" requirement, There is typically along delay between the population enumeration and the date when the tabulated results become avail- able, and there is generally a 10-year interval between censuses. Yet it may be possible to im- prove on both of these features. Continued re- search on character-reading methods of data processing, computer tabulation, and the intro- duction of sampling methods at various stages of census processing may appreciably improve the rapidity with which census results are obtained. The increasing practice of taking quinquennial censuses will cut the intercensal interval in half, On the other hand, the weakness of the census method with regard to subject content pertinent to measuring population change is likely to be more intractable. As a major national statistical collection method the general census must serve many purposes and the cost of a total census is so large that there must necessarily be a tight restriction on the number of items included. It is hardly conceivable that a population census could ever give the subject coverage required for KAP studies or could cover any large number of the items recommended for fertility surveys by the International Union for the Scientific Study of Population. Each method can be examined carefully not only to determine its defects as the method is now usually used but to probe for features that can most feasibly be improved. A concentration of theoretical and applied research on such fea- tures may more rapidly develop the method so that it is more efficient in meeting the types of requirements described in this paper. However, the summary analysis presented here cannot serve fully as a basis for either designing the overall national measurement sys- tem or identifying completely the areas of needed methodological research. These objec- tives require a more detailed and comprehensive systems analysis of each method for which this summary analysis can ‘serve as a framework or background. Such a more detailed analysis for each method should go beyond consideration of the broad measurement factors considered here, and examine the available evidence as to the technical validity of the method as well as to survey more completely the administrative experience that has been obtained for each method. Such studies would then identify in detail the needed theoretical and operations research which could lead to the design of more ade- quate methods to meet the urgent needs outlined in the first part of this paper. REFERENCES United Nations: World population prospects. Population Studies No. }1. New York, 1966. 2Durand, J.D.: The modern expansion of the world’s popu- lation. Proceedings of the American Philosophical Society I11(3):136-159, June 1967. 3Notestein, F. W.: The population crisis, reasons for hope. Foreign Affairs Oct. 1967. p. 171. 4Corsa, L.: United States, Public policy and programs in family planning. Studies in Family Planning No. 27. Popula- tion Council, Mar. 1968. pp. 1-4. Cohen, W.J.: Freedom of choice. Studies in Family Plan- ning No. 23. Population Council, Oct. 1967. pp. 2-5. bsystems analysis, no panacea for nations domestic prob- lems. Science 158:1028-1029, Nov. 24, 1967. "Systems Analysis and Policy Planning, edited by E.S. Quade and W. I. Boucher. New York. American Elsevier Pub- lishing Company, Inc., 1968. 830m, R. K.: On recall lapse in demographic studies. In- ternational Population Conference. Vienna, 1959. pp. 50-61. Som, R. K., and others: Preliminary estimates of birth and death rates and of the rate of growth of population, 14th round, July 1958-July 1959. National Sample Survey Draft No. 36. Indian Statistical Institute. 1958-60. 1080, R. K.: Response bias in demographic enquiries. Proceedings of the World Population Conference 111:179-182, 1965. Hynited Nations Economic Commission for Africa: Tech- nical paper on non-sampling errors and biases in retrospective demographic enquiries, E/CN14/ CAS.4/VS/3. 128abagh, G., and Scott, C.: An evaluation of the use of retrospective questionnaires for obtaining vital data, the ex- perience of the Moroccan multi-purpose sample survey of 1961-1963. Proceedings of the World Population Conference I11:203, 1965. Horvitz, D. G.: Problems indesigning interview surveys to measure population growth. Proceedings of the Social Sta- tistics Section, American Statistical Association. 1966. pp. 245-249. Mnternational Union for Scientific Study of Population: Variables for Comparative Fertility Studies. June 1967. Iophe Population Council: Selected Questionnaires on Knowledge, Attitudes, and Practice of Family Planning. 2 vols. Aug. 1967. 16.8. Bureau of the Census: Atlantida, a case study in household sample surveys, unit 11, content and design of household surveys. Series ISPO-1. Washington, D.C., 1965. ynited Nations: Principles and recommendations for the 1970 population censuses. Statistical Papers, Series M, No. 44. United Nations, New York. 1967. 18ynited Nations: Recommendations for the improvement and standardization of vital statistics. E/CN.3/388/Add.1, Jan. 1968. Dgeyfitz, N.: Newdevelopments in measurement and anal- ysis of factors of population growth and structure. Proceed- ings of the World Population Conference 1:65-69, 1965. 12 20united Nations: Methods of estimating basic demo- graphic measures from incomplete data. Manual IV. Manuals on Methods of Estimating Population, ST/SOA/Series A/42, Sales No. 67.XIII. 2, 1967. 21rhe Population Council: 4 Handbook for Service Statis- tics in Family Planning Programs. New York. The Population Council, 1968. 