Series 2 - Number 45 VITAL and HEALTH STATISTICS DATA EVALUATION AND METHODS RESEARCH (NN Z A 0 APY ny 0) dd h/ $9 3 RST Reporting Health Events in Household Interviews: Effacts of Reinforcement, Question Length, and Reinterviews U. S. DEPARTMENT OF HEALTH, EDUCATION, AND WELFARE Public Health Service Health Services and Mental Health Administration Vital and Health Statistics-Series 2-No. 45 For sale by the Superintendent of Documents, U.S. Government Printing Office, Washington, D.C. 20402 - Price 70 Series 2 DATA EVALUATION AND METHODS RESEARCH Number 45 TY © o U.S.8. Reporting Health Events in Household Interviews: Effects of Reinforcement, Question Length, and Reinterviews A methodological study designed to test the effective- ness of certain questionnaire designs and interviewing techniques used in the collection of data on health events in a health interview. DHEW Publication No. (HSM) 72-1028 U.S. DEPARTMENT OF HEALTH, EDUCATION, AND WELFARE Public Health Service Health Services and Mental Health Administration National Center for Health Statistics Rockville, Md. March 1972 NATIONAL CENTER FOR HEALTH STATISTICS THEODORE D. WOOLSEY, Director PHILIP S. LAWRENCE, Sc.D., Associate Director OSWALD K. SAGEN, Ph.D., Assistant Director for Health Statistics Development WALT R. SIMMONS, M.A., Assistant Director for Research and Scientific Development JAMES E. KELLY, D.D.S., Dental Advisor EDWARD E. MINTY, Executive Officer ALICE HAYWOOD, Information Officer DIVISION OF HEALTH INTERVIEW STATISTICS ELIJAH L. WHITE, Director ROBERT R. FUCHSBERG, Deputy Director RONALD W. WILSON, Chief, Analysis and Reports Branch KENNETH W. HAASE, Chief, Survey Methods Branch Vital and Health Statistics-Series 2-No. 45 DHEW Publication No. (HSM) 72-1028 Library of Congress Catalog Card Number 72-610272 ACKNOWLEDGMENT This study was conducted by the Survey Research Center, Institute for Social Research, The University of Michigan, under contract with the National Center for Health Statistics. Technical assistance was provided by Dr. Thomas Tharakan, Mrs. Barbara King, Mrs. Marion Wirick, Mrs. Carol Berlin, and Mrs. Diane Schoeff. The staff of the Community Health Association in Detroit supplied infor- mation on the sample for the study. Dr. George LeRoy, Director of Metropolitan Hospital, Mr. John Kerr, Assistant Director, and the medical staff extended special assistance. FOREWORD This study, conducted by contractual arrangement with the Survey Research Center, Institute for Social Research, The University of Michigan, is the second ina series of three studies designed to investigate the effects of some experimental interviewing techniques on the amount and quality of in- formation obtained during a health interview. (The other two studies are de- scribed in Series 2, Numbers 41 and 49.) The plan for this series was moti- vated by the findings of an earlier study on interviewer-respondent behavior also completed by the Survey Research Center. The basic study, which is described in Vital and Health Statistics, Series 2, Number 26, indicated that reporting in an interview can be more effectively improved by increasing the behavioral interaction of the respondent and the interviewer during the interview than by changing the basic attitudes of the respondent or increasing his levels of information. In view of this finding, it seemed thatimproved reporting might be obtained by the introduction of techniques used by the interviewer to encourage re- spondent reaction during the interview which would stimulate maximum recall. This approach, however, varied substantially from the usual practice of train- ing interviewers to behave in a standardized manner during an interview. The standardized manner; which was restricted to the asking of questions and re- cording of responses, was an attempt to reduce the known biasing influence on survey data that has been attributed to interviewer performance. The design of this series of studies has taken advantage of the fact that int2rviewers can influence respondents, and it has attempted to bring the potentially biasing behavior cues under control--in effect, to incorporate them as a part of the "standardized" behavior. Through the interaction be- tween the interviewer and the respondent it was expected that the systematic changing of the interviewer's technique would change the activity level of the respondent, thereby increasing the amount and quality of reported health in- formation. Because of the complex relationship between methods of interviewing, the performance of interviewers, and the reporting of respondents, the problem of obtaining accurate data in a household interview is not a simple one. The findings from this investigation of experimental interviewing techniques indi- cate that verbal "reinforcement of the respondent (i.e., appreciative com- ments by the interviewer following fruitful recall efforts by the respondent), question length, direct memory probing, an intensive interview, a diary pro- cedure, and a reinterview canhave importanteffects on survey interview data. More investigation is needed to determine the appropriateness of specific techniques for the collection of certain types of health information and to evaluate their effectiveness in terms of the validity, reliability, and amount of data reported. Elijah L. White Director Division of Health Interview Statistics CONTENTS Page Acknowledgment - === momo meee cmc ee iii Foreword -=e- mmm meee eee ee eemecmmmcme mem v Introduction == meme m ec eeceeeee eee ee 1 Summary of Main ResultS-==ememme cece caer mmme mmm mmm 4 Reinforcement == - occ eee ee eee 4 Question Lengthe seam come eee eee 5 Reinterviews m= comm cm cee eee eee 5 Pilot Studies-===== =m camo eee eee eee 6 Effects of Question Length on Reporting— Pilot Study I--=-ecc-cccccaaaooo 6 Effects of Reinforcement and Question Length on Reporting-Pilot Study II-- 7 Methods and Procedures-==e cocoa mmo eeeemccmcme mem 8 Experimental and Data Analysis DesSign--=-=cececocmo oma ommocmcceaoe 8 Sample Design---- ccm mmm moc eee eee 9 Response Rates === - mmm comme eee emcee 10 Interviewers dnd Training--e-e e-em om oo momo eee 10 Interview TechniqUeS-we~ewmcmcrcmcmcorcmmm meee rccenece meaner ———— 11 Reinterviews-- o-oo cece eee eee eee 12 Respondent Education-=e=eem eco mm omc emeeee 12 Dependent Variables--==-m me cm ome moc eeeeeem ee 12 Experimental Results-o=m meme oom coon eeee 16 Validity ==cccommm mm com cee m 16 Reliability === mmm meme eee emma -- 33 Number of Health Events Reported-eeeeee oom caecmcccccccccecceen 38 Discussion === eee m mmm eee eee eemcecmmmmeee 44 Education Differences in Ability and Bias--==--eccmmmmmccmcccccccaeeae 44 Social Status, Rapport, and Task Orientation---=eeccccmccocmccmcmaeoan 46 Summary of DISCUSSION === mm meen m eo eo 48 RETCTCIICES mmm om msm mt oo oo 5 0 2 Sw 0 0 0 0010 49 Appendix I. Sample Pages From Questionnaires----eceececeacacccmacaoo- 50 Interview Procedure A—Short Questions Without Reinforcement----=ee-- 50 SYMPtOMS= mmm mmm meee eee ee eer meee ——————— 50 Chronic Conditions--=ceeeccmcmcmmcmccmere remem annem e mn ———- 51 Physician VisitS-==ceeemcmc meme ma es 53 Interview Procedures B--Short Questions With Reinforcement--=eee-aaa- 54 SYMPLOME === mmm msm mm eee meee eee eee cccccmmeme—ce—————— 54 CNT ONIIC CC OTIATLT TNE wr 0-00 00 mm os es i sw 55 Physician Visit§~«eweccaccmcmmnmnmrnnnnencennmcnnanancnnneemenn—- 57 vii CONTENTS —Con Page Interview Procedure C--Long Questions Without Reinforcement--------- 58 SYMPLOMS === === === mmm mmm mmm mmm mmm mem meme em mom — 58 Chronic ConditionS------=cememmec meme m meme mmm ecm mmm mmm 59 Physician ViSit§----====mmmmeccm mmm —— eee mm ee mn 61 Interview Procedure D--Long Questions With Reinforcement------------ 62 SYMPLOMS === === == == == =m = mom mem mmm mmm mmm mmm ose mees 62 Chronic ConditionS===-==eccmcmmemm emcee creme meme m mmm 63 Physician Visit§---==---ccrmmmcommmmomco momo moomoo moomoo 65 Appendix Il. Forms Used In Study «mmm mmm mm mm mmm mmm mon 0 0 0 mm ammo 66 Interviewer's FOrm------ecmmmcmmmmme cece cme mmc meme mmm mmm mmm 66 Observation FOrme=-=--cecmom macro 67 Physician Summary Form-----e-remmmeme meme meme mmm mmm mm mmm mm 68 Original Interview Illness Table-~-----commmmmmmmee emo mmm mee 69 Reinterview Illness Table---eccmomme mc mcccm cme em meme me meme em — 70 viii REPORTING HEALTH EVENTS IN HOUSEHOLD INTERVIEWS: EFFECTS OF REINFORCEMENT, QUESTION LENGTH, AND REINTERVIEWS Kent H. Marquis, Ph.D., Charles F. Cannell, Ph.D., and André Laurent, Doct. (Sorbonne), Survey Research Center, Institute for Social Research, The University of Michigan INTRODUCTION Since its inception, the National Center for Health Statistics (NCHS) has undertaken a con- tinuing research program designed to assess the quality of its data and, more basically, to under- stand the survey interview process. Previous research has indicated that respond- ents make errors when relating factual infor- mation in interviews, and that such errors may be influenced by at least three classes of variables: (1) the nature of the information to be reported, (2) the method by which the information is re- quested, and (3) the response of the interviewer to the reported data, This study is concerned with the second and third variables. Several NCHS studies indicate some kinds of errors respondents tend to make when de- scribing their health. In recent research of this type, Madow! measured response error in re- porting chronic illnesses and conditions by per- sons known to have at least one chronic condition. The records of a large group health insurance plan were used to establish the fact of chronic illness. The data indicated that approximately 45 percent of those conditions shown in the medical records were not reported in the interview. Various ap- proaches to questionnaire construction had no great overall effect on the validity of these data, One of the major predictors of underreporting was a lack of impact of the condition on the respondent, High impact events included worry, long duration of the illness, imposed work limits, and medication, among others. Conditions accom- panied by these high impact events were much more likely to be reported than conditions with low impact. The Survey Research Center (SRC) of The University of Michigan in cooperation with NCHS has undertaken a new series of studies designed to study variables other than the nature of the information reported which might be responsible for variations in response error. One major study in the new series by Cannell, Fowler, and Marquis? was designed to investi- gate the effects of some psychological character - istics (attitudes, opinions, motives, perceptions, interest, health information, and abilities) of in- terviewers and respondents on the number of events reported.® The investigators were unable to find any significant association between these Several earlier validity studies compared respondents’ declared hospitalizations or visits to doctors with known records. These studies revealed that those respondents who were more accurate in their reporting tended also to report a larger number of health events. Since higher reporting does seem to suggest more accurate reporting, the number of reported events has been used as a dependent variable in later studies where better validity data have not been available. psychological-cognitive variables and the amount reported, Needless to say, this was contrary to expectations derived from both common sense and accepted interview theory. The research also involved an effort to col- lect detailed data on the kinds of behavior ex- hibited by questioner and respondent during the interview.” Their behavior was noted on special forms by an observer who took no other part in the household interview. When the behavioral data were related to the number of items re- ported, after correcting as much as possible for factors causing spurious correlations, it was ob- vious that the greater the rate of behavioral ac- tivity on the part of the respondent, the greater the number of items she reported. This held true regardless of whether the behavior was task oriented or irrelevant to the interview. Because of this correlation between behavior and number of items reported, the inferences were made that the validity of the data had been improved and that the improvements in the reporting were a function of things that went on during the inter- view rather than the more remote psychological and demographic characteristics of the respond- ent and interviewer. The study was descriptive rather than ex- perimental, so it was impossible to isolate vari- ables which caused or accounted for the general activity level of the respondent, However, there was a very high correlation between interviewer behavior activity level and that of the respondent. This relationship suggested that systematically changing the interviewer's technique would change the respondent's activity level, thereby increasing both the amount and quality of reported health information. bThe kinds of behavior noted were interpersonal and interview or task oriented behavior of both respondent and interviewer. Respondent interpersonal behavior might include laughing, joking. and furnishing unnccessary information, while the interviewer interpersonal behavior included similar items such as asking nonhcalth questions of the respondent, flattery, and praise. The respondent's task-oriented behaviors mentioned were such things as elaborations and asking for clarifications. The interviewer's task-oriented behaviors noted were directive and nondirective probes, clarification of questions, and so forth. The overall pattern of findings from this first study led to hypothesizing a ''cue-search’ model of the household interview, It was assumed that a household interview was so out of the ordinary that respondents could not generalize a well-de- fined set of previously held attitudes, feelings, or expectations toward it, Therefore, the respond- ent looked to the interviewer as a source of cues about how to behave. Thus, in a subsequent study inthe SRC-NCHS series by Marquis and Cannell’ the investigators attempted to manipulate systematically two sets of these cues and to assess their effects on the number of health items reported. The first set of cues consisted of positive reinforcement state- ments given by interviewers in response to infor - mation supplied by the respondent. Although this feedback is generally overlooked in considering the interview process, it was felt that it might be a major source of uncontrolled variation. Bringing these cues under some systematic con- trol could greatly influence the respondent and the quality of the information he provided subsequent- ly. Results indicated that a technique in which the interviewer used a positive reinforcing state- ment after the respondent mentioned a health item resulted in a 25 percent increase in the number of such items reported. This increase occurred for most classes of items-——recent and past, medically attended and personally treated, em- barrassing and neutral—as well as for items reported on a ''yes-no'' list or in response to less structured questions. The second set of cues was inserted at the beginning of the experimental interviews in the form of a symptom list, a list of items contain- ing a wide variety of health symptoms and con- ditions (serious and minor, embarrassing and neutral, frequent and rare), which was read to the respondent, The interviewer asked her to respond ''yes' to each item she had ever ex- perienced, and "mo" to the others. The purposes of this experimental procedure were to accustom the respondent to thinking about her health and to convey the idea that a broad range of health items would be covered. However, the number of re- ported items was not increased by this prepara- tory procedure. The present research was designed to pro- vide some clarification of the results of the pre- vious reinforcement experiments. One hypothesis tested was that an interview using reinforcement reduces response error to a greater extent than an interview not employing reinforcement. An adequate test of this hypothesis required a de- parture from two important procedures used in previous research. The reinforcement technique mentioned in connection with a previous study? involved at least three variables, In addition to verbal reinforcement, the interviewer smiled and looked up at the respondents in the experi- mental group but neither smiled nor looked at the control group respondents, Also, the ques- tionnaire for the experimental group contained long questions and section introductions aug- mented by clichés and redundancies; the con- trol questionnaires did not. In the present study the question length variable was manipulated in- dependently of reinforcement, and the interviewer used eye contact and smiles for respondents of both groups. An alternative hypothesis was also advanced. It stated that reinforcement merely produces more "'yes'' answers to list-type questions than when reinforcement is not used and that the net effect of this reinforcement on the reporting of listed chronic conditions is not only to reduce the under - reporting but also to increase overreporting sub- stantially. Previous validity studies on health reporting had been designed mainly to test the extent of underreporting. The present study was deliberately designed to examine overreport- ing as well. With the resulting data it therefore was possible to measure changes in both over- and underreporting rates and any ''tradeoff" occurring between them. Research has suggested that question length may be an important determinant of reporting be- havior. A recent series of studies, carried out under the direction of Matarazzo and Saslow, is of special interest. They investigated the effects of interviewer behavior on respondent behavior during different types of interviews. Among these studies is a subset dealing with the relationship between the length of time the interviewers spoke and the length of time the respondents answered. * Briefly, the results indicated that when interview- ers talked more, so did respondents. (The find- ings seem similar to those from the behavior observation health studies described above show- ing the amount of total interviewer and respondent behavior to be highly correlated.) The data also suggested that respondents might talk more in answer to longer questions, thereby giving more health information. In the present study the in- vestigators also explored the possibility that in- creasing interviewer speech duration (by man- ipulating question length) would yield additional, valid health information from the respondent. Finally, while not directly derived from prior theory or research, the effects of a reinterview on the validity of chronic condition reporting were tested in this research. It is often assumed that the reinterview, sometimes conducted to provide continuing data in panel studies and sometimes done to check the accuracy of interviewer's per- formance, produces more valid information than that obtained in the original interview, The little research in the area, most of which has been done by Ferber] does not show clear-cut find- ings. The classic study in the field by Neter and Waksberg$ suggests that reinterviews obtain less accurate data than do original interviews, This present study was designed to explore the comparative validity of original interviews and reinterviews and to provide data on effects, if any, which the experimental variables might have on respondent performance when interviewed a sec- ond time. cNeter and Waksberg® uncovered what was termed a “conditioning effect” of reinterviews. This effect is not to be confused with the effect of verbal reinforcement or verbal conditioning tested in this study as the two are distinct phenomena. Neter and Waksberg found a 9 percent loss in the number of jobs reported between the second and third interview. Both interviews had asked for the same information. SUMMARY OF MAIN RESULTS Previous published research and pilot studies indicated that variables having the greatest ef- fects on response accuracy in personal interviews are probably the variables most closely asso- ciated with the interview process itself rather than with more remote phenomena such as participant attitudes and demographic and social character - istics. This study was designed to test the effects of interviewer reinforcement, length of question, reinterviews, and respondent education on the validity, reliability, and quantity of respondent reports of health information in a household in- terview, A special respondent sample not repre- senting a national cross section was used. A measure of validity of respondent reports was approximated by comparing interviewee infor- mation on chronic conditions with information obtained from physicians. The results obtained were not as anticipated initially. They indicated that the effects of dif- ferent ways of conducting interviews were medi- ated by the level of education of the respondent. Procedures which increased accuracy and re- duced bias for one education group had opposite effects for the other education group. Care should be taken in generalizing the re- sults of this study since the sample was not typi- cal of the U.S. population. In addition, many of the findings are based on trends in the data, the components of which are not statistically signi- ficant. Nevertheless, it appears to be a sound conclusion that small variations in either the asking of questions or reinforcement procedures can produce variations in the quality of data. The effects of the experimental procedures are summarized below. dThe initial report from this study did not include education of the respondent as a variable in the analysis. However, further analysis of the same data, which was supported in part by a general purpose research grant from the National Center for Health Services Research and Development (PHS Grant No. HS00252), indicated the importance of education as a variable in explaining the effects of the different interviewing procedures. Subsequently, the initial report underwent modifications to introduce the effects of the respondent’s education. Reinforcement The interviewer's use of a positive rein- forcing verbal statement following reports of morbidity increased the accuracy and reduced the bias of chronic condition reporting only for the group of respondents who had not completed high school. It had the opposite effect of decreas- ing accuracy and increasing bias for respondents who had graduated from high school. Reasons for these effects are not clear, although it is possible to rule out one plausible hypothesis. It might be expected that less educated respondents were not as able to report their chronic condi- tions as were more educated respondents and that, therefore, reinforcement had the effect of increasing the ability of less educated respond- ents to recall and report their chronic sickness. The high education respondents already possessed the ability to recall and report accurately, and hence, would not benefit from the reinforcement procedure. However, it is possible to infer from the data that the two education groups did not differ in ability to report chronic conditions accurately. Therefore, the different reinforce- ment effects appear not to be due to their influ- ence on different initial levels of ability to report. Several untested hypotheses are offered to account for the mediating effects of education on the relationship between reinforcement and re- porting accuracy. Less educated respondents interviewed by more educated (and, consequently, high status) interviewers ''maturally' establish a task-oriented interpersonal atmosphere rather than a personal or social relationship. Rein- forcement for adequate performance in this task-oriented context has the effects predicted by theory because reinforcement is perceived by the respondent as appropriate and necessary. For the more educated respondent, the in- terviewer-respondent relationship is hypothe- sized to be more interpersonally oriented and the respondent less concerned about adequate task performance, possibly because of the ap- parent simplicity of the reporting requirements. Reinforcement, because it is friendly and sup- portive, may accentuate the tendency to per- ceive the relationship as personal rather than professional or it may be perceived as inappro- priate behavior for an equal status relationship. Either or both perceptions may result in biased or inaccurate reporting. Question Length Increasing the length of questions about chronic conditions by introducing redundancies and irrelevant material (not by furnishing more pertinent information such as clarifying phrases or examples) significantly increased reporting accuracy and reduced bias for the more educated respondent group. In contrast, these long ques- tions tended to decrease accuracy and increase bias for the less educated respondent group. The positive and negative effects of long questions as mediated by respondent education cannot be fully explained. It is hypothesized that the additional presentation of key elements of a question and the additional time a long question allows for the consideration of an answer are useful to the higher educated respondent who may need his attention directed more exclusively to the information-giving aspects of the interview situation. Why long questions produced less accu- racy and greater bias for the low education group is more conjectural, Possibly the long questions, including the irrelevant statements, merely con- fused the less educated respondent. Reinterviews Accuracy and bias of reporting did not differ to any great extent in reinterviews and initial interviews, but again, the observed differences were mediated by respondent education. Gener- ally, but with some exceptions, the more edu- cated group reported more accurately during the reinterview than during the initial interview. Asin the original interview, the use of long questions without reinforcement produced the highest aver- age accuracy for this group. The less educated group showed no improvement during the rein- terview. It may be that the more educated respondents benefited from a second opportunity to think about their chronic conditions and were thereby able to overcome initial tendencies to respond "no" to a question when they were unsure of their answer. PILOT STUDIES In order to test some basic assumptions and the feasibility of the method and techniques to be used in the main study, several pilot studies were conducted, Two of these preliminary studies are described briefly here. The first study was designed to see how the use of long questions affected the length of respondent answers and the amount of information given by respondents. The purpose of the second pilot test was to ex- amine the effects of both long questions and reinforcement on answers given by respondents, Effects of Question Length on Reporting—Pilot Study | To explore the possibility that increasing the duration of interviewer speech by manipu- lating question length would yield more health information from the respondent, a pilot study was conducted, using the services of two inter- viewers on the Survey Research Center staff. The interviewers received about 8 hours of training on the use of the questionnaires and the special procedures for selecting households and respondents. They were instructed to ask the questions exactly as worded and to give clari- fication only when it was absolutely necessary and then only by repeating the relevant part of the question. Furthermore, they were told to accept the respondent's first answer rather than to probe for more information and were instructed to give no feedback to the respondent. Respond- ents were chosen from blocks selected atrandom from two census tracts in Jackson, Michigan. The tracts contained intact white families, of moderate income, with a low proportion of persons over 65 years of age and a high proportion of native-born citizens. Interviewer speech duration was varied by the use of long and short questions. An interview with 28 questions was created, covering a wide variety of health topics and a wide variety of types of questions (for example, open and closed, specific and general). In an attempt to rule out the possibility that respondents gave longer answers to long questions merely because these asked for more information, length was added to these questions in three ways: (1) by redun- dancy or asking the same thing twice, with vari- ation in the grammatical structure of each re- quest, (2) by inserting clichés, such as '""The next item is...," and (3) by introducing extraneous information, for example, '"We ask this of all respondents’ or ''The health service wants to know about...." The exact words used in a short question always appeared intact in a long one, usually at the end of the question. The pilot study design employed three ques- tioning procedures. Twenty-seven interviews were obtained, nine in each of these three pro- cedural groups (A, B, and C). The basic ques- tionnaire was divided into four roughly equal parts. In procedure A the first and third parts of the questionnaire contained long questions, while the second and fourth parts asked short questions. For procedure B short questions were used in the first and third parts, and long questions in the other two parts, In procedure C short ques- tions were used in all parts. A possible fourth procedure, employing only long questions, was not tested as it was assumed to be potentially detrimental to useful respondent performance. Two dependent variables, each thought to be affected by the length of the question asked, were tested: (1) the length of time respondents talked, defined as the number of seconds from the end of the question to the end of the response minus any interviewer speech which intervened; and (2) the percent of questions for which at least one item of information was reported. Table 1 shows the average length of time the respondent took to answer a question in relation to the question length. In the interviews using both short and long questions, question length did not seem to affect the time duration of the respondent's answer, Interviews using only short questions, however, did obtain an average answer length which was slightly lower than that obtained in the other interviews, but this difference is considered to be inconsequential. This pattern of findings is not consistent with the results of other research, but no reason for the discrep- ancy can be given at this time. Table 2 describes the effect question length had upon the number of questions for which one or more health items were reported. Again, the Table 1. Number of respondents and average duration (in seconds) of answers per question, by question length, pilot study I Number Lrennpe ¢ of re- uration o Question length spond- GHEVGLS ents (in sec- onds) Long questions in interviews with both long and 5.58 short questions--- sl : 1 Short questions in interviews with both long and short questions--- 5.69 Short questions in interviews with short questions only-=====m=m====- 9 5.31 results were somewhat different from those ex- pected. In interviews using