Public Health Service Publication No. 1000-Series 2-No. 5 For sale by the Superintendent of Documents, U.S. Government Printing Office Washington, D.C., 20402 - Price 25 cents NATIONAL CENTER| Series 2 For HEALTH STATISTICS | Number § VITALand HEALTH STATISTICS DATA EVALUATION AND METHODS RESEARCH An Index of Health: Mathematical Models Mathematical models of the distribution of illness episodes, of illness duration, and of mortality in a population. An index of health reflecting both mortality and morbidity is proposed. Washington, D.C. May 1965 U.S. DEPARTMENT OF HEALTH, EDUCATION, AND WELFARE Public Health Service Anthony J. Celebrezze Luther L. Terry Secretary Surgeon General CA I OR PUBLIC HEALTH R NATIONAL CENTER FOR HEALTH STATISTICS FORREST E. LINDER, Pu. D., Director THEODORE D. WOOLSEY, Deputy Director 0. K. SAGEN, Pu. D., Assistant Director WALT R. SIMMONS, M.A., Statistical Advisor ALICE M. WATERHOUSE, M.D., Medical Advisor JAMES E. KELLY, D.D.S., Dental Advisor LOUIS R. STOLCIS, M.A, Executive Officer OFFICE OF HEALTH STATISTICS ANALYSIS Iwao M. Moriyama, Pu. D., Chief DIVISION OF VITAL STATISTICS RoBerT D. Grove, Pu. D., Chief DIVISION OF HEALTH INTERVIEW STATISTICS Puivte S. LAWRENCE, Sc. D., Chief DIVISION OF HEALTH RECORDS STATISTICS MoxrokE G. SIREN, Pu. D., Chief DIVISION OF HEALTH EXAMINATION STATISTICS ARTHUR J. McDowell, Chief DIVISION OF DATA PROCESSING SioNEY BINDER, Chief Public Health Service Publication No. 1000-Series 2-No. 5 Library of Congress Catalog Card Number 65-60058 RA oq Usa wo, 5-8 PUBLIC HEALTH LIBRARY PREFACE This study is presented as a contribution to the method- ology of measuring health status. Among the objectives of the National Center for Health Statistics is the development of new techniques in health measurement. Although it is generally conceded that mortality statistics no longer provide an adequate measure of the health status of a population, no generally accepted method of measuring health in terms of both mortality and morbidity has emerged. Dr. C. L. Chiang of the School of Public Health, University of Cali- fornia, was invited to develop mathematical models which might serve as the basis for a general index which reflects morbidity as well as mortality. The models which have been developed represent one of many possible approaches to the problem. It is hoped that the publication of his work will lead to more in- tensive investigation of both the conceptual and the mathemat- ical problems involved in constructing such an index. As pointed out by the author, further testing is needed to determine whether the models presented provide a good de- scription of observed data. 433 CONTENTS Page PrefaC@e-mmmm momo eee mmm meme meme eee iii I. IntroduCtion==-=-m==c cme m meee meme mmm mmm i II. Probability Distribution of the Number of Illnesses-----=--=---- 2 III. Probability Distribution of the Duration of Illness--=----===----- 5 IV. Time Lost Due to Death---=-mmcmmmmm ome 9 V. Index of Health-==- coco mmm emma e oo 10 VI. Remark— Adjusted Index of Health---=-==ccemmmomcmmmomnnmaea 12 References -==---- common 12 Detailed TableS-=====cec mmm meee meme 15 Fr AN INDEX OF HEALTH: MATHEMATICAL MODELS C. L. Chiang, Ph.D, School of Public Health, University of California, Berkeley I. INTRODUCTION The state of health of a nation is one of the most important aspects in the study of a human population; but the lack of quantitative measures to assess health has always been a problem in the field of public health and welfare activities. The purpose of this study is to suggest mathemat- ical models for describing the state of health of a well-defined population over a given period of time, such as a calendar year. The health of a population is in part a func- © tion of such demographic variables as age, sex, ~ and possibly race. People of different ages and sexes have different susceptibilities to diseases, and diseases may act differently upon them. To describe the health status adequately, the popula- tion should be divided into subpopulations accord- ing to these variables. For convenience of presen- tation, however, these subpopulations will be as- sumed homogeneous with respect to all demo- graphic variables except age. From the public viewpoint, a simple and comprehensive index of the current state of health is most desirable. Be- cause of the complexity of the problem, however, a satisfactory approach should begin with detailed investigations of the basic component variables. * The state of health is best measured by the fre- quency and duration of illness, by the severity of illness, and by the number of deaths. These components taken together give a comprehensive picture of health; separately, each describes an aspect of the state of health. To measure the frequency of illnesses, we need to know the number of illnesses occurring in a calendar year to each individual of a given age group and the distribution of the subpopulation with respect to this variable. A mathematical model will be developed in section II. Although the model is not specifically developed for a particular type of illness, the general line of approach applies equally well for any specific disease. The derived probability distribution characterizes the pattern of proneness and susceptibility of a subpopulation to disease; it also provides an easy means of cal- culating incidence and prevalence rates. The mean number of illnesses and the corresponding stand- ard deviation will serve to measure the average proneness and its variability for each subpopula- tion. Furthermore, all these measures can be used for comparing subpopulations or summarized for the entire nation. As a test the suggested model is fitted to actual data from a sickness survey. The severity of an illness varies with the dis- ease and the individual concerned. Itdoes not lend itself to quantitative measures except as it isre- lated to duration or to termination in death. In section III we present the derivation of a general model for the duration of illness, which again applies either for a particular disease or for all illnesses. Because of lack of data, no attempt is made to find a specific function; however, alter- native approaches are described in detail. Since death must be related to ill health in a population, a study of health is notcomplete with- out considering the mortality rate. Mortality is evaluated from the standpoint of health in sec- tion IV. While studies of the component variables give a more detailed picture of the state of health, development of a single measure summarizing the information for the entire population is also es- sential. Based on the ideas in the preceding sec- tions, a health index is derived in section V. An adjusted index is suggested in section VI. II. PROBABILITY DISTRIBUTION OF THE NUMBER OF ILLNESSES Consider the time interval of 1 year (0,1), and for each t,0 <¢<1, let the random variable N (t) be the number of illnesses that an indi- vidual has during the time interval (0,¢t), with N (0)=0. The purpose of this section is toderive the probability distribution of the random variable Nt), B(®)=Pr{N(t)=n | NO) = 0}. (1) This probability function is an idealization of the proportion of people in the population having n illnesses, for n=0,1,..., during the interval (0,t). When an explicit form for the probability function is derived and computed for each value of n, we have a mathematical representation of the state of health of the population in terms of the number of illnesses. An assumption underlying the probability dis- tribution (1) is that the probability of occurrence of an illness during the infinitesimal interval (t,t +h) equals \h+o(h), where x, is afunc- tion of time ¢ and o(h)is a negligible quantity when h tends to zero. In essence, this means that the probability of an illness occurring within an infin- itesimal time interval is a function of time and is independent of the number of previous illnesses. This assumption leads to a system of differential difference equations for BR, (t). Consider twocon- tiguous time intervals, (0,¢) and (¢, t+h). Exactly n illnesses can occur in the interval (0, t+h) in three mutually exclusive ways: (a) n illnesses will occur in (0, ¢) andnone in (¢, t +h) witha prob- ability A, (t) [1-X\;h =o (W)]; (b) n—1 illnesses will occur in (0,) and one in (¢,t+h) witha prob- To indicate explicitly that the models are developed for a sub- population, say age group x, a subscript x should be added in the appropriate places. For the sake of simplicity of presentation, however, such a subscript will not be used in this section or in section III. ability P,_,(t) [A A+o (A)];and (c) n-2 illnesses or less in (0,t) and two or more in (t, t+h), with a probability of o(h). Taking these possibili- ties together we have the formula: P(t+h)=PF@®)[1- A h—o(h)] (2) + Bo [Nh +o®]+ om). Transposing P,(t), dividing through by h, and taking the limit as h tends to zero, yield a sys- tem of differential difference equations. A pe) = =\P,® dt (3) = Po = =\Py®+NPi(®, n=12,... t The first equation has the solution t - Npdr Pt)=e ° 3 (4) the remaining equations are solved successively to give the probabilities _ [ra [vf ©) e ° ° P= sn=l2, , n! For a period of 1year i.e., for t =1 the random variable N has the distribution n=0,1, ... (0) Within a period of 1 year, the instantaneous probability \, h +o(h) of occurring illness need not be dependent upon time ¢, and A, may be assumed to be constant. Under this assumption, we have the ordinary Poisson distribution = Ayn e £3 =n}= eee ’ n=012 ... (7) The constant A in formula (7) signifies an individ- ual's susceptibility to diseases and, as such, is a measure of the degree of his health. In fact, A is the expected number of illnesses oc- curring to an individual during a periodof 1 year. The larger the value of x, the more illnesses the individual may be expected to have. The value of A varies from one individual to another. To describe mathematically the health status of a population, we shall study the probabil - ity distribution of Xx. The distribution of x will be denotedby g (N\) dN, the theoretical proportion of people having the specified value x. Since the sum of the proportions of individuals is unity, the function g satisfies the condition Sevdr= 1, where the integral extends over all possible values of A. The probability distribution of illnesses will be a weighted average of the probability func- tion (7) with the density function g (\) dx em- ployed as weights; that is, —- A n 8 P trent= [20 d0VdA , n=0.1, ... (8) n! Roughly, formula (8) may be interpreted as follows: The expected proportion of in- dividuals who will have n illnesses during the year is equal to the sum of the products of (a) the proportion of individuals having a specific value of », and (b) the probability that an individual with the specific value of \ will have n illnesses during the year, where the sum is taken over all possible values of x. Choice of function g(x)d A» is dependent upon the health condition of the particular group of people in question. It appears, however, that the following function may describe the distribution in general: a 2& -1 = BX Nd \—nrr—————— &(N) re ©) where the gamma function I'(a) is defined by co Ia) = Fh 0 l oY gy (10) The ranges of the constants for which (9) is defined are a 2 0 and g>0. The function g (A) starts at Xx =0, increases as \ increases at a rate of Lg) = gO [e-7 -8] , (11) and reaches a maximum of («=p ' ge” @~V - 12 gM) T@ (12) at a-1 A 3 (13) After reaching the maximum value, g(\) de- creases as \ increases and assumes a value of zero as \ tends to infinity. The expectation and variance of x may be di- rectly computed from (9): I Bg a—1 EO) = A A T (a) er ax- = (14 and 2 : a2 BE a—1 —@B\ 75 - [o- -) Te e dN. 0 a = 5 (15) 8 Thus the ratio « /B measures the average health of a population, and the reciprocal of « is the rel- ative variance, 1 oN a eB 1 a [En]? ’ ( 6) which is a measure of variation of health among individuals in the subpopulation relative to the mean health. Assuming (9) as the function underlying the distribution of the population with respect to health condition, we have from (8) the probability func- tion of the number of illnesses during the year: "Xx a _ e A B a—-1_—fBX P {N=n}= [EE ra e dx (17) - LR a+)" oa, This probability is the expected proportion of in- dividuals in the population having nillnesses dur- ing the year, taking into account the variability among individuals in the population as described by formula (9). The expected number of illnesses occurring to an individual in the subpopulation is given by E(N) = -- ' (18) and the variance by oye 2D ee : (19) Formula (17) represents a family of infinitely many probability distributions, depending upon the constants « and B. The health status of a sub- population may best be described as a member of the probability distribution family for which a and B assume particular values. In order to estimate these values, it is necessary toknow the observed frequency distribution of the number of illnesses occurring to the individuals of the sub- population from which the mean N and variance sk of the number of illnesses are computed. Substituting N and S§ for E (N) and od, re- spectively, in (18) and (19) and solving the result- ing equations for a and 8 give the estimates Ay 2 N Cw 20; and ~ N = Se 21 # (s3-N) (21) Using the estimated values a and Bin (17), we have [(n+a) AR PN-n}- SHS BT ae) (HE) n=0,1, ... (22) The probability (22) multiplied by the total num- ber of individuals in the subpopulation is the ex- pected number of individuals having n illnesses during the year, for n=0,1,2,. . . This ex- pected frequency distribution may be compared with the observed frequency distribution by means of the chi-square test to determine whether the model described by (17) is an adequate measure of the state of health. Material collected by the Canadian Sickness Survey, 1950-1951, is used for this purpose (see references 3 and 4). The data in the Survey were based on a sample of approximately 10,000 households? inflated to give the national figures as appeared inthe publi- cations. Thus the published figures are much . greater than the actual counts in the sample. Not knowing the exact number, we take 13,538 as the sample size and each thousand in the published data as a single count (the total popula- tion size is 13,538,000, see table 3). Since the actual sample size is probably larger than 10,000 the exact chi-square values in our test should be somewhat greater. Two indirect measures of illness were used— the number of doctors' calls and clinic visits and the number of complaint periods that an individual had during the year. For the number of doctors’ calls and clinic visits the model is 2See page 17 of Reference 3. fitted for six age groups-—under 15, 15-24, 25-44, 45-64, 65 and over, and all ages. The results for the first two age groups and for all ages are pre- sented in tables 1, 2, and 3, and figures 1, 2, and 3, respectively. In each of the fitsttwocases, the fit is quite good. For all ages, however, the chi- square value exceeds the critical value at the 1 percent level of significance. Data on the number of complaint periods were divided into only four age groups—under 15, 15-64, 65 and over, and all ages. Only the age group under 15 is well described by the present model as shown in table 4 and figure 4. Although neither of the underlying random variables is that of our model, the chi-square tests show promising prospects when the age intervals are not too large. Itis hoped that more appropriate material will be made available for further testing. lll. PROBABILITY DISTRIBUTION OF THE DURATION OF ILLNESS Let random variable T be the duration of an illness so that Pr(Tt) (25) the probability that an individual will be ill for a period longer than ¢, and the time in- terval (¢,¢t+ At). In order for anindividual tobe ill for a period longer than (¢t + At,) he must be ill for a period longer than ¢ and not recover within the period (t,t+ At). Accordingly, the corre- sponding probabilities have the following rela- tionship: Q(t+a)=Q() [1-v,at-o(at)] . (26) Subtracting Q(t) from both sides of (26), dividing (24) through by A ¢t, and letting At tend to zero, we have the differential equation 4 9W=-9 Mv . (27) The general solution of (27) is easily found to be t —- vp dr +c Qt) = e° ’ (28) where c is the constant of integration. Since QO)=Pr(T>0)=1 for t=0 the constant of inte- gration must be 0, and we have the desired so- lution t —- vedr Q(t)=¢ ) (29) and the probability t -— ve dr Pr (T (40) i= For application later, we also compute the second and the third moments about the origin, Ea = fetveriae- 2 v2 0 and (47) E(TY)- IG ve-vt de = Oo Vv 0 Using the estimated value 9 from either method, the expected relative frequency of ill- nesses with a truncated duration between, say ¢, and t, may be computed from t A / 2psg A ee et _e=P - 2 = 1—e for any interval (¢,, t,). Multiplying (48) by M gives the expected number of illnesses. The dis- crepancy between the observed and expected num- ber of illnesses for each interval (¢,, t,) may be evaluated by the chi-square test to determine the sufficiency of the model in formula (36). Equation (36) may be applied either to all diseases as a whole or separately to individual diseases. When all diseases are taken as a single group, the computations involved are quite simple. When individual diseases are studied separately, they can be summarized by the compound distri- bution £*(t) dt = Deft de (49) where x. is the known proportion of illnesses of the jth disease and f;*(t) is the corresponding probability density function as given in (36). In this case, the overall expected duration of illness is given by E(T"=2 T Bb _ S . (50) 1—eVi Since =; is known and v; can be estimated for each disease, the estimate of E (T*) can be com- puted from (50). Illnesses are often classified as acute or chronic. A mathematical model todescribe illness from this viewpoint is essential but difficult to formulate, because the demarcation line between acute and chronic illness is not always well de- fined,and the exact proportions of illnesses inthe two categories are never known. Inthe discussion to follow, we shall use the duration of the illness as acriterion of classification and make an attempt to solve the problem. According to our formulation, a general math- ematical model in this case may be represented by the probability density function fre) dt = (51) t t =f vi dr -f vor dr v,, e° vy e° We cog # Qed: | il, 1 ~f vy dt 0 1~e : l1-e When v= v, and v,=v, are assumed to be in- dependent of time t, the probability function be- comes (52) vie "i! vye—'2 — + Q-x) ——— | dt. 1-e=" 1-e—"2 fr(t)dt= |» Here = is an unknown proportion and may be in- terpreted as the probability that an illness will be acute with the severity signifiedby v,. A similar interpretation holds for (1 - x) ahd v,. Con- sequent to our arbitrary classification of ill- nesses as acute or chronic, a graphic represen- tation of the model will show a bimodal curve. The statistical problem is to fit sucha model to empirical data and to estimate the parameters rm, vi, and v, in the formula. In this case the simplest approach is to consider the complete duration of a sample of M illnesses and to use the method of moments, which allows the parameters to be estimated from the first three sample mo- ments u,,u,, and uj (see equations (45) and (47) ) 2, 1-5 u, ss —— 4 1" H 2% % 201-7) Of, ee Pee Ue TP (53) 6% 601-7) Ug™ = mp These equations may now be solved for DN and 9. Eliminating # from the first two equa- tions in (53) gives 2-2u, (+0) + u,M%; = 0 (54) and from all three equations gives 6u,— 3u,(W + + uM vy =0 (55) Solving (54) and (55) simultaneously, we have N+P, = Suu, - uy (56) ) 3u? - 2u,u, and 12u?2 — bu nw, = 1 2 (57) 3u?l = 2u,u, These can be computed from the sample mo- ments. Now let n+ = a, (58) and D0) = ay 4 (59) and formulate a quadratic equation and Y'- a Y+a,=0. (60) The two roots of (60) are 4} and ¥;. Substi- tuting these values into the first equation in (53), we obtain the estimate of =: » (=u) V. al (61) » _AN 1" "2 Using the estimated values of «x, v,, and v,, we can compute the expected relative frequency from t / £08) dt = (62) for each interval (t,, ¢t,). The expected number of illnesses of durations between ¢, and ¢,isob- tained by multiplying (62) by the total number of illnesses. IV. TIME LOST DUE TO DEATH In general, a high frequency of mortality indicates ill health within a population and, con- sequently, the death rate is one basic measure of the state of health. When the amount of illness in a calendar year is studied, not only the length of the illness before death must be considered, but also the period from the time of death to the end of the year. This period will be referred to as "the time lost due to death' and will be denoted by ¢. We want to determine the probability dis- tribution of ¢,and the expected value of ¢ among the deaths occurring during the year. The time lost due to death is determined by the time of death. Although deaths may be subject to seasonal variation, as an approximation we as- sume that they take place uniformly throughout the year. Consequently the random variable ¢£ also has a uniform distribution within the inter- val (0, 1), and its distribution function is given by 6 Ple - o NN. 2 a 8 o w o g g 5 < > © z » —. ~ ” o 8 8 I" a year. Figure |. Observed and expected number of persons under 15 years of age by number of doctors' calls or clinic visits in cli Figure 2. Observed and expected nu mber of persons 15-24 years of age by number of doctors' calls or a year. visits in 13 3 NONE NE e 3 a o = 5 © ° <+ » 0 8 2 S s ° © a z - «4 3 « © o & ° ou 5 4 2 Q " " OQ wo ° - « & a © wv = LN | sd z £ < ) < : @ g oa a 3 ~ n ao ° | | visits mber of doctors' calls or clinic Figure 3. Observed and expected number of persons of all ages by nu in a year. Figure 4. Observed and expected number of persons under |5 years of age by number of complaint periods in a year. 14 Table X. DETAILED TABLES Observed and expected number of persons under 15 years of age, by number of doc- tors' calls or clinic visits in a year-------------me-c-ucmcomcecce meee mem Observed and expected number of persons 15-24 years of age,by number of doctors’ calls or clinic visits in a year----=-=-----c--cooooomcmoomm mmm mmm mmm mmm mm Observed and expected number of persons of all ages, by number of doctors' calls or clinic visits in a year---=--------m-omommemomo mem m meme m om mm mmm mmm mmm Observed and expected number of persons under 15 years of age, by number of com- plaint periods in a year---=-====-=---------- o-oo mmem-oeooooooooososoooe Page 16 17 18 19 15 Table 1. Observed and expected number of persons under 15 years of age, by number of doctors' calls or clinic visits in a year Number of doctors' calls or clinic Observed Expected Difference (fa— FR)? visits, n fy F, fo — Fy F, Number of persons in thousands! TIGER] we i 0 re me 4,116 4,116 0 '3.899 0 call or visit--=---cemcccc mcmama 2,367 2,379 -12 0.060 1 call or visit=----cccccmmmaaaaaao 749 715 +34 1.617 2 calls or visitS--------cccceccca-oo 350 372 -22 1.301 3 calls or visits=-----=c-cc-ceooo-o- 222 221 +1 0.004 4 calls or visits------------coaeoa-- 136 140 -4 0.114 Subtotal, 5-9 calls or visits-- 239 242 -3 0.037 5 calls or visitS=------ccccmcmaaanao 95 91 +4 6 calls or visitS-------c--cooomaaaoo 64 61 +3 7 calls or visits-----c-c-ccomacaoaao 41 41 0 8 calls or visits------=c-cccocecanao 25 29 -4 9 calls or visitS===----ccceoeao ooo 14 20 -6 Subtotal, 10+ calls or visits-- 53 47 +6- 0.766 10 calls or visit§e=--=--c-cececacaaao 12 14 -2 11 calls or visits--------cccceaoooo- 11 10 +1 12 calls or visit§=----=--=coceeaao__- 9 7 +2 13 calls or visitS----------coceoo-noo 8 5 +3 14 calls or visitS-=----mc--cmecccaaoan 5 3 +2 15+ calls or visits------------c-oo-- 8 8 0 N=1.163 @ = 0.405 x?= 3.899 \ S?= 4.500 B = 0.348 d.f. = 4 Source: Observed f, were calculated from percent distributions shown in table 2=C, page 27 of Reference 4, and population totals shown in table 114, page 193 of Reference 3. To estimate the parameters involved in the model, subtotals shown were distributed for each n in the respective groups. by the number of calls or visits to obtain f, For justification in using thousand as a single count in computing the x?, see text on page 4. Table 2. Observed and expected number of persons 15-24 years of age, by number of doctors' calls or clinic visits in a year Number of doctors' calls or clinic Observed Expected | Difference ty — Fo) visits, n t F, £,~F, I Number of persons in thousands! Foals wnwmwrmen enn mmm sm 2,050 2,050 0 16.955 0 call or visit--------c-cecmmccaanna- 1,326 1,327 -1 0.001 1 call or visit=----cceecemmmceao-——- 248 254 -6 0.142 2 calls or visits--=----m-e-ecmenenn—- 141 136 +5 0.184 3 calls or visitS-----esecceeeeccaoo- 88 86 +2 0.047 4 calls or visitS-=--=---m---cceeo-—-- 68 59 +9 1.373 Subtotal, 5-9 calls or visits-- 113 133 -20 3.008 5 calls or visits---=--ecememmoeecoo—- 47 43 +4 6 calls or visits---=--c-ccmcoceaanm- 28 32 -4 7 calls or visitS-===e-=-----cco--oo- 17 24 -7 8 calls or visits--------seemmoocan-n 12 19 -7 9 calls or visitS---=-=-ee--c-co-coon 9 15 -6 Subtotal, 10+ calls or visits-- 66 55 +11 2.200 10 calls or visitS=====-c-------c-o-m- 8 11 -3 11 calls or visits--==----ec-ecec-ca-- 7 9 -2 12 calls or visitS=--=ecmecmmmccecan-—-- 7 7 0 13 calls or visits----=---ccceeo----- 6 6 0 14 calls or visits--=-=--c-ceeee-a--- 5 5 0 15 calls or visits---==---cceceeaaan-- 4 4 0 16 calls or visits----===--s-ee-ncao-- 4 3 +1 17 calls or vigligmm=re=mwnsenmummsum 3 3 0 18 calls or visitS=-==--=se-eec-ccc-o-- 3 2 +1 19 calls or visiLg-—mw——== 5.392 $= 1.102 d.f.= 8 Source: Table 31, page 122 of Reference 3. To estimate the parameters involved in the model, the subtotal shown was distributed by the number of complaint periods to obtain f, for each n in the group. lFor justification in using thousand as a single count in computing the x2, see text on page 4. 19 # U.S. GOVERNMENT PRINTING OFFICE : 1965 0—770-162 ETE TT hme. Asie Ce; REPORTS FROM THE NATIONAL CENTER FOR HEALTH STATISTICS Public Health Service Publication No. 1000 Series 1. Programs and collection procedures No. 1. Origin, Program, and Operation of the U.S. National Health Survey. 35 cents. No. 2. Health Survey Procedure: Concepts, Questionnaire Development, and Definitions in the Health Interview Survey. 45 cents. No. 3. Development and Maintenance of a National Inventory of Hospitals and Institutions. 25 cents. Series 2. Data evaluation and methods research No. 1. Comparison of Two-Vision Testing Devices. 30 cents. No. 2. Measurement of Personal Health Expenditures. 45 cents. No. 3. The One-Hour Glucose Tolerance Test. 30 cents. No. 4. Comparison of Two Methods of Constructing Abridged Life Tables. 15 cents. No. 5. An Index of Health: Mathematical Models. Series 3. Analytical studies No. 1. The Change in Mortality Trend in the United States. 35 cents. No. 2. Recent Mortality Trends in Chile. 30 cents. Series 4. Documents and committee reports No reports to date. Series 10. Data From the Health Interview Survey No. 1. Acute Conditions, Incidence and Associated Disability, United States, July 1961-June 1962. 40 cents. No. 2. Family Income in Relation to Selected Health Characteristics, United States. 40 cents. No. 3. Length of Convalescence After Surgery, United States, July 1960-June 1961. 35 cents. No. 4. Disability Days, United States, July 1961-June 1962. 40 cents. No. 5. Current Estimates From the Health Interview Survey, United States, July 1962-June 1963. 35 cents. No. 6. Impairments Due to Injury, by Class and Type of Accident, United States, July 1959-June 1961. 25 cents. No. 7. Disability Among Persons in the Labor Force, by Employment Status, United States, July 1961-June 1962. 40 cents. No. 8. Types of Injuries, Incidence and Associated Disability, United States, July 1957-June 1961. 35 cents. No. 9. Medical Care, Health Status, and Family Income, United States. 55 cents. No. 10. Acute Conditions, Incidence and Associated Disability, United States, July 1962-June 1963. 45 cents. No. 11. Health Insurance Coverage, United States, July 1962-June 1963. 35 cents. No. 12. Bed Disability Among the Chronically Limited, United States, July 1957-June 1961. 45 cents. No. 13. Current Estimates From the Health Interview Survey, United States, July 1963-June 1964. 40 cents. No. 14. Illness, Disability, and Hospitalization Among Veterans, United States, July 1957-June 1961. 35 cents. No. 15. Acute Conditions, Incidence and Associated Disability, United States, July 1963-June 1964. No. 16. Health Insurance, Type of Insuring Organization and Multiple Coverage, United States, July 1962-June 1963. No. 17. Chronic Conditions and Activity Limitations, United States, July 1961-June 1963. Series 11. Data From the Health Examination Survey No. 1. Cycle I of the Health Examination Survey: Sample and Response, United States, 1960-1962. 30 cents. No. 2. Glucose Tolerance of Adults, United States, 1960-1962. 25 cents. No. 3. Binocular Visual Acuity of Adults, United States, 1960-1962. 25 cents. No. 4. Blood Pressure of Adults, by Age and Sex, United States, 1960-1962. 35 cents. No. 5. Blood Pressure of Adults, by Race and Region, United States, 1960-1962. 25 cents. No. Heart Disease in Adults, United States, 1960-1962. 35 cents. No. Selected Dental Findings in Adults, United States, 1960-1962. 30 cents. - Series 12. Data From the Health Records Survey No reports to date. Series 20. Data on mortality No reports to date. Series 21. Data on natality, marriage, and divorce No. 1. Natality Statistics Analysis, United States, 1962. 45 cents. No. 2. Demographic Characteristics of Persons Married Between January 1955 and June 1958, United States. Series 22. Data from the program of sample surveys related to vital records No reports to date. This report was originally published in the series ''Health Statistics from the U.S. National Health Survey,' which has since been replaced by the ''Vital and Health Statistics' series. It presents findings from a methodological study pertaining to improved techniques in data collection in the Health Interview Survey. Because this material is of continuing importance, and is relevant to data currently being released from the Survey, the report is being reprinted in its present form. Public Health Service Publication No. 1000-Series 2-No. 6 For sale by the Superintendent of Documents, U.S. Government Printing Office Washington, D.C. 20402 - Price 4) cents. NATIONAL CENTER| Series 2 For HEALTH STATISTICS | Number 6 VITALand HEALTH STATISTICS DATA EVALUATION AND METHODS RESEARCH Reporting of Hospitalization in the Health Interview Survey A methodological study of several factors affecting the reporting of hospital episodes, Washington, D.C. July 1965 U.S. DEPARTMENT OF HEALTH, EDUCATION, AND WELFARE Public Health Service Anthony J. Celebrezze Luther L. Terry Secretary Surgeon General NATIONAL CENTER FOR HEALTH STATISTICS FORREST E. LINDER, PH. D., Director THEODORE D. WOOLSEY, Deputy Director OSWALD K. SAGEN, PH. D., Assistant Director WALT R. SIMMONS, M.A., Statistical Advisor ALICE M. WATERHOUSE, M.D., Medical Advisor JAMES E. KELLY, D.D.S., Dental Advisor LOUIS R. STOLCIS, M.A., Executive Officer OFFICE OF HEALTH STATISTICS ANALYSIS Iwao M. Moriyama, Pu. D., Chief DIVISION OF VITAL STATISTICS RoserT D. Grove, Pu. D., Chief DIVISION OF HEALTH INTERVIEW STATISTICS PuiLip S. LAWRENCE, Sc. D., Chief DIVISION OF HEALTH RECORDS STATISTICS MonroE G. SIREN, Pu. D., Chief DIVISION OF HEALTH EXAMINATION STATISTICS ArtHUR J. McDoweLL, Chief ~ DIVISION OF DATA PROCESSING SipNEY BINDER, Chief COOPERATION OF THE SURVEY RESEARCH CENTER, THE UNIVERSITY OF MICHIGAN AND THE BUREAU OF THE CENSUS Under legislation establishing the National Health Survey, the Public Health Service is authorized to use, insofar as possible, the services or facilities of other Federal, State, or private agencies. I'he methodological study in this report was performed under a con- tractual arrangement with the Survey Research Center, Institute for Social Research, The University of Michigan, The Bureau of the Census also participated actively in the planning and conduct of the research. Public Health Service Publication No. 1000-Series 2-No. 6 PREFACE The outstanding importance of reliable na- tional statistics on hospitalizationhas led the U.S, National Health Survey to give high priority to the problem of securing such statistics. The first step in this program was taken when plans were being made for the Health Interview Survey in the fall of 1956. At that time hospitalization was designated as one of the basic topics with which that Survey should concern itself, The Health Interview Survey is one of three major data collection programs of the U. S, Nation- al Health Survey. Based upon sampling of house- holds throughout the United States, it seeks to gather by means of interviews various types of health information from which national and re- gional statistics canbe derived. The statistical de- sign and procedures used in the survey are de- scribed in detail in two National Health Survey Publications.!' 2 The datacollected include illness and accidental injuries, chronic conditions and im- pairments, disability, hospitalization, the use of medical and dental care, and related health topics. The information about hospitalization experience is collected by asking about instances when per- sons in the household were confined to a hospital overnight or longer within the 12-month period ending at the beginning of the interview week. Questions are then asked about each hospital epi- sode, including: month of admission, length of stay in days, diagnosis, operations performed, and name and address of the hospital. Since the questions cover only persons living in the household at the time of interview, the sta- tistical data developed from the interviews exclude the experience of persons who would have been ly, 3. tistical Design of the National Health Survey. The Sta- Heal th Household- Interview Survey. Health Statistics. Se- ries A-2. PHS Publication No. 584-A2. Pub- lic Health Service. Washington, D.C., July 1958. 2y. S. National Health Survey. Concepts and Definitions 1n the Health Household- Interview Survey. Health Statistics. Series A-3. PHS Publication No. 584-43. Public Health Service. Washington, D.C., Septem- ber 1958. living in the sampled households had they notdied in the year prior tothe interview. A method of es- timating the volume of this hospitalization of de- cedents has been developed and the reporthas been published,3 Paralleling its programs of data collection the U. S. National Health Survey has undertaken to evaluate the reliability of its own statistics through a series of research studies. Since the hospitaliza- tion information was considered to rate high in importance, plans were made soon after the Health Interview Survey got under way to test the relia- bility of reporting of hospital episodes ina series of contract studies. The first of these, of a preliminary nature, made use of data collected in an earlier survey and laid the groundwork for later studies. The results were not published. The second investigation was conducted as a part of a study with broader objectives. The Health Insurance Plan of Greater New York (H.I.P.) sam- pled its enrollees and, for each person in the sam- ple, produced a chronological record of medical services received from the Plan and of hospitali- zations incurred during a period of a year, Inter- views were then conducted in the households in which the sampled enrollees lived. The interview was the same one being used throughout the Nation in the Health Interview Survey. Responses in these interviews were compared with information from the medical records, thus permitting a direct measure of the extent of underreporting in this particular population, The findings with regard tounderreporting of hospitalization in the study of H.I.P. enrollees are contained in a forthcoming publication. Before the study of H.I.P. enrollees was com- pleted plans were made for a study that would in- clude a larger sample of hospitalizations and be devoted entirely to problems of reporting of hos- pital episodes in the Health Interview Survey. For 3u.s. National Health Survey, Hospital Utili- zation 1n the Last Year of Life. Health Statis- tics. Series D-3. PHS Publication No. 584-D3.Pub- lic Health Service. Washington, D. C., January 1961. this purpose a contract was made with the Univer- sity of Michigan's Survey Research Center, Insti- tute for Social Research, and this is the study the results of which are described in the present re- port. The relationships between the staff of the U. S. National Health Survey and the Institute for Social Research were very close, permitting the Survey to gain the maximum profit from observation of the data collection and participation in the analy- sis. The Bureau of the Census staff, too, concerned as they are with the quality of data which they are collecting for the Survey, participated in all phases of the study. The Bureau's participation, in ways which will be described in the report, was also essential inorder to ensure comparability between the interview results from the study and those ob- tained in the national survey. Dr. Abbott Ferris and Mrs. Katherine Capt carried the primaryre- sponsibility for the Bureau of the Census. Of crucial importance in the present study was the assistance of Dr. Vergil N, Slee, Director of the Professional Activity Study of the Commission on Professional and Hospital Activities, Inc. Ar- rangements were made by Dr. Slee for the sam- pling of the discharge records of hospitals partici- pating in the Professional Activity Study (P.A.S.), and these records formed the main basis of the criterion source against which interview results were checked. For those unfamiliar with the na- ture of the Professional Activity Study a briefde- scription of this useful organization will be found in Appendix III of this report. Also to be found in Appendix Ill isa list of the hospitals participating in the P.A.S. which agreed to allow their records to be used for the study. Having been assured that the information from the hospital files would be accorded confidential treat- ment, 21 of the 23 hospitals selected in the sample gave their permission. This assistance is grate- fully acknowledged. For the "Developmental and Evaluation Stud- ies' which are carried out at its expense but are not directly conducted by the National Health Sur- vey, a staff member is assigned for liaison with the research organization doing the study. In addition to keeping closely informed on the study progress and conveying the National Health Survey's view- point in decisions on study methodology, the liai- son person edits the final research report for pub- lication in Health Statistics, Series D. For this study, Mr, Earl Bryant discharged these respon- sibilities, CONTENTS Prefac@ ==-mmmmmmc meee een ——————— - INET OQ UCTION = or ve rr er ce cr rc me et ot tm Objectives and Scope of the Research ======emmoommomommmmoooo The Study Design —====== mmm meee Abstracting Hospital Records =========m mmm mm mmm meee eee Coding the Data =======m mmm meee eee meee Matching Hospital Record With Interview Report =-======cecace-- The Time Reference -=-===mmemm momen Definitions of Certain Terms Used in This Report-=-=-=====-=c--- Type of Respondent =-===mmemm ecm meee eee eee mmm Description of the Sample ========m emo m mame meee Underreporting Based on a Comparison of Interview Reported and Hospital Recorded Episodes =====mmmmmmomam ammo Underreporting by Type and Demographic Characteristics of Respondents ===== == mmm eee Underreporting by Relationships Between Sample Person and Respondent = === === comme Family Relationship ========oemmmm mn meee Age Comparison === === == mmm Sex COmMPAariSON ====m= =m ee eee eee eee eee Underreporting by Health Characteristics-=======ceecammmce eee Underreporting Based on Matched Hospital Episodes--=====cccmmaeaoooo Underreporting by Length of Stay======m =m ommm ooo Underreporting by Diagnostic and Operation Classes----=-====—=cecau- Underreporting by Time Interval Between Interview and Hospital Discharge —-======== =o cm ee Relationship Between Diagnostic Rating, Length of Stay, and Time Between Interview and Hospital Discharge---=-===mecommmaaoo__ The Follow-up Interview====- == mmm eee eee Introduction == == == =m mm em ee ee al Underreporting by Characteristics of the Respondents-=---====ccooeo- Reaction to the Illness and the Hospital Episode-==--===aemocmmmmoon Reaction to the Interview and the INI VIEWET == we ema mma m——mm—————— A Discriminant Analysis of the Follow-up Interview Data------------= Misreporting Characteristics of HospitalizationS-=====e=mecmcmmccaaean Accuracy of Reporting the Length of Stay=--===ceeeemaaaamaaoooo Misreporting of Diagnoses and Operations Causing Hospitalization Appendix I: Part 1: Sampling Errors--------eeee cmoomommmccoao. Part 2: Analysis of Nonresponse-=--= -=--eeeccommaano Appendix II: Rating of DiagnoSesS=-====mmmm mmm m cece meme Rating of Operations=======m mm omm moomoo Appendix III: Hospitals Which Co-operated in the Study--------=--= The Professional Activity Study-=-===-ecceeccacaa_ Appendix IV: Forms and Questionnaires Used in the Study--------- — Orn Us sD WO 14 14 15 16 17 20 24 31 36 39 40 43 50 52 52 54 57 58 60 61 62 62 64 SYMBOLS Magnitude of the sampling error precludes showing separate estimateS----=========-- (%) Percentage is greater than zero but is less than 0,05 ===cm=mmmm mre erm * REPORTING OF HOSPITALIZATION in the Health Interview Survey The following research report was prepared by the Survey Research Center, Institute for Social Research, The University of Michigan, under contract with the U. S. National Health Survey. The finaings anu conclusions are those of the Survey Research Center. Cnartes F. Cannell, Pn.D., directes tne project for the Survey Research Center. He was assist- ed by uwordvon Fisner and Thomas Bakker. lrs.Charlotte Winter and irs. Doris Muehl helped to develop codes and supervised the coding. Leslie Kish, Ph.D., proviced guidance and assistance on statisti- cal proolems. INTRODUCTION OBJECTIVES AND SCOPE OF THE RESEARCH There were three major objectives under- lying this research. 1. To obtain estimates of the amount of un- derreporting of hospital episodes in house- hold interviews in order to provide a rough approximation to underreporting in the U. S. National Health Survey, 2. To analyze the types of hospitalizations which were underreported and to investi- gate some of the factors relating to un- derreporting. 3. To study some response errors and to explore factors associated with these er- rors for hospitalizations which were re- ported. As the objectives indicate, this was a study of response error. In addition to estimates of the type and magnitude of underreporting, it included the study of some correlates of underreporting. For example, Are the underreports character- istic of particular types of respondents? What kind of information is most subject to underre- porting? What are some of the situations in which underreporting is likely to occur? Unlike many studies of response error which are concerned with the performance of the inter- viewer, this study focused on the respondent. However, many aspects of response error are clearly the result of complex psychological forces generated by the interaction between the inter- viewer and the respondent. Some of these forces will be examined in this report. In this study, as in the National Health Sur- vey household interviews, the respondent was asked to report the hospitalizations he and other members of his family experienced during the year preceding the interview. He may or may not have reported the hospital episodes, and the in- formation about those which he did report may be correct or incorrect, complete or incomplete. Several reasons can be listed for theseinaccura- cies and omissions. The respondent may not have known that a hospitalization had occurred and therefore could not report it. For example, he may nothave known about his father's hospitalization if his father only recently went to live with the family, Or the respondent may have known about his mother-in- law's hospitalization but did not know the type of operation which was performed. The respondent may not have understood what was wanted in the question or he misunder- stood the concept underlying it. Thus when the interviewer specified that she was talking about the year preceding the week of the interview, the respondent may have thought of the calendar year, or his ''year of recollection may have extended several weeks or even months beyond the actual year, The respondent may have once known the in- formation requested but may have forgotten it. Minor events in the past are easily forgotten. The respondent may have remembered the information, but recalled it inaccurately and therefore reported it inaccurately. Perceptual distortion over time may have diminished sig- nificantly the ability to recall the event with ac- curacy. The respondent may have remembered the information accurately but reported it inaccu- rately or not at all because it was embarrassing to him, Two major variables underlie these potential sources of error and are given special attention in this report. These are memory and motivation. Early experiments in the psychological lab- oratory demonstrated two principles of memory which are important to the present study. The first principle is that memory is better for re- cent events than for those having a greater time lapse. The second principle is that events having a great impact on the person will be remembered better than those having only a minor impact. Such principles coincide with everyday experience. In terms of this study one would expecthospitali- zations occurring close to the date of the inter- view to be reported more accurately and more completely than those occurring earlier. One would also expect longer or more serious hos- pitalizations to be reported more completely than short, less serious ones. There are some special cases of these prin- ciples in this study. In general, a hospitalization may be expected to be less important and less salient to a respondent who is reporting for some- one else than when the hospitalization was his own. A routine appendectomy should have much more of an impact on the patient than on members of his family, The patient, if he has had several hospitalizations, is likely to remember the more serious and forget the minor. In general one may expect a ''decaying' of experiences over time, depending upon the seri- ousness of the event, and the closeness of the event to the respondent. But to consider memories as fixed and "life- less" is unreal and misleading, Memory is an active, dynamic process which follows predict- able patterns. One of the most important forces in memory is motivation. There is a tendency to integrate events into one's psychological life in such a way that they fit most comfortably with past experi- ences and with an image or perception of one- self and one's world. Numerous experiments testify to the selectivity and distortion which occur in the recollection of an event. In working on consumer economics, the Sur- vey Research Center found that if one wants to learn respondents’ current incomes, motivation is important, Most people know, at least approxi- mately, what their current income is but whether or not they will communicate this information depends on their willingness to do so. If one wants to know their incomes for past years, the problem is more difficult. However, willing they may be to answer, many will have forgotten and many will "remember' so inaccurately that the usefulness of the information they offer is se- verely restricted. Those who do answer tend to report their earlier financial situation in a more favorable light. Memory, in short, is not a simple process by which the events of the present recede uni- formly into the past. This kind of decay does occur, but it is modified by a number of other factors, including the meaningfulness of the ini- tial experience, the degree to which it was "learned," and the interference of other experi- ences. In addition, the way in which things are remembered depends upon their congruence with the individual's other experience and with his image of himself. Such factors determine whether or not we remember at all, and in what system- atic ways our recollections differ from events as they actually occurred. Thus far the motivational forces which are closely related to the psychological life of the respondent have been discussed. They determine whether information can be reported accurately. There is another constellation of motives of a different type which is also relevant, and which influences whether the information will be re- ported accurately. From the most elementary point of view, the motivational level of the respondent will deter- mine how much effort he is willing to make to give an accurate report. In order to report ac- curately the respondent must relive or review carefully his experience, constantly checking his own memory, or he must resort to records of the event. The farther away the event is in time or the less importance it has, the greater the energy required to recall it, Frequently respondents give inaccurate information merely to avoid the work required to respond accurately. Particularly when the hospital experience has been embar- rassing or unpleasant or especially threatening, either physically or psychologically, the respond- ent may be unwilling to dwell on the event enough to be able to report it accurately, But perhaps a more serious type of problem occurs when the motives or goals of therespond- ent are served better by inaccurate reporting. For example, the respondent who has been hos- pitalized for alcoholism, mental disorder, or venereal disease may not be motivated to report the hospitalization because of its presumed anti- social nature, Other conditions, such as breast amputations, reproductive organ disorders, and the like may be embarrassing to the respondent, may threaten her self-image, or may be con- sidered too personal to discuss, It is likely that such hospital episodes will be suppressed. Suppression refers to the tendency of the respondent to withhold information which he is able to report because it puts him in an unfavor- able light; either because of his self-image or because of his perceptions of others' attitudes toward him. Information may be suppressed for fear that disclosure of the information would result in an unfavorable attitude toward him. Examples of this are: hospitalizations for mental or nervous dis- orders, venereal disease, alcoholism, et cetera, or for other disorders which are attributed to or associated by folklore with mental or moral deviations. Information may be suppressed because of embarrassment due to the personal nature of the problem; for example, various ''female troubles are not discussed by some segments of the popu- lation. Information may be suppressed due to threat to self-image. Examples: a hysterectomy may change perception as a ''complete woman''; am- putation or loss of other organs may result in changed perception of self which is psychologically threatening and therefore suppressed. Respondents may react to questions about such conditions by refusing to grant an interview, refusing to report an embarrassing condition, or by misreporting the condition in such a way as to make it more acceptable. Thus the respondent may be willing to report "female troubles' when she would not report the specific problem. In addition to the subject matter, lowered «motivation may also occur because of negative reactions to the survey, its objectives and spon- sorship, or to the interviewer. To participate in an interview requires that the respondent accept the goals of the survey and react in a positive way to the interviewer. A negative reaction to either may be expected to result in inaccurate data. The effect of memory and the types of motivation which have been discussed would be expected to result in a net underreporting rather than an overreporting of hospital episodes. There are few motives which would be expected to lead the respondent to overreport his hospitalizations. In summary several hypotheses about factors leading to underreporting are as follows: 1. The 12-month period of reference used by the National Health Survey for hospi- talization data is arbitrary and the an- chorage of the date, a year ago, may be so vague that a person remembers his hospitalization which occurred within the 12 months as having occurred earlier. The reverse is also true. Some hospital- izations which occurred prior to the year will be remembered as being within the year. This type of error may be random but it is likely that the effects of motiva- tion will lead to misplacing a hospital- ization backwards rather than bringing it forward. 2. The greater the time interval between the hospitalization and the interview, the less well it will be reported. Particularly, minor hospital episodes are more likely to be underreported as the time span be- tween the hospitalization and the interview increases. 3. Some hospital episodes can be expected to be suppressed or withheld because they place the respondent in an unfavorable light. 