2Horvit, D. G.; Giesbrcht, F.: and Lachenbruch, P. A.: Micro-simulation of vital events in a large population. Paper presented at meeting of Biometrics Society, Atlanta, 1967. Research Triangle Institute and University of North Carolina. 23 inder, F. E., and Simmons, W. R.: Objectives and re- quirements for measurement of vital rates. Proceeding of the Social Statistics Section, American Statistical Association, 1966. pp. 240-244. 24United Nations: Studies in methods. Handbook of Vital Statistics Methods, Series F, No. T, 1955. 25United Nations: Studies in methods. Handbook of Popu- lation Census Methods, Vol. 1, Series F, No. 5, Rev. 1, 1958. 26United Nations: Methodology and evaluation of continu- ous population jregisters. U.N. Statistical Commission, 12th Session, Dec E/CN.3/293, 1962. p. 41. 2THofsten, E. V.: Population registers and computers, new possibilities for the production of demographic data. Review of the International Statistical Institute 34(2):186-194, 1966. 28Hauser, P.M.: The use of sampling for vital records and vital statistics. Bulletin of the World Health Organization 11(1-2):5-24, 1954. 2cavanaugh, J. A.: Sample vital registration experiment. Proceedings of the International Population Conference 2:363- 371. New York, 1961. 30L auriat, P.: Field experience in estimating population growth. Demography 4(1):228-243, 1967. 31Ahmed, N., and Krotki, K. J.: Simultaneous estimations of population growth—the Pakistan experiment. Quarterly Jour- nal of the Institute of Development Economics III(1):37-65. Karachi, Pakistan. Spring, 1963. 32United Nations: Guanabara Demographic Pilot Survey, a Joint Project of the United Nations and the Government of Brazil. United Nations, New York. (Undated) 33Manceau, J. N.: Sample survey of the registration areas of deaths and fetal deaths in the Brazilian Northeast Region. Review of the International Statistical Institute 36(1):43-67, 1968. yells, H. B., and Agrawal, B. L.: Sample registration in India. Demography 4(1):347-387, 1967. 35Kodlin, D., and Thompson, D. J.: An appraisal of the longitudinal approach to studies in growth and development. Monographs of the Society for Research in Child Development, Vol. XXIII, Serial 67, No. 1, 1958. 36United Nations: Report of the 12th Session of the Sta- tistical Commission. Doc. E/CN.3/304, May 1962. 3TIndian Statistical Institute: The use of the National Sample Survey in the estimation of current birth and death rates in India. Proceedings of the International Population Conference 2:395-402. New York, 1961. APPENDIX DETAILED NOTES ON RATINGS IN TABLE A Table A. Various methods of measuring population change rated according to the degree to which they satisfy broad measurement requirements [The numbers in each cell of the table correspond to paragraph numbers in this appendix, which is a justification of the ratings given] Statistical methods for Required measurement factors measuring population change 2 3 Population Subject content | Geographic area composition Time Conventional population census------=- 1. weak 2. Strong 3. strong | 4. weak Conventional vital registration SyStem==-===============-=-----=----- 5. weak 6. strong 7. weak 8. medium Continuous population register------- 9. medium 10. strong 11. weak 12. strong Sample registration system----------- 13. ‘medium 14. medium 15. weak 16. strong Sample interview systeém---~==mmm==c-=- 17. strong 18. medium 19. strong |20. strong Panel longitudinal follow-up--------- 21. medium 22. weak 23. weak 24, medium Demographic laboratory--------------- 25. strong 26. weak 27. strong |28. strong 1. Census - content (weak): The subject-matter content of the traditional population census is limited by the large cost of collecting data on many variables for a total population. The fact that most censuses are taken as official governmental projects under penalties for nonresponse, tends also to restrict the census con- tent to those few items which the government is willing to consider so important as to justify a mandatory re- sponse. The census, as United Nations recommendations for the 1970 World Census! show, is typically designed as a multipurpose statistical instrument, providing most essential national data on a variety of subject topics. The content coverage in terms of the number of items included for any one topic (e.g., fertility) is therefore quite restricted. The newer census practices of covering the total population with only a few basic items and expanding the subject coverage, utilizing subsamples of the population, greatly increase the scope of the census, Censuses of this design might be con- sidered as "medium" with regard to coverage. 2. Census - area (strong): The area coverage of the census is its strongest characteristic. By definition the term 'census'' connotes a national census and this is universally true except insofar as a few censuses in special cases may omit by design some relatively minor sector of the national territory because of rea- sons such as inaccessibility. Since by definitiona 'cen- sus" covers 100 percent of the population? the volume of data is adequate to permit tabulation and detailed cross-tabulation of items for major and minor civil divisions. Census data is almost universally tabulated in some detail for states or provinces, subnational regions, individual large cities, counties, and some- times even for smaller geographic units, such as census tracts. If the subject-matter scope of a census is increased through the use of population subsamples, the validity of tabulating these items for smaller geographic areas is weakened. 