both long and short questions respondents gave affirmative answers © to 40 percent of the long questions and 38 percent of the short questions. Therefore, within an inter- view using questions of varying length, the length of a specific question did not appear to have any effect on reporting. However, interviews using only short questions obtained affirmative answers to only 29 percent of the questions. This amounts to a drop in the affirmative answer rate of about 25 percent, Therefore, the pilot study suggested that while question length per se may have some effect on respondent answering behavior, it is the combination of Jong and short questions that yields the most health information. Within an interview using different question lengths, the length of an individual question has little or no influence on the answers as long as the question has been preceded by a series of long questions. c“Affirmative answer” is defined as the reporting of at least one health item. Table 2. Number of respondents and per- cent of questions for which any health information was reported, by question length, pilot study I Percent of Number questions : of re- for which Question length spond - information ents was re- ported Long questions in interviews with both long and short questions--- 40 18 Short questions in interviews with both long and short questions--- 38 Short questions in interviews with short questions only--==--cec-ou=u- 9 29 Effects of Reinforcement and Question Length on Reporting—Pilot Study | In a second pilot study consideration was given to the effect of question length with and without reinforcement on the number of health items reported. Three procedures were tested: (1) short questions with reinforcement, (2) long questions’ with reinforcement, and (3) long ques- tions without reinforcement. The dependent vari- able was the number of questions for which at least one health item was reported. Two experienced interviewers queried a total of 48 respondents residing in Jackson, Michigan. Respondents were sampled by a procedure es- sentially similar to that described for the first pilot study, while the reinforcement procedures used were similar to those described below for f Actually, following the results of the first pilot study, a mixture of long and short questions was used. This procedure is given the abbreviated designation “long questions” for convenience of presentation. the main study. However, only about 1 hour of training was devoted to the use of the reinforce- ment procedures, in contrast to the several days of instruction given for the main study. Twelve questions were asked of all respond- ents. Since answer duration was not measured, the interviewers were instructed to probe and clarify when necessary to obtain answers which met the objectives of the questions. The percent of questions for which at least one health item was obtained by the three inter- viewing procedures is shown in table 3. The data indicated that interviews in which long questions were used with or without reinforcement obtained more affirmative responses than did interviews with short, reinforced questions. The main finding was that the effects of reinforcement and long questions were additive—they could be com- bined into one interview procedure and would thus produce more affirmative answers than either reinforcement or long questions used alone. Table 3. Percent of questions for which health information was reported, by in- terview procedure, pilot study II Percent of questions for Interview procedure which infor- mation was reported Short questions with reinforcement -------- 66 Long questions! with reinforcement -------- 84 Long questions without reinforcement -------- 72 IConsisted of a mixture of long and short questions. METHODS AND PROCEDURES This section describes in detail the design and procedures used in the main experimental study. Experimental and Data Analysis Design The research was designed as a 2 x 2 fac- torial experiment with one interview for half the respondents and two interviews for the other half (see figure 1). The data were analyzed sepa- rately for two education groups. The independent variables used were rein- forcement and question length. Reinterviews and respondent education served as mediating vari- ables. The dependent variables included indexes of accuracy (under- and overreporting, total error, and net bias) and amount of reported health information. More detailed descriptions of these variables are given later in this sec- tion. | | Not reinterviewed | 7, (2nd wave of respondents) | | Reinterviewed | 7s wave of respondents) | ¥ Number of respondents= Figure I. 50 per cell Short questions Long questions Without With reinforcement reinforcement Experimental design. Sample Design Sample size.—It was determined from esti- mates of the frequency of occurrence of 13 selected chronic conditions inthe U.S. population’ and estimates of over -and underreporting rates’ that approximately 600 respondents were needed to establish the reliability at the .05 level of a 25 percent difference on the dependent variable (due to the main effects of reinforcement and question length). It was then assumed that the number of respondents could be reduced some- what because of the experimental controls and sample weighting described below. Since a sam- ple size of 400 for the original interviews was within budget limitations, this number became the final sample size. Respondents, —Respondents were selected from a population of persons who belonged to a prepaid health insurance plan and who came to one of the plan's several clinics between February and July 1968. To reduce extraneous sources of variation in health ‘reporting, a somewhat homogeneous population of respondents was selected. The sample was restricted to white females, aged 17-60, who were residents of the greater Detroit metropolitan area. In an attempt to increase the relative amoun. of reporting variance attribut- able to the experimental interviewing technique, a weighted selection of "sick" respondents was used. One-third of the persons in the original population had none of the special chronic con- ditions; two-thirds had one or more. But in the final sample chosen for the study, 88 percent of the women had at least one chronic condition and 12 percent had none of the conditions of interest, Most of the analyses which follow in- clude only those respondents who had at least one of the chronic conditions. Fifty percent of the respondents (half of each of the four experimental groups) selected for an interview were also scheduled for rein- terviews. To facilitate data collecting, all re- spondents selected for original interviews during the first half of the data collection period were designated as the reinterview group. A small number of respondents whose scheduled original interviews came during the second half of this period were also reinterviewed in order to bring the reinterview sample sizes to approximately 50 persons for each of the four interviewing procedures. Other controls, —In the sample design several variables were stratified or controlled. Respond- ents were assigned to 15 geographic area groups, four age groups, and four ''sickness' groups. Sickness was defined by the number of chronic conditions checked 'yes' on the Physician Sum- mary Form (see appendix II). Using these groupings, clusters with four respondents each were formed. The clusters were geographic, based on large areas corre- sponding roughly in size to one or more of the metropolitan census tracts. Within each of the clusters the four interviewing procedures were assigned at random, and within each interviewer's assignment an attempt was made to have a balanced variety of interviewing procedures to use as well as various age groups and sickness levels among the respondents interviewed. The design effect on the variances due to clustering is assumed to be zero and has not been calcu- lated for this study. While it would have been desirable to con- trol interviewer variance by random or strati- fied assignments, this was not financially feasi- ble as the interviewing area covered several hundred square miles. In this study interviewer variance might also be attributed to geographic location of the respondent. All interviewers used all combinations of procedures (short questions, long questions, re- inforcement, no reinforcement, no reinterview, and reinterview). Thus, interviewer differences were not seriously correlated with the effects of the interviewing procedures used. On the other hand, requiring interviewers to learn and use a variety of techniques probably introduced some error which might have reduced the contrast among the procedures. With this compromise design and a less than optimal sample size it was expected that con- trast between the experimental interviewing methods would be small. Specifically, the effects of reinforcement and question length might appear somewhat unstable; the reinterview effects would show up but be even more unreliable; and it would be difficult to detect and test for an inter- action effect of the procedures. Response Rates The overall response rate, shown in table 4, was 90 percent for the original interview and 92 percent for the reinterview, The response rates for the four procedures ranged from 84 to 95 percent, Interviewers and Training Ten white female interviewers were employed in this research, Four had extensive experience with the Survey Research Center (SRC), two had been on the staff several months prior to this study, and four were new additions to the perma- nent staff. Of the last group one had extensive experience in market research interviewing. All 10 had training in basic SRC interviewing pro- cedures prior to this study. Interviewers received about 1 week of class- room training covering the basic definitions and procedures of health interviewing and were drilled in each of the four experimental interviewing methods. Considerable classroom work was directed toward acquiring a thorough understand- ing of the concepts underlying both the question- naire and the techniques basic to all of these interviews. However, most of the classroom sessions were devoted to supervised role-playing interviews, after which immediate feedback was given by the research staff. The interviewers were not told that medical information was available about the respondents. Following class- room training, interviewers spent about 1% weeks in practice interviewing and tape-recording of these practice interviews. Practice interview protocols and tapes were reviewed soon after completion, and the interviewer was given a critique of her performance shortly after the material was received. Comments were made, usually via telephone, and immediately after- ward a written evaluation was sent to the inter- viewer along with her original protocols. Practice with reinterviews occurred 2 weeks after the start of interviewing and was handled in essen- tially the same way as the training for original interviews, Table 4. Number of sample persons and response rates for original interviews and rein- terviews, by interview procedure Original interview Reinterview Interview procedure Origi- | Eligible a Re- Eligible Respond- Ro- nal respond- fritens sponse | respond- {tar sponse sample ents viewed rate ents viewed rate Per- Per- Number of persons Number of persons cent cent All procedures-- 511 448 404 90 224 205 92 Short questions: Without reinforce- ment---=-===-=-—-- 128 117 105 90 56 50 89 With reinforce- ment-----======--- 127 107 99 93 53 49 92 Long questions: Without reinforce- ment------==-====-- 130 114 96 84 57 53 93 With reinforce- ment------==-=-=--- 126 110 104 95 58 53 91 10 Throughout the entire interviewing period for the study an attempt was made to observe each interviewer on at least two occasions each week. A member of the staff accompanied the interviewer during these two interviews, took systematic notes on strong and weak points of her interview performance, and transferred the notes to an Observation Form (see appendix II) for the interviewer's use. The form served as a basis for discussing the completed interview with the observer. After the interviewer finished studying the Observation Form, she was asked to return it to the central office for the records. Although observations were thought to be nec- essary and helpful, there was concern about their effect on the data, as about half the obser- vations were made by males and both interviewer and respondent were female. Sex of observer was not recorded, but the data indicate no re- lationship between the presence or absence of an observer and the accuracy of reporting chronic conditions. This lack of relationship held true for all experimental procedure combinations within each respondent education group (see table 16). Interview Techniques A description of the interviewing techniques, reinforcement and question length, is given be- low. The basic features of each interviewing technique used were actually written into the questionnaires. Further detail regarding each procedure is available in the sample pages from the questionnaires in appendix I. Reinforcement,—In those interviews where reinforcement was used, each time the respond- ent reported an instance of illness or health service utilization (up through question 14 of the original interview and through question 17 of the reinterview), the interviewer recorded the answer, looked up at the respondent, and used one of several reinforcing statements, After question 14 or 17, reinforcing statements were used but were worded and scheduled at the interviewer's discretion. The reinforcing statements were printed in the following form at the top of each applicable page of the reinforcement questionnaires, Thank you We're interested in that. Mm-hmm That's the kind of information we need. I see (REPEAT ANSWER JUST GIVEN) Yes That's important, 0O.K. We need to know that. The interviewer combined a statement in the first column with one in the second column to form a complete reinforcing statement. Infre- quently, a large number of consecutive reinforcing statements were required, and the interviewer was given the option of occasionally using only a short statement from column one to avoid ‘monotony. A great deal of training and observation time was devoted to the problem of making the reinforcing statements sound natural. The inter- viewers eventually were able to do this satisfac- torily. The natural effect would seem desirable for the interview situation, yet, there is some feeling in retrospect that it is not essential for successful application of the reinforcement pro- cedure. Question length, —The contrast in question length was basically between an interview using only short questions and an interview using a mixture of short and long questions, heavily weighted with the latter. The following examples show how short questions about chronic condi- tions were lengthened: Short questions Long questions a. Asthma? a. Asthma is the first one we need information about, Have you had asthma? b. Hay fever? b. What about hay fever? Have you had that? c. Thyroid trouble c. Another areaofinterest or goiter dur- to the Health Service ing the past 12 is thyroid and goiter months? trouble, Have you had any thyroid trouble or goiter during the past 12 months? For every lengthened question an attempt was made to add only words which neither changed the meaning of the question nor clarified what was wanted by the question. For the long question interviews an attempt was made to intersperse short questions at differ - ent intervals, since pilot study data indicated this technique was effective, However, any probes that were used to follow up a main question were the same length for all interview procedures. Because of the nature of the probe questions, a respondent who volunteered information about illness in response to an open question was asked proportionately more of these short ques- tions than one who either did not furnish infor- mation on the open questions or who reported illness only on the structured, checklist ques- tions. This particular aspect of the design would tend to minimize the effect of question length on numbers of items reported—the more infor- mation reported, the more the long question interview became like the short question inter- view, Reinterviews Whenever possible, the reinterviews (with approximately half the respondents in the origi- nal sample) took place 2 weeks after the initial interview. The reinterview content was very similar to that of the original interview except for the addition of some "filler" questions about health insurance and a slight change in the order of the questions. An attempt was made to hold everything else constant, Respondent Education The educational level of the respondent was used as a mediating variable in the data analy- sis. The amount of education was coded from the respondent's answers to the following questions asked near the end of the original interview: 18. What was the highest grade you attended in school? 18a. Did you complete that grade? Respondents who had not completed the 12th grade (had not graduated from high school) were classified as "less educated." Those who had completed the 12th grade or more were classi- fied as "more educated." Respondent education was not used as a stratification control in the sample design be- cause this information was not available on the health records which served as a basis for sam- ple selection. Therefore, respondent education was not distributed evenly among the experi- mental procedure groups. The analyses of vari- ance employing respondent education as a control variable were calculated using an unweighted means solution in order to compensate for un- equal cell sizes arising from the procedure x education sampling variation. Dependent Variables Several types of dependent variables were used for this study. The main emphasis was on the validity of reporting 13 selected chronic con- ditions. The other dependent variables were the average frequency of reporting a particular kind of health event and the correlations between the number of events reported on the original inter- view and on a reinterview, Undev- and overveporting of chronic con- ditions.—The basic dependent variables were indexes reflecting the degree of agreement be- tween the respondent and a physician about the presence of each of 13 chronic conditions. ®Chronic condition items. The 13 chronic conditions were selected from those reported frequently in the Health Interview Survey. For this study they were included in a list of 19 con- ditions and placed approximately in the middle of the interview. The respondent was asked if she had any of the following during the preceding 12 months: asthma, chronic bronchitis, hay fever, chronic skin trouble, hemorrhoids, hernia, ulcer, and varicose veins. She was then asked if she had ever had arthritis, heart trouble, stroke, hypertension, or diabetes. ® Physician information. Information about the respondent's status on the 13 chronic con- ditions was obtained from the Physician Sum- mary Form (see appendix II). For each of the 13 conditions, the physician checked one of these categories: 1. Definitely or probably present (scored as "yes" in data analysis) 2. Definitely or probably not present (scored as "mo" in data analysis) 3. Don't know, no information For each condition and for each respondent the data from the interview and from the Sum- mary Form were combined and assigned to one of five match categories: . Respondent, yes - physician, yes Respondent, no - physician, yes Respondent, yes - physician, no Respondent, no - physician, no Missing data (a response other than ''yes" or 'no'" from either source) Ax WN = Comparing the respondent's answers to the phy- sician data provided a measure of reporting validity. However, it should be pointed out that the physician data were assumed to be only ap- proximations of the ''truth' about respondent chronic conditions. There were several reasons for this: (1) Even though each physician filled out the Physician Summary Farm immediately after the respondent left the clinic and the re- spondent's complete medical record was in the hands of the physician, time constraints undoubt- edly prevented a thorough scrutiny of the record's contents; (2) some of the chronic condition diag- noses were tentative at the time the form was filled out; (3) physicians probably differed with respect to how certain a diagnosis must be before it was entered as a "probably present’ on the Physician Summary Form; and (4) some time elapsed between the date the physician's form was filled out and the date of the house- hold interview. In some cases the time interval was as great as 6 months. Both health and diag- nostic precision could be expected to change within these time intervals.” e Indexes of reporting error. For statisti- cal purposes it was assumed that some propor - tion of respondent-physician disagreement was valid and that this kind of disagreement was not confounded with assignment to experimental in- terviewing procedure groups. It was also assumed that some of the disagreement represented re- spondent reporting error and that the extent of this error was reflected in an attenuated form within the higher education group of respondents there was a slight negative correlation between the recency with which the physician filled out the Physician Summary Form for the respondent and reporting error. That is, for these respondents, the amount of reporting error was slightly less for those cases in which physician data were recent relative to the time of the interview. This relationship was extremely small, confined only to the more educated group of respondents, and was not confounded with any of the other experimental procedures. Because of the existence of this and other kinds of bias in the matching procedures, the reader should not attempt to interpret the error scores presented herein as exact or even close approximations of the true rates of misreporting of chronic conditions. On the other hand, the existence of this kind of bias, distributed evenly among procedures, does not preclude making meaningful comparisons among the reinforcement and question length experimental procedures for their relative effects on reporting error. by the calculated mismatch rates described be- low. Two basic mismatch rates were calculated based on the following scheme: RESPONSE TO EACH CHRONIC CONDITION ITEM Respond- Physi- Match ent cian cate- says: says: gory: Yes Yes A No Yes B Yes No C No No D (1) Underreporting, eh the number of condi- tions for which the respondent said 'mo" and the physician said 'yes' divided by the total number of conditions for which the physician said "yes." (2) Overreporting, 5p; the number of condi- tions for which the respondent said "yes" and the physician said ''mo' divided by the total number of conditions for which the physician said "no." The underreporting and overreporting index scores were combined into two addi- tional indexes of reporting error: an index of total error and an index of net bias. It was felt that these additional indexes gave the clearest picture of the underlying dy- namics of recalling and reporting chronic condition information. (3) Total error, 248 + C+D ; the unweighted sum of the rate of underreporting and the rate of overreporting. In this study the oppor - tunities to overreport were much greater than the opportunities to underreport. Therefore, the index of total error contained a ''dispro- portionate'' representation of observed un- derreporting errors. An index of total error which represents each type of error accord- ing to the actual frequency of occurrence, not Soupited here, would have the form iB + A+B+C+D * 14 (4) Net bias, 12 - ~ ; the rate of underreporting minus the rate of overreport- ing. Again, this index gave equal weight to both overreporting and underreporting er- rors, making it easier to talk about the underlying psychological processes involved in recall and reporting. However, itdistorted the fact that opportunities for overreporting were much more frequent than for underre- porting. A correctly weighted net bias score could be computed for this or any other single study as its . These dependent variable index scores were proportionate scores. The underreporting and overreporting index scores could range from .00 to 1.00. The total error score had a possible range from 0.00 to 2.00. The net bias score had a theoretical range from -1.00 through 0.00 to 1.00. Practically speaking, however, the ranges of the total error and net bias scores were much less. In the absence of completely consistent, deliberate lying the underreporting and overre- porting scores should not be expected to average above 0.50. Therefore, the practical range of the net bias score was between -0.50 and +0.50, Average scores much beyond these practical ranges would represent the phenomenon of re- spondents having available accurate information to report to the interviewer but reporting answers exactly opposite of the truth. Signal detection theory,” developed in the field of sensory psychology, uses special statis- tical and experimental procedures which combine the total error score and the net bias score along with certain theoreticalassumptions to make esti- mates of two psychological parameters involved in reporting: the accuracy or sensitivity of recall and the bias in the decision about how to report recalled information. By using certain experi- mental procedures and statistical techniques and by making certain assumptions about human sen- sation and memory, itis possible toestimate when an experimental interviewing procedure has ac- tually changed the respondent's ability to recall information, This is opposed to the case in which an experimental technique has merely changed the respondent's decision to say ''yes' or 'mo'' when he is uncertain of the correct answer, Whilé it would have been desirable to make these two kinds of estimates for the experimental procedures in this study, the data necessary to make empirical estimates of the two parameters were not collected, and it was not feasible to make the necessary assumptions about human memory in order to derive the separate esti- mates from theory. Therefore, when changes in accuracy of recall were observed as a function of the independent variables in this research, it could not be determined if these changes had been mediated solely by a change in the bias (what to report when uncertain) parameter. Improvement in reporting accuracy may also (or solely) have been a function of improvement in the actual ability of respondents to recall chronic condition information, At present, it seems that the ob- served relationships between the independent and dependent variables provide useful- knowledge despite the problem of pinpointing the exact me- diating processes. All cases for which the physician indicated that none of the 13 chronic conditions was pres- ent were deleted from the chronic condition (validity) analyses, Thus, the sample for whom the indexes of reporting accuracy and bias were computed consisted only of respondents who were potential over- or underreporters, Quantity rveporvted.