4. Negative attitudes toward the interviewer, the survey, or its sponsors may result in underreporting. THE STUDY DESIGN Since the study was focused primarily on problems of underreporting of hospital episodes rather than on overreporting, the sample con- sisted of persons who were known to have been in a hospital. The sample was a probability se- lection of persons with one or more discharges during the period, April 1, 1958-March 31, 1959 from 21 hospitals”. Stratification by month of discharge was used in order to obtain a propor- tionate number of persons discharged each month during the sampling time interval, The surnames, addresses, and telephone numbers of the sample persons were assigned to a group of experienced Census interviewers, all of whom were regular interviewers for the Na- tional Health Survey. The procedures in the field were essentially the same as those used in the National Health Survey's health interview survey. The basic questionnaire was the same; the interviewing instructions and procedures were the same, This was important since a major purpose of the study was to evaluate the amount of underreporting of hospital episodes in the National Health Survey. The interviewers were not told the purpose of the study because such knowledge could cause them to probe harder for hospital episodes, or in some way change their usual National Health Survey interviewing methods. The study design was sufficiently different, however, from that of the National Health Survey that interviewers would likely guess the purpose of the study in the early stages of interviewing. * The hospitals participating in the study were members of the Professional Activity Study (PAS). A list of the hospitals and a brief description of PAS are given in Appen- dix 111, During the first two weeks of the study, the interview assignments consisted of about 300 names and addresses that were chosen from the general population in the study areas and 100 ad- dresses of sample persons. Thus during this pe- riod the proportion of hospitalizations reported in the interviews was somewhat similar to that normally reported in the National Health Survey; consequently there was a good chance that the true purpose of the study would be concealed. A com- parison of hospitalizations reported for sample persons in these interviews was made with those reported in interviews taken later. This compari- son showed no difference in reporting rates. The field work was carried out by 27 inter- viewers working in 18 primary sampling areas located in 14 states. Interviewing started Aprill, 1959 with assignments each week through June 1959. Interviewers were instructed to follow the standard National Health Survey procedure at each sample address. Each adult who was found at home was interviewed about himself. Informa- tion for adults who were absent and for all un- married children under 18 years of age was ob- tained by interviewing a responsible family mem- ber. This means that the person whose hospital record was drawn into the sample might be inter - viewed about himself or another family member might report for him. In order to obtain additional information on characteristics of underreporting, and, hopefully, reasons why hospitalizations were not reported, a follow-up interview using a specially designed questionnaire was conducted with families who did not report all hospitalizations of sample per- sons, and with a l10-percent sample of families who correctly reported the sample persons’ hos- pital episodes. These interviews were conducted by Census' Regional Supervisors. The sample hospitals are scattered through- out the East and Midwest of the United States with a couple in the Mountain States. None is in the Far West or deep South. The hospitals vary widely in size, the smallest having 3,000 annual discharges and the largest 28,000. The 21 sample hospitals were chosen from some 95 participating in the Professional Activity Study. They were selected on a subjective basis, mostly to provide the widest possible geographic distribution. Also only hospitals were chosen which were in or near sampling areas used on the National Health Sur- vey where experienced interviewers were located. ABSTRACTING HOSPITAL RECORDS The hospitals were asked to complete Case Abstract Forms for all discharges that sample persons experienced between January 1, 1958 4 and June 30, 1959 (see Appendix IV for the form and questionnaires used in this study). Records, therefore, were obtained on all discharges (de- pending on the thoroughness of the record search) during the reference period of one year before the week of interview, Obtaining records back to January 1, 1958 made it possible to identify some erroneously reported hospitalizations which ac- tually occurred more than a year before the interview. It is not possible to know whether all the discharge records for the sample persons were abstracted or not. There is evidence however, that, at most, only a very few records were not abstracted. For persons who experienced only one hospital episode during the reference year (about 90 percent of the sample), it is known that abstracting was complete. The results of control methods used gives assurance that the vast ma- jority of records for persons with multiple hospi- talizations were also abstracted. CODING THE DATA The information reported in the original, or basic interview, was coded by the Bureau of the Census using standard National Health Survey procedures, thus making the data comparable in this respect to that obtained inthe National Health Survey. Except for medical coding, the follow-up in- terviews and the Case Abstract Forms were coded by the Survey Research Center. Coding of diagnoses and operations reported in interviews was done by the Bureau of the Census; medical coding on hospital records was done by the hos- pitals. MATCHING HOSPITAL RECORD WITH INTERVIEW REPORT After coding by the Bureau of the Census, all questionnaires and Case Abstract Forms were sent to the Survey Research Center. The first task was to match the person whose hospitaliza- tion was drawn into the sample with the same person on the interview. The two forms were matched, independently, by two persons using the name, address, age, sex, and race. In most cases the matching was accomplished easily and independent matching proved highly reliable, Fewer than 1 percent of the attempted matches were doubtful. Final de- cision on the problem cases was made by two supervisors. Similarly, it was necessary to match the hospital episode reported in the interview with that on the hospital abstract. Matching of episodes was done on a subjective basis rather than on some strict criteria. Most sample persons had only one episode and usually for such cases this matching was readily apparent. In cases of mul- tiple hospitalizations, particularly where the di- agnosis was the same for all episodes, matching was more difficult,” Undoubtedly errors were made. Some cases were classified as. matched that were not and others that were actually the same episode were considered to be unmatched. For this reason, as far as is possible, the analysis is based on all episodes reported in the interview and all those recorded from the hospital records. Thus the effect of errors due to mismatching was kept to a minimum, THE TIME REFERENCE Interview assignments were made for a par- ticular week, and in most instances were com- pleted within that week. Those which could not be completed were taken the following week or were reassigned to a later week. This analysis includes hospital discharges occurring one year prior to the Sunday of the week of assignment. Atthe time of the analysis it was understood that the inter- viewer asked about hospitalizations during the 12 months prior to the Sunday night of the week of assignment, This was erroneous. In fact, the interviewers asked about all episodes during the 12 months beginning with the Sunday of the week in which the interview was taken. Fortunately, this difference in time periods affected only a very few cases and were found ina special analy- sis to make no changes in the findings. DEFINITIONS OF CERTAIN TERMS USED IN THIS REPORT Several descriptive terms used inthis report are defined as follows: Matched case.—A matched case is one in which both the interview report and the hospital record were considered to refer to the samehos- pital episode and both documents indicated that the episode occurred during the reference year. Underreport.— A hospital episode is an un- derreport if the hospital record showed the epi- sode to be within the reference year and there was no matching episode reported in the house- hold interview, *Both tn sampling and in matching, the techniques and control methods were much more elaborate than it appears from this brief The reader who wishes to know more about the methods can obtain them by National Health Survey or to the Survey Research Center. description. writing to the Overreport.—A hospital episode is an over- report if it was reported in the interview to have occurred in a sample hospital during the refer- ence year, and there was either no hospital rec- ord for the episode or the hospital record showed that the episode actually occurred outside the reference year, All episodes.—Many of the tables in this re- port refer to "all episodes." From the interview "all episodes'' were the matched cases plus the overreports. For hospital records, "all episodes" included the matched cases plus the underreports. Number of episodes recorded.—This term is used throughout the report to mean the num- ber of hospital episodes for sample persons for which hospital abstracts were obtained. TYPE OF RESPONDENT In the initial interview all adults who were found at home were interviewed about themselves, The exceptions are adults who were ill or in- competent. Adults who were not present were re- ported for by another adult. All children under 18 years of age unless married were reported for by an adult. Many of the tables differentiate self-respondents from others as follows: Self-respondent.—The respondent is the sample person. Proxy child.—The respondent is an adult member of the family reporting for a sample person under 18 years of age. Proxy adult.—The respondent is an adult member of the family reporting for a sample person 18 years of age or over. DESCRIPTION OF THE SAMPLE Interviews were completed on 1,505 sample persons. Fourteen of these reported single epi- sodes for which there were no corresponding hospital records. The remaining 1,491 persons experienced 1,833 discharges according to hos- pital records. In the interviews, 1,645 episodes were reported. Of these reported episodes, .1,600 were matched with hospital records, and 45 were not matched. There were 233 hospital-recorded episodes that were not reported inthe interviews. The 45 episodes reported in the interviews which could not be verified from hospital records are considered as overreports. For each such episode reported a second search of hospital records was made. Nevertheless, these overre- ports should not be interpreted as an accurate estimate of overreporting, even for this special sample. The respondent may have reported the episode correctly in the interview but perhaps misnamed the hospital, or maybe because of other kinds of errors the episode was misclas- sified. Based on the matched interview reports, 12 percent of the hospital episodes were not re- ported. When the 45 overreports are included, the net proportion underreported reduces to 10 percent. Some of the important characteristics of the sample are shown in table 1. The data on hospi- talizations were taken from hospital records, thus overreports are not shown. The demograph- ic characteristics of the sample were taken from the interview report since these data appear to be more appropriate for this purpose than those contained in the hospital record. Of particular importance is the distribution according to the status of the respondent who re- ported for the sample person. Females were more likely to be at home when the interviewer called than males, Consequently females reported for themselves much more frequently. About three quarters of the females were self-respondents compared with one quarter of the males. In all, 58 percent of the sample persons reported for themselves. The only diagnostic category which repre- sents any sizeable proportion of the total is de- liveries, accounting for over one fifth of all epi- sodes. Since people make a special attempt to remember birth dates of their children, it can be expected that hospitalizations for deliveries will be reported more accurately than those for other reasons. Because of this, and since a sizeable proportion of all hospitalizations is for deliveries, many of the tables in this report are divided into two sections, one for all episodes and the second excluding deliveries. According to the hospital records, about 45 percent of the hospitalizations involved an oper- ation. If deliveries are included with operations, this proportion increases to two thirds. Other than deliveries, the most frequently performed operations were tonsillectomies, reduction of fractures and dislocations, and for female genital disorders. Table 1. Number of sample persons and hospital episodes recorded by characteristics of the sample Number of Characteristic Number of api pisodes persons recorded Totalemmm mm meme meee eee - 1,491 1,833 Sex Male====m mmm me ee ee eee 507 613 Female=======c ccm eee 984 1,220 Respondent status Self-respondent===-===c-cc moon meee 879 1,092 Proxy respondent for children under 18----------a- 302 349 Proxy respondent for adults---=-=---cccccmmmmaaa.. 310 392 Type of hospitalization Single-==-=mcm momma 1,236 1,236 Multiple--=-m=-omm enema een 255 597 Operations performed Total excluding deliveries--=-----ccccccmmmcaaaaao 708 813 Deliveries===-==mmcoem comme eee 358 359 Education College graduate-======-=cemcemcccmc ccc 103 114 Some college-========-cc cme een 127 152 High school graduate---===-e-ecccccmmmmmcccccaaaa 417 511 Less than high school graduation-----=-=ccceacaaa- 565 726 Under 14 and education unknown---«=----cecceaaaa-o 279 330 Family income Under $2,000=======c=-cc mma emo 120 154 $2,000-3,999 === === mm mmm meee eee 238 301 $4,000-6,999 === cme eeeee 623 750 $7,000-9,999 === m mmm eee 230 272 $10,000 == =m mmm mm mm ee ee ee eee 196 248 UnKnown === === === mm ee eee eee eee eee 84 108 Age 0-1 mmm mmm ee eee 18 23 1-9 eee 209 244 10-17 mmm mmm ee ee ee eee em 77 84 18-34m mmm mmm mm eee eee 522 631 35-5 mm mmm mmm me eee 405 507 55-64 mm mmm mm mm me ee eee 119 156 65-7 hmmm mmm mm me eee 102 141 Jt mmm mmm me eee em 39 47 UNDERREPORTING BASED ON A COMPARISON OF INTERVIEW REPORTED AND HOSPITAL RECORDED EPISODES One questionthat needs to be answered is how the information obtained in household interviews differs from that recorded in hospitals. Is the accuracy of reporting different for respondents or sample persons with different characteristics? Do respondents of some ages report better than those of other ages? Do men report as well as women? This section compares the information asre- ported in the interviews with that from hospital records. On this basis, the underreporting amounts to 10 percent. The hospital records in- clude all hospitalizations from the sample hospi- tals which occurred during the reference year. All episodes reported as occurring in sample hospitals within the reference year are included whether or not they actually occurred within the year. These are the data which would usually be available to the analyst. The ratios shown in this report are weighted to adjust for unequal probabilities used in the sam- ple selection. Sampling errors which may be at- tached to these ratios are presented in Appendix I. Although sampling errors were computed and frequently tests of hypotheses are made, much of the analysis is based on meaningful patterns which may not pass such statistical tests because of small numbers involved. This was done because one of the important purposes of the study was to develop hypotheses which may be important for further research. UNDERREPORTING BY TYPE AND DEMOGRAPHIC CHARACTERISTICS OF RESPONDENTS As might be expected, there was a marked difference in the amount of underreporting de- pending on whether the sample person reported for himself or whether someone else reported for him. Sample persons reporting for themselves un- derreported by 7 percent while the rate for both proxy adults and proxy children was twice ashigh (table 2). When deliveries are excluded the total underreport increases from 10 to 12 percent. Only 8 fewer deliveries were reported than recorded in the hospitals, anunderreporting rate of only 2 per - cent. This low rate was expected since the birth of a child is a dramatic event, and the circumstances and dates are likely to be remembered. Since it is usually the women's role to care for sick or recuperating family members it would be expected that women are less likely to under- report hospitalizations than men; this idea isre- futed in table 2. Thereis little difference between men and women in reporting either for themselves or for other members of the family. The second half of the table, excluding deliv- eries, shows that women were slightly poorer re- porters for themselves than were men for them- selves. However, women reported somewhat bet- ter for other adults than did men. Table 3 indicates that there may be a slight tendency for underreporting to increase with in- creasing age of the respondent. When deliveries are excluded, however, the level of underreporting is raised in the age groups under 55 years of age so that there is no general upward trend with in- creasing age. Ages 65-74 show a larger underre- port than other ages; however the largest differ- ence shown (10 percentage points) is not statisti- cally significant at the five percent level. When ages 55 and over are combined (table 4) the dif- ference between the percentage underreporting in this age group and that of the best reporting age group (under 35) is statistically significant. Re- spondents under 35 years of age reported their hospitalizations more often than did other s because of the large proportion in this age group reporting for themselves and because of a large number of deliveries in the group. The best response by proxy was by persons 35-54 years of age. An age-sex comparison revealed no statisti- cal significance in differences in underreporting by male and female respondents. White respondents reported hospitalizations more accurately than did nonwhites (table 5). This tendency is apparent for all types of respondents except when reporting for another adult. These differences may well be a reflection of other var- iables, such as educational level and income. The relationship between education and amount of underreporting of hospital episodes shows an unusual pattern. Table 6 indicates that the respondents who graduated from high school or college report more accurately than those who have less than a high school education or who started but failed to complete college. The same pattern is observed for all episodes and for the episodes exclusive of deliveries. A tenuous hypoth- esis to account for this is that accurate report- ing is partially a matter of motivation. Those people who are highly motivated report more ac- curately than those whose motivation is low. Ac- complishment in school is also related to motiva- tion. Success in school requires diligence, re- Table 2. Percent underreporting of hospital episodes by type and sex of respondent and sex of sample person, including and excluding deliveries’ Type of respondent All respondents Proxy respondent for: Self-respondent Sex of respondent Children under 18 Adults and sample person Yunber Percent ” r r NTThar Percent der- ercen ercent der- episodes un er Tumber ) under- Bumbe® | vndes= episodes | "Toor cecorded re episodes 9 episodes recorded re- ported | recorded Te recorded re- ported ported ported All episodes Total==mnw== 1,833 10 349 14 392 14 1,092 7 Male respondent--- 311 11 52 12 98 16 161 8 Female respondent- 1,522 10 297 14 294 14 931 7 Male sample per- SON===m=mmmm———— 613 12 204 12 248 14 161 8 Female sample Person-========= 1,220 9 145 16 144 16 231 7 Excluding deliveries Total======- 1,474 12 348 14 367 15 759 10 Male respondent--- 290 11 51 12 78 19 161 8 Female respondent- 1,184 12 297 14 289 14 598 10 Male sample per- SON=======ccema= 613 12 204 12 248 14 161 8 Female sample PersSOn-====m==m=== 861 12 144 16 119 19 598 10 ihe percentages shown in this table and in tables 3-13 are based on all hospital discharges for sample persons renorted in the interviews to nave veen from sample hospitals, and on the number of discharges abstracted from nosni-— tal records. The percentages are appropriately weighted to reflect each individual's chance of being selected in tne sample. In cases of multiple episodes the personal characteristics accompany each episode. Thus the respondent charac— teristics are the came as the number of episodes, since some persons are included more than once. Table 3. Percent underreporting of hospital episodes by age of the respondent, including and excluding deliveries All episodes Excluding deliveries Age of respondent Number Percent Number Percent episodes under - episodes under - recorded reported recorded reported Total=m=-mmcmm meme mmc mcm m 1,833 10 1,474 12 18-34=m mmm meee ccc am 792 8 487 12 35-54 mcm mcm ecm me cme ———— 691 10 638 11 55=64==m meme cmc ccccmcc cm ————— 169 13 168 13 El a 128 18 128 18 75mm mmm cmc mccmm mmm —————— 53 14 53 14 Table 4. and excluding deliveries Percent underreporting of hospital episodes by age and type of respondent, including Type of respondent All respondents Proxy respondent for: Self-respondent Age of respondent Pendant Children under 18 Adults Number Percent under - Percent Percent Number episodes ZE~ Number | oo. Number | gor under - recorded episodes episodes episodes re- ported recorded Iga recorded Tee recorded orted oc ported ¢ ported P All episodes Total====-== 1,833 10 349 14 392 14 1,092 7 18-34~=cmmcmcmana 792 8 176 16 108 13 508 4 35-54mmmmmnmmn nme 691 10 164 9 174 11 353 11 55t-mmmmm meme ——— 350 15 9 (*) 110 22 231 10 Excluding deliveries Total=====--- 1,474 12 348 14 367 15 759 10 18-34==-mcmmcnnna— 487 12 175 16 87 16 225 6 35-54mmmmmmmmnnna 638 11 164 9 171 1 303 12 55 ann. ———— 349 15 9 (*) 109 22 231 10 Table 5. Percent underreporting of hospital episodes by race and type of respondent Type of respondent All respondents Proxy respondent for: Self-respondent Race of respondent Children under 18 Adults Number | Percent Number | Percent episodes | under- Number | Fercent | yumber | Percent | episodes under. recorded fe episodes under - episodes under- | recorded i ported | recorded Lo recorded re-~ porte ported ported Total--===-- 1,833 10 349 14 392 14 1,092 7 White-=====c==ca=-= 1,723 10 329 13 376 14 1,018 7 Nonwhite==-=======- 110 16 20 24 16 14 74 14 Table 6. including and excluding deliveries Percent underreporting of hospital episodes by education and type of respondent, Type of respondent All respondents Proxy respondent for: Self-respondent Education of Tw Suh Percent Children under 18 Adults Cber Pevosnt sponden umber : Aap episodes bodes Number Perens Number ho episodes es recorded episodes unde episodes u recorded ported re- re- ported recorded recorded ported ported All episodes Total======- 1,833 10 349 14 392 14 1,092 7 Less than high school gradua- tion=====cecea-- 829 13 141 14 173 20 515 10 High school graduate-===---- 646 7 139 11 130 9 377 4 Some college-=-=---- 180 16 42 30 38 11 100 1 College graduate Or MOre=====-=--- 155 5 27 2 40 15 88 2 Unknown-=========-- 23 (*) 0 0 11 (*) 12 (*) Excluding deliveries Total-=-==-= 1,474 12 348 14 367 15 759 10 Less than high school gradua- tion-=-=mmcmenan 698 14 141 13 158 22 399 11 High school graduate-=-=------ 488 10 138 12 123 10 227 8 Some college------ 149 18 42 30 38 11 69 14 College graduate Or MOre========= 116 5 27 2 37 12 52 3 Unknown==--====-== 23 (*) 0 0 11 (*) 12 (*) Table 7. Percent underreporting of hospital episodes by education of respond- Type of respondent All respondents Proxy respondent for: Self-respondent Education and Children under 18 Adults family income Percent of respondent Number . Number Percent episodes | YM9er- Number Porcam: Number | Percent | opjgodes | under- recorded res episodes | Under= | ohig0des | under- | recorded re- ported | recorded re- recorded re~ ported ported ported All episodes Total Total-=---=-=--- 1,833 10 349 14 392 14 1,092 2 Under $2,000------- 154 18 19 45 34 18 101 15 $2,000-3,999------- 301 13 50 12 46 26 205 10 $4,000-6,999-~---~- 750 10 167 10 142 13 441 6 $7,000-9,999-=====~ 272 8 48 18 68 10 156 3 $10,0004----~----~=~- 248 8 49 7 72 13 127 6 Unknown============ 108 14 16 (%) 30 16 62 10 High school graduate or less Total----===- 1,475 10 280 13 303 15 892 7 Under $2,000------- 146 18 19 45 29 18 98 12 $2,000-3,999------- 273 14 47 7 42 36 184 11 $4,000-6,999--~----- 641 8 145 11 120 11 376 6 $7,000-9,999------- 190 7 29 12 50 8 1 le 5 $10,000+---=-======= 137 9 25 4 41 18 71 5 Unknown==------=-==-= 88 11 15 (*) 21 (*) 52 11 Some college or college graduate Total-=-=--==-=-= 339 11 69 18 78 13 188 7 Under $2,000------- 3 (*) 0 0 0 0 3 (*) $2,000-3,999------~- 27 14 3 (*) 3 (*) 21 16 $4,000-6,999-===--- 104 12 22 8 21 22 61 10 $7,000-9,999------- 80 11 19 26 18 16 43 (%) $10,000+-=-=--======= 109 7 24 10 29 5 56 6 Unknown==-========-- 12 (*) 1 (*) 7 *) 4 (%) Education unknown Total-==-====- 23 (*) 0 0 11 (*) 12 (*) ent, family income, and type of respondent, including and excluding deliveries Type of respondent All respondents Proxy respondent for: Self-respondent Zdugacion and Children under 18 Adults amily income Percent Percent of res dent Number _ Number : ponash episodes under Number Teen Number Pe oent episodes aden recorded 2 episodes | YN9€T" | episodes | YP9T" | recorded re d pores recorded res recorded i ports ported ported Excluding deliveries Total Total-===--=-- 1,474 12 348 14 367 15 159 10 Under $2,000------- 134 20 19 45 32 16 83 15 $2,000-3,999------- 243 14 50 9 43 28 150 12 $4,000-6,999------- 569 11 166 11 127 13 276 10 $7,000-9,999------- 226 10 48 18 67 12 111 5 $10,000+-=-======-~- 209 9 49 7 70 13 90 7 Unknown--===--=----- 93 14 16 (%) 28 16 49 10 High school graduate or less Total----=-=-- 1,186 12 279 13 281 16 626 10 Under $2,000------- 126 20 19 45 27 16 80 15 $2,000-3,999--==--- 217 14 47 7 39 34 131 1] $4,000-6,999-=-==-- 491 11 144 13 108 13 239 9 $7,000-9,999-===--- 162 9 29 12 49 10 84 7 $10,000+--------~--- 116 10 25 4 39 Li 52 5 Unknown--=-======a-- 74 12 15 (%) 19 (*) 40 12 Some college or college graduate Total-------- 265 12 69 18 75 11 121 9 Under $2,000------- 3 (*) 0 0 0 0 3 (*) $2,000-3,999---==-- 25 15 3 (*) 3 (*) 19 18 $4,000-6,999---~--- 73 12 22 8 18 18 33 12 $7,000-9,999------- 62 14 19 26 18 16 25 (*) $10,000+--=---==---- 91 8 24 10 29 5 38 9 Unknown==-=========- 11 (*) 1 (*) 7 (*) 3 (*) Education unknown Total-==-=--- 23 (*) 0 0 11 (*) 12 (*) Table 8. Percent underreporting of hospital episodes by family size and type of respondent, including and excluding deliveries Type of respondent All respondents Proxy respondent for: Self-respondent Family size Children under 18 Adults Number | Percent Number | Percent episodes under- Number Percent Number Percent episodes under- recorded re- episodes | Under- | episodes wder= 1% orded re- ported | recorded ree recorded LG ported ported ported All episodes Total--=====- 1,833 10 349 14 392 14 1,092 7 Ro cnn ms si 0, mo 75 12 0 0 0 0 75 12 ll 316 10 3 (*) 96 12 217 9 3 OF L4mwmmmnmeee 760 12 137 16 170 20 453 7 5 OF b=-========m== 494 7 146 12 96 4 252 5 mis AE 188 12 63 14 30 22 95 7 Excluding deliveries Total--==-==== 1,474 12 348 14 367 15 759 10 lemon mmm 75 12 0 0 0 0 75 12 0 0 ER 313 10 3 (*) 96 12 214 9 3 Or 4mmmmmemmmmmn- 598 14 137 16 161 21 300 9 5 OF 6=========ee== 354 10 145 12 86 4 123 13 Tt wwii mim me ne 134 14 63 14 24 24 47 7 sponsibility, and compliance with authority. Per- UNDERREPORTING haps these are some of the same traits required for accurate reporting of hospitalizations. Since a relationship exists between education and income, the income data for all episodes and for two educational groups are presented sepa- rately in table 7. There is a clearly observable pattern which indicates that accuracy of reporting increases with income, both for all episodes and excluding deliveries. Both educational groups demonstrate similar effects. Of the two variables, income has the major effect on the accuracy of re- porting. Table 8 shows the amount of underreporting for various family sizes. It might be expected that the larger the family the less accurate the report because it is more likely that the sample person would be reported for by a proxy respondent. Such does not appear to be true. Over-all, there was little difference in accuracy between large and small families. However, self-respondents of larger families reported more accurately. This is probably a reflection of age. One and often two- member families are characteristically composed of people in the older age groups where reporting tends to be worse. 14 BY RELATIONSHIPS BETWEEN SAMPLE PERSON AND RESPONDENT In addition to the characteristics of respond- ents or sample persons which might be expected to be related to how well episodes are reported there are some relationships between these two persons which could be expected to have some bearing on how well episodes are reported. For example, the closer the family relationship be- tween the respondent and the sample person the more one would expect the respondent to know about the hospitalizations of the sample person. The closer the ages the more accurate one might expect the information to be. The remainder of this section reports some of the effects of these factors on reporting of hospital episodes. Family Relationship Table 9 shows that the closer the relationship between the respondent and the sample person the more accurately hospital episodes were reported. Self-respondents, as the data in this report con- sistently show, were the most accurate reporters. Table 9. Percent underreporting of hospital episodes by relationship of sample person to respondent and type of respondent, including and excluding deliveries Type of respondent All respondents Proxy respondent for: Relationship of sample Children under 18 Adults person to respondent Number | Percent episodes under. Number Fefouny Number Beylent recorded re d episodes | YNY€T™ | episodes | UM9eT- porte recorded re recorded Te-= ported ported All episodes Total--==-ececccennnna- 1,833 | 10 | 349 | 14 | 392 | 14 Self-respondent----===-=-=--- 1,092 7 0 0 0 0 Sample person is spouse of respondent----====cccccaaa- 275 10 2 (*) 273 10 Sample person is child of respondent’ --==asnmsmnnnnmn 386 14 330 12 56 28 Sample person is other relative--------cccccmaao-- 78 22 15 28 63 21 Sample person is unrelated-- 2 (*) 2 (*) 0 0 Excluding deliveries Total====mcmemcccaaam- 1,474 12 348 14 367 15 Self-respondent=-=====-oc---- 759 10 0 0 0 0 Sample person is spouse of respondent------mmeecmeeaa- 255 11 1 (*) 254 10 Sample person is child of respondent !--eeeeeacannaaoo 385 14 330 12 55 29 Sample person is other relative-=---cc-ccccmccaooo 73 24 15 28 58 23 Sample person is unrelated-- 2 (*) 2 (*) 0 0 Luchild" does not refer to age The next most accurate group was the respondent reporting for his spouse; respondent reporting for a child was third. The least accurate was the re- spondent reporting for some other relative. This pattern also holds when deliveries are excluded. Looking at the accuracy of reporting for off- springs, it is seen that when they are under 18 years of age they were reported for with about the same accuracy as were spouses, However, epi- sodes for adult offsprings were underreported ata considerably higher rate. This probably reflects the greater independence of adult offsprings from the family and, conversely, the greater responsi- bility of parents for younger children. It means that the sample person is an offspring of the respondent. Age Comparison Table 10 shows age differences between re- spondents and sample persons. Only proxy re- spondents are included in this table. The hypothesis here is that the greater the age differential between the respondent and the sample person the less likely they are to be in close com- munication about their personal lives. This should be particularly true when the sample person is older than the respondent, The table indicates are- lationship but the differences are not statistically significant, It appears that the nearnessofthere- lationship between the sample person and respond- IS Table 10. Percent underreporting of hospital episodes by the age dif- ference between respondent and sample person, proxy respondents only Age difference Number Percent between respondent | episodes under - and sample person | recorded | reported Total=======- 741 18 Respondent is younger by 10 years or more--- 71 15 Respondent and sample person are within 10 yearg-===-====== 255 18 Respondent is older by 10 or more years=----- 415 20 Table 11. ent is more related to accuracy of reporting of hospital episodes than is the age difference. Sex Comparison Table 11 shows underreporting of hospital ep- isodes when proxy respondents report for sample persons of the same or different sex, All of the household interviews were taken by women interviewers. Thus, if there is a problem of reporting about hospital episodes between men and women it may be reflected in differences between sample persons and respondents or between re- spondents and interviewers. Considering all episodes, reporting was most complete when both the respondent and the sample person were male, Underreporting was highest when the respondent was male and the sample person female. A man may be embarrassedtore- port to a female interviewer about female hospi- talizations. It may be noted in table 11 however, that the differences are accounted for by ahigher rate of underreporting for children, For adults these tendencies are not present. The results are inconclusive and the sex relationship is apparently not a strong determinant to reporting hospital ep- isodes. Percent underreporting of hospital episodes by sex of respondent and sample person and type of respondent, proxy respondents only Type of respondent All respondents Proxy respondents for: Sex Of vesponnent zi Children under 18 Adults Number Percent episodes under - Number Percent Number Percent recorded | reported episodes under - episodes under - recorded | reported | recorded | reported Toll asscnmsnannas 741 14 349 14 392 14 Both respondent and sample person are male~-mmmccccmce amen 39 8 29 (*) 10 (*) Both respondent and sample person are femalem=mmmmmmcccmee me 178 15 122 14 56 16 Respondent is male; sample person is femalem==mmmcmemcm eae 111 17 23 24 88 14 Respondent is female; sample person is male- 413 13 175 14 238 13 Erie —— UNDERREPORTING BY HEALTH CHARACTERISTICS In the interview, questions were asked about the frequency of chronic and acute conditions, It was thought that accuracy of reporting might differ according to whether the sample person was ''very healthy," i.e., not suffering from chronic or acute conditions or was 'not healthy," suffering from several conditions, Since the number of conditions might be expected to increase with age, such a comparison was made for three age groups. Information on the number of conditions was obtained by counting the frequency of report of either chronic or acute conditions. All responses were divided into three groups: those mentioning no chronic or acute conditions, those mentioning one or two conditions, and those reporting three or more, In table 12 a strong relationship appears to exist between the number of chronic and acute con- ditions or both reported in the interview and the accuracy of reporting hospital episodes. The fewer conditions reported, the greater the underreport- ing of hospital episodes. In an attempt to under- stand these data, similar statistics were obtained for the respondent. The reasoning was that if the variable was actually related to the health of the sample person then the conditions which the re- spondent suffered would show no relationship with accuracy of reporting hospital episodes of others. From table 13 it is apparent that the relation- shipis as strong for the conditions of the respond- ent as it is for those of the sample person. It ap- pears that there is some factor other than health which is affecting the accuracy of reporting of hospitalizations. The best hypothesis is that the factor is motivation. The respondent who has a low level of motiva- tion to participate in the interview slides through the interview in such a way as to make the least demands on his time and energy. Thus he does not work very hard to report his hospitalizations and by the same process fails to report physical con- ditions suffered by himself and other members of the family. Hence, in addition to the usual problems of forgetting hospitalizations, there may be a strong factor of motivation accounting for some of the un- derreporting. Table 12. chronic and acute Percent underreporting of hospital episodes by age of the sample person, number of conditions reported in the interview for the sample person, and type of respondent, including and excluding deliveries Type of respondent All respondents Proxy respondent for: Self-respondent Age and number Children under 18 Adults of conditions Surber Percent Percent Percent Number Percent reported episodes under - Number under= Number Grider episodes | under- recorded Des episodes re- episodes re- recorded re ported | recorded ported recorded ported ported All episodes All ages Total------- 1,833 10 349 14 392 14 1,092 7 None---======mme-- 528 15 190 16 85 33 253 9 EE — 945 8 146 11 241 9 561 8 3mm mmm mmm 360 4 13 (*) 66 7 278 4 Under 18 Total------- 351 14 349 14 0 0 2 (*) NOE wwe mmimmcmn 191 16 190 16 0 0 1 (%) 1 or 2------mmmmmn 147 11 146 11 0 0 1 (*) Femme = 13 (*) 13 (*) 0 0 0 0 18-44 Total------- 896 8 0 0 202 14 694 6 None---==-=-======-- 290 13 0 0 69 31 221 7 1 or 2-==mmeccmee- 471 6 0 0 122 6 349 6 Femme mmm 135 3 0 0 11 (*) 124 4 ast Total-=====-= 586 1) 0 0 190 14 396 9 None-=-=--==-======== 47 28 0 0 16 44 31 20 lor 2--=-mm-mmm-- 330 12 0 0 119 14 211 10 Hmmm mmm meee 209 6 0 0 55 8 154 5 Excluding deliveries All ages Total--=-=--== 1,474 12 348 14 367 15 759 10 None-==-========nu- 370 20 189 16 68 42 113 15 l or 2----=mmmmm-- 782 10 146 11 233 9 403 10 Hemme mma 322 6 13 (*) 66 7 243 6 Under 18 Total------- 348 14 348 14 0 0 0 0 None-============-= 189 16 189 16 0 0 0 0 l or 2--==-mmmnnea- 146 1X 146 11 0 0 0 0 Herm 13 (*) 13 (%) 0 0 0 0 18-44 Total-=-=----- 542 12 0 0 177 16 365 10 None--===========- 135 25 0 0 52 42 83 13 l or 2-===mmmmmmun 306 8 0 0 114 4 192 10 Hemme 101 8 0 0 11 (*) 90 9 4x Total-==---~- 584 12 0 0 190 15 394 10 None---==-=-======- 46 28 0 0 16 44 30 20 lor 2---mmmmmemn- 330 Ii 0 0 119 14 211 10 Hemmmmmmmm—————— 208 6 0 0 55 8 153 5 Table 13. Percent underreporting of hospital episodes by age of the respondent, number of chron- ic and acute conditions reported in the interview by the respondent about himself, and type of respondent Type of respondent All respondents Proxy respondent for: Self-respondent Age and number Children under 18 Adults of conditions reported Number Dergers Percent Percent | Number | Percent episodes | Under- Number | ypder Number | ynder episodes | under- recorded re- episodes re- episodes _— recorded re- ported | recorded ported | recorded | ported ported All ages Total ====-- 1.833 10 349 14 392 14 1,092 8 None=-====m=cmeman 548 3 132 17 163 18 253 9 1 or 2-=--ecceenaa 913 9 174 10 178 14 561 8 mm mm ti 372 6 42 9 51 3 278 4 18-44 Total======- 1,193 9 308 14 191 13 694 6 None--======om-ca- 424 12 118 16 85 17 221 7 1 or 2---cmecmanaan 590 8 151 11 90 8 349 6 Femme 179 9 39 19 16 (*) 124 4 45+ Total-=-=---- 637 12 41 12 200 16 396 9 None--======mceou- 122 20 14 27 77 19 31 20 l or 2----mmemen-- 322 12 23 2 88 19 211 10 Femme meee 193 4 4 (*) 35 0 154 5 es for persons und UNDERREPORTING BASED ON MATCHED HOSPITAL EPISODES Since there is a particular interest in prob- lems of underreporting, the analysis in this section is based on those episodes of hospitalization re- ported in the household interview which could be matched with hospital records. The ratios of un- derreporting based on matched cases represent the maximum percentage of underreporting. It is entirely possible that a number of the cases re- ported in the interview that could not be matched with hospital records were actually the same epi- sodes as recorded. The matching procedure used makes it possible for an episode to appear as an underreport (when there was no interview report classified as a clear match with the hospital rec- ord) and also as anoverreport (when there was no clear match for the interview in the hospital rec- ords). In fact, there were five such cases. The in- dications are, however, that the matching criteria were quite good. Generally speaking, the distribu- tions presented in the preceding section where no matching criteria were used are about the same as similar distributions based on matched episodes. The following analysis is based on three fac- tors which are likely to be relatedtoone's ability to remember; namely, the seriousness of thehos- pitalization, the reason for hospitalization, and the time interval between the interview and discharge from the hospital. In planning the analysis of these data several measures of seriousness were considered. These were discussed with doctors who pointed out prob- lems in each measure contemplated. There was general agreement that length of stay inthe hospi- tal would provide a reasonably good index of seri- ousness. By seriousness is meant the level of physical threat or trauma which is involved. For example, it is generally true that the more serious the operation the longer the hospitalization. The same tendency usually is characteristic of non- operative cases. On this basis an analysis was made comparing three lengths of stay: 1 day is considered minor, 2-4 days is somewhat more severe, and 5 days and over is considered to be serious. These time periods are arbitrary but it was thought that they would show fairly high agree- ment with classifications of "major" and "minor" hospitalizations. Diagnoses and operations were used inthe in- vestigation of the hypothesis that respondents sup- press or withhold information about hospitaliza- tions which may place them in an unfavorable light. A test of this hypothesis requires ana priori classification of diagnoses and operations which differentiates between those that are embarrass- ing or threatening and those that are not. While in 20 the literature some discussion was found of spe- cific diagnoses and operations which cause psy- chological trauma, no over-all classification sys- tem was located and the writers devised their own classification. The diagnostic classification was a three- point scale based on the judgment of the research- ers as to what extent the diagnosis would be threatening or embarrassing. Alldiagnostic clas- gifications which, in the opinion of the raters, would be very embarrassing or threatening were placed in Rank 1. Rank 3 included the groups which were judged not embarrassing and non- threatening. Rank 2 contained a mixture of cate- gories which were thought to be somewhat threat- ening, or in which some diagnoses would be threatening and others would not. Thus Ranks 1 and 3 were kept as pure as possible, with 2 con- taining some of the uncertain categories. No claim is made for the validity of this scale, nor for the method of classification. While other people were consulted as to the ranking of diagnoses itis based on a subjective judgment of the authors. A similar scale was used to rank operations. Here, however, only Ranks 1 and 3 were used. It was felt by the raters that operations were much easier to rank, because of the specificity. Thus all operations were ranked either embarrassing or threatening, or those that were not. The ratings are shown in Appendix II. It was expected that these ratings would be positively correlated with the seriousness of the diagnosis or illness, and consequently, to some ex- tent correlated with the length of stay in the hos- pital. In devising ratings, like these, there are two important considerations. The first is that theor- dering of the items should be predominantly cor- rect and the second is that the average value of those items placed in one grouping should differ from that of another. The rating of any one item may be inaccurate. The main test of the usefulness of the scales is whether or not they help to differ- entiate and understand the differences in reporting hospital episodes. UNDERREPORTING BY LENGTH OF STAY Table 14 clearly indicates that underreporting of hospitalizations is related to the length of stay in the hospital. The only reversal of the trend is for episodes lasting longer than a month. Thisre- Table 14. Percent underreporting of hospital episodes by length of stay shown in hospital records and type of respondent, excluding overreports Type of respondent All respondents Proxy respondent for: Self-respondent Length of sta ji nla Children under 18 Adults Percent P t Number Percent Percent | Number green episodes under Number under Number under episodes under - recorded For episodes re- episodes ra recorded LQ ported | recorded recorded ported ported ported Total-==---- 1,833 12 349 16 392 18 1,092 9 lemme mmm eo 150 26 85 24 8 - 57 28 2-fmmmmccccm————— 646 14 141 14 125 23 380 11 ———— 456 10 64 11 98 24 294 6 8 = Lm mmm mom som sai si 352 10 43 11 100 10 209 10 15321 mmm mse 111 6 4 (*) 34 9 73 5 22-30-=====-mm-nn- 58 2 3 (*) 19 (*) 36 1 Blob mm mmm wn 46 8 7 (*) 8 (*) 31 {*) Unknown-====ee-oo- 14 (*) 2 (*) 0 0 12 {%) Table 15. Percent underreporting of hospital episodes by age of sample person and length of stay shown in hospital records, excluding overreports Length of stay (in days) All stays 1 2-4 5+ Age of sample person Nite Sue Nunibor Tercens Number Pacers Nitibap Ponusne episodes | "Mer" | opigodes | under- episodes | YM9€T" | opigodes | Under- recorded re- recorded Te- recorded Tos recorded re- ported ported ported ported Total'==-=-- 1,833 12 150 26 646 14 1,023 9 0-18=mmcmcmcemeeee 351 16 85 24 143 14 121 13 18-34=ccccmaceaa 631 8 35 26 288 10 305 5 35-54 mmcmmmmaaas 507 13 15 33 153 18 334 11 S55tmmmmmm meee 344 15 15 31 62 24 263 12 There are 13 episodes from interview reports and 14 from n spital records for which the lenyjth of stay was unknown. Totals add to 1,587 and 1,819. Table 16. Percent underreporting of hospital episodes by age of the respondent and length of stay shown in hospital records, excluding overreports Length of stay (in days) All stays 1 2-4 5+ Age of respondent Percent Percen t Percent Percent Number NRT Number UB Number vnder— Number unders episodes Tie episodes Hg episodes ra= episodes ras recorded ported recorded ported recorded ported recorded ported Total semen 1,833 12 150 26 646 14 1,023 9 18-34-cccccmcaaa 792 10 ZL 25 342 10 375 7 35-54mmmmmcmeaaae 691 13 61 24 238 16 386 10 55+-mmmmmmm emma 350 15 18 37 66 28 262 13 Yhere are 13 episodes from interview reports and 14 from nospital records for which the length of stay was unknown. Totals add to 1,587 and 1,819. 21 Table 17. Percent underreporting of hospital episodes by relationship of sample person to the respondent and length of stay shown in hospital records, excluding overreports Length of stay (in days) Relationship of S11 supys ! 2 7 sample person to respondent Number Peraenc Number Number Popoent Number Sercent episodes | "P9¢T" | episodes under- | ohjigo0des | er- | episodes | “MCT” recorded al recorded res recorded To= recorded Te ported ported ported ported Total’ --==-= 1,833 12 150 26 646 14 1,023 9 Self-respondent--- 1,092 9 57 28 380 11 643 6 Sample person is spouse===-====== 275 15 6 29 78 16 191 14 Sample person is child---======-=- 386 16 81 23 161 16 142 13 Sample person is other relative-- 78 26 4 (*) 27 35 47 20 Sample person is unrelated-----=-- 2 (*) 2 (*) 0 0 0 0 lrnere r | oc | from intervi renor sy 14 from hospital records for which the lengtn of stay unk n etal t > 37 an |,819. Table 18. Percent underreporting of hospital episodes by relationship of sample person to re- spondent and length of stay shown in hospital records, excluding overreports Length of stay (in days) Relationship of All stays 1 2-4 5+ sample person to respondent Number Parzen Number Tesgen: Number Pepoant Number Pasgarne episodes | "" er- | episodes | "" €r- | episodes under- | opjgodes | one recorded Tg recorded Te- recorded To- recorded re ported ported ported ported Total’ ~==--- 1,833 12 150 26 646 14 1,023 9 Self-respondent--- 1,092 9 57 28 380 11 643 6 Sample person is spouse~-=-=-===== 275 15 6 29 78 16 191 14 Sample person is child-=========~ 386 16 81 23 161 16 142 13 Sample person is other relative-- 78 26 4 (*) 27 35 47 20 Sample person is unrelated------- 2 (*) 2 (*) 0 0 0 0 vars are 13 episodes from interview reports and \4 from nospital records for whicn the lengtn of stay was unknown. Totals add to 1,587 and |,819. 22 versal is due to a difference of three episodes in the proxy child group which happen to have large weights associated with them. A similar pattern is observed for each type of respondent. The better reporting of self-respond- ents is again apparent and is seen for all lengths of stay with the exception of stays of one day. In table 15 the general increase in underre- porting with age of the sample personis apparent. However, the introduction of the additional vari- able of length of stay shows an interesting pattern. In every age group the longer the stay the better the report. It is also apparent that the effect of length of stay on the accuracy of reporting is greater than the effect of age of the sample per- son. In the 2-4 and 5 days and over stays the low- est underreporting occurs at ages 18-34 years. This is the group with the highest number of de- liveries. As other tables show, most delivery cases appear in the categories of 2-4 days'stays and 5 days and over. Accordingly, these cells show the smallest amount of underreporting. Since length of stay is related to age, con- trolling for length of stay should make the age effect more pronounced. For the 1 day and the 2-4 days' stays the episodes of the younger per- sons were reported better than for older persons. This tendency was not found in stays of 5 days or longer. Table 16 shows information based on the age of the respondent. The pattern is similar to that seen in table 15. The longer the stay the better the respondent reported the episode. Again it appears that the pattern of underreporting can be under- stood in terms of the interaction of three factors: age, length of stay, and better reports of deliv- eries, To strengthen the idea that reporting im- proves with increasing lengths of stay, it can be seen in tables 17 and 18 that almost without ex- ception the trend is consistent with the hypothesis. No matter how close or distant the relationship of the sample person to the respondent, reporting improves with increasing length of stay. The same is generally true for each family income group (table 19). In the $10,000 or more group the re- porting is better for 1-day than for 2-4 days' stays. Table 19. Percent underreporting of hospital episodes by family income and length of stay shown in hospital records, excluding overreports Length of stay (in days) All stays 1 2-4 5+ Family income Number Percent Number Percent Nunber Percent Number Percent episodes | Under- episodes | under- episodes under- episodes | under- recorded re-~ recorded Tes recorded I8- recorded ros ported ported ported ported Totall------ 1,833 12 150 26 646 14 1,023 9 Under $2,000-==--- 154 19 11 *) 47 19 93 12 $2,000-3,999---=-- 301 17 19 35 102 13 178 18 $4,000-6,999-==~~- 750 10 73 20 267 12 404 7 $7,000-9,999--=-=~ 272 11 20 26 107 17 143 5 $10,0004=====mamun 248 9 22 9 84 15 142 6 Unknown-=====m===== 108 16 5 (*) 39 12 63 1.7 There are 13 episodes frominterview reports and 14 from nospital records for whicn the length of stay was unknown. lotals add to 1,587 and 1,819. 23 UNDERREPORTING BY DIAGNOSTIC AND OPERATION CLASSES The analysis presented in this section is based on diagnoses and operations recorded in hospital records. In many cases more than one diagnosis was recorded and occasionally more than one operative procedure was listed. Thus one hospital discharge might list a fractured arm, diabetes, and a heart condition. Or the description of an operation might include a hysterectomy and an appendectomy. Since all of the sample hospitals were par- ticipants in the Professional Activity Study, whose function it is to make analyses of hospital rec- ords, all hospitals were instructed to list first the diagnosis leading immediately to the hospitaliza- tion. Thus the first diagnosis or operation listed in the discharge record was regarded as the pri- mary cause of the hospitalization. Only the first diagnosis or operation listed was used in the analysis presented in this report. For all classes of respondents nonsurgical cases were more seriously underreported than surgical ones (table 20). Within the surgical groups, deliveries were reported more completely than other surgery. Differences in reporting for proxy children and proxy adults were not great, The most important difference was between re- porting by self-respondents and proxy respond- ents. The most seriously underreported episodes were mental and personality disorders (table 21). The probability is that this diagnosis was suffi- ciently embarrassing that the respondent avoided discussing it by not reporting the episode.The next poorest reports were for pre- and post-natal con- ditions, benign and unspecified neoplasms, and "all other conditions." The pre- and post-natal conditions can be accounted for by adifferent fac- tor. Many of these were false labor in which the woman was in the hospital for a short time and then discharged, usually to return soon for the de- livery. To her this short stay was probably either considered as part of the main hospitalization or was so minor as not to be considered an actual hospitalization. Thus the hypotheses are thatthere are two reasons for underreporting, one because of embarrassment or threat and the other because of the minor nature of the episode. The best reporting was for arthritis and dis- eases of the gallbladder, which were reported per - fectly, and deliveries, Arthritis and gallbladder conditions are both serious, in that hospital stays are usually long, and the disorder is physically threatening in terms of discomfort. Yet neither condition is embarrassing. Delivery dates are easily remembered since they are associated with 24 a child's birthday and usually the event is re- called as a happy occasion. Table 22 shows the percent of underreporting of hospital episodes by type of operation. The highest rate of underreporting for any surgical group was 18 percent, while there were several diagnostic groups which showed a higher propor- tion of underreported episodes. This may reflect, again, the importance of the seriousness of the event. The operations for which the episodes were reported best were deliveries, gallbladder, appen- dectomies, and repair of hernias. Those with the highest underreporting of episodes were eye op- erations, hysterectomies, and operations on the bladder and on the intestines, It appears that the important factors in de- termining whether or not a hospitalization will be reported are, first of all, its seriousness and, second, how embarrassing or threatening it is. Tables 23-27 show underreporting of episodes by diagnostic ratings. It seems clear thatreport- ing varies with the amount of threat represented by the diagnosis (table 23). The largest difference is between the most threatening and the somewhat threatening groups. This relationship holds for proxy adults and chil- dren as well as for self-respondents. The relative drop in underreporting by degree of threatisless for children than for the other groups. The underreporting of episodes for the most threatening group rises with the age of the sample person, except for the youngest ages (table 24). However, the somewhat threatening and the not threatening groups do not follow a consistent pat- tern. This seems to indicate that the reaction to threat is independent of the age of the sample per- son. Why underreporting in the middle category drops for the ages 55 years and over is not clear. It may reflect the fact that the middle ratings were in part a miscellaneous grouping which did not readily fit into one of the other two groups. Under- reporting among persons in the most threatening group was also highest for each respondent age group (table 25). Since all of the initial household interviews were done by female interviewers, it was possible to investigate whether or not reporting com- pleteness differs for various combinations of the sex of the respondent, sample person, and inter- viewer. Is a female respondent less reluctant than a male respondent to talk about an embarrassing type of diagnosis? Do females report certain types of diagnoses better for female sample persons than they do for males? Table 26 indicates that female respondents report better for males than for fe- males and that male respondents report better for females, regardless of the diagnosis. The evidence is not conclusive, however, since the number of cases in each group is small. Table 20. Percent underreporting of hospital episodes by type of treatment shown in hospital records and type of respondent, excluding overreports Type of respondent All respondents Proxy respondent for: Self-respondent Type of treatment Children under 18 Adults Percent Percent Number cides Number Devoe Niiiber Percent | Number under- $Siseles re- episodes Heer” episodes 9niepe Splsojes re- recorde = - recorde ported | recorded ported recorded ported ported Total-=----- 1,833 12 349 16 392 18 1,092 9 Deliverieg=---=---=- 359 3 1 (*) 25 8 333 3 Other surgical---- 813 12 199 12 200 16 414 9 Nonsurgical------- 661 19 149 22 167 22 345 16 Table 21. Percent underreporting of hospital episodes by diagnostic categories, excluding overreports Number episodes Percent under- Disgnostic wategory recorded reported Total-mm mmm mmm ee ee ee een 1,833 12 Infective and parasitic diseases-==---=--ccoommaooo___ 19 22 Malignant neoplasms ===--======cmommomo 59 11 Benign and unspecified neoplasms----=-m-cmmmomaeooooo. 