3. Census - composition (strong): Since one of the primary purposes of the census is to determine the composition of the population this is one of the census’ strong attributes, Many of the census items relate to the economic, social, and educational status of the popu- lation, To the extent that the census also includes items of more formal demographic interest (age struc- ture, marital history, and fertility history) the census provides an extremely rich source of data for demo- graphic analysis. Again, as in the case of tabulation £Based on the U.N. definition of a census, the term ‘“‘sample cen- sus’ is really a contradiction of terms. 13 by areas, the large volume of census data permits tabulation and cross-tabulation of demographic vari- ables in great detail according to the population com- position groups, 4, Census - time (weak): As a statistical tool for measuring population change the census is weak in two ways regarding the time factor. As a major and expensive nationwide statistical effort the usual practice is to take a census at 10-year intervals, While a sequence of decennial censuses provides an excellent base for long-range studies of historical demography and perhaps the most complete base for analyzing changes over the past decades, the long interval be- tween censuses does not provide a sensitive method of measuring current population change which is required to evaluate population action programs, The decennial census interval is also too long for the effective use of census data for other national purposes and consequently more and more countries are establishing a tradition of quinquennial censuses. However, for measurement of current population change a S-year interval is also too long. The second time defect of the nationwide census arises from the bulk of data processing required with the result that in ordinary circumstances census tabu- lations, except for a few total figures, are typically not available for 2 to 3 years after the date of the enumera- tion, In many cases the publication of the census results extends over a longer time span, This means that the most recent available decennial census data are usually 3 to 13 years old. With an increasing pattern of quin- quennial censuses and the use of electronic computers, the conventional national census procedure is being greatly improved, but even so, the census is not an adequate statistical method from the standpoint’ of the time variable for measuring current population change, 5. Vital registration - content (weak): The vital registration system exists originally and primarily as an official method of establishing the legal fact of birth and death, The statistical uses of the system are sec- ondary byproducts. Accordingly many items on the registration document are for the purpose of establish- ing identity and family relationship and other legal characteristics of the vital event and have little or no value as analytical variables in studying population change. This priority of items, in combination with the necessity for administrative reasons to keep the total number of items limited, results in a registration document with a paucity of information relating to factors affecting population change. Those demographic items which are included are usually highly relevant to the analysis of population change and, of course, the fact of birth or death established by a registration system is the datum of greatest importance to this question, Nevertheless, the statistical content of the vital registration system is one of its weak aspects. 6. Vital registration -avea (strong): As withthe census, the area factor of the registration system is a strong point, A registration system, presumably cover- ing all of the population of a nation, can be the source of birth and death information for every civil subdivision even to the smallest village, This strong feature of the registration system--the ability to provide data on vital events for subnational geographic areas—is some- what dampened because the gross number of births and deaths for an area has little meaning unless it can be related in the form of rates to the populations from which the births and deaths arise (the population ex- posed to risk), Because many births or deaths occur in areas other than those in which the principals con- cerned usually reside, it is difficult in many cases to tabulate the data so that there is a correct relationship between the population exposed to risk and the vital events, This factor is negligible for data tabulated for national totals but becomes more and more serious as the geographic tabulation unit becomes smaller and smaller, It is, of course, particularly serious in the case of small geographic units containing medical facilities that attract births and deaths across boundary lines from adjacent geographic units. Most modern vital statistics systems attempt to correct this defect by a reallocation of nonresidents. 7. Vital wvegistvation - composition (weak): Na- tional vital registration systems are usually weak inso- far as information on the registration record which would permit compilation of the data according to the economic, social, and demographic composition of the population, The latest draft of United Nations recom- mendations!'