— Another objective of this study was to ascertain the effect of reinforcement, question length, and reinterview upon the amount of illness and health service utilization reported. A description of the dependent variables used in these analyses is given here, ® Chronic conditions, In the original inter- view (questions 6 and 7) and reinterview (ques- tions 9 and 10) there were two lists of chronic conditions, Together, they contained 19 chronic conditions, 13 of which were used in connection with the validity analysis described above, For analysis of quantity reported, the number of ""yes'' answers on both lists were combined to form a variable, regardless of whether answers were obtained to all 19 items, ® Total chronic and acute illness. All re- ports of injuries, present effects of old injuries, and other kinds of conditions, either chronic or acute and present within 2 weeks of the original interview or 4 weeks of the reinterview, were eligible for inclusion in this comprehensive variable, All instances of conditions which met the objectives of the question for which they were reported were included. Symptoms of such mal- adies were included only if the respondent could not report the underlying cause of the symptom. Otherwise, the cause of the symptom was in- cluded and the symptom itself was deleted, Mul- tiple symptoms relating to the same cause, even if the cause was vague, were deleted and the cause retained. Also omitted were redundant re- ports of sickness, reports of normal menstruation and pregnancy, and symptoms reported only in question 1 (symptoms list). If the interviewer was in doubt about whether to retain an item, she asked a series of screening questions to determine eligibility, The screening questions on the originial interview are questions T1-T3 on the Original Interview Illness Table (see appendix II), They are questions TO-T3 on the Reinterview Illness Table (see appendix II), Answers to these questions were used to deter- mine if the health item was retained. ® Symptoms, The first health question in both the initial interview and reinterview asked the respondent if she ever had had each of 19 health symptoms. The list may be found in appendix I, The respondent's "symptoms score consisted of the number of "yes'' answers given to items on the list, If one or more items for a respondent had answers which could not be clas- sified either '"yes' or ''no," that respondent was excluded from data analysis for the entire symp- tom variable, ® Medicine and treatment taken in the past 2 weeks. In the original interview each respondent was asked if she had taken medicine or treat- ment for any condition during the past 14 days and, if so, what she took, At the end of the inter- view the respondent was again asked about med- icines or treatments taken during the preceding 2 weeks, In coding, the total number of undupli- - cated medicines and treatments was recorded, Data on the numbers and names of medicines and treatments taken were not obtained in the rein- terview, ® Physician visits, A complex set of proce- dures and definitions was used by interviewers to ascertain the number of times a respondent received the services of a physician during an approximate 3-month period, For the initial interviews the respondent circled the dates on a calendar on which she talked to a doctor or went to a doctor's office or clinic for herself during the 10-week period prior to the 2-week reference period of the in- terview. The interviewer cleared up any ambi- guities or misunderstandings and confirmed the correct number of circles for the 10-week ref- erence period. This number, subject to modifi- cation by the following series of probe questions, was the value the respondent received on the variable "physician visits, 10 weeks." Because the definition of ""use' in regard to physician services might not have been entirely clear to the respondent, the interviewer asked whether the respondent had done any of the fol- lowing during the reference period: saw a doctor in an emergency room, saw a doctor at respond- ent's home, or talked to a doctor over the phone (excluding calls only to make appointments). If the respondent answered 'yes' to any of these, she was asked for the date of the contact, The interviewer entered any new information on the calendar and corrected the numbers entered for the 10 weeks on the questionnaire. The procedure was somewhat simplified for the reinterviews. The respondent was handed a calendar with the original 10 weeks outlined in blue and the 4 weeks following that in red. If the reinterview took place more than 2 weeks after the original, the 4 weeks outlined in red ended some time before the day of the rein- terview. Thus, the recall reference period could differ for each respondent. Although data were obtained on physician contacts in the period between the original interview and the reinter- view, these data were not used in the present study. The physician visit data are those for the 10-week reference period which ended 2 weeks prior to the original interview and at least 4 weeks prior to the reinterview. ® Dentist visits for 12 months, Original in- terview question 13 and reinterview question 17 were worded identically, In these questions the respondent was asked how many times she had gone to the dentist during the 12 months prior tothe interview. Therefore, the days included in the 12-month reference periods differed between the original interview and the reinterview by about 2 weeks. For the reinterview score the respond- ent was asked to state how many dentist visits she made during the most recent 2 weeks, This number was subtracted from the total figure given for 12 months to obtain the reinterview dentist visit number. This resulted in the rein- terview reference period being 11% months and the original interview reference period 12 months. This difference is assumed to be inconsequential when original interview data are compared with reinterview data, ® Hospitalizations in 12 months, The re- spondent was asked how many times she had been in a hospital or nursing home overnight or longer during the 12 months preceding the in- terview, This information was obtained from question 12 in the original interview and ques- tion 13 for the reinterview, The equating of the 12-month reference periods was done in the same way as previously described for dentist visits, EXPERIMENTAL RESULTS validity Experimental treatment effects on the ac- curacy of chronic condition veporting in the original intevview,—The data presented here indicate that reinforcement and question length had effects both on the error and on the bias in- volved in reporting of the 13 selected chronic conditions. The effects of the experimental vari- ables were mediated by respondent education so that the variables which improved reporting for the less educated respondents interfered with good reporting for the more educated respond- ents. Conversely, the procedures which aided 16 the reporting of the respondents with more ed- ucation had a detrimental effect on the reporting of the respondents with less education. The underreporting, overreporting, total error, and net bias index scores are given in tables 5-8 and are shown graphically in figures 2 and 3. The reader is cautioned that these dependent variable scores cannot be projected to any other population because a special sample was used in the study, and the physician ''crite- rion'' data are not to be considered as completely valid. These indexes were computed for the pur- pose of examining comparative effects of the dif- ferent experimental interviewing procedures. UNDERREPORTING 70 60 S0 443 0 387 wo 7 p 7 Short questions Long questions Short questions Long questions OVERREPORTING sem 7 With reinforcement INDEX SCORE Without reinforcement N\ INDEX SCORE > o Short questions Long questions Short questions Long questions LESS EDUCATION MORE EDUCATION Figure 2. Index scores of underreporting and overreporting of chronic conditions in original interviews by re- inforcement, question length, and respondent education. MEAN TOTAL ERROR 70 644 564 rg 7 537 w 468 3S .386 & 338 > Vy 5 o Short questions Long questions Short questions Long questions NET BIAS or reinforcement 70 7 a5 Vv With reinforcement .506 & 50 7 3 352 7 729 a» 40 322 : 7 8 30 244 230 244 2 2 20 a2 7 o 10 % 0 % WY 7 Short questions Long questions Short questions Long questions LESS EDUCATION MORE EDUCATION Figure 3. Mean total error and net bias index scores for the reporting of chronic conditions in original inter- views by reinforcement, question length, and respondent education. The underreporting index scores (table Sand figure 2) indicate that, for respondents who had not completed high school, the use of reinforce- ment in interviews significantly reduced under- reporting error, However, reinforcement signifi- cantly increased underreporting error for the more educated respondents, These data also in- dicate that question length affects underreporting. For less educated respondents long questions tended to increase underreporting (not statis- tically significant), while for more educated re- spondents the long questions significantly reduced underreporting. Although education interacted with reinforcement and question length to affect underreporting, within each education group the two experimental procedures had additive effects and did not interact to determine error, The overreporting errors (table 6 and figure 2) were generally low (7 to 14 percent) and were much less affected by the experimental proce- dures. The small, nonsignificant trends in the data suggest that reinforcement reduced over- Table 5. Number of respondents and index scores of underreporting of chronic conditions in original interviews, by reinforcement, question length, and respondent education Less education More education Interview procedure Number of Index Number of Index respondents! score respondents score Short questions: Without reinforcement----- 42 .387 53 L443 With reinforcement-=-==---- 50 225 39 +575 Long questions: Without reinforcement----- 44 L443 40 «305 With reinforcement=---=-=--- 36 .349 46 L463 INot included are three respondents for whom education was not ascertained and>5l persons for whom the physician indicated there were no chronic conditions present, Summary of analyses of variance Less education More education Source of variation d.%. M.S. F value d.f. M.S. F value Reinforcement (R)==--====-==-== 1 .696 bg, 77 1 .921 h5,78 Question length (Q)-======== 1 .342 2.34 1 .688 b4.32 R Xx Qu====== tattle 1 . 049 0.33 1 .008 0,05 EXrOF=mmmmmemisac sam ammm mma ee 168 . 146 174 .159 bp < .05. NOTE: d.f. = degrees of freedom; M.S. 18 = mean square, reporting in both education groups. This trend runs counter to what might have been expected. One hypothesis was that reinforcement or reward on the part of the interviewer every time the re- spondent reports a sickness will result in the respondent "making up" sicknesses so that she will gain further approval of the interviewer, The data from this study indicate that the rein- forcement procedure did not produce this effect. On the contrary, for both education groups, re- interview tended to produce fewer false positive reports of chronic conditions than interviews which did not include reinforcement, Table 7 and figure 3 present the total error index scores which represent the unweighted sum of the individual underreporting and over- reporting index scores, Since underreporting scores had a higher mean and variance than overreporting scores, the total error scores tended to show the same effects of experi- gardless of question length, the reinforcement mental procedures and respondent education Table 6. Number of respondents and index scores of overreporting of chronic conditions in original interviews, by reinforcement, question length, and respondent education Less education More education Interview procedure Number of Index Number of Index respondents! score respondents! score Short questions: Without reinforcement ------ 41 143 53 .091 With reinforcement--==w-w-- 50 113 38 .069 Long questions: Without reinforcement------ 44 121 40 .081 With reinforcement=-e=eee-- 36 .119 46 .074 INot included are three respondents for whom education was not ascertained, 51 per- sons for whom the physician indicated there were no chronic conditions present, and two persons for whom physician data were incomplete, Summary of analyses of variance Less education More education Source of variation d.f. M.S. F value! | d.f. M.S. F value! Reinforcement (R)===ve=ecee--- 1 .011 0.58 1 .009 1.19 Question length (Q)====ee-=-e- 1 .003 0.15 1 .000 0.04 R X Qee-mmemmmmcc ccc eee 1 .008 0.41 1 .003 0.32 Error------eceeecccmcccccaaaa- 167 .019 173 .008 INone of the F values is statistically significant (p<.05). NOTE: d.f. = degrees of freedom; M,S. = mean square. Table 7. Number of respondents and mean total error index scores for the reporting of chronic conditions in original interviews, by reinforcement, question length, and respondent education Less education More education Interview procedure Number of Index Number of Index respondents! score respondents’ score Short questions: Without reinforcement------ 41 .530 53 .534 With reinforcement--------- 50 .338 38 644 Long questions: Without reinforcement------ 44 .564 40 .386 With reinforcement -=------- 36 468 46 «537 INot included are three respondents for whom education was not ascertained, 51 per- sons for whom the physician indicated there were no chronic conditions present,and two persons for whom physician data were incomplete. Summary of analyses of variance Less education More education Source of variation d.E. M.S. F value d.£. M.S. F value Reinforcement (R)=======w==== 1 .934 b5,96 1 . 749 bg, 47 Question length (Q)=====-===== 1 .205 1.59 1 L724 4, . 32 R X Qe--cmmmmmmmm meen emcee 1 .115 0.73 1 .017 0.10 Error-----=-eeeeeececnocannax 167 +157 173 .168 bp <.05. NOTE: d.f. = degrees of freedom; M.S. = mean square. observed in connection with underreporting. For less educated respondents, reinforcement signifi- cantly lowered total error while long questions increased it (the increase was not significant), For the more educated respondents, reinforce- ment significantly increased total error and long questions significantly reduced it. Within education groups, the experimental procedures did not interact but produced their effects as additive main effects. 20 The net bias index scores are shown in table 8 and figure 3. They indicate that higher educated respondents tended to exhibit more bias in the interview than respondents who had not graduated from high school. The effects of reinforcement and question length on net bias were also mediated by respondent education and were somewhat unstable for the less educated group. For the more educated group, long ques- tions reduced bias and reinforcement increased Table 8. chronic conditions in original interviews, respondent education Number of respondents and mean net bias index scores for the reporting of by reinforcement, question length, and Less education More education Interview procedure Number of Index Number of Index respondents score respondents! score Short questions: Without reinforcement------ 41 . 244 53 .352 With reinforcement--==----- 50 .112 38 .506 Long questions: Without reinforcement ------ 44 .322 40 L244 With reinforcement --------- 36 .230 46 .389 INot included are three respondents for whom education was not ascertained, 51 per- sons for whom the physician indicated there were no chronic conditions present, and two persons for whom physician data were incomplete. Summary of analyses of variance Less education More education Source of variation d.f. M.S. F value d.f. M.S. F value Reinforcement (R)--=-==c===ao 1 +374 3.32 1 1.123 b6.67 Question length (Q)===vv-w--- 1 .366 2.12 1 0.664 b3,94 R X Quemmememmcccc cm ccccaaaea 1 .026 0.15 1 0.001 0.01 Error----—-—ceeemmmcccmeceeeeo 167 .173 173 0.169 bp <.05. NOTE: d.f. = degrees of freedom; M,S. = mean square. it. For the less educated group, the trends were in the opposite direction; long questions increased bias and reinforcement reduced it, To summarize, the more educated group tended to underreport at a higher rate and exhibit a higher level of total error and more net bias than did the less educated group. It should be noted, however, that when conventional interviewing procedures were used (short ques- tions without reinforcement), the rounded total error scores were the same for both education groups—,53, It was only when the experimental procedures were introduced that the groups showed different levels of response error and bias, It can also be seen that the two groups achieved the same low levels of error under interviewing conditions most appropriate (on the basis of findings in this study) to their education levels: the lowest average total error score was ,34 for the less educated group, 21 achieved by using short questions and rein- forcement; the lowestaverage total error achieved by the more educated group was .39, a result of the long question, no reinforcement interview procedure, Experimental treatment effects on the ac- curacy of chvonic condition veporting in lhe reintevview,— A second interview covering the same topics as the original interview was con- ducted for about half (205) of the respondents. The reinforcement and question length proce- dures to which the respondent was exposed originally were repeated during the second in- terview, The data presented here show the cumulative effects of the reinforcement and question length procedures. Because of the small sample sizes, the effects were not statistically significant. The trends in the reinterview data were, in general, similar to those observed in connection with the original interview: for less educated respondents, reporting accuracy was generally improved: by reinforcement; for more educated respondents, accuracy was generally increased by long questions. The best reporting performance for the less educated group was obtained by using short questions with reinforcement. For the more educated group, the best reporting was obtained by the interviewer using long ques- tions without reinforcement, The average index scores on the four measures of variation for the eight experimental groups are given in table 9. Reinforcement reduced underreporting for less educated respondents and increased it for more educated respondents. Long questions tended to increase underreporting among the less educated group but decrease it among the more educated group, especially in the case when reinforce- ment was not used. In general, average under- reporting was relatively higher for the more educated respondents, The overreporting data were quite unreli- able but suggested a somewhat different picture from that observed in connection with the orig- inal interview, Because of the extremely low F values, these conclusions should be consid- ered highly tentative, As in the original inter- view, for less educated respondents overreport- ing in the reinterview was highest in the short question, no reinforcement group. The other 22 combinations of procedures obtained approxi- mately the same level of overreporting (lower than the short question without reinforcement procedure in the less educated group but higher than any of the combinations for more educated respondents). However, in contrast to the orig- inal interview, the reinterview data suggest a possible interaction between reinforcement and question length on overreporting among both the more educated and less educated respondents. In the reinterview, reinforcement tended to lower overreporting when short questions were asked and to raise it when long questions were used. The total error data suggest that the previ- ously observed independent main effects of rein- forcement and question length were similarly operative here for the less educated respondents, Reinforcement lowered total error for this group and long questions increased it, The two vari- ables were additive and did not interact. For the more educated group, reinforcement and ques- tion length appeared to interact in determining total error, The long question without reinforce- ment procedure produced an extremely low level of total error in comparison with all other combinations, which produced total error at approximately the same level (between .603 and L615). The effects of the reinforcement, question length, and respondent education variables on net bias were similar to those observed previ- ously in connection with underreporting and total error, For the less educated respondents, the net bias index score declined when rein- forcement was used and increased when long questions were used. There appeared to be no interaction between these two experimental in- terviewing procedures within the less educated group. Net bias scores were uniformly high for the more educated group in all interview proce- dure combinations except long questions without reinforcement. In this procedure the net bias score was about half of what it was in the other procedures (table 9). Thus, the cumulative effects of the experi- mental interviewing procedures of reinforcement and long questions were essentially the same in the reinterview as in the original interview, The one possible exception to this trend was that in reinterviews of the more educated group of Table 9. Number of respondents and underreporting, overreporting, total error, and net bias index scores for the reporting of chronic conditions in reinterviews, forcement, question length, and respondent education by rein- Index score for chronic condition reporting Respondent education and Number of | interview procedure respondents Under - Over- Total Net reporting | reporting | error? | bias? Less education Short questions: Without reinforcement ---eeeeeee-- 20 .400 .166 .566 .234 With reinforcement-==-eececececaax 25 .230 .131 .361 .099 Long questions: Without reinforcement------cee-ao 22 .504 121 .625 .383 With reinforcement--=-==ececececeea- 21 .380 . 140 +521 .240 More education Short questions: Without reinforcement----veeee--- 27 495 .120 .615 .375 With reinforcement --ee-eeccacaaaac 20 .521 .082 .603 .438 Long questions: Without reinforcement ---eeecee--- 26 .286 .086 .373 .200 With reinforcement -eeeeeecececcaaac 24 .509 .099 .608 L410 Includes only respondents who were reinterviewed, but excludes whom education was not ascertained and 18 persons there were no chronic conditions present. 2May not add or subtract exactly due to rounding error. Summary of F values from analyses of variance for whom the 1 two respondents for physician indicated Source of variation for Source of variation for respondents with respondents with less education? more education Dependent variable Reinforce- Question R x Q Reinforce - Question qx ment (R) length (Q) ment (R) length (Q) Q Underreporting----- 3.04 2,27] 0.08 2.10 1.66 1.32 Overreporting------ 0.07 0.40 | 0.92 0.54 0.25 2.18 Total error-------- 3.33 1.66 0.36 1.66 1.87 2.02 Net bias----------- 2.24 2,43 0.00 2.42 1.34 0.70 INone of the F values is statistically significant (p <.05). Degrees of freedom = 1,84, 3pegrees of freedom = 1,93. 23 respondents, the beneficial effects of long ques- tions were obtained only when reinforcement was not used. When reinforcement was combined with long questions in the reinterview, underreporting, total error, and net bias index scores were at approximately the same high levels as when long questions were not used. Validity of reinterview data compared with original interview data.—In order to compare the various indexes of reporting obtained in the original interviews and reinterviews, data in tables 10 and 11 are presented for only those respondents who were included in both interviews. One of the main conclusions from these data was that reinterviews tended to reduce the amount of underreporting error made by the more educated respondents. Reinterviews did not reduce under- reporting error (in fact, tended to increase it) for less educated respondents, There were some exceptions to these trends: the short question with reinforcement reinterview did not produce an increase in underreporting with less ed- ucated respondents; and both the short ques- tion without reinforcement reinterview and the long question with reinforcement reinterview did not reduce underreporting error for the more educated respondents. Finally, although reinterviews did help the more educated re- spondents reduce underreporting error, this type of error was still generally higher in their re- interviews than in those of the less educated group. The effects of the reinterview on overre- porting error were less clear. The data in- dicated that overreporting index scores tended to increase in reinterviews for both education groups when the standard type of interviewing procedure (short questions without reinforce- ment) was used. The trend was in the same direction for most of the other types of re- interviews. From these data it seems reason- able to conclude that overreporting error was not reduced in an important way and showed a tendency to increase when reinterviews were employed. As mentioned above with respect to under- reporting, reinterviews brought about a slight improvement for higher educated respondents who were interviewed using either one of two 24 of the experimental procedures. However, when interviewing was conducted by the conventional procedure (short questions without reinforce- ment), the reinterview did not produce better data than did the original interview regardless of the respondent's educational level. Since these find- ings are contrary to what might be expected, the next section of this report explores the possibility that original interview and reinterview data may be combined in some way to produce lower in- dividual and total error index scores than would be obtained from original interview data only. The comparative total error and net bias index dara for original interviews and rein- terviews are presented in table 11. Total error was not reduced in any of the four experimental groups in reinterviews with the less educated respondents. Between the two interviews the total error increased most when the short question without reinforcement procedure was used. There were intermediate levels of in- crease when long questions were used either with or without reinforcement. Almost no change occurred when the short question with reinforce- ment procedure was used. The lower educated group showed the largest increase in net bias scores in reinterviews using long questions. However, as for the total error scores, there was almost no change in net bias scores when the short question with reinforcement procedure was used in the reinterview. For the group of respondents with more education, the data showed that total error was reduced in reinterviews for two of the four experimental procedures. When short questions and reinforcement were used, total error in the reinterview was .60 compared to a score of .70 for this same group in the original in- terview. The most interesting result, however, concerns the reduction in totalerror for the long question without reinforcement procedure, This procedure yielded the lowest rate of total error for these respondents in the original interview (.46). When these respondents were reinterviewed using long questions without reinforcement, the total error dropped to .37. This is the lowest average rate of total error for any of the ex- perimental procedures for the higher educated group. It is almost identical to the lowest rate Table 10. Number of respondents and underreporting and overreporting index scores for the report- ing of chronic conditions in original interviews and reinterviews, by reinforcement, question length, and respondent education Underreporting index score Overreporting index score Number Respondent education of and interview i i respond- | Original _ Original _ procedure ents! inter- Reiss Difference inter- id Difference view view view View Less education Short questions: Without reinforce- ment--=-=-=-c-=c--= 20 .346 L400 -.054 .134 .166 -.032 With reinforcement-- 25 .230 .230 .000 L126 L131 -.005 Long questions: Without reinforce- ment--==cemmcemc=co- 22 .452 . 504 -.052 L131 L121 +.010 With reinforcement-- 21 +331 .380 -.049 .138 .140 -.002 More education Short questions: Without reinforce- ment=---=-=-=cc=---- 27 .488 .495 -.008 .097 .120 -.023 With reinforcement -- 20 .638 .5321 +.117 .065 .082 -.017 Long questions: Without reinforce- ment-------e-u--=-- 26 .364 .286 +.078 .094 .086 +.,008 With reinforcement-- 24 L454 .509 -.055 .090 .099 -.009 Includes nly respondents who were reinterviewed, but excludes two respondents for whom edu- cation was not ascertained and 18 persons for whom the physician indicated there were no chronic conditions present, Summary of F values from analyses of variance! Source of variation for respond- | Source of variation for respond- ents with less education? ents with more education? Variable Reinforce- Question R x Q Reinforce=- Question R x Q ment (R) length (Q) ment (R) length (Q) Underreporting: Original interview-------- 2,03 1.55 0.00 1,90 3.15 0.12 Reinterview---c=ceececaooo- 3.04 2,27 0.08 2,10 1.66 1.32 Overreporting: Original interview---=----- 0.00 0.02 0.07 1.04 0.37 0.64 Reinterview------cmecneano 0.07 0.40 0.92 0.54 0.25 2,18 INone of the F values is statistically significant (p<.05). 2Degrees of freedom = 1,84. 3Degrees of freedom = 1,93. 25 Table 11, and respondent education Number of respondents and total error chronic conditions in original interviews and reinterviews, by reinforcement, and net bias index scores for the reporting of question length, Respondent education Total error index score Net bias index score and interview oy, Co. Lo. prodedure Original Reinter- | Differ- original Reinter- | Differ- irre - view ence ph view ence Less education Short questions: Without reinforcement-- 20 .480 .566 -.086 L212 .234 -.022 With reinforcement----- 25 .356 +361 -.005 .104 .099 +.005 Long questions: Without reinforcement -- 22 .584 +625 -.041 322 .383 -.061 With reinforcement----- 21 470 «521 -.051 +192 L240 -.048 More education Short questions: Without reinforcement-- 27 +335 .615 -.030 .390 «375 +.015 With reinforcement ----- 20 .703 .603 +.100 +372 L438 +.134 Long questions: Without reinforcement-- 26 «437 373 +.084 +270 .200 +.070 With reinforcement ----- 24 «543 .608 -.065 .364 L410 -.046 ! Includes only respondents who were reinterviewed, cation was not ascertained conditions present. but excludes two respondents for whom edu- and 18 persons for whom the physician indicated there were no chronic Summary of F values from analyses of variance 1 Source of variation for Source of variation for respondents with less education? respondents with more education’ Variable Reinforcement (R) Lensih (© R x Q| Reinforcement (R) DEAE R x Q Total error: Original interview- 1.83 1.52 0.00 1.41 2,78 0.03 Reinterview-------- «33 1.66 0.36 1.66 1.87 2.02 Net bias: Original interview- 1.79 1.26 0.01 2,28 3.26 0.23 Reinterview-------- 2.24 2.43 0.00 42 1.34 0.70 "None of the F values “Degrees of freedom Degrees of freedom 26 = 1,84. 1,93. is statistically significant (p <.05). of total error achieved by the ''best' group of less educated respondents in the reinterview (short questions with reinforcement—,36). Re- interviews of the more educated group decreased the net bias score for three of the interviewing procedures but increased it for the long question with reinforcement procedure, Net bias was lowest in the long question without reinforcement rein- terview, the procedure which also produced the lowest rate of total error, The data suggest that meaningful improve- ment in reporting accuracy can be obtained by reinterviewing more educated respondents if long questions are asked and reinforcement is not used. Reinterviews with less educated re- spondents did not improve reporting accuracy. Special combinations of original interview and reintevview data,— Analyses were made of various ways of combining data from original interview and reinterview sources to produce error and bias rates which were lower than those from the best of either source alone. Results of two ways of combining data from both interviews are shewn in this section: a procedure which attempts to minimize underreporting or false negative errors and a procedure designed to minimize overreporting or false positive errors. ® Combination to minimize false negative errors, As long as the respondent did not say "no" when asked about the presence of a selected chronic condition in both the original interview and the reinterview, the response to that item was scored ''yes." A 'yes' was scored if the respondent said ''no'' in the original interview and "'yes'' in the reinterview, ''yes' in the firstinter- view and "no" in the reinterview, ''yes'' in both interviews, or ''yes' in one interview and ''don't know" in the other interview, However, a respond- ent who said "don't know'" in both interviews was not scored "yes." ® Combination to minimize false positive errors, Data were also combined so that a respondent received a score of "mo" on any chronic condition item unless she reported the condition in both interviews, Data using these scoring methods were matched against the information provided by the physician, and the four dependent variable index scores were recalculated, The underreporting and overreporting data using each of these two combination scoring procedures are shown in table 12, Table 13 contains the total error and net bias scores for the two combination ap- proaches, In tables 14 and 15 these total error and net bias scores are compared with those from the original interviews and reinterviews separately. Data in tables 12 and 13 indicate that the recalculation of the dependent variable scores did not alter the previously observed effects of respondent education and interviewing pro- cedure on reporting errors. As expected, the statistical procedure designed tominimize under- reporting or false negative error produced lower underreporting index scores than those calculated by a procedure intended to minimize overreport- ing or false positive errors, The same kind of phenomenon occurred with respect to the over- reporting index data. The calculation procedure designed to minimize overreporting errors yielded lower average overreporting index scores than did the procedure designed to minimize underreporting errors. In order to understand the net effects of these apparent tradeoffs between overreport and underreport data, it is necessary to look at the total error index scores and the net bias index scores (table 13), With one minor ex- ception, the calculation which minimized false negative reports produced lower total error index scores and lower net bias scores than did the calculation which reduced false positive errors, While the latter procedure reduced over- reporting error in the data, it did so at the expense of introducing more than a compensat- ing amount of false negative errors. Regardless of which of the two calculation procedures were used, the relative effects of reinforcement, ques- tion length, and respondent education on total error and net bias remained as observed in previous analyses: reinforcement improved the reporting of less educated respondents and hindered that of the more educated respondents; the long question interview interfered with op- timal performance in the less educated group and promoted it in the more educated group. The fact that net bias scores were lowest when data from the original interviews and re- interviews were combined to minimize false negative reporting seems to reflect the possi- 27 Table 12. conditions in original Number of respondents and index scores of interviews and reinterviews underreporting and overreporting of chronic (combined to minimize false negative and false positive errors), by reinforcement, question length, and respondent education Underreporting index score Overreporting index score Respondent education Feunee and interview eg ponds Combined to Combined to Combined to Combined to procedure Sepon minimize minimize minimize minimize false negative | false positive| false negative| false positive errors errors errors errors Less education Short questions: Without reinforce- ment --=-----eeooooo 20 333 A412 «172 «129 With reinforcement-- 25 +217 L243 146 112 Long questions: Without reinforce- ment --=----=-cme_c-o 22 W417 . 