60 23 Allergic, endocrine, and metabolic disorders---------- 57 12 Mental and personality disorders---------amommeocooooo 25 32 Intracranial lesions==-====-cmme oom ____ 12 9 Diseases of nervous system and sense organs===-=-==----- 85 17 Heart diseases=========m cm eeeeeeeeel 61 13 Hemorrhoids == === moomoo eee __ 23 12 Other circulatory diseases=-=-====meocomoooomoo ooo ___ 36 17 Upper respiratory condition§======m-moccommmomooaaaooo 127 14 Other respiratory conditiong==-====ecmmommmmomooo__._ 70 12 Ulcer of stomach and duodenum--==-==ccacommmooaoo____ 31 19 Appendicitig-=m-m mmm em __ 29 5 Hernian === === mo eee ee eo 54 4 Diseases of the gallbladder=========cocoommmmoaoaooooo 44 0 Other digestive system conditions-----=-==ccmmmmcaeooo- 116 16 Female breast and genital disorders---------ecmeeeeco- 96 21 Other genitourinary conditiong====-===-ecoooemmoaooao_. 66 11 Deliveries-=-=m mmm eo 359 3 Pre- and post-natal conditions=-=====ecoeooomoooo_._ 89 23 Diseases of the skin=-=--=-aoommmmme 29 19 Arthritis ooo mm cm eee ee ee 13 0 Other musculoskeletal disorder§==-=-=--cmcomcmooce____ 65 9 Fractures and dislocations--===m=-meeomoooomaoooooo___ 50 17 Other current injuries--=--=mmmmeeeme ooo. 66 13 Observation only====== comme 9 (*) All other conditiong======cmcmmmom oe. 70 23 No diagnosis=== m= === momo. 13 0 25 Table 22. Percent underreporting of hospital episodes by type of operation, excluding overreports Number Percent Type of operation episodes under- recorded reported Total-==========memmmee-eseece--e--=-==-=-========= 1,833 12 Operations on the brain and skull----=-=-=----=-=-======-= 2 (*) Eye operations----=-============sss=-ee-====e=sooooo=—-- 31 18 Varicose veins---======-=c-cs-m-e-mem--cs-----oo---===-= 8 (*) Tonsillectomy and adenoidectomy=-=-=--========-=-==========< 94 12 Operations on the Stomach~====s=veesuummmwe nme oom nmnm= 8 (*) Appendectomieg-===========--====--==-==------=sssosssosoos 26 3 Repair of hernias--===m======-ceesmmee~—-=enemn—=——— 49 3 Operations on the intestines--=--=-----=-==========-=-==-=°-=" 23 16 Operations for hemorrhoids--=-==-===-=--=-==-============-=" 18 10 Operations on the gallbladder--=-=--====--==-==========="= 38 0 Operations on the Kidneyge~====ss=scwmmmmme~~oecosn=s 5 (%) Operations on the bladder-----====-===----c==-===-==-=-===-==" 35 16 Operations on the male genital system=-----==-=-=====-==< 31 12 Hysterectomi@s--====mm=m==-====smee=e=—======coo==——==== 20 17 Other female genital operations-< (11) PLACE SEEN H.I.P. Coding Section GROUP ADMINISTRATIVE OFFICE “o.emvaE OFFICE | | or. Hose TOTALS => LM _ SUMMARY MADE BY "1. MEDICAL CENTER ~ | OP. OTHER APPROVED BY 2. Home. of 20 = DATE 3 - HOSPITAL CAES. “Ee " DATE FWD. TO HIP T Figure 1. period became available for comparison with mor - bidity and hospitalization information obtained through National Health Survey interviews! A special study on a subsample of the inter- viewed population was carried out in the course of this project. This consisted of physicians inter- viewing the H.I.P. physicians who had rendered the services for specified conditions to persons in the subsample. These physician interviews attempted to relate the comparisons of physician and survey reports both to the clinical chart and to the ex- pressed judgments of the physician in response to the interview questionnaire. Analysis of the re- sults of these physician interviews is nota part of this report, but reference is made to some of the findings pertinent to the record comparison. 1only minor modifications of the NHS schedule were adopted for this survey: identification of physicians named as attending illness or as “usual doctor,” and obtaining information on occupation and on pregnancy history. The pertinent questions used appear as Appendix I. 2 This study was carried out in only one geo- graphical area, New York City, in a populationre- ceiving medical care ina special setting. The field operation, although done by the Regional Office of the Bureau of the Census, which is responsible for the regular NHS interviews inthe area, differed in some details from the normal enumeration pro- cedure. Another limitation, discussed below in some detail, is introduced by the nature of the criterion document for physician reports, partic- ularly with respect to definitions of "chronic" conditions. It is accordingly not possible to use the results here presented as measures of ''un- derreporting' in the total National Health Survey, or any part of it. The findings are, however, use- ful in any attempt to clarify major problems in the interpretation of morbidity data derived from household interviews. A pilot comparison of data inferred from the H.I.P. physician reports (Med 10's) and household interview information collected on a sample of the H.I.P. population in 1952 served both to emphasize the need for a study with the actual NHS interview and to demonstrate the feasibility of the proce- dures necessary to process such data. The 1952 interviews were carried out by a private research agency under contract to "The Committee for the Special Research Project in the Health Insurance Plan of Greater New York."! The questionnaire had been designed to elicit information about health and medical care in the 8-week period pre- STUDY SETTING AND The Health Insurance Plan of Greater New York is a prepaid comprehensive medical care plan, organized on a group practice basis. En- rollees in the Plan are entitled to receive medical care from family physicians and specialists in the office, home, or hospital. Coverage is for preven- tive and diagnostic medical services and for treat- ment of illness. There are no waiting periods for service or exclusions from enrollment because of preexisting conditions, and no limitations on the number of services or duration of medical care. Medical services are provided by physicians as- sociated with 31 medical groups distributed throughout New York City and Nassau County, and one medical group in Columbia County. On June 30, 1957, shortly after the startof the second decade of H.I.P.'s operations, and the date for selection of the sample for this study, there were 513,052 persons enrolled in the Plan. About 67 percent were employees of New York City and their dependents, 19 percent were insured through health and welfare plans established by labor groups, 7 percent were persons who had converted to individual policies, and the remaining 7 per- cent came from a variety of small employment groups and housing projects. EnrollmentinH.I.P. is on a group basis, the usual requirement being that at least 75 percent of the eligibles enroll. Contracts with these groups ordinarily provide for coverage of the employee, spouse, and dependent children under 18 years of age. A typeof contract providing coverage only for the employee, under- taken by a number of union health and welfare plans, accounted for 7.2 percent of the enrollment on the specified date. The independent record source in this study consists of the basic reporting document which H.I.P. physicians are required to submit to the central office in the normal course of the opera- tions of the Plan. The entries on a single line of 1{ealth and Medical Care in New York City, A Report by The Committee for the Special Research Project in the Health Insurance Plan of Greater New York, Harvard University Press, Cambridge, Mass., 1957. ceding interview, about the existence of a selected list of chronic conditions, and about hospitalization in the calendar year 1951. When the interview data were compared with the H.I.P. physician reports for the 8-week period preceding date of interview, it was found that only 42 percent of the conditions inferred from the physician reports were corre- spondingly reported by the household respondents. MEDICAL RECORDS this "Med 10" form (fig. 1) represent a single face- to-face contact between a patient and a physician. H.I.P. physicians also make entries in clinical charts, so that medical records relating toH.IL.P. enrollees exist in the files of the H.I.P, medical group centers and, frequently, in the private of- fices of H.I.P, practicing physicians. The question may reasonably be raised why the Med 10 was chosen as the criterion record source for this study rather than the clinical record, since the Med 10 does not provide detailed information which one might expect to find in a clinical rec- ord. The Med 10 gives nomedical history, no eval- uation of symptoms or disability, and no weighing of differential diagnoses. Diagnostic entities must be inferred from the Med 10's by examining the terminology used by physicians in the context of the dates and places of service and the identifica- tion of the physician-specialties of those rendering the services. Error can be introduced in the nu- merical identification of the patient either at the source, where the Med 10 is originally filled, or in the course of processing to collate all services for a given individual. It was nevertheless more feasible to use the Med 10 as the basic record source rather than the clinical chart. Because of the wide geographic dispersal of the medical groups, and the variety of methods of recordkeep- ing, great difficulties would have been encountered in an effort to examine all physician entries for a given individual. In addition, administrative diffi- culties would have been raised through the de- mands on group centers and private offices of phy- sicians to make records available. Because the Med 10's have served as the source of data for a number of studies made in the Division of Research and Statistics of H.I.P., evi- dence has accumulated on their reliability. All observations made in the paston the validity of the Med 10, both with respect to the clinical records existing in the medical groups and physicians’ of- fices and with respect to more general considera- tions, have indicated that the Med 10 is a reliable document for the statistical purposes for which it has been used. A systematic study to validate the 3 Med 10's with respect to the clinical records was part of an earlier research project, which ex- amined the enrollment, morbidity, and utilization experience of a 10 percent sample of H.I.P. en- rollees over the years 1948-1951. Here it was found that the total number of services reported on the Med 10's was slightly greater than that found in the clinical records, with the largestdis- crepancy produced by failure to enter home visits on the clinical charts. Inferences on number of cases of specified diagnostic categories were sub- stantially the same from both record sources, ex- cept that more respiratory conditions were in- ferred from Med 10 reporting (a reflection of more complete entering of home visits), and more symptomatic complaints were inferred from the clinical records (possibly a reflection of the re- quirement that the physician enter a diagnosis, definite or tentative, on the Med 10). Later studies made in the Division of Re- search and Statistics have substantiated infer- ences made from the Med 10's on hospitalization, on the prevalence of cancer, and on the number of deliveries in H.I.P. When estimates of prevalence of specific diagnostic entities made from the Med 10's are compared with similar data from other sources, generally good agreement is found.’ The interviews with H.I.P, physicians, noted above as a special development of this study, were directed toward illuminating the circumstances under which respondents in the household inter- view either reported or failed to report conditions inferred from the Med 10's. They thereby furnished information relating the inference made from the Med 10's to the knowledge that the physician, aided by his clinical chart, had regarding the patient's illnesses. The results of these interviews with 280 H.I.P. physicians, about 600 conditions in 341 pa- tients, again provide strong evidence of therelia- bility of the Med 10's. Over 98 percentof the diag- noses inferred from the Med 10's appeared at some time in the clinical record, and only 4 percent of the inferred conditions had in fact been ruled out by the physicians after the entry had been made on the Med 10. METHODOLOGY The Sample The sample for this study provided about 1,400 interviewed families. Sampling was restricted to subscribers and their covered dependents who were enrolled in H.I.P. on June 30, 1957 under family coverage and affiliated with a medical group. Persons who were not continuously insured for the 12 months preceding date of household in- terview were excluded from the sample. In order to increase the volume of chronic conditions for study the sample was stratified as follows: Stratum 1: families in whicn one or more personshad received one or more Med 10-reported services re- lated to a selected list of condi- tions during the 6-month period April 1 through September 30, 1957, Stratum 2: families in which no person had received such services in the stated period. The selected list of conditions consisted of medi- cal terminology which approximated the conditions on the NHS interview checklists (fig. 2). Stratum 1 was sampled roughly three times as intensively as stratum 2. The stratum 1 families submitted to the Regional Office of the Bureau of the Census constituted roughly 2.0 percent of all H.I.P. sub- scribers in this category, while the stratum 2 families were approximately 0.7 percent of sub- scribers as defined.” The tables presented inthis report are all based on frequencies inflated to the extent necessary to give each element equal weight (referred to as the weighted sample), in order to present a representative picture of the segment of the H.I.P. population defined above. Each study family submitted to Census for in- terviewing was identified by a 5-digit serial num- ber which was a translation of the H.I.P. 8-digit certificate number. Addresses obtained from the H.I.P. enrollment files had been confirmed by a preinterview mailing of a piece of educational ma- terial (provided by the American Heart Associa- tion), with arrangements made with the Post Office for notification on changes of address. 1Densen, P.M.; Balamuth, E.; and Deardorff, N.R.: Medical Care Plans as a Source of Morbidity Data. The Prevalence of Illness and Associated Volume of Service. The Milbank Memorial Fund Quarter ly 38: 48, January 1960. 2precise details of universal delineation and sampling fractions on which computation of weights was based are not presented in this report because of space limitations but are available and may be obtained upon request. Appendix III contains a few illustra- tive sampling errors. Card A NATIONAL HEALTH SURVEY Check List of Chronic Conditions 1. Asthma 14. Stomach ulcer + Any allergy 15. Any other chronic stomach trouble 3. Tuberculosis 16. Kidney stones or other kidney trouble 4. Chronic bronchitis 17. Arthritis or rheumatism 5. Repeated attacks of sinus trouble 18, Prostate trouble 6. Rheumatic fever 19. Diabetes 7. Hardening of the arteries 20. Thyroid trouble or goiter 8. High blood pressure 21. Epilepsy or convulsion of any kind 9. Heart trouble 22. Mental or nervous trouble 10. Stroke 23. Repeated trouble with back or spine 11. Trouble with varicose veins 24, Tumor or cancer 12. Hemorrhoids or piles 25. Chronic skin trouble 13. Gallbladder or liver trouble 26. Hernia or rupture Card 3 NATIONAL HEALTH SURVEY Check List of Impairments 1. Deafness or serious trouble with hearing 2. Serious trouble with seeing, even with glasses 3. Condition present since birth, such as cleft palate or club foot 4. Stammering or other trouble with speech 5. Missing fingers, hand, or arm 6. Missing toes, foot, or leg 7. Cerebral palsy 8. Paralysis of any kind 9. Any permanent stiffness or deformity of the foot or leg, fingers, arm or back Figure 2. National Health Survey Check Lists of chronic conditions. The Field Operation Interviewing of sample families was carried out by the Regional Office of the Bureau of the Census over the 9-week period from May 2 through July 6, 1958. The interview questionnaire (selected sections appear as Appendix I) was sub- stantially the same as the schedule in use at the time by the National Health Survey. Several items on the regular questionnaire not readily amenable to record check in the context of this study were omitted from the study questionnaire, while others necessary for the record check—such as the name of attending physician—were added. The enumerators were for the most part not the regular NHS interviewers employed in the area, but were specially employed for this study. Interviewers attended the standard Census inter- viewer training course as modified for the study. They were supervised by the regular Census re- gional supervisory personnel, and a reinterview rate of 20 percent was scheduled. In order to meet the time requirements of the field operation and to provide the basis for dealing with the nonresponse, a subsample was drawn by the Census Bureau in the final stages of the enu- meration. This subsample, 105 of 309 uncom- pleted households, was composed of families who had moved from the original H.I.P. address, other noninterviews, and unassigned questionnaires. Completed interviews were obtained on 3,937 H.L.P. enrollees insured continuously under family contract for the full 12 months preceding inter- view and distributed in 1,388 households. The weighted count, adjusting for stratification of the sample and nonresponse, was 6,609 individuals in 2,488 households. General Outline of Data Processing The completed interviews collected by the Census Bureau were sent to Washington and sub- jected to the usual NHS processing— editing, cod- ing, and punching of the standard decks of cards. Both the processed schedules and the decks of punch cards were then forwarded to H.I.P, Here the first task was to identify individuals in the study both on the NHS schedules and in the H.I.P. sample file. A "Demographic File'' was created by combining selected data from three sources: (1) the Persons File routinely punched by the National Health Survey, (2) the H.I.P, sample file of insured persons, and (3) the NHS schedules. This file be- came the source of data on personal characteris- tics established on household interview and on in- surance information for all decks of cards used in the study tabulations. All Med 10 reports on interviewed persons for the period April 1, 1957 through June 30, 1958 were processed to produce a listing of services to each individual in chronological order. These list- ings, which included the physician's terminology for diagnosis, abbreviated in accordance with a standard glossary, were edited to eliminate all identifiable error. Services in the 12 months pre- ceding interview (''study year'') and those in the 2 weeks specified in the NHS schedule (ending on the Sunday preceding interview) were marked off for coding. Figure 3 gives a specimen of the listing as prepared for coding for two persons in the study. Information on morbidity, medical care, and hos- pitalization experience as reflected in the Med 10 reports was coded. The coded Med 10 listing was then examined in conjunction with the matching NHS interview, and correspondence in household interview reporting of medical conditions and hos- pitalizations was coded. Conditions and hospitali- zations reported on interview which did not cor- respond with data inferred from the Med 10's were also entered from the interview schedules. Additional data on hospitalizations were obtained from individual hospitals and from the files of As- sociated Hospital Service of New York (AHS) (Blue Cross) and coded. Abbreviated Med-10 Patient ternivalony Place | pate "| airtn Con- | certificate wedicat | of 2 of identi- datad tract AURBP group hysician's original terminology, MED 10 1-59172' 5/202 service® | goryice | fication" number SMITH MARY TOOTH ABSC R JAW 2 05 05 7 | 2310 01 | 5107 13 4000 | 8967493 304 Tooth abscess, right jaw SMITH MARY INDIG POSS 63 DIS I 07 30 7 2310 01 51 07 13 4000 8967493 304 Indigestion, possible gallbladder disease SMITH MARY CHR CH 8 CHLITH 1 08 027 | 2310 01 | 5107 13 4000 | 8967493 304 Chronic cnolecystitis and cholelithiasis SMITH MARY X GB SER | 08 077 | 1736 11 | 5107 13 4000 | 8967493 304 sallbladder series SMITH MARY GAST COMPL REF TO X 31 8 3A | 08 237 | 1319 50 | 5107 13 4000 | 8967493 304 Sastric complaints, refer to X-ray, Gl and barium SMITH MARY X BA EN | 08 30 7 | 1736 11 | 51 07 13 4000 | 8967493 304 3arium enema SMITH MARY X GI SER 1 09 117] 1735 11 | 51 07 13 4000 | 8967493 304 Gl series SMITH MARY CHLITH 8 CHR CH | 09 27 7 | 2310 01 | 51 07 13 4000 | 8967493 304 Cholelithiasis and chronic cholecystitis SMITH MARY X CST I Oy 14 8 1736 11 51 G7 13 4000 8967493 304 Chest X-ray oo smn MARY PX NEG | 05 12.8 | 2310 01 | 5107 13 4000 | 8967493 304 Physical exam. negative ks SMITH MARY HYAL 1 05 25 8 | 2310 01 51 07 13 4000 8367493 304 Myalgia 1-41663' 5/132 sTupy |JONES JOE COR INSUF | 06 05 7 | 1u8w 01 [11 oy 91 4405 | 8900639 102 Coronary insufficiency YEAR JONES JOE EKG 1 06 14 7 2559 01 oy 91 4405 8900639 102 EKG JONES JOE COR INSUF | 07 177 | usu 01 [11 Gy 91 4405 | 8900539 102 Coronary insufficiency JONES JOE PAROX FIBRILLATION 2 01 03 8 14s4 01 1 Gy 91 4405 8900639 102 Paroxysmal fibrillation 2 JONES JOE MYOC DAM 2 01098 [1484 01 [11 04 31 4405 | 8900639 102 Myocardial damage Wk su ONES JOE HYDRO | 05 03 8 1484 OI 11 Oy 91 Yuus 3900639 102 ilydrocele JONES JOE EXG 1 05 15 8 1081 01 104 91 4405 8900639 102 EKG 'Stratum and serial numoer Zpate of nousenold interview 3Place of service: | - office: 2 - home “First 4 digits identify individual 40: last 2 digits identify pnysician specialty: 01 - family p. sician 11 - radiologist: 50 - Seneral surgeon SFirst 2 digits are sex and family status: 51 - female, spouse of suoscriver; || - male subscriber Last 4 digits are month and year of pirtn NOTE: All data shown are exactly as they appear on the listing except for fictitious patients' names and certificate numbers. Figure 3. Specimen of coding listing. Coding: General Considerations and Classification of ISC-PHS Codes The central interest of this study was to com- pare morbidity from chronic conditions as in- ferred from a criterion source—physician re- ports—with that reported by respondents in the NHS interview. The essence of the proposed com- parison can be summarized: Is the diagnosis of an H.I.P. physician, as es- tablished through his reporting on the Med 10's regarding a specified medical condition, reflected in any way by the reporting of the re- spondent in the NHS interview? Secondarily, if this diagnosis is reflected in the interview, how closely does the condition as reported correspond to the physician's diagnosis? In order tounderstand the meaning of answers to these questions obtained from processing data from the two sources a number of considerations must be carefully weighed. The mostimportantof these are the circumstances under which a re- spondent report on a given condition might be ex- pected, and conceptual differences in classifica- tion of morbidity from the two sources. The design of the NHS schedule is such thata report on inter- view of any condition, chronic or not, which pro- duced symptoms or the need for medication or treatment in the two weeks preceding interview could be elicited through the battery of questions about these two weeks (Questions 11-14, Appen- dix I). But if no symptoms were present during the 6 two weeks, and there was no related therapy or medication, reports of conditions would have to be elicited either through Question 15, which probes for conditions that have been present '"a long time," through the checklist questions (16 and 17), or through the question on hospitalization in the year preceding interview (25). The physician reports onthe Med 10's furnish no information on date of onset. Itis therefore not possible to define which conditions inferred from physician reports are to be considered chronic in terms of the duration of the symptoms or diag- noses. NHS practice in classifying interview-re- ported conditions is to consider a condition chron- ic either if it appears on the checklists of chronic conditions and impairments (fig. 2), or if the onset of the condition is stated by the respondent to be three months or more before the date of interview (except for pregnancies). But in the case of condi- tions inferred from the physician reports it was necessary to define chronicity solely on the basis of the terminology used. Physicians on the re- search staffs of both NHS and H.I.P. came to agreement on a list of all ISC-PHS codes which were to be accepted as descriptive of conditions which, in the clinical experience of physicians, could be chronic. It should be noted that the di- rection of judgment was to include the maximum number of conditions in the "possibly chronic" list, and a minimum number in the 'nonchronic" list. The diagnoses expressed by these ''possibly chronic" codes which were to be inferred from H.L.P. physician reports for the year preceding household interview were grouped into the follow- ing classes: Class 1: those which are covered by NHS terminology for the checklists with- out any qualifications introduced by modifying adjectives. Class 2: those which might be suggested by checklist terminology, but thereare qualifications arising for the most part from the use of modifying ad- jectives (''repeated," "chronic," etc.). Class 3: those which would not in any obvious way be suggested by checklist termi- nology, but which had been judged ""chronic'" or "possibly chronic' on the basis of the clinical experience of physicians. These three "classes of condition'' constitute a major axis of analysis inthe study. A priori they present varying probabilities of eliciting re- sponses with the NHS schedule, not only between the classes, but within a given class. Class 1, for example, or ''checklist unqualified," includes mainly conditions in connection with which misun- derstanding on the part of the respondent is un- likely—such as diabetes, asthma, high blood pressure, and heart trouble. On the other hand, because of item 9 on Card B (any permanent stiff- ness or deformity of the footor leg, fingers, arm, or back), NHS impairment codes for "specified de- formity of limbs, trunk, or back' were classified as Class 1 conditions. Flatfoot thereby became a Class 1 condition, but it is worth pondering what proportion of respondents would think of flatfoot as a ''deformity." Similarly, item 3 on Card B specifies "condition present since birth, such as cleft palate or clubfoot." For this reason all con- genital malformations were classified as Class 1. But it again remains a question whether, for ex- ample, such conditions as extroversion of the bladder or cryptorchism, coded as congenital malformations according to ISC rules, would be suggested to a respondent by this probe. Conditions were assigned to Class 2, or ""checklist qualified," usually on the basis of ad- jectives in the checklist terminology which might produce different subjective responses among re- spondents. Examples are "repeated trouble with back or spine," "any other chronic stomach trou- ble." Other qualifications could arise from the classification of a disease as a checklist condition because of a reasonable inference about an im- pairment that would be produced by the disease diagnosed by the physician, for example, glaucoma assumed to produce "serious trouble with seeing." Still other qualifications could stem from the dif- ferential meaning which conditions assignable to the same ISC code could have for laymen and for physicians: for example, a physician-diagnosed "'fibrositis," although codable to the same ISC code as 'rheumatism," is not necessarily the kind of condition suggested to the layman by the term "rheumatism'' which appears on the checklist. The conditions included in Class 3 ("non- checklist") are those which were judged by the NHS and H.I.P. physicians to be ''possibly chronic for which no obvious probe appears on NHS cards A and B. Reports of these conditions could there- fore be theoretically expected on interview only if (1) they produced symptoms, et cetera, in the two weeks preceding interview, (2) the respondent was reminded of them on the basis of Question 15 (con- ditions present for "a long time"), or (3) they had produced a hospitalization during the 12 months preceding interview. This class includes a wide variety of conditions, some of them unquestionably chronic (multiple sclerosis, peripheral vascular disease), but others actually ailments delimited in time—acute conditions—which the respondents should not have reported in response to any of the questioning during the interview (for example, an acute bursitis experienced several months before interview, with no residual symptoms). It is clear that the selection of conditions to be considered "possibly chronic' and the grouping of these conditions into the three defined classes contain many arbitrary elements. The classifica- tion nevertheless provided a useful analytical tool and may serve to suggest more refined designs for future studies of this nature.! Because of the potentially wide variety of con- ditions included within each class, another axis of classification of conditions was introduced. This combined the "class of condition" concept with two other variables: the number of physician services in the study year and the time spread over which services for the condition were ren- dered during the year. Four grades of condition were defined as follows: Grade I: Identical with Class 1, checklist without qualification. Grade II: Class 2 conditions, checklist with qualification, for which more than one service was rendered in the study year and for which there was an interval of more than one month between the first and last related service, Grade III: Class 3 conditions, nonchecklist, with the same stipulations as to volume and spread of related serv- ices as for Grade II. 1The detailed categories included in each of the three classes of conditions discussed above, and those considered nonchronic for purposes of this study, may be obtained upon request. Grade IV: Conditions assignable either to Class 2 or Class 3 (checklist with qualification or nonchecklist) with either only one related service in the study year or a time spread of one month or less between the first and last related service. Coding: Specific Procedures All possibly chronic conditions inferred from Med 10's—coding from listing of Med 10 serv- ices.—Each possibly chronic condition, as defined above, which was inferred from the listing of Med 10 services rendered in the year preceding inter - view of a person in the study, was characterized by a 4-digit code. Selection of this code followed current (July 1, 1958) NHS coding practice as closely as possible, using the 1955 Revision of the International Statistical Classification (ISC) as modified by the Public Health Service (PHS). In inferring conditions from the listing of Med 10 services, a not infrequent problem was to decide how many different ISC-PHS codes were required to describe the total morbidity experience. Here the principle followed was to lean in the direction of the greatest economy in the number of ''condi- tions" to which code numbers were assigned. In those cases where terminology assignable to dif- ferent codes appeared on the listing, and it was reasonable to conclude, in the context of the list- ing, that such diverse terminology applied to the same set of complaints, the choice of code was determined by examining the relative date whena term was used and the specialty of the physician who used the term. Diagnoses made later intime— that is, in the natural course of establishing a definitive diagnosis—were preferred over those made earlier in time, and diagnoses made by specialists were preferred over those made by family physicians. Within the framework of these two considerations, a more fundamental diagnosis was preferred to a symptom consistent with the diagnosis. All diagnoses which remained tentative at the date of interview were identified as such in the coding. For each possibly chronic condition coded in this way to the ISC-PHS code, additional coding specified the class of condition and an indication of the volume of related Med 10 services in the study year, of the time spread between the first and last services for the condition in the study year, and of the time interval between the last service for the condition and the date of the house- hold interview. Coding of correspondence of survey data with possibly chronic conditions inferred from Med 10's.—A determination was now made of whether any condition was reported on interview to corre- spond with each possibly chronic! condition in- ferred from the Med 10's. This decision was made from parallel examination of the interview sched- ule and the coded Med 10 listing, and, wherever a survey-reported condition was judged in corre- spondence, the type of match was also noted. de- fined as follows: Type 1 match: A condition is reported on household interview (HHI) which was coded by NHS to the same Recode No. 1? category as the H.I,P.-coded Med 10 services, or the Recode No. 1 codes from the two sources would have been the same had it not been for arbitrary coding decisions made for the H.IL.P. data.’ Type 2 match: A condition is reported on household interview which was coded by NHS to a different Recode No. 1 category but to the same Recode No. 3 cate- gory as the H.I.P.-coded Med 10 services.* Type 3 match: The NHS schedule contains no report coded by NHS to the same Recode No. 3 category as the H.I.P.-coded Med 10 serv- ices, but there is a description in the schedule of a condition or symptom which is consist- ent with or could be associated with the diagnosis inferred from the Med 10's. It is to be noted that the criteria for match made it possible to judge a survey-reported condition not coded chronic by NHS to be in correspondence with a “possibly chronic” condition inferred from the Med 10’s. Of the total survey-reported conditions matched to those inferred from the Med 10’s, 14 percent fell into this cate- gory--7 percent having been coded “nonchronic” by NHS and 7 per- cent not. having been designated either “chronic” or “nonchronic.” The last group, not designated by NHS, consisted almost entirely of survey-reported hospitalizations, matched to conditions inferred from the Med 10’s, but not entered as illness reports on Table I of the interview schedule. 2Recode No. 1--278 titles defining selected ISC-PHS codes and groups of such codes. 3In all instances where NHS coding practice differentiates be- tween codes to be selected on the basis of stated date of onset, it was impossible to pattern H.I.P coding practice in this way. For purposes of coding the Med 10 services the “chronic” code was usually preferred over the “nonchronic” code. If the condition was reported on household interview and coded by NHS to the nonchronic code because of onset within the 3 months preceding household in- terview it was nevertheless still considered a Type 1 match. All possible instances where such discrepancies could arise had been organized into a special chart to facilitate coding. 4Recode No. 3--43 categories with specified Recode No. 1 inclu- sions. Nothing is reported on survey which could correspond in any way with the condition inferred from the Med 10's. To make the determination on type of match defined above the entire schedule was examined. The procedure was to start with the conditions in the Schedule's Table I which had been coded by NHS and to proceed to examine all other entries on the schedule related to the given individual— hospitalizations, enumerator notes, et cetera. The listing of Med 10 services was referred to in order for the coder to have clearly in mind the termi- nology which had been used by physicians. Once a condition reported on interview was judged to be in correspondence with a condition inferred from the Med 10's, selected data relating to the condition were transcribed from the inter- view schedule to the code sheet. These included the question number which produced the household in- terview report, related medical care and disabil- ity, reported date of onset of the condition, the NHS code for type of condition (chronic, nonchron- ic), date of last doctor consultation, and H.I.P. status of last doctor consulted, as previously coded on the schedule. Provision was made for noting the number of Med 10 conditions to which a given survey-reported condition was matched. In summary, it should be noted that the cur- rent analysis of survey-reported morbidity in comparison with that inferred from physician-re- porting is directed primarily toward ascertaining whether the respondent told the enumerator any- thing which can reasonably be judged to correspond with the diagnostic entity inferred from the physi- cian's Med 10 entries. In terms of the coding con- ventions adopted, this can be expressed as the total proportion of conditions inferred from the Med 10's matched by survey-reported conditions in any way—i.e., the sum of all Types 1, 2, and 3 matches. Of secondary interest is an examination of the way in which what the respondent told the enumerator, as coded by NHS, corresponds to the data coded from physician reports. Coding of nonchronic conditions.—Med 10 services codable to ISC-PHS codes which had been classified as nonchronic in the review of codes by staff physicians were assigned codes only if the dates of these services fell within the two-week period ending on the Sunday preceding the house- hold interview. Provision was made for classify- ing these conditions into five broad diagnostic groups and for indicating the number and place of related services in the two-week period. The in- terview schedules were examined in conjunction with the coded Med 10 listing to determine whether any condition or symptom was noted by the enumer - ator which was consistent with or could be asso- ciated with the nonchronic condition inferred from No match: the Med 10's. No attempt was made to distinguish different types of match for these conditions. If a corresponding condition was reported on inter- view, selected information about the condition was transcribed from the interview schedule, as de- scribed above for chronic conditions. Coding of morbidity and medical care ex- pressed as experience of persons. —The coding so far described was designed to make possible com- parison of conditions reported on survey with those inferred from the criterion source. The issues in this study may also be posed in terms of the experience of persons, For example: what propor - tion of persons with medical services for one or more chronic conditions reported at least one of these conditions on interview? To make possible analysis along these lines a card summarizing morbidity experience during the study year was coded for each person. Information coded into this card included number of chronic conditions in- ferred from the Med 10's and number corre- spondingly reported on interview, and data on the number of H.I.P. physician services received during the study year and during the two-week period. An indication of whether a physician con- tact during the two-week period had been reported on survey was entered from examination of the in- terview schedule. Coding of survey-reported conditions not in correspondence with conditions inferred from Med 10's.—Information on all survey-reported condi- tions was coded from the interview schedules, in- cluding the H.I.P. status of the physician reported by the respondent to have last attended the condi- tion. The nature of the study materials restricts the investigation to determining the extent to which conditions medically attended in a defined setting were correspondingly reported to the enumerators on survey. This one-way process is concerned with "underreporting." There is, of course, a gen- eral interest in the reverse process—the extent to which respondents report chronic conditions to be present which through some independent source could be shown never tohave been medically diag- nosed. Since some H.I.P, members seek medical care outside the Plan, conditions reported on sur- vey for which no corresponding condition was in- ferred from the Med 10's cannot be assumed to be overreports. Although the data do not provide the basis for an analysis of overreporting to parallel that of underreporting, certain characteristics of the unmatched survey-reported chronic conditions are described in the findings of this report. Coding of hospitalization experience.—Since the place of each service reported onthe Med 10's is noted by the H.I.P. physician (home, office, or hospital), episodes of hospitalization for given conditions can be inferred from the listing of Med 10 services. An opportunity was thus provided to examine the extent to which hospitalizations es- tablished from an independent record source were reported by the respondents on interview. The Med 10's do not, however, provide exact dates of admission and discharge, since the physician re- ports only the dates on which he sees the patient in the hospital. For this reason the study design provided for a follow-back to hospital records and, in some cases, to the records of the Associated Hospital Service, in order to obtain accurate data on duration of stay. In all instances where the hos- pitalization inferred from the Med 10's had been correspondingly reported on interview, the hos- pital queried was that named by the respondent. Where there was failure to report a hospitaliza- tion on survey, either the name of the hospital to be addressed was obtained from the physician's clinical record, or all hospitals in which the given physician had admitting privileges were queried. Of all episodes of hospitalization inferred from Med 10 services reported rendered in the hospital, the hospital follow-back confirmed 95 percent as involving at least one night in the hos- pital. The hospitalizations thus confirmed were used (a) to examine underreporting of the fact of hospitalization on household interview, and (b) to study accuracy in reporting the duration of stay on interview in comparison with the primaryrec- ord source. Hospitalizations reported on interview which had not been inferred from the Med 10's were also checked against an independent record source—the hospital named by the respondent or Associated Hospital Service files. Through this follow-back, part of the problem of overreporting of hospitalization was studied, and an additional set of records became available to examine ac- curacy in reporting duration of hospital stay. Inquiries were directed to a total of 112hos- pitals (97 in New York City and 15 outside the City), and, with intensive follow-up by mail and telephone, responses were obtained from all of them. Episodes for which the hospital was unable to locate an admission were further cleared with AHS files. Data pertaining to all hospitalizations in- ferred from the Med 10's which had been con- firmed by the hospital or AHS records as involv- ing at least one night in a hospital in the study year were coded. Diagnosis was coded from the listing of Med 10 services, duration of stay and date of admission from the hospital or AHS rec- ord. The interview schedule was examined for re- ports of hospitalizations corresponding to those inferred from the Med 10's. A hospitalizationre- ported on survey was considered in correspond- ence with that inferred from the Med 10's if the respondent's stated reason for hospitalization was judged consistent with the Med 10-inferred diag- nosis producing the hospitalization. Interview-re- ported data relating to the matched hospitalization were transcribed (survey-reported diagnosis, date of admission, and duration of stay). All sur- vey-reported hospitalizations not matched to epi- sodes inferred from the Med 10's were also coded. Data on diagnosis, date of admission, and duration of stay were coded from the hospital record, and interview-reported data were again transcribed. Survey-reported hospitalizations for which there was failure to confirm the fact of hospitalization from the independent record source were identi- fied. GENERAL CHARACTERISTICS OF THE STUDY POPULATION The study population is a representative sam- ple of H.I.P. subscribers and their covered de- pendents insured under family contract for the full 12 months preceding household interview in the spring of 1958. This is essentially a population of New York City residents in the labor force and their dependents. Some of the more important de- mographic characteristics are shown in tables 1 and 2, while table 3 gives the proportion of persons with specified characteristics for whom one or more possibly chronic conditions were inferred from the Med 10's. The H.I.P. population contains a slightly high- er proportion of males than of females (table 1), whereas the reverse is true for New York City as a whole. As would be expected in a working popu- lation, the H.I.P. subscribers and their families 10 are younger than the total City population. H.I.P. is substantially deficient in persons aged 65 and over (3 percent compared with 9 percent found in the 1957 special census of New York City), and has a higher proportion of children under 15 (31 per- cent compared with 23 percent for the City). The deficiency in aged persons is especially marked in the case of women—only 1 percent of the H.I.P. population are women aged 65 or older, compared with 5 percent for New York City. The distribution of H.I.P.'s population by race is practically identical with that found for New York City as a whole in the special 1957 census, where 12.6 percent of the population was classified as nonwhite, Slightly more than one fourthofthe H.I.P. en- rollees are members of families in which the fam- ily head had completed at least one year of college; less than one fourth are in families whose head had not completed one or more years of high school (table 2).! H.I.P. families have a somewhat higher in- come than families in New York City as a whole. More than half of H.I.P. enrollees (56 percent) are in families whose income is between $5,000-9,999. Except for enrollees aged 65 and over, the propor - tion of H.I.P. members in families with incomes under $4,000 is roughly 10 percent (table 2). But 22 percent of the aged enrollees are in this cate- gory. If the subscribers, rather than total en- rollees, are distributed by family income, com- parison is possible with the special census of New York City carried out in 1957, which provided a distribution of families by 1956 income.’ Of all H.I.P. families reporting income only 13 percent reported incomes under $4,000 inthe current NHS survey (1957-1958 income), whereas 27 percent of the New York City families fell into this category for 1956. Approximately the same proportion of H.L.P. and New York City families fell into the $4, 000-4999 bracket, while the proportion of H.L.P. families was greater in all income classes of $5,000 and over. The percentage of enrollees in white families in which the family head had completed more than 12 years of schooling was twice that found in the nonwhite group (29 percent compared with 14 per- cent), Roughly five times as many nonwhite H.I.P, members were in families with incomes under $4,000 (37 percent compared with 7 percent). Practically all males aged 25-64 in the study population were working in the year preceding in- terview, while slightly over athird of those 65 and over were retired. Almost one third of the women aged 25-44 were in the labor force, and this pro- portion rose to some 45 percent in women aged 45 to 64. Roughly one fourth of all H.I.P, members are in families where the subscriber's occupation is classified as professional or managerial. One third of all enrollees are in families whose sub- scriber is classified as a professional, manage- rial, clerical, or sales worker. Less than 5 per- cent are in families headed by laborers, while 18 percent are in families for whom the subscriber is a fireman or policeman. FINDINGS Correspondence in Household Interview Reporting of Possibly Chronic Conditions Inferred From Med 10's General considerations.—A total of 4,648 possibly chronic conditions was inferred from H.I.P. physician reports on the interviewed pop- ulation for the year preceding interview, Some 40 percent of these were Class 1 conditions (check- list unqualified), slightly more than one fourth were Class 2 (checklist qualified), while one third were Class 3 (nonchecklist). The proportion of conditions in these classes correspondingly re- ported by the respondents on interview in no case reached half of those inferred from theMed 10's— 44 percent of the Class 1 conditions were re- ported, 28 percent of Class 2, and 20 percent of Class 3 conditions. Un the 1952 household survey which compared a sample of the H.I.P. population with a representative sample of New York City, 27 percent of the H.I.P. enrollees aged 25 or over had completed more than 12 years of schooling, compared with 14 percent of the comparable N.Y.C. group. “Health and Medical Care in New York City.” Harvard University Press, 1957. 2 Characteristics of the Population of New York State, 1956 and 1957, Interdepartmental Committee on Low Incomes, Bulletin No. 1 (Part 1), October 1958. Seventy-five conditions for which the diagnosis inferred from the Med 10’s was designated as tentative or questionable as of the date of interview are not included in the analysis. Discrepancies of such magnitude immediately raise a question about the possibility that factors unrelated to the accuracy of reporting may have been responsible. Within the setting of this study, such factors might be, for example: 1. Conditions inferred from the Med 10's may in fact have been errors, or diagnoses no longer maintained by the physicians for their patients. 2. A substantial proportion of the conditions classified as "possibly chronic" may in reality have represented nonchronic con- ditions for which the NHS schedule was not designed to elicit reports. 3. The deficiency in survey-reported condi- tions corresponding to those inferred from the Med 10's may reflect a poor quality of enumeration in this survey, at- tributable to the relative lack of training or experience of the interviewers. The reliability of the Med 10 relative to the physician's clinical record has already been dis- cussed briefly. It will be recalled that the inter- views of H.I.P. physicians carried out as adevel- opment of this study showed that there was no mention in the physician's record of the Med 10- inferred condition in less than 2 percent of the conditions about which physicians were ques- tioned. The interviewed physicians characterized 86 percent of the conditions about which they were asked as definite diagnoses, 8 percent as tentative, and 4 percent as ruled out after being considered tentative. Accordingly, the finding that only a low proportion of the conditions inferred from the Med 10's were reported on interview cannot be attributed to Med 10 error or unreliability. Questions related to the chronicity of the conditions under discussion emphasize the dis- tinct character of the two universes of conditions being compared. On the one hand there is the uni- verse of conditions inferred from physician re- ports on the Med 10's in terms which, in the judgment of NHS and H.I.P. staff physicians, are likely to represent largely chronic—i.e., long- standing, or continuously present, or recurring— disease. On the other hand, there is the universe of conditions reported on interview inresponse to a questionnaire which probes about (1) conditions which produced symptoms or the need for medi- cation or medical care in the two weeks preceding interview; (2) conditions which have been present "for a long time'' or which produced a hospitaliza- tion in the year before interview; and (3) a spe- cific list of conditions and impairments. In comparing information from the two sources an assumption that complete reporting by respondents would duplicate the universe of condi- tions inferred from physician reports can never be made. There are nevertheless two considerations relevant to the problem raised here. The first is that the results of the interviews of H.I.P. physi- cians, while not definitive, suggest that relatively few of the conditions unreported by respondents were considered 'monchronic'' by the patients’ physicians.! The second is that some control on chronicity is provided throughout the analysis by examining the data in relation to axes of classifi- cation which tend to segregate nonchronic condi- tions included—such as, class of condition, num- ber of related physician services, and specific diagnosis. The reinterview program carried out by the Bureau of the Census provided some data for IThe interviewed physician’s judgment on chronicity was not systematically obtained for every condition subject to interview. In- terviewed physicians volunteered opinions on chronicity in connec- tion with the two following questions on the physician-interview schedule: Question 1: As of (date of household interview) was the diag- nosis definite, tentative, or ruled out? Question 10: As you know, our chief interest in this study is in discovering, if we can, reasons why people might not report chronic conditions in an interview re- garding health. That’s what my questions have been directed to. In such an interview the patient did not report.... Considering the things we have talked about and any other reasons that might sug- gest themselves to you, what do you think is the possible explanation for the patient’s not men tioning these to the interviewer? examining the quality of the enumeration as a fac- tor influencing the proportion of Med 10-inferred conditions reported on interview. The data are fragmentary, since only 80 reinterviews were available on persons for whom one or more possi- bly chronic conditions had been inferred from the Med 10's. One can only state that the order of magnitude of improvement obtained after recon- ciliation of these reinterviews with the original interviews would not account for any substantial part of the underreporting found. The evidence on hand therefore supports the general conclusion that it is not possible to ex- plain the failure of respondents to report such a large proportion of the conditions inferred from the Med 10's as the effect of recognizable factors unrelated to reporting. Class of condition,—The gradient foundinre- spondent reporting of conditions in the three classes (44 percent for Class 1, 28 percent for Class 2, and 20 percent for Class 3) is consist- ently maintained no matter what other variableis simultaneously examined. Whenever one com- pares the percentage of Class 1 and Class 3 con- ditions reported, the proportion for unqualified checklist conditions is one and a half totwo and a half that for nonchecklist conditions. This is true for a large number of variables with which class of condition has been crossed: volume of service; interval between first dnd last related service; in- terval from last service to household interview; relationship to respondent, and sex and place of birth of respondent; age, sex, and race; education of family head, family income and family size; whether or not permission toreview medical rec- ords was granted. (Many of these may be examined in detail in tables 4 through 16.) The proportion of Class 2 conditions (checklist with qualification) re- ported in relation to these other variablesisusu- ally somewhere between that for Class 1 and Class 3, although on occasion there is little difference between the percentages for qualified checklist conditions and those considered to be nonchecklist. There is no question that in this study the re- spondents reported most completely a group of medically attended conditions about which the in- terviewer had asked specifically and unequivo- cally. But it does not follow from this that the production of an interview report to correspond with a medically attended condition inferred from a physician record source is a simple matter of including specific terminology for that condition on an interview checklist, The design of the NHS schedule is such that the first probes are for conditions which produced symptoms, or disability, or the need for medica- tion or treatment within the two weeks preceding interview (Questions 11-14). Next, the respondents are questioned about any ailments or conditions that have lasted "a long time" (Question 15), and then they are queried with the checklists (Ques- tions 16 and 17). Thus itmay be argued that chron- ic illnesses which produce symptoms, or for which medication is regularly taken, or which have been present for relatively long periods of time, have a better chance of being elicited from the respondent than conditions without these charac- teristics even before the checklists have been mentioned by the interviewer. It is worth examining here the proportion of the survey-reported conditions in correspondence with Med 10-inferred conditions for which the in- terview report was in response to the checklist questions, shown in table 20. One third of all in- terview-reported conditions which were matched to conditions inferred from the Med 10's were mentioned by the respondents in response to the checklist questions. Although this percentage dif- fered somewhat in the three classes of condition (Class 1, 32; Class 2, 41; Class 3, 29), the more important observation is that the use of the check- lists improved correspondence in reporting for all classes of condition. This is true even for those conditions (Class 3) where no obvious stimulus to response can be identified on the checklists. It is possible to estimate what the percent of con- ditions correspondingly reported on interview would have been in the absence of checklist ques- tions by assuming that all conditions mentioned in response to the checklist questions would then have remained unreported: Percent ‘f Med 10-inferred conditions correspondingly reported Without Observed | questions 16 and 17 Class l-==-=======- 44.1 29.8 Class 2-=--=-==-==-= 27.6 16.4 Class 3-=-========= 20.4 14.6 Clearly, the conditions grouped in Class 1— checklist without qualification—would have been reported roughly twice as efficiently as those in the other classes even without the use of any checklist question. The general significance of the checklists is that without these probes the pro- portion of conditions correspondingly reported on interview would have been substantially lower in all three classes of conditions. In summary, it is important to bear in mind throughout this report that, although the three classes of condition were set up originally on an a priori basis related to the terminology of the NHS checklists, the gradient found in correspond- ence of reporting cannot be simply interpreted as a reflection of the efficacy of checklist terminol- ogy. Rather, we are dealing with three groups of’ conditions which are reported with varying effi- ciency for many complex reasons. Class 1 is heavily weighted with very specific diagnostic en- tities, predominantly chronic, many of which have a high probability of producing disability or symp- tomatic complaints and of thereby producing the need for more intense medical care. Succeeding sections in this report present data specifically related to these issues. Volume of related physician services.