® for the contents of the birth record in- clude only a few priority factors of this kind, Typically vital statistics are not available by income, occupation, education, literacy, and other important variables corresponding to population composition groups, 8. Vital registration - time (medium): Since the registration system operates continuously it does not suffer from the same time deficiencies as the periodic census, Also since the total volume of individual records for births and deaths to be processed in any 1 year would be only about 2.5 percent to 7.0 percent of the volume of census records the system potentially could produce annual reports on the number of vital events with an acceptable timelag intabulationand publication. Provisional totals are sometimes available ona monthly or quarterly basis with a timelag of only 3 to 4 months, With a periodic census as a base, current birth and death data permit the continuous calculation of a national population total, assuming international migration is negligible or, as is usually the case, its net effect is known, The census and the registration system, if bothare accurate and timely, thus provide a strong up-to-the- minute measure of overall population change for a country as a whole. This strong time factor largely disappears for the estimation of vital rates or net rates of population change for subnational areas since in most cases, internal migration is not negligible nor is its net magnitude known in post-census periods. Therefore, for other than national totals, the vital statistics system inherits the time deficiency of the periodic census. In certain countries there is another serious time-defect in the registration system, In these countries vital events are tabulated notaccording to the month or year in which they occurred but according to the month or year in which they were registered, Since substantial numbers of births may be registered years after the events took place, this method of compilation may conceal a significant trend in the rate, 9. Continuous population register - content (me- dium): The continuous national population register based on some form of a national census, continuously updated by an input of information from a registration system and periodically verified by a census check, would be expected to be more complete in terms of content than either of the two systems which support it. In general, these registers include clear information on the fact of a vital event, on the geographic move- ment of people from residence in one geographic civil division to another, and, in some cases, ona few census- type variables relating to population composition, If the registers are maintained on a family basis, many other items of demographic interest, such as marital history and fertility history, can be derived from the information on the record. Because of this linking of discrete items of information the population register can be richer in content than either of its supporting systems. In spite of such technical advantages it may be noted that the United Nations statistical commission has expressed a rather negative opinion of population registers as a statistical data collecting method. ?% 10. Continuous populationrvegister - area (strong): One of the original and primary purposes of the popu- lation register is to establish a population figure (or even a roster) for small geographic civil administra- tive units. For this reason the fact of change of resi- dence from one local jurisdiction toanother is currently noted. Hence data on vital events, together with figures on the corresponding base population is potentially available for cities and small geographic areas, Popu- lation registers are often stronger in providing data for local areas than in methods for collating suchdata, ex- cept for a few summary figures, into totals for more and more comprehensive subnational or national fig- ures, 11. Continuous population vegister - composition (weak): Since population registers are periodically re- vised or verified by some form of population census, they could have the same strength relative to data on population composition as the census itself. However, the current input information from the registration system is weak on this type of information and as a consequence the continuous register will rapidly de- teriorate with reference to its accuracy on these vari- ables as the original or revised census data become obsolete. Furthermore, the population register collates input data relating to the base census information, the current registration information, and reported infor- mation on migration, The necessity for economical and efficient procedures for recording these data tends to eliminate from the population registers variables which were not essential to their original purpose. As stated, these purposes usually have been more closely con- cerned with local administrative problems than com- piling useful national or subnational statistics. 12. Continuous population register - time (strong): The continuous population register has all of the ad- vantages of the vital registration system with regard to the time factor, and because the base population is continuously updated, the system overcomes the dis- advantages relating to time which are suffered by the periodic census. Therefore from the standpoint of measuring current population changes and current changes in vital rates, the population register has outstanding advantages particularly since this time advantage is accompanied by an equally strong area advantage. These strong points would seem to indicate that the population register method approaches an ideal system for measuring current population change. However, the method is highly sophisticated and serious defects in the system arise from a complex of technical and administration problems. The detailed considera- tion of these problems requires a separate analysis of the operational requirements of population register systems. 