332 L147 «107 With reinforcement-- 2% .283 L428 «152 +129 Move edugation Short questions: Without reinforce- ment =====-c-cmeaooo 27 .432 +351 .128 .092 With reinforcement-- 20 «321 .638 .083 L065 Long questions: Without reinforce- ment -===--c=mceoon=o 26 .286 .364 .105 073 With reinforcement-- 24 L454 .509 .109 .080 Includes only respondents who were reinterviewed, cation was not ascertained and conditions present. Summary of F values from analyses of variance! but excludes two respondents for whom edu- 18 persons for whom the physician indicated there were no chronic Source of variation for respond-| Source of variation for respond- ents with less education? ents with more education? Variable Reinforce- Question R x Q Reinforce- Question R x Q ment (R) length (Q) ment (R) length (Q) Underreporting: Combined to minimize. false negative errors---------- 2.34 0.85 0,01 2.29 1.59 0.21 Combined to minimize false positive errors---------- 2,52 3.11 0,14 1.76 3.25 0.11 Overreporting: Combined to minimize false negative errors---------- 0.11 0,08 0.25 1,17 0,01 1.74 Combined to minimize false positive errors---------- 0.01 0.01 0.52 0.46 0.00 1.01 None of the F values is statistically significant (p<.05). Degrees of freedom = 1,84, 3Degrees of freedom = 1,93. 28 Table 13. Number of respondents chronic conditions in original interviews and rei and false positive errors), by reinforcement and total error and net bias index scores for the reporting of nterviews (combined to minimize false negative , question length, and respondent education Total error index score Net bias index score Respondent education Nanier and interview PESDOAd~ Combined to Combined to Combined to Combined to procedure sp 1 minimize minimize minimize minimize ents false negative | false positive | false negative | false positive errors errors errors errors Less education Short questions: Without reinforce- ment----mm-cemmemean 20 . 505 . 542 .162 .283 With reinforcement-- 25 .362 .356 .071 L131 Long questions: Without reinforce- ment---e--ceemoeaao 22 . 564 .639 «270 L425 With reinforcement -- 21 L436 .556 L131 .299 More education Short questions: Without reinforce- ment --=---comeamaa_- 27 . 560 .642 .304 L459 With reinforcement-- 20 .604 +703 L438 «572 Long questions: Without reinforce- ment----s==cemceeu= 26 .392 439 .182 .288 With reinforcement-- 24 .563 .589 . 344 L429 Includes only respondents who were reinterviewed, cation was not ascertained conditions present. Summary of F values from analyses of variancel but excludes two respondents for whom edu- and 18 persons for whom the physician indicated there were no chronic Source of variation for respond- | Source of variation for respond- ents with less education? ents with more educations Variable Reinforce- Question R x Q Reinforce- Question R x Q ment (R) length (Q) ment (R) length (Q) Total error: Combined to minimize false negative errors---------- 2,56 0.80 0.01 1.74 1.59; 0.38 Combined to minimize false positive errors---------- 2,36 3.33 0,35 1.48 3.20 | 0.26 Net bias: Combined to minimize false negative errors---------- 1.61 1,02 0.07 2,78 1.31 0.03 Combined to minimize false positive errors---------- 2,27 3.21 0.02 2,04 2,91 0.02 !None of the F values Degrees of freedom = Degrees of freedom = is statistically significant (p <.05). 1,84. 1,93. 29 bility that a 'no''-saying response bias—the tendency to deny to the interviewer the pres- ence of existing health conditions—was a fairly predominant characteristic in chronic con- dition reporting for all respondents, Procedures which reduced the effects of this tendency to underreport, whether they were experimental interviewing strategies or a special statistical strategy, yielded lower net bias scores. In- spection of these data also revealed that the lowest net bias scores within education groups Table 14. chronic conditions in original interviews, Number of respondents and total error index scores were obtained by the same experimental combi- nation of interview procedures which produced the lowest total error scores, Tables 14 and 15 present, for summary purposes, the main response error and re- sponse bias findings of this research. Here the relative effects of the major analysis vari- ables (reinforcement, question length, respond- ent education, reinterviews, and special statis- tical procedures for minimizing error) can be seen, for the reporting of reinterviews, and in original interview- reinterview data combined to minimize false negative and false positive errors, by reinforcement, question length, and respondent education Total error index score Respondent education mer fad Sasgrulen respond - Combined to Combined to P ents! Original Reinter- minimize minimize interview view false negative | false positive errors errors Less education Short questions: Without reinforce- ment--=---cemcoocnmna- 20 .480 .566 .505 .542 With reinforcement --- 25 «356 .361 +362 .356 Long questions: Without reinforce- ment--=---=c-=ce---- 22 . 584 “B25 . 564 .639 With reinforcement --- 21 470 «5321 L436 .556 More education Short questions: Without reinforce- ment -=--=---comcmmmce- 27 . 585 .615 . 560 L642 With reinforcement--- 20 +703 .603 .604 .703 Long questions: Without reinforce- ment ----==----c-e--- 26 457 +373 .392 L439 With reinforcement --- 24 . 543 .608 +363 .589 Includes only respondents who were reinterviewed, but excludes two respondents for whom education was not ascertained and 18 persons there were no chronic conditions present. for whom the physician indicated NOTE: Lowest scores for each procedure are underscored. 30 Within each education group, the range of total error and net bias index scores was quite wide, indicating that the way in which the interview was conducted and the method of data analysis had particularly large effects on the quality of survey results, The highest total error index score was almost twice as high as the lowest total error index score. Within education groups, the highest net bias index score was about six times greater than the smallest net bias index score, The data also indicate that the experi- mental procedures interacted or combined in different ways to influence reporting quality, The only main effect observed was that com- bining the data from first and second inter- views to minimize false negative reporting (underreporting) always resulted in the lowest average net bias scores. This was true for all interviewing procedures, regardless of respond- ent education, Table 15. Number of respondents and net bias index scores for the reporting of chronic conditions in original interviews, reinterviews, and in original interview-reinterview data combined to minimize false negative and false positive errors, by reinforcement, question length, and respondent education Net bias index score Respondent education Mier and interview respond = Combined to Combined to procedure espe Original |Reinter- minimize minimize interview view false negative | false positive errors errors Less education Short questions: Without reinforce- } ment ===-eemcmee————— 20 .212 .234 .162 283 With reinforcement --- 25 .104 .099 L071 .131 Long questions: Without reinforce- ment cece ccccccn ccc 22 «322 .383 .270 L425 With reinforcement--- 21 .192 . 240 L131 .299 More education Short questions: Without reinforce- ment =eeececccccnane 27 .390 .375 .304 .459 With reinforcement --- 20 .572 L438 .438 .572 Long questions: Without reinforce- ment —=-—-—ecemecmaaa—— 26 .270 .200 182 .288 With reinforcement =--- 24 .364 L410 .344 .429 Includes only respondents who were reinterviewed, but excludes two whom education was not ascertained and 18 persons there were no chronic conditions present. NOTE: Lowest scores in each procedure are underscored. respondents for for whom the physician indicated 31 This method of combining original inter- view and reinterview data also resulted in rel- atively low total error index scores within experimental groups, but only when a question length procedure was used which was found to be least effective for the education group. Thus, the analysis minimizing false negative error yielded lowest total error index scores among less educated respondents asked long questions and among more educated respondents asked short questions, If one wished to minimize the total error index scores for chronic condition reporting, he would interview less educated respondents only once, using reinforcement and short ques- tions. He would interview more educated re- spondents twice, using long questions without reinforcement and disregarding data from the original interview, Were these procedures to be followed, the total error rates for each ed- ucation group would be about the same-—.36 for less educated respondents and .37 for more educated respondents (table 11). It is interesting to note, however, that if less ed- ucated respondents were interviewed using the procedure optimal for more educated respond- ents, and vice versa, maximum total error rates (almost twice as high) would result, One additional conclusion from these optimal procedure data may be that better data would result from redesigning procedures of inter- viewing rather than trying to introduce statis- tical corrections into survey data already col- lected. Other correlates of accuracy and bias of chronic condition veporting.—In addition to the effects of education, reinforcement, question length, and reinterviews, the data from this study may be inspected for other correlates of reporting error and bias, The relationship of reliability and of number of items reported to validity will be discussed in the next section. In this section, the effects of personal charac- teristics of respondents and of various adminis- trative considerations on four of the dependent variables are explored. Of the five personal characteristics shown in table 16, only "health rating" revealed a 32 consistent pattern of effects and then only for the less educated respondents. Respondents rated themselves on a four-point scale describing their general state of health as either excellent, good, fair, or poor. The data indicated that those less educated respondents who rated themselves toward the poor’ end of the scale (essentially those who did not claim excellent health): (1) underreported less, (2) overreported more, (3) had lower total error scores, and (4) lower net bias (''no"'-saying bias) scores, The direction of these effects was similar for the more ed- ucated respondents, but only the effect on net bias was statistically significant. The second part of table 16 shows the correlations of the dependent variables with various administrative variables, One would hope that none of the correlations would be signifi- cant, and almost none were, As mentioned earlier, there was a negative relationship between the length of time since the physician record was made and total error for the more educated group. This indicates that slightly less report- ing error was detected when the physician data were recent. The effect was small and was not confounded with interview procedure effects. The only other linear relationship which met the criterion of statistical significance was the positive relationship between length of inter- view and overreporting in the less educated group. Longer interviews apparently had a slightly greater tendency to contain chronic condition overreports than did shorter interviews if the respondent had not completed high school. It is interesting to note the lack of linear influence of the other variables on error and bias. It apparently made no important difference if the interview was observed, if the interviewer had to make a number of callbacks, or if the interview occurred in a particular week. Re- spondent reporting error was not influenced by the number of conditions the person had accord- ing to the physician, This lack of effect increases the possibility of generalizing the present findings to a more representative cross section of the population in which the incidence of occurrence of the 13 selected chronic conditions is considerably lower than that in this research sample. Table 16, Product-moment correlation coefficients between characteristics of respondents in original interviews and the underreporting ,overreporting,total error, and net bias index scores by respondent education Less education More education Variable Under - Over - Total Net Under - Over- Total | Net reporting | reporting | error bias reporting | reporting | error | bias Personal characteristics: Age------cmmemmmmmm mmm -,05 .07 -,02 -.06 -.03 +10 -.01 -.05 Education-----===cec=== -.01 +01 .01 .00 .00 -.05 -.01 «01 Income====mmmemeceeee=-= .01 .10 .04 -.02 -.05 -.01 -.06 -.05 Number of children----- .04 .07 .06 .01 .00 -.08 .01 .02 Health rating---------- a-,29 2,21 2-,20 | 2-.34 -.16 .18 | -.12| 2-,20 Administrative variables: Interval since Physi- cian Summary Form was filled out=-=======-=- .04 14 .09 -.01 .16 a.,21| 2-,20 -.11 Number of conditions on Physician Summary Form------- mmmmm—————— .05 -.02 .04 .05 13 -.08 «ll 14 Length of interview---- -.12 a,22 -.04 -.18 -.09 add -.06 -.13 Week of interview------ .00 .00 .01 .01 -.05 .10 -.05 -.05 Number of calls to obtain interview------ .10 -.03 .09 «11 + 12 -.04 Wi J: | .12 Observer present at interview---==--==ce-- -.07 -.03 -.08 -.06 .02 .05 .03 «01 ip <,01. Reliability correlation coefficient is reported. The coeffi- Effects of experimental interviewing proce- dures on reliability of interview data,— Since the same questions were asked in the original inter- view and the reinterview, it is possible to examine the effects of reinforcement and ques- tion length on reliability. The following data approach reliability from two different points of view: (1) for chronic conditions (the 19-item list and the 13-item subset which was validated) and symptoms, the calculations are in terms of the consistency of reporting the individual items on the lists; and (2) for the other health variables (doctor con- tacts, dentist visits, hospital episodes, total chronic and acute conditions, and health rating) and secondary variables (height, weight, edu- cation, and income), a Pearson product-moment cient is based on the total number of items re- ported in each interview, regardless of the consistency of reporting each item making up the total, However, the latter approach may mask the existence of compensating errors, For ex- ample, three physician visits might be reported in both interviews even though they might not be the ''same'' three visits in both interviews, Tables 17 and 18 present the average num- ber of responses in the original interview and reinterview match categories in each educa- tion group for the interviewing procedures for reporting of items on the symptom list and on the chronic condition list, In the far right column of the tables is the overall original interview- reinterview agreement rate; the other agree- ment rate columns contain two other coeffi- cients called ''yes' match rate and ''no' match 33 Table 17. Number of respondents, number of reported symptoms by match category, and rates of original interview-reinterview agreement, tion length, and respondent education original interview-reinterview by reinforcement, ques- Original interview- Average number of symptom : : Total BS br ¢ Feintervien Jaresment Respondent Salen hi- education and 3 i interview spond- symp "Yes" | "No" Overall procedure ein tom Ty Match | Match | Match | Match match | wagon | 2gree- 1 tems! in cate- | cate- | cate- | cate- rate rate ment ar gory gory gory gory ’ D rate, © 2 BZ ce D2 = | ATR A+D A+B+C | BHCHD | AipiciD Less education Short questions: Without reinforce- ment=---=-=----=-- 21 19.01 0.05 6.57 1.29 1.24 9.86 L722 .796 «357 With reinforcement- 26 19.02 0.08 85 1,35 1.12 | 10.62 . 703 L811 .870 Long questions: Without reinforce- MeNt==-===cn===u=~- 23 19,00 0.04 6.09 1.65 1.61 9.61 .651 L747 .828 With reinforcement- 25 19.00 0.28 5.56 1.44 1.20 | 10.52 .678 .799 +359 More education Short questions: Without reinforce- ment----==---=-=u- 29 18.99 0.03 5.35 Ys 1.41 | 10,79 .679 . 805 .862 With reinforcement- 23 19.00 0.17 3.7 0. 1.87% 12.353 «5383 .821 .857 Long questions: Without reinforce- ment=---m=--e=coan 28 19.00 0.00 4,79 0.71 1.86 | 11.64 .651 .819 . 865 With reinforcement- 28 19.01 0.61 4,04 0.96 1.36 | 12,04 .635 . 838 .875 Symptom items may not add to 19 due to rounding error. 2 . . . » . . . “Codes for original interview-reinterview match categories are as follows: Respondent Respondent answer in ; Match category original answer in inte oview reinterview A Yes Yes B No Yes C Yes No D No No see p. 33. For further discussion of agreement rates, 34 Table 18. Number of respondents, number of reported chronic conditions by original interview- reinterview match category, and rates of original interview-reinterview agreement, by reinforce- ment, question length, and respondent education Average number of chronic condi=- Syiginal intervie. Total tion items reported AREoRYToN Trecmeny R Number number rates espondent of of Sdncesion and re- chronic ""Yes' | "No" |Overall DLSFYiSw spond - condi- | Migs- | Match | Match | Match | Match | match | match | agree- P ents tion ing cate- | cate-| cate- | cate- | rate, | rate, ment items! data gory gory gory gory D rate, A2 B2 c2 D2 A+B+C | B+C+D | __A+D A+B+C+D Less education Short questions: Without rein- forcement ----- 21 13,01 0.10 2,38 0.48 0.14 9.91 «193 L941 +952 With reinforce- ment-=-eeee-v- 26 13,00 0,04 2,65 0,23 0.19 9.89 .863 .959 .968 Long questions: Without rein- forcement----- 23 13.00 0.09 2.39 0.26 0.52 9.74 .754 .926 . 940 With reinforce- ment---=-- ————— 25 13.00 0.04 2,28 0.16 0.32] 10.20 .826 .955 .963 More education Short questions: Without rein- forcement=----- 29 13.01 0.07 1.76 0.35 0.17 | 10.66 W772 .954 .960 With reinforce- ment -=-====--- 23 13.00 0.13 1.04 0.35 0.00 | 11.48 L748 .970 +973 Long questions: Without rein- forcement----- 28 13.00 0.00 1.86 0.25 0.187 10,71 .812 .961 .967 With reinforce- ment----c=-=-= 28 13,01 0,11 1,75 0.18 0.18 | 10.79 .829 .968 .972 IChronic conditions included are the 13 validated items. Items may not add to 13 due to round- ing error, 2Codes for original interview-reinterview match categories are as follows: Pasponasat Respondent Match category : answer in original i i interview Tetnteiview A Yes Yes B No Yes C Yes No D No No 3For further discussion of agreement rates, see p. 33. 35 rate. The "yes" match rate is the average num- ber of items for which the respondent said "yes" in both interviews divided by the number of items for which the respondent said 'yes' in either interview, The ''mo" match rate was computed in a similar manner for the ''no' answers to items on the symptom and chronic condition lists. The tables indicate only small differences among the interview procedures on the consis- tency of respondent reports. The rates of agree- ment were all very high—higher for chronic condition reporting than for symptom reporting and higher for 'mo'" answers than for ''yes" answers, The data for chronic condition report- ing show that differences among interviewing procedures were not very large, indicating that reinforcement and question length did not have especially strong effects on consistency of ill- ness reporting. For the less educated group of respondents, reinforcement tended to be associated with higher coefficients of agreement. Long questions tended to produce lower rates of consistency than did short questions. Thus, within the less educated group of respondents, the effects of the experi- mental procedures on consistency of reporting followed the same pattern as the effects on the validity of reporting, For more educated respondents the effects of the interviewing procedures on consistency were not identical to the effects on validity, While the differences in consistency among in- terviewing procedures were small, the general pattern for the 'mo'-match rate and the over- all agreement rate was the production of higher levels of consistency between interviewing pro- cedures in responding to specific items when reinforcement was used. Thus, it appears that reinforcement produced slightly more reliable symptom and chronic condition reporting in both education groups. In table 19 product-moment correlation coefficients are presented showing the consis- tency of reporting in the original interview and the reinterview of the number and/or quantity reported of various types of information, The table organizes the reporting into three general types: illness, utilization of health services, and miscellaneous items, All three types of reporting followed similar patterns, First, the coefficients of agreement were 36 almost uniformly high, providing very little opportunity for the interviewing procedure to show any effects, Second, for less educated respondents there was a slight tendency for the coefficients of agreement to be higher for interviews in which short questions and rein- forcement were used than for other kinds of interviews, Thus, within this education group, the interviewing procedures which produced the most valid reporting of chronic conditions also showed a slight tendency to produce more reli- able data when original interview and reinter- view responses were compared, Among more educated respondents interviews in which long questions and no reinforcement were used did show a slight tendency to produce more reli- able reporting for illness items, However, this pattern did not carry over to any great extent when health service reporting and miscellaneous item reporting were examined. For the more educated group of respondents the most reli- able reporting of health service use and mis- cellaneous items was obtained, on the average, by the long question and reinforcement proce- dures, To summarize, several different reliability rates were calculated for the reporting of health- related and nonhealth-related variables. The overall level of agreement between reporting in the original interview and the reinterview was high, leaving little variance upon which the ex- perimental interviewing procedures could act. No large differences between procedures were found, although there was a very slight tendency for reliability to be highest among less educated respondents when short questions and reinforce- ment were used, While these data are far from definitive, it would appear that the relationship between reliability and validity of reporting health information in survey studies is fairly small and that persons or agencies who employ reliability of response as an index of the validity of collected data might benefit from further ex- perimentation concerning this relationship. Physician Summary Form veliability.,—Re- spondent reports of chronic conditions were compared with data provided by physicians on the Physician Summary Form (PSF) in order to calculate the error and bias rates discussed previously. Table 19. Product-moment correlation coefficients of agreement between quantities of va rious health items reported in original interviews and reinterviews, by reporting item, reinforcement, question length, and respondent education Short questions Short questions Long questions Long questicns without reinforce- with reinforce- without reinforce- with reinforce- ment ment ment ment Reporting item Paired Correla- Paired Correla- Paired Correla- Paired Correla- number tion number tion number tion number tion coeffi- coeffi- coeffi- coeffi- cient cient cient cient Less education Illness: Validated chronic conditions-----=--= 21 .88 26 .96 23 . 89 25 «92 All 19 chronic conditions--------- 21 +39 26 +95 23 .90 25 .89 Total chronic and acute conditions--- 21 .85 26 9% 23 +37 25 .76 Symptoms -=-=====------ 20 +85 24 +93 22 .78 22 .83 Health services: Physician visits, 10 weeks=--c=mcmncomnn 18 .56 26 .88 21 .81 25 .28 Dentist visits, 12 monthg-=--=-ouco-n- 21 «78 26 1.00 23 .90 25 1.00 Hospitalizations, 12 monthg--=--cc-=u--- 21 .93 26 .90 23 +99 25 .69 Miscellaneous: Health rating------- 21 .49 25 .89 23 .61 25 e773 Height------emcmuumn 20 1.00 26 .98 22 99 23 .98 Weight-----cmeecamnn 20 1.00 26 .99 23 .99 24 1.00 Education-=----=----- 20 1.00 26 .96 23 99 24 +98 Income--====-----==-= 20 .96 23 .96 22 «95 24 .92 More education Illness: Validated chronic conditions-=-=-=-=--- 29 . 84 23 wl2 28 .89 28 +93 All 19 chronic conditions----=-==-- 29 .88 23 .81 28 .89 28 .88 Total chronic and acute conditions--- 29 «52 23 old 28 «81, 28 .68 Symptoms --=-==-==-== 29 +76 20 wll 28 .88 25 .78 Health services: Physician visits, 10 weeks ==-mmmmmecmmaan 25 +43 19 +91 26 74 23 74 Dentist visits, 12 months=--=eemaeeeaa- 29 .87 23 .94 28 . 94 28 .96 Hospitalizations, 12 months=-----===cco=- 29 1.00 23 .69 28 .94 27 1.00 Miscellaneous: Health rating------- 29 .83 22 .86 28 .81 28 77 Height -=-a==eecemua- 29 .98 23 .99 26 .92 27 +95 Weight -=--cmeeocea—o 29 «99 23 .95 28 .94 28 .99 Education------ ————— 29 .99 22 .96 28 .84 27 .99 Income==--===uee=uo- 28 .86 22 .61 26 .87 26 .98 37 Also, as discussed previously, it was antic- ipated that the physician data could not adequately represent the entire "truth" about respondent chronic condition status as it existed at the time of the interview with the respondent, How- ever, it was assumed that record "error" would be distributed randomly across interviewing pro- cedures and thus would have no systematic effect on the comparisons between procedures. The effect of random error in the records, therefore, would be to attenuate the coefficients of respond- ent-physician agreement so that they would be lower than those expected if the physician records were truly valid. An effort was made to obtain a rough esti- mate of the degrees of attenuation in the match rate coefficients due to record instability, Physi- cians filled out the chronic condition summary form when a patient was seen at a clinic, There were 88 persons for whom at least two physi- cian records of chronic condition status were available. These records were made at two dif- ferent times, but it is not known whether they were filled out by different physicians. Incon- sistencies in the records may represent a re- finement of diagnosis over time, a new appear- ance or a cure of a chronic condition, or a variance introduced in the interpretation of the medical records of the patient, When more than two PSF's were received for a person, com- parisons were made between pairs of forms contiguous in time (namely, first with second, second with third, third with fourth, fourth with fifth), No more than five forms were received for any one patient, By adding these paired forms to the original 88 instances in which at least two forms were available, a total of 97 pairs of forms could be analyzed, The distributions of physi- cian responses in terms of average numbers of conditions in each response category were as follows: Average number of chronic conditions reported by Toral Miss- : response category | J 0 ag condi- Form 1: Yes | No Yes No tions Form 2: Yes | Yes No No 1.21 .78 | 10.70 «31 13.00 38 The overall agreement rate between the two (1.21 + 10.70) (13.00 — .3D percent. The average consistency of checking 1.21 1.21+.78) * or about 61 percent."The corresponding figure for checking "no" was 93 percent, One cannot attach a great deal of meaning to these figures, The 88 persons or 97 dupli- cate cases were not altogether representative of the sample actually used. (The actual sample contained a higher proportion of persons with at least one condition checked ''yes'" on the PSF.) If the figures were representative and if respond- ents were completely truthful, then a lower limit which the mismatch coefficients could reach is expressed by 1.00 minus VPSF reliability, This figure would be .22 for 'yes' answers and .04 for "no" answers. It is interesting to note that the underreporting scores achieved by less ed- ucated respondents interviewed with short ques- tions and reinforcement approached this lower limit quite closely. forms was , Or approximately 94 mn" yes" for a particular condition was Number of Health Events Reported In this section the effects of the experimental interviewing procedures and respondent education on amount of illness and use of health services reported are examined. In addition, the results of this study are compared with those of a previous study on reinforcement, and the possibility of using the number of chronic conditions reported as an index of accuracy of reporting is investi- gated, Effects of interviewing procedures on the amount of health items wveported.—In table 20 the average number of health items reported in original interviews and reinterviews is shown for each of the experimental groups, The F values greater than 2,00, indicating the statistical signifi- cance of the various trends, are also shown in this table. Generally speaking, for original in- hThe Pearson product-moment correlation coefficient for the total number of items checked “yes” on both forms is, 77. Comparing the total number of items checked “yes” is not a sensitive measure to the existence of “compensating errors’ on individual items. Therefore, the Pearson. product-moment coefficient is not preferred as a measure of reliability here. Table 20. Average number of health items reported in original interviews and reinterviews, by reinforcement, question length, and respondent education Less education More education Short questions Long questions Short questions Long questions Health item With- | yieh With- | yith Wigs With With- 1 yien v rein- rer rein- PEt rein- rein rein- rein. force- force- force- - force- force- RENE force- rs force- ent force- nent ment ment ment ment Average number of items reported Original interview: Validated chronic conditiong-=---=-=cco-u-- 2.40 2.80 2.66 2.69 2.08 1.59 2.08 1.98 Total chronic and acute conditions 5.02 6.02 5.57 5.56 4.34 3.97 4.58 4.39 Symptoms =====-=c-ceooonao 7.00 7.54 1+55 7.92 6.45 5.90 6.50 3.33 Physician visits, 10 WeeKS =m==cmmmmcmcmm meme 4,10 2.58 2.48 2.44 2.04 2.68 2.50 2.59 Dentist visits, 12 months =---cecooomecaaaao 0.98 0.86 1,50 2,58 1.74 1.79 3.30 1.80 Hospitalizations, 12 months -=---ceccccccaaoano 0.24 0.26 0.16 0.19 0.09 0.08 0.18 0.24 Reinterview: Validated chronic conditions-=-=vc=eeeaanaan 2.95 3.00 2.64 2,81 2.22 1.55 2.19 2.04 Total chronic and acute conditions-------- 5.90 5.96 5.68 5.24 4.56 3.60 4,27 4.33 Symptoms =====-==eeoaaaoo- 8.16 7.22 7:52 7.39 6.89 4,06 5.77 4.95 Physician visits, 10 wWeekS===-mmcccm maaan 2.00 2.16 6.90 1.90 2.30 2.06 2.30 3.10 Dentist visits, 12 months =---ccmcaooanaannao 1,20 1.24 1.23 3.05 1.89 2.05 3.58 2.13 Hospitalizations, 12 months ==--ceccaacaacaaaa 0.36 0.32 0.23 0.00 0.07 0.00 0.19 0.35 Summary of F values greater than 2.00 from analyses of variance Source of variation Variable and Rein- Ques- health item force- tion R x Q ment length (R) Q Less education Reinterview: Hospitalizations, 12 months=-=--- 3.48 More education Original interview: Validated chronic conditions--- 2.24 Symptoms===-========cccoococoano b3,96 Dentist visits, 12 months------ 2.69 2.61 Hospitalizations, 12 months---- 3.17 Reinterview: Validated chronic conditions--- 2.72 Symptomg-=-========mmmmm——————— b9.57 2.97 Dentist visits, 12 months------ 2.32 bp <.05. 