— There is a strong relationship between the num- ber of physician services rendered for a speci- fied condition in the year preceding interview and the probability of that condition's being reported by the respondent. Among Class 1 conditions, 3 out of 10 (27 percent) seen only once by a physi- cian were reported on survey, compared with 9 out of 10 (88 percent) which had required 10 or more services, Intermediate levels of service show intermediate rates for correspondence in reporting (table 4). Similar relationships prevail for conditions in Classes 2 and 3. The data in table 4 also clarify one of the issues discussed above—the nature of Class 1 conditions as compared with those in the other classes, It will be noted that almost three fourths of the conditions with 10 or more Med 10 serv- ices are Class 1 conditions. About a fourth of the Class 1 conditions received at least 5 medical services during the year, but only a tenth of the conditions in the other two classes had this many services. Variability in correspondence cannot, however, be explained by a comparatively simple factor such as number of physician services. Within each level of service category there is evidence of the gradient in percent of conditions correspondingly reported between the three classes of condition. Time interval: date of last service to date of interview.—Conditions last attended by a physi- cian shortly before the date of interview are bet- ter reported by the respondents than those last at- tended at earlier times in the study year. Two thirds of Class 1 conditions seen by a physician within the two weeks preceding interview were re- ported in contrast with one third of these condi- tions last seen by a doctor four months or more before the date of interview (table 5). For condi- tions in Classes 2 and 3 there was an even larger disparity between the proportion of recently at- tended conditions reported and that for conditions seen at the longer interval from the interview date. It is worth noting that, although the NHS schedule could be expected to elicit reports ofall conditions medically attended in the two weeks I3 preceding interview, without regard to the ques- tion of chronicity, a very substantial proportion of such Med 10-inferred conditions remained unre- ported—almost one third of the Class 1 condi- tions, one half of the Class 2 conditions, and 58 percent of the Class 3 conditions. Table 5 shows a strong inverse relationship between correspondence in reporting and the du- ration of the time interval between last physician service for a condition and date of household in- terview. This, however, is greatly affected by dif- ferential distributions in volume of service. As would be expected by chance alone, conditions with comparatively high volumes of service are more likely to have services on dates close to the household interview date. Actually, almost halfof the conditions for which the last service had been rendered within the two weeks preceding inter- view had received five or more services during the study year. The corresponding proportion for conditions last attended four months or longer from the date of interview was one tenth, while that for the intermediate time interval was one fifth. The gradient in reporting between the three classes of condition is maintained when these classes are examined for specific time intervals between last physician service and household in- terview. Checklist unqualified (Class 1) conditions are reported more than twice as efficiently as nonchecklist (Class 3) conditions in both of the longer interval classes, and more than one and a half times as efficiently when the last serviceoc- curred in the two weeks preceding interview. Time spread from first to last related serv- ice.—In planning the variables to be examined in this study there was interest in any axis of clas- sification which might throw light on the difficult question of defining chronicity. It was reasoned that a condition requiring physician services over a relatively long period was more likely to be "chronic" than one for which the physician serv- ices were concentrated in a brief time span. For this reason a dichotomy was provided for distin- guishing conditions for which the interval from first to last service was one month or less from those with an interval of more than one month, This dichotomy is, of course, inapplicable to con- ditions for which only one physician service was noted during the study year—49 percent of all Med 10-inferred possibly chronic conditions. Conditions with physicians’ services spread over a period of more than one month constituted 61 percent of all conditions with more than one service. These conditions were somewhat better reported on interview than those for which all services rendered were concentrated within a period of one month or less (table 6). But there is, of course, a relationship between the time span over which services are rendered and the actual number of services. Almost one fourth (24 percent) of the conditions with services spread over more than one month had received 10 or more related services in the study year, while only 2 percent of the conditions with concentra- tions of services within one month or less had re- ceived this many services. While there is attimes more complete reporting of conditions for which services are spread over the longer interval within a given volume of service level (table 6), the total number of services related to a condition seems to be a much more important factor in re- lation to the reporting on household interview. In summary it can be said that the arbitrarily defined variable here discussed is not viewed as one of great intrinsic importance, as studied in the above context. In this study its chief useful- ness has been to serve as one factor in the defi- nition of ''grade of condition." Grade of condition.— Reference has already been made to the introduction of the concept of "grade of condition' as a means of combining class of condition with both volume of services and the time spread for services in the study year. In this classification Grade I is identical with Class 1 (checklist without qualification). Grade II selects from Class 2 (checklist with qualification) those conditions which received more than one service in the study year spread over a period of more than one month, Conditions are selected from Class 3 (nonchecklist) on the same basis to constitute Grade III, and Grade IV is made up of the conditions from Classes 2 and 3 which either had only one service or, if more than one, a time spread between first and last service of one month or less. The distribution of all possibly chronic conditions inferred from the Med 10's by class and grade is shown in the following: Schematic relationships between grade and class of condition Class 1 (checklist without qualification)--- Class 2 (checklist with qualification) Grade 1 Grade II Grade III Grade 1V It is seen from examination of the first line of table 7 that the over-all effectof this reclassi- fication on correspondence in reporting isto wipe out the differential previously noted between Classes 1 and 2. The percentage of Grade II con- ditions correspondingly reported is 42 compared with 44 percent for Grade I (or Class 1), whereas only 28 percent of all Class 2 conditions as a whole are reported. There is an improvement of similar magnitude in the reporting of Grade III conditions in comparison with that previously noted for Class 3 conditions (31 percent compared with 20 percent). When correspondence in reporting conditions in Grades I, II, and III is examined for specific levels of service, it is apparent that the effect described above is largely a reflection of the re- moval of conditions with only one service from Classes 2 and 3 in order to produce Grades II and III. The gradient in correspondence between Grades 1, II, and III is very similar to that for Classes 1, 2, and 3 (table 4) at all applicable service levels. With regard to Grade IV, the im- portant point to bear in mind is that71 percent of the conditions in this category received only one service in the study year. Respondent status,— Self-respondents report Med 10-inferred chronic conditions on interview to a somewhat greater extent than proxy respond- ents (table 4). But while there is some improve- ment in reporting when a person responds for himself, the change is not major. It would appear .that the low over-all correspondence in reporting cannot be attributed to the fact that over half of the possibly chronic conditions were in persons for whom a relative responded on household in- terview. Despite the moderate over-all influence of respondent status there are a number of relation- ships worth noting. When conditions are examined in specific volume of servicé categories, the largest differential in reporting between self-re- spondents and others is found for conditions with only one physician service in the study year. It is of interest too that the largest difference between self-respondents and others is found for Class 2 conditions. These are conditions which had been classified as ''checklist with qualification' be- cause of the use of modifying terminology which might produce different subjective reactions in different respondents. And these conditions were reported almost twice as well by self-respond- ents as by proxy respondents—36 compared with 20 percent. Class 1 conditions were reported only slightly more completely for self-respondents than for others (48 compared with 41 percent, but the difference is statistically significant!), and there was no differential at all withrespectto re- spondent status in the case of nonchecklist condi- tions (Class 3). Differentials in correspondence by respond- ent status for conditions in the four specified grades (table 7) are consistent with these findings. Here the effect of better self-respondent report- ing of Class 2 conditions with only one service is to produce the largest difference between self-re- spondents and others for Grade IV conditions. Relationship to respondent.—Classification of the conditions in relatives of respondents by actual relationship to the respondent (table 8) lotatements on statistical significance throughout the text refer to a probability level of 0.05: 15 shows that the differential favoring self-respond- ents over others is largely a matter of poorer re- porting of conditions in children. Nonchecklist conditions (Class 3) are reported very similarly in self-respondents, spouses, and children (21, 21, and 18 percent, respectively). Class 1 condi- tions (checklist without qualification) are re- ported equally well in spouses and in self-re- spondents, but somewhat less well in children. The largest differential, in Class 2 conditions, shows these to have been reported twice as well in self-respondents as in children, and one and a half times as well in self-respondents as in spouses. When reporting by relationship to re- spondent is examined for the four grades of con- dition, the differential between self-respondents and children is reflected both in Grade II and Grade IV conditions (table 9). It is clear that limiting the NHS interviewing to persons responding for themselves would have produced no impressive increase in the percent of conditions inferred from the Med 10's which were correspondingly reported on interview. The differential found between self-respondents and children cannot, moreover, be interpreted as a simple reflection of the status of the respondent, since the chronic conditions to be found in children are apt to have a very different diagnostic distri- bution from those found in adults. Age and sex.—Chronic conditions inferred from the Med 10's in mature and older adults are reported more completely than in children and young persons. This is true to agreater or lesser extent for all three classes (table 10), It is also true for all four grades of conditions. The gradient with age is more pronounced for conditions other than those classified as checklist unqualified (Class 1). Conditions in Class 2 (checklist with qualifi- cation) are better reported on interview in females than in males, the direction of the difference favoring females at every age except 15-24, This difference is of statistical significance for the age groups 25-44 and under 15 years. The re- verse applies to Class 1 conditions, which are somewhat better reported in males at every age, although the difference is statistically significant only for ages 15-24, There is little difference be- tween the sexes in the case of nonchecklist con- ditions (Class 3). Respondent status and age.—Examination of completeness of reporting of the three classes of condition by age crossed with respondent status (table 11) emphasizes the earlier observation on the relative efficiency of reporting by self-re- spondents of Class 2 conditions (checklist with qualification). The differential in favor of self- respondents is found at every age, and holds also for both male and female self-respondents. Inter - 16 pretation of the relatively more accurate report- ing on Class 2 conditions in females, noted above, must be made bearing in mind the large prepon- derance of females responding for themselves as well as for others. Race.—There is no consistent difference in the reporting of chronic conditions between whites and nonwhites. Conditions in white children are better reported in this study than those in non- white, but the difference lacks statistical signifi- cance. Education.—There is no consistent pattern in percent of possibly chronic conditions reported when examined in relation to the years of school - ing completed by the family head (table 12). Nor is there any apparent relationship between corre- spondence in interview reporting and the educa- tional attainment of the individual with the condi- tion (table 13). Family income.—No regular relationship be- tween family income and the percent of chronic conditions reported on interview is found whichis applicable to all three classes of condition (table 14). Class 2 conditions are, however, distinctly better reported in families of the lowest income class (under $4,000) than among all other fami- lies, while Class 1 conditions are somewhat bet- ter reported in the lowest income families. This finding is a reflection of the much better reporting of Class 2 conditions and the somewhat better re- porting of Class 1 conditions by self-respondents in the lowest income class, and of the relatively higher proportion of conditions reported on by self rather than proxy respondents in comparison with all other income classes. The families with income under $4,000 are known to contain ahigh- er proportion of aged persons than those with higher incomes. The relatively high proportion of persons responding for themselves also suggests that the conditions in this income class are more heavily weighted with disabling illness. Family size.—The number of family mem- bers covered on the H.I.P. policy as ofthe date of drawing the original sample (June 30, 1957) is examined in relation to correspondence in re- porting in table 15. Since H.I.P. enrollees, es- pecially young unmarried adults, can be living with their parents and other family members who are not members of H.I.P. and therefore not in the study population, the "number of H.I.P.-cov- ered persons'' cannot be strictly equated to fam- ily size. With this qualification in mind, it is noted that the data show a decline in correspond- ence in reporting nonchecklist (Class 3) condi- tions with increasing number of H.I.P.-covered persons, from two persons on. That this decline is not solely due to the larger proportion of chil- dren in the larger family units is seen from the fact that the decline occurs for self-respondents as well as for others. It is therefore possible that the need to respond for a large number of indi- viduals reduced the probability that a condition inferred from the Med 10's for which there was no specific probe in the schedule would be re- ported to the interviewer. But perhaps of greater interest in this connection is the negative finding: the fact that there is no loss in percent of condi- tions correspondingly reported for any conditions appearing on the checklists (Classes 1 and 2) with increasing number of persons in the household. Permission to review medical records.—In planning for the interviews with H.I.P. physicians already mentioned it was recognized that these in- terviews would have to be restricted to patients who authorized the release of findings from their medical records to the National Health Survey Each respondent was accordingly asked to sign an authorization form. Permission for review of medical records was granted for 89 percentofthe persons interviewed, and 87 percent of the pos- sibly chronic conditions inferred from the Med 10's were in these persons. Completeness of re- porting was examined in relation to whether the authorization had been signed, since it was rea- soned that refusal to grant such permissionmight be positively correlated with a generally unco-op- erative attitude on the part of the respondent. The data do suggest (table 16) slightly higher corre- spondence in reporting in persons for whom the requested permission was granted. This improve- ment applies almost wholly to nonchecklist condi- tions (Class 3), where the rates were 21 percent for those with permission granted, and 14 percent for those with refusals, but the difference is not statistically significant. One must conclude that any influence on completeness of reporting which this variable may reflect is of a comparatively minor nature. Diagnosis.— The focus in presentation of the findings up to this point has been on relatively broad classifications of disease categories which had been designated ''possibly chronic' in the a priori review of all ISC-PHS codes by staff physi- cians described above. By using the broad group- ings (class or grade of condition) ithas been pos- sible to examine completeness of reporting inre- lation to aspects of the medical care received from H.I.P. physicians (number of services, dates of service) and in relation to demographic and other characteristics of the interviewed popula- tion. At the same time, the variation in the find- ings among the three different classes of condi- tion has served as a constant reminder of the in- fluence of the differing diagnostic contents of each class on the percent of conditions correspondingly reported on interview. It has also been pointed out that within each of the three classes there is a wide variety of inclusions, with varying distri- butions in the population in relation to age and sex. Some interest, therefore, attaches to analy- sis in terms of more specific disease entities. This interest centers both on more detailed exam- ination of the diagnostic content of each of the three classes of condition and on the findings for certain specific diagnoses of public health im- portance. The diagnostic data to be presented are for the most part organized into the categories of NHS's Recode No. 3! as modified for the study. While many of the Recode No. 3 categories are very specific entities defining a single 3- or 4- digit code according to the ISC (such as diabetes), others are still relatively broad groupings which may include a heterogeneous collection of dis- ease entities. For example, ''other diseases ofthe digestive system'' includes both ulcerative colitis, a serious chronic disease, and any symptoms re- ferable to the abdomen or gastrointestinal tract. The diagnostic tables are presented specifically by class of condition as well as by NHS Recode No. 3 category. When the inclusions in a single Recode No. 3 category fall into more thanone of the three classes of condition, the inclusions within each class are shown separately. Some of the frequen- cies which result are very low, and the general principle of not computing a percentage corre- spondence for a total of less than 15 conditions in- ferred from the Med 10's has been followed. Be- cause of the low frequencies in some of the diag- nostic categories shown, care must be exercised in interpreting some of the differences found. For example, the difference between correspondence in reporting asthma (269 cases, 76.2 percent cor- respondingly reported) and that for diabetes (60 cases, 61.7 percent) is not of statistical signifi- cance. But the difference between the figure for asthma and that for heart disease (162 cases, 60.5 percent) is statistically significant. In general, examination of correspondence in interview reporting of specific disease categories emphasizes once again the substantial number of possibly chronic conditions which remain unre- ported by respondents. There are only eight diag- nostic categories (Recode No. 3) for which more than half of the conditions inferred from the Med 10's were correspondingly reported on interview: INHS Recode No. 3, as modified for this study, consists of 43 selected categories of chronic conditions. Class Percent . of correspond- Diagnosis condi- ingly tion reported Asthma and hay fever------------ 4 76.2 Diseases of gall- bladder-----=----- 1 66.7 Bronchitis (chronic) 2 65.0 Diabetes mellitus- 2 61.7 Heart disease----- 1 60.5 Ulcer of stomach and duodenum----- 1 60.0 Back conditions--- 2 56.4 Hernia------------ i 54.4 In terms of the finest diagnostic breakdowns avail- able, hay fever was the best reported disease, with 79 percent of the cases inferred from the Med 10's correspondingly reported on interview. This is a very specific disease entity, with identical physician and layman terminology, for which itis possible that a good number of the persons so diagnosed were receiving desensitization treat- ment at a time not too far removed from the date of household interview, Correspondence in household interview re- porting of Med 10 conditions in detailed diagnostic terms is shown in table 17. The large variation in percent of conditions reported by respondents within a given NHS Recode No. 3 category assigned to a single class of condition is readily apparent. For example, 41 percent of the cases of benign neoplasm of the uterus and other female genital organs (predominantly fibroid uterus) were corre- spondingly reported, compared with only 13 per- cent of all other benign and unspecified neoplasms. Heart disease, cited above as one of the best re- ported categories, shows variation when examined by specific etiology, with the degenerative types better reported than either rheumatic heart dis- ease or ''other' heart disease. The distortions which can be introduced through the weighting of a specified NHS Recode No. 3 category in a given class of condition with a large number of one very poorly reported condi- tion are illustrated by examination of NHS Recode No. 3 category 30 (Other conditions of muscles, bones, and joints) in table 17. For this category as a whole it appears that conditions assigned to Class 1 are reported less completely than those in either of the other two classes (19 percent com- pared with 32 and 26). Examination of the detailed inclusions shows that this is due to the fact that more than half of the cases in this category as- 18 signed to Class 1 are cases of flatfoot, only 6 per- cent of which were correspondingly reported on interview. The other Class 1 conditions falling into this recode are reported to the same extent as the Class 2 conditions. In general, however, the gradient found in percent of conditions reported in each of the three classes, discussed earlier in this report, is again found for specific Recode No. 3 categories which contain conditions assigned to more than one class. For example, "arthritis," considered a Class 1 condition, and "rheumatism," assigned to Class 2, are combined in NHS Recode No. 3 cate- gory 28. Forty-eight percent of the cases of arth- ritis were correspondingly reported, but only 18 percent of the cases of rheumatism. Or, in the case of category 26, "other diseases of genitouri- nary system,’ 47 percent of the diseases of the kidney, ureter, and prostate (Class 1) are re- ported, compared with 18 percent of the variety of conditions in this grouping which were assigned to Class 3. Qualifications similar to those which have been expressed above about the interpretation of correspondence rates for specific categories of disease of course apply to all statistical consider- ations of morbidity, even in the finest possible groupings of disease entities. A case of coronary heart disease where the diagnosis rests solely on electrocardiographic evidence obtained in the course of routine examination is not to be equated with a hospitalizing myocardial infarction. The total complex of factors which may influence re- porting in the household interview cannot be lost sight of even in the most detailed diagnostic ap- proach to the data available. Ranking the diagnostic categories within each class of condition by level of correspondence in interview reporting (table 18) demonstrates the wide range in percent correspondingly reported to be found within each class of condition. Class 1 conditions range from a correspondence rate of 76 percent for asthma and hay fever to 4 percent for ill-defined mental and nervous trouble, Per- cent correspondingly reported for Class 2 condi- tions ranges from 65 to 0, while that for Class 3 conditions is from 34 to 0. More than half of the specific disease categories in Class 1 havecorre- spondence rates of 40 percent or more (11 out-of 20), while only 4 out of the 13 Class 2 categories, and none of the 14 Class 3 categories are reported this well on interview. Differentials-in completeness of reporting of specified disease categories by respondent status are presented in table 19. If the categories are classified into three groups defined with respect to the magnitude of the difference between the per - cent reported by self-respondents and others, itis seen that the rate for self-respondents is above that for relatives to the defined extentin 21 of the 32 categories. fined above (see Methodology), but may be briefly summarized here: Correspondence ratios by magnitude and class of condition Percent for relatives exceeds No difference of that for specified magnitude self-respondents by | between the two rates 25 percent or more Percent for self-respondents exceeds that for others by 25 percent or more CLASS 1 Diabetes mellitus Mental illness Heart disease Benign and unspecified neoplasms Arthritis Ulcer of stomach and Other allergies Asthma and hay fever Hypertension without | Hernia heart involvement Hemorrhoids Varicose veins duodenum Other conditions of muscles, etc. Ill-defined mental CLASS 2 Skin infections and dis- |[Infective and eases parasitic NEC Impairment of hearing Back conditions Other diseases of diges- tive system Other conditions of mus- cles, etc. Rheumatism Sinusitis Other diseases of genito- urinary system CLASS 3 Other diseases of respira-| Other diseases of Obesity tory system digestive system Diseases of eye Anemia Other conditions of | and ear NEC Diseases of brain, spinal muscles, etc. cord etc. Other diseases of genito- urinary system Headache and migraine The differential in favor of self-respondents with respect to Class 2 conditions, discussed earlier for the class as a whole, is seen to apply to all individual diagnostic categories within the class with the single exception of "infective and parasitic diseases NEC." Specificity of Match Between Conditions Correspondingly Reported on Interview and Those Inferred From Med 10's It has already been mentioned that, although the central interest of this study was an examina- tion of the extent to which respondents in the household interview reported anything in corre- spondence with conditions inferred from the Med 10's, there was a secondary interest inthe way in which these corresponding reports matched the conditions which had been inferred from the physi- cians' terminology. For this reason each interview report matched to a Med 10 condition was charac- terized by one of three types of match, defined in relation to the inclusions in NHS Recodes No. 1 and No. 3. The types of match are precisely de- Type 1 match: A condition is reported on household interview which falls into the same Recode No, 1 category as the H.L.P,- coded Med 10 services, Type 2 match: A condition is reported on household interview which falls into a different Recode No. 1 category but into the same Recode No, 3 category as the H.I.P.-coded Med 10 services, Type 3 match: The NHS schedule contains no report coded by NHS to the same Recode No, 3 category as the H.I.P.-coded Med 10 services, but there is a de- scription in the schedule of a condition or symptom which is consistent with or could be associated with the diagnosis inferred from the Med 10's, It must be recognized that the proportion of survey-reported matching conditions which fall into "Type 1'"is influenced not only by the specific- ity of the respondent in describing the nature of the condition, but also by the number of inclusions in the specified recode category as constituted by NHS. For example, NHS Recode No. 3 category 06 includes only diabetes mellitus, which is also a discrete Recode No. 1 category. Itis apparent that if the respondent reports anything to match this condition as inferred from the Med 10's, the probability of that report being a '"Type 2'' match is zero, since by definition the Recode No. 1 and No. 3 categories are identical. On the other hand, an NHS category with a wide range of inclusions, such as "other conditions of the muscles, bones, and joints' offers many chances for a survey-re- ported condition to be matched to the Med 10 con- dition in ways that would be designated '"Type 2" or "Type 3." The examination of ''type of match" for survey-reported conditions should therefore be viewed not as any absolute rating of the accu- racy of respondents, but rather as anindicationof the way in which respondent specificity affected the matching of survey-reported conditions tothe groupings of disease categories used by NHS in tabulating. Data pertinent to this question are presented in Appendix II. Survey-reported conditions which were matched to conditions of each class and specified diagnostic category inferred from the Med 10's are distributed by type of match in table A. It is seen that Class 1 conditions have a higher proportion of type 1 matches than those in Class 2 or 3. But within Class 1thereis great variation in this distribution: all matches for diabetes, hemorrhoids, and hernia are type 1 matches, ] while of the mental illness correspondingly re- ported on interview only 18 percent is couched by the respondent in terms assignable to the same Recode No. 1 category as the physician's termi- nology on the Med 10's. It is of some interest to examine the net effects of differences in physician and respondent terminology applied to chronic conditions reported on household interview. Table B of Appendix II presents the frequencies in each NHS Recode No. 3 category obtained from the two sources—H.L.P. physician and respondent—with the categories arranged in order of decreasing ratio of frequency from the respondent source to frequency from the physician source. Such a nettablecanonlygive an impression of the over-all effect of differences in terminology from the two sources. For example, if one views the relative preponderance of cases of mental illness (Recode No. 3 category 09) from the physician source together with the relative preponderance of cases of ill-defined mental and nervous trouble (Recode No. 3 category 10) from the respondent source, itis reasonable to conclude that the more precise terminology of physiciansis being replaced on interview by vague terminology which nevertheless refers to the same condition. To a certain extent this is true, but a more com- plete understanding of whathashappenedinthein- terview process can be obtained by case study of respondent and physician terminology in specific instances. ! While there are many ramifications demon- strated by the "matched' diagnoses from the two sources, it is possible in some cases to make some generalization about the differences arising in this study from application of a set of coding rules to respondent terminology on the one hand and to that of physicians on the other. For exam- ple, a wide variety of orthopedic conditions —char- acterized by physicians as osteomyelitis, Paget's disease, sacroiliac sprain, degenerative disc syndrome, or undiagnosed and referred to by phy- sicians simply as "pains in legs' or "'metatarsal- gia"—are matched in the household interview with lgor example, of the 55 cases of mental illness (Recode No.3. category 09) inferred from physician reports on the Med 10’s and matched by some respondent report, 45 were “type 3” matches. Ref- erence to the terminology used by the respondents shows that 13 of these 45 were reported on interview in terms codable to ill-defined mental and nervous trouble (Recode No.3 category 10). The remain- .ing 32 “type 3” matches in this category were reported by respond- ents with a variety of terminology codable to headache and mi- graine, hypertension, other diseases of the digestive system, meno- pausal disorders, other diseases of the genitourinary system, endo- crine disorders, et cetera. Conversely, for all conditions reported by respondents in terms codable to mental illness, it was found that in 7 instances these reports were judged to correspond to physician diagnoses of a variety of gastrointestinal conditions (duodenal ulcer, gallbladder disease, gastritis, mucous colitis, spastic colon), and in one instance to a physician’s diagnosis of contactdermatitis. 20 a respondent report of arthritis. Many specific dermatological diagnoses made by physicians (eczema, seborrheic dermatitis, contactdermati- tis, psoriasis, neurodermatitis, lichen simplex, alopecia) were matched on survey by a respondent report of a skin disorder stated to be due to allergy. An appreciable number of cases of ar- teriosclerotic and/or hypertensive heart disease were matched on interview by respondent reports, simply, of high blood pressure. Further examples of this sort can be adduced from study of the un- published detail. Duplication of match.—In establishing the criteria for judging a survey-reported condition in correspondence with a possibly chronic condition inferred from the Med 10's, there was no insist- ence in this study on a one-to-one match. One condition reported on interview was theoretically allowed to be matched to any number of conditions inferred from the Med 10's, provided that the sub- stantive criteria were fulfilled. Data onthe extent to which one condition reported on interview was matched to more than one condition inferred from the Med 10's are to be found in table A of Appen- dix II. Some 14 percent of all the matched Med 10 conditions were considered matched by survey- reported conditions which had also been matched to other Med 10-inferred conditions. This per- centage was lowest for Class 1 conditions and highest for Class 3 conditions. In assessing the importance of duplication of match it is necessary to have in mind the distri- bution of multiple diagnoses inferred from the Med 10's among the persons in the study. Of all per- sons sustaining a diagnosis of a possibly chronic condition inferred from Med 10 reporting during the study year, 62 percent had only one such con- dition, two conditions were inferredin 25 percent, and three or more conditions in 13 percent of these persons. There were 1,116 persons for whom more than one condition was inferred from the Med 10's; for 616 of these at least one condi- tion was correspondingly reported on household interview. Duplication of match was a factor in only 92 of these persons. Study of the content of the duplicated matches shows a wide diagnostic range and establishes that these matches are not attributable to any systematic error which could be characterized as 'overcoding'' from the Med 10 services, The Total of Chranic Conditions Reported on Household Interview This report has so far dealt with possibly chronic conditions inferred from physician re- ports and those conditions reported on household interview which were judged in correspondence with them. Although it is the relationship between these two frequencies that furnishes the main focus of the study, some interest also attaches to respondent-reported conditions—possibly chron- ic in accordance with the criteria applied to phy- sician reporting—which remained unmatched to any diagnoses inferred from the Med 10's. Such survey-reported conditions could theoretically be any of the following types: 1. Conditions for which no H.I,P, physician service was rendered in the study year and for which no H.IL.P. physician service in this period was reported on interview, a, Conditions reported on interview as not medically attended, or medical care not stated, b. Conditions reported on interview as last medically attended before the study year. c. Conditions reported on interview as last medically attended in study year by physician unaffiliated with H.LP, 2, Conditions for which no Med 10 service was reported in study year, but appear- ing on survey as last attended by H,I.P, physician in this time period, a. Errors (respondent or interviewer) in date of last physician service— i.e,, actually last seen by H.I.P, physician before study year. b. Errors (respondent, interviewer, or coder) in status of last physician seen—i.e,, condition actually at- tended by non-H.I.P, physician, c. Errors (respondent or interviewer) in substance—i.e., condition not present in this person, d. Med 10 underreports—i.e,, condi- tion actually seen by H.I.P. physi- cian in study year but Med 10's as listed for coding failed to reflect this fact. In contrast with the conditions reported onin- terview which remained unmatched to any inferred from the Med 10's, those survey-reported condi- tions judged to correspond to Med 10-inferred conditions were all presumably attended by H.I.P, physicians within the study year. One would there- fore expect that the two groups of conditions, matched and unmatched, would differ in important respects. The data support this cénclusion when comparison is made between the two groups along whatever lines are possible. Some of these differences are shownin tables 20-22. The source of the household interview re- port was a checklist question (Questions 16 and 17) for 56 percent of the conditions remaining unmatched to Med 10 conditions, compared with 33 percent of the matched reports (table 20). Con- versely, the battery of questions on the two-week period (Questions 11-14) produced the respondent reports for 37 percent of the matched conditions, but for only 20 percent of the unmatched. Both for Class 1 and Class 2 conditions a higher propor- tion of the unmatched survey reports were obtained in response to the checklist questions, and a lower proportion in response to the questioning on the two-week period. The chief differences between matched and unmatched Class 3 conditions were found in the proportion mentioned in response to the questions on hospitalization (15 percent of the matched conditions but none of the unmatched) and in the relatively larger number of unmatched reports produced by question 15 ("ailments or conditions that have continued for a long time"). Findings on medical care and disability for the matched and unmatched survey reports are consistent with those on the source of the survey report. It would seem reasonable that a higher proportion of nonmedically attended conditions would fail to produce symptoms or the need for medication in the two-week period, and, conse- quently, would be elicited on interview by the checklist questions.! A distinctly higher percent- age of the unmatched survey-reported conditions was in fact stated never to have been medically attended—12 percent, compared with 2 percent for the matched group (table 21). As would be ex- pected, many more of the unmatched conditions were reported to have been last seen by a physi- cian at some time before the study year —29 per - cent, compared with 6 percent for the matched group. Care by a non-H.I.P, physician was re- ported as the last medical contact for 16 percent of the unmatched conditions but for only 7 percent of the matched conditions reported on interview. A higher proportion of the matched survey re- ports had associated disability, bed disability, and time lost from work or school in the two weeks preceding interview than was the case for theun- matched survey-reported conditions (table 22). In summary, the findings on the general na- ture of the survey-reported possibly chronic con- ditions which were not in correspondence with any conditions inferred from the Med 10's are consist- ent with the theoretical description of the possible contents of this group of conditions. They are con- ditions less likely to be mentioned on interview in response to questioning about symptoms or medi- cation in the two-week period than the matched survey-reported conditions. A higher proportion of them are not medically attended at all, or last medically attended before the study year, or last medically attended by a non-H.I.P. doctor. A low- er proportion of them is associated with disability lrabulation of the soyrce of the household interview report sepa: rately for medically attended conditions and others was not avail- able. 21 or time lost from work or school. Consistent with these findings also are the differences seen in diagnostic distribution between the two groups of survey-reported conditions. Frequencies by in- dividual diagnosis, as reported by the respondent, are shown for the matched and unmatched sur- vey-reported conditions in table 23. Categories relatively more heavily represented in the group of matched conditions include neoplasms, dia- betes, and heart disease. The unmatched group has relatively more varicose veins, hemorrhoids, sinusitis, bronchitis, back conditions, deafness, and headache and migraine. In the main, the latter group contains conditions which are more likely either to be self-diagnosed or to remain unattend- ed by a physician for relatively long periods of time. The description of underreporting which is documented in this study by the percentage of survey-reported conditions in correspondence with possibly chronic conditions inferred from the Med 10's cannot be complemented by an equally direct examination of overreporting. The NHS schedule attempted to elicit respondent reports of conditions whether medically attended or not, and, if medically attended, whether by H.I.P. physi- cians or others, in the study year or before the study year. Under these circumstances, before a condition could be classified as an "overreport," it would have been necessary to check medical charts in H.I.P. for physician services at times preceding the study year and to have access to the records of physicians not associated with H.L.P. Such an investigation was beyond the scope of the present study. Relationship Between the Diagnostic Experience of Persons and the Percent of Med 10-Inferred Conditions Reported on Interview Correspondence between interview reporting and physician entries on the Med 10's has so far been expressed as the percent of inferred condi- tions with specified characteristics which were correspondingly reported on household interview. Some characteristics of the persons in whom the conditions were diagnosed have been considered in relation to these proportions—age, sex, education of family head, family income, respondent status, et cetera. But a number of questions may be posed which require analysis other than one re- stricted to qualifying the conditions by the char- acteristics of the persons in whom they are found. To what extent does the over-all low propor - tion of conditions correspondingly reportedonin- terview reflect the existence of a group of per- sons for whom there is failure to reportany con- 22 dition on interview? What is the relationship be- tween the number of conditions for which persons have received care and the proportion of persons for whom no condition is reported to the enumer- ator? Does the number of conditions medically attended during the year influence the proportion which is correspondingly reported on survey? Before dealing with these questions, it is worthwhile examining the distribution of the pos- sibly chronic conditions inferred from the Med 10's among the persons in the study. The 4,648 conditions were diagnosed in a total of 2,934 pa- tients, or 44 percent of the total number of inter - viewed H.1.P. members. Of these persons who had sustained a diagnosis by an H.I.P. physician of a possibly chronic condition during the study year, 62 percent had only one such condition, 25 percent had two, while three or more conditions were in- ferred in 13 percent. The persons with only one condition accounted for 39 percent of all the Med 10-inferred conditions; persons with two condi- tions contributed 32 percent, while those with three or more conditions accounted for 29 percent of the total conditions. Considerable light on the issues raised here is provided by the data in tables 24 and 25. Per- sons for whom there were no reports on house- hold interview of Med 10-inferred chronic condi- tions represent major segments of the total group with Med 10 conditions. The figures are 60 per- cent for all persons, 53 percent for self-respond- ents, and 64 percent for relatives of respondents. The need for examining this situation becomes even clearer when it is realized that 78 percent of all the unmatched Med 10-inferred conditions are attributable to these persons. To a considerable extent the percentages of persons with no corresponding reports of Med 10 conditions on interview are influenced by the com- paratively large group with only one Med 10 con- dition. Obviously, for this category either all con- ditions are reported or no condition is reported. However, even when attention is directed at per- sons with more than one Med 10-inferred condi- tion, it is found that a substantial proportion did not report any of these conditions on household in- terview, As seen in table 24, the percentage for whom no matching conditions are reported on in- terview does decline with increasing number of diagnosed conditions, but it does not fall below 25 percent even for those persons with five or more conditions inferred from the Med 10's. It might be argued that the percentages of persons with no conditions in correspondence with those inferred from the Med 10's only reflect a generally poor relationship between physicianre- ports and survey data, and that there is therefore no special concentration of persons for whom the interview process produces no corresponding in- formation. This is examined below by comparing Percent of persons reporting none or all of Med-10 inferred chronic conditions by num- ber of conditions and respondent status Percent of persons with specified number of Med 10 conditions correspondingly reported on household interview Number of Med 10 chronic conditions None All Expected’ Observed Expected! Observed All persons 2m mmm eee eam 47.1 51.1 9.9 13.9 3mm mmm meee eee 29.0 36.7 3.9 8.9 fmm mmm mmo 19.2 26.8 1.3 4.1 Self-respondents EE tt aati 39.6 43.8 13.8 18.0 mmm mm mmm eee eee 24.6 33.1 5.2 11.3 fmm mmm mmm eae 14.5 17.2 2.2 3.1 Relatives 2mm mmm meee eam 54.6 57.9 6.8 10.1 3mm mmm meme eee em 35.2 40.8 2.5 5.8 a tatatate 31.6 45.5 0.4 6.1 expected! and observed proportions of persons with no reports for Med 10-inferred possibly chronic conditions. The data show that one might well expect a fairly high proportion of persons with no conditions reported, even among those for whom four conditions were inferred from the Med 10's. However, the expected figure is not as high as the observed in any of the cells examined. It would therefore appear reasonable to conclude that the comparatively high percentages that failed to have any condition in correspondence are more than chance phenomena. It is also of interest to examine the data for any special tendency for persons to report all of the Med 10-inferred conditions. This is the other end of the scale in correspondence. The above table indicates that here too the observed per- centages are consistently greater than might be expected by chance—i.e., given the over-all per- centages of conditions in correspondence. lExpected” proportions are obtained from the binomial distribu- tion (p + q)?, where p = percent of Med 10 condition corresponding- ly reported within each category of persons. That is, p = 31.4 for all persons with 2 Med 10-inferred conditions, 33.8 for persons with 3 or 4 Med 10-inferred condition, et cetera. See table 25. Table 25 demonstrates that the number of conditions for which a person has received H.I.P. physician services during the year has no influ- ence on the over-all completeness of reporting of conditions. The proportion of all conditions in- ferred from the Med 10's which are correspond- ingly reported by respondents remains quite con- stant no matter how many Med 10 diagnoses were sustained by the given individual, Correspondence in Reporting Nonchronic Conditions Inferred From Med 10's on Household Interview Although the chief interest of the study cen- tered on problems of chronic disease, advantage was taken of the opportunity to examine accuracy of reporting of nonchronic conditions for which H.I.P. physician services had been reported as rendered in the two weeks preceding date of inter - view. Nonchronic conditions were defined as all conditions codable to ISC-PHS codes which had not been designated "'possibly chronic" in the re- view of codes made prior to processing the data, Services recorded by H.I.P. physicians onthe Med 10's for dates falling within the two-week period 23 ending on the Sunday preceding interview were examined and coded to such conditions whenever appropriate. A total of 143 (unweighted) nonchronic condi- tions for which service appeared on the Med 10's in the specified time interval was inferred in the original coding. There was failure to report any condition in correspondence in 58 of these 143. The weighted figure was a failure toreportin 106 out of a total of 233 nonchronic conditions in- ferred. Because of the possibility that errors in entries on the Med 10's for nonchronic conditions, often attended only once by the physician, might play a larger role than inthe case of chronic con- ditions, for which the relevant period of service was the whole year preceding interview, the 58 cases in which there was failure to obtain an in- terview report were all searched in the clinical records at the medical groups or physicians’ of- fices in an effort to confirm the diagnosis in- ferred from the Med 10's. In this way the Med 10- inferred diagnosis was confirmed for the date specified in 44 of the 58 cases. The 14 cases for which confirmation was not obtained (no entry was found in the clinical chart in 6 of these; in 8 a dif- ferent diagnosis appeared) were eliminated from the analysis. There remained a total of 129 un- weighted, or 201 weighted, conditions which had been inferred from the Med 10's and confirmed in the clinical record in all instances where there was failure to report the condition on interview! Correspondence in reporting these 201 condi- tions on household interview is given in tables 26 and 27. Of the total, 63 percent were correspond- ingly reported by respondents. Conditions for which one or more services in the two weeks had been rendered at home or in the hospital were bet- ter reported (77 percent) than those for which services had been rendered only in the physician's office (56 percent). And conditions seen by the phy- sician more than one time within the two weeks were somewhat better reported than those which had been seen only once. By broad diagnostic cat- egory, the best reporting was for acute respira- tory conditions (73 percent), and the least com- plete for acute conditions of eye and ear (40 per- cent). Nonchronic conditions medically attended within the two weeks preceding interview werere- ported best in children (67 percent) and least ac- I is recognized that had the clinical records been checked for verification of all conditions inferred from the Med 10’s in the two weeks, whether survey reported or not, the total number of matched conditions might have been reduced to some extent. But it was not possible to carry out the same checking procedure for such a large number of conditions. The effect is therefore to give a somewhat higher figure for correspondence in reporting nonchronic conditions than would have been obtained had more conditions been eliminated from both the numerator and the denominator of the correspondence rate. 24 curately in spouses (47 percent); 60 percent of these conditions in self-respondents were corre- spondingly reported on survey (table 27). In evaluating these data on nonchronic condi- tions medically attended in the two-week period, it is worth noting that possibly chronic conditions for which Med 10 services were rendered in the same time period were reported on interviewtoa similar extent—>58 percent of the total conditions in Classes 1, 2, and 3 (table 5), compared with the over-all figure of 63 percent for the nonchronic conditions. In other words, roughly 40 percent of all conditions inferred from H.I.P. physician re- ports for the two weeks preceding interview re- mained unreported by the respondents. Reporting of Medical Care on Household Interview Data from the study provide information on the proportion of persons stated by H.I.P. physi- cians to have received a service in the two weeks preceding interview for whom a doctor contact in this period was reported by respondents. There- port of having seen a doctor in the two weeks is not in any way tied to reports onillness. All H.I.P. physician services on the Med 10's within this time period were noted— whether associated with a diagnosis, or simply a physical examination, or any other type of service. Correspondence in sur- vey reporting of physician contact was judged solely on the basis of whether any doctor contact in the two weeks was reported (answers to Ques- tion 18 of NHS schedule). There was no require- ment that the physician named on interview be identified as an H.I.P. physician, nor that the reason for the doctor contact (Question 19: "What did you have done?') be matched in any way to the nature of the Med 10 service reported by the H.I.P. physician. Of all persons for whom H.I.P. physicians noted a service during the two weeks on the Med 10's, 64 percent were reported on interview to have seen a doctor in this period (table 28). There was no difference in this proportion between males and females, and no over-all difference between proxy and self-respondents. Neither was there any clearcut variation with age. Similarly, the data were processed to deter- mine the extent to which persons noted on the Med 10's as having received at least one H.L.P. physician service during the year preceding inter - view reported their last contact with any doctor as within the study year. The survey report on date of last physician contact was derived from answers to Question 20: "How long has it been since you last talked to a doctor?" Eighty-one percent of the persons for whom H.1.P. physicians entered services on the Med 10's in the study year were reported on interview to have had their last doctor contact within this period. This propor - tion shows little variation when examined in con- nection with a number of demographic character - istics. A slightly higher proportion of females than of males (83 compared with 79 percent) were reported as having last seen a doctor within the year, and the percentage for children under 15 was slightly higher (86 percent) than that for per- sons of other ages. Negroes for whom H.I.P. physicians had noted Med 10 services reported their last doctor contact as within the year to a greater extent than whites (88 compared with 80 percent, a statistically significant difference). The proportion of persons for whom doctor con- tact within the year was correspondingly reported varied directly with the education of the family head—from 75 percent where the family head had completed less than nine years of schooling to 88 percent where more than 12 years of schooling had been reported. There was no variation inthis percentage with family income. One may conclude that there is some under- reporting of physician contacts both in the two- week period and in the year preceding household interview. Data from the current study do not how- ever provide any measures of the contrary ques- tion—the extent to which medical care is reported on household interview as occurring in a given time interval although in fact it was not received during that period. Reporting of Hospitalization on Household Interview In contrast with the level of correspondence in reporting medically attended conditions, hos- pitalization experience was very well reported by the respondents in this study. A hospitalization is here defined as an episode involving one or more nights in a general or allied hospital ("'short-stay"' institutions) in the study year. Of such episodes which had been inferred from the Med 10's and confirmed by the hospital or Associated Hospital Service record as meeting the study definition, 87 percent were correspondingly reported on household interview. There is a difference of only 2 percent between average duration of stay as computed from the dates of admission and dis- charge furnished by the record source (hospital or Blue Cross) and that obtained from respondent reporting. The number of nights in the hospital was exactly stated on interview for almost half the episodes (49 percent) reported on interview, and was in agreement by plus or minus one night with the duration obtained from the record source in an additional 35 percent of the survey-reported episodes. Agreement on duration of stay within one hospital day is therefore shown for almost 85 per - cent of the episodes reported. Correspondence in reporting the fact of hos- pitalization.