13. Sample vegistration system - content (me- diwm): As has been stated above in connection with the definition of the statistical methods considered here, the concept of the sample registration system is not yet precisely defined. If the sample registration system is, in effect, an integral part of the regular nationwide legal registration system then it tends to be limited in content to the same items that are included in the regular system. Some new experimental data-collection procedures which are sometimes referred toas sample registration schemes are more in the natureofa series of successive interview surveys of the same sample areas. To the extent that this is so, the procedures can have a wider content scope than the conventional regis- tration system, Because it is on a sample basis and therefore more economical than a complete system, there is no intrinsic reason why a sample registration system could not expand its content coverage to some extent even if the system is closely related to the reg- ular registration organization. 15 14. Sample registvation system - avea (medium): The sample registration system is appropriate either for collecting data for a small defined area, such as a minor civil administrative unit, or for larger geographic areas such as states or provinces or for the territorial expanse of a whole country. For example, the Indian sample registration scheme?! is planned for the whole country, but it is initially being put into operation in only some of the states and within some of these states it has also been operating on a more intensive basis than in others with increased sample sizes in selected regions, A sample registration system, whether it covers only a small local area or the total country, is of course based on data collected from a number of small geographic sample units. This exposes the system to some technical and administrative problems related to reallocation of data on a residence basis, Another prob- lem arises in the assumption that there is correspond- ence between the population in an area and the vital events recorded in the area, 15. Sample wvegistrvation system - composition (weak): The sample registration system suffers from the same limitations as the conventional registration system in the restriction of the number and scope of items on the registration document that permit an analysis of the information on vital events according to composition groups of the population, Unless established as an independent and duplicating system in addition to the regular system in the sample areas, the sample system will be constrained somewhat by the underlying legal orientation of the conventional system. 16. Sample wvegistration system - time (strong): Since under a registration system the data are collected currently and since under a sample system the total volume of data need not be large, the time aspect of the sample registration scheme is one of its strong factors. However, the gross number of vital events becomes meaningful as a measure of population change only when it is converted into rates by being related to a correct population base, Therefore a sample registration sys- tem, by itself, has some of the same time deficiencies as the conventional registration system, That is, it is most valuable at the time of, or for a short period after, the regular census. However, if the primary purpose of the sample registration system is to pro- vide a current measure of change in birth and death rates it need only cover a population of sufficient size to bring the sampling errors down to a size consonant with the magnitude of the rate changes it is desired to detect, Under this sampling scheme, it therefore be- comes administratively feasible to take ad hoc or reg- ular periodic censuses of the sample areas as frequently as required to establish a satisfactorily accurate popu- lation base, 16 17. Sample interview survey - content (strong): The feasible content coverage of the sample interview survey as a method of measuring population change and related factors is one of its strong aspects. Since it is on a sample basis, the volume of field and processing work may be held at a manageable level so that there is no intrinsic limit to the scope of items determined by work volume considerations, Usually there are no legal or policy considerations, as in the case of the census or the vital registration system, that restrict the scope of the items included. In this type of data-collection proc- ess it is so easy to add items that this may turn into somewhat of a disadvantage rather than an advantage. The Population Council has recently compiled a col- lection of fertility survey schedules!” that containa be- wildering array of topics. In an attempt to introduce some order and standardization into fertility survey practice the International Union for the Scientific Study of Population has prepared several recommended lists of topics but even the recommended core list contains over 80 items and subitems., The census and the vital statistics system may be too weak on item coverage but the items that are included in these official pro- cedures have had the advantage of many decades of appraisal, testing, and international recommendation.h This sifting process has barely begun for items in- cluded in fertility surveys. 