39 terviews the effects of interviewing procedures on numbers of items reported were unstable and did not show many consistent patterns, Further analysis, not shown here, has suggested that the effects of the interviewing procedures differ for some types of reporting depending upon whether the respondent was scheduled for a reinterview. As described earlier, the selection of respondents for reinterviews was not random, Original in- terviews with respondents scheduled for rein- terviews were generally taken in the first few weeks of the 6-to-8-week interviewing period, and some clinics from which records were sampled were represented less frequently in the reinterview group. Rather than offer tentative conclusions about the effects of procedures on amount of health reporting here, it is suggested that further research is needed, employing designs which do not contain the confounding mentioned above. Fortunately, the procedural effects on the amount of reporting of the 13 validated chronic conditions were not mediated by the reinterview selection to any meaningful extent. Comparison of veinforcement effects on amounts reported with previous research.—In a previous study by Marquis and Cannell” a rein- forcement procedure of interviewing was found to produce about 25 percent more symptoms and chronic and acute conditions reported than were obtained without reinforcement, The previ- ous research confounded reinforcement and ques- tion length because respondents receiving rein- forcement were also asked questions that were longer (contained extra words) than were questions asked of nonreinforced respondents. In addition, the sample consisted of a cross section of adult white women living within the city limits of Detroit which differs somewhat from the sample of women in the present research, A reanalysis of the data from the previous research, controlling for respondent education (shown in the first section of table 21), shows that respondent education had little effect on amount of reporting when reinforcement was not used. When reinforcement was used, less ed- ucated respondents reported more health items, in general, than did more educated respondents. Data from the present study are correspond- ingly rearranged and are shown by question length in the lower section of table 21. A similar ed- ucation-reinforcement interaction can be seen 40 when short questions were used, but only for symptom reporting when long questions were used. The lack of a reinforcement-education interaction for the long question procedure in the present study probably reflects differences between the two studies in the actual length of questions. In the present study, 'long' ques- tions were considerably longer and occurred more frequently than in the previous study. By examining these two sets of data, it can also be seen that in the first study, the rein- forcement effects on amount of reporting were generally confined to the less educated group. Moreover, reinforcement did not decrease amounts reported by the more educated respond- ents to any important extent, whereas in the present study, a definite reinforcement suppres- sing effect on illness reporting was often ob- tained for the more educated group. Relationship between amount veporvied and accuracy of veporting,.—The amount of health items reported is of interest because it has been used as an indicator of accuracy of re- porting when validity data were not available (for example, in the study by Cannell, Fowler, and Marquis? and in the pilot studies for the present research), The correspondence between the number of selected chronic conditions reported and the ac- curacy of reporting them is shown for the various experimental groups in table 22, When the experimental group was the unit of analysis, the rank order correlation between amount and accuracy of reporting was reasonably high (Spear- man's rho = .64; p< .05). When the long ques- tion without reinforcement procedures for both education groups were removed, the rank order correlation was 1,00, This suggested that, for this kind of chronic condition reporting, dif- ferences among groups in amounts reported (especially those not asked long questions) tended to reflect, albeit imperfectly, group differences in validity. On the other hand, it would probably not be safe to conclude that the increased level of reporting produced by long questions reflected increases in accuracy of reporting, Why increases in number of conditions reported brought about by reinforcement reflected increases inaccuracy, while increases in reporting produced by long questions did not, is a question which deserves further research, Table 21. Average number of conditions and symptoms reported, with percent difference by education in the present study, by reinforcement, question length, and respondent education, compared with data by respondent status derived from an earlier study Without reinforcement With reinforcement Variable and : Less More Percent Less More Percent health item educa- educa- differ- | educa- educa- differ- tion tion ence! tion tion ence! ha Average number of items reported Chronic conditions: For selfe=eeeeaa- 1.46 1.09 34 1.59 1.44 10 By proxy--------- 1.01 0.91 11 1.84 0.81 127 Total chronic and acute conditions: For self-=-eeea-- 2.14 2.19 -2 2.81 2.68 5 By pProxy=-—==-=—---- 1.46 1.39 5 2.41 1.36 76 Symptoms ========aa= 5.23 5.07 3 7.42 5.51 35 Data from present study Short questions: Validated chronic conditions ===--- 2.40 2.08 15 2.80 1.59 76 Total chronic and acute condi- tiong====-=meuu- 5.02 4.34 16 6.02 3.97 52 Symptoms =======-- 7.00 6.45 9 7.54 5.90 28 Long questions: Validated chronic conditiong====-- 2.66 2.08 28 2.69 1.98 36 Total chronic and acute condi- tiong-==-=meuea- 5.57 4,58 22 5.56 4.39 27 Symptoms ======-=- 7.55 6.50 16 7.92 5.33 49 Computed as: L = Xu M Where X, = mean number reported by less educated respondents and X, = mean number reported by more educated respondents, 41 Table 22. Average number of chronic con- ditions reported and average accuracy score, by reinforcement, question length, and respondent education Average Average ; a accuracy Interview score chronic a procedure cond El {i 002 tota reported error) Less education With reinforce- ment: Short questions- 2.80 .66 Long questions-- 2.69 +33 Without reinforce- ment: Long questions-- 2.66 44 Short questions- 2.40 47 More education Without reinforce- ment: Short questions- 2.08 47 Long questions-- 2,08 +B. With reinforce- ment: Long questions-- 1.98 46 Short questions- 1.59 .36 Listed by rank order of column 1, average number of chronic conditions re- ported. NOTE: Spearman's rho = .64; p< .05. 42 Finally, data in table 23 show the relation- ships of number of chronic conditions reported (out of 13 validated items) and the four indexes of reporting error and bias when the individual respondent is used as a unit of analysis. A marked tradeoff between overreporting and underreporting can be seen, Correlations of amount reported with underreporting were large and negative, Correlations of amount reported with overreporting were also large but in a positive direction. The net effect of this pattern on the total error score was quite small, in- dicating that decreases in underreporting errors were potentially canceled by increases in over- reporting errors, It should be pointed out that the total error index gave a disproportionate weight to overreporting errors, so that for practical purposes the reduction in total error can be assumed to be somewhat greater than that indicated by the correlation coefficients. For the entire sample, the correlation be- tween the number of chronic conditions (of the 13 validated items) reported and total error is -.24 (p<.0l1). This indicates that the previously assumed positive relationship of quantity and ac- curacy of health reporting is confirmed but that only a small amount of the accuracy variance is explained by reporting quantity (r?=.0576). Thus, for future studies which are designed to explore for variables influencing accuracy of reporting (that is, studies that are not planned for compari- sons among groups defined on an a priori basis) and which use quantity of reporting as anindex of accuracy, large sample sizes are needed. The relative costs of increasing sample sizes and collecting actual validity data should be consid- ered in the study plan. Table 23. Product-moment correlation coefficients between number of chronic conditions reported and index scores of underreporting, overreporting, total error, and net bias, by reinforcement, question length, and respondent education £ Correlation coefficient Respondent education Rumer 9 Ral Later les conditions Under- Over- Total Net P reported reporting reporting error bias Less education Short questions: o 2 a Without reinforcement------ 41 -.48 .72 -.16 a=+69 With reinforcement-----=---- 50 8..46 4,83 -.15 -.68 Long questions: a Without reinforcement------ 44 b-.40 2.76 -.15 a=+63 With reinforcement----===-- 36 -.37 77 -.15 -.57 More education Short questions: a 4 2 Without reinforcement------ 53 b= 37 a,82 -.39 a=+/0 With reinforcement--=-=---=- 38 -.39 RR) -.30 -.50 Long questions: b Without reinforcement------ 40 hE °.82 -.13 2-53 With reinforcement--=--=---- 46 -.59 .63 fo bb -.72 Combined education groups Long and short questions: Without and with reinforce- a 2 a ment-=---ememm cece memo - 348 -.47 8,74 -.24 -.64 &p <.,01. bp <,05. 43 DISCUSSION Education Differences in Ability and Bias Earlier research in the NCHS-SRC series demonstrated that there is variation in the ac- curacy with which health information is reported by respondents, The initial studies suggested that some of the response error could be reduced by asking only for health information that was easy to recall accurately (for example, relatively recent and of high impact). Since changing the nature of the events asked about is not always feasible, a search for other correlates of response error in health reporting was undertaken, This search indicated that the most immediate influences on response bias were to be found in the conduct of the interview itself rather than in the attitudes or demographic characteristics of respondents. The results of the present research confirmed the earlier conclusion that response error can be affected greatly by even minor changes in the conduct of the interview, Making only two changes at the major points of leverage in the interview-—the wording of the questions and the way in which the interviewer reacts toanswers-— had significant effects on information accuracy and response bias. A major surprise, however, was to find that these proximal influences were mediated by respondent education, These highly reliable effects suggest again that efforts to improve the data collected by personal inter- views should take into account the fact that the interview is an instance of the two-person social interaction, potentially governed by the vari- ables important in such interactions, Any changes in the behavior of one participant, the inter- viewer or the respondent, will have potentially complex effects on the behavior of the other. As the following discussion infers, the thinking in regard to variables affecting response ac- curacy in health interviews has come almost full circle, Earlier research failed to confirm the hy- pothesized importance of the main effects of re- spondent cognitive and demographic (for ex- ample, education) variables on reporting accu- racy, Current results and interpretations in- dicate that these variables are indeed important, but only as mediators of the effects of different 44 interviewing practices, rather than as prime causal variables in the "main effect" sense, An attempt is made in the material presented below to construct a social-psychological frame of reference which will provide a setting for the current results and which may be used as a basis for future research, In the discussion certain assumptions derived from signal de- tection theory are made—changes in total error scores reflect changes in ability to recall and changes in net bias scores reflect changes in reporting bias tendencies, Ability to wvecall.—It is generally assumed that a significant correlation exists between education level and general ability, Itis assumed, too, that more able people tend also to be more educated, A corresponding hypothesis is that less educated respondents are less able to recall their chronic conditions than are more educated persons, The data from the current study, however, do not support this hypothesis, When conventional interviewing procedures (short questions without reinforcement) were used, the total error scores of the two education groups were ap- proximately the same, In addition, the minimum total error index scores of each group were similar, although achieved with different inter- viewing conditions. Total error index score Interview procedure Less More educa- | educa- tion tion Conventional interviewing pro- cedure (short questions with- out reinforcement): Original interview---------- «33 «33 Reinterview-=-=--=---ccceaoaoo «37 62 "Best" procedure (minimum total error index score): Original interview, short questions with reinforce- ment-==----=-cem meme +36 Reinterview, long questions without reinforcement------ .37 If reinforcement improved the ability of less educated respondents to recall, reinter- view total error index scores would be lower than those obtained in an initial interview, The data show opposite trends for the less educated group. Thus, assuming total error scores to be an index of ability to recall, the different ex- perimental procedure effects on reporting could not be attributed to different ability levels of the education groups. Furthermore, the dif- ferential reinforcement and question length effects could not be attributed to their effect on dif- ferent ability levels, Reporting bias,— The current data do suggest that the two education groups differed somewhat in their tendency to deny to the interviewer the presence of existing health conditions—the ''no''- saying response bias, Under conventional pro- cedures (short questions without reinforcement) the net bias scores were lower for the less educated group than for the more educated group. Moreover, ignoring the statistical cor- rection analysis (attempts to minimize over- or underreporting statistically), the lowest average net bias index score obtained by the less educated group was about half of the lowest average net bias index score achieved by the high school graduate group: Net bias index score Interview procedure Less More educa- | educa- tion tion Conventional interviewing pro- cedure (short questions with- out reinforcement): Original interview----=--=---- «24 .35 Reinterview---====mmemmenmm- «23 .38 "Best' procedure (minimum net bias index score): Original interview, short questions with reinforce- MEI = wim om om om mm mm .10 Reinterview, long questions without reinforcement------ 220 One hypothesis relevant to understanding the bias problem is that the education groups differ in the extent to which confidence or certainty is required before sickness or health service use is reported. In other words, the threshold of certainty for responding ''yes' may differ be- tween groups, Thus, when a more educated re- spondent is unsure of an answer, he may be inclined to search his memory for further con- firmation of his tentatively recalled facts and respond ''yes'' only when he is fairly sure this is the correct answer. Procedures, therefore, which stimulate and allow extra time for this confirmation memory search (such as long ques- tions and reinterviews, both of which present the question or search stimulus a second time) may increase the accuracy of reporting of persons who are 'conservative' in the sense of not ad- mitting sickness, or other facts, when there is some uncertainty about the answer, Reinforcement, coming as it did immedi- ately after an answer, may ''cut off" additional memory search or answer-confirmation activity and thereby maintain the existing conservatism bias of higher educated respondents, Future re- search might be directed toward testing the hy- pothesis that the education groups differ in the degree of certainty required before reporting health information and toward exploring the idea that increased memory-search or answer-con- firmation time will aid accuracy in reporting of more educated respondents, If the conservatism bias hypothesis is confirmed, other interviewing procedures might be developed to reduce it directly, There is some independent evidence that higher educated respondents might be more reluctant to report sickness to an interviewer, Phillips and Clancy!? found that higher status respondents (status defined in terms of ed- ucation and income) rated more items on a 22- item mental illness symptoms list as "less desirable to have" than did lower status re- spondents, If the social undesirability of re- porting symptoms of physical illness showed the same interaction effect with education as did the reporting of mental illness symptoms, then the conservatism or high level of '""no'-saying bias observed for the higher education group in this research may be a product of a special sensitivity to the undesirability of mentioning illnesses in public. 45 However, if the observed bias of the higher educated group is a social desirability bias, it is not at all clear how it might be reduced. Clinical interviewers, suchas psycholgists, might claim that this type of bias is reduced by creating more rapport: a warm, supportive, friendly atmosphere of communication. As will be dis- cussed below, however, one hypothesis suggested by the data from this study is that there is too much rapport in interviews with higher educated respondents and that it is this rapport which leads to increases in the ''no''-saying bias tend- encies. Social Status, Rapport, and Task Orientation Hyman and associates’! have pointed to the distinction between task involvement and total involvement in characterizing the relationship between the interviewer and respondent during an interview. They suggested that valid data may be more a function of task involvement, Too much rapport or involvement in the nontask aspects of the relationship may detract from accuracy of reporting when a "hen party at- mosphere is maintained. Back and Gergen'? made a somewhat similar distinction between the "information game'' and the "ingratiation game," suggesting that the in- terviewer's problem is to maximize information giving and minimize ingratiation, These writers theorized that as social distance between the in- terviewer and respondent decreases (that is, they are more alike in demographic and other status characteristics), the tendency increases for the respondent to give ingratiating rather than ac- curate, task-oriented answers. One series of studies (Dohrenwend, Williams, and Weiss?) focusing on the possible inter- action effects of rapport and social distance on accuracy and bias in the personal interview dif- fered on a number of dimensions and often produced different conclusions, These studies did seem to suggest, however, that response bias is a complex function of the difference (or similarity) of status between the interviewer and respond- ent and the amount of rapport created in the interview. For example, Weiss' data! indicated that response inaccuracy is inversely related to 46 social status similarity, and that when status differences are small, rapport decreases ac- curacy of reporting. To clarify the relevance of the above dis- cussion, several things about the variables in the current study should be mentioned. The in- terviewers were all "high status' in terms of education, All had completed high school and many had some college training. Hence, when we talk about variation in respondent education we are also talking about variation in the simi- larity of interviewer-respondent educational status: social distance was relatively large when the respondent had not graduated from high school and relatively small when the respondent had graduated. Second, the reinforcement proce- dure might be regarded as one which was some- what high in rapport, since it was generally friendly and accepting of reports of illness, If the above reinterpretations of education and reinforcement are made, the results of this study tend to agree with those of the other researchers in that, when social differences are small (high educated respondent and high ed- ucated interviewer here), an increase in rap- port (here, using reinforcement) also increases bias. Fowler!” made an extensive reanalysis of data from a study of health reporting by Cannell, Fowler, and Marquis and has presented per- haps the most elaborated theory to date con- cerning the relationship between rapport and respondent education, and between cognitive ap- proaches and respondent education, Fowler theorized that because highly ed- ucated and less educated respondents differ on a number of dimensions, the personal inter- view should be conducted to emphasize different things for the two education groups. Less ed- ucated respondents are seen as having less skill or ability to report accurately, although Fowler's analysis suggested that interviewers can and do compensate for skill differences invarious ways, such as probing or clarifying inadequate answers, The reporting performance of more highly ed- ucated respondents is largely under their own cognitive control, Reporting accuracy, therefore, is mainly a function of how well the highly ed- ucated respondent understands what he is sup- posed to do. The reporting performance of less educated respondents is not internally controlled but much more dependent upon guidance and help from the interviewer, Explaining to the less educated respondent about the reporting requirements of the research (for example, complete accuracy and coverage) is hypothesized to have no effect on reporting accuracy since this performance is not strongly governed by the degree of ''understanding' achieved, Finally, the nature of the social interaction is hypoth- esized to differ according to the level of re- spondent education. Because survey research interviewers tend to be high school graduates or even college graduates, a respondent who has not completed high school is at a clear disadvantage vis-a-vis the interviewer in the interview social interaction, Fowler hypothesized that this status discrepancy presents a problem to the lower status respondent and that his re- porting performance is a function of how this problem is resolved. In some unpublished pilot studies carried out at the Survey Research Center, Wood!® was able to experimentally control three independent vari- ables (ability to perform a recall task, reinforce- ment, and cognitive control over performance) and to measure the effects of combinations of these variables on recall accuracy. Her results (which were not always statistically significant due to the small sample sizes used in the pilot research) were somewhat parallel to those ob- tained in the present study. For the low recall ability group, both reinforcement and under- standing of the task requirements had a benefi- cial effect on recall accuracy, But for those with high recall ability, reinforcement and task understanding worked against each other. Re- spondents who understood the recall task and had the ability to perform it reasonably well actually did worse when the interviewer reinforced correct answers than when the interviewer did not use reinforcement, Most of the studies cited above produced the common finding that the status relationship between the interviewer and respondent has an effect on reporting accuracy and bias. Some writers go further to state that social status interacts withthe performance of the interviewer to influence the data, The implication made by most of these writers is that the natural relationship between a lower status respondent and a higher status interviewer is not a personal one, Back and Gergen!? infer that this is good because the information gain can be maximized. Fowler!® Weiss,!* and others also see a potential benefit in a large social distance because, under these conditions, if the interviewer is of higher status than the respondent, her behavior has a great influence on the respondent's performance. Fowler 15 and deKoning!7 concluded that probing and question clarification are necessary and beneficial. Fowler went even further to suggest that abstract explanations will have no effect.The present study showed that the redundant long questions and reinterviews (all task-oriented experimental phenomena) had negative effects on the less educated respondents, All of these findings suggest to the present writers that interview situations where there is a wide social discrepancy call for task-oriented interview behavior which (1) is contingent on respondent behavior and (2) has a friendly, supportive component, The reinforcement procedure Seems to meet both criteria, It should be noted, however, that reinforcement is not the only way to achieve these two goals, As Fowler theorized, more traditional interviewing styles employing effective probing strategies and occasional, noncontingent interpersonal behavior (for example, a pleasant personal comment, a joke, and some laughter) should achieve the same effects, The question now becomes whether effective probing (essen- tially negative reinforcement because probes are used when the respondent fails to perform his role properly) plus some positive but irrelevant comments can serve the same function as a schedule of positive reinforcement for the less educated respondent group. Caution is advised against accepting the idea that noncontingent supportiveness really 47 has any effect on accuracy of reporting. In a study by Marquis et al.'® which was designed specifically to test the effects of a supportive versus challenging atmosphere on reporting ac- curacy the data clearly indicated that respondents enjoyed the supportive atmosphere more and thought they did a better job of reporting. How- ever, accuracy of recall was not even slightly in- fluenced by the contrast in interrogation atmos- pheres, The researchers and theorists referred to above point out that a different problem exists when the status of the interviewer and respond- ent are similar. The natural relationship in this case is not a task-oriented one, but an inter- personal, friendship one, If the interviewer does anything to support this perception, the resulting atmosphere will encourage minimum attention to the reporting taskas well asanswersappropriate to a hen party (socially acceptable, ingratiating answers). The problem with the equal status relationship appears to be one of too much rapport (total involvement). Apparently, in this situation, an interviewer must be attentive to establishing a task-oriented attitude for the respondent, Fowler suggests that this can be done cognitively by the interviewer presenting a clear explanation of the repondent role. The current results show that long, redundant ques- tions and reinterviews also seem to achieve the desired effects. Neither of these procedures is contingent on the behavior of the respondent, The above findings imply that in situations where the interviewer and respondent are of the same status (in this case high education women), a reinforcement procedure will accentuate a natural tendency to establish an interpersonal rather than task-oriented relationship. The more friendly the interview situation becomes, the worse the data become. A final and alternative hypothesis (based on the work of Wood mentioned above) concerns the negative effects of reinforcement in the high status relationship. The problem becomes one of not creating too much rapport with rein- forcement but requiring the interviewer toengage in behavior which is inappropriate or unwanted in an equal status situation, Reinforcement may be perceived as unnecessary and awkward after 48 every report of illness by the highly educated respondent. Such a respondent undoubtedly feels she understands what the interviewer wants and that she is perfectly capable of performing the apparently easy reporting task without the in- terviewer's constant reminders that performance is adequate.' Possibly it is the confident, highly educated respondent who might benefit from negative reinforcement for inadequate perform- ance, The major benefits for this respondent would seem to be derived from changing her perception that minimal effort is needed to report accurately and that the reporting task is only a minor part of her interactive re- lationship with the interviewer. Summary of Discussion It is hypothesized that the effects of rein- forcement, question length, and reinterviews differ among education groups for the following reasons: Less educated respondents rely on inter- viewer cuestodirect their reporting performance. Thus, appropriate use of reinforcement, probes, and other feedback by the interviewer can aid recall and reporting accuracy a great deal, Because of the social status discrepancy, feed- back is perceived as appropriate and actually welcomed. Abstract explanations are probably irrelevant to performance for this group, and long questions appear only to serve as a source of confusion. Performance is about as good as it can be in an appropriately conducted initial interview; there appears to be no additional benefit from a reinterview, More educated respondents carry out the reporting task largely according to their own understanding of it; they do not rely on cues from the interviewer, Reinforcement under these circumstances may encourage an interpersonal (versus task) orientation; may be perceived as It should be noted that respondents’ perceptions of reporting requirements can be wrong. In one study? 