— The episodes of hospitalization un- der the care of H.I.P. doctors which had been in- ferred from the Med 10's were confirmed from hospital or AHS records through the hospital fol- low-back procedure described in the Methodology section. A total of 350 such episodes! was avail- able as the base for examining correspondence in reporting the fact of hospitalization (or rate of un- derreporting) presented in tables 29 and 30. There is no difference in the proportion of hospitalizations correspondingly reported in self- respondents (88 percent) and others (87 percent), nor isthere any demonstrable variation in relation to the sex ofthe respondent or the specific rela- tionship to the respondent (table 29). There is also little over-all variation with respect to the age of the hospitalized person, but self-respondents aged 45 and over show a somewhat higher correspond- ence (89 percent) than others of this age (77 per- cent). Hospitalizations among women are slightly better reported by female respondents (89 per- cent) than by male respondents (81 percent). None of these differences is statistically significant, Education of the family head shows no con- sistent pattern with the percent of hospitalizations correspondingly reported by self-respondents, but the percent reported by proxy-respondents ap- pears to increase with increasing education of the family head. Hospitalizations in families in the lowest in- come class, less than $4,000, were reported less completely than those in all other families—73 percent compared with roughly 90 percent for families in all other income groups (table 30). This difference reflects less complete reporting for both proxy and self-respondents in the lowest income class. The time interval elapsing from date of ad- mission to the hospital to date of household in- terview has a distinct influence on the proportion of episodes reported on survey. Admission to hos- pitals eight months or more before the date of in- terview were deficiently reported both in self-re- spondents and in others (table 30). Only half of the admissions before July 1957 (10 to 11 months be- fore interview) were reported on. interview, com- pared with four fifths of those from July-Septem- ber 1957, and with 97 percent of all the remaining (more recent) admissions. IThree of these episodes were actually not inferred from the Med 10’s, but were reported by the hospitals queried for dates of admis- sion and discharge for the 347 episodes which had been obtained from the Med 10’s. For convenience, the total 350 are referred to in the report as those inferred from the Med 10’s. 25 Exactly the same duration of stay from the record source and the household interviews is found for 49 percent of all hospitalizations here examined; for hospitalizations in children this proportion is 61 percent. The difference between duration of stay as computed from the two sources is no greater than one hospital day in 86 percent of the hospitalizations of self-respondents, 81 percent of those of relatives as a whole, and 94 percent of those of children of respondents. Fe- male self-respondents reported duration more ac- curately than males responding for themselves, with 90 percent of their hospitalizations differing in reported duration from the record source by no more than one day, compared with 74 percent as the comparable figure for male self-respondents. In summary, the distribution by number of nights in the hospital of all episodes reported on interview and confirmed by the record source is substantially the same whether based onthe dura- tion obtained from the record or the interview source (table 32). Overreporting of hospitalization.—A total of 470 hospitalizations in general and allied hospitals which were reported on interview was confirmed through the hospital follow-back procedure as in- volving one or more nights in the hospital in the study year. Of these episodes, 306 had been in- ferred from the Med 10's. These represented 87.4 percent of the 350 episodes which had been in- ferred from the Med 10's and confirmed by an in- dependent record source as involving at leastone night in the hospital in the study year. If it is as- sumed that the same extent of underreporting applied to the hospitalizations attended by non- H.I.P. physicians—that is, that the 164 such epi- sodes reported on interview represented 87.4 per - cent of a total of 188 such hospitalizations—then the estimated total universe for this population in the study year is 538 hospitalizations. There were 17 interview reports of hospitali- zations as having occurred in general or allied hospitals in the study year for which the independ- ent record sources provided no confirmation. Of these, four were overreports which had been tele- scoped into the study year from the preceding year! while no confirmation of any kind could be obtained for the remainder. Relating these over- reports to the estimated universe of hospitaliza- tions produces an overreporting rate of 3 percent. Net reporting of hospitalization and of hos- pital days on interview.—It is apparent that even with the relatively good correspondence in re- porting the fact of hospitalization which was found in this study, overreporting was so small that a net underreporting rate of episodes of hospitali- zation of 9 percent remains. Net underreporting of total nights in the hospital in the study year is somewhat lower (5 percent) because of the slight inflation in duration of stay for reported episodes. 11 is of interest that for 3 of the 4 telescoped episodes the month of admission to the hospital was correctly stated by the respond- ents, who erred in reporting the year as 1957 rather than 1956. Nights in hospital by source of hospital episode Nights in hospital Source of hospital episode From hospital|Reported on house- record hold interview Total===-----c-cmmec ccc mccm cece mmm 5,016 4,780 Episodes confirmed by hospital or AHS record Inferred from Med 10's and reported on household interview-=--=-----ccccecccccncccnaox0- 2,376 2,397 Inferred from Med 10's, not reported on household interview-----------ccemccccccccaaa—- 258 - Not inferred from Med 10's, reported on household interview-------=----eccccmccccenaa—- 2,149 2,223 Estimated episodes not inferred from Med 10's, not reported on household interview” (estimated underreports, hospitalizations by non-H.I.P. doctors)--=-=-=---cemcmceoomeeceeee——- 233 - Overreports (reported household interview, not confirmed by hospital or AHS as involving time in hospital in Study Jean) s=mwsis mmm seme - 160 ‘Estimate is made by assuming same relationship between unreported and reported days as for hospitalizations by H.LP. doctors, ie, x = 258 :x=233 2,149 2,376 26 SUMMARY AND DISCUSSION This report has presented a methodological study undertaken by the U, S. National Health Sur- vey in an effort to improve understanding of the data obtained from its household interview survey on health. The study, carried out under contract by the Division of Research and Statistics, Health Insurance Plan of Greater New York, was focused in the main on an examination of the relationship between the conditions reported in the household interview and conditions diagnosed by physicians among these persons as they received medical care during the year prior to interview. The chief emphasis of the study is a compari- son of chronic conditions inferred from a set of physician reports with survey-reported condi- tions. Additional data are presented on corre- spondence in reporting acute conditions attended by H.I.P. physicians in the two weeks preceding interview, the reporting of medical care, and the reporting of hospitalization experience. The avail- able data lend themselves best to expressions of the underreporting on interview of H.I.P.-medi- cally attended conditions. Since it is known that some medical care is obtained outside the H.I.P. setting by persons enrolled inH.I.P., the total uni- verse of medically attended illness to be inferred from physician records was not available for com- parison with the total universe of survey-reported conditions. The data therefore do not provide the possibility of an analysis of overreporting to par- allel that presented for underreporting. The study population is a stratified sample of families enrolled in the Health Insurance Plan of Greater New York who were residents of the five counties of New York City and Nassau County. H.I.P. is a prepaid insurance plan providing med- ical care through group practice of 31 medical groups in the geographical area specified. Physi- cian reports on medical services to these persons in the year preceding household interview were ob- tained from the routine H.I.P. physician reports on medical services to insured persons, submitted in the H.I.P. reporting document known as the Med 10. Household interviews were obtained with an in- terview schedule containing minor modifications of the regular NHS document. Interviewing was carried out by the Regional Office of the Bureau of the Census which is responsible for the regular NHS interviews in the area. The procedures adopted provided for a priori selection of all International Statistical Classifi- cation-Public Health Service codes to be classified as ''possibly chronic." A determination was then made, for each interviewed individual, of all con- ditions codable to these categories which could be inferred from the Med 10 services in the year preceding household interview, Correspondence in survey reporting of these categories was estab- lished on the basis of a case-by-case comparison of the coded Med 10 services and the interview schedule. Chronic conditions reported on inter- view which had not been inferred fromthe H.I.P.- physician reports were also noted and their char- acteristics compared with those survey-reported conditions judged to match those inferred from physician reporting. A separate study, referred to briefly in this report, consisted of interviews of H.I.P. physicians who had rendered the services for specified conditions to a subsample of the in- terviewed population. The purpose of this study was to relate the comparison of physician record and survey data to the physician's total knowledge of the patient. For analytical purposes all chronic conditions were grouped into three classes, defined with re- spect to the checklist questions on the NHS sched- ule, as follows: Class 1: conditions covered by NHS terminol- ogy for the checklists (Cards A and B) without any qualifications intro- duced by modifying adjectives Class 2: conditions which might be suggested by checklist terminology, but there arz qualifications arising for the most part from the use of modifying adjectives (''repeated,'" ''chronic," ere.) Class 3: conditions which would not in any ob- vious way be suggested by checklist terminology, but which had been judged ''chronic''or "possibly chron- ic" on the basis of the clinical ex- perience of physicians Highlights of Findings Chronic conditions Respondents on household interview furnished statements in correspondence with the "possibly chronic medically attended conditions inferred from H.I.P.-physician reports for the year pre- ceding interview in the following proportions: 44 percent of Class 1 conditions (checklist unquali- fied), 28 percent of Class 2 conditions (checklist qualified), and 20 percent of Class 3 conditions (nonchecklist). This relationship of level of re- porting to class of condition persisted when re- porting was examined by demographic variables such as age, sex, respondent status, socioeco- nomic status, and by a number of variables re- lated to medical care received. One third of all interview-reported conditions which corresponded to conditions inferred from 27 the Med 10's were mentioned in response to the checklist questions. The use of the checklists im- proved correspondence in reporting for all classes of condition, even for those (Class 3) where the likelihood of stimulating response from the check- lists seemed comparatively small. There was a strong relationship between the number of physician services rendered for a given condition in the year preceding interview and the probability of that condition's being reported by the respondent. Eighty percent of conditions for which 10 or more physician services had beenre- ceived were reported on interview. For Class 1 conditions (checklist unqualified) 57 percent of those with more than one service were reported, 27 percent of those with only one service. Chronic conditions last attended by a physi- cian within the two weeks preceding interview were better reported than those with the last phy- sician service further removed in time. Of the former group 58 percent were reported, compared with 24 percent of those for which no service had been given in the four months preceding inter- view, While reporting of chronic conditions by per - sons responding for themselves was somewhat more complete than that by persons responding for other family members, the magnitude of the dif- ferential was small. Poorer reporting of condi- tions in children was largely responsible for the differential. Little difference was found between male and female respondents in the percent of chronic con- ditions reported in correspondence. Somewhat more complete reporting of conditions in mature and older adults was obtained than for children and young persons. Conditions in males and fe- males were reported to the same extent, although some differences by sex appeared when specific classes of condition were examined. There was no difference in the percent of chronic conditions reported on household inter- view by race; nor was any consistent pattern found by education of the family head or education of the individual with the condition. The figure varied little in all income classes except the low- est (under $4,000), where a somewhat higher per- cent of chronic conditions was correspondingly re- ported. The percent of Class 1 and Class 2 conditions reported was not affected by family size, but a decreasing percent of Class 3 (nonchecklist) con- ditions was reported with increases in family size. Permission to review medical records was granted for almost 90 percent of the persons in- terviewed. Completeness of reporting of chronic conditions did not appear to be associated with this variable. 28 Great variation was shown in percent of con- ditions reported on interview by specific diagnos- tic categories. This variation appeared in each of the three classes of condition, with no clear pat- tern discernible in relation to diagnostic termi- nology alone. There were relatively few cate- gories for which more than half of the conditions inferred from the Med 10 physician reports were correspondingly reported. Conditions reported on interview which cor- responded to those inferred from the Med 10's were further characterized by ''type of match'— an indication of the degree of similarity of the terminology used by the physician and the layman. Of all matching survey-reported conditions, 37 percent were reported by the respondents in terms which did not permit coding to the same diagnostic category (NHSRecode No. 3)as the physician'sre- port. The proportion of matches of this type ("Typed") varied greatly with the specificity of the particular disease category. There were many conditions reported on in- terview, chronic according to the ISC-PHS code designation, which did not correspond to any diag- noses inferred from the H.I.P. physician reports. Sixty percent of the total 3,739 interview-re- ported chronic conditions fell into this category. In comparison with the survey-reported condi- tions matched to Med 10-inferred conditions, a larger proportion of these unmatched conditions were not medically attended, or were reported last medically attended before the study year or by a non-H.I.P. doctor. A smaller proportion of the unmatched conditions was associated with dis- ability or time lost from work or school in the two weeks preceding interview. This study did not pro- vide for the examination of any medical records which might be applicable to this group of un- matched conditions. There was somewhat greater concentration both of persons for whom no chronic condition was correspondingly reported on interview and of per - sons for whom all chronic conditions were re- ported than might have been expected by chance alone. The number of conditions for which a person received H.I.P.-physician services during the study year had no influence onover-all complete- ness of reporting. Nonchronic conditions Respondent reports in correspondence with 63 percent of nonchronic conditions, attended by H.I.P. doctors in the two weeks preceding inter- view, were obtained on survey. This figure did not differ by much from the 58 percent of chronic con- ditions, attended by H.I.P. doctors inthis interval, which were reported on household interview. Medical Care A doctor contact within the two preceding weeks was reported on interview for 64 percent of the persons for whom an H.I.P. physician service had been noted in this time period. Of the persons for whom physician services were noted on the Med 10's in the year preceding household interview, 81 percent were reported to have had their last doctor contact within this pe- riod. Hospitalization Eighty-seven percent of the episodes of hos- pitalization under the care of H.I.P. physicians during the study year were reported by the re- spondents on household interview. A distinctly lower proportion of hospitaliza- tions which had taken place relatively long before the interview was reported than for more recent admissions. About 97 percent of admissions with- in eight months of the date of interview were re- ported, compared with 50 percent of those which had taken place almost a year before the inter- view date. Duration of hospital stay was reported witha high degree of accuracy, with the mean duration of stay as computed from interview reports only 2 percent higher than that computed from the rec- ord source. Agreement on duration of hospital stay within one hospital day was shown for almost 85 percent of the episodes reported on survey. Overreporting of the fact of hospitalization was estimated to be very small, leaving anet un- derreporting of 9 percent of the episodes which took place in the year preceding interview. Net underreporting of total nights in the hospital was somewhat lower (5 percent) because of the slight inflation in duration of stay for the episodes re- ported, The results of the current study illustrate the complex problem posed by attempts to interpret data on chronic diseases collected through the household interview process. They suggest strong- ly that the survey information does not conform even moderately well to the universe of conditions inferred from physician reporting. It would ap- pear that this lack of conformity cannot be ex- plained by simple population attributes and char- acteristics of the interview situation. Age, sex, socioeconomic status, respondent status, ethnic background, and other conventional demographic attributes exert surprisingly little influence on the degree to which the knowledge that a physician has about the existence of illness is reflected in a household interview. Furthermore, the fact thata physician has recorded a diagnosis of a disease usually thought of as serious, or containing all the elements of chronicity (for example, diabetes or heart disease), by no means gives assurance that the condition will be identified by the respondent in an interview. There is reason to conclude from this study that the lack of conformity does not result from an extreme reluctance on the part of the respondent to talk about illness. Such reluctance is difficult to postulate in the face of the large numbers of chronic conditions which were in fact reported on household interview. It will be recalled that only 40 percent of all chronic conditions reported by respondents were matched to conditions inferred from the Med 10's. The design of this study pre- cluded an analysis of the 60 percent unmatched in- terview-reported conditions in relation to relevant independent physician record sources. While re- spondent failure cannot be dismissed as an im- portant reason for lack of correspondence between the H.I.P. physician and the household interview information, the specific circumstances which ac- company an illness may exert an evengreater in- fluence. The study does throw some light on this issue. The completeness and accuracy with which hos- pital episodes are reported is particularly rele- vant. Here, the respondent is asked to report a circumstance which can only be considered as a fact. There is no speculation about the question, nor are there differences in interpretation or changes in circumstance that may confuse theis- sue. Either the person spent a night inthe hospital or he did not. Added to this factual context is the unusual, dramatic character of the event. The combination of unequivocal meaning and sharp deviation from the ordinary pattern of living is not often present in illness that does not require hos- pitalization. But it is important that in situations where such combinations are likely a high degree of correspondence does result. For example, a very high proportion of conditions for which large volumes of medical services had been rendered were reported.! When such conditions (with 10 or more physician services) were those specified on the checklists without qualification (Class 1), almost 9 out of 10 were correspondingly reported on interview. Under what circumstances would an individual see a physician many times in relation to one condition during a year? Certainly, in most such cases the condition must have been very ac- 1Good correspondence in survey reporting of these conditions was not merely a reflection of the high proportion of hospitalizing conditions in this group. Although one third of all chronic condi- tions for which 10 or more services were rendered did involvea hospitalization, correspondence for those without related hos pitali- zation was still 74 percent. 29 tive and the patient must have acknowledged his illness as a fact. It seems reasonable, too, that such a patient would in most instances have viewed his experience as unusual, either inrelation to his own past history or in relation to thatof his fam- ily and friends. The preceding observations must be viewed as tentative until additional information can be gathered in other settings. A number of conditions that prevailed in the current study make this par- ticularly important, First, the use of comparative- ly unseasoned interviewers raises the question of the extent to which interviewer failure contributed to the poor correspondence between physician and interview reports of chronic diseases. Although the fragmentary evidence on this issue suggests that this factor isnotofgreat significance, further testing is required. Another special characteristic of this study is the population covered—a cross-section of fami- lies enrolled in a health insurance program inthe New York area. Coverage by health insurance of itself is not a limitation. But it would be hazardous to generalize from the experience with a popula- tion in one urban area to the national sample cov- ered by the National Health Survey. Also, while H.I.P.'s population includes a wide range of occu- pations and ages, it is drawn largely from one type of employment group (municipal employees). Perhaps the most important qualifications arise from the nature of the criterion source and the restricted scope of the study. The criterion source for physicians' diagnoses in this study was the H.I.P. Med 10 form. This is not the physician's record on the patient, but a secondary document in which the physician notes the diagnosis, definite or tentative, associated with each face-to-face contact with an insured H.I.P, member. Much processing of these routine Med 10 reports is necessary to collate all medical care rendered within H.I.P. over a given time period. Although the reliability of the Med 10 as a statistical docu- ment has been demonstrated in the past, and was again emphasized by the results of the physician interviews carried out in connection with this study, it is still true that theMed 10's cannot pro- 30 vide details which one might expect to obtain from a complete clinical record. Information onthe history of a given condition, on treatment and progress of associated symptoms or disability, questions related to differential diagnosis, obser - vations which might make posecible a less arbitrary definition of chronicity than that of necessity used in this study—all these cannot be provided by the Med 10's. Further, the fact that the criterion sourcere- flects only a part of all medically attended illness in the study year, and, by definition, none of the illnesses unattended during the year, results in a restricted 'one-way' analysis. The meaning of survey-collected information can never be fully understood until the conditions reported onhouse- hold interview and not found on any doctors' rec- ords relating to the reference period are carefully studied. Additional opportunities for testing the find- ings inthis study are needed. Further observations would be desirable in settings permitting some changes from the design of the present study— notably, the direct use of physicians' detailed clinical notes rather than a secondary summary document for establishing the universe of physi- cian-reported conditions; provision for study of respondent reports of illnesses, the existence of which is not apparent from the physician's record covering the reference period; and extension of the inquiry into the role of the interviewer. New studies should also break into presently unex- plored areas. Until now the emphasis in methodo- logical study has been on determining how well the household interview reports mirror the reports of physicians. But if this relationship should, onre- peated study, prove to be a poor one, the need to know what it is that survey information does in fact reflect will still remain, Through follow-back studies to physicians and patients some under- standing could be obtained regarding the influence on respondent reporting of doctor-patient commu- nication, the assessment and interpretation the pa- tient made of his illness, and the circumstances that make the respondent aware of and ready to report a given condition in an interview situation. Table 1: 10. 5 12, 13. 14. DETAILED TABLES POPULATION CHARACTERISTICS Number and percent distribution of interviewed H.I.P. enrollees by age and sex-- . Percent distribution of interviewed H.I.P. enrollees by selected demographic characteristics and age--=-===-==---c-cecccmmmcocmmcccooeococmmecceeoocoooooone . Percent distribution of H.I.P. enrollees with specified characteristics for whom possibly chronic conditions were inferred from Med 10 services in the study year by type of condition Inferreds—= = issmsmammnese cos oe ow wows dso mm me mmm CORRESPONDENCE IN REPORTING CHRONIC CONDITIONS - FACTORS OTHER THAN DIAGNOSIS - Percent of possibly chronic conditions inferred from Med 10's reported on household interview . Percent of possibly chronic conditions inferred from Med 10's reported on house- hold interview by number of related Med 10 services in study year, respondent status, and class of condition-er=errrrrsrrrrmrcenecccncnvnn nn nmnn nnn nn . Percent of possibly chronic conditions inferred from Med 10's reported on house- hold interview by interval between last related service and household interview, number of related Med 10 services in study year, and class of condition--------- Percent of possibly chronic conditions inferred from Med 10's reported on house- hold interview by number of related Med 10 services in study year, interval be- tween first and last related service, and class of condition----=-------cc-ccce-- . Percent of possibly chronic conditions inferred from Med 10's reported on house- hold interview by number of related Med 10 services in study year, respondent status, and grade of CONALLione=rmmmmmmmm mmm nisin www ens wnwe nnn w= . Percent of possibly chronic conditions inferred from Med 10's reported on house- hold interview by relationship to respondent, sex of respondent, and class of condition-=-==emmeecccmmccccceceeceeeememmen memes esessesssossssso—ss oe Percent of possibly chronic conditions inferred from Med 10's reported on house- hold interview by relationship to respondent, sex of respondent, and grade of Fle) eV b A oh Me « BE er Percent of possibly chronic conditions inferred from Med 10's reported on house- hold interview by age and sex of person with the condition and class of condi- Percent of possibly chronic conditions inferred from Med 10's reported on house- hold interview by age and respondent status of person with the condition and class of condition-====-ececccccccccccccccrcccece meme e mee mc ec cccesc oso Percent of possibly chronic conditions inferred from Med 10's reported on house- hold interview by education of family head, respondent status, and class of condition-===m=e-cemmccccccccec cee em meee m meee mem ees ecesesesesoscsss-—ooooooo Percent of possibly chronic conditions inferred from Med 10's, persons aged 15 years or older, reported on household interview by age and education of person with the condition and class of condition---=-==---cceccccccccecnmncnnnonconnannx Percent of possibly chronic conditions inferred from Med 10's reported on house- hold interview by family income, respondent status, and class of condition------ Page 34 34 35 36 37 38 39 40 41 42 43 44 45 46 CORRESPONDENCE IN REPORTING CHRONIC CONDITIONS - FACTORS OTHER THAN DIAGNOSIS—Con. DETAILED TABLES--Continued Table 15. 30 16. 17. 18. 19. 20. 21. 22. 23. 24, 25. 26. 27. 28. Percent of possibly chronic conditions inferred from Med 10's reported on house- hold interview by number of H.I.P.-insured persons in household, respondent sta- tus, and class of condition===-==cc emo eee ea - Percent of possibly chronic conditions inferred from Med 10's reported on house- hold interview by respondent status, indication of permission to review medical records, and class of condition-====mem ome eee CORRESPONDENCE IN REPORTING CHRONIC CONDITIONS - DIAGNOSIS Percent of possibly chronic conditions inferred from Med 10's reported on house- hold interview by diagnostic category and class of condition-==--=c-ecececcaaaao Percent of possibly chronic conditions inferred from Med 10's reported on house- hold interview by Recode #3 categories ranked within each class by percent of conditions correspondingly reported and diagnosis and class of condition-------- Differentials in percent of Med 10 conditions reported on household interview by respondent status--Recode #3 categories within each class of condition by magni- tude of correspondence ratio between proxy- and self-respondentS----------ecac-- ALL CHRONIC CONDITIONS REPORTED ON HOUSEHOLD INTERVIEW Percent distribution of all survey-reported conditions by question number pro- ducing household interview report--all possibly chronic conditions according to ISC designation--by class of condition and whether or not matched to conditions inferred from Med 10'S======-= cmc m meee Medical care reported in relation to conditions reported on household inter- view--all possibly chronic conditions according to ISC designation--by class of condition and whether or not matched to conditions inferred from Med 10's-=----- Percent of conditions repoited as producing disability, bed disability, and time lost in two weeks preceding household interview--all possibly chronic conditions according to ISC designation--by class of condition and whether or not matched to conditions inferred from Med 10's-======== me mmm eee eee memcceccem ema All household survey-reported conditions coded "chronic" by National Health Sur- vey by diagnosis reported on household interview and by whether or not matched to conditions inferred from Med 10's======--- Hm mm ee meme EXPERIENCE OF PERSONS Persons classified by number of possibly chronic conditions inferred from Med 10's by number of these reported on household interview and respondent status--- Percent of possibly chronic conditions inferred from Med 10's reported on house- hold interview in persons classified by number of conditions inferred from Med 10's and respondent StatuS====mmmemeeeeceemeececcccececcmmmcmcmmcemmeee———————— NONCHRONIC CONDITIONS Percent of nonchronic conditions inferred from Med 10's for two weeks preceding interview reported on survey by broad diagnostic category, volume, and place of Percent of nonchronic conditions inferred from Med 10's for two weeks preceding interview reported on survey by volume of service, relationship to respondent, and sex of respondent-=--=---c comme eee cme mem eee REPORTING OF MEDICAL CARE Correspondence in reporting doctor contact in the two weeks preceding household interview by respondent status, sex, and age--========s-ceecmmcccemmcmmccccanao=n Page 47 48 49 54 56 57 58 59 60 61 62 63 64 Table 29. 30. 3. 32. DETAILED TABLES--Continued HOSPITALIZATION Percent of hospitalizations reported on household interview by age of person hospitalized, respondent status, and sex of respondent---=======eesemsmmmo==—=—- Percent of hospitalizations reported on household interview by family income, date of hospital admission, duration of hospital stay, and respondent status---- Comparison of average duration of hospital stay from record source with that from household interview reports by selected characteristics--=-=-=-==-=-=-----===== Comparison of percent distribution of survey-reported hospitalizations by dura- tion of stay from record source and from respondent reports----=--------==-=-=-°° Page 65 66 67 68 33 Table 1. Number and percent distribution of interviewed H.I.P. enrollees by age and sex Age Sex Both sexes Male Female Both sexes Male Female Number of persons Percent distribution All ages-======c-eecceccceceaao 6,609 3,358 3,251 100.0 50.8 49.2 Under 15----=-m-ccmmccmcccm cement 2,046 1,060 986 31.0 16.0 14.9 15-24 mm mmm mmm eee 466 215 251 7.1 3.3 3.8 25-44mcmcmm meme eee eee 2,281 1,096 1,185 34.5 16.6 17.9 45-64 mmm mmm eee eee 1,632 871 761 24.7 13.2 11.5 65t-—mmmmmmmcm mcm 184 116 68 2.8 1.8 1.0 Table 2. Percent distribution of interviewed H.I.P. enrollees by selected demographic characteristics and age Age Demographic characteristic ALL | Under | 5.54 | 25-44 | 45-64 | 65+ ages 15 All persons Number-----=e-e-eecc cece cece meee ee 6,609 2,046 466 2,281 1,632 184 Percent-=-===-cecemmcc cmc cme me eee ae 100.0 100.0 100.0 100.0 100.0 100.0 Percent distributior Education of family head Under 9 years==---==c-coccommcoccncoonnoo 22.7 15.2 22.3 17.1 37.6 46.2 9-12 years==---=----ececccccmemeeaeenan 47.0 55.9 45.5 51.8 32.0 25.5 12+ yearg-=-=-==-mecccccememcm cena 26.8 26.0 29.0 27.8 26.1 23.9 Unknown or unreported----==-==-==c------- 3.4 2:9 3.2 3.3 4.2 4.3 Family income Under $4,000-===-==cccmoommmmamcc ence 10.6 10.4 22.4 9.3 10.8 22.3 $4,000-4,999--==-=mcmmmmcm mcmama 15.3 17.6 9.2 17.1 11.8 15.2 $5,000-6,999-====cmccc mmm e eee 37:2 45.8 31.3 40.3 25.7 20.1 $7,000-9,999--==-mccmmmmmm meee 18.5 13.8 21.7 18.1 23.7 22.3 $10,000 === ==mmmom mmm e ee 11.7 7.4 14.2 9.3 20.0 8.2 Unknown or unreported------=----ce-e--a- 6.6 4.9 11.2 5.8 8.0 12.0 Race White-===---emee ccm cece ccc ccc eee 87.3 85.6 86.9 84.3 92.8 96.2 Nonwhite--=--==--c-oecccccc ccc ccc ccceeen 12.7 14.4 13.1 15.7 7.2 3.8 Occupation of subscriber Professional, managerial 22.2 18.6 26.6 20.8 27.1 23.9 Clerical and saleS-=====cecccccccccnnna- 12.5 10.3 14.2 11.5 16.3 10.3 Craftsmen---===-cccmeccmccccncmacc nena 13.3 11.9 14.2 11.0 18.2 12.5 Transit operatives----=---c-cccccecacaa- 7.2 8.8 9.0 6.2 6.6 3.8 Other operatives-=--=-=---cccccmccccceanan. 12.5 13.3 9.2 15.3 9.7 0.5 Firemen, policemen-=---==c-ccccceccacanax 18.0 25.4 14.2 22.3 6.0 - Other service, private household-------- 4.2 3.6 3.4 3.7 5.5 7.1 Laborers==-===--eecececmcmcn cence 4.7 4.7 5.2 3.8 5.8 6.0 Unknown or unreported------=-==c-cecec-- 1.0 0.6 1.3 1.3 1.2 - Not working-----=-cecemcmccmccccccmceeaa 4.3 2.8 2.8 3.9 3.7 35.9 34 Table 3. Percent distribution of H.I.P. tion inferred enrollees with specified characteristics for whom pos- gibly chronic conditions were inferred from Med 10 services in the study year by type of condi- All persons One or more possibly No chronic conditions in- nd Characteristic — 3 ferred from Med 10's condition umber | Percent Total Checklist | gi per inferred un- only from qualified Med 10's Percent distribution All persons-------=e------ceo-ao--- 6,611 100.0 | 44.4 23.07 21.3 55.6 Education of family head Under 9 year§-=---=-=-==---ceeeccecomeooooon 1,501 100.0 | 40.8 22.9 18.0 59.2 9-12 years----------=m--e-cecmeemceeea———— 3,112 100.0 || 42.4 21.3] 21.2 57.6 12+ years----===me-memememmeeeee————————— 1,771 100.0 f 51.5 27.3] 24.2 48.5 Unknown or unreported-=----=-=-=-e--c--ce--- 227 100.0 38.8 15.4 | 23.3 61.2 Sex Male----====-c--emmmem meee mmc cm me 3,360 100.0 | 43.5 23.3] 20.2 56.5 Female-==-===---cocecmcmcc meme meme mm 3,251 100.0 | 45.2 22.7] 22.5 54.8 Age Under 15 years------e-e--eeeeeeeeeeeeeaa- 2,046 100.01 33.2 15.2 18.0 66.8 15-44 years---------m-mmeeccceeccee—————— 2,751 100.0 | 45.9 20.7; 25.2 54.1 45+ yearg-------=--m----eeceemcececcooaa- 1,814 100.0 | 54.6 35.4 19.2 45.4 Relationship of respondent Self-respondent8----=-==---memccceccaao=- 2,428 100.0 | 51.9 28.1 23.8 48.1 Relativeg------===-eccmmcecmccc ccc ce mae 4,140 100.0 | 40,1 20.1| 20.0 59.9 Spouse-=---==-=---memmmmmmme—meeo————oooo 1,411 100.0 || 49.3 26.6 | 22.6 50.7 Child-=-=--=c---ccmmmmcc cmc c ccm mmm eo 2,429 100.0 || 34.9 15.0 19.9 65.1 Other relative------=-c-e-mcecccccaacn-" 300 100.0 || 39.0 31.0 8.0 61.0 Unrelated and unknown relationship------- 43 100.0 34.9 16.3 18.6 65.1 Survey report on hospitalization, study year YeS-mmmmeommmmmmeme meee memee— eon 471 100.0 | 60.5 37.8) 22.7 39.5 NO-=mmmmmm mmm emccc meme cemee mmm mmm 6,090 100.0 | 43.2 21.97 21.3 56.8 Unknown or unreported-------=-e-===-==a-- 50 100.0 | 38.0 22.0] 16.0 62.0 Permission to review medical records YeS=mmmmmmmmm meee emcee mmm — mma 5,882 100.0 | 43.9 22.31 21.6 56.1 NO==-mmmmmmccm mc ccecmme ecm ec ecm mm 729 100.0 | 47.9 28.7] 19.2 52.1 35 Table 4. Percent of possibly chronic conditions inferred from Med 10's reported on household in- terview by number of related Med 10 services in study year, respondent status, and class of con- dition Number of related Med 10 Number of conditions in- ferred from Med 10's Percent correspondingly re- ported on household interview services and respondent Checklist Checklist status Non- Non- Un- Qualified checklist Wii aliiad checklist u u qualified | “¢].5s 2 Class 3 || qualified Rua 9 Class 3 Class 1 Class 1 All services Total----=-----cmuuum- 1,872 1,231 1,545 44.1 27.6 20.4 Self-respondents--=--=--=----- 878 605 739 47.7 35.7 21.1 Relativeg----=-=---c-cco---- 987 621 801 41.0 20.0 19.9 1 service Total------------==-=-- 802 685 796 27.2 20.0 14.3 Self-respondents------------ 377 323 376 30.0 27.2 18.6 Relatives--------cc-coceuan- 421 358 415 24.7 13.7 10.6 2-4 services Total--=--ccecccmcnca= 594 414 541 40.9 33.3 18.1 Self-respondents-=-===-=-====== 266 212 244 45.9 41.0 15.6 Relatives----=c-ccecccmeeaaax 327 201 297 36.7 25.4 20.2 5-9 services Total---==-c-cccmenan- 210 114 131 62.4 48.2 45.8 Self-respondents---=-=---=-=---- 116 58 69 67.2 58.6 42.0 Relatives--=--=-=ccemmcceaxa- 93 56 62 57.0 37.5 50.0 10+ services Total-----==cemeceeea= 266 18 77 88.0 55.6 55.8 Self-respondents-=--========- 119 12 50 89.1 (*) 38.0 Relativeg-=-===cccccccanaaa- 146 6 27 87.7 (*) 88.9 36 Table 5. Percent of possibly chronic conditions inferred from Med 10's reported on household in- terview by interval between last related service and household interview, number of related Med 10 services in study year, and class of condition Interval between last Number of conditions in- ferred from Med 10's Percent correspondingly re- ported on household interview service and household Checklist Checklist interview and number of Non- 7 Non- related Med 10 services Un~ 4 | Qualified checklist WT Qualified | checklist qualifie Class 2 class 3 [9 Class 2 Class 3 Class 1 Class 1 Two weeks or less Totale====m=cmmemeenean 246 87 124 67.9 50.6 41.9 1 service--------cccccccaaa- 45 32 43 33.3 46.9 30.2 2-4 services------=-cceoooo- 53 28 42 50.9 46.4 28.6 5-9 services-=-=--c-cc-ca-o- 26 22 20 53.8 ‘50.0 75.0 10+ services---------cccce-- 122 5 19 91.0 (*) 63.2 More than two weeks but less than four months Total-==--cceccncanann 714 413 602 49.3 34.1 22.1 1 service--=-===-cccccenoo-- 247 174 304 28.3 21.3 17:1 2-4 services---===--c--e---- 262 184 204 43.1 40.8 20.6 5-9 services=====----ecceao- 109 47 52 74.3 53.2 40.4 10+ services-=--=-=---ccc---- 96 8 42 91.7 (*) 42.9 Four months or more Total-====cocmmcenaaan 912 731 819 33.7 21.2 15.9 1 service-===-=ccmmcncccana- 510 479 449 26.1 17.7 10.9 2-4 services-------==-meen-- 279 202 295 36.9 24.8 14.9 5-9 services-------=--ce---- 75 45 59 48.0 42.2 40.7 10+ services--------=--cee-- 48 5 16 72:9 (*) 81.3 37 Table 6. terview by number of related Med 10 services in study year, related service, and class of condition Percent of possibly chronic conditions inferred from Med 10's reported on household in- interval between first and last Number of related Med 10 Number of conditions inferred from Med 10's Percent correspondingly re- ported on household interview services and interval Chathiiat Checklist e e s tol oh) bs tent he Non- : Non- Un- checklist Un- checklist qualified Qigligag Class 3 | qualified Quslisin Class 3 Class 1 Class 1 All services Total----=-ccccccieaana= 1,872 1,231 1,545 44.1 27.6 20.4 One service-------=-=-=-==--- 802 685 796 27.2 20.0 14.3 More than one service---- 1,070 546 749 56.8 37:2 26.8 One month or less------------ 313 . 250 352 35.1 32.0 22.4 More than one month---=---=--- 757 296 397 65.8 41.6 30.7 2-4 services Total----=---===-=co-== 594 414 541 40.9 33.3 18.1 One month or less----====---- 272 232 323 32.7 31.5 18.3 More than one month---------- 322 182 218 47.8 35.7 17:9 5-9 services Total----=-=-cccmecaea- 210 114 131 62.4 48.2 45.8 One month or less-=-==--====-- 30 18 24 53.3 38.9 70.8 More than one month-----=----- 180 96 107 63.9 50.0 40.2 10+ services Total-----=-=-cceecuau- 266 18 77 88.0 55.6 55.8 One month or less-=----=-=-=-- 11 - 5 (*) a (*) More than one month--====---- 255 18 72 89.8 55.6 55.6 38 Table 7. Percent of possibly chronic conditions inferred from Med 10's reported on household interview by number of related Med 10 services in study year, respondent status, and grade of condition Number of conditions inferred from Med 10's Percent correspondingly reported on household interview Number of related All other All other Med 10 services and respondent Checklist >l service and >1 1 service Checklist >1 service and >1 1 service status no quali- month from lst to only or 1f quali month from lst to |only or 1 fication last oEh B fication last RO Grade I de I ede Checklist Non- lst to Grade Checklist Non- 1st to qualified | checklist last qualified | checklist last Grade II | Grade III | Grade IV Grade II | Grade III | Grade IV All services Total------- 1,872 296 397 2,083 44.1 41.6 30.7 19.7 Self-respondents-- 878 162 187 995 47.7 47.5 28.3 24.3 Relatives--------- 987 134 210 1,078 41.0 34.3 32.9 15.6 1 service only Total-==-=---- 802 1,481 27.2 16.9 Self-respondents-- 377 699 30.0 22.6 Relatives---====--- 421 773 24.7 12.0 2 to 4 services Total------- 594 182 218 555 40.9 35.7 17.9 23.8 Self-respondents-- 266 97 85 274 45.9 41.2 16.5 25.9 Relatives--------- 327 85 133 280 36.7 29.4 18.8 21.8 5 to 9 services Total---=---- 210 96 107 42 62.4 50.0 40.2 57.1 Self-respondents-- 116 53 57 17 67.2 56.6 40.4 58.8 Relatives--------- 93 43 50 25 57.0 41.9 40.0 56.0 10+ services Total------- 266 18 72 5 88.0 55.6 55.6 (*) Self-respondents-- 119 12 45 89.1 (*) 35.6 (*) Relatives--==------ 146 6 27 - 87.7 (*) 88.9 hho: 3 39 Table 8. Percent of possibly chronic conditions inferred from Med 10's reported on household in- terview by relationship to respondent, sex of respondent, and class of condition Number of conditions in- ferred from Med 10's Percent correspondingly re- ported on household interview Relationship to respondent and sex of respondent Checklist Checklist Non- Non- Un- Qualified | checklist Un- Qualified | checklist qualified | c1ags 2 Class 3 [qualified | 1.65 2 Class 3 Class 1 Class 1 All conditions Totalm====meomcmeeee eee 1.872 1,231 1,545 44,1 27.6 20.4 Male respondent-=-======e-e-- 560 333 369 46.6 27.6 21.7 Female respondent=-----ceca-- 1,308 895 1,169 43.1 27.7 19.8 Self-respondent Total-====m-mccecmeeeenm 878 605 739 47.7 35.2 21.1 Male-=====mmcomccc cece eee 299 184 142 52.5 29.9 19.7 Female-=-===mmccecccccccenaa- 579 421 597 45.3 38.2 21.4 Spouse Total-=====meemeecaeaao 462 283 343 46.8 22.6 21.0 Male respondent-=-==--=ccee--- 128 65 106 43.0 23,1 21.7 Female respondent=-==-=-=-====-- 334 218 237 48.2 22.5 20.7 Child Total-=---==ceccmceeenn 403 293 411 36.5 16.7 18.0 Male respondent=====---ceca--- 73 54 89 37.0 27.8 19.1 Female respondent=---=-==-=---- 330 239 319 36.4 14.2 16.9 Other relative Totale==cmecec ccc em 122 45 47 34.4 24.4 27.7 Male respondent=---=-====---- 60 30 32 36.7 23.3 37.5 Female respondent----=====n-- 62 15 15 32.3 26.7 6.7 40 Table 9. Percent of possibly chronic conditions inferred from Med 10's reported on household interview by relationship to respondent, sex of respondent, and grade of condition Number of conditions inferred Percent correspondingly reported on from Med 10's household interview All other All other Belasisnghin io Checklist | >1 service and >1 1 service [Checklist | >l service and >l 1 service Zeypen en go Sex no quali- month from lst to [only or 1 [no quali- month from lst to | only or 1 of responden fication last month or | fication last month or Grade I less from || Grade I less from Checklist Non- lst to Checklist Non- 1st to qualified | checklist last qualified | checklist last Grade II Grade III | Grade IV Grade II Grade III | Grade IV All conditions Total-=----- 1,872 296 397 2,083 44.1 41.6 30.7 19.7 Male respondent--- 560 104 114 484 46.6 34.6 18.4 23.8 Female respondent- 1,308 192 283 1,589 43.1 45.3 35.7 18.4 Self-respondent Total------- 878 162 187 995 47.7 47.5 28.3 24.3 Male---========--- 299 56 39 23 32.5 32.1 17.9 25.1 Female-=====-===-=--= 579 106 148 764 45.3 55.7 31.1 24.1 Spouse Total------- 462 68 105 453 46.8 29.4 38.1 16.8 Male respondent--- 128 20 46 105 43.0 30.0 19.6 21.9 Female respondent- 334 48 59 348 48.2 29.2 52.5 15.2 Child Total------- 403 47 96 561 36.5 31.9 29.2 14.3 Male respondent--- 73 13 21 109 37.0 38.5 19.0 21.1 Female respondent- 330 34 75 449 36.4 29.4 32.0 12.0 Other relative Total------- 122 19 9 64 34.4 37.9 (* 18.8 Male respondent--- 60 15 8 39 36.7 46.7 (*) 28.2 Female respondent- 62 4 1 25 32.3 (*) (*) 4.0 41 Table 10. Percent of possibly chronic conditions inferred from Med 10's reported on household in- terview by age and sex of person with the condition and class of condition Number of conditions in- ferred from Med 10's Percent correspondingly re- ported on household interview Age and sex Checklist Some Checklist Non- ae lad Qualified Tasokley: A Qualified Seite class 1 Class 2 Class 1 Class 2 All ages Total-=-m-m-mmmmmemmeae 1,872 1,231 1,545 44.1 27.6 20.4 Male--=-===ceccacmccccacnnna- 944 555 610 46.8 22.7 19.7 Female-=====ceccccccc ccc cece 928 676 935 41.4 31.7 20.9 Under 15 years Totale-mmemmemmeme———————— 344 191 326 36.0 17.3 17.5 Malg-======sm=seemcem———————— 214 91 173 39.3 9.9 18.5 Female======cccmcccc ccc ccc ee 130 100 153 30.8 24.0 16.3 15-24 years Totals mrarunnanuannnnne 81 112 96 38.3 19.6 10.4 Male====cmcccccc ccc ccc ccc cna 33 36 40 54 +5 19.4 10.0 TRIER Lm mo mm mm om wm me 48 76 56 27.1 19.7 10.7 25-44 years Total===c-ccmccccccaaaa 600 469 625 46.7 29.2 19.4 Male=======eemmeem————————— 252 181 188 48.4 20.4 19.1 Female~=======memcecemeanea=- 348 288 437 45.4 34.7 19.5 45-64 years Total-=====emeeeemamm=—- 707 381 455 44.8 32.3 25.5 Male====-cemccccc ccc ccc ccc mmm 359 202 179 47.9 29.7 22.9 Female----===c-cccmmmmmcmanon 348 179 276 41.7 35.2 27.2 65+ years Total-=-eecemc cece ccc 140 78 43 52.9 32.1 25.6 Male-=-ccmccc cece ccc ccc cc — 86 45 30 53.3 28.9 23.3 Female====-mccccccaccccccccaa 54 33 13 51.9 36.4 (*) 42 Table 11. Percent of possibly chronic conditions inferred from Med 10's reported on household in- terview by age and respondent status of person with the condition and class of condition Number of conditions in- Percent correspondingly re- ferred from Med 10's ported on household interview Checklist Checklist Age and sespondent v Non- 5 Non- stecus n- hecklist l= checklist qualified Qualities Ra 3 [qualified Qualigied Class 3 Class 1 | Class Class 1 ass All ages Total----===========-== 1,872 1,231 1,545 44.1 27.6 20.4 Self-respondents===-======-===~ 878 605 739 47.7 35.7 21.1 Relatives-=--=-====ccemca--=- 987 621 801 41.0 20.0 19.9 Under 15 years Total--=--=-=-========c-- 344 190 326 36.0 17.3 17.5 Self-respondents--=-==-====-=-=- - | 3 “s (*) (*) Relatives-----========c------ 340 189 319 36.5 17.5 16.9 15-24 years Total----=-============ 81 112 96 38.3 19.6 10.4 Self-respondents========--=== 22 25 26 22.7 24.0 0.0 Relatives-=---======ceccc---- 58 85 70 43.1 18.8 14.3 25-44 years Total-===========c===== 600 469 625 46.7 29,2 19.4 Self-respondents--====--==-==== 340 299 414 47 .4 37.1 20.3 Relatives=-==--===-==ceeo-n=-- 259 169 211 45.6 15.4 17.5 45-64 years Total---=--============ 707 381 455 44.8 32.3 25.5 Self-respondentg============= 421 226 271 46.3 35.4 23.2 Relatives-==--======-em=c=-o- 285 154 183 42.8 27.9 29.0 65+ years Total--===-============ 140 78 43 52.9 32.1 25.6 Self-respondents-=-=======-=== 95 54 25 61.1 35.2 24.0 Relatives--=-====c-=eoecc-—== 45 24 18 35.6 25.0 27.8 43 Table 12. Percent of possibly chronic conditions inferred from Med 10's reported on household in- terview by education of family head, respondent status, and class of condition Education of family Number of conditions in- ferred from Med 10's Percent correspondingly re- ported on household interview head and respondent Checklist Checklist status Non- Non- Un- checklist Un- checklist qualified | @allfied| “eyo "3" | qualified | Qallfied] Tp) oy Class 1 285 Class 1 age Under 9 years Total=====ecmccceanaana- 448 254 318 44.6 29,1 22.0 Self-respondents---====ccca-- 243 126 184 46.9 42.9 21.2 Relatives=====cecccmmccaaaaoo 205 128 134 42.0 15.6 23,1 9-12 years Total-=-===-ccmccccaca- 807 557 690 39.2 28.0 16.8 Self-respondents---------c--- 369 272 310 42.8 32.0 17.1 Relatives====-eccecccacoannono 434 282 376 36.2 24.5 16.8 12+ years Total-=====-=cemceunun- 571 370 490 49.9 25.9 22,2 Self-respondents----=-==-a--- 242 194 225 55.8 35.1 24.0 Relatives====c=cccccccaoaaaoo 326 174 264 45.7 16.1 20.8 44 Table 13. or older, reported on household interview by age class of condition Percent of possibly chronic conditions inferred from Med 10's, persons aged 15 years and education of person with the condition and Number of conditions in- ferred from Med 10's Percent correspondingly re- ported on household interview Age and education of person Checklist Non- Checklist ons with condition Un- Qualified checklist Non- Qualified checklist qualified Class 3 | qualified Class 3 Class 2 Class 2 Class 1 Class 1 All ages-15+ TOLAL mn mmm mm mm = mm e 1,528 1,040 1,219 45.9 29.5 21.2 Under 9 years--====-==-==---- 388 222 242 48.2 28.4 27.3 9-12 years----===========---- 690 521 586 41.0 30.1 17.4 12+ years----==-==-===-=------ 420 283 371 51.0 28.3 22.6 15-24 years Total-===-===========-== 81 112 96 38.3 19.6 10.4 Under 9 years---------=-====-=-=- 1 6 2 (*) (*) (*) 9-12 years==========-==-====-= 67 93 80 3.3 17.2 12.5 12+ years----==-========-=--- 13 13 14 (*) (*) (%) 25-44 years Total=-==-====ce======= 600 469 625 46.7 22.2 19.4 Under 9 years=-===--=-==-=-==---== 57 44 65 52.6 20.5 15.4 9-12 years-----=-=======------ 361 279 365 41.8 31.5 15.9 12+ years--=---==-=====------- 180 141 186 54.4 27.7 26.3 45-64 years Total-===-==-cc=cee=m==m 707 381 455 44.8 32.3 25.5 Under 9 years-==-=-=-=-----=----== 270 144 160 46.7 31.9 32.5 9-12 years------==-=========- 237 132 128 40.9 35.6 24.2 12+ yearg----==========-=----- 183 101 156 46.4 25.7 19.9 65+ years Total-====--=c========= 140 78 43 52.9 32.1 25,6 Under 9 years------========-- 60 28 15 50.0 28.6 26.7 9-12 years-----===========---- 25 17 13 48.0 35.3 (*) 12+ years----=---==-=-====---=-= 44 28 15 54.5 32.1 26.7 45 Table 14. Percent of possibly chronic conditions inferred from Med 10's reported on household in- terview by family income, respondent status, and class of condition Number of conditions in- ferred from Med 10's Percent correspondingly re- ported on household interview Family income and respondent status Checklist Non- Checklist Non- Un- checklist Un- checklist qualified Qualified Class 3 qualified Qualified Class 3 Class 2 Class 2 Class 1 Class 1 Under $4,000 Total=-==ecccmmccanaaa- 223 130 167 49.3 44,6 16.8 Self~respondents=====ceacmuax 125 88 108 59.2 56.8 12.0 Relatives==--ccccmcmccccaaano 98 42 59 36.7 19.0 25.4 $4,000-4,999 Total===--cemccccceaaae 292 182 186 46.6 19.2 22.6 Self-respondentg-====eeeeee-n 162 96 91 50.6 28.1 25.3 Relatives-=--ecmcmmcccccmeao 130 86 95 41.5 9.3 20.0 $5,000-6,999 Total===vcoccccccaaana- 577 437 546 43.5 27.7 22.3 Self-respondents--====eemeoa- 246 176 239 46.3 36.9 24.7 Relatives====eccccmmmmaaaaao 327 259 303 41.6 21.6 20.8 $7,000-9,999 TOLER Lv ~mwmmw swam ewie 380 234 316 44.5 30.8 17.4 Self-respondentg====mcecaeoo- 168 126 178 38.7 31.7 19.7 Relatives===--ecccmcaccaaaaa. 