18. Sample interview surveys - area (medium): With regard to area coverage, sample surveys may be considered to have medium or variable strength, The sample investigation can be focused on any defined small area that may be of interest in evaluating popu- lation action programs or, as in the case of the Indian National Sample Survey,37 the area coverage may be nationwide, While adjustable in this way, surveys in practice have one feature that usually limits their utility for area analysis, If designed for nationwide coverage, a sample survey cannot be designed to also yield estimates for states or provinces without sub- stantially increased costs, If designed to produce esti- mates for a single state or province, the survey cannot yield estimates for subprovincial units without large increased costs, Theoretically, of course, a nationwide survey could be designed to produce national, state, and local estimates, but in this case the operation would become so massive that it would approach census- like magnitude and would necessarily take on some of the census deficiencies with regard to content coverage, time required to produce results, and the gap between successive surveys as dictated by budget considera- tions. Work on international standards for mortality classification was initiated at the first International Statistical Congress in 1853. The firstinternational recommendations for a population census were adopt- ed by the International Statistical Institute in 1872. 19. Sample interview survey - composition (strong): A sample interview survey if properly de- signed can produce population estimates in the proper categories for which it is desired to compute vital rates and other population-based proportions. Since there is no rigid limit on adding content items it is practical to produce population estimates for the de- sired composition factors, This factor is therefore one of the strong features of sample interview surveys. 20. Sample interview survey - time (strong): The time factor is a strong aspect of the sample interview survey. Due tothe relatively small volume of documents, data processing can be completed without serious de- lays and aside from budgetary and administrative con- siderations sample surveys can be repeated at frequent intervals. Of course budgetary and administrative con- siderations are also the only barriers to a frequent repetition of a census, but inthe case of sample surveys these considerations are not so restrictive because of the much smaller cost in funds and other resources required for sample investigations. 21. Panel longitudinal - content (medium): The panel longitudinal followup study method potentially can cover a wide variety of topics or items, With a relatively small number of persons comprising the panel there is no strict barrier to the addition of a wide variety of topics to the scope of the data collected. This apparent freedom is somewhat restricted by the longitudinal character of the method which then results in a sequence of observations on the same items for the same person, thus introducing problems of reconciling inconsistencies and dealing analytically with a number of individual time sequence patterns. Because of these difficulties data collected as longitudinal series are often analyzed merely as successive cross-section studies, The analytical difficulties also make it im- practical to study the interrelationship of a great many items. The wide possible subject coverage of this method is more apparent than real, 22. Panel longitudinal - avea (weak): One of the fundamental problems with the panel followup method of study is the difficulty of tracing and identifying members of the panel that move from their original place of residence during the course of the period of the longitudinal study. A related difficulty is to identify the fact of death for panel members or to record the fact and reason for other disappearances from the roster under active study. These problems are greatly increased as the geographic scope of the panel study is extended. Longitudinal studies have many intrinsic operational limitations even if confined to a small geographic area where intensive inter- mittent observations of the panel are possible, For a study extending over a large area the longitudinal study is an impractical data-collecting method. 23. Panel longitudinal - composition (weak): This method of study is weak with regard to its potential for analyzing population change and related factors by various components of the population, The area limita- tions and the necessary limitation on the overall size of the panel is not likely to permita panel with adequate representation of many different population components in adequate size for analysis. The method however per- mits the collection of a wide range of items and with- in the above restrictions the data can be analyzed by population composition, A major difficulty arises, however, as the study proceeds over a longer period of time. Even if the panel at its initiation contains the adequate or desired elements for a study of variables by the composition of the population represented, the gradual erosion of the panel by death and other losses produces doubt on the validity or representativeness of any analysis by the initially established categories. 24, Panel longitudinal followup - time (medium): The necessity of periodic observation of or interviews with the panel members introduces a time delay in this method which does not exist, for example, inthe sample registration system where the recording of vital events is continuous, However, because the panel is likely to be small and located within a compact geographic area the interval between successive observations need not be long. The analysis of simple events, suchas the number of births, requires no great time for processing, but the analytical difficulties of interpreting a set of complex series means that the method isnot likely to yield quick and clearly interpretable measures of population change. 