45 percent of the respondents had incorrect ideas about accuracy of re- porting requirements; in another study3 41 to 69 percent of respondents misunderstood the requirements to varying ex- tents. inappropriate, unnecessary, or even conde- scending; and may, at least for open-ended questions, ''cut off'' the additional search and confirmation time needed to avoid underreporting. The more educated respondent appears to have a stronger tendency to underreport chronic con- ditions, possibly due to a stronger conservatism or social desirability bias. This type of respond- ent may feel the need to rethink answers and be very sure of their accuracy before admitting the existence of a chronic condition, Long questions and reinterviews apparently provide the additional cues necessary to search memory to confirm ambiguous answers and, thereby, provide the confidence level required to report a chronic condition, These highly speculative hypotheses are offered as a guide to further research on sur- vey interview reporting accuracy. Hopefully, they highlight some of the important areas to consider when survey data validity is of concern, These "human'' characteristics involving memory, re- call, cognition, social status, and interpersonal interaction are often overlooked in methodo- logical studies, yet appear to be very crucial in any meaningful understanding of the personal interview process, REFERENCES 1National Center for Health Statistics: Interview data on chronic conditions compared with information derived from medical records. Vital and Health Statistics. PHS Pub. No. 1000-Series 2-No. 23. Public Health Service. Washington. U.S. Government Printing Office, May 1967. 2National Center for Health Statistics: The influence of interviewer and respondent psychological and behavioral variables on the reporting in household interviews. Vital and Health Statistics. PHS Pub. No. 1000-Series 2-No. 26. Public Health Service. Washington. U.S. Government Printing Office, Mar. 1968. 3National Center for Health Statistics: Effect of some experimental interviewing techniques on reporting in the health interview survey. Vital and Health Statistics. PHS Pub. No. 1000-Series 2-No. 41. Public Health Service. Washington. U.S. Government Printing Office, May 1971. 4Matarazzo, J. D., Saslow, G., and Wiens, A. N.: Studies in interview speech behavior, in L. Krasner and L. P. Ullman, eds., Research in Behavior Modification. New York. Holt, Rinehart and Winston, 1965. SFerber, R.: Collecting Financial Data by Consumer Panel Techniques. Urbana. Bureau. of Economic and Business Research, University of Illinois, 1959. 6Neter, J., and Waksberg, J.: Response Errors in Collection of Expenditures Done by Household Interviews: An Experimental Study. Washington. U.S. Department of Commerce, Bureau of Census Technical Paper No. 11, 1965. 7National Center for Health Statistics: Chronic conditions and activity limitation, United States, July 1961-June 1963. Vital and Health Statistics. PHS Pub. No. 1000-Series 10-No. 17. Public Health Service. Washington. U.S. Government Printing Office, May 1965. 8U.S. National Health Survey: Health interview responses compared with medical records. Health Statistics. PHS Pub. No. 584-D5. Public Health Service. Washington. U.S. Government Printing Office, June 1961. 9Green, D. M., and Swets, J. A.: Signal Detection Theory and Psychophysics. New York. John Wiley and Sons, 1966. LOphillips, D. L., and Clancy, K. J.: Response biases in field studies of mental illness. Am. Sociol. Rev. 35 (3): 503-515, June 1970. 11Hyman, H. H., Cobb, W. J., Feldman, J. J., Hart, C. W., and Stember, C. H.: Interviewing in Social Research. Chicago. University of Chicago Press, 1954. 12Back, K., and Gergen, K.: Idea orientation and ingratiation in the interview: a dynamic model of response bias. Proceedings of the Social Statistics Section of the American Statistical Association. Technical Report No. 13, 1963, pp. 284-288. 13Dohrenwend, B. S., Williams, J. A., and Weiss, C. H.: Interviewer biasing effects: toward a reconciliation of findings. Pub. Op. Quart. XXXIII (1): 121-129, 1969. 14Weiss, C. H.: Validity of welfare mothers’ interview responses. Pub. Op. Quart. XXXII (4): 622-633, 1968. 15Fowler, F. J., Jr.: Education, Interaction, and Interview Performance. Doctoral dissertation, The Michigan, 1965. 16Wood, L.: The role of task difficulty in operant verbal conditioning. Personal communication. University of 17deKoning, T. L.: Interviewer and Respondent Interaction in the Household Interview. Doctoral dissertation, The University of Michigan, 1969. 18Marquis, K. H., Marshall, J., and Oskamp, S.: Accuracy and completeness of testimony as a function of kind of question, interrogation atmosphere, and item content. J. Appl. Soc. Psychol. 1972, in press. 000 4y Symptoms 50 APPENDIX | SAMPLE PAGES FROM QUESTIONNAIRES INTERVIEW PROCEDURE A—Short Questions Without Reinforcement Now I'm going to ask you some questions about your health. Have you EVER had------- ? (DO NOT ENTER THIS INFORMATION ON TABLES) Yes P45970B 1-A No Bad headaches? Coughed up blood? Had fainting or blackout spells? Bad sore throats? Shortness of breath? Serious backaches? Ever felt your heart beating hard or acting funny? Pain in or around your heart or chest? Gas in your stomach? Bad stomach cramps? Loose bowels? Pain or soreness in the female organs? Pain or burning when you go to the bathroom? Ever had painful or swollen joints? Any broken bones? Itching skin? Mental illness? Ever had trouble sleeping? Ever had any venereal disease? Chronic Conditions (Procedure A) 45970B 6-A 6. Please tell me if you have had any of these conditions during the past 12 months?: Yes No a. Asthma? (IF YES) What kind of asthma is it? b. Hay fever? c. Thyroid trouble or goiter during the past 12 months? (IF YES) What kind of thyroid trouble is it? d. Repeated bronchitis? e. Repeated skin trouble? (IF YES) What kind of skin trouble is it? f. Paralysis of any kind? g. Hemorrhoids or piles during the past 12 months? h. Hernia or rupture? i. Repeated gall bladder or liver trouble? (IF YES) What kind of trouble is it? j. Peptic or stomach ulcer? k. Varicose veins? 1. Have you had repeated trouble with your back or spine? (IF YES) What kind of back trouble is it? 52 7. 45970B 7-A Please tell me if you have ever had any of these conditions: Yes No m. Arthritis or rheumatism? n. Rheumatic fever? o. Heart disease or any heart trouble? “eR Em wm em mm ae mm RENEE SEEN RE b= “ (IF YES) What kind was it? p. Ever had a stroke? Sl ESE EEE RE ERE EERE Se es ws L ~~ - (IF YES) How does it affect you now? q. Hypertension or high blood pressure? r. Hardening of the arteries? 8. Ever had diabetes? Physician Visits (Procedure A) 10. 11. During the past 14 days, did you talk to a doctor or go to a doctor's office or clinic for yourself? [7] Yes 9b. 9c. 1 the last 14 days. (NOTE IN CALENDAR MARGIN: { No (SKIP TO Q10) According to what you have circled, you saw a doctor Is that correct? Please take this pencil and circle the date of each visit on the calendar. time(s) during MORE THAN ONE VISIT ON A SINGLE DAY ) WEEK OF VISIT IF EXACT DAY UNKNOWN ) During the three months outlined in blue on the calendar, did you talk to a doctor or go to a doctor's office or clinic for yourself? [1 Yes ™ No (SKIP TO Ql1) 10b. 10c. (NOTE IN CALENDAR MARGIN: Please circle the date of each visit on the calendar. According to what you have circled, you saw a doctor the three months outlined in blue. Is that correct? time(s) during MORE TIAN ONE VISIT ON A SINGLE DAY ) WEEK OR MONTH OF VISIT IF EXACT DAY UNKNOWN) (Not counting the visits you have already mentioned) During the times outlined in red and blue, did you: Yes See a doctor in an emergency room? No When was this? ® At your home? Talk to a doctor over the phone? CALLS ONLY TO MAKE APPOINTMENTS) (DO NOT COUNT *INTERVIEWER: CIRCLE ANY ADDITIONAL DATES OF CONTACTS MENTIONED 53 INTERVIEW PROCEDURE B—Short Questions With Reinforcement Symptoms R STATEMENTS - USE SHORT AND LONG ——————P45970 2-B Thank you We're interested in that, Mm- hmm That's the kind of information we need. I see (REPEAT ANSWER JUST GIVEN) Yes That's important. 0.K. We need to know that. 1. Now I'm going to ask you some questions about your health. Have you EVER had==--- =? (DO NOT ENTER THIS INFORMATION ON TABLES) Yes No a. Bad headaches? R b. Coughed up blood? R c. Had fainting or blackout spells? R d. Bad sore throats? R e. Shortness of breath? R f. Serious backaches? R g. Ever felt your heart beating hard or acting funny? R h. Pain in or around your heart or chest? R i. Gas in your stomach? R j. Bad stomach cramps? R k. Loose bowels? R 1. Pain or soreness in the female organs? R m. Pain or burning when you go to the bathroom? R n. Ever had painful or swollen joints? R o. Any broken bones? R p. Itching skin? R q. Mental illness? R r. Ever had any trouble sleeping? R s. Ever had any venereal disease? R 54 Chronic Conditions (Procedure B) R STATEMENTS - USE SHORT AND LONG 45970B Thank you We're interested in that, 6-B Mm- hmm That's the kind of information we need. I see (REPEAT ANSWER JUST GIVEN) Yes That's important, 0.K, We need to know that. 6. Please tell me if you have had any of these conditions during the past 12 months?: Yes No a. Asthma? (IF YES) What kind of asthma is it? b. Hay fever? c. Thyroid trouble or goiter during the past 12 months? (IF YES) What kind of thyroid trouble is it? d. Repeated bronchitis? Ec] J 2 2 OE EC e. Repeated skin trouble? (IF YES) What kind of skin trouble is it? f. Paralysis of any kind? g. Hemorrhoids or piles during the past 12 months? h. Hernia or rupture? Ec LI EE I EE i. Repeated gall bladder or liver trouble? 1 1 1 1 1 1 ' 1 i 1 1 1 1 1 1 1 1 1 1 1 ' 1 1 1 1 1 1 1 1 1 1 i 1 1 1 1 | (IF YES) What kind of bladder trouble is it? j. Peptic or stomach ulcer? k. Varicose veins? FF |= |-F 1. Have you had repeated trouble with your back or spine? 1 1 1 1 1 1 1 ! 1 1 1 1 1 1 1 1 1 1 1 1 ' 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 (IF YES) What kind of back trouble is it? = 56 — R STATEMENTS - USE SHORT AND LONG Thank you We're interested in that. Mm-hmm That's the kind of information we need. I see (REPEAT ANSWER JUST GIVEN) Yes That's important. 0.K. We need to know that. 7. Please tell me if you have ever had any of these conditions 4 59708 1-5 Yes m. Arthritis or rheumatism? = n. Rheumatic fever? = o. Heart disease or any heart trouble? (IF YES) What kind was it? = p. Ever had a stroke? (IF YES) How does it affect you now? = [7] q. Hypertension or high blood pressure? r. Hardening of the arteries? s. Ever had diabetes? Physician Visits (Procedure B) R STATEMENTS - USE LONG ONLY 45970B Thank you We're interested in that. 9-3 Mm- hmm That's the kind of information we need. I see (REPEAT ANSWER JUST GIVEN) Yes That's important. 0.K. We need to know that. 9. During the past 14 days, did you talk to a doctor or go to a doctor's office or clinic for yourself? [J Yes [J No (SKIP TO Ql0) 9b, Please take this pencil and circle the date of each visit on the calendar. [R] 9c. According to what you have circled, you saw a doctor time(s) during the last 14 days. Is that correct? (NOTE IN CALENDAR MARGIN : MORE THAN ONE VISIT ON A SINGLE DAY) WEEK OF VISIT IF EXACT DAY UNKNOWN ) 10. During the three months outlined in blue on the calendar, did you talk to a doctor or go to a doctor's office or clinic for yourself? [] Yes [] No (SKIP TO Q11) 10b. Please circle the date of each visit on the calendar. [R] 10c. According to what you have circled, you saw a doctor time(s) during the three months outlined in blue. Is that correct? (NOTE IN CALENDAR MARGIN : MORE THAN ONE VISIT ON A SINGLE DAY ) WEEK OR MONTH OF VISIT IF EXACT DAY UNKNOWN) 11. (Not counting the visits you have already mentioned) During the times outlined in red and blue, did you: Yes No When was this? * i ? See a doctor in an emergency room? R [|] At your home? [Rr] [R] Talk to a doctor over the phone? (DO NOT [|] ® COUNT CALLS ONLY TO MAKE APPOINTMENTS) *INTERVIEWER: CIRCLE ANY ADDITIONAL DATES OF CONTACTS MENTIONED, 57 INTERVIEW PROCEDURE C—Long Questions Without Reinforcement Symptoms P45970 2-C 1. Now I'm going to ask you some questions about your health. By asking these questions, The Public Health Service can get a good picture of the nation's health. Have you EVER had--==-- ? (DO NOT ENTER THIS INFORMATION ON TABLES) Ves No a. Bad headaches? b. Coughing up blood is the next one. Have you coughed up blood? c. How about fainting spells or blackout spells. Have you had these? d. Now a question about bad sore throats. We're looking for information about these. Have you had bad sore throats? ed e. What about shortness of breath? f. The next item is serious backaches? Have you had serious backaches? g. Ever felt your heart beating hard or acting funny? — h. What about pain in or around your heart or chest? Have you had that kind of pain? : . . . 1 i. Gas in the stomach is the next item. This is another health problem we're interested in. Have you had gas in your stomach? j. We'd like to know about bad stomach cramps. Have you had them? 1 k. The next question is about loose bowels? Have you had loose bowels? | 1. Have you had pain or soreness in the female organs? m. What about pain or burning when you go to the bathroom? n. How about painful or swollen joints. These are other items we would like to know about. Have you ever had painful or swollen joints? 0. Broken bones is the next item. Have you had any broken bones? oe p. Now a question about itching skin. Have you had that? — q. What about mental illness. This is another kind of condition we need information about. Have you had any mental illness? r. Have you ever had trouble sleeping? 8. Venereal discase is the last item of this list. We'd like to get 58 an estimate of this condition in the population. Have you ever had any venereal disease? Chronic Conditions (Procedure C) P45970B 6. Next, I'm going to read a list of some health conditions of special interest to the Health Service. during the past twelve months. Please tell me if you have had any of these conditions Yes No Asthma is the first one we need information about. Have you had asthma? (IF YES) What kind of asthma is it? What about hay fever? Have you had that? Another area of interest to the Health Service is thyroid and goiter trouble. Have you had any thyroid trouble or goiter during the past 12 months? (IF YES) What kind of thyroid trcuble is it? The next item is repeated bronchitis. Have you had that? We are also interested in trouble with your skin. Have you had repeated skin trouble? (IF YES) What kind of skin trouble is it? Have you had paralysis of any kind? Hemorrhoids and piles is the next item we need information about. Have you had hemorrhoids or piles during the past 12 months? We are looking for information about hernias and ruptures. Have you had a hernia or rupture? Have you had any repeated gall bladder or liver trouble? (IF YES) What kind of trouble is it? We are also interested in how many people had peptic or stomach ulcers. Have you had a peptic or stomach ulcer? What about varicose veins? Have you had varicose veins? The last item on this list asks about repeated back trouble. Have you had repeated trouble with your back or spine? (IF YES) What kind of back trouble is it? _ 59 60 1. P45970 B 8-C Here are some other conditions we want to ask you about. This time we want to know if you have ever had any of them. Please tell me if you have ever had any of these conditions. [Yes No m. The first item is arthritis or rheumatism. We're interested in how many people have this condition. Have you had arthritis or rheumatism? n. The next one is rheumatic fever. Have you had that? 0. Heart disease or heart trouble is the next item. This is another area of special interest to the Health Service. Have you had heart disease or any heart trouble? (IF YES) What kind was it? p. The next item is stroke. Have you ever, at any time, had a stroke? (IF YES) How does it affect you now? q. We also need information on hypertension or high blood pressure. Have you had hypertension or high blood pressure? r. What about hardening of the arteries. Have you had that? s. The last item is diabetes. Have you ever had diabetes? Physician Visits (Procedure C) 10. 11, P45970 B 10-C The next questions are about doctors, hospitals, and other places people use in connection with their health. The health service is very interested in how people use the services of doctors and clinics. During the past 14 days, did you talk to a doctor or go to a doctor's office or clinic for yourself? [J Yes [J] No (SKIP TO Q10) 9b. Please take this pencil and circle the date of each visit on the calendar. 9c. According to what you have circled, you saw a doctor time(s) during the last 14 days. Is that correct? (NOTE IN CALENDAR MARGIN: MORE THAN ONE VISIT ON A SINGLE DAY) WEEK OF VISIT IF EXACT DAY UNKNOWN ) Now we are interested in the last three months. We ask people about the times they have talked to a doctor during this period. During the three months out- lined in blue on the calendar, did you talk to a doctor or go to a doctor's office or clinic for yourself? [J Yes [J No (SKIP TO Ql1) 10b. Please circle the date of each visit on the calendar. 10c. According to what you have circled, you saw a doctor time(s) during the three months outlined in blue. Is that correct?-: (NOTE IN CALENDAR MARGIN: MORE THAN ONE VISIT ON A SINGLE DAY) WEEK OR MONTH OF VISIT IF EXACT DAY UNKNOWN) In addition to seeing a doctor in his office, there are other places wlifere people see doctors about their health. During the times outlined in red and blue (but not counting what you've already mentioned) did you: * es No When was this? See a doctor in an emergency room? How about a telephone call. Did you talk with a doctor about your health over the telephone? (DO NOT COUNT CALLS ONLY TO MAKE APPOINTMENTS) We are also interested in the times the doctor came to your home in connection with your health. Did you talk to a doctor at your home during the times outlined on the calendar? %* INTERVIEWER: CIRCLE ANY ADDITIONAL DATES OF CONTACTS MENTIONED 61 Symptoms 62 INTERVIEW PROCEDURE D—Long Questions With Reinforcement R STATEMENTS - USE SHORT AND LONG Thank you We're interested in that. Mm-hmm That's the kind of information we need. I see (REPEAT ANSWER JUST GIVEN) Yes That's important gk, We need to know that. Now I'm going to ask you some questions about your health. By asking these questions, The Public Health Service can get a good picture of the nation's health. Have you EVER had--=-- ? (DO NOT ENTER THIS INFORMATION ON TABLES) res P45970 a. Bad headaches? b. Coughing up blood is the next one. Have you coughed up blood? =] c. How about fainting spells or blackout spells. Have you had these? = d. Now a question about bad sore throats. We're looking for information about these. Have you had bad sore throats? e. What about shortness of breath? R! f. The next item is serious backaches? Have you had serious backaches? g. Ever felt your heart beating hard or acting funny? (=| [= h. What about pain in or around your heart or chest? Have you had that kind of pain? (=| i. Gas in the stomach is the next item. This is another health problem we're interested in. Have you had gas in your stomach? [=] j. We'd like tc know about bad stomach cramps. Have you had them? El k. The next question is about loose bowels? Have you had loose bowels? 1. Have you had pain or soreness in the female organs? m. What about pain or burning when you go to the bathroom? Fl= | = n. How about painful or swollen joints. These are other items we wculd like to know about. Have you ever had painful or swollen joints? o. Broken bones is the next item. Have you had any broken bones? Pp. Now a question about itching skin. Have you had that? FH, =H] = q. What about mental illness. This is another kind of condition we need information about. Have you had any mental illness? (= r. Have you ever had trouble sleeping? = s. Venereal disease is the last item of this list. We'd like to get an estimate of this condition in the population. Have you ever had any venereal disease? [= 2-D lL : Chronic Conditions (Procedure D) R STATEMENTS - USE SHORT AND LONG Thank you We're interested in that. Mm-hmm That's the kind of information we need. I see (REPEAT ANSWER JUST GIVEN) Yes That's important. 0.K. We need to know that. P45970B 7-D 6. Next, I'm going to read a list of some health conditions of special interest to the Health Service. Please tell me if you have had any of these conditions during the past twelve months. No a. Asthma is the first one we need information about. Have you had asthma? (IF YES) What kind of asthma is it? Ee b. What about hay fever? Have you had that? c. Another area of interest to the Health Service is thyroid and goiter trouble. Have you had any thyroid trouble or goiter during the past 12 months? (IF YES) What kind of thyroid trouble is it? (= |=] d. The next item is repeated bronchitis. Have you had that? e. We are also interested in trouble with your skin. Have you had repeated skin trouble? (IF YES) What kind of skin trouble is it? = | =] =, f. Have you had paralysis of any kind? g. Hemorrhoids and piles is the next item we need information about. Have you had hemorrhoids or piles during the past 12 months? =| =| =, h. We are looking for information about hernias and ruptures. Have you had a hernia or rupture? i. Have you had any repeated gall bladder or liver trouble? (IF YES) What kind of trouble is it? j. We are also interested in how many people had peptic or stomach ulcers. Have you had a peptic or stomach ulcer? k. What about varicose veins? Have you had varicose veins? 1. The last item on this list asks about repeated back trouble. Have you had repeated trouble with your back or spine? (IF YES) What kind of back trouble is it? [=], FF = =F] = 64 7. R STATEMENTS - USE SHORT AND LONG Thank you We're interested in that. Vie Yom That's the kind of information we need. I see (REPEAT ANSWER JUST GIVEN) Yes That's important. 0.K. We need to linow that. P45970 B 8-D Here are some other conditions we want to ask you about. This time we want to know if you have ever had any of them. Please tell me if you have ever had any of these conditions. es No m. The first item is arthritis or rheumatism. We're interested in how many people have this condition. Have you had arthritis or rheumatism? n. The next one is rheumatic fever. Have you had that? o. Heart disease or heart trouble is the next item. This is another area of special interest to the Health Service. Have you had heart disease or any heart trouble? (IF YES) What kind was it? p. The next item is stroke. Have you ever, at any time, had a stroke? (IF YES) How does it affect you now? q. We also need information on hypertension or high blood pressure. Have you had hypertension or high blood pressure? r. What about hardening of the arteries. Have you had that? s. The last item is diabetes. Have you ever had diabetes? Physician Visits (Procedure D) R STATEMENTS - USE LONG ONLY ———————————— Bo] B Thank you We're interested in that. Mm-hmm That's the kind of information we need. I see (REPEAT ANSWER JUST GIVEN) Yes That's important. O.K, e need to know that. 9. The next questions are about doctors, hospitals, and other places people use in connection with their health. The health service is very interested in how people use the services of doctors and clinics. During the past 14 days, did you talk to a doctor or go to a doctor's office or clinic for yourself? [J Yes [J] No (SKIP TO Q10) 9b. Please take this pencil and circle the date of each visit on the calendar. [R] 9c. According to what you have circled, you saw a doctor time(s) during the last 14 days. Is that correct? (NOTE IN CALENDAR MARGIN: MORE THAN ONE VISIT ON A SINGLE DAY) WEEK OF VISIT IF EXACT DAY UNKNOWN ) 10. Now we are interested in the last three months. We ask people about the times they have talked to a doctor during this period. During the three months out- lined in blue on the calendar, did you talk to a doctor or go to a doctor's office or clinic for yourself? [] Yes [[] No (SKIP TO QLl) 10b. Please circle the date of each visit on the calendar. R] 10c. According to what you have circled, you saw a doctor time(s) during the three months outlined in blue. Is that correct? (NOTE IN CALENDAR MARGIN: MORE THAN ONE VISIT ON A SINGLE DAY) WEEK OR MONTH OF VISIT IF EXACT DAY UNKNOWN) 11. In addition to seeing a doctor in his office, there are other places where people see doctors about their health. During the times outlined in red and blue (but not counting what you've already mentioned) did you: * Yes No When was this? See a doctor in an emergency room? How about a telephone call. Did you talk with a doctor about your health over the telephone? (DO NOT COUNT CALLS ONLY TO MAKE APPOINTMENTS) We are also interested in the times the doctor ® ® came to your home in connection with your health. Did you talk to a doctor at your home during the times outlined on the calendar? * INTERVIEWER: CIRCLE ANY ADDITIONAL DATES OF CONTACTS MENTIONED —o0O0O0——— 65 APPENDIX II FORMS USED IN STUDY INTERVIEWER’S FORM P45970B RESPONDENT'S NAME : ADDRESS : PHONE NO: SERIAL NUMBER FORM LETTER OF OF INITIAL INTERVIEW: INITIAL INTERVIEW: (A,B,C, or D) DATE OF INITIAL INTERVIEW: DATE AND TIME OF REINTERVIEW APPOINTMENT: REMARKS : OBSERVATION FORM P. 45970B Observed Interviewer Name: Interview Number: Observer Initials: Interview Form: A[{B|C|D Specific Observation Question or item origin - in Forms A-C: Gives erroneous feedback - in Forms B-D: Misses reinf. statement Gives inadequate reinf. - in all Forms: Should have probed more Uses inadequate probe General Observation Introduction at door Asking Questions Looks up in middle of question Eligible Conditions Doesn't look up at end of question Rushes in reading questions Inconsistent pace for short and long Using Reinforcing Statements Doesn't lock up for R. stat. Doesn't pause after R. stat. ____ Uses tco many single R. stat. __ Lack of variety in selecting R. stat. ___ Lack of naturalness in using R. stat. Lack of R. after Q. 14 Other Comments 67 68 PHYSICIAN SUMMARY FORM Survey Research Center The University of Michigan Project 45970B March 1968 i | | a I Address: Date: Definitely or Definitely or Don't know, probably probably no Condition present not present information Chronic skin trouble O J O 0 OJ Cl Hernia, rupture a 0 Any heart disease » Hypertension a0 Arthritis, rheumatism 10 [3 Hay fever J Ml tH) Hemorrhoids J ] 7 Peptic ulcer 7 Diabetes mm m Varicose veins Ml ] I Asthma a Chronic bronchitis 1 C1 Stroke 0 [ NOTE: The terminology used is that of the questionnaire as asked of respondents. Please fill out for patients who are: Female White Age 17-60, inclusive ORIGINAL INTERVIEW ILLNESS TABLE Interviewer Initials Serial Number 45970B Page _ of __ NAME OF CONDITION OR SYMPTOM: Q FIRST MENTIONED # CHECK ALL APPLICABLE CHECK Q. CHECK Q. CHECK Q. SOURCE BEGIN Q'S| SOURCE BEGIN Q'S |SOURCE BEGIN Q'S [10.2-7, 000 T80 (]Q.2-7.....T4 Q.2-7.....Th QiBisnsnraTh Q.8.......T1 QBassniasTd Other.....Tl Other,....Tl [lother.....T1 Tl. Did you have...(was your /YBS/ ASK T4 /YES/ ASK T4 /YES/ ASK T4 ...present) during the rey . prs pute Us done /NO/ ASK T2 /NO/ ASK T2 /NO/ ASK T2 12, Did you have it during /YES/ ASK T3 /YES/ ASK T3 /YES/ ASK T3 - ? ee sev— ——— the past 12 months? /NO/STOP_ QUESTIONS NO/STOP QUESTIONS| /NO/STOP QUESTIONS T3. For how long did ... last? (GET DURATION OF | CONDITION, NOT SINGLE : EPISODES) [[]3 months or more Less than 3 months [Jon cClist-ASK T&4 [Not on CC list STOP_QUESTIONS Less than 3 months [Jon CClist-ASK T4 [INot on CC list STOP QUESTIONS []3 months or more ASK T&4 Less than 3 months [Jon cClist-ASK T4 [Not on CC list STOP QUESTIONS ] has it caused you? | JAlmost none fis [JAlmost none T4, Did you ever talk to JYES/ ASK T4b /YES/ ASK T4b /YES ASK T4b ? rs pli 3 ‘doctor MONE «1s? [M0] ASK TS /NO] ASK TS [NO] ASK TS rm eh NNR EEE hE) em ww om ow ww dog mw afm mmo wow om T4b, What did the doctor say it was--did he = rr Ty N NO give it a medical name? Ed Jef L220 TS During the pany 12 [YES] ASK Sb JYES] ASK TSb [YES] ASK ’TSb months did you have to Cai A —— cut down on your usual /NO/ ASK T6 /NO/ ASK T6 NO/ ASK T6 activities because of 4.43 T5b, How many days? days days days T6. During the past 14 [Jvery much [Jvery much Very much days, at its worst how Much Much Much ) = much pain or discomfort | ] Some [ ] some [ ]Some [J] Almost none 69 REINTERVIEW ILLNESS TABLE Interviewer Initials Serial Number_ 459708 Page____ _ of ____ NAME OF CONDITION OR SYMPTOM: # Q FIRST MENTIONED ! SOURCE OF ILLNESS SOURCE BEGIN Q's} SOURCE BEGIN Q's! SOURCE BEGIN Q's (CHECK ALL APPLICABLE) | 5. 4-7.8-9-10.76 |[]Q 2-4-7-8-9-10.T4 |[]Q 2-4-7-8-9-10..T4 Q Winiimsniome T1 Qll.v.vun.... 11 Q 1l...... IN, | Other.cs ws miivs TO || |Other.......... TO ORE ws vv ninin vin un TO LEAVE BLANK [R2] [28] [B] R2/ [7B] [B] [R2] [28] [B] [57 [TTT [OFF [57 [TCT [BET [87 [CTT [oe] TO. Did you first notice . during the [Z weeks / ASK T&4 | ££ weeks/ ASK T&4 [2 weeks] ASK T&4 past 2 weeks or bef ASK T1 bef ASK T1 bef ASK T1 before that tine? | L2Efore’ [before] [before] Tl. Did you have . — ——— ms HEE FOAL » = » [YES/ ASK T4 YES / ASK T4 YES / ASK T&4 cresent) during the | /55 ASK T2 | /NO ASK T2 ASK T2 past 4 weeks? = | 7 7 T2. Did you have it /XES / ASK T3 | /YES/ ASK T3 | /YES/ ASK T3 during the past 12 months? /NO/ STOP QUESTIONS /NO/ STOP QUESTIONS /NO/ STOP QUESTIONS T3. For how long did . last? (GET DURATION OF CONDI - TION, NOT SINGLE EPISODES) {]3 months or more ASK T4 Less than 3 months On CC list ASK T4 Not on CC list... STOP QUESTIONS |[[]3 months or more | Less than 3 months i On CC list ASK T4 Not on CC list... STOP QUESTIONS []3 months or more ASK T4 Less than 3 months On CC list ASK Té4 Not on CC list.. STOP QUESTIONS T4. Did you ever talk — ER Ira to & dostor dbout [YES / ASK. T4b | /YES/ ASK T4b | /YES/ ASK Té4b ol [NO/ ASK T5 xo NO/ ASK T5 | /NO/ ASK T5 Bo Th Ly fh ga Ted sak saa a TL LT rs wl say it was -- did a , - — he give it a ! [No/ No/ [No/ medical name? T5. During the past 12 months did you have ! /YES/ ASK T5b | /YES/ ASK T5b | /YES/ ASK T5b to cut down on your usual activities /NO/ ASK T6 | /NO/ ASK T6 | /NO/ ASK T6 oso RR08 OE 0.8, tio fr ses te mn ome sons nen a 0 2 0 100 000 05 500 05.00 he 505 05 ER Se ERE 8 5 ‘5b. How many days? days days days y y y Te. During the past 4 Very much Very much _]Very much weeks, at its worst Much ™ Much Much how much pain or Some . | Some Some disconfort fas ft Almost none =) Simoes none Almost none caused you? 70 oo O # U.S. GOVERNMENT PRINTING OFFICE : 1972 482-005/13 Series 1. Series 2. Series 3. Series 4. Series 10. Series 11. Series 12. Series 13. Series 14. Series 20. Series 21. Series 22. VITAL AND HEALTH STATISTICS PUBLICATION SERIES Formerly Public Health Service Publication No. 1 000 Programs and collection procedures.— Reports which describe the general programs of the National Center for Health Statistics and its offices and divisions, data collection methods used, definitions, and other material necessary for understanding the data. Data evaluation and methods reseavch.— Studies of new statistical methodology including: experi- mental tests of new survey methods, studies of vital statistics collection methods, new analytical techniques, objective evaluations of reliability of collected data, contributions to statistical theory. Analytical studies.—Reports presenting analytical or interpretive studies basedon vital and health statistics, carrying the analysis further than the expository types of reports in the other series. Documents and committee reports.