211 108 138 48.8 29.6 14.5 $10,000+ Total==-eemcmcmmcaaaa a 271 166 205 42.8 24.1 23.9 Self-respondentg-====vecnca-- 127 88 84 48.8 27.3 29.8 Relatives-==ecccccmmmccaeaana- 142 75 120 38.0 21.3 20.0 46 Table 15. tion Percent of possibly chronic conditions inferred from Med 10's report terview by number of H.I.P.-insured persons in household, respondent status, ed on household in- and class of condi- Number of conditions in- ferred from Med 10's Percent correspondingly re- ported on household interview Number of H.I.P.-insured persons (as of 6/30/57) Checklist Non- Checklist Non- and respondent status Un- checklist Un- checklist qualified | Qualified | Class 3 | qualified | Qualified] class 3 Class 1 Class 2 Class 1 Class 2 One person Total-=-=-----====ee==== 228 141 144 43.0 31,2 18.1 Self-respondents===========--- 151 94 95 45,7 34.0 14.7 Relatives=--==-=-=====emoemmo- 76 47 49 36.8 25.5 24.5 Two persons Total--=---====-e=====-= 554 295 365 45.3 33.9 23.8 Self-respondents===-=========- 335 192 236 47.8 38.0 24.2 Relatives=========c-ocecnu-u= 219 103 129 41.6 26.2 23.3 Three or four persons Total-===-==-=========o== 761 879 703 41.8 23.8 21s) Self-respondents=-====-======-== 305 252 277 47 +5 32,5 22.4 Relativeg-=--=-=======c==ce---- 454 325 423 38.1 17.2 20.3 Five or six persons Total-=-=--============== 279 176 269 49.8 24.4 18.2 Self-respondents==--=====-==-= 69 57 108 52.2 33.3 18.5 Relatives====--=====emmeemeo=- 208 117 160 49.5 20.5 18.1 Seven + persons TOLAL=e me mmm 50 40 64 40.0 37.5 7.8 Self-respondents============- 18 10 23 50.0 (*) 13.0 Relatives-=-=-=-cemeemcaaea=- 30 29 40 33.3 17:2 5.0 47 Table 16. Percent of possibly chronic conditions inferred from Med 10's reported on household in- terview by respondent status, indication of permission to review medical records, and class of condition Number of conditions Percent correspondingly reported inferred from Med 10's on household interview Respondent status and Checklist Checklist indication of permission Non- Non- to review medical records Unqualified | Qualified | checklist Unqualified | Qualified | checklist Class Class Class Class Class Class 1 2 3 1 2 3 All respondents Total======cececaaa- 1,872 1,231 1,545 44.1 27.6 20.4 Permission granted-------- 1,619 1,070 1,374 44.8 27.5 21.2 Permission not granted---- 253 161 171 39.5 28.6 14.0 Self-respondents Total-====mceeeaaa- 878 605 739 47.7 35.7 21.1 Permission granted------- 749 528 656 48.6 36.2 22.5 Permission not granted--- 129 77 83 42.6 32.5 14.5 Relatives Total-===-=ccccuaaa 987 621 801 41.0 20.0 19.9 Permission granted------- 863 537 713 41.7 19.2 20.6 Permission not granted--- 124 84 88 36.3 25.0 13.6 48 Table 17. Percent of possibly chronic conditions inferred from Med 10's reported on household in- terview by diagnostic category and class of condition Nurber of condicions Percent correspondingly inferred from Med 10's reported on household interview Diagnostic category . i el ie . Checklist Non- Checklist Non- ISC broad classification All check || All check NHS Recode #3 (NHS Recode #1) class- Unduel- Qual-| 3155 || c1ags- Unqual- | Qual- | ;; ified [ified | 1.49 ified | ified | -y.q5 es Class | Class 3 Li Class | Class 3 1 2 3X 2 Infective and parasitic diseases- 70 7 46 17 12.9 (*) 13.0 - 01 Tuberculosis, all forms---------- 4 4 - - (*) (*) 51 Infective and parasitic diseases NEC---==========-c=-----====- 66 46 17 12.1 (*) 13.0 - Dermatophytosis (039) ----=------ 46 - 46 - 13.0 €x) 13.0 - All other (005,029,038,040,041) 20 - 17 10.0 (*) — - Neoplasms==-»======-=cecwenmnmn-= 171 171 - - 23.4 23.4 02 Malignant neoplasmg----=======---- 33 33 - - 33.3 33.3 03 Benign and unspecified neoplasms- 138 138 - - 21.0 21.0 Uterus and other female genital organs (063,064)----------=--- 42 42 - - 40.5 40.5 Other (060-062,065-080)-------- 95 95 - - 12.5 12.5 Allergic, metabolic, endocrine, nutritional----=-=m==ee=-meeeu- 684 485 - 199 47.7 62.5 11.6 04 Asthma and hay fever------------- 269 269 - - 76.2 76.2 Asthma (082)---=--====--=mmm=-- 97 97 - - 71.1 71.1 Hay fever (08l)-------======--- 172 172 - - 79.1 79.1 05 Other allergies-----==========--- 125 125 - - 37.6 37.6 06 Diabetes mellitug~~~—r~wewnmnnmm~ 60 60 - - 61.7 61.7 with 58 ObeSity~==r=wm==rmocunwnnn sme 177 = - 177 9.6 oi 9.6 52 Endocrine, metabolic and nutri- tional diseases NEC----=--=---- 53 31 - 22 37.7 45.2 27.3 Diseases of thyroid (087-089)-- 31 31 - - 45.2 45.2 ew Other (091-096) ----=-------==--=-- 22 - - 22 27.3 27.3 Diseases of blood and blood-form- ing Oorganse=r-em===cceseaen- 52 - " 52 17.3 17.3 07 Anemia---===rmrmme=-——mreee———— 49 - - 49 18.4 18.4 53 Other--=--==-======------==-==-==---- 3 - 3 (*) (*) Mental, psychoneurotic, person- ality disorders-------------- 285 285 ~ - 20.4 20.4 09 Mental illness---=-------========- 214 214 - - 25.7 25.7 10 Ill-defined mental and nervous trouble-----=====c===-------- 7% 71 - - 4.2 4.2 Diseases of nervous system and Sense OrgaNS======-===wemm=== 506 31 105 370 22.7 35.5 39.0 17.0 54 Diseases and conditions of brain, spinal cord and nerves NEC, including impairments due to them, except paralysis------- 147 3 6 110 31.3 35.5 (*) 26.4 Vascular lesions of the central nervous system (107)--------- ds 15 - - 40.0 40.0 49 Table 17. Percent of possibly chronic conditions inferred from Med 10's reported on household in- terview by diagnostic category and class of condition—Continued Number of conditions inferred from Med 10's Percent correspondingly reported on household interview Diagnostic category Checklist Non- Checklist Non- ISC broad classification check check NHS Recode #3 (NHS Recode #1) | Unguale (Quals | 1400 a Cena | Tort { 14ss ified | ified Class ified ified Class es Class | Class 3 es Class | Class 3 1 2 1 2 54 Dis. and cond. of brain--Con. Sciatica, neuritis, and neural- gia (113-115) ===--ocmmcuanam- 80 - - 80 28.8 28.8 Other (108-110,112,116,227, X10-X19) =====cmeommmcnee eee 52 16 6 30 32.7 31.3 (*) 20.0 31 Impairment of vision------==----- 33 - 33 - 33.3 “a 33.3 cos 32 Impairment of hearing------------ 34 - 34 - 41.2 41.2 .ie 55 Diseases of eye and ear NEC------ 292 - 32 260 15.1 31.3] 13.1 Diseases of circulatory system--- 457 422 - 35 47.9 49.1 34.3 11 Heart disease-=-==------ccccccaaoo 162 162 - - 60.5 60.5 Chronic rheumatic heart dis- ease (128)------ccmccmccaaaa- 24 24 - - 54.2 54.2 Arteriosclerotic heart dis- ease (129)---=--cmcocennanon 91 91 - - 68.1 68.1 Hypertensive heart disease (133) 26 26 - - 73.1 73.1 Other heart disease (131,132)-- 21 23 - - 19.0 19.0 12 Hypertension without heart in- volvement-=-=====eemecocoaaanan 118 118 - - 45.8 45.8 13 Varicose veing---=-=----cccoooooo 52 52 - - 42.3 42.3 14 Hemorrhoids=====---ccccccmcaaonoo 76 76 - - 38.2 38.2 15 Other diseases of circulatory SyStem---------mmececeee———ean 49 14 - 35 32.7 (*) 34.3 Diseases of respiratory system--- 360 - 84 276 31.4 52.4] 25.0 16 Sinusitis--===--=cecccconoaaoo 64 - 64 - 48.4 48.4 17 Bronchitig======-mccocmconanmaoao 20 - 20 - 65.0 65.0 18 Other diseases of respiratory SyStem-=====-eememmmmcmememn 276 - - 276 25.0 25.0 Chronic tonsillitis (153)------ 80 - - 80 47.5 47.5 Chronic pharyngitis, naso- pharyngitis and laryngitis (154) ===mmmmmmmmmcm eee 76 - - 76 11.8 11.8 Other diseases of upper re- spiratory tract (156)-------- 43 - - 43 18.6 18.6 Pleurisy (157)---=--=--cceecan- 19 - - 19 - - Symptoms referable to respira- tory system (229)-=--==------ 51 - - 51 21.6 21.6 All other diseases of the re- spiratory system (159)------- 7 - - 7 (*) (*) Diseases of digestive system----- 422 160 163 99 35.5 58.8] 19.0 25.3 19 Ulcer of stomach and duodenum---- 60 60 - - 60.0 60.0 20 Hernia========c-ecccocaccmaoaaaaao 57 57 - - 54.4 54.4 21 Diseases of the gallbladder------ 33 33 - - 66.7 66.7 . 22 Constipation======-=cecccccaaaaoo 17 - 17 - - w - 23 Other diseases of the digestive system-----=----e-mcccecccea- 255 10 146 99 23.9 (*) 21.2; 25.3 Diseases of teeth, buccal cav- ity, esophagus (161,162) ----- 42 - - 42 4.8 ‘nin 4.8 Gastritis and duodenitis (164)- 47 - 47 - 6.4 6.4 re 50 Table 17. Percent of possibly chronic conditions inferred from Med 10's reported on household in- terview by diagnostic category and class of condition—Continued Number of conditions inferred from Med 10's Percent correspondingly reported on household interview Dizgnestle category Checklist Checklist ISC broad classification All NO All Bon NHS Recode #3 (NHS Recode #1) rE ual asl Coe | ciass~ una a re €s Class | Class [Class es Class | Class | Class 1 2 3 1 2 3 23 Other diseases of the digestive system Disorders of function of stomach (165) --===--=cmemeaa- 51 - 51 - 17.6 17.6 Chronic enteritis and ulcer- ative colitis (169)-----=---- 8 - 8 - (*) (*) Other functional disorders of intestines (171)--=--=------- 40 - 40 - 46.2 46.2 Symptoms referable to abdomen and gastrointestinal tract (233) ===-mmmmmmme mmm mmm ——— 17 - - 17 41.2 7 od 41.2 All other (173,174,178)-------- 50 10 - 40 42.0 (*) 40.0 Diseases of genitourinary system 349 38 33 278 21.8 47.4 3.0 20.5 24 Menstrual disorders----==--===-===- 40 - - 40 25.0 25.0 25 Menopausal disorders-----=-=-=-=---= 37 - - 37 29.7 29.7 26 Other diseases of genitourinary SyStem-=---==-=-mcmommonoon 272 38 33 201 20..2 47.4 3.0 17.9 Diseases of kidney and ureter (179,180,183) --memmmmmmmnnn—- 22 22 - - 54.5 54.5 Diseases of the prostate (184)- 16 16 - - 37.5 37.5 Other male genital, male breast (185,186) ---=--=--=--== 30 - - 30 13.3 13.3 Female breast conditions (187)- 21 - - 21 42.9 42.9 Diseases of the ovary, Fal- lopian tube and parametrium (188) -=-==-mmmmmmm mmm mmm 22 - - 22 22.3 27.3 Diseases of the uterus (189)--- 108 - - 108 12.0 12.0 Other diseases of the female genital system (192) --------- 19 - - 19 21.1 21.1 Symptoms referable to genito- urinary system (234)--------- 21 - 21 - 4.8 4.8 pp All other (194,X38)----=--=-=---- 13 - 12 (*) (*) (*) Diseases of skin and cellular tissue 446 - 446 - 19.5 19.5 27 Skin infections and diseases----- 446 - 446 - 19.5 19.5 Other dermatitis (not due to plants) (206) -==========eeucu-=- 132 - 132 - 21.2 21.2 Other diseases of skin (207) --- 314 - 314 - 18.8 18.8 Diseases of bones and organs of movement-=-=======mecemmee———— 72 255 354 162 33.7 34.1 36.7 | 26.5 28 Arthritis and rheumatism--------- 229 114 115 - 33.2 48.2 18.3 Arthritis, all forms (210)----- 114 114 - - 48.2 48.2 Tr Rheumatism (212) ---=--====-====- 15 - 115 - 18.3 I. 13.3 29 Back conditions----------=-=ccm-- 137 4 133 - 56.2 (*) 56.4 Displacement of intervertebral disc (213) =-mmmmmmmmmm—————— 5 - 5 - (*) (*) Nonparalytic orthopedic impair- ment back (X70,X71)-========- 128 - 128 - 54.7 54.7 Specified deformity of back (X80,X81)--=--mmmmmmm mmm 4 4 - - (*) (*) 51 Table 17. Percent of possibly chronic conditions inferred from Med 10's reported on household in- terview by diagnostic category and class of condition—Continued Number of conditions inferred from Med 10's Percent correspondingly reported on household interview Diagnostic category Checklist Checklist ISC broad classification 11 No All it} A chec ec NHS Recode #3 (NHS Recode #1 Unqual- | Qual- . | Unqual- | Qual- R 8 g ) class- “ified | ified iist || class=] “ified | ified 25s es Class | Class | © os 5 Class | Class ss 1 2 1 2 30 Other conditions of muscles, bones, and joints-----=------ 393 128 106 159 25.4 18.8 32.1 26.4 Nonparalytic orthopedic im- pairment, except of back (X73-X76) ==========uu- ——————— Il - 11 - (*) ren (*) Flatfoot (X82)---------cemoouon 70 70 - - 5.7 5.7 ‘oo Specified deformity, limbs or trunk (X83-X89)---=---=------ 55 55 - - 30.9 30.9 si Synovitis and bursitis (215)--- 120 - - 120 25.0 —_— 25.0 Symptoms referable to limbs and back (235)--=-=====mmemencann 95 - 95 - 31.6 & wd 31.6 pa All other (X31,214,216,217,251) 42 3 -]le 39 35.7 (*) wie 30.8 33 Paralysis of extremities and/or trunk--------m-mm meme eee 9 - - (*) (*) 3.500 57 Residuals of injuries NEC-------- 3 - - (*) % 58 (*) Congenital malformations--------- 18 18 - - 27.8 27.8 56 Congenital malformations--------- 18 18 - - 27.8 27.8 Symptoms and ill-defined condi- tions-=--=---mommmm meme eee 57 - - 57 24.6 24.6 08 Headache and migraine---=-=-------- 47 - - 47 14.9 14.9 59 Symptoms and ill-defined condi- tions NEC--=====---moccmaenan 10 - - 10 (*) (*) 52 Table 18. Percent of possibly chronic conditions inferred from Med 10's reported on household inter- view by Recode #3!categories ranked within each class by percent of conditions correspondingly re- ported and diagnosis and class of condition Conditions on National Health Survey checklist Without qualification (Class 1) With qualification (Class 2) Number Percent Number Percent Recode #3 category of con- | correspond- Recode #3 category of con- | correspond- ditions ingly re- ditions ingly re- inferred | ported on inferred ported on from Med | household from Med | household 10's interview 10's interview 04 Asthma and hay fever------ 269 76.2 17 Bronchitis----- 20 65.0 21 Diseases of the gall bladder---===-=cccaceaa- 33 66.7 (29) Back condi- 06 Diabetes mellitus--------- 60 61.7 tions--====~-- 133 56.4 11 Heart disease--========--- 162 60.5 19 Ulcer of stomach and 16 Sinusitis------ 64 48.4 duodenum-=-====cccemenu= 60 60.0 32 Impairment of 20 Hernia------=------cccoo-o 57 54.4 hearing------ 34 41.2 (28) Arthritis and rheumatism-- 114 48.2 31 Impairment of (26) Other diseases of genito- vision------- 33 33.3 urinary system-------=-- 38 47 .4 (30) Other condi- 12 Hypertension without heart tions of involvement-=--==-====--- 118 45.8 muscles, (52) Endocrine, metabolic, and bones and nutritional diseases NEC 31 45.2 jointg~»=m==- 106 32.1 13 Varicose veins------------ 52 42.3 (55) Diseases of 14 Hemorrhoids-----=-=====--- 76 38.2 eye and ear 05 Other allergies 125 37.6 NEC-=-==-====- 32 31:3 (54) Diseases and conditions (23) Other diseases of brain, spinal cord of the diges- and nerves NEC, in- tive system-- 146 21.2 cluding impairments due 27 Skin infections to them, except paraly- and diseases- 446 19.5 ER EEE EEE EEE EE 31 35.5 (28) Arthritis and 02 Malignant neoplasms-=--=-=-=-- 33 33.3 rheumatism--- 115 18.3 56 Congenital malformations-- 18 27.8 (51) Infective and 09 Mental illness---==-----=- 214 25.7 parasitic 03 Benign and unspecified diseases NEC- 46 13.0 neoplasms-===--=-=c-c-wc=n 138 21.0 (26) Other diseases (30) Other conditions of of genitouri- muscles, bones and nary system--- 33 3.0 jointg-===-ccemmccnaaa-- 128 18.8 22 Constipation-- 17 - 10 Ill-defined mental and nervous trouble--------- 71 4.2 Conditions not on National Health Survey Checklist (Class 3) Percen rre din Recode #3 category Number of conditions ay ay : inferred from Med 10's : interview (15) Other diseases of circulatory system---------- 35 34.3 25 Menopausal disorders--------ceeeccccmmccacanan 37 29.7 (52) Endocrine, metabolic and nutritional diseases NEC === mm mm mmm mmm mmm mmm meme een 22 27.3 (30) Other conditions of muscles, bones and joints- 159 26.4 (54) Diseases and conditions of brain, spinal cord and nerves NEC, including impairments due to them, except paralysig==-=--=c-cceccccacaaa- 110 26.4 (23) Other diseases of the digestive system-------- 29 25,3 18 Other diseases of the respiratory system------ 276 25.0 24 Menstrual disorders-------e-eccccccceccceaann- 40 25.0 07 Anemig-----=cc-ceccmceemmeemee ccc cceeeeaea 49 18.4 (26) Other diseases of genitourinary system-------- 201 17.9 08 Headache and migraine--=------ccccccacamcnnanx 47 14.9 (55) Diseases of eye and ear NEC----==-=-=cececea-- 260 13.1 58 Impairments NEC (predominantly obesity) ------- 177 9.6 (51) Infective and parasitic diseases NEC--------- 17 - IRecode No.3 categories within a given class of condition with less than 15 conditions inferred from the Med 10’s have been omitted from this table. { )Recode No. 3 category components of which have been assigned to more than one class of condition. 53 Table 19. Differentials in percent of Med 10 conditions reported on household interview by re- spondent status—Recode #3 categories! within each class of condition by magnitude of corre- spondence ratio between proxy- and self-respondents Number of conditions inferred from Med 10's Correspondence on household interview Class of condition; Reported Boia, ae diagnostic category (Recode #3) Sells total | gelf- spondence, Total | respond- Rela- respond- Rela- | relatives hts tives CHES tives to self- respondents Checklist without qualification (Class 1) 05 Other allergies--------=---- 125 39 81| 37.6 25.6 43.2 1.69 12 Hypertension without heart involvement-----=-cccceu-o 118 67 51 45.8 40.3 52.9 1.31 14 Hemorrhoids-----==cccmccaaaa 76 52 24 38.2 36.5 41.7 1.14 13 Varicose veins=----=-ccecaa- 52 29 23] 42.3 41.4 43.5 1.05 04 Asthma and hay fever-------- 269 85 183 | 76.2 77.6 76.0 0.98 20 Hernia-----c-cccccacaccmana- 57 20 37| 54.4 55.0 54.1 0.98 06 Diabetes mellitus------===-- 60 35 25] 61.7 68.6 52.0 0.76 09 Mental illness-=-=-==-==cecee-- 214 113 101 25.7 29,2 21.8 0.75 11 Heart disease-----=-=-=cc-o- 162 80 82 60.5 71.3 50.0 0.70 03 Benign and unspecified neo- plasms==--=c-cccmmmccaaaa 138 74 64 21.0 25,7 15.6 0.61 (28) Arthritis and rheumatism---- 114 61 53 | 48.2 59.0 35.8 0.61 19 Ulcer of stomach and duo- denum=----===--coccmcmacaao 60 37 23 | 60.0 73.0 39.1 0.54 (30) Other conditions of the muscles, bones and joints- 128 24 104 | 18.8 45.8 12.5 0.27 10 Ill-defined mental and nervous trouble-----===--- 71 44 27 4,2 6.8 0.0 0.00 21 Diseases of the gallbladder- 33 23 10 66.7 69.6 (*) (%) (26) Other diseases of genito- urinary system------------ 38 27 11 | 47.4 51.9 (*) (%) (52) Endocrine, metabolic and nutritional diseases NEC-- 31 22 9 45.2 45.5 (*) (*) (54) Diseases and conditions of brain, spinal cord and nerves NEC, including im- pairments due to them, except paralysis---------- 31 6 25 35.5 (%) 20.0 (%) 02 Malignant neoplasms--------- 33 14 19: 33.3 (*) 31.6 (*) Checklist with qualification (Class 2) (51) Infective and parasitic diseases NEC-------=-=c-n=- 46 25 21 13.0 8.0 19.0 2,38 27 Skin infections and diseases 446 168 276 19.5 22.6 17.8 0.79 32 Impairment of hearing------- 34 17 17 | 41.2 47.1 35.3 0.75 (29) Back conditions===========-= 133 88 45 56.4 63.6 42.2 0.66 (23) Other diseases of the digestive system-----====- 146 87 58 21.2 27.6 12.1 0.44 (30) other conditions of the muscles, bones and joints- 106 60 44 | 32.1 43.3 18,2 0.42 (28) Arthritis and rheumatism---- 115 51 64 18.3 29.4 9.4 0.32 16 Sinusitis------=-ccceceoana- 64 28 36 | 48.4 85.7 19.4 0.23 (26) Other diseases of genito-- urinary system------------ 33 18 15 3.0 5.6 0.0 0.00 17 Bronchitis=---=-cccceeaaaaao 20 5 15 65.0 (*) 723.3 (*) 31 Impairment of vision=------- 23 22 11 33.3 36.4 (*) (*) (55) Diseases of eye and ear NEC- 32 21 111 31.3 33.3 (*) (*) 54 Table 19. spondence ratio between proxy- and self-respondents--Continued Differentials in percent of Med 10 conditions reported on household interview by re- spondent status—Recode #3 categories! within each class of condition by magnitude of corre- Number of conditions Correspondence on household interview inferred from Med 10's Class of condition; Reporred Bette, pez- c “ diagnostic category (Recode #3) Self- vale | Tevet] sense Solu: | SROBAEI Total | respond- tives respond- tives | relatives ents ents to self- respondents Nonchecklist (Class 3) (23) Other diseases of the digestive system=-=-=--=--=--= 99 32 66 25.3 18.8 28.8 1.53 (30) Other conditions of the muscles, bones and joints- 159 59 100 26.4 22.0 29.0 1.32 58 Obesity (impairments, NEC)-- 177 104 73 9.6 9.6 9.6 1.00 (55) Diseases of eye and ear NEC- 260 101 158 13.1 13.9 12.7 0.91 18 Other diseases of respira- tory system------=---c--=- 276 77 196 | 25.0 29.9 23.5 0,79 07 Anemia------=-=--ccccaconao- 49 27 22 18.4 22.2 13.6 0.61 (54) Diseases and conditions of brain, spinal cord and nerves NEC, including im- pairments due to them, except paralysis---------- 110 66 44 26.4 31.8 18.2 0.57 (26) Other diseases of genito- urinary system--------=--- 201 141 60 17.9 20.6 11.7 0.57 08 Headache and migraine------- 47 22 25 14.9 31.8 0.0 0.00 (15) Other diseases of circu- latory system------------- 35 14 2) 34.3 (%*) 47.6 (*) 25 Menopausal disorders-------- 37 28 9 29.7 21.4 (*) (%) (52) Endocrine, metabolic, and nutritional diseases NEC-- 22 19 3 27.3 26.3 (*) (*) 24 Menstrual disorders--------- 40 29 11 25.0 31.0 (*) (*) 1 Categories with less than 15 conditions in both self-respondents and relatives of respondents have been omitted from this takle. O Recode #3 category components of which have been assigned to more than one class of condition. 55 Table 20. Percent distribution of all survey-reported conditions by question number producing household interview report—all possibly chronic conditions according to ISC designation—by class of condition and whether or not matched to conditions inferred from Med 10's Question number producing household interview report Class of condition and Total correspondence with Med 10's number | 1; 121] 13] 14 15 16 17 25 | Other Percent distribution All possibly chronic condi- tions---=-----ccccccmmnne oo 3,739 |10.0}0.3|2.0)14.3]23.8139.4] 7.5] 2.7 0.1 Matched to Med 10's----===-------- 1,481 ;15.510.5(10.,2] 19.9] 23.3 § 30.0 3.4{ 6,3 0.1 Unmatched to Med 10's-----=------- 2,258 | 6.3|0.1)2.7| 10.6 | 24.1 |45.6 10.1} 0.3 0.1 Checklist without qualification (Class 1)----=mcccccmcmcnaaa- 2,185) 7.6{0.2{0.8| 16.1 }122,7 (45.3 3.3 2.0 0.0 Matched to Med 10's-------=------- 826 (13.6 0.20.6 24.3 | 23.8|31.6| 0.8| 4.8 0.1 Unmatched to Med 10's====-=mcecau- 1,359 4.0/0.2|0.9| 11.0 22.0|53.6| 7.9] 0.3 - ‘hecklist with qualification (Class 2)-===cc-cccmmccccacnan 898 | 9.4 -|13.5] 6.2|25.6 (41.1 13.3 1.0 - Matched to Med 10's-=====cccacaaan 340 | 15.9 -{1.8(10.3(29.7|33.8| 6.8 1.8 - Unmatched to Med 10's----==------- 558 5.4 -|4.5| 3.8[23.1(45.5|17.2| 0.5 - Nonchecklist (Class 3)-=---=--- 656 | 18.6 | 0.8 | 4.1 | 19.4 | 24.8 | 17.7 7.0 7.2 0.5 Matched to Med 10"s-=-=cecmemeaa-x 315 20.31.6| 1.0] 18.7 { 14.9] 21.9 6.71] 14.9 - Unmatched to Med 10's-==-=ccceeaa- 341 (17.0 -17.0[19.9(34.013.8| 7.3 - 0.9 Question 11. Were you sick at any time last week or the week before? 12. Last week or the week before did you have any accidents or injuries, either at home or away from home? 13. ‘Last week or the week before did you feel any ill effects from an earlier accident or injury? 14. Last week or the week before did you take any medicine or treatment for any condition (besides ... which you told me about)? 15. At the present time do you have any ailments or conditions that have continued for a long time? (If ““No’’) Even though they don’t bother you all the time? 16. Has ... had any of these conditions during the past 12 months? (Card A) 17. Does ... have any of these conditions? (Card B) 25. During the past 12 months has ... been a patient in a hospital overnight or longer? 56 Table 21. Medical care reported in relation to conditions possibly chronic conditions according to ISC designation—by class of condition and whether not matched to conditions inferred from Med 10's reported on household interview—all or Percent of survey-reported conditions Last seen in Total | Medically Last seen stud ear b Class of condition number aptended by doctor? eto re eo and correspondence of ever ified status’ with Med 10's condi- tions Within | Before Yes No | study | study | H.I.P. | Non- year year H.1.P. All possibly chronic conditions----- 3,739; 91.4; 7.9 58.8 19..7 44.3 12.6 Matched to Med 10's-=-------=-ccccmmuaan= 1,481] 97.6 1.7 74.8 6.1 65.6 2.2 Unmatched to Med 10's------====-eccceeana- 2.258) 87.4] 12.0 48.3 28.6 30.3 16.2 Checklist without qualification (Class 1)-- 2,185( 90.9 8.3 62.3 19.5 47.5 12.9 Matched to Med 10's---=------ccccceemanm- 826 98.4 0.8 81.4 6.8 73.8 5.9 Unmatched to Med 10's-----=-=-cccceeacaaa- 1,359] 86.4 | 12.9 50.8 27.2 31.5 17.1 Checklist with qualification (Class 2)----- 898 | 90.9 8.9 52.3 24.4 37.3 13.0 Matched to Med 10's--------=-ccomcomaonan 3401 97.9 1.8 72.9 7.1 59.7 10.3 Unmatched to Med 10's--=--------cccmeauan- 558 | 86.6 | 13.3 39.8 34.9 23.7 14.7 Nonchecklist(Class 3)------=-----meeceenman 656 | 93.9 5.0 55.8 14.0 43.1 11.1 Matched to Med 10'g----~-----cccmcemmaan- 315] 95.2 3.8 59.7 3.5 50.5 7.9 Unmatched to Med 10's-----=-==cccemeeea-- 3411 92.7 6.2 52.2 23.8 36.4 15.0 I percent for which®fact of medical attendance was unknown or unreported is not shown in table. 2 Percent last seen by doctor within study year, plus percent last seen by doctor before study year, plus percent for which date of last doctor contact was unknown or unreported, not shown in table, equal total medically attended conditions. 3 Percent last seen in study year by H.L.P. doctor, plus percent last seen in study year by non-H.I.P. doctor, plus percent last seen in study year by doctor of unknown H.I.P. status, not shown in table, equal total conditions last seen by doctor in study year. 57 Table 22. in two weeks preceding household interview—all possibly Percent of conditions reported as producing disability, chronic conditions according to ISC bed disability, and time lost designation —by class of condition and whether or not matched to conditions inferred from Med 10's Percent of conditions Total Disability Bed-disability Time lost Class of condition and number Un- correspondence with of Kk oh Med 10's condi- Hom tions Yes No Un- Yes No Un- Yes No or in- known known appli- cable! All household interview- reported conditions------- 3,739 8.7] 87.7 3.6) 4.6) 91.6 3.8] 3.2] 0.6 96.2 Matched to Med 10's-======-= 1,481 | 11.2 | 81.7 71 6.2] 86.5 7:3 5.0 0.9 94.1 Unmatched to Med 10's------- 2,258 7.1]91.6 1.3 3.6} 95.0 1.57 .2.0 0.4 97.6 Checklist without qualifica- tion (Class 1)---=======n- 2,185 6.6 90.2 3.2 2.0] 93.5 3.5 2.6 0.3 97.1 Matched to Med 10's-=====--= 826 9.8 | 84.5 5.7 4.5] 89.5 6.1 4.4 0.6 95.0 Unmatched to Med 10's------- 1,359 4.6 193.7 1.7 2.1] 96.0 2.0 1.5 0.1 98.4 Checklist with qualification (Class 2)-====-c-cmmmnenax 898 | 11.2 | 87.5 1.2 6.31 92.4 X.2 3.9 0.3 95.8 Matched to Med 10'g-======-- 340 | 15.0 | 82.6 2.4 9.41 88,2 2.4 7:1 0.6 92.4 Unmatched to Med 10's------- 558 9.0 90.5 0.5 4.5] 95.0 0.5 2.0 0.2 97.8 Nonchecklist (Class 3)----=---- 656 | 12.3 | 79.6 8.1 7.81 84.1 8.1 4.3 1.8 93.9 Matched to Med 10's--=--=-=-- 315{10.8 7 73.3 15.9 1:3} 76.8 15.9 4.4 1.9 93.7 Unmatched to Med 10's------- 341 | 13.8 | 85.3 0.9 8.21 90.9 0.9 4.1 1.8 94.1 L4Time lost” is inapplicable if no disability was associated with condition, if person (adult) with condition would not have been working or going to school, or if person with condition was less than 6 years of age. 58 Table 23. All household survey-reported conditions coded '"chronic' by National Health Survey by diagnosis reported on household interview and by whether or not matched to conditions inferred from Med 10's Matched Unmatched . to Med 10- | to Med 10- Survey-reported diagnosis (Recode #3) Total tnferred inferred conditions | conditions All household survey-reported conditions coded chronic- 3,523 1,275 2,248 01 Tuberculosis, All formawerwmmmmmeims mimeo mm mm meme meee - 12 - 12 02 Malignant neoplasmg--==-=-=-==-=====-==-=--=-oe-=--==--=====-- 5 4 1 03 Benign and unspecified neoplasmg--=-===-===--===-c-c-----=== 45 24 21 04 ASEIE SNA DAY LONE www mmm mmom cures om co sn wn 0 1 470 228 242 05 OLher BLLETELle Sem mmm mim mim comm wr wm at ww wm mw 0 03 0 0 161 64 97 06 Diabetes mellitug-========m===-------eccece=ooocoocooooo-- 53 37 16 07 Anemia--=-----=====meec-memm---coese-osssssoooooooossosssss 17 7 10 08 Headache Gnd MUGT ALINE ww mm mmm mis wo sno om 0 8 0.0 8 ke 0 60 10 50 09 Mental illness~=~ po : . to cut N in col.(c) - record respondent’s| . vr a and 6 years | (d-1), circle *X” without in cols. (d-1) through Na |Quesfror es in (d-1) circle “X” without : . > 2 down on of [tion |p oe description) ashing the question) old or over, | asking the question) (d-4), circle "X” without | your usual S per{ Na R 2s (If ill-effects of earlier acci- 2 7 ask): asking the question) activities wed > f much] y [son dest lsat Tonle.) (If accident or injury, fill |. ow EB For an accident or injury oc- Table A) an you roy ~ curring during past 2 weeks, ask: read oe = What part of the body was hurt? ordinary ( ore What kind of injury was it? owl Pee ise Anything else? pride wit : 8 glasses? col. [-» (Also, fill Table A) (k)) @) | ®) | (c) (D-1) (d-2) (d-3) (d-4) (d-5) (e) |(f) [J¥es [] Yes X 1 CLs X J ne X X Table Il - HOSPITALIZATION DURING PAST 12 MONTHS Col When.did How amany Tonle ames: What was the matter? es : you enter the | days were How many of How many of Was this person hi : ’ : g {Non [Ques hospital? you in the th d y these--days still in the Anything else? ; 2 | of 160 ese-days [o_o during hospital last (Record each condition in same detail as called for in o | per- | No. h iy were in the |¢he pase 2 night? Table I. If condition is result of accident or injury, .£ | son (Month, Year)f not counting past 12 weeks, ending (Verify that no also fill Table A) i the day you months? last Sunday hosp. days after > : (a) (b) (©) leith oy (e) (f) Sere in col. d (h) 1 Mo. [J All or [J Yes Year Nays Days Days [J No Card A NATIONAL HEALTH SURVEY Check List of Chronic Conditions 1. Asthma 14. Stomach ulcer 2. Any allergy 15. Any other chronic stomach trouble 3. Tuberculosis 16. Kidney stones or other kidney trouble 4. Chronic bronchitis 17. Arthritis or rheumatism 5. Repeated attacks of sinus trouble 18. Prostate trouble 6. Rheumatic fever 19. Diabetes 7. Hardening of the arteries 20. Thyroid trouble or goiter 8. High blood pressure 21. Epilepsy or convulsion of any kind 9. Heart trouble 22. Mental or nervous trouble 10. Stroke 23. Repeated trouble with back or spine 11. Trouble with varicose veins 24. Tumor or cancer 12. Hemorrhoids or piles 25. Chronic skin trouble 13. Gallbladder or liver trouble 26. Hernia or rupture MEDICAL CARE--Continued 20. If “No” to q. 18a, ask: How long has it been since you last talked to a doctor? Mos. or Yrs. [] Less than 1 mo. [ [Never 21. Do you have a doctor you USUALLY go to? [1] Yes J No If "Yes" (b) What is his name and address? (Full name and street address, borough or town Enter State if outside New York) TT 23. How long has it been since you went to a dentist? Mos. or Yrs [[JLess than 1 mo. [[JNever HOSPITAL CARE 25. (a) DURING THE PAST 12 MONTHS has anyone in the family been a patient in a [JYes (Table II) [|No hospital overnight or longer? If “Yes”: TTT TT-T-TTT= === (b) How many times were you in the hospital? No. of times 26. (a) During the past 12 months has anyone in the family been a patient in a nursing [] Yes (Table II) []No home or sanitarium? If "Yes": [TTT -— == (b) How many times were you in a nursing home or sanitarium? No. of times 27. During the past 12 months in which group did the total income of your family fall, that is Group No. your's, your --'s, etc.? (Show Card H) Include income from all sources, such as wages, salaries, rents from property, pensions, help from relatives, etc. Table 1 - ILLNESSES, IMPAIRMENTS AND ACCIDENTS If 6 years old Did you first notice ... To inter- | . . ; How How man er, ask; | DURING THE PAST 3 viewer: Did you first When did | what is the doctor's name Have you many | of these As: MONTHS or before that time? NOLICE .;: you last | 554 address? talked to days | days | weekor | faye {If Col. DURING THE [talk to a any other includ-| were you the week| | Check one | Did ... start (k) is PAST 12 doctor doctors ing in bed betiny i): during the past Sheghied MONTHS or about ...? | (Enter full name and street addresgabout ... the 2 | all or you have Before [During| 2 Weeks or ald before that and borough or town. Enter State [during the week- | most of | peep How miznyt ~ g 3 | before that condition time (Month and] jf oueside New York) past 12 ends? | the day? working days did | monchs {months | time? is on year = months? ats jop | keep either one | (If during past |Year only or busi- | you from {35 (If during past |of Cards |12 months, ask):|if prior to a work a 2 weeks, ask): [A or B, 1956) (going to (going to |€O! Which week, last| continue; | Which month? school) | school)? (n)) week or the otherwise; Freep) week before? STOP (8) (h) (i) (j) (k) (1) (m) (aa) (n) (0) (X) (Y) —Days —— DAYS [JLast[ | Before Mo. [Mo. [No Dr T Yes] No ree or Yes or 2 wks. WT cimmciscsnane, NE commer TF mt 1 i gee ef et ee Days |[J None [[] No [J None [JWeek before [Before []Birth| [JNo Dr. _|CINo Dr. Table II - HOSPITALIZATION DURING PAST 12 MONTHS Were any operations performed on you during What is the name and address of the hospital you 3 this stay in the hospital? were in? 2 If “Yes”: . (Enter name, borough or town and State, if outside New York) y (a) What was the operation? = (b) Any other operations? (i) G) [J Yes 5 [J] No 1 Card B NATIONAL HEALTH SURVEY Check List of Impairments Deafness or serious trouble with hearing Serious trouble with seeing, even with glasses Condition present since birth, such as cleft palate or club foot Stammering or other trouble with speech Missing fingers, hand, or arm Missing toes, foot, or leg Cerebral palsy Paralysis of any kind Any permanent stiffness or deformity of the foot or leg, fingers, arm or back CPN Suwon 71 APPENDIX II: DETAILED DIAGNOSTIC TABLES Table A. Specificity of match and duplication of match— percent distribution of household inter- view-reported conditions in correspondence with Med l0-inferred conditions by type of match, and by number of Med 1l0-inferred conditions to which household interview report was matched, each class of condition and diagndstic category: Total Number of other furdber Type of satel? Med 10 conditions Class of condition and diagnostic category of matched by household (recode #3) condi- interview report tions 1 2 3 None z 2 3 All Med 10-inferred conditions matched by household interview reports------ 1,48. % $).5¢ 11.3 37.2% 86.1 )13.310.4| 0.2 Checklist without qualification (Class 1)-------mcmcmmcceeea 826 63.3 11.0 [25,75 89.6 9.80.4) 0.2 03 Benign and unspecified neoplasms-------- 29 41.41 20.7 (37.9 75.9117.2| 3.4 3 04 Asthma and hay fever----------coccecaooo 205 76.6] 13.2 | 10.2 94.6 5.4 - - 05 Other allergies-====-===ceccoaooonnaaoao 47 3 27.7 -172.3| 85.1 14.9 - - 06 Diabetes mellitus---------ceeccomcmcaaan 37 || 100.0 - - || 100.0 - - - 09 Mental illness---=-=-cccmccmmmmcmcccaaa 55 18.2 - 181.8 80.0(18.2| 1.8 - 11 Heart disease--=-=-=--c-cocmcmmcmcceeaoo 98 43.9129.6 26.5 98.0 2.0 - - 12 Hypertension without heart involvement- 54 || 85.2 -114.8) 87.0}|11.1] 1.9 - 13 Varicose veing-------ccccmmccmccccnaaaa- 22 | 86.4 13.6 77.3 {22.7 - - 14 Hemorrhoids------==-cccccmmcc cee 29 || 100.0 - - 96.6 3.4 - - 19 Ulcer of stomach and duodenum----------- 36 72.2 - 127.8 83.3 [16.7 - - 20 Hernia-=====-cccmcmmm meee eee 31 || 100.0 - - | 100.0 - - - 21 Diseases of the gallbladder------------- 22 72.7118.2 7 9.10) 95.37 4.3 - - (26) Other diseases of genitourinary system-- 18 | 55.6 22.2 (22.2 94.4 | 5.6 - - (28) Arthritis and rheumatism----===ccceecaaax 55 69.1 5.51.25.5 83.6 | 16.4 - - (30) Other conditions of muscles, bones and joints=mmmm meen meee 24 29.2 | 41.7] 29.2 75.0 125.0 - - Checklist with qualification (Class 2)-- 340 34.7 {15.3 |.50.0 83.5 (16.5 - - 16 Sinusitis-=----cccccccmmcceeooo 31 74.2 - 125.8 87.1)12.9 - - (23) Other diseases of the digestive system-- 31 6.5 |48.4 | 45.2 87.1112.9 - - 27 Skin infections and diseases---=--=----- 87 19.5 8.0] 72.4 88.5 | 11.5 - - (28) Arthritis and rheumatism------=cece-ceeu-- 21 4.8 | 42.9 | 52.4 66.7 | 33.3 - - (29) Back conditiong------=ceccccmmmmoanaaoo 75 | 44.0 9.3 146.7 86.7 | 13.3 - - (30) Other conditions of muscles, bones and RE BoE ol EE atatatate 34 14.7 | 14.7 | 70.6 82.4 | 17.6 - - Nonchecklist (Class 3)---=--mcmcccaa_ oan 315 38.7 7.6 153.7 79.7 119.0 1.0] 0.3 (54) Diseases and conditions of brain, spinal cord and nerves NEC, including impair- ments due to them, except paralysis--- 29 6.9 3.4 | 89.7 62.1 | 37.9 - - (55) Diseases of eye and ear NEC----=====c=-= 34 38.2 8.8 | 52.9 64.7 | 35.3 - - 18 Other diseases of respiratory system---- 69 | 56.5 | 8.7 |34.8|| 85.5 |14.5 - - (23) Other diseases of the digestive system-- 25] 20.0 | 16.0 [64.0 96.0 | 4.0 - - (26) Other diseases of genitourinary system-- 36 | 27.8 (13.9 (58.3 72.2 (19.4 |5.6 | 2.8 (30) Other conditions of muscles, bones and Joints==-mmme meee 42 || 45.2 | 4.8 | 50.0 90.5 | 9.5 - - 58 Impairments NEC---=--=c--ccccccmmaaaaann 17 | 29.4 70.6 || 82.4 | 17.6 - - lrecode #3 categories within a given class of condition with less than |5 conditions reported on household inter- view in correspondence with Med |0-inferred conditions have been omitted from this table. 2pefinition, type of match: Type | — Survey—reported condition falls into the same recode #| category as the Med [0 diagnosis. Type 2 — Survey-reported condition falls into the same recode #3 category as the Med |0 diagnosis, but not into the same recode #| category. Type 3 — Survey—reported condition or symptom is consistent with or associated with the Med I0 diagnosis, but is not codable to the recode #| or #3 category to which the Med |I0 diagnosis belongs. ( ) Recode #3 category components of which have been assigned to more than one class of condition. 12 Table B. Comparison of frequencies of specified diagnostic categories, physician's diagnoses, and respondent diagnoses—all possibly chronic conditions inferred from Med 10's for which condi- tions were correspondingly reported on household interviews, coded chronic by National Health Survey, ranked by magnitude of ratio between number from respondent and number from physician, each diagnostic category! Number of conditions in specified Ratio, household Diagnostic category (recode #3) category according interview fre- [Rank to diagnosis from quency to Med 10 frequency Med 10's | Household interview 10 I1l-defined mental and nervous trouble--=--=---- 3 22 7.33 1 25 Menopausal disorders----------cccemcmmmncnanan 8 20 2.50 2 15 Other diseases of circulatory system---------- 10 20 2.00 3 28 Arthritis and rheumatism-----=====mememeeeen——" 73 137 1.88 4 32 Impairment of hearing--------=--=emeeececenaa- 14 24 1.72 5 "52 Endocrine, metabolic, and nutritional diseases NECrmmmrmmm ee ————————————— e n m n e e m 16 27 1.69 6 05 Other allergies------=--e-c-ececcc cece 46 64 1.39 7 12 Hypertension without heart involvement-------- 51 70 1.37 8 16 Sinusitis-==--=----mmeemee cece 28 38 1.36 9 14 Hemorrhoids---=====mmrcceccccac ccc cence ——— 26 35 1.35 10 03 Benign and unspecified neoplasms----==-==c--=- 18 24 1.33 11 13 Varicose veins----==--cmccceccccnc cece een 19 23 1.21 12 20 Hernia----======-mcecccc ccc cece mcm 23 27 1.17 13 04 Asthma and hay fever------------cceccccceeea-- 204 228 2.12 14 06 Diabetes mellitug-~===wmmeswmmmreemnennmm nn 37 37 1.00 15 21 Diseases of the gallbladder------=---c-cce---- 20 20 1.00 15 19 Ulcer of stomach and duodenum---------=oc-c--- 34 31 0.91 16 35 Other chronic conditions----=-=---ceccccceccun-- 129 108 0.84 17 ‘54 Diseases and conditions of brain, spinal cord and nerves NEC, including impairments due to them, except paralysis----=--cccccceee-a- 42 33 0.79 18 26 Other diseases of genitourinary system-=---=---- 37 29 0.78 19 11 Heart disease-------==-cmecmcccccccc ccc 93 71 0.76 20 29 Back conditions---==-=--e-ccccmc cece 70 50 0.71 21 30 Other conditions of muscles, bones and joints- 85 50 0.59 22 18 Other diseases of respiratory system-=----=----- 35 20 0.57 23 “55 Diseases of eye and ear NEC--===--=-ccencconnn 39 22 0.56 24 23 Other diseases of the digestive system-------- 45 21 0.47 25 27 Skin infections and diseases---------ccceoaaa- 67 30 0.45 26 09 Mental illness-=-=====--ecceccccecocccaccc cna 45 17 0.38 27 omitted are diagnostic categories with less than |5 conditions from both physician-source and respondent-source. * Subdivision of category 35 of recode #3. 73 APPENDIX II SAMPLING Most statistics in the study are combined ratio esti- mates of the form r = X' where X' and Y' are estimates of universe Y! aggregates. In many cases this will be the proportion of conditions of a specified type reported on household in- terview. The appropriate statistical model for variance estimation is, therefore, a stratified sample of families with a combined ratio estimate statistic. The estimating formulas used are fully discussed in Section 4, Chapter 5, of Sample Survey Methods and Theory, Volume I, Hansen, Hurwitz, and Madow, and in other modern sta- tistics textbooks. Modern electronic processing equipment (UNIVAC) was used to accumulate the data necessary for variance estimation and to perform the necessary computations. A general UNIVAC program was supplemented by a series of short instruction programs which specified the variable or variables to be processed. This specifica- tion usually required about five minutes of programing time for each variance. Using this method, it was pos- sible to produce variances for those variables which seemed most useful for such examination as indicated by the basic punch card tabulations. Although a large number of variances were com- puted for use in specific areas of the analysis, those shown in the following table are sufficient to indicate the ranges of values commonly encountered in the study. In general, the magnitudes of these were satisfactory, making it possible to consider differences having rel- variances of less than one percent for many groups of interest. Correspondence in reporting on household interview and variances, selected classifications Number of Proportion chronic reported on Variance of Chronic conditions conditions household proportion in sample” interview *¥eported Class 1 among male respondent§=====-=ce-ceceaanax 560 L465 .000686 Two weeks or less between last service and household interview-===--ccccmmmmmmccceeaas 457 .575 .000880 Ten or more related Med-10 services during Study year=-==-==--ece meena 361 «795 .001295 Asthma and hay fever-=------ccmcmmmmcoccccceeeaan 269 .762 .001008 Diabetes mellitug===-===-ccccmm mma 60 .617 .008008 Menopausal disordersg-===-----ccmmmcmcaccccceaaan 37 +297 .012847 Proportion of people Number of reporting Variance of Persons persons in doctor proportion sample” contact in reporting corresponding period Self-respondents age 45 and over seen by H.I.P. physician during two-week period prior to household interview-====----commmmocccceaa 167 +545 .002690 Persons in families in which the family head completed less than 9 years of schooling, seen by a H.I.P. physician during study year--- 1,013 +753 .000521 Number of Pe ” o 14 oportion Hospitalization episodes 2 Spleen reported on Variance of household proportion episodes reported in sample® interview Among males for whom female responded-=--=======- 94 .862 .001640 Among females for whom male responded--=-====a--= 31 .805 .012494 *Replicated to give each unit equal weight. 74 U. S. GOVERNMENT PRINTING OFFICE : 1965 O - 776-144 REPORTS FROM THE NATIONAL CENTER FOR HEALTH STATISTICS Public Health Service Publication No. 1000 Series 1. Programs and collection procedures No. 1. Origin, Program, and Operation of the U.S. National Health Survey. 35 cents. No. 2. Health Survey Procedure: Concepts, Questionnaire Development, and Definitions in the Health Interview Survey. 45 cents. No. 3. Development and Maintenance of a National Inventory of Hospitals and Institutions. 25 cents. No. 4. Plan and Initial Program of the Health Examination Survey. Series 2. Data evaluation and methods research No. 1. Comparison of Two-Vision Testing Devices. 30 cents. No. 2. Measurement of Personal Health Expenditures. 45 cents. No. 8. The One-Hour Glucose Tolerance Test. 30 cents. No. 4. Comparison of Two Methods of Constructing Abridged Life Tables. 15 cents. No. 5. An Index of Health: Mathematical Models. No. 6. Reporting of Hospitalization in the Health Interview Survey. No. 7. Health Interview Responses Compared With Medical Records. No. 8. Comparison of Hospitalization Reporting in Three Survey Procedures. No. 9. Cooperation in Health Examination Surveys. No. 10. Hospital Utilization in the Last Year of Life. Series 3. Analytical studies No. 1. The Change in Mortality Trend in the United States. 35 cents. No. 2. Recent Mortality Trends in Chile. 30 cents. Series 4. Documents and committee reports No reports to date. Series 10. Data From the Health Interview Survey No. 1. Acute Conditions, Incidence and Associated Disability, United States, July 1961-June 1962. 40 cents. No. 2. Family Income in Relation to Selected Health Characteristics, United States. 40 cents. No. 3. Length of Convalescence After Surgery, United States, July 1960-June 1961. 35 cents. No. 4. Disability Days, United States, July 1961-June 1962. 40 cents. ; No. 5. Current Estimates From the Health Interview Survey, United States, July 1962-June 1963. 35 cents. No. 6. Impairments Due to Injury, by Class and Type of Accident, United States, July 1959-June 1961. 25 cents. No. 7. Disability Among Persons in the Labor Force, by Employment Status, United States, July 1961-June 1962. 40 cents. No. 8. Types of Injuries, Incidence and Associated Disability, United States, July 1957-June 1961. 35 cents. No. 9. Medical Care, Health Status, and Family Income, United States. 55 cents. No. 10. Acute Conditions, Incidence and Associated Disability, United States, July 1962-June 1963. 45 cents. No. 11. Health Insurance Coverage, United States, July 1962-June 1963. 35 cents. No. 12. Bed Disability Among the Chronically Limited, United States, July 1957-June 1961. 45 cents. No. 13. Current Estimates From the Health Interview Survey, United States, July 1963-June 1964. 40 cents. No. 14. Illness, Disability, and Hospitalization Among Veterans, United States, July 1957-June 1961. 35 cents. No. 15. Acute Conditions, Incidence and Associated Disability, United States, July 1963-June 1964. 40 cents. No. 16. Health Insurance, Type of Insuring Organization and Multiple Coverage, United States, July 1962-June 1963. 35 cents. No. 17. Chronic Conditions and Activity Limitations, United States, July 1961-June 1963. 35 cents. Series 11. Data From the Health Examination Survey No. 1. Cycle I of the Health Examination Survey: Sample and Response, United States, 1960-1962. 30 cents. No. 2. Glucose Tolerance of Adults, United States, 1960-1962. 25 cents. No. 3. Binocular Visual Acuity of Adults, United States, 1960-1962. 25 cents. No. 4. Blood Pressure of Adults, by Age and Sex, United States, 1960-1962. 35 cents. No. 5. Blood Pressure of Adults, by Race and Region, United States, 1960-1962. 25 cents. No. 6. Heart Disease in Adults, United States, 1960-1962. 35 cents. No. 17. Selected Dental Findings in Adults, United States, 1960-1962. 30 cents. No. 8. Weight, Height, and Selected Body Dimensions of Adults, United States, 1960-1962. 9 No. 9. Findings on the Serologic Test for Syphilis in Adults, United States, 1960-1962. Series 12. Data From the Health Records Survey No reports to date. Series 20. Data on mortality No reports to date. Series 21. Data on natality, marriage, and divorce No. 1. Natality Statistics Analysis, United States, 1962. 45 cents. No. 2. Demographic Characteristics of Persons Married Between January 1955 and June 1958, United States. 35 cents. Series 22. Data from the program of sample surveys related to vital records. No reports to date. This report was originally published in the series ''Health Statistics from the U.S. National Health Survey,’ which has since been replaced by the "Vital and Health Statistics'' series. It presents findings from a methodological study pertaining to improved techniques in data collection in the Health Interview Survey. Because this material is of continuing importance, and is relevant to data currently being released from the Survey, the report is being reprinted in its present form. Public Health Service Publication No. 1000-Series 2-No. 8 For sale by the Superintendent of Documents, U.S. Government Printing Office Washington, D.C. 20402 - Price 40 cents. NATIONAL CENTER| Series 2 For HEALTH STATISTICS | Number 8 VITALand HEALTH STATISTICS DATA EVALUATION AND METHODS RESEARCH Comparison of Hospitalization Reporting in three survey procedures A study of alternative survey methods for collection of hospitalization data from household respondents. Washington, D.C. July 1965 U.S. DEPARTMENT OF HEALTH, EDUCATION, AND WELFARE Public Health Service Anthony J. Celebrezze Luther L. Terry Secretary Surgeon General NATIONAL CENTER FOR HEALTH STATISTICS FORREST E. LINDER, PH. D., Director THEODORE D. WOOLSEY, Deputy Director OSWALD K. SAGEN, PH. D., Assistant Director WALT R. SIMMONS, M.A., Statistical Advisor ALICE M. WATERHOUSE, M.D., Medical Advisor JAMES E. KELLY, D.D.S., Dental Advisor LOUIS R. STOLCIS, M.A., Executive Officer OFFICE OF HEALTH STATISTICS ANALYSIS Iwao M. Moriyama, Pu. D., Chief DIVISION OF VITAL STATISTICS RoBERT D. Grove, Pu. D., Chief DIVISION OF HEALTH INTERVIEW STATISTICS Puirip S. LAWRENCE, Sc. D., Chief DIVISION OF HEALTH RECORDS STATISTICS Monroe G. SirkEN, Pu. D., Chief DIVISION OF HEALTH EXAMINATION STATISTICS ArTHUR J. McDoweLL, Chief DIVISION OF DATA PROCESSING SiDNEY BINDER, Chief COOPERATION OF THE SURVEY RESEARCH CENTER, THE UNIVERSITY OF MICHIGAN AND THE BUREAU OF THE CENSUS Under legislation establishing the National Health Survey, the Public Health Service is authorized to use, insofar as possible, the services or facilities of other Federal, State, or private agencies. The methodological study in this report was performed under a contractual arrangement with the Survey Research Center, Institute for Social Research, The University of Michigan. The Bureau of the Census also participated actively in the planning and conduct of the research. Public Health Service Publication No. 1000-Series 2-No. 8 PREFACE This report is the second inthe U, S, National Health Survey's methodological series on the sub- ject of hospitalization reporting in the Health In- terview Survey, both of which were conducted by the Survey Research Center of The University of Michigan under contract with the U, S, National Health Survey and in co-operation with the Bu- reau of the Census. These studies are part of a program of the National Health Survey to evaluate the reliability of its statistics and to develop and test improved methods for collection of data. (Prior publications inthis developmental and eval- uation series are listed inside the back cover of this report.) The statistical design and procedures used in the Health Interview Survey of the U. S. National Health Survey are described in two Na- tional Health Survey publications.! ? ly. s. National Health Survey. The Statistical Design of the Health [Household-Interview Survey. Health Statistics. Series A-2. PHS Publication No. 584-A2. Public Health Service. Washington, D. C., July 1958. 