25. Demographic measurement labovatory - con- tent (strong): As defined, the demographic measure- ment laboratory is a contiguous area for whicha variety of methods are used on a continuing basis to determine the demographic characteristics of the population resi- dent in the laboratory. Accordingly the subject scope and content of the data collected can be extensive, The content coverage can include not only- typical census and vital statistics items but also many other vari- ables concerning factors related to population change. The danger may be that the laboratory concept en- courages the collection of a plethora of subject items many of which may be of tangential value to the main issues under study. 26. Demographic measurement laboratory - area (weak): A concept involving data collection by any method as an intensive process in a defined and limited area has the major weakness of an area limitation in the interpretation and application of its results. The laboratory concept, implying in effect a continuous intensive field investigation, is not feasible ona nation- wide basis. Conceptually it is possible to visualize a network of such laboratories distributed throughout a country to adequately represent the heterogeneous spread of a nation's population, It is entirely possible that this area disadvantage could be overcome inthis way and perhaps at a cost of more or less the same order 17 of magnitude as the annual cost of the decennial census program or of the vital registration system, However, cost and other operating data for an illustrative demo- graphic laboratory are not yet available to allow an evaluation of the practicality of such a national network of laboratories. Even if the laboratory idea were ex- tended to a national network, it would still be inadequate to provide data for other political subdivisions such as states, Experience and the intrinsic concept of demo- graphic laboratory indicate that the area factor is a weak aspect of the laboratory idea. 27. Demographic measurement labovatory - com- position (strong): Since the laboratory concept visu- alizes the frequent periodic census-——a current collec- tion of data on vital events and various ad hoc surveys-— the system has the potential of permitting detailed analysis of population change for various components of the population, The restrictions on these analytic possibilities do not arise from the feasibility of col- ¥¢ U. S. GOVERNMENT PRINTING OFFICE: 1969—342046/45 18 lecting information on the variables of interest, but rather from the fact that the area limitation of the laboratory itself may mean that the population resident in the laboratory does not include an adequate number of the various population component groups which are of interest, Nevertheless, if the laboratory site is selected with this matter in mind, its usefulness in permitting the study of population changes for different components of the population is a strong factor. 28. Demographic measurement labovatory - time (strong): The time factor is one of the strong aspects of this concept, A frequent census, with current col- lection of information on vital events, involving a relatively small volume of data, permits a rapid col- lection and analysis of current material with little intrinsic timelag, The time delay in noting population changes can be reduced as much as desired so that it is limited only by administrative adeptness in the sta- tistical operations, Oo 0 o—— Series 1. Series 2. Series 3. Series 4. Series 10. Series 11. Series 12. Series 13. Series 14. Series 20. Series 21. Series 22. OUTLINE OF REPORT SERIES FOR VITAL AND HEALTH STATISTICS Public Health Service Publication No. 1000 Programs and collection procedures.—Reports which describe the general programs of the National Center for Health Statistics and its offices and divisions, data collection methods used, 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, 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, based on data collected in a continuing national household interview survey. Data from the Health Examination Survey.—Data from direct examination, testing, and measure- ment of national samples of the 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 Institutional Population Surveys.—Statistics relating to the health characteristics of persons in institutions, and on medical, nursing, and personal care received, based on national samples of establishments providing these services and samples of the residents or patients. Data from the Hospital Discharge Survey.— Statistics relating to discharged patients in short-stay hospitals, based on a sample of patient records in a national sample of hospitals. 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 manpower occupations, hospitals, nursing homes, and outpatient and other inpatient facilities. Data on imortality.—Various statistics on mortality other than as included in annual or monthly reports——special analyses by cause of death, age, and other demographic variables, also geographic and time series analyses. Data on natality, marriage, and divorce. — Various statistics on natality, marriage, and divorce other than as included in annual or monthly reports—special analyses by demographic variables, also geographic and time series analyses, studies of fertility. Data from the National Natality and Mortality Surveys. —Statistics on characteristics of births and deaths not available from the vital records, based on sample surveys stemming from these records, including such topics as mortality by socioeconomic class, medical experience in the last year of life, characteristics of pregnancy. ete, For a listof titles of reports published in these series, write to: Office of Information National Center for Health Statistics U.S. Public Health Service Washington, D.C. 20201 U.C. BERKELEY LIBRARIES LT (021206016