—Final reports of major committees concerned with vital and health statistics, and documents such as recommended model vital registration laws and revised birth and death certificates. Data from the Health Interview Survev.—Statistics on illness, accidental injuries, disability, use of hospital, medical, dental, and other services, and other health-related topics, based on data collected in a continuing national household interview survey. Data from the Health Examination Survey.—Data from direct examination, testing, and measure- ment of national samples of the civilian, noninstitutional population provide the basis for two types of reports: (1) estimates of the medically defined prevalence of specific diseases in the United States and the distributions of the population with respect to physical, physiological, and psycho- logical characteristics; and (2) analysis of relationships among the various measurements without reference to an explicit finite universe of persons, Data from the Institutional Population Surveys — Statistics relating tothe health characteristics of persons in institutions, and their medical, nursing, and personal care received, based on national samples of establishments providing these services and samples of the residents or patients. Data from the Hospital Discharge Survey.—Statistics relating to discharged patients in short-stay hospitals, based on a sample of patient records in a national sample of hospitals. Data on health resources: manpower and facilities.—Statistics on the numbers, geographic distri- bution, and characteristics of health resources including physicians, dentists, nurses, other health occupations, hospitals, nursing homes, and outpatient facilities. Data on mortality.—Various statistics on mortality other than as included in regular annual or monthly reports—special analyses by cause of death, age, and other demographic variables, also geographic and time series analyses. Data on natality, marriage, and divorce.,—Various statistics on natality, marriage, and divorce other than as included in regular annual or monthly reports—special analyses by demographic variables, also geographic and time series analyses, studies of fertility. Data from the National Natality and Mortality Surveys.— Statistics on characteristics of births and deaths not available from the vital records, based on sample surveys stemming from these records, including such topics as mortality by socioeconomic class, hospital experience in the last year of life, medical care during pregnancy, health insurance coverage, etc. For a list of titles of reports published in these series, write to: Office of Information National Center for Health Statistics Public Health Service, HSMHA Rockville, Md, 20852 Vital Signs aL Cg, ST, EF, ° NCHS 2 ES & < (4 LASTS CNH Tg) U.S. DEPARTMENT OF HEALTH, EDUCATION, AND WELFARE Public Health Service Health Services and Mental Health Administration Vital and Health Statistics-Series 2-No. 46 For sale by the Superintendent of Documents, U.S. Government Printing Office, Washington, D.C. 20402 - Price 30 cents Series 2 Data Evaluation and Methods Research Number 46 Vital Signs Present at Birth Report of a study of vital signs present at birth as observed in the delivery rooms of five hospitals, and a study of the relation- ship of these signs of life to definitions of live birth and fetal death used for vital registration purposes; comparison of rates based on the study data according to alternate definitions which include various combinations of vital signs present at birth. DHEW Publication No. (HSM) 72-1043 U.S. DEPARTMENT OF HEALTH, EDUCATION, AND WELFARE Public Health Service Health Services and Mental Health Administration National Center for Health Statistics Rockville, Md. February 1972 NATIONAL CENTER FOR HEALTH STATISTICS THEODORE D. WOOLSEY, Director PHILIP S. LAWRENCE, Sc.D., Associate Director OSWALD K. SAGEN, Ph.D, Assistant Director for Health Statistics Development WALT R. SIMMONS, M.A., Assistant Director for Research and Scientific Development JAMES E. KELLY, D.D.S., Dental Advisor EDWARD E. MINTY, Executive Officer ALICE HAYWOOD, Information Officer OFFICE OF HEALTH STATISTICS ANALYSIS IWAO M. MORIYAMA, Ph.D., Director DEAN E. KRUEGER, M.S., Deputy Director Vital and Health Statistics-Series 2-No. 46 DHEW Publication No. (HSM) 72-1043 Library of Congress Catalog Card Number 76-610273 CONTENTS Introduction Purposes of Study Study Procedures Results of Study . . . . Definition of Live Birth Effect on Vital Statistics Rates Registration . Discussion Conclusion . References Appendix I. Recording Form Appendix II. Computation of Rates Page nN NO DW 13 16 17 18 19 SYMBOLS Data not available amma serves Category not applicable-------rremreeeeee QANLIEY ZT rnnrermnnn nn asomosmsmemmmmmssesonsss sass Quantity more than 0 but less than 0.05----- 0.0 Figure does not meet standards of reliability or precision -——---eeeemeeeeeeees VITAL SIGNS PRESENT AT BIRTH Helen C. Chase, Dr. P.H., Louis Weiner, E.E., and Joseph Garfinkel, M.P.H.2 INTRODUCTION In demographic and epidemiologic studies of infant mortality, vital statistics rates such as fetal death rates or infant or neonatal mortality rates are often compared. The basic data needed for compiling and computing such rates are the numbers of infant, neonatal, and fetal deaths and the number of live births. The basic data are derived from vital records which, in the United States, are kept on permanent file in the States in accordance with State statutes. In some recent international studies of perina- tal and infant mortality, considerable attention has been paid to the comparability of data and definitions.!-3 It was recognized that the defini- tions of “live birth” and “fetal death” were matters of basic importance to the statistics. At time of birth, the attendant must decide, on the basis of certain evidence, if the infant is live born. His decision is reflected in a live birth or a fetal death certificate, and he determines the category to which the event is allocated statis- tically. Fortunately, the attendant has no diffi- culty in arriving at a decision for the great majority of deliveries: the offspring is unques- tionably born alive and survives. At times, apr, Chase is Staff Associate (Biostatistics) at the Health Services Study, Institute of Medicine, National Academy of Sciences. At the time this study was conducted, she was Statistician (Health) at the National Center for Health Statistics. Mr. Weiner, presently retired, was Director, Bureau of Records and Statistics, City of New York Department of Health. Mr. Garfinkel was Research Analyst in the same Bureau and is presently Senior Research Scientist, Birth Defects Institute, New York State Department of Health. however, the attendant must refer to the defi- nitions of vital events to arrive at a decision regarding the type of vital record which must be filed. Official definitions of live birth and fetal death were approved and recommended for use in all countries by the Third World Health Assembly in 1950 and were recommended for use in all States of the United States by the Surgeon General of the Public Health Service. Following this action, the international recom- mendations were incorporated into the laws and regulations of almost all of the States in this country and are regarded as the official defini- tions to be followed for vital registration. As a natural consequence, these particular definitions affect the vital statistics of individual States and the country as a whole. The recommended definition of live birth is: “Live birth is the complete expulsion or extraction from its mother of a product of conception, irrespective of the duration of pregnancy, which, after such separation, breathes or shows any other evidence of life such as beating of the heart, pulsation of the umbilical cord, or definite movement of voluntary muscles, whether or not the umbilical cord has been cut or the placenta is attached; each product of such a birth is considered live born.” 4 The definition of fetal death is the complement of the definition of live birth: “Foetal death is death prior to the complete expulsion or extraction from its mother of a product of conception, irrespective of the duration of pregnancy; the death is indicated by the fact that after such separation, the foetus does not breathe or show any other evidence of life such as beating of the heart, pulsation of the umbilical cord, or definite movement of voluntary muscles.”# At the time a pregnancy terminates, each con- ceptus should be assigned to one of the two groups based on the vital signs present at time of birth: it is either a live birth (liveborn infant) or a fetal death (all other pregnancy terminations). The terms “live birth” and “fetal death” relate to each conceptus so that a single pregnancy may result in one live birth and one fetal death when twins are involved. It is important to note that the definition of live birth stipulates that if any one of four vital signs is observed at birth, the infant shall be registered as a liveborn infant. In practice, the same event could be inter- preted somewhat differently, or the definition could be applied slightly differently by different physicians. When one considers that over 3.5 million live births and over 50,000 fetal deaths with gestation periods of 20 weeks or more are registered annually throughout the United States, occasional differences of opinion are bound to occur as to whether certain births should be registered as live births or as fetal deaths. Although these occasional disparities exist, they assume practical importance in a statistical sense insofar as they occur often enough to affect the vital statistics rates. It is important, therefore, to estimate how often such events occur and to estimate their potential effect on fetal death rates or on neonatal, perinatal, or infant mortality rates. In 1966 the National Center for Health Statistics (NCHS) contracted with the Medical and Health Research Association of New York City, Inc., to investigate the problems associated with the definitions of “live birth” and “fetal death,” and the resulting rates. That city was selected as the site for the study because of its longstanding reputation as having one of the most complete registration systems in the coun- try insofar as live births and fetal deaths are concerned. Two exploratory efforts were undertaken. First, 2,300 fetal death records for pregnancies with gestation periods of 28 completed weeks or more were reviewed to determine whether there were Indications anywhere on the records that the infants were born alive. Only one such record was found, and for that birth, a fetal death, live birth, and death record had all been filed. In addition, two fetal death records were found for each of nine deliveries. Thus among these 2,300 fetal death records, 10 superfluous fetal death records were found, but only one appeared to be related to some question about whether the infant was born alive. A second exploratory effort was undertaken to determine whether the desired information could be obtained retrospectively from existing hospital records. Small groups of patients’ records in three hospitals were reviewed in detail. These included records for 238 live births and 172 fetal deaths with gestation periods of 28 weeks or more. Specifics regarding signs of life were not found in sufficient detail in the hospital records. It was concluded from these two reviews that neither the existing vital records nor the existing hospital records contained the desired informa- tion in a satisfactorily complete and consistent fashion. Following the two exploratory investi- gations, it was decided to pursue the problem in a prospective study by direct observation of a group of deliveries, including the careful record- ing of vital signs observed at birth. PURPOSES OF STUDY The present investigation was undertaken with a number of objectives in mind. One objective was statistical, i.e., to determine the quantitative effect of using different combina- tions of vital signs to define live births. To accomplish this purpose, it was necessary to obtain information on the four specified vital signs for a group of births. With careful observa- tion, one could document the vital signs for a consecutive series of births and use the observed results to compute rates for analytical purposes. Another objective was methodological, i.e., to demonstrate the feasibility of conducting such studies amid the daily routine of hospitals. The vast majority of live births and fetal deaths with gestation periods of 28 weeks or more occur in hospitals. For international comparisons of infant and perinatal mortality, this group of pregnancy terminations are of particular rele- vance because only fetal deaths with gestation periods of 28 weeks or more are required to be registered in other countries. STUDY PROCEDURES The study method consisted of the direct observation of a number of deliveries to record the signs of life observed within 2 minutes after birth. To avoid selection of cases by the ob- servers, personnel were provided on a round-the- clock basis to witness consecutive series of deliveries. Five hospitals in New York City participated in the study: three municipal and two voluntary hospitals. The first delivery was observed in May 1967 and the last in July of the same year. Observers were instructed to witness every delivery in the delivery rooms and every cesarean section in the surgical suites. Since the observers were required to be present in the delivery rooms at the time of delivery, it was not possible to use individuals unrelated to the hospital staffs. Instead, a member of the obstetric team was assigned to act as observer, and that member was often a resident, intern, or medical student. In one of the hospitals, the observer was the accoucheur. The observations were recorded independ- ently of the regular record-keeping activities of the usual delivery room staff. Although objec- tive and completely independent observations were desired, it was recognized that because of the selection of observers as described above, this would not be entirely possible. Further- more, it was not always possible for the observer to be close enough to the infant at the moment of delivery to detect the signs of life firsthand, and some unmeasured amount of indirect infor- mation was included. By verbal report, such instances were rare. The four signs of life which were to be specifically noted were those included in the official definition of live birth, i.e., breathing, heartbeat, pulsation of the umbilical cord, and definite movement of voluntary muscles. The observers determined the respiratory effort and movement of voluntary muscles by visual means. The presence or absence of heartbeat was observed by palpation at the apex of the heart, and if no heartbeat was detected, an attempt was made to detect the heartbeat by ausculta- tion. Pulsation of the umbilical cord was deter- mined by palpation. All of the signs were required to be observed within 2 minutes after the fetus was separated from the body of its mother, and pulsation of the umbilical cord was recorded only up to the time immediately after the cord had been tied if that occurred before 2 minutes had elapsed. A vital sign was considered absent if it failed to manifest itself within 2 minutes following delivery. A copy of the recording form is shown in Appendix I. The completed study forms were forwarded to the City of New York Department of Health at weekly or biweekly intervals. In addition to the form which was completed by the observer, a coding sheet was used at the Department of Health to abstract information from the corresponding official vital records. The Health Code of the City of New York requires the registration of all products of conception irrespective of gestation; therefore, each of the events observed in the delivery rooms was required to have a vital record on file. Comparison of the study forms with the vital records for the study period was included to determine how well the definitions of the World Health Organization (WHO) were applied in the everyday working situation. Following the receipt of the study forms in the Department of Health, a search was made for the corresponding birth or fetal death certificates. In addition, reference was made to the weekly lists of deaths among children under 1 year of age which are prepared by the Department. From the lists for the five study hospitals, all deaths which occurred in the first 7 days of life among the study infants born in the same hospitals were abstracted, and the informa- tion was entered on the coding sheet. In this cross-check, the assumption was made that deaths which occur during the first week of life occur in the same hospital in which the birth occurs. This assumption was felt to be reason- able by the project staff. Later, vital signs were coded using a maxi- mum 16-unit code which covered all possible combinations of the four vital signs. In some tables in this report, vital signs are denoted by a single letter representing each of the signs: respiratory effort (R), pulsation of the umbilical cord (P), movement of voluntary muscles (M), and heartbeat (H). All data on the presence of vital signs refer only to the 2 minutes after separation of the fetus from its mother or for pulsation of the umbilical cord the shorter interval described earlier. For some infants, signs which were not observed during this 2-minute interval appeared subsequently and will be com- mented on later in this report. The ethnic groups are classified for the purpose of this study as follows: White. —Includes 1,015 infants; excludes those with mothers born in Latin America. Negro. —Includes 946 infants; excludes those with mothers born in Latin America. Other. —Includes 661 infants: 535 to mothers born in Puerto Rico, 106 to mothers born in other Latin American coun- tries, and 20 others. This classification does not correspond exactly with the categories which are generally used for demographic or health statistics, nor with the ethnic code generally used in the Department of Health. RESULTS OF STUDY A total of 2,629 events which came within the scope of the study occurred in the study period at the five selected hospitals; seven of these were excluded. One case was excluded because a vital record was not filed and six because a study form was not filed on time, or was incomplete. Because of the desire to com- pare information from the two sources of information, these seven incomplete cases were rejected. Table 1. Number and percentage distribution of deliveries by ethnic group, and WHO classification based on vital signs: five selected hospitals in New York City, May-July, 1967 Ethnic group All Live births | Fetal deaths deliveries Number Total . . . .. 2,622 2 565 57 White! 1,015 1,007 8 Negro’ wis ww Buy wd 946 9156 31 Other? . . ...... 661 643 18 Percentage distribution Total . . . .. 100.0 297.8 22 White! LL 100.0 99.2 08 Negro! . . . .. . .. 100.0 96.7 33 Other? . | 100.0 97.3 2.7 ' Except those with mothers born in Latin America. 2 Includes 535 mothers born in Puerto Rico, 106 in other Latin American countries, 20 others. The remaining 2,622 vital events, when clas- sified by the WHO criteria, represented 2,565 live births and 57 fetal deaths of all gestations. The distribution of the events by ethnic group is shown in table 1: the two largest groups were white (39 percent) and Negro (36 percent), with the residual category constituting 25 percent of the total study group. Fetal deaths form a higher proportion of the deliveries for the Negro and other groups than for the white. The distribution of all deliveries by hospital and WHO definition are shown in table 2. The proportions of fetal deaths are greater among the deliveries in the three municipal hospitals than among those in the two voluntary hos- pitals, probably due to the greater proportion of Negro and Latin American clients in munici- pal hospitals. Table 2. Number and percentage distribution of deliveries by hospital, and WHO classification based on vital signs: five select- ed hospitals in New York City, May-July, 1967 Hospital ail . Live births’ Fetal deaths? deliveries Number Total 2,622 2,565 B57 Municipal ALLL 320 314 6 BB. . sis aus 568 548 20 Ci in sssiws 619 594 25 Voluntary Bow osen won 856 851 5 Bl. is va vw 259 258 1 Percentage distribution Total 100.0 97.8 22 Municipal ALLL 100.0 98.1 19 B ..wv.oo us 100.0 96.5 35 Cuvier wsny 100.0 96.0 4.0 Voluntary Dosen oma 100.0 99.4 0.6 E vv oe viniw on 100.0 99.6 04 ' One or more vital signs present (WHO classification). 2No vital sign present. A complete tabulation of all recorded com- binations of vital signs which were present among the 2,622 deliveries is shown by hospital in table 3. For the great majority of cases in each hospital, all four vital signs were present. Combinations which are not listed did not occur. The total number of times each of the four vital signs was present is shown in table 4. Among the 2,622 deliveries, there were 57 fetuses which exhibited none of the four vital signs. These deliveries were classified as fetal deaths according to the WHO definition; the remaining 2,565 deliveries were classified as live births since one or more of the four vital signs were present. Of the four signs, heartbeat was recorded as present for 2,564 of the 2,565 live births. Next in order of magnitude was respira- tory effort, which was present for 97.4 percent of live births, followed by movement of volun- tary muscles (97.2 percent) and pulsation of umbilical cord (96.8 percent). Table 3. Combinations of vital signs present for all deliveries, by hospital: five selected hospitals in New York City, May-July, 1967 Hospital Vital sign present’ Total A B Cc D E Toad .....: «s+: 2,622 320 568 619 856 259 None” . . . . . . «oo 57 20 25 5 1 Oneormore” . . . .. ..... 2,565 314 548 594 851 258 RPMH . ........... 2,389 286 529 550 819 205 PMH . .....c005%% &3 24 2 6 1 1 4 RMH ............ 65 - - 21 2 42 RMHP)* ©... 1 6 : . ; 5 MH ins mems oven 5 - 4 - 1 RPH . ..:v:5:emi3 5 B82 5 1 12 - PH .....ss@cusmsn 27 8 2 5 1 RH ....si3:08®3@ 1 - 1 - - H eo. msci@imsmnyd 10 - 4 2 - Prema mem HE 1 - - - - ! R=respiratory effort; P=pulsation of umbilical cord; M=movement of voluntary muscles; H=heartbeat. 2CJassified as fetal deaths (WHO definition). 3 Classified as live births (WHO definition). 4 For these 11 cases, pulsation of umbilical cord was not recorded as present or absent. Table 4. Number of times each vital sign was present for all deliveries: five selected hospitals in New York City, May-July, 1967 Vital sign Vital sign present Vital sign Buseriation Percent Percent Present Absent of all of live deliveries births All deliveries. . . . . o.oo... 2,622 2,565 57 100.0 Livebirths . . . . . .. ....... 2,565 100.0 Respiratory effort . . . . ......... 2,565 2,498 67 95.3 97.4 Pulsation of umbilical cord? LL... ... 2,554 2,473 81 94.7 96.8 Movement of voluntary muscles . . . . . . 2,565 2,494 71 95.1 97.2 Heartbeat . . . « + 5 5 5 so + 5 «wv» 2,565 2,564 1 97.8 100.0 I Classified as fetal deaths (WHO definition), no vital sign present. 2 Excludes 11 cases for which pulsation of umbilical cord was not recorded as present or absent. The vital signs noted at birth are shown for the total deliveries by gestation in table 5, and by weight at birth in table 6. The distributions of deliveries with no vital sign present were skewed toward the lower ends of both the gestation and birth weight scales. Among the deliveries with vital signs present, there was relatively little difference among the distribu- tions in either table. The largest difference, although it is not large, was between respiratory effort and heartbeat according to birth weight: small infants, those weighing 1,000 grams or less at birth, constituted 0.5 percent of the group with observed respiratory effort and 1.1 percent of the group with an observed heartbeat. Among these small infants, respiratory effort was either not as readily observed or was not established as soon after birth as heartbeat. The differences at other weight groups were of small order. Definition of Live Birth The definition of live birth and its effect on the registration of vital events has long troubled vital statisticians.’-'3 When the WHO definition was constructed for worldwide use, it was recognized that the vast majority of deliveries in some countries occur in hospitals, while in others quite the reverse is true. To be interna- tionally useful, the definition had to be appli- cable to this wide diversity of situations. The WHO definition of live birth encompasses all infants who demonstrate any evidence of life at time of birth even though they may die very soon thereafter. The definitions of live birth and fetal death were clear enough, it was felt, so that if they were applied uniformly in all countries, comparable birth and perinatal statistics could be produced for international comparisons. If, on the other hand, countries were to continue to Table 5. Vital signs present among deliveries by period of gestation: five selected hospitals in New York City, May-July, 1967 One or more . . p . } . Pulsation of | Movement of Period of gestation No vital vital signs Respiratory . umbilical voluntary Heartbeat (completed weeks) sign present | present (WHO effort 1 Shy cord muscles definition) Total . .............. 57 2,565 2,498 2,473 2,494 2,564 Under20 . ............... 7 3 3 1 2 3 2027 ise ies 21 20 13 18 12 20 2B uses ms Erm EWE LD EE 1 272 256 257 263 271 kT 10 1,122 1,105 1,092 1,105 1,122 DBE, i. smi he sms mee 3 743 733 721 732 743 42andover . .............. 5 344 333 325 334 | 344 Notstated . ........ Leite ne ee ig - 61 55 59 56 | 61 "Excludes 11 cases for which pulsation of umbilical cord was not recorded as present or absent. Table 6. Vital signs present among deliveries by weight at birth: five selected hospitals in New York City, May-July, 1967 One or more . No vital vital sian RTO aiEE Pulsation of | Movement of Birth weight (grams) . UD! Sons D Y umbilical voluntary Heartbeat sign present | present (WHO effort 1 ti cord muscles definition) Total ons me msn bmn 57 2,565 2,498 2,473 2,494 2,564 10000rless . . ............. 27 28 13 23 14 27 1001-1800 .......00cnueue 7 32 30 30 28 32 1601-2500 ............... 10 239 228 228 224 239 25014000 ............... 8 2,149 2,112 2,081 2313 2,149 4001ormore . ............. 2 116 115 110 115 116 Notstated . .. ............. 3 1 - 1 - 1 "Excludes 11 cases for which pulsation of umbilical cord was not recorded as present or absent. use different criteria for determining live birth, it could affect not only the number of live births but the number of fetal deaths and neonatal deaths as well. It is obvious that, as long as all pregnancy terminations are divided into a dichotomy of live births and fetal deaths, the definitions will affect the number of fetal deaths as well as the number of live births. The effect of the defini- tions on the number of neonatal deaths is less obvious: if a liveborn infant dies soon after birth (within minutes, perhaps), that death must be considered a neonatal death as well. Thus, once a determination is made regarding the birth, the number of neonatal deaths is also affected. In the present study, the application of the WHO definition which accepts one or more of four vital signs to be indicative of live birth resulted in 2,565 liveborn infants, 57 fetal deaths, and 40 deaths in the first week of life (table 7). A definition which considers only heartbeat as a necessary criterion of live birth would have yielded virtually the same results: 2,564 live births, 58 fetal deaths, and 39 deaths in the first week of life. Another older defini- tion, which mentioned only respiratory effort as indicative of live birth, would have yielded quite different results: 2,498 live births, 124 fetal deaths, and 22 deaths in the first week of life. The last line in table 7 presents the data when even another definition for live birth is used— i.e., that all four vital signs must be present in the first 2 minutes of life before an infant is deemed to be liveborn. Using all four criteria, the data would change markedly: the number of live births would decline to 2,389, and “fetal deaths” would increase to 222. Among the latter, 57 had no vital sign present at birth and are unquestionably fetal deaths. In the remain- ing 165 cases with 1-3 signs present, 18 died during the first week of life and 147 survived that period; these are obviously not fetal deaths. This combination of vital signs as a definition of live birth is, therefore, unacceptable. The number of live births and fetal deaths resulting from every combination of the four signs of life is shown in table 8. The last three columns contain the ratios of the number of events classified according to each of the defini- tions to the number of events classified accord- ing to the WHO definition. Varying combina- tions of vital signs to be observed for alternate definitions of live birth affects the number of live births relatively little: at most, they are understated by 3 percent. In some instances, however, the number of fetal deaths is more than doubled, and the number of deaths in the first week of life is reduced almost by half. Table 7. Allocation of deliveries to live births, fetal deaths, and deaths in the first week of life according to vital signs included in various definitions of live birth: five selected hospitals in New York City, May-July, 1967 Vital signs missing among all deliveries Alternate definitions Tol Live Specified All signs missing Deaths of live birth deliveries || births Total ian in first 59 20 weeks | 20-27 | 28 weeks |" missing Total week or less weeks or more Specified vital sign required: Respiratory effort . . . . .. .. 2,622 || 2,498 124 67 57 7 21 29 22 Pulsation of umbilical cord’ 2611 || 2473 138 81 57 7 21 29 33 Movement of voluntary muscles 2,622 || 2,494 128 71 57 7 21 29 24 Heartbeat . . . .. ........ 2,622 | 2,564 58 1 57 7 21 29 39 One or more vital signs (WHO definition) . . . ....... 2,622 || 2,565 57 - 57 7 21 29 40 All vital signs! ow mE ow ke 2611 || 2,389 222 165 57 7 21 29 18 1Excludes 11 cases for which pulsation of umbilical cord was not recorded as present or absent. Table 8. Number of vital events determined by alternate vital signs and ratio to events determined by WHO definition: five selected hospitals in New York City, May-July, 1967 Vise) wignis) wo oe Ratio to events identified observed for alternate Live Fetal Deans by WHO definition finiti f irth d de nitions ° births Ll week Live Fetal Deaths in live birth i . births deaths first week Only one sign required (R) Respiratory effort . . . . . . . . 2,498 124 22 0.97 2.18 0.55 (P) Pulsation of umbilical cord! 2,473 138 33 0.97 2.42 0.83 (M) Movement of voluntary muscle 2,494 128 24 097 225 0.60 (H) Heartbeat . . . . . . . .... . . 2,564 58 39 1.00 1.02 0.98 Any of the following Ror? vem mensnsms 2,550 72 33 0.99 1.26 0.83 RorM . ............... 2527 95 28 0.99 1.67 0.70 RorH . . .......... .. 2,564 58 39 1.00 1.02 0.98 PorM 2,554 68 37 1.00 1.19 0.93 PorH . ........... .. .. 2,565 57 40 1.00 1.00 1.00 MOF ws 58% 26 + ons wn 2,564 58 39 1.00 1.02 0.98 RPOrM inns ns man 2,555 67 37 1.00 1.18 093 B.POrH ov 5 svc nonommn ovis 2,565 57 40 1.00 1.00 1.00 R,M,orH . . . _._ . .. .... . 