2y. S. National Health Survey. Concepts and Definitions in the Health Housebold-Interview Survey. Health Statistics. Series A-3. PHS Publication No. 584-A3. Public Health Service. Washington, D. C., September 1958. The study was a co-operative project of the staffs of the Bureau of the Census, the Survey Research Center, and the National Health Survey, each organization actively participating in all phases of the study. The sample was designed by Harold Nisselson of the Bureau of the Census. Katherine Capt and George Kearns of the Bureau of the Census were responsible for the prepara- tion of interviewing manuals, training of inter- viewers, and general quality control of the field operations. An important contribution was also made by John Tharaldson, Edward Knowles, and John Campbell of the Detroit Regional Office of the Bureau of the Census, who helped in selecting the sample from the hospitals and carrying out the field procedures. Charles F, Cannell, Ph.D., and Floyd Fowler were the principal investigators for the Survey Research Center. In addition to developing a spe- cial experimental procedure and questionnaire for the collection of hospitalization data, they were also responsible for the report presented here. Earl Bryant of the U. S. National Health Sur- vey staff had the responsibility of co-ordinating the activities of the participating organizations and conveying the National Health Survey view- point in decisions on methodology. He also edited the contractor's report for the present publication. - - j- i - > B . n a . } B a ! - i . : > Ny - . | l F | : 1 } - 4 b a . - . > = fi - : . . sae Sm = ) . - oF Sg opt == 3 Earh . . a H A CONTENTS Preface-=--cmmemmc cece creer mem SUMMA Y === mmm mmm mm meme em mm meme em Objectives of the Study and Description of the Research Design======mememmmcmcmcmcme comme IntroduCtion=======cmmme ce emc emcee The Pilot InvestigationS=======mmmmmmeeoacacaaaaax The Sample Design-=====mmmmmmmmmoccccaecaccmeeee Description of the Procedures--------=====cccueou- The Interviewers--=-===e-c- cccmcmcecccme mem ———— Assignment of Interviewers-----------ccceeecaanan Follow-up Techniques--======cmaecmmcmoccaaaaaaa- Deviations From the Design---====cm=caaccaacaaao- Editing, Matching, and Coding-=--========mmccceaaou Comparison of Underreporting in the Three Procedures by Characteristics of the Sample Person-------====== Comparison of Underreporting in the Three Procedures by Characteristics of the Episodes-==mmmmmmmmmaaaaax Follow-up to Procedures B and C--------=--cccceuu-- Comparison of Underreporting in the Three Procedures by Characteristics of the Respondents-----====ceuu== Accuracy of Reporting Length of Stay and Date of Discharge------= ccc mmmmm meee cme me Detailed TableS-====-cmmmmm mmc cee mem Appendix I. Part 1: Sampling EXrors----==--ceceameomacmaoe —- Part 2: Analysis of Interview AssignmentS-=-==-=--=--- Appendix II, Forms and Questionnaires Used in the Procedure A Questionnair€----ceceecccccccccana- - Procedure B Questionnaires--------- --=-=ccco-ou- Procedure C Questionnaires==----========coc-ozo- Letter and Brochure----==ececmmmcmccccmcmcaaaano OOOO Ub NN 11 12 16 17 18 SYMBOLS AND NOTES Magnitude of the sampling error precludes showing separate estimates-------=-eauu- NOTE: Due to rounding detailed figures within tables may not add to totals COMPARISON OF HOSPITALIZATION REPORTING in three survey procedures The following research report was prepared by the Survey Research Center, Institute for Social Research, The University of Michigan, un- der contract with the National Health Survey Division, National Center for Health Statistics. Charles F. Cannell, Ph.D. and Floyd Fowler, of the Institute for Social Research, directed the project and were responsible for the analysis and the reporr presented here: Leslie Kish, Ph.D., provided guidance on statistical problems and was responsible for the variance analysis. Valuable assistance was also given by Thomas Bakker during the pilot investigations and by Mrs. Doris Muehl who supervised the editing and coding procedures. SUMMARY The objective of this research was to com- pare the effectiveness of two experimental pro- cedures with the standard Health Interview Survey of the U, S. National Health Survey procedure in obtaining information about hospital stays. Pro- cedure A, the control, used the standard Health Interview Survey (HIS) questionnaire and proced- ures. Procedure B was arevisedinterview sched- ule which was followed by a mail form in which any information about hospital stays thathad been overlooked in the interview was to be recorded by the respondent. Procedure C eliminated the ques- tions about hospitalizations from the interview; the requested information was to be entered on a self-administered form which was given to the respondent by the interviewer at the close of the interview. The follow-up forms in Procedures B and C were to be mailed to the Regional Office of the Bureau of the Census. The design of the study and interpretation of results must be judged in the context of primary purposes of the undertaking. Previous research had suggested a considerable variety of steps and techniques which might constitute improvements. The prime effort was to construct a total proced- ure which included a number of these potential improvements, and to test this procedure against the current standard. The key decision would be whether the new procedure was better than the old, with only secondary consideration being given to which of several factors were chiefly responsi- ble for any net improvement that should appear. This new or consolidated procedure was the one designated Procedure B. During the course of planning the study, the possibility arose that a more streamlined self-administered approach might yield most of the benefits hoped for from the consolidated procedure. Accordingly, Proced- ure C was included in the test. Thus the analysis puts primary emphasison over-all net effectiveness of the three procedures. It does not include comparative costs of the dif- ferent processes. Further, it is important to note that the total effect from Procedure Bis the prod- uct of arather intensive interview routine followed by a self-administered process; while the effect from Procedure C is the consequence of a self- administered process followed by a telephone and personal visit interview for a substantial number of nonrespondents. Care must therefore be taken in ascribing the cause for different results to any single feature of the procedures. For several reasons the study does not pro- duce a representative measure of underreporting, and Procedure A does not produce a valid esti- mate of the level of the underreporting errors for estimates shown in publications of the Health In- terview Survey of the National Health Survey. Prominent among their reasons are (1) restric- tion of the study to Detroit; (2) elimination of hos- pital episodes for deliveries, which previous studies have shown to be very well reported; and (3) the fact that NHS publications currently are based on a six-month-recall period. The net effect of these differences is an implied over- statement of underreporting by several percent- age points for NHS published data. A stratified sample was selected from Detroit hospitals of residents of the Detroit area whohad had one or more hospital stays during the year preceding the interviewing. Those whose only hos- pital stays were for normal deliveries were ex- cluded from the sample. The following are some of the significant findings of this study: The proportions of the known sample of hos- pital episodes which were not reported were 17 percent for Procedure A, 9 percent for procedure B, and 16 percent for Procedure C, The difference in the reporting in experimental Procedure Band the control Procedure A is significant at the 0.05 level of confidence. When apparent overreports were included, the rate of underreporting was decreased by two or three percentage points for each procedure. There was an increase in the underreporting rate for all three procedures asthe lengthof time between the hospital discharge and the interview increased. There was an especially sharp in- crease in underreporting for all procedures when the discharge preceded the interview by more than 40 weeks, However, the relationship was some- what weaker in Procedure B for episodes which occurred within 40 weeks of the interview. One-day stays were reported very poorly, with the underreporting rates being almost the same for all three procedures. For all other stays, however, the reporting in Procedure B showed marked improvement, For all three procedures the degree of social threat or embarrassment of the diagnosis leading to hospitalization was negatively related to the rate of reporting. Episodes which involved surgical treatment were reported significantly better in all three procedures than those which did not. There was a consistent relationship in all three procedures between the number of chronic and acute conditions reported for the sample per- son and the reporting rate; therate improved with an increase in the number of conditions. In all three procedures, the reporting for per- sons with three or more episodes in the sample was considerably poorer than for persons with only one or two. For all procedures, the underreporting rate was higher for nonwhite than for white persons. 2 In Proceaures B and C the reporting for per - sons in low income families was significantly poorer than it was for those in higher income families. The same pattern was found in Proced- ure A. In Procedure A, episodes for persons with higher education were reported somewhat better than those for persons with lower education, This bias is even more apparent in Procedure C, but is essentially eliminated by Procedure B. Respondents reported their own episodes con- siderably better than they reported the episodes of others in Procedure A, This tendency is re- duced in Procedure C and eliminated in Proced- ure B. A large proportion (30 percent) of the hos- pital episodes not reported in the direct interview for Procedure B was obtained in a mail-follow-up procedure. It was found that the promptness with which respondents replied to the follow-up was directly related to the quality of reporting in both Pro- cedures B and C. Month of discharge was reported equally well in all three procedures. Procedure C proved to be significantly better than Procedure A in obtaining correct reports of the number of days involved in hospital episodes. The most outstanding finding, of course, was the significant improvement of reporting found in Procedure B. In this improvement, one clear fac- tor was the better reporting for proxy-respond- ents; another was the reduction of underreporting for persons in the lower educational brackets. While it is not possible to specify the reasons for these improvements, several aspects of the procedure were designed to "motivate" respond- ents. As the study yielded considerable evidence that the level of motivation of the respondent is an important determinant of how well he reports, it is suggested that the success of Procedure B may be largely attributable to its effectiveness in encouraging and directing increased effort to re- port. OBJECTIVES OF THE STUDY AND DESCRIPTION OF THE RESEARCH DESIGN Introduction In 1959 the Survey Research Center, under contract with the National Health Survey, and working co-operatively with the National Health Survey and the Bureau of the Census, conducted a study which compared hospitalizations reportedin household interviews with those recorded in hos- pitals.? The purpose of the study was to estimate the magnitude of underreporting of hospital epi- sodes in the Health Interview Survey of the Na- tional Health Survey, to investigate some of the patterns of underreporting, and to develop hypoth- eses relating to the mechanisms of underreport- ing. For ease of reference the 1959 study is re- ferred to in this report as Special Study No. 8, which was used as the working title. Based largely upon the findings of Special Study No. 8, another study was carried out, de- signed to test new methods of collecting hospitali- zation data from household respondents. Thisre- port presents the analysis of the experimental study. Since Special Study No. 8 provided the basis for the development of new experimental tech- niques for improving the reporting of hospital data in the Health Interview Survey, a brief sum- mary of the hypotheses developed from the re- sults of Special Study No. 8 is given for back- ground information. In addition, proposed changes which were the basis for the development of ex- perimental procedures are described. In Special Study No. 8, respondents were asked to report hospitalizations which occurred during the 12 months preceding the Sunday night of the week in which the interview was conducted. Such a time period is confusing to the respondent and apparently creates problems of time refer- ence. The marked underreporting of episodes occurring near the beginning of the 12-month pe- riod suggested that when the respondent was in doubt he preferred to recall the episode as having taken place before the beginning of the year and, thus, did not report it. The proposed solution to this problem was to use as a reference period that partof the calendar year preceding the interview and the preceding calendar year. The analysis then could be based on the 12 calendar months preceding the month in which the interviewing took place. There is a lack of positive motivation on the part of respondents to devote the amount of energy required to report hospital episodes. To recall 3U. S. National Health Survey. Reporting of Hospitalization in the Health Interview Survey. Health Statistics. Series D-4. Public Health Service Publication No. 584-D4. Public Health Service. Washington, D. C., May 1961. hospitalizations over a period of one year re- quires the respondent to exert some effort, Many respondents are not so motivated and are inclined to complete the interview as rapidly and as easily as possible, reporting only those events which are most salient at the moment. Proposed solutions were to: a. Ask more probe questions to stimulate the respondent to work harder. b. Ask questions about hospitalizations which research showed to be most com- monly unreported—minor episodes and those which occurred several months prior to the interview, c. Ask about each individual separately in- stead of about the entire family, d. Send a brochure to the household prior to the interview to stress the impor- tance of the survey. e. Use the respondent as an interviewer to collect information from other mem- bers of the family through a self-ad- ministered form. There is a tendency for people to conceal or distort their memory of episodes which are em- barrassing or physically threatening. This may be because respondents have reservations about reporting certain types of problems or because the emotional nature of the episode has resulted in distortion or suppression of the memory of the experience. The assumption is that most episodes are not so threatening or stressful that they cannot be ob- tained by an interviewer, but that a greater moti- vational force is required to obtain the informa- tion. A follow-up interview which was part of Spe- cial Study No. 8 supports this conclusion. Proposed solutions to this problem were to: a. Use a self-administered form where it might be easier for the respondent to report episodes which would not bere- ported readily to an interviewer. b. Include introductory statements in the questions to stress the importance of the data. c. Use the brochure mentioned above. The viewpoint taken in this research is that problems of memory can be understood and dealt with more successfully if they are considered in terms of motivation. For instance, a hospitaliza- tion of one day's duration whichoccurrednearlya year ago is not actually an inaccessible memory, but greater effort and, therefore, a higher level of motivation is required for the respondenttore- port it. In the same way, an operation which is surrounded with intense emotion is not actually repressed, but it requires a higher level of moti- vation for the respondent to be willing to discuss it. This concept is not derived solely from theory but conforms closely to the data from Special Study No. 8, especially those from the follow-up interviews. The Pilot Investigations Because of time and budgetary limitations, it was not possible to set up an elaborate experi- mental design to test all the variables separately. Instead, several small pilot investigations were conducted, each built upon the preceding one, and each one testing one or more new concepts. For the most part, the evaluation of these investiga- tions was subjective, although tabulations were made of the major variables. The number of cases in each pilot study was small so that no statistical tests were attempted. Each pilot study consisted of between 25 and 50 interviews; a total of six investigations were carried out, Five interviewers were employed in the pilot studies, each of whom had considerable experi- ence in pretesting questionnaires and new field ideas. The sample for the pilot investigations was selected from persons discharged from two Detroit hospitals. It was selected by a random process and covered hospital discharges during the preceding 18 months. The interviewers were told that someone in each family assigned to them had been hospital- ized within the past two years. This was neces- sary, because it was important to use the inter- viewers' experience to evaluate the various pro- cedures. A questionnaire was prepared for each pilot investigation. Interviewers were asked to record verbatim the responses given to each question, and to note anything that might be relevant to the problem of reporting hospital stays. In addition, interviewers were asked to explore, on their own initiative, new questions which they thought might be useful in eliciting unreported hospital episodes. Such exploration was undertaken only after the specified sequence of questions was asked. After each pilot investigation, a meeting was held with all interviewers. Interviewers' ideas as to how to improve the questionnaire were dis- cussed, and each interviewer's experience with each question was reviewed in detail. Following the discussions the interviews were analyzed, searching for ways to improve the re- porting of hospitalizations. The following is a summary of findings of the pilot investigations. The frame of reference of the respondent in reporting hospitalizations.—In the first pilot study two frames of reference were observed. If left free to report hospitalizations for themselves and their families, some respondents first talked about the more serious episodes for all family members and then the minor episodes for all fam- ily members. Other respondents tended to report systematically for each member of the familyin turn, regardless of whether the episodes were major or minor. In later pilot studies the problem was to discover which frame of reference seemed to predominate and to make use of it in the ques- tionnaire design, the assumption being that the closer the questioning conforms to the respond- ent's way of attacking the problem, the better the reporting. In subsequent pilot investigations, both approaches were used independently, The conclusion reached was that for small families or families with a small number of epi- sodes, the first method was satisfactory. For large families, particularly where several mem- bers had been hospitalized, a systematic ques- tioning about each family member produced more complete reports. The second approach was used in the final questionnaire. Use of additional questions.—Interviewers tried various additional questions or probes to obtain more complete reporting of episodes. In the first test, interviewers were asked to use whatever follow-up questions seemed most appro- priate to obtain more complete reporting. Addi- tional episodes were obtained by the use of these questions and several were standardized for the successive pilot investigations. Three types of follow-up questions were tried. The first, general probes, of the type, '"Did you have any other hos- pital stays?" The second, questions about possible types of hospitalization; for operations, for obser - vations, to have a baby, etc. The third type focused on minor episodes and those occurring several months prior to the interview, Most families have only one or two episodes to report. Thus respondents tended to become irritated at being asked a series of questions, since they felt they had reported all of their epi- sodes in response to the original question. Rap- port tended to suffer, and respondents developed a fixed response—they answered 'mo' without really considering the question. A lengthy series of probes, therefore, defeated its own purpose, and it was concluded that only a few probes should be used. Since the major problem of un- derreporting was for minor episodes, and those removed in time from the interview, it was de- cided to focus the probes on these issues. It was found that telling the respondent the reason for asking the questions helped to counter - act negative reactions. The probes, therefore, were introduced with the statement '"We find that people tend to forget . ..etc." With these changes the respondents appeared to tolerate the additional probes, and these changes resulted in picking up episodes previously unreported. The reference period for reporting.—For reasons described in the review of hypotheses in the previous section, respondents were asked about episodes occurring at any time during the calendar year 1959 and that part of 1960 prior to the interview. (The pilot study interviewing was done in the fall of 1960, so respondents were re- porting for 22 or 23 months.) The analysis period was the 12 calendar months preceding the month of the interview. Accuracy of reporting admission and dis- charge dates.— Various methods of obtaining dates of admission and discharge were tried insucces- sive pilot studies. The objective was to find the most accurate method of obtaining the discharge date, which was basic to the analysis. The discharge date can be obtained either by asking for the month of discharge inthe interview or by calculating the month of discharge by use of the admission date and the length of hospitali- zation. In the first pilot studies respondents were asked the month and day of admission, the length of stay, and the month and day of discharge. A comparison of these reports with hospital rec- ords revealed that respondents were fairly ac- curate on the month of admission or discharge, but inaccurate as to the day of admission. The re- port of the month of admission was slightly more accurate than the month of discharge. Of the two methods, it was found that the reported discharge month was considerably more accurate than the computed discharge date using the date of ad- mission and the length of stay. It was found also that handing the respondent a calendar before asking about dates improved reporting accuracy. Procedures to motivate the respondent,— Special Study No. 8 plus many other related stud- ies provide evidence that special attempts needed to be taken to motivate the respondent to report accurately. Several techniques were attempted in the pilot studies. Introductions to the National Health Survey, which were designed to stress the importance of accurate data for health planning and to educate the respondent in some of the uses made of the in- formation, were used by interviewers. These statements were later incorporated into a bro- chure and mailed to each household prior to the interview. In addition to the general introduction, spe- cial phrases were used to preface the hospital questions, The objective of these questions was to provide the respondent with some added stim- ulation to report episodes. Special problems.—During the pilot studies, some of the questions were reworded. Two changes are sufficiently interesting to be reported here. The word '"hospitalization'' was confusing to some respondents. Some failed to understand the word, and for others the implication was of a ''serious or long stay in the hospital." Hence the final ques- tionnaire used the awkward but meaningful phrase hospital stay." The word ''patient' also gave trouble, again because respondents tended to asso- ciate the word with severe illness. The word was therefore dropped. As a result of these pilot studies, techniques gradually evolved which appeared to increase the probability of obtaining a higher proportionofre- ports of hospitalizations than did the standard Na- tional Health Survey household interview. These techniques were then used in this experimental study. The design of this study is described below. The Sample Design Since the major interest inthis study wasin a comparison of procedures for collecting hospitali- zation data, rather than in population estimates as such, it was decided to conduct the study in a single, compact area. The efficiencies which re- sulted saved considerable money. A sample of 20 general or short-stay hos- pitals was chosen from those listed for the Detroit urbanized area by the American Hospital Association and the American Osteopathic Hos- pital Association. The hospitals were selected with probability proportional to the number of discharges they had during 1960 (exclusive of dis- charges for deliveries and for deaths). Sixteen of the twenty hospitals agreed to participate in the study. Replacements were selected for three of the four. Two of these replacements agreed to co-operate, making a total of 18 sample hospitals. The second-stage-sample selection was of persons discharged from the hospitals between May 1, 1960 and March 31, 1961. The sampling fraction for each hospital was such that the prod- uct of the first-stage-sampling ratio (of selecting hospitals) and the second-stage ratio was con- stant. The sample persons were selected system- atically after a random start from a list of dis- charges routinely maintained by the hospitals. To maintain the desired constant sampling fraction for each sample person, a subsample of persons with multiple discharges was taken, pro- portional to the number of discharges they had during the sampling time interval. Restrictions were put on the sample design to exclude the following: Persons who lived outside the Detroiturban- ized area. Persons whose only episode during the year was for a normal delivery. This restriction was placed since it was found in Special Study No. 8 that 97 percent of the deliveries were reported, and it was desired to weight the sample toward the less readily reported epi- sodes. Hospital episodes with stay of less than over- night. This conforms with the specifications of the National Health Survey. Persons who died in the hospital. Persons who were found to have moved out- side the Detroit urbanized area. If the sample person no longer lived at the address given on the hospital record and could not be lo- cated, it was assumed that he had moved out of the area. After the person was chosen for the sample, abstracts of all his episodes terminating between May 1, 1960 and the date of interview were ob- tained. (The interviews were conducted during the five-week period beginning May 1, 1961.) Since the sample was of persons discharged during the period, May 1960-March 1961, abstracts showing discharge dates during April, May, and June were for persons readmitted to the hospital and dis- charged during this period. Special Study No. 8 showed that discharges which had occurred near the date of interview were reported more accu- rately than those which had occurred earlier. Thus, by design, the sample consisted of rela- tively few discharges near the date of interview. A Latin Square design was used consisting of four orthogonal, completely randomized Latin Squares which generated the interviewing assign- ments, These assignments consisted of approxi- mately 18 interviews per week per interviewer. * The design was worked out by Harold Nisselson of the Bureau of Census. The design used as two major sources of variance the week of the interview and the region of the city. These were randomized, with the effects of their interactions assumed to be bal- anced or negligible, The city was divided into five geographic re- gions, and as has beenmentioned the interviewing was conducted in five weeks. Twenty interviewers were divided randomly into two groups. One group used the control procedure (Procedure A)andone experimental procedure (Procedure C), while the other group used the two experimental procedures (Procedures B and C). (These procedures arede- scribed in the following section.) This division in assignments was necessary because of the par- ticular procedures to be tested. Thus, the Pro- cedure C interviews were taken by 20 interview- ers; Procedures A and B interviews were taken by different groups of 10 interviewers. Each in- terviewer was assigned twice as many A or B in- terviews as C interviews. The following table, one of the four Latin Squares, will illustrate the design. Region Region | Region Region Region I n III Iv Vv A,C inter- Week Week Week Week Week viewer #1 5 2 4 1 3 A,C inter- Week Week Week Week Week viewer #2 2 1 3 4 5 A,C inter- Week Week Week Week Week viewer #3 1 3 5 2 4 A,C inter- Week Week Week Week Week viewer #4 4 5 2 3 1 A,C inter- Week Week Week Week Week viewer #5 3 4 1 5 2 It may be seen that there were five possible patterns of interviewing assignments. Taking in- terviewer No. 1, for Week 5 all of her interviews fell into Region I of the city. Two thirds of these interviews were Procedure A and one third, Pro- cedure C. Since there were 20 interviewers, three other interviewers were working in the same re- gion during Week 5, one other A,C interviewer and two other B,C interviewers. The patterns were such that no interviewer worked in any region for more than one week; and no two interviewers worked together in the same region more than once. Region of the city was selected as a major source of variance for three reasons. First, since a given hospital tends to serve persons in its immediate area, control on region, to some extent, controlled the variance between hospitals. Second, there was some evidence in Special Study No. 8 that socioeconomic status is related to the rate with which hospitalizations are reported. Controlling the region of the city, to some extent, made it possible to isolate the variance attribut- able to this relationship. In addition, restricting the sample to five regions seemed to give optimum spread without substantially increasing travel costs per interview. Description of the Procedures As was described in the section on the re- search design, three procedures were used in this study; one control procedure and two experi- mental procedures. The questionnaires and forms used can be found in Appendix II. Procedure A—the control interview.—The survey procedure referred to as "Procedure A" in this report was essentially the standard pro- cedure used in 1961 by the Health Interview Sur- vey of the National Health Survey, except that some minor changes were made in anticipation of the 1962 NHS questionnaire, Prior to the interview, a letter was sent to each Procedure A household informing the family that a Bureau of Census interviewer would visit their home in a week or two. This letter and questionnaires used in the study are shown in Appendix II. In the interview the hospital questions were asked about each family member separately, rather than about the family group as a whole as has been the procedure used in the National Health Survey in the past, Procedure B—an experimental interview and follow-up self-administered questionnaire.—Pro- cedure B consisted of a direct interview and a mail follow-up questionnaire, The direct inter- view questionnaire was developed as a result of the pilot investigations described earlier. The questions are identical to those used in Proced- ure A except for marked differences in the hos- pitalization section, These differences are as follows: Hospital questions were expanded to include additional probe questions. The reference period was 1960 and that part of 1961 prior to the interview rather than the 12 months prior to the week of interview as used in Procedure A. Respondents were asked to report month and year of discharge rather than month and year of admission, Special explanatory statements were included in the section. This procedure was also different from Pro- cedure A in that a special brochure was enclosed with the letter which is ordinarily sent to the households prior to the interviews. The brochure is reproduced in Appendix II. Following the interview the questionnaires were edited in the Census Regional office. As soon as the editing was completed, a self-administered form was mailed to the family. This form con- tained the family composition as reported to the interviewer and a record of the hospitalizations as reported in the interview. Respondents were asked to answer a few questions designed to elicit additional hospitalizations and return it to the Bureau of the Census office. If the form was not received within one week after the date of the first mailing, a follow-up form was mailed, con- taining the same questions but a different letter from the Census Regional supervisor. If neither form was returned, an attempt was made to obtain the information by telephone. If telephoning was not possible, a personal visit was made and the data collected by interview, Procedure C—the experimental self-admin- istered questionnaire.—In this procedure the in- terview questionnaire was identical to that used in Procedure A except that no questions on hos- pitalizations were included. Instead of being ques- tioned about hospitalizations, a form to be filled out by the family was left with the respondent. Nonresponses were followed up using the same techniques as for Procedure B. The Interviewers Twenty interviewers were employed for this study. Most of them had had a limited amount of interviewing experience, largely on the Decennial Census. The decision to use new interviewers was based on several considerations. The existing Census staff in the Detroit area was fully occu- pied. In addition, it was felt thatnew interviewers would be less likely to perceive that the rate of hospitalizations in the sample was abnormally high. Of greatest importance, however, was the need for training interviewers in new techniques without having them recognize that the techniques were different from the usual National Health Survey interview procedures. It was felt to be very important to keep the interviewers from knowing that this was a study of hospitalizations, since they might probe with greater zeal. Specif- ically, it was feared, the knowledge that there was at least one hospitalization for each family would have motivated them to probe until a hospitaliza- tion was reported. Interviewers were trained by the Bureau of the Census using, in general, their usual training procedures. The interviewers were divided ran- domly into two groups; one for Procedures A and C, and the other for Procedures B and C. The training for the two groups was made as com- parable as possible. Since it was expected that interviewers would improve their skill with experience, the week of interviewing was used as one of the con- trols in the research design. Assignment of Interviewers Interviewers were given assignments to be completed within the week. They were given the family name and address from the hospital rec- ords. In cases where the family name was found to be different from that assigned, no interview was taken at that address. The usual quality con- trols used by the Bureau of the Census on Na- tional Health Survey data were used also on this study. Questionnaires were edited for missing in- formation and inconsistencies. Where necessary the missing information was obtained by telephone or a personal visit, Follow-up Techniques Procedures B and C included self-admin- istered questionnaires: the Procedure C inter- viewer leaving the questionnaire at the household at the completion of the interview, and the Pro- cedure B, self-administered questionnaire, being mailed to respondents. The Procedure B inter- viewers were presumably unaware that the follow- up was being conducted, at least until the third week when one interviewer was employed to fol- low up nonresponses. All self-administered forms were edited upon reaching the office. Maximum use of the telephone was made to obtain missing data. When respondents had no telephone, personal visits were made. Nonresponse was followed up by: first, a mail inquiry to those who had not responded with- in a week of initial contact, and second, personal visits or telephone calls to those not responding to the mail inquiry. Deviations From the Design The study, as it was carried out, deviated from the design in three ways. First, if a sample family was found to have moved to another region of the city, the interviewer to whom the assign- ment was originally made was instructed to follow that family and conduct the interview. Second, in some cases, if the family was not found at home or if the assignment could not be completed dur- ing the week in which it was assigned, the family was interviewed during the following week. Third, two interviewers were unable to complete the study assignments. One was dropped during the fourth week, and another did not interview during the fifth week. In each case, the incompleted in- terviews were reassigned to another interviewer who was working in the same region and who was using the same procedures. Editing, Matching, and Coding The editing and coding was carried out by a trained group of coders on the Survey Research Center staff. Three distinct tasks were involved in the editing: the matching of persons, the re- editing of episodes, and the matching of episodes. To determine whether or not the person whose hospitalizations were sampled was included in the household, age, race, sex, and name were used as criteria. In general, this was not a com- plex task, as it was usually clear whether or not the sample person was in the household. Because the interviewing took place over the period of a month, some of the episodes fell out- side of the reference year. The reference year differed for the procedures. For Procedure A the year was the 365 days preceding the Sunday night of the interviewing week. For Procedures Band C the year was the 12 months preceding the month in which the interviewing took place. To be in the sample the hospital discharge had to be within the reference year. Other episodes were excluded from the sample for other reasons. (For instance, a woman who was hospitalized twice, once for a delivery and once for an episode which proved to be outside of the reference year, was excluded from the sample, since her only episode during the reference year was for a delivery.) All hos- pital discharges were edited to ascertain that they truly were within the scope of the study. In matching episodes, it was occasionally difficult to determine whether or not the some- time-vague and inaccurate reports found in the interview actually referred to the episode for which there was a hospital discharge record. The length of stay, month, diagnosis, name of hospital and, in the case where surgery was performed, the type of operation, were all used as criteria for matching. When three of these characteristics were reported with reasonable accuracy and the other two were not too inconsistent, the episodes were considered to be matched. If there was amajor inconsistency, especially if the hospital seemed to have been reported in- correctly, the decisions were made by the super- visors. For every interview, the editing and matching was checked independently by one of the researchers or the coding supervisor. Disagree- ment was resolved by consensus. Although the process was of necessity somewhat arbitrary, 85 percent of the cases included only one episode for a person, and in these cases it was usually clear whether or not the episode had been re- ported. The coding was unusually accurate. Incheck- ing about 15 percent of the coding, it was found that the reliability was 0.99, when calculated in terms of the percent of variables which were coded correctly. This small percentage of error was further reduced by intensive consistency checks of the cards. COMPARISON OF UNDERREPORTING IN THE THREE PROCEDURES BY CHARACTERISTICS OF THE SAMPLE PERSON The primary purpose of this study was to compare two experimental procedures with a con- trol procedure, i.e., the one used by the National Health Survey, to determine whether either or both show a significant improvement in the level at which hospitalizations are reported and to in- vestigate ways in which underreporting rates for the procedures differ in relation to the charac- teristics of persons who are hospitalized. To gain added confidence that results ob- tained were not due to differences between sam- ples rather than differences between procedures, demographic characteristics of the three samples were compared. Those differences found were well within chance fluctuation, as would be ex- pected from any probability sampling design carefully carried out. The rates of underreporting of hospital epi- sodes in the three procedures are compared in table A. The difference between the net under- reporting rate of 6 percent for Procedure B and a rate of 14 percent for both Procedures A and C is statistically significant. (Standard errors of estimates may be found in Appendix I.) The re- porting rate’for Procedure B includes the epi- sodes reported in the mail follow-up. The results of the follow-up procedures are discussed in the following section. When the overreports are excluded, the un- derreporting rate is 17 percent for Procedure A, 9 percent for Procedure B, and 16 percent for Procedure C. Considering only the direct inter- view for Procedure B, the underreporting rate was 12 percent. Table 1'*shows that Procedure B produced a sizable reduction in underreporting compared with Procedures A and C for both males and fe- males. The underreporting rate was lower for males than for females (4 percent and 7 percent, respectively). Similarly, table 2 indicates Pro- cedure B was superior to Procedure A for all age groups. The largest difference is for the group 55 years or older where there was a net underreporting rate of zero in Procedure B. How- ever, differences for all age groups are signifi- cantly lower in Procedure Bthanin Procedure A. The underreporting for white and nonwhite sample persons is compared in table 3. For all proced- ures the rate of underreporting for nonwhite was about twice that for white persons. While Proced- ure B showed a substantial reduction in under- reporting for both groups, the same two to one ratio is found in all procedures. Table 4 shows the comparisons of under- reporting by family income. Procedure B showed a significant improvement in reporting episodes for both low and high income groups (those above and below $7,000), Within Procedures B and C persons with family incomes above $7,000 were significantly lower in underreporting than those in lower income groups. The pattern is observed also within Procedure A, Here, as in table 5, it can be observed that while Procedure B showed This rate takes into consideration the episodes reported in the interviews that could not be matched with hospital records; these unmatched reports are referred to as ‘‘overreports.’’ Experience in Special Study No. 8 suggests that a number of the episodes were classified as overreports in error due to failure to locate the rec- ords in the hospitals. *% . . . Tables designated by arabic numerals are shown in the sec- tion following the text. Table A. Percent of hospital episodes underreported in the survey, by survey procedure Hospital discharges Number of episodes . reported in the Percent Survey procedure Rubel i Percent | survey not corre- | underreported roral i Batehe under- spondingly matched (including with inter- overreports) view report reported (overreports) Pp Armrmme mmm —————— 521 90 17 17 14 Brommrmmmmnnmm mmm 558 48 9 16 6 Crmmmmmr mmm. 546 87 16 12 14 *This percentage is the ratio of total unreported episodes plus overreported episodes to total hospital discharges. improvement, the patterns of underreporting re- mained consistent between the groups. The relationship between education of the sample person and reporting rates can be seen in table 5. Combining the groups, as shown in table B, the underreporting rates in Procedures A and B for persons who had not graduated from high school were about the same as the rates for those with higher education. For Procedure C, however, hospital episodes were reported better for those with at least a high school education, than for those in the lower educational group. This relationship possibly reflects a greater ease of handling self-administered forms by persons with higher education. Table 6 shows the level of underreporting by the relationship of the sample person to the re- spondent, In Procedure A, respondents reported better for themselves than they did for others. This seems to be true for Procedure C respond- ents also, but the picture is not clear. In Procedure C the data are confused by the fact that a number of people did not sign the fol- low-up forms; and often the interviewers did not record the name of the person with whom they talked when they had to follow-up via telephone or personal visit, This group, probably the leastco- operative and the least willing to report, is most prone to underreport; their underreporting rate being about 50 percent higher than the nexthighest rate. For those cases in which the respondent could be identified, respondents reported best for themselves. In contrast, the relationship observedin Pro- cedure A is eliminated by Procedure B. Persons Table B. Percent of hospital episodes underreported in the survey for persons 17 years of procedure and age and education of the sample over, by survey person, including and excluding overre- ports Survey procedure and education of sample person-17+ years Percent under- Procedure A Less than high school graduate-- High school grad- uate Or MOLE===iww Procedure B Less than high school graduate-- High school grad- uate or morg= === Procedure C Less than high school graduate-- High school grad- uate Or more=~wwww reported Includ- Exclud- ing ing over - over- reports reports 14 19 13 16 5 10 6 8 16 15 10 12 reported just as well for others as they did for themselves. Indeed, this is one of the obvious ways in which Procedure B was an improvement over Procedures A and C, and offers one answer to the question of what was accomplished with Procedure B, which enabled the underreporting to be reduced so drastically. In conclusion, it is worth noting thatthe over- all reporting in Procedure B was significantly better than in Procedures A and C. COMPARISON OF UNDERREPORTING IN THE THREE PROCEDURES BY CHARACTERISTICS OF THE EPISODES Turning from characteristics of sample per- sons to a consideration of some characteristics of the episodes, table 7 shows that all three pro- cedures resulted in better reporting for episodes involving longer hospitalization. Procedure B was superior to the other pro- cedures in evoking reports of episodes of greater than one day. Procedure C showed a decrease in underreporting as the stays became longer, but the underreporting was consistently higher than for Procedure B. The pattern in Procedure A is not entirely clear, probably because of the small number of episodes in some categories. Proced- ure B did not result inimproved reporting of one- day stays, but there was an obvious improvement in the reporting of stays longer than a day. The one-day stays, however, were reported as poorly in Procedure B as they were in Procedures A or C. The "diagnostic rating'' in table 8 refers toa subjective scale of the degree of threat which is involved in a given diagnosis." Included in this are two concepts, physical threat, or the medical seriousness of the diagnosis, and psychological and social threat, especially the social accepta- bility of a problem. For example, havinga baby is quite socially acceptable, and therefore would be easily reported, even to a stranger such as the in- terviewer; but a psychotic breakdown or delirium tremens would detract from one's social image, and therefore would be less readily reported. "This rating was devised for Special Study No. 8 and a more detailed description of the ratings can be found in the report of that study. As can be seen, the effects of threat were marked in all three procedures. The underreport- ing rate for all degrees of threat was lowered with the use of Procedure B but the pattern was the same as for Procedures A and C, i.e., an in- crease in underreporting with an increase in the level of threat. It was hoped that a self-administered form would make it easier for the respondent to report an embarrassing episode, since writing about it would seem to be easier than reporting it to a stranger; but the pattern was not changed with the use of Procedure C. Table 9 shows a comparison of the three procedures for hospitalization with and without accompanying surgery. The differences between the underreporting rates for surgical and nonsur- gical treatment are statistically significant for all three procedures. Although the reporting for both types of episodes was improved in Proced- ure B, the pattern between type of treatment re- mained. It is undoubtedly true that episodes in- volving surgery have greater emotional impact on the person and his family than nonsurgical hospitalizations, and are therefore more readily recalled. Surgical episodes are also likely to in- volve longer hospitalizations and, longer stays are reported more completely as shownintable?7. Preceding tables have shown that underre- porting of hospital episodes varies with the im- pact of the episodes on the respondent. Another variable closely related to impact is the recency of the event. It has been found repeatedly that events closer to the present are recalled more accurately than those farther back. Table 10 shows a comparison of episodes by the elapsed time between the hospital discharge and the in- terview. All three procedures showed anincrease in underreporting as the time between the hos- pitalization and the interview became longer. The differences between the underreporting rates for the first 30 weeks and the remaining weeks are statistically significant. Procedure B was somewhat different from the others in that the rate of underreporting was relatively flat through 40 weeks, with a rise in the period over 40 weeks. It should be recalled here that the reference period presented to the respondent was different for Procedure A than for Procedures B and C. In Procedure A the period was one year preceding the interview week. For Procedures B and C it was the part of 1961 which preceded the interview plus all of 1960. The hope was that this change would help substantially to overcome the large underreporting rate of episodes which terminated near the end of the reference year. Both Proced- ures B and C showed animprovement in this year- end effect, but in neither procedure was the effect eliminated. Table 11 shows the underreporting of hos- pitalizations by the number of hospital recorded episodes experienced by the sample person during the reference year. In all procedures when the sample person had three or more episodes during the reference period, the underreporting rate was higher than for fewer episodes. Interestingly enough, there is very little difference in reporting rates for persons with one and two hospitaliza- tions. Again it is noted that the pattern in Pro- cedure B is similar to that found in Procedures A and C, but the rate is lower for each group. In conclusion, this section has presented convincing evidence for the importance of the characteristics of the episodes themselves in problems of reporting. All of these characteris- tics which would make a hospital stay less psycho- logically relevant—one-day stays, nonsurgical stays, and time-distant stays—are reported very poorly. The one contradictory bit of evidence is that high threat episodes are reported more poorly, even though they should have more impact on the respondent. Two solutions to this latter point are presented. First, it may be explained by stating that persons remember such episodes, but do not want to talk about them with an interviewer. Sec- ond, one can draw upon personality theory and postulate that the person does not even think about some threatening illnesses; thathe keeps them out of consciousness to the point thatitis difficult for him to recall them in an interview situation. The latter is consistent with findings of this study in relation to the other types of episodes thatare not reported. In all probability, the consistent patterns found with threat ratings was due toa combination of both of these factors. Procedure B shows a consistently lower rate of underreporting and significantly improved re- porting in certain subgroups. It was not successful however in eliminating some patterns of under- reporting, such as episodes involving one-day stays, and those episodes 40 weeks or more prior to the interview. FOLLOW-UP TO PROCEDURES B AND C Both Procedures B and C included self-ad- ministered forms for the reporting of hospitaliza- tions. Procedure C relied entirely onthe self-ad- 12 ministered form for information on hospitaliza- tion. In Procedure B, however, all households where interviews had been completed were mailed a questionnaire for the purpose of eliciting hos- pitalizations which were not reported in the in- terviews. A brief description of both procedures is given in the first section of this report. The questionnaires used are shown in Appendix II. Follow-up tothe self-administered procedure for nonresponse included one mail inquiry to all sample households not responding within a week of initial contact. Further follow-up to those not responding to the mail inquiry was made by tele- phone where possible and by personal visits when a telephone contact could not be made. The reason for using a follow-up question- naire in Procedure B was the finding from Special Study No. 8 that a personal follow-up interview was successful in obtaining episodes not originally reported. It was felt that a mail follow-up might achieve the same results and be financially feasi- ble in the National Health Survey. Table C shows that for Procedure B, 96 per- cent of the episodes finally obtained were reported during the interview. The follow-up procedures produced an additional 21 episodes. This resulted in a reduction of 3 percentage points inthe under - reporting rate, from 9 percent to 6 percent includ- ing overreports, or 12 percent to 9 percent, ex- cluding overreports (table D). Along most dimensions the 21 episodes which were reported in the follow-up for Procedure B were evenly distributed. There were, however, several groups for which the follow-up procedure was particularly effective in reducing the under- reporting. The most obvious of these is that1l of the 21 episodes were reported by parents for children under 17 years of age. This reduced the underreporting rate for children from 13 percent without the follow-up to 6 percent when the follow- up episodes were added (table E). Note also in table E that self-respondents reported no better in the direct interview part of Procedure B thanthey did in the other procedures. However, the Pro- cedure B interview was especially effective in eliciting hospital episodes from respondents answering for other adults. The second largest reduction in the under- reporting rate was for nonwhite sample persons. The rate for white persons was only slightly affected, but the nonwhite underreporting rate was reduced from 21 percent to 10 percent when the follow-up reports were added. Two income groups show marked improve- ment as a result of the follow-up report. The un- derreporting rate for persons with an annual in- Table C. Number and percent distribution of hospital episodes reported in Procedure B, by manner in which hospitalization report was obtained, including and excluding over- reports Manner in which hospitalization report Including overreports | Excluding overreports was obtained Number Percent Number Percent TOLALr mmm mmm mw mm om mm om mm wm om om om mm mm mm mn 526 100 510 100 Household interview------==-=ccecccca-- 505 96 490 96 First mail form----------=cc-cccecoo-- 10 2 10 2 Second mail form---------====eemmeena- 3 1 3 1 Telephone or personal follow-ups------ 8 1 7 1 Table D. Cumulative number and percent cedure B, by manner in which hospital cluding overreports of underreporting of hospital episodes in Pro- ization report was obtained, including and ex- Including overreports | Excluding overreports Manner in which hospitalization Cumulative | Cumulative Cumulative | Cumulative percent . ercent report was obtained interview endeg~ interview Pp reports * reports under - reported reported Household interview----------------- 505 9 490 12 First mall form-====e==secacnsmmnnsm 515 8 500 10 Second mail form----------===>=eue-- 518 7 503 10 Telephone or personal follow-ups=-=--- 526 6 510 9 *The cumulative percentage of 558 hospital episodes from hospital records which had not been reported after each respective step was completed. Table E. Comparison of underreporting of hospital episodes for Procedure B, with and without follow-up, with Procedures A and C, by type of respondent Procedure A Procedure B Procedure C Percent Type of respondent Percent underreported Percent underreported underreported With Without follow-up | follow-up SEL rm wm mmm mom om om mm mm ee en a mm a wm om 10 6 9 9 Proxy for adult----=--===mmmuum=- 21 6 8 16 Proxy for child----==-=====wmu=- 15 6 13 12 come of less than $2,000 was reduced from 25 percent to 11 percent by the addition of the fol- low-up reports; the underreporting rate for per- sons in the $7,000-10,000 category dropped from 8 percent to 3 percent. A consideration of the follow-up reports in terms of diagnostic rating reveals no differences between high and low threat episodes. For all three categories, about one third of the episodes not reported in the interview were reported inthe follow-up. It was thought that the follow-up might help pick up the very short stays which tend to be for- gotten. In fact, the opposite was true. Of the 32 five or more stays not reported in the interview, 11 were reported in the follow-up; but only 2 of 14 unreported one-day stays were obtained in the follow-up. It can be concluded from this that re- spondents generally did not consult records to fill out the follow-up questionnaire, that the kinds of episodes which were reported in the follow-up were important episodes which were not likely to be forgotten. Actually since the numbers are small, no definite conclusions are made. But at least it seems safe to state that the short, easily forgotten stays, which the respondent isnot likely to remember on the spur of the moment, were not well reported in the follow-up in Procedure B. In regard to the interval between the hospital discharge and the interview, an interesting phe- nomenon occurred. No hospital episodes within 10 weeks of the interview were reported in the fol- low-up. And, although there were 46 underreports after the interview among episodes which occurred 31 weeks or more before the interview, only 10 were reported in the follow-up. The greatest im- provement in reporting, therefore, pertained to episodes which occurred 10 to 30 weeks prior to the interview. For these, the underreporting rate was reduced from 8 percent to 3 percent when the follow-up reports were added. These data indicate that the follow-up ques- tionnaire of Procedure B is capable of reducing substantially the number of hospital episodes not reported in household interviews. In general, the follow-up was most effective among groups in which the underreporting rate was still high after the interview, The exception was among hard- core-like episodes with one-day duration, '"threat- ening" diagnoses, and episodes which occurred more than 30 weeks prior to the interview. This suggests that the follow-up would have produced more striking effects than it did had it been used in connection with less successful Procedure A. For the self-administered form in Procedure C, table F shows the percent distribution of re- turns. Three fourths of the questionnaires left with the respondent by the interviewer were re- turned without follow-up. As shown in table G, had no follow-up been made, over one third of the episodes would not have been reported. Tables H and I, show underreporting rates by the manner in which the hospitalization report was obtained. Underreporting rates by the person who filled out the follow-up forms for both Procedures B and C are shown in table 12, It is felt that these tables relate more to the characteristics of re- spondents than they do to the follow-up proced- ures. Tables H and I indicate that the persons who mailed in the first or second forms were much more inclined to reporthospitalizations than those who had to be contacted a third time, either by telephone or by a personal visit. The implications of these tables seem to be apparent. Persons who Table F. Number and percent distribution of hospital episodes reported in Procedure C, by manner in which hospitalization report was obtained, including and excluding over- reports Manner in which hospital- Including overreports Excluding overreports Yzation'seport Wes dbtatesd Number Percent Number Percent Total-====m=memcceecan- 471 100 459 100 First mail form--=-====-==eee-- 349 74 343 75 Second mail form--=--=---==---- 65 14 64 14 Telephone or personal follow-up--=====mcccccccccca- 57 12 52 11 Table G. cluding overreports Cumulative number and percent of underreporting cedure C, by manner in which hospitalization report was of hospital episodes in Pro- obtained, including and ex- Including overreports | Excluding overreports Manner in which hospitalization Cumulative Cumulative report was obtained Cumulative percent | Cumulative percent interview under - interview under- reports reported reports reported First mail form----------ccccecccaca- 349 36 343 37 Second mail form-------==-ceccceaaa- 414 24 407 25 Telephone or personal follow-up-=---- 471 14 459 16 Table H. Number and percent of underreporting for procedure B, by the manner in which hospitalization report was obtained, including and excluding overreports Including overreports Excluding overreports Manner in which hos- pigalication was Interview | Hospital percent Interview | Hospital Peyeant reports records reported reports | records reported Total-=mmmmm mmm 526 558 6 510 558 9 First mail form------- 371 388 4 361 388 7 Second mail form------ 68 72 6 66 72 8 Telephone or per- sonal follow-up--=---- 85 95 11 81 95 15 Unknown--=======m=m==- 2 3 (*) 2 3 (*) Table I. Number and percent of underreporting of hospital episodes in Procedure C, by the manner in which hospitalization report was obtained, including and excluding overreports Including overreports Excluding overreports Manner in which hos- pitalization report | rpterview | Hospital Depsany Interview | Hospital Fevcent was obtained reports | records uncer reports | records Unger. reported reported Total-=========- 471 546 14 459 546 16 First mail form==----- 349 394 11 343 394 13 Second mail form------ 65 75 13 64 75 15 Telephone or person- al follow-up=====-=-- 57 77 26 52 77 32 were prone to co-operate with the study would do so both by reporting hospitalizations thoroughly and by returning the mail form promptly. Those who had to be contacted repeatedly seemed to be less interested and unwilling to be helpful. In a similar vein, table 12 shows thatit makes considerable difference whether or not the sam- ple person or the person who was originally in- terviewed completed the self-administered form. One obvious hypothesis is that a respondent who was interested in a study would sit down and fill out the form herself, while a less interested re- spondent might give it to someone else to com- plete. Another relevant point would seem to be that the original interview respondent would be more familiar with the reasons for which the study was being conducted through contact with the in- terviewer than, for instance, her husband, and therefore might do a more thorough job of filling out the form. An added by-product of the follow-up to Pro- cedure B was the use of thedatato correct infor- mation obtained in the interview. Thirty-six of 490 interview reports (7 percent) were corrected in some significant way by the use of information obtained in the follow-up. Most of these correc- tions related to reported length of stay, month of discharge, or diagnosis. COMPARISON OF UNDERREPORTING IN THE THREE PROCEDURES BY CHARACTERISTICS OF THE RESPONDENTS This study was not designed to permit avery satisfactory analysis of the reasons why one pro- cedure performed better than another inobtaining hospitalizations. Except for the mail follow-up to Procedure B, this was an "all or none' design; that is, if one procedure was significantly better than the other, this procedure would need to be adopted in its entirety since the factors leading to improvement could not be isolated. However, cer- tain tendencies in the data do support hypotheses as to the reasons for the outcome of the various procedures. In this section the focus is on the characteristics of respondents to see whether sig- nificantly different patterns of reporting are ob- tained by the three procedures. It should be re- membered that about 40 percent of the respond- ents were reporting for themselves and the re- mainder for some other family member. 16 Table 13 shows the reporting rates by survey procedure and sex of the respondent. For all pro- cedures the underreporting rates were lower for female respondents than for males. However, ex- cept for Procedure C, the differences are not statistically significant. For both men and women respondents, Procedure B shows a considerably lower rate of underreporting than Procedures A and C. Another point of interest is thatahigher pro- portion of respondents in Procedure C weremale; 28 percent for Procedure C compared with 19 percent for Procedure A and 17 percent for Pro- cedure B. For 73 episodes the sex of the respond- ent was unknown. However, there is no indication that these were predominantly female, It can be hypothesized that filling in a questionnaire is part of the role of the male family head. If this hypoth- esis is true, a sizable number of persons filling in the mail form on Procedure C were different from the respondents in the interview part of Procedure C. Thus, if the interviewer did anything to instruct the interview respondent or motivate him to fill out the mail form, the effort was either wasted or at best transmitted indirectly to the person who actually filled out the form. This could account for the higher underreporting rate for male respondents in Procedure C. Table 14 shows that in Procedure A, older respondents tend to be poorer reporters of hos- pital episodes than younger respondents. This is consistent with previous findings in Special Study No. 8. The relationship with age disappears in Procedures B and C. The reason for this differ- ence is not apparent. The relationship between the education of the respondent and the underreporting of hospital epi- sodes can be seen in table 15. There is a clear pattern in Procedure A—the higher the educational level, the better the reporting. This pattern does not show up in Procedures B and C. Infact, one of the impressive differences between Procedures A and B was the significantly better reporting of hospitalizations among the lower educational groups in Procedure B, in which there were essen- tially no differences in underreporting attributable to the educational level. The findings for Procedure C are not clear because of the large group for which the respond- ent could not be ascertained. Disregarding this, the education of the respondent has only a slight effect on reporting of hospitalizations. In the interviews, respondents were asked to report chronic and acute conditions experienced by members of the family. Table 16 shows the relationship between the number of these condi- tions reported for the sample person and therate of underreporting of hospital episodes. It seems clear for all three procedures, that the underre- porting rate decreased as the number of condi- tions reported for the sample person increased. This relationship is also evident, but to a lesser degree, when underreported episodes are dis- tributed by the number of conditions reported for the respondent, There are several factors which might ex- plain this relationship. (1) A respondent who tends to be particularly conscious of health con- ditions of himself and his family may be more likely to recall illnesses as well as hospitaliza- tions; (2) the sample person may be considered to be 'ailing' and the reporting is a reflection of this perception; (3) that persons for whom sev- eral conditions were reported tended to have had recent (and, hence, better reported) hospitaliza- tions; and (4) that reporting both hospitalizations and illnesses is an index of how hard the re- spondent tries to give information. If this is so, then reporting can be considered as an indication of the level of motivation of the respondent to report. Except for the mail follow-up of Procedure B where a reduction in the underreporting rate of 3 percentage points was obtained, itisnot clear how much difference each change in procedure made in reporting of episodes. From the patterns re- ported above, it may be that asking about each family member individually and asking additional probes were useful to stimulate memory and im- proved reporting, particularly among proxy re- spondents. Also it may be that these factors assisted older persons inrecalling episodes more readily. Reduction in underreporting for episodes of short duration and for those some time prior to the interview may be attributable to the added probes, one of which specifically asked for short stays and distant episodes. ACCURACY OF REPORTING LENGTH OF STAY AND DATE OF DISCHARGE The preceding analysis has included only one type of reporting accuracy, the completeness with which persons report hospitalizations. There is another aspect of reporting which is also im- portant in evaluating field procedures, namely, the accuracy with which details of hospitaliza- tions are reported. One aspect of this question would be the accuracy of reported diagnoses but unfortunately there are very few cases in any given diagnostic category, thus the data are not very meaningful. Another consideration istheac- curacy with which the month of discharge and length of stay were reported. It was expected that a self-administered form, such as was used in Procedure C, would provide an opportunity for persons to refer to records, consult other members of the family, and generally give more time and thought to their responses. While Procedure C did not substan- tially increase the percentage of hospitalizations reported, tables 17 and 18 show that the informa- tion that was obtained about hospital episodes tended to be more accurate than the information in either Procedures A or B. The tables are generally self-explanatory. Slightly better reports on the month of discharge was obtained with Procedure C, and the improve- ment over Procedure A is even more marked in the reporting of the length of stay. Slightly more accurate reports with Procedure B wereobtained on the length of stay, than Procedure A, but was essentially no more accurate than Procedure A on the month of discharge. An interesting feature of table 17 is that misreporting of the month of discharge in Pro- cedure A tended to err in the direction of under- stating the interval of time that had lapsed since the hospitalization, while in Procedure B the re- verse seemed to be true. The numbers involved, however, are quite small. There is a consistent tendency in all three procedures for the length of stay to be exaggerated. With respect to accuracy of information col- lected, Procedure C seemed to be superior to both of the other procedures, supporting the hy- pothesis that respondents who take the time tofill out a self-administered form can do a better job than those who respond to an interview, The data suggest that the primary obstacle in Procedure C is to motivate respondents to take the time to complete the form. Table 1. 10. 11. 12. 13. 14. 15. 16. 17. 18. DETAILED TABLES Percent of hospital episodes underreported, by survey procedure and sex of sample person, including and excluding overreports=-------=-ceemocmme ccc cccece eee Percent of hospital episodes underreported, by survey procedure and age of sample person, including and excluding overreportg--=---cemccecccmmccccccccccccccc————— Percent of hospital episodes underreported, by survey procedure and race of sam- ple person, including and excluding overreportg-----------eececcoccecccccccccaann Percent of hospital episodes underreported, by survey procedure and family in- come, including and excluding overreports=---------ceecccmmmcm coccinea Percent of hospital episodes underreported, by survey procedure and education of sample person, including and excluding overreportS----------c-ccccccccccccacaaaaa- Percent of hospital episodes underreported, by survey procedure and relationship of sample person to respondent, including and excluding overreports-------------- Percent of hospital episodes underreported, by survey procedure and number of days in hospital, from hospital records, excluding overreports--------=-=ceecca-- Percent of hospital episodes underreported, by survey procedure and diagnostic rating of diagnosis, from hospital records, excluding overreport§--=-----------c--- Percent of hospital episodes underreported, by survey procedure and type of treat- ment, from hospital records, excluding overreports---------ceccccmccmcccccaooaoaao Percent of hospital episodes underreported, by survey procedure and number of weeks between hospital discharge and interview, excluding overreports------------ Percent of hospital episodes underreported, by survey procedure and number of hospital recorded episodes during the reference year for the sample person, in cluding and excluding oOverreportsS===-= === cc mce commence ccm meee Percent of hospital episodes underreported, by survey procedure and relationship of respondent for the self-administered questionnaire to the respondent for the household interview, including and excluding overreports---------cecccccccccoaoaoo Percent of hospital episodes underreported, by survey procedure and sex of re- spondent, including and excluding overreports=-------cccocccmm mmm ccccccceea Percent of hospital episodes underreported, by survey procedure and age of re- spondent, including and excluding overreportsS=-=----c-cecccm como mmccm ccc cccean Percent of hospital episodes underreported, by survey procedure and education of respondent, including and excluding overreports----------cececcccmccccccccccaccaan Percent of hospital episodes underreported, by survey procedure and number of chronic or acute conditions reported for the sample person, including and ex- cluding OVerreportS======mee mmc me cece cece eee eee eee meemccmmmcae man Number and percent distribution of reported hospital episodes, by accuracy of re- porting month of discharge and by survey procedure, excluding overreports-------- Number and percent distribution of reported hospital episodes, by accuracy of re- porting length of stay and by survey procedure, excluding overreports------------ Page 19 19 20 20 21 22 23 24 24 25 26 27 28 28 29 30 31 31 Table 1. Percent of hospital episodes underreported, by survey procedure and sex of sample person, including and excluding overreports Survey procedure and sex Including overreports Excluding overreports of sample person Interview | Hospital Percent Interview | Hospital Percent reports records underreported reports records | underreported Procedure A Total-============ 448 521 14 431 521 17 Male-===-=cmecccccccec== 194 229 15 184 229 20 Female--======eeee=-aan= 254 292 13 247 292 15 Procedure B Total--=========== 526 558 6 510 558 9 Male-=======--==se====== 231 240 4 223 240 7 Female==--===ceccccccec=a= 295 318 7 287 318 10 Procedure C Total--=========== 471 546 14 459 546 16 Male-======-=ccecmeamn=== 220 255 14 217 255 15 Female-===-=ceccemecmna=== 251. 291 14 242 291 17 Table 2. Percent of hospital episodes underreported, by survey procedure and age of sample person, including and excluding overreports Survey procedure and age Including overreports Excluding overreports of sample person Interview | Hospital Percent Interview | Hospital Percent reports records underreported | reports records | underreported Procedure A Total-===========- 448 521 14 431 521 17 0-17 years=============- 138 162 15 136 162 16 18-34 years======-=------ 100 111 10 98 111 12 35-54 years==-===-====== 111 132 16 104 132 21 55 years or over-------- 99 116 15 93 116 20 Procedure B Total-=-s==ee===e-= 526 558 6 510 558 9 0-17-=-===mmmmcmccceaa=- 151 161 6 149 161 7 18-34 years----=-======-=- 114 121 6 112 12% 7 35-54 years----========= 149 163 9 144 163 12 55 years or over-------- 112 112 0 105 112 6 Unknown--=======-======== 0 1 (*) 0 i (*) Procedure C Total========c===- 471 546 14 459 546 16 0-17 years========-==--= 136 156 13 133 156 15 18-34 years----====----- 88 103 15 88 103 15 35-54 years-===-=-====== 146 172 15 141 172 18 55 years or over-------- 101 115 12 97 115 16 Table 3. Percent of hospital episodes underreported, by survey procedure son, including and excluding overreports and race of sample per- Survey procedure and Including overreports Excluding overreports race of sample person Interview | Hospital Percent Interview | Hospital Percent reports records underreported reports records underreported Procedure A TOLAl=~=======mm—— 448 521 14 431 521 17 WH ow ees ie mnprn 398 454 12 382 454 16 Nonwhite---=--ccccmaaaao 50 67 25 49 67 27 Procedure B Total-=====ceec-u- 526 558 6 510 558 9 White--=--coococacaaaao 457 481 5 444 481 8 Nonwhite-----=ceccccaaao 69 77 10 66 77 14 Procedure C Total-=======c-u-n 471 546 14 459 546 16 White wm snwmmmmm mens wee 409 464 12 399 464 14 Nonwhite-==e=eccccccaaan 62 82 24 60 82 27 Table 4. Percent of hospital episodes underreported, by survey procedure and family income, in- cluding and excluding overreports Survey procedure and Including overreports Excluding overreports family income Interview | Hospital Percent Interview | Hospital Percent reports records underreported reports records underreported Procedure A Total===-=ceececan- 448 521 14 431 521 17 Under $4,000=~=sswmwnene 92 103 11 87 103 16 $4,000-6,999-====-cuuum- 166 199 17 158 199 21 $7,000-9,999--===cnmuna- 108 120 10 105 120 12 $10,000 or over--------- 73 85 14 72 85 15 Unknown=--=-ccemeaaaaaax 9 14 36 9 14 36 Procedure B Total-====cvecooa- 526 558 6 510 558 9 Under $4,000--=====e-u-- 109 119 8 104 119 13 $4,000-6,999-=-=cemuenn- 215 226 5 207 226 8 $7,000-9,999--=ccemauca- 106 109 3 105 109 4 $10,000 or over--------- 84 89 6 82 89 8 Unknown======-cceccmanan. 12 15 20 12 15 20 Procedure C Total-=====eeeeo--- 471 546 14 459 546 16 Under $4,000---=-=-eec-x 115 131 12 110 a3 16 $4,000-6,999--=--cmceua- © 140 173 19 138 173 20 $7,000-9,999--=-ccmcunan 111 122 9 108 122 11 $10,000 or over----==-=--- 85 96 11 84 96 12 Unknown=-=-=ecccceeaaaan 20 24 17 19 24 21 20 Table 5. Percent of hospital episodes underreported, by survey procedure and education of sample person, including and excluding overreports Survey procedure and education of Including overreports Excluding overreports Interview | Hospital Percent Interview | Hospital Percent sample person reports | records | underreported | reports | records |underreported Procedure A Total-=--======c=== 448 521 14 431 521 17 0-8 years elementary school~======mmemecnnan 88 103 15 83 103 19 1-3 years high school--- 84 96 12 79 96 18 4 years high school----- 98 108 9 94 108 13 1 year of college or MOYE@======m======cea=== 42 53 21 41 53 23 Inappropriate (child under 17)---=-==cc===== 135 158 15 133 158 16 Unknown-=--==========-== 1 3 (*) 1 3 (*) Procedure B Total-==-==cemec=n 526 558 6 510 558 9 0-8 years elementary school--===-ccncmcncnnu- 118 128 8 113 128 12 1-3 years high school--- 96 98 2 91 98 7 4 years high school-sw-- 1.2 120 7 110 120 8 1 year of college or MOre=========c==eeeee== 49 51 4 47 51 8 Inappropriate (child under 17)-===--=-==-=e-- 145 155 6 143 155 8 Unknown===========mem=== 6 6 (*) 6 6 (*) Procedure C Total-========w="- 471 546 14 459 546 16 0-8 years elementary school--=====ceemeacen- 99 120 17 95 120 21 1-3 years high school--- 89 105 15 86 105 18 4 years high school----- 94 108 13 93 108 14 1 year of college or MOre=======m===ememee=— 52 55 5 51 55 7 Inappropriate (child under 17)--========-==-- 133 153 13 130 153 15 Unknown----=====c==ee==- 4 5 (*) 4 5 (*) 21 Table 6. Percent of hospital episodes underreported, by survey procedure and relationship of sam- ple person to respondent, including and excluding overreports Including overreports Excluding overreports Survey procedure and relationship of sample person to respondent Interview | Hospital Percent Interview | Hospital Percent reports records underreported reports records underreported Procedure A Total-=====vecceau- 448 521 14 431 521 17 Self-respondent--------- 209 231 10 202 231 13 Sample person is child of respondent---------- 134 158 15 132 158 16 Sample person is adult but not self-respondent--===--- 100 126 21 92 126 27 Unknown==-====ccceeaaaan 5 6 (*) 5 6 (*) Procedure B Total-==-=-=canaa-- 526 558 6 510 558 5 Self-respondent---=------ 241 257 6 231 257 10 Sample person is child of respondent-------=--- 146 155 6 144 155 7 Sample person is adult but not self-respondent--====-- 136 144 6 133 144 8 Unknown=-====-=ceeeeaconx 3 2 (*) 2 2 (*) Procedure C TOA Lm mmmnmmmnnse 471 546 14 459 546 16 Self-respondent--------- 162 179 9 160 179 11 Sample person is child of respondent---------- 119 135 12 116 135 14 Sample person is adult but not self-respondent--===--- 129 153 16 125 153 18 Unknown-===<-cecccaaaaa- 61 79 23 58 79 27 22 Table 7. Percent of hospital episodes underreported, by survey procedure and number of days in hospital, from hospital records, excluding overreports Survey procedure and number of days in hospital Excluding overreports Interview Hospital Percent reports records underreported Procedure A Total-==--eecemccc ccc ccc ccc ce cme ———— 431 521 17 1 day--~-===-c-cememecccccmcece ccc c ccm ecm meme 39 49 20 2-4 dayS--=-==cecceecccccecmme emcee eceec meee maa 122 154 21 5-7 dayS--==-~=-eemeccemmceecmeeeececceccccece—ea——— 105 125 16 8-14 days-======m=-mcmceeccccccecmccecc cece cea ——— 111 127 13 15 days Or more---=-===s-e-eme-cecececcccceecececece———- 54 66 18 Procedure B Total-===-==e-cecccce ccc cece cece —————— 510 558 9 1 day--==-=--ccecccccccmmcccceemmcecec meee eme nae 44 56 21 2-4 days--==—===c-=mceccece ccm cececccccceecec————— 169 184 8 5-7 dayS-====s=cmccceeccecec cece ccceeccecce—ce———— a 98 109 10 8-14 dayS=======meme-ceecmcecccccecccecee cece ——aa— 130 136 4 15 days Or more--=----es=-eee-eccccccccccccececece——-- 69 73 5 Procedure C Total=-===-=cecccmccem cece ccc meen — ecm —————— 459 546 16 1 day-=======scecceccccceeeeececceeccceee— meme 34 42 19 2-4 dayS--w==eeemccccccccccccccccccccece emcee 145 178 19 5-7 dayS=====ee-cecmcccecccmcecccemcecccece— mea 102 119 14 8-14 dayS-==-==-eme-mmcmccec ce ccccccc mcm —— na 107 126 15 15 days or more~-==-=--r=mmemmmm_ eee ———————————— 71 81 12 23 Table 8. Percent of hospital episodes underreported, by survey procedure and diagnostic rating from hospital records, excluding overreports Excluding overreports Survey procedure and diagnostic rating Interview Hospital Percent reports records underreported Procedure A Totale=mmmmm mmm mm ee eee eee 431 521 17 Most threatening=======cccm oom m eee 66 84 21 Somewhat threatening=-==-==-=cocccmammmm ceo 92 111 17 Not threatening======-c-me momma eee 272 325 16 UNKNOWN == == == = = me ee ee eee ee eee 1 1 (*) Procedure B Total===mm mmm me eee ee eee ec eeem 510 558 9 Most threatening===-===-ccccmmm mmc 97 110 12 Somewhat threatening----=--=ecemcmmcmcmccc acca 117 127 8 Not threatening-==-=--==cecmm mma 292 315 7 UNKNOWN === == =m mee ee ee ee emma 4 6 (*) Procedure C Total=mm mmm om ee ee eee em 459 546 16 Most threatening=======ecc cma mace 70 89 21 Somewhat threatening=-=---===ececmm momen 85 102 17 Not threatening==-====m-= oo come cmc 302 353 14 Unknown-=--==ceecmeeax mm een 2 2 (*) Table 9. Percent of hospital episodes underreported, by survey procedure and type of treatment, from hospital records, excluding overreports Excluding overreports Survey procedure and type of treatment Tuterview Hospital Percent reports records underreported Procedure A Totalemmmm meen m mee ee eee eee 431 521 17 Surgical ====c omnes 262 297 12 Nonsurgical==== === momma 168 223 25 UNKNOWN == == = = = em ee ee ee ee eee 1 1 (*) Procedure B Total mmm mmm mmm mmm me eee ee eee 510 558 9 Surgical-===--=m cme eam 313 334 6 Nonsurgical-====m mmo comm cece ee eee 193 218 11 UNKNOWN == == == = re ee eee ee eee eee 4 6 (*) Procedure C Totalmm mmm mmm mm me ee eee 459 546 16 Surgical======= comm eee 286 326 12 Nonsurgical=====-= =m mamma eee 171 218 22 UNKNOWN === = == = = oe ee mee eee ee eee 2 2 (*) 24 Table 10. Percent of hospital episodes underreported, by survey procedure and number of weeks be- tween hospital discharge and interview, excluding overreports Excluding overreports Survey procedure and number of weeks between hospital discharge and interview Interview Hospital Percent reports records underreported Procedure A OLA mmm mmm mt mm mm rm mm mm mm 431 521 17 1-10 weekS====m==coccccmcccn mene m cme mmm 49 52 6 11-20 weekS-=====memecmmecem cece mm emcee nee 112 123 9 21-30 week§-=====--cecemmcccccme mmm mcm — mma 89 100 11 ET — 97 122 20 41-53 weekS-==--=ccccccmcnc cc cmmce mcm meme mmm na 84 124 32 Total-=-===-c-eccmcem ccc ccc c cmc c ccc m mmm mem 510 558 9 1-10 weekS======-cecccmcc ccc mc ccm c emma 47 51 8 11-20 wWeekS======mmccccc cece 126 130 3 21-30 weekS====---mccccccccmmccccmcc cmc m eee 114 118 3 31-40 weekS==-===cccmcmmccc cmc e cmc mmm mma 115 126 9 41-53 weekS-=-=-cccceccm cece cece neem mma 108 133 19 Procedure C Total-==-=-=c-mmcmecc ccc ccc ccc ———— 459 546 16 1-10 weekS===-c-eccmemcec mecca 48 56 14 11-20 weekS======cccccccecec meee meee 116 129 10 21-30 weekS=====mcmccccc mmc c mmm 91 104 12 31-40 weekS-=====cecccmmmcecmmce cece 99 122 19 41-53 weekS--=-mmrrmmmrmmeee sree eee ———————————— 105 135 22 25 Table 11. Percent of hospital episodes underreported, by survey procedure and number of hospital recorded episodes during the reference year for the sample person, including and excluding over- reports Survey procedure and number of hospital Including overreports Excluding overreports recorded episodes Interview | Hospital Percent Interview | Hospital Percent reports records underreported | reports records underreported Procedure A Total-=---==-=--=--- 448 521 14 431 521 17 1 episode--===-ccmceacooo 354 410 14 339 410 17 2 episodes--=-=-ccccea-- 78 90 13 76 90 16 3 episodes or more------ 16 21 24 16 21 24 Procedure B Total-======cee--- 526 558 6 510 558 9 1 episode-====ccccacaaa- 364 381 4 353 381 7 2 episodes--=---ccccca-a 105 114 8 103 114 10 3 episodes or more------ 57 63 10 54 63 14 Procedure C Total-===c==ceuea- 471 546 14 459 546 16 1 episode-===-=ceccecaa- 350 401 13 342 401 15 2 episodes--===--cccca-- 92 102 10 90 102 12 3 episodes or more------ 29 43 33 27 43 37 26 Table 12. spondent for the including and excluding overreports Percent of hospital episodes underreported, by survey procedure and relationship of re- self-administered questionnaire to the respondent for the household interview, Survey procedure and re- lationship of respondent Including overreports Excluding overreports for self-administered questionnaire to Interview | Hospital Percent Interview | Hospital Percent household interview reports records underreported | reports records underreported respondent Procedure B Total-==========-- 526 558 6 510 558 9 Self-respondents, same respondent as in interview---====-==---- 113 116 3 109 116 6 Proxy-respondent, same respondent as in interview---=-====c==-- 118 122 3 115 122 6 Sample person, not interview respondent--- 36 36 0 35 36 3 Neither sample person nor interview respondent-=-====-=-=--=--- 66 75 12 65 75 13 More than one person---- 23 24 4 22 24 8 Unknown-==============-= 170 185 8 164 185 Ti Procedure C Total---========== 471 546 14 459 546 16 Self-respondent; same respondent as in interview--=--=c==-ce-- 136 152 11 134 152 12 Proxy-respondent, same respondent as in interview--=-=ccccca-u- 159 180 12 156 180 13 Sample person, not interview respondent--- 25- 26 4 25 26 4 Neither sample person nor interview respondent=========-=== 64 79 19 62 79 22 More than one person---- 23 27 15 22 27 19 Unknown=-========e=cae=-x 64 82 22 60 82 27 27 Table 13. Percent of hospital episodes underreported, including and excluding overreports by survey procedure and sex of respondent, Survey procedure and Including overreports Excluding overreports sex of respondent Interview | Hospital Percent Interview | Hospital Percent reports records underreported reports records underreported Procedure A Total »mewceecwanns 448 521 14 431 521 17 Male===cccccmmeccccnnaan 80 97 18 77 97 21 Females=enmee=ensunnnne- 360 415 13 346 415 17 Unknown===-==ccaccaaaaan 8 9 (*) 8 9 (*) Procedure B TOLALmmmmmmumeeeen 526 558 6 510 558 9 Male--=ceccccmcccc cca 86 94 9 82 94 13 Female=====v=-emmmmnnnn- 437 462 5 426 462 8 Unknown=-====eeceaeaaaoo 3 2 (*) 2 2 C0) Procedure C Total-====eeeeaaan 471 546 14 459 546 16 BQ im rm rm mmr 124 152 18 120 152 21 Female---===vcccccaaaano 289 321 10 285 321 11 Unknown======cceecccaaoo 58 73 2) 54 73 26 Table 14. Percent of hospital episodes underreported, by survey procedure and age of respondent, including and excluding overreports Survey procedure and Including overreports Excluding overreports age of respondent Interview | Hospital Percent Interview | Hospital Percent reports records underreported reports records underreported Procedure A Totale=e=emcemena-n 448 52) 14 431 521 17 0-17 years======--c=ca-- 2 2 {%) 2 2 (*) 18-34 years------ce----- 166 187 11 162 187 13 35-54 years=------co-e-- 180 213 15 173 213 19 55 years or over=-=----- 94 112 16 88 112 21 Unknowne======eeececeauan- 6 7 (*) 6 7 (*) Procedure B Total-=-=v-ceecan- 526 558 6 510 558 9 0-17 years--==--===ec-u- 3 3 (%) 3 3 (*) 18-34 years------c-ca--- 181 197 8 178 197 10 35-54 years--------c---- 238 254 6 233 254 8 55 years or over=--=---=- 101 101 0 94 101 7 Unknown==-=-===eeeacauax 3 3 (*) 2 3 (*) Procedure C Total-====eceenona- 471 546 14 459 546 16 0-17 yoagsesmewwwiniwnse 4 5 (*) 4 5 (*) 18-34 years---=--cccca-oo 157 177 11 154 177 13 35-54 years-----eccecaa- 186 218 15 182 218 17 55 years or over=------- 82 89 8 78 89 12 Unknown=====eccececmaaa- 42 57 26 41 57 28 28 Table 15. Percent of hospital episodes underreported, spondent, including and excluding overreports by survey procedure and education of re- Including overreports Excluding overreports Survey procedure and education of Interview | Hospital Percent Interview | Hospital Percent respondent reports records underreported reports records underreported Procedure A Total-===mmmem==== 448 521 14 431 521 17 0-8 years elementary school-===memmmmcneennxn 105 129 19 97 129 25 1-3 years high school--- 127 154 18 122 154 21 4 years high school----- 153 170 10 149 170 12 1 year college or more-= 54 59 8 54 59 8 Unknown=======m=e=eeee=x 9 9 (*) 9 9 (*) Procedure B Total--==========- 526 558 6 | 510 558 9 0-8 years elementary school-===memameneeaa—- 153 159 4 145 159 9 1-3 years high school--- 124 134 7 122 134 9 4 years high school----- 165 175 6 161 175 8 1 year college or more-- 83 89 7 81 89 9 Unknown=====m====co===== 1 1 (*) 1 1 (%) Procedure C Totale====mem=mm=- 471 546 14 459 546 16 0-8 years elementary school====cenmammeeea—— 88 98 10 84 98 14 1-3 years high school--- 109 125 13 104 125 17 4 years high school----- 145 168 14 145 168 14 1 year college or more-- 63 71 11 63 71 11 Unknown-========meee-c== 64 81 21 61 81 25 Inappropriate---======== 2 3 (*) 2 3 (*) 29 Table 16. Percent of hospital episodes underreported, by survey procedure and number of chronic or acute conditions reported for the sample person, including and excluding overreports Survey procedure and Including overreports Excluding overreports number of chronic or acute conditions re- ported for sample person Interview | Hospital Percent Interview | Hospital Percent reports records underreported reports records underreported Procedure A Totale====mmmmunn- 448 521 14 431 521 17 None---==meeomccccacanan 98 129 24 97 129 25 1 or 2 conditions---=---- 205 236 13 199 236 16 3 conditions or more---- 145 156 7 135 156 13 Procedure B TOLAL~=~m a mmmmmmme 526 558 6 510 558 9 None-=-=e=cmmemcmneceee- 212 126 11 112 126 11 1 or 2 conditiong-==----- 276 293 6 270 293 8 3 conditions or more---- 138 139 1 128 139 8 Procedure C Total--===-mmceuea-x 471 546 14 459 546 16 None-=--=-memmmcoccaeeee 112 148 24 111 148 25 1 or 2 conditions------- 222 253 12 216 253 15 3 conditions or more---- 137 145 6 132 145 9 30 Table 17. Number and percent distribution of reported hospital episodes, by accuracy of reporting month of discharge and by survey procedure, excluding overreports Excluding overreports Accuracy of reporting month Procedure A Procedure B Procedure C of discharge Number | Percent Number | Percent Number | Percent Total-========-reemeeee——eeeee————— 431 100 510 100 459 100 Reported to have occurred before actual month of discharge’ --========c=cecccec=- 41 9 76 15 35 8 Reported to have occurred after month of discharge-----=======-=mmececcesmecom=oo= 59 14 46 9 37 8 Reported in actual month of discharge---- 331 77 386 76 380 83 Unknown=======m===seeemmemeee-ceooem—=——— 0 0 2 0 7 1 *The month of admission was asked for in Procedure A. The month of discharge was then computed for this table. Table 18. Number and percent distribution of reported hospital episodes, by accuracy of reporting length of stay and by survey procedure, excluding overreports Excluding overreports Accuracy of reporting length Procedure A Procedure B Procedure C of stay Number | Percent | Number | Percent Number | Percent Total-=mm==e=emmmeemeeem cece —e————— 431 100 510 100 459 100 More than actual days reported----------- 138 32 152 30 138 30 Fewer than actual days reported---------= 110 26 115 23 61 13 Exact number of days reported------=----- 181 42 241 47 259 57 Unknown========m=cee-ememmmecoaooen—————— 2 0 2 0 1 0 31 APPENDIX | PART 1: SAMPLING ERRORS The standard errors for the estimates in this re- port were calculated by use of the following equation: (The standard error is the square root of sa )- 2 2 m 2 2.2 SR "Tg Cy *t R°®% - Ro, ,) R m-T),2 Vy, X; Y, X; C2 m 2 where: i 7 a iZ1 (Y, - XY" 2 oy and oy x. are defined similarly. m = number of interviews for a given pro- cedure, Y.= number of episodes not reported for sam- ple persons in households interviewed by gh 5 the i interviewer. X. = total number of hospital episodes, based on hospital records, for all sample per- sons in households interviewed by the i interviewer. X=2z X = total number of hospital episodes, based on hospital records, for all sample persons in a survey procedure. IY, R = 3 = =hospitalization underreporting rate, i As may be seen, this formula treats the interview- ers for a procedure as clusters. From the underreport- ing rates for the 10 clusters for Procedure A or Pro- cedure B and 20 for Procedure C, the variance of esti- mates of underreporting may be generated, either for the total sample or for subgroups within the sample, The estimate of the variance follows the standard procedure for cluster sampling. (e.g., W.G. Cochran; 32 Sampling Techniques, New York: Wiley, 1953, p. 119.) This model was used on the advice of Dr. Leslie Kish, as a practical, useful approximation that fitted well enough, though not completely, the actual design which was somewhat more complicated. Dr. Kish supervised the computations and the construction and use of the tables of the sampling errors. For purposes of comparing Procedures A and B, since the interviewers were different and yet randomly assigned to procedures, it was possible to assume that the report rates in the two procedures were independ- ent. In comparing Procedures A and C, however, since half of the interviewers in Procedure C were also the interviewers for Procedure A, it was necessary to compute the covariance between Procedures A and C for estimating the variance of differences. Table I shows standard errors for selected char- acteristics of the sample as well as standard errors of differences between percentages for Procedures A and B, and between Procedures B and C. The Procedure B estimates include the results of the mail follow-up form. In general, and as demonstrated in table I, the standard error of one statistic is different from that of another statistic, even when the two come from the same survey. Since it was not feasible to compute standard errors for each of the many statistics in the report, ratios of the standard errors shown in table I to the standard errors of binomial variates, assuming simple random sampling, were computed. The ratios ranged from a low of about 0.7 to a high of about 2.2, The median value was 1.4. Rough estimates of standard errors of percentages shown in this report, which should be sufficiently accurate for most purposes, may be ob- IL 1.4 PQ/n, where P is the proportion of hospital episodes underreported, Q is the proportion reported, and n is the number of episodes in the sample. If a more conservative estimate of the variance is desired, use the upper limit of the ratio instead of the median as the constant multiplier, tained by the equation o Table I. Standard errors of underreporting percentages shown in this report and standard errors of differences between Procedures A and B and between Procedures A and C, for selected charac- teristics of the sample Characteristic of sample Standard errors of underreporting percentages X 107 Standard errors of differences X 10-2 Procedure A | Procedure B | Procedure C 9(A-B) 9(A-C) Total=mrme summon 2.5 1.0 Income Under $7,000=»=mm==mececnnen 3.0 1.3 2.2 3.3 4.1 B75 O00 Awe mo mmm mmc cai i i om 0 2.1 1.5 2.3 2.5 2.9 Type of respondent Self---==cemcmmmmem mmm 1.9 1.4 2.5 2.4 2.5 Proxy-=====================-- 3.0 2.0 2.3 3.6 2.6 Race White-=======cecemeececmecen= 2.0 1.6 1:8 2.5 2.0 ONL CE wwe mom 0.00 ctr 8 7.4 4.3 5.7 8.6 7.9 Age Under 1l7-=-=-ceecmmeccneenea= 4.0 1.0 3.4 4.4 5.2 18-54mmmmmmmm mmm mmm mmm mmm 2.1 2.1 2.0 3.0 2:9 55tm mmm mmm mmm mmm mmm ——— 4.6 1.9 4.2 5.0 4.5 Sex Male-====-==m===—ceereme————- 2.8 1.4 3.4 3.1 4.5 Female-=======mmmccccecmnn——— 2.4 2.5 1.2 3.5 1.8 Threat rating Most threatening--====-==w==- 4.3 4.1 5.2 5.9 9.4 Somewhat threatening------=--- 5+5 2.0 4.1 5.9 1.3 Not threatening---=-========- 23 1.7 2:5 2.9 3.3 Time interval between discharge and interview Under 30 week§=============== 1.4 0.9 X.9 1:7 1.8 31-53 weekg=======meemcemon== 3.5 2,9 2.5 4.5 4.3 Type of treatment Surgical-============-ce===-- 2.0 1.4 o7 2.4 2.3 Nonsurgical-=-=========wee==- 3.6 2.5 8 4.4 3.4 33 PART 2: ANALYSIS OF INTERVIEW ASSIGNMENTS The initial sample was composed of 600, 598, and 597 persons for Procedures A, B, and C, respectively, However, the data shown in this report is based on 462 persons for Procedure A, 456 for Procedure B, and 465 for Procedure C. The difference between the two sets of figures is due to nonresponse and deletions from the sample. Table II shows the results of the interview assignments, and lists the reasons for nonresponse and Table II. edit deletions. The deletions should not be considered part of the sample as they would nothave been included in the initial sample if they could have been detected. The major reason for nonresponse was that the assigned family could not be located. Follow-up of fam- ilies who had moved outside of the Detroit urbanized area was not attempted, Disposition of interview assignments, by survey procedure Disposition of interview assignment Survey procedure A B Cc Number of interviews assigned--======--eccacacaao_ 600 598 597 Number of interviews completed--=-====ceceaecaaoooo 516 492 500 Number of interviews not completed----====w-eo--- 84 106 97 Refusal-===mm mmm ee ee eee 8 9 8 Not at home-======= mmo ecco 15 22 32 Family not located========c-cccmmomoeeeeoo 59 71 53 Other====--mmmm occ ee eee eee 2 4 4 Number deleted during editing---=====eeeccocoao. 54 36 35 Sample person not listed in interview---------- 32 28 26 No hospitalizations for sample person during reference year=------ecoecmmeaaaaooooo 20 7 6 Other =====m- cm eee eee eee 2 L 3 Total number of persons included in the analysig--== mmm eal 462 456 465 34 APPENDIX II FORMS AND QUESTIONNAIRES USED IN THE STUDY Budget Bureau No. 68-6017; Approval Expires June 30, 1961 FORM NHS-S-14-1 (2-24-61) U.S. DEPARTMENT OF COMMERCE BUREAU OF THE CENSUS CASE ABSTRACT FORM 1. Serial No. 2. No.of discharges 3a. Name of hospital b. Location 4 PATIENT a. Name of patient g. Hospital No. of patient b. Address (Enter house No.; street; apt. No. or other description; h. Discharge (Month, day, year, time) city (or county); State) A.M. P.M. c. Telephone No. d. Age e. Sex f. Race i. Admission (Month, day, year, time) [} Male (T] White Femal, Noawhi AM. (CZ) Female (] Noawhite PM 5. NEAREST a. Name of nearest relative b. Relationship RELATIVE c. Telephone No. d. Address (Enter house No.; street; apt. No. or other descHption; city (or county); State) OR, If same as 4b, check here: | | 6. Discharge diagnosis (List in same order as shown on record) LEAVE BLANK 7. Operations 8. Remarks USCOMM-DC 13729 P-61 35 36 PROCEDURE A QUESTIONNAIRE tion. Such repetitive spaces are omitted in this illustration. The items below show the exact content and wording of the basic questionnaire used in the nationwide household survey of the U. S. National Health Survey. The actual questionnaire is designed for a household as a unit and includes additional spaces for reports on more than one person, condition, accident or hospitaliza- The National Health Survey is authorized by Public Law 652 of the 84th Congress (70 Stat. 489; 42 U.S.C. 305). All information which would permit identification of the individual will be held strictly confidential, will be used only by persons engaged in and for the pur- poses of the survey, and will not be disclosed or released to others for any other purposes (22 FR 1687). FORM NHS-S-14-A U.S. DEPARTMENT OF COMMERCE (3-20-61) BUREAU OF THE CENSUS ACTING AS COLLECTING AGENT FOR THE U.S. PUBLIC HEALTH SERVICE 1. Questionnaire of NATIONAL HEALTH SURVEY Questionnaires 2. (a) Address or description of location 3. Iden. [3.(a) 4. Sub- 5. Sample | 6. PSU 7. Segment 8. Serial No. Code | Reg. sample No. No. office weight A Code (b) Mailing address if not shown in (a) (¢) Type of ! (d) Name of special dwelling place ; Code living | [J Housing unit ! quarters | } |] Other ! A 9. Are there ony other living quarters, occupied or INSTRUCTIONS FOR Q. 9, 10 AND 11 vacant, in this building (apartment)?................... [J Yes [J No If “Yes, to questions 9, 10 or 11 apply definition of a housing unit to determine whether one or more additional questionnaires should be filled. 10, Does anyone else living in this building use YOUR ENTRANCE to get to his living quarters? ............. 0 Yes CI Ne 12 What is the telephone 13. In case I've overlooked anything, number here? what is the best time to call? Ask at all units except apartment houses: 11. Is there any other building on this property for people to live in- either occupied or vacant?........... [J Yes [No [J No telephone 14. RECORD OF CALLS AT HOUSEHOLDS Item 1 Com. 2 Com. 3 Com. 4 Com. 5 Com. Date Entire household [os ” =] == meesl ~~ De=rawssmms) (peesmmmess] = beewewewed paws seses Time Date 2 Col. No. Time § 3 § Date Col.No. _____ [-"-""f "See >» [-—-------"1 t+! F-------1 F------- = ¢ Time 23 Date =D col No. = Fr---pewenemadandiid oo... | bo. ....... feooo-....] rmmne «wwe 33 Col. No. Time < E Date Col. No. Time |= sd 41 1 Ir——1 rr 15. REASON FOR NON-INTERVIEW TYPE A 8 z [] Refusal (Fitt item 16) [C] Vacant - non-seasonal Interview not obtained for: [] No one at home- [] Vacant - seasonal repeated calls ®4l) pom [[] Usual residence elsewhere Cols Reason: | LJ Temporarily absent [J Armed Forces cn, Sor, Sas fies [7] Other (specity) [_] Other (specity) because: 16. Reason for refusal 17. TYPE "A" FOLLOW-UP PROCEDURE If final call results in a Type A non-interview (except Refusals) take the following steps: - Contact neighbors (caretakers, etc.) until you find someone who knows the family. 2. Find out the number of people in the household, their names and approximate ages; if names of all members not known, ascertain relationships. Record Teo information . .. in the regular spaces inside the questionnaire. 3. Find out if anyone in the housing unit is now in a hospital as a patient; if so, which person it is. This is done by asking the following question: 4. Is anyone in the household now in the hospital? [] Yes [J Ne [J] Don’t know [J No contact made (a) lt ‘Yes,’ -- Who? (Enter name) (Col. No.) 18. Signature of Interviewer 19. Code 1. (a) What is the name of the head of the household? (Enter name in first column) (b) What are the names of all other persons who live here? (List all persons who live here.) Last name (1) ~~ (") Do any of the people in this housshoid have a home sisewhere? No (leave on questionnaire) [] Yes (apply household membership rules; if not a household member, delete) (c) Do any (other) lodgers or roomers live here? [1 No [J Yes (List) (d) Is there anyone else who lives here who is now temporarily in a hospital? [CI No [] Yes (List) ———ee— (e) Away on business? [No [] Yes (List) meee (f) On a visit? [J] No [] Yes (List) =——————> (g) Is there anyone else staying here now? [1 No [] Yes (List) =m First name and initial (Check one box for each person) 2. How are you related to the head of the household? (Enter relationship to head, for example: head, wife, daughter, Relationship grandson, mother-in-law, partner, lodger, lodger’s wife, etc.) Head Age 3. How old were you on your last birthday? = Dade year White Negro 4. Race (Check one box for each person) tJ = — 5. Sex (Check one box for ea ch person) [CC] Male [7] Female If 17 years old or over, ask: [] Under 17 years Married i 6. Are you now married, widowed, divorced, separated or never married? 0] Artie 72 Diinmid [] Widowed [| Separated [] Never married If 17 years old or over, ask: 7. (a) What were you doing most of the past 12 months -- (For males): working, or doing something else? (For females): working, keeping house, or doing something else? If *Something else’’ checked, and person is 45 years old or over, ask: (b) Are you retired? [C] Under 17 years [] Working [CT] Keeping house [_] Something else NOTE: Determine which adults are at home and record this information. Beginning with question 8 you are to interview [1 Under 17 years ** ask: (b) Who was this? (c) How many times were you in a nursing home or sanitarium? for himself or herself, each ‘adult person who is at home. [J Athome™ [7] Noe at ome 8. Were you sick at any time LAST WEEK OR THE WEEK BEFORE? (That is, the 2-week period which ended [] Yes [No last Sunday)? (a) What was the matter? (b) Anything else? 9. Last week or the week before did you take any medicine or treatment for any condition (besides . . . which [] Yes [J No you told me about)? (a) For what conditions? (b) Anything else? 10. Last week or the week before did you haye any accident or injuries? (a) What were they? [CJ Yes [CI Ne (b) Anything else? 11. Did you ever have an (any other) accident or injury that was still bothering you last week or the week before? [] Yes [No (a) In what way did it bother you? (b) Anything else? 12. AT THE PRESENT TIME do you have any ailments or conditions that have lasted for a long time? [1 Yes [LJ No (If *'No’*) Even though they don’t bother you all the time? (a) What are they? (b) Anything else? 13. Has anyone in the family - you, your --, etc. - had any of these conditions DURING THE PAST 12 MONTHS? [] Yes [] No (Read Card A, condition by condition} record any conditions mentioned in the column for the person) 14. Does anyone in the family have any of these conditions? [] Yes [C1 No (Read Card B, condition by condition; record any conditions mentioned in the column for the person) 15. (a) Have you been in a hospital at any time DURING THE PAST 12 MONTHS? [] Yes [No Hes TTT Tere (b) How many times were you ir the hospital overnight or longer? No.of times 16. If baby under one year listed as a household member, ask: | — 1 (a) Was the baby bom in a hospital or at home? (Check proper boxes for A Hospieal ul Mone —_— If *‘hospital’’ in q. 16 and 1 or more in q. 15 ask: both mother and child.) 5 ! (b) Was this hospitalization included in the number you just gave me? C3 Yes, [CINe 7. 49). During the past 12 months has anyone in the family been a patient in a nursing home or sanitarium? [] Yes [No es No. of times For persons 17 years old or over, show who responded for (or was present during the asking of) q. 8 - 17. If person responded for self, show whether entirely or partly. For persons under 17 show who responded for them. R (for q. 8-17) [J] Responded for self-entirely 1Col.____Resp.; this person: [] Present and reported [] Present -did not report [C] Not present (or child) 37 Table I - ILLNESSES; IMPAIRMENTS AND INJURIES. \ Did you Ask for all illnesses and Ask if the entry in Col. (d-1) |Ask only if:|Ask for any entry in Ask only for: EVER present effects of old injuries: | is: 6 years old gol {1) oe 5 . {1:2 Impairments and injuries at any . : : or over and jthat includes the words: . Hone (a) If doctor talked to: An impairment, Slindgess. And oi a IIHR talk to | What did the doctor say it was? or poor vision, |Allergy®* Tumor aes Note a doctor] - - did he give It 0 medical a Symptom, or eye Asthma ‘Condition’ Bleedin Ne about name? or trouble of [Cyst “py: ” . . ang kind: [Geoweh Disease Blood Clot Pains (b) If doctor not talked to: 4 came from question 11 or 14: Stroke *'Trouble’’ Dorls Sores i Record original entry an aski(d-2)- (4-5) as What wos. the cause of ... 2 [Canyon |r vind ob... is it? og ok 5 Tequirec. (If "'Cause’’ is an injury, also | ®"ovg °F 11 i Infaction Weakness w Yok 3 1 * d or an allergy or stroke § 3 pe for all injuries during fill Table A) ordinary | ask: What part of the body is affected? = newspaper Ll) ec} hat pony of the body was hurt? print with |How does the allergy Ear or eye - (one or both) gl g What kind of injury was it? glasses? (stroke) affect you? Head - (Skull, scalp, face) al z g Anything else? Back - (Upper, middle, lower) Y 3 $ (Also, fill Table A for all Arm - (Shoulder, upper, elbow, Aloo injuries) lower, wrist, hand; one or both) Leg - (Hip, upper, knee, lower Sos, oot; one or both) (a) J(b (©) d-1) (d-2) (d-3) (d-4) (d-5) x x x x [J] Yes [J] Yes ! CI Ne [No [] Yes x 0 Vos x x 2 {No [J No Table Il - HOSPITALIZATION DURING PAST 12 MONTHS Col. Ques- | When did | How To Interviewer What did they say af the hospital the condition was vr No. tion many did they give it a medical name? of No. nights How many| Will you |How many] Was this . [per pital? were you [of these | need to [of these | person (If *‘they’’ didn’t say, ask): Y]son n the -- nights | ask Cols.|-- nights | still in E (Month, hospital? | were in (f) and were last | the hos- What did the last doctor you talked to say it was? a year) the past | (g)? week or ital on 12 the week | last (Entry must show '‘Cause,’” Kind,” and "Part of & months? before? Sunday Body” in same detail as required in Table I) —- night? (a) (b (c) (d) I. (x) f (g) (h) Mo: Lia [] Yes pv [Yes 1 I ights Yeo Nights Nights [Ne {_] None [J Ne Al Mo: 1A d [J Yes | ———u| [] Yes 2 N Nights CIN Yr: | Nights Nights [CJ No [] None ©; 18. (a) | have some questions about health insurance. We don’t want to include insurance that pays ONLY for accidents, [] Yes [No ut we are interested in all other kinds...Do you, your--have insurance that pays all or part of the bills when you |=. __~ ________‘— 7" | go to the hospital? Name of plan(s) If “Yes,” (b) What is the name of the plan (or plans)? Any other plans? (€) Who is covered by this plan (each plan)? (Check '*Yes,' in 18(a) for each person covered) 19. (a) Excluding insurance that pays ONLY for accidents, do you, your---have insurance that pays all or part [C] Yes [J No of the surgeon's bill for an operation? ~~ 0 EEE (Les LJRe If “Yes,” Name of plan(s) (b) What is the name of the plan (or plans)? Any other plans? (c) Who is covered by this plan (each plan)? (Check “Yes, in 19 (a) for each person covered) 20. (a) Do you, your---, etc. have insurance that pays any part of doctors’ bills for home calls and [] Yes [C1No office visits? 1 “Yes,” (b) Does it pay for home calls and office visits for most kinds of sickness? Name of plan(s) If "Yes," (c) What is the name of the plan? (d) Who is covered by this plan? If 17 years old or over, ask: [] Under 17 years 21. (a) What is the highest grade you attended in school? De | ; 3 ‘ 5678 i : ”" Ag he (Circle highest grade attended or check *'None'’) Colleges 123 435+ [] None mre i TR ees | (b) Did you finish the- -grade (year)? [] Yes [CJ No 38 Table | - TLLNESSES, TMPAIRMENTS AND TNJURIES. LAST WEEK |How |How many |If6- 16 If 17 years| Did you first notice . . . To About If 1 or Ask after completing last OR THE many [of these years old | old or over| (did it happen) durin, the how more days] iti WEEK BE- |days, |-- days ask: ask: iid 3 Dopp OR re Sten: | many in Col. condition for each person FORE did includ-| were you that time? viewer: | days (0) and J, If 1 « » .. cause you| ing in bed all H LAST during Col (e) see " ? if Yes to cut down |the or most of — ne WEEK the past |is check- res a 12" or in Col. on your usual Satur- | the day? ors th WEEK Check one |Did... start CON- 12 ed, ask: 3 Sou he, in |(q): activities for| days ou oep € R during the past | TINUE | months, pn | eo ol. (pX: Which? as much os a|and Jou om BEFORE, efore | Dur- | 2 weeks or if Col. has... |How Jecte, day? Sun- school how many 3 |ing |before that time?|(k) is kept you [many of Is this lost week | days did [Then tell | because days? ry mos. | 3 checked, | in bed these (Enter or ses boop mos. | (If during past lor the | for all days eo which Jof any |X on wee you from 2 weeks, ask): |condi- |or most |were® tatement | of the |{ine for before? work? tion of the during its you |condi- |each Check one For Go Which week, is on day? last post, in | tions condi- - emales to last week or Card A week erms of | you tion 3 No | Yes add) Col. the week or is an or the onli me named) | not count- | (n) re impair- old me (Go ing work l= ment; Votre? [Cards C- | about? : to around the other: F, as e Col. house? wise, jappro- V1 (k) STOP priate) (e) (0H (8)