2,564 58 39 1.00 1.02 0.98 P,M,orH . .... .._ ..... .. 2,565 57 40 1.00 1.00 1.00 R,P,M, or H (WHO definition) . . .. ...... 2,565 57 40 1.00 1.00 1.00 ' Excludes 11 cases for which pulsation of the umbilical cord was not recorded as present or absent. The purpose of the structured WHO defini- tions of live birth and fetal death is to assist in determining at time of birth whether a delivery is to be classified as a live birth or a fetal death. If the status of the infant is to be determined at time of birth, it seems axiomatic that the observations need to be made as soon as the infant is completely separated from its mother. For the purpose of this study, a 2-minute interval was assigned to the observers as the period within which the signs of life were to be recorded as signs of life “at time of birth.” It was felt that if specific signs did not manifest themselves within that period, they would re- main absent. However, the data demonstrated that some of the vital signs do not manifest them- selves within 2 minutes of birth in all infants, even in those infants who survive the first week of life (table 9). For example, 67 of 2565 infants who showed any of the four signs of life failed to demonstrate respiratory effort in the first 2 minutes of life. Of the 67 infants, six died within an hour of birth, 10 died in the remain- der of the first day, two died in the remainder of the first week, and 49 survived the first week of life. It is obvious, therefore, that breathing within 2 minutes of birth, taken as the sole criterion of life, is unsatisfactory. The data in table 9 demonstrate that if either pulsation of umbilical cord or movement of voluntary muscles in the first 2 minutes is the sole determinant of life at birth for registration purposes, even greater numbers of questionable cases are encountered than when respiratory cffort is the sole criterion of live birth. The only vital sign which discriminated well was heart- beat: only one of the 2,565 infants failed to demonstrate a heartbeat within 2 minutes of Table 9. Vital signs which were missing during the first 2 minutes of life among selected categories of 2,565 live births by survival of infant: five selected hospitals, New York City, May-July, 1967 No No pulsation No movement i Le No Event respiratory of umbilical of voluntary heartbeat effort cord muscles Live birth (WHO definition) . . . .. . .. 67 81 71 1 Death: Underone hour . . .......... 6 2 8 1 V2BHOUIS . ui vv ois mv mon mm 10 5 7 - 16days ................ 2 1 Survived firstweek . . . ....... .. 49 74 55 - birth, but showed another sign of life during that interval. That infant died later in the first hour of life. Effect on Vital Statistics Rates The data shown in tables 1-8 include all 2,622 births in the study group. According to the WHO definition, 2,565 were live births and 57 were fetal deaths of all gestations. The require- ment to register fetal deaths of all gestations is not typical of the registration requirements of most States or of other countries. National statistics and the statistics for most States in the United States include only those fetal deaths with gestation periods of 20 completed weeks or more. When the data from the present study were rearranged in accord with these criteria, it became apparent that very few fetal deaths of less than 20 completed weeks of gestation were observed in delivery rooms (table 7). The large number of fetal deaths which were expected with gestation periods of less than 20 weeks were apparently born elsewhere—either outside hospitals or occasionally in other parts of hospitals (e.g., surgery, emergency rooms)—and are therefore not part of this study. In registra- tion areas requiring the registration of fetal deaths irrespective of gestation, the omission of these vital events is of serious statistical conse- quence. The consideration of definitions of live birth is not entirely an end in itself. Its purpose here is in relation to the vital statistics rates which are produced. According to earlier definitions of live birth, breathing at time of birth was the only specified criterion of life. In the preceding section, it was demonstrated that if this criterion was strictly adhered to, 67 of 2,565 infants who were classified as live births according to the WHO definition did not breathe in the first 2 minutes of life (table 9) and would not be considered as having been born alive. How would these 67 vital events have been registered? If they were not liveborn, by definition they were fetal deaths, and their registration would depend on local requirements. In most States, if the period of gestation is 20 completed weeks or more, they should be registered as fetal deaths; if less than that period, they need not be registered at all. Although the weakness of depending on definitions based on single criteria of life has been mentioned, it is interesting to assess the effect the definitions would have on some of the commonly used vital statistics rates. These rates are not shown in this report as illustrations of rates in actual populations. They are presented to demonstrate the implications of strict applica- tion of the criteria of live birth to specific vital events and strict application of these events within the structure of vital statistics rates. The methods of computation are shown in Appen- dix II. The data in table 10 are presented in relationship to registration practice in effect for most of the United States; that is, they include only those fetal deaths with gestation periods of Table 10. Fetal and early neonatal mortality rates and ratios for each combination of vital signs: five selected hospitals in New York City, May-July, 1967 — Births with Births with all Biialie with all Vital sign(s) to be one or more specified sign(s) Seine ob sme) Fetal death Fetal death Deaths in Mortality rate observed for alternate specified sign (s) absent, others 9 9 rate’ ratio first week in first week es 1 (fetal deaths) definitions of present present isebinn cols. B+C cols. B+C col. D col. A col. B col. C cols. A+B+C col. A col. D py Only one vital sign Number Rate per 1,000 Number Rate per 1,000 (R) Respiratory effort . . . . ....... 2,498 67 50 44.7 46.8 22 8.8 (P) Pulsation of umbilical cord? ew 2,473 80 50 499 52.6 33 133 (M) Movement of voluntary muscles 2,494 70 50 45.9 48.1 24 9.6 (H) Heartbeat . . . . . . ........« 2,564 1 50 19.5 189 39 15.2 Any of the following ROFIP . od soi s sms wows simu sw 2,550 15 50 249 255 33 129 ROFM 55 smo s ss smn sms vor 2,527 38 50 337 348 28 1.1 ROFH ga: ane vamos soma » ame no 2,564 50 195 199 39 15.2 POIM Gis sms vimn oo wows sms su 2,554 1 50 233 239 37 145 POIH ws cmme nms van some vow 2,565 - 50 19.1 19.5 40 15.6 MOF H oon wmin a vow mows mime ey 2,564 1 50 195 199 39 15.2 RPOTM oncu cma nimi soma & 5818 3 2,555 10 50 29 235 37 145 R,P,orH . ................ 2,565 - 50 19.1 195 40 15.6 RMorH ..........‘¢ccouu- 2,564 1 50 195 199 39 15.2 PMorH ................: 2,565 50 19.1 195 40 15.6 R,P,M,or H (WHO definition) . . .......... 2,565 50 19.1 195 40 15.6 ! Gestation periods of 20 completed weeks or more. 2 Excludes 11 cases for which this item was not recorded as present or absent. 20 completed weeks or more. Included as live births (column A) are all births that exhibited the specified combinations of vital signs shown in the stub of the table. Those infants that did not exhibit the specified vital signs would be registered as fetal deaths only if they completed 20 weeks of gestation (column B). The number of infants who exhibited no vital sign but whose periods of gestation were 20 completed weeks or more would remain constant (column C). The fetal death rates and ratios are computed from the first 3 columns. From the table, it is obvious that the greatest numerical variation is in col- umn B, which reflects the variation one might expect for infants who exhibited one or more, but not all, signs of life. If respiratory effort were the only vital sign which would be considered to classify an infant as liveborn and all other infants were classified as fetal deaths, the fetal death rate would be 44.7 per 1,000 live births and fetal deaths having gestation periods of 20 completed weeks or more. For the next two vital signs (pulsation of the umbilical cord, movement of the voluntary muscles) the rates were somewhat, but not much, higher—49.9 and 45.9, respectively. How- ever, if heartbeat were the only vital sign considered in defining live birth, the fetal death rate would be less than half the rate for any of the other three signs of life—19.5 per 1,000. As expected, this rate is rather close to the rate obtained when any of the four vital signs is considered to signify life; adhering to the WHO definition, the fetal death rate would be 19.1 per 1,000. The mortality rate for the first week of life would also show wide variation. The mortality rate when heartbeat was the only vital sign considered to define live birth (15.2) was close to the rate obtained using the WHO definition (15.6). The greatest differences were for the definitions which considered only respiratory effort or movement of voluntary muscles to define live birth—8.8 and 9.6, respectively. In addition to the wide variation in the rates which would be introduced by strict adherence to observing only certain sign(s) of life, the problem of what to do with the births exhibiting other signs of life remains. The births shown in column B of table 10 would not be considered live births under the specified criteria, but neither are they fetal deaths. The larger the number of births with some, but not all, of the signs of life (column B), the less satisfactory is that definition of live birth. It has long been accepted that differences in rates are found depending on the vital signs used to define a live birth, but the magnitude of the differences has been largely unknown. It has been thought that the major discrepancy was due to allocating some live births to fetal deaths and wvice versa. To overcome this difficulty, perinatal rates (or ratios) have been proposed. For the numerator of a perinatal rate, fetal deaths of specified gestation periods and early neonatal deaths of specified ages are summed. The rationale is that by summing them, border- line decisions about live birth and fetal death are avoided. The denominator is the sum of the live births and the fetal deaths which are represented in the numerator. One perinatal mortality rate which is com- monly used for international comparisons com- bines fetal deaths with gestation periods of 28 completed weeks or more with deaths in the first week of life for the numerator and com- bines the live births and fetal deaths with gestation periods of 28 completed weeks or more for the denominator. By convention, the result is multiplied by 1,000. This rate has been advocated within the framework of using the WHO definitions of live birth and fetal death. The present study provides an opportunity to examine the perinatal mortality rates under strict application of each of the combinations of vital signs, and the components of the vital statistics rates. Perinatal mortality rates and ratios for the present study are shown in table 11. As before, when heartbeat alone or heartbeat in combina- tion with other vital signs were deemed to qualify an infant as liveborn, the rates were very close; in fact, for perinatal mortality, the rates were identical (26.6). However, when any one of the other three vital signs alone was deemed sufficient to define a live birth, the perinatal mortality rates were markedly higher—respira- tory effort, 42.9; pulsation of the umbilical cord, 53.9; and movement of voluntary muscles, 44.5. The ratio of the highest perinatal mortality rate to the lowest rate was 2.0, and the marked difference was due to the numerical variation shown in column B. As in the previous table, the larger the entry in this column, the less satisfac- Table 11. Selected mortality rates and ratios for each combination of vital signs: five selected hospitals in New York City, May-July, 1967 Births with Births with all § i Deaths in first i i . Vi 3 one or more specified sign (s) Births with # week with one Fetal death | Fetal death Mortality rate Perinatal Berirgtsl ital sign(s) to be ecified sign (s) sent, othen 4 signs absent r more specified oe ratio in first mortality mortality observed for alternate spec! #9nis . s (fetal deaths) 9 OT SReCH| 8 week rate ratio definitions of five birth present present sign(s) present cols. B+C cols. B+C col. D cols. B+C+D |cols. B+C+D col. A col. B col. C col. D mmm; || stmm——— ———— meme | Sem—— cols. A+B+C col. A col. A cols. A+B+C col. A Only one vital sign Number Rate per 1,000 (R) Respiratory effort... ... 2,498 60 29 22 34.4 36.6 8.8 429 44.4 (P) Pulsation of umbilical cord? . . . . . 2,473 77 29 33 41.1 429 13.3 53.9 56.2 (M) Movement of voluntary muscles . 2,494 62 29 24 35.2 36.5 9.6 44.5 46.1 (H) Heartbeat . . . ..... .. ... . . 2,564 1 29 39 116 1.7 16.2 26.6 26.9 Any of the following RorP 2,550 12 29 33 15.8 16.1 129 286 29.0 BOFM yi 62m 55 sammsmenns 2,527 32 29 28 236 24.1 1.1 34.4 35.2 BOM ovum men manny 2,564 1 29 39 11.6 11.7 16.2 26.6 26.9 POEM . viic cna m nn nmms 2,554 9 29 37 14.7 14.9 14.5 289 29.4 PorH .................. 2,565 - 29 40 11.2 1.3 15.6 26.6 26.9 MOF 4 1 is 5a vn mw eine rm ow ow 2,564 1 29 39 1.86 11.7 15.2 26.6 26.9 FLPOPM i 5 45 0h vivo 2 im wom 3 we 8 2,555 8 29 37 143 145 14.5 28.5 29.0 BPO ow vvvnwsmis vines 2,565 - 29 40 11.2 13 15.6 26.6 26.9 RMorH ................. 2,564 1 29 39 116 11.7 16.2 26.6 26.9 PMOrd «vo vo mssmsmanimai « 2,565 - 29 40 11.2 13 15.6 26.6 26.9 R,P,MorH (WHO definition) . . ........ . . 2,565 29 40 11.2 11.3 166 26.6 26.9 ! Gestation periods of 28 completed weeks or more. * Excludes 11 cases for which this item was not recorded as present or absent, 1 tory were the specified criteria as a definition of live birth. From these data, one must conclude that perinatal mortality rates and ratios do not entirely overcome the statistical artifacts intro- duced by differences in definitions of live birth based on selected vital signs. The variation between rates shown in table 11 reemphasizes the necessity for standard logical definitions and their uniform application to produce compar- able statistics. Registration A secondary objective of the present study was to compare the vital events collected in the study with the events registered in the official vital records in the City of New York Depart- ment of Health. The Health Code requires that every pregnancy termination shall be registered, irrespective of the duration of pregnancy. There- fore, each event which was identified in the study should have been registered as either a live birth or a fetal death. All official live birth and fetal death records for the 2,622 events which occurred in the five selected hospitals during the study period were linked to the study records. The results of the linking operation are shown in the following table: Vital event by Vital records WHO definition . Death No Live Fetal og : for study records : in first vital birth death week record Live birth (Some vital sign) . . . . . . . 2,561 4 - - Fetal death (No vital sign) . . . .. ... - 57 - 1 Norecord . . vs 5 6 ov v2 v0 0s 13 13 - - Total deaths in first week identified from all sources . . . . . .. - 36 4 1 Excluded because a vital record was not filed or because a study form was not filed on time or was incomplete. In all, 2,629 cases were identified by either mechanism, and of these, seven were excluded from the study. One case with no vital sign present had both a fetal death and live birth record filed for the same event, and the live birth record was not used in the study based on the information on the study form. In four cases where the study form indicated that a sign of life was present, fetal death rather than live birth records were prepared by the hospital and filed with the Department of Health. Significant characteristics of these four cases were the following: Item Case | Case Il Case 11 Case IV Hospital . . ..... +... A A C D Type of hospital . . . . .. ....... Municipal Municipal Municipal Voluntary Birth weight (grams) . . . . . . . . ... Unknown 312 624 1700 Gestation (completed weeks) . . . . . . 36 29 27 35 Yes: Malformation . . . . . . . . os type None None Anencephaly unspecified Age at death (min)... ... ..... 1 Unknown 10 5 Type of delive Cesarean Spontaneous Spontaneous Cesarean ype of delivery . . . . . .. section Breech Breech section Pulse i IONS © 4 vv he ee ee eee ee 1 i Vital signs Heartbeat Pulse Heartbeat All 4 signs 12 Three of the four infants were known to be of low birth weight (2,500 grams or less); all four infants were preterm (less than 37 completed weeks of gestation); all had other than a normal delivery; and three of the four infants died within 10 minutes of birth. One fetus was described as being congenitally malformed of unspecified type and another as being anen- cephalic. The irregularities in registration for these four infants are no doubt associated with the problems in the delivery rooms which accompanied these “complicated” cases. Overall, the 2,565 live births identified in the study were all registered although four were erroneously registered as fetal deaths, represent- ing an underregistration of 0.2 percent of live births. This deficiency is similar in magnitude to that shown in other studies.8:10,12 The statis- tical effect of the failure to register these four events as early neonatal deaths is more serious. Since the study identified 40 deaths in the first week of life, the understatement of four cases represents an understatement of 10 percent of the mortality in the first week of life. During the course of the study, 57 fetal deaths were identified based on the WHO definitions, but 61 fetal death records were filed. Four of the recorded fetal deaths were determined to be live births on the basis of the observer’s record, and therefore each should have had a live birth and a death record on file instead of a fetal death record. The overregistra- tion which was found for this group of fetal deaths is not an estimate of the degree of completeness of all fetal death registration in the City of New York. The Health Code requires the registration of all pregnancy terminations irre- spective of the period of gestation, but most early fetal deaths are not delivered in delivery rooms and consequently were not included in this study. The study demonstrated a relatively small degree of error in the registration of live births in the five hospitals. The misregistration of four live births as fetal deaths was documented as being associated with early termination of preg- nancy, death in the first 10 minutes of life, and with complicated deliveries. Thus despite the fact that the study observers and the study staff were not entirely independent, one source of inaccurate vital registration and vital statistics was detected—i.e., the failure to register some very early neonatal deaths as live births and deaths, and their misregistration as fetal deaths. DISCUSSION The conduct of this study has demonstrated the feasibility of conducting studies of vital signs at birth in the delivery rooms of hospitals. The study site, New York, has a history of collabora- tive obstetrical studies, and close cooperation has developed between the Department of Health and the obstetrical departments of hos- pitals. Through the assistance of the Depart- ment’s obstetrical consultants, the cooperation of the heads of the obstetrical departments of the five hospitals was secured. The presence of observers in the delivery room was accepted by the hospitals and seemed to cause no problems. The fact that they were selected by the heads of the obstetrical departments from residents or interns of the same hospitals probably facilitated the conduct of the study. The matter of independent observations of each of the vital signs was not completely solved. Observers were from the same hospitals in which the study was conducted and may have been influenced by common training. After the fact, it can only be said that the information recorded on the study forms represented a synthesis of the observations of the observer and some unknown input from the accoucheur through the observer. In future studies which may be conducted in other locales, it may be possible to establish a team of observers to rotate among a group of study hospitals to overcome some of the reservations which were experienced. The study form was fairly simple and ap- peared to cause no problems. The routing of the forms to the Department of Health at weekly intervals helped to keep the hospitals aware that the forms had to be transmitted promptly. With regard to vital signs, it was found that heartbeat in the first 2 minutes of life could serve as the only criterion of live birth virtually as well as the WHO definition. Only one of the 2,565 live births would have been missed if heartbeat were the only vital sign considered (table 9). Pulsation of the umbilical cord was poorest in this respect: 81 of the 2,565 live births would have been missed. In general, when there is a strong heartbeat in the newborn 13 infant, a pulsation of the umbilical cord can be detected. However, when the heartbeat is weak, pulsation of the cord may be difficult to detect. Pulsation of the cord was not recorded as present or absent for 11 cases and was either overlooked by the observer or was not recorded due to an oversight. These 11 instances occurred in two of the five hospitals (table 3). Since the WHO definition is intended for use under a great range of circumstances, it would not be desirable to extrapolate from the present experience in a highly structured hospital setting to effect a change in definition for use in a worldwide range of nonhospital settings by persons with a wide range of medical experience. In this study, heartbeat was detected by highly trained medical personnel under favorable cir- cumstances including the use of palpation and auscultation, but these conditions may not readily be duplicated on a worldwide basis. For international comparisons, it seems quite evident that the current WHO definitions of live birth and fetal death are superior to any based on single vital signs. Of the 2,565 infants classified as live births by the WHO definitions, 67 failed to demonstrate any respiratory effort in the first 9 minutes of life (table 9). Of these, 18 died before the end of the first week, and 49 survived the first week of life. Thus respiratory effort in the first 2 minutes of life is highly unsatisfactory as an only indicator of live birth. After the fact, it is easy to say that even though these 67 infants did not breathe within 2 minutes after birth, their demise later in the first week after birth or their survival past that point is evidence that they were born alive. Such reasoning implies either that other signs of life must have been used as a criterion of life at birth, or that observations at some unspecified point in time following birth played a part in determining whether an infant was alive at time of birth. Although such practices would provide rational solutions to questionable cases, they demonstrate that breathing alone is not a satis- factory criterion for establishing live birth and reinforce the need for other signs of life to determine whether an infant is alive at time of birth. The results of this study demonstrate the wisdom of including a number of possible signs of life in the definition of live birth. The WHO 14 definition contains the words “... breathes or shows any other evidence of life such as beating of the heart, pulsation of the umbilical cord, or definite movement of voluntary muscles, whether or not the umbilical cord has been cut or the placenta is attached;. . . 2’ The structure of the definition indicates that these four signs are illustrative and not necessarily exhaustive. The problem of defining “life” currently faces the medical and legal professions and is still subject to change. What is life? At what point in time does an individual die? The questions are presently of particular importance with regard to organ transplants. In that connection, it has been suggested that the point at which the brain ceases to emit impulses is the true time of death. Such new concepts may play a role in defining ‘qife” in the future, and they may impinge on the WHO definition of “live birth” as well. However, their incorporation into a definition of live birth for worldwide usage may not be practical for many years to come. The alternative definitions of live birth which can be constructed using observations of one or more of the four signs of life during the first 2 minutes after birth, and without regard to whether the signs appeared later, resulted in marked differences in vital rates. The smallest differences from rates based on the WHO defini- tions were found for those combinations of signs which included heartbeat. These results clearly demonstrate that statis- tical manipulation of data, as in the perinatal rate, cannot overcome the problems caused by conceptual differences in definitions which al- most inevitably (1) allow events subsequent to the time of the birth to influence its classifica- tion as a fetal death or live birth, and (2) foster the failure to register a live birth and infant death when the early postpartum course is other than survival. One purpose of this study was to examine the statistical effect of the possible alternate combi- nations of signs of live birth on vital statistics. From a statistical view, the matter of definitions would not matter much if the errors were of such a magnitude that the rates would be affected little or, perhaps, not at all. The comparison presented in tables 10 and 11 is carried out within the framework of the dichotomy between live birth and fetal death as defined by the World Health Organization. In this report, the data can be arranged to segregate this class, and its magnitude in regard to vital statistics rates can be examined. The effect of acceptance of any of the four signs of life as an indication of live birth, in contrast with most of the other combinations which are shown, is to include more of the births as liveborn infants and fewer as fetal deaths (table 8). This affects not only the number of fetal deaths in the numerator of a rate and the number of live births in the denominator but the number of deaths in the first week of life as well. Thus one effect of the application of the WHO definition would be to lower the fetal death ratio and to increase the mortality rate in the first week of life (tables 10 and 11). The basic reason for the unsatisfactory nature of the rates other than those involving heartbeat is that a number of events failed to manifest one of the specified signs of life in the first 2 minutes of life, but were alive at the end of the first week of life: Specified sign Infant alive Vital sign absent in first at end of 2 minutes of life | first week Respiratory effort . . . . . . 67 49 Pulsation of umbilical cord ............ 81 74 Movement of voluntary muscles... . 71 55 Heartbeat . . . . . . . . specific event, but because it was thought that they would also reduce some of the problems associated with differing practices and education of accoucheurs in various countries. The present study does not address itself directly to interna- tional comparisons, but examines the data for New York City to determine to what degree the perinatal rates solve the problem of wide varia- tion which was noted in the fetal death ratios and mortality rates in the first week of life (table 10). The data in table 11 address themselves to this point. In this table, fetal deaths are limited to those with gestation periods of 28 completed weeks or more, as is the custom in other countries. The range of the rates based on the several definitions of live birth and fetal death and computed as shown in Appendix II, and the ratios of the highest to the lowest rates are as follows: Ratio of Rate (ratio) Range Differ-| highest to ence | lowest rate (ratio) Fetal death rate "112411 299 367 Fetal death ratio "113429 316 3.80 Mortality rate in first week of life . . . 8.8-115.6 6.8 1.77 Perinatal mortality rate . . . . 1266-539 27.3 2.03 Perinatal mortality ratio 126.9-66.2 29.3 2.09 Onc of the fundamentals in constructing fetal death ratios and mortality rates for the first week of life is the absolute dichotomy: if a vital event is not a live birth, it is a fetal death. To overcome problems caused by this stipulation, it has been widely suggested that perinatal mortal- ity rates would be preferable to fetal death rates and neonatal mortality rates. For the present study, perinatal deaths were obtained by sum- ming all fetal deaths with gestation periods of 28 completed weeks or more and early neonatal deaths in the first week of life. Perinatal mortality ratios have been particu- larly advocated for use in international compari- sons, not only to avoid the problems of differen- tiating between live birth and fetal death for a ! Rates based on WHO definitions of live birth, fetal death. The arithmetic differences between the maxi- mum and minimum rates are only slightly lower for perinatal than for fetal mortality. However, the ratios between the highest and lowest rates for perinatal deaths are markedly lower than for fetal deaths. Moreover, the ratios for perinatal rates and ratios are closer to the ratio for first-week mortality than the ratios for fetal mortality. Thus the use of perinatal mortality rates or ratios helps to reduce the range of rates and to offset the differences in one direction for fetal death rates (ratios) and in the other direction for the first-week mortality rates. Yet, despite the achievements due to recombi- nation of the basic data, the importance of the 15 basic definitions and their proper application at time of birth remains. The data which have been presented demonstrate that the presence of one or more vital signs at time of birth is a better definition of live birth than older definitions which relied on observing only one sign of life—i.e., respiratory effort. If applied properly, the WHO definitions of live birth and fetal death tend to promote more uniform statistics than any of the other definitions which were con- sidered here. The uniform application of the definitions of live birth continue to play an important part in quantitative measures of pregnancy loss. The definitions of live birth and fetal death must be applied consistently by persons responsible for providing the information for vital records (physicians, nurses) as well as by those who actually complete the documents (medical rec- ords librarians, and other record room person- nel). In the hurried affairs of physicians’ daily activities and the busy routines of hospital record rooms, basic matters such as definitions of vital events can easily be overlooked. There- fore, it is all the more important that these individuals be aware that their practices have direct bearing on the resulting statistical infor- mation. The completion of a single fetal death record in place of a live birth record for an infant who lives only a few minutes may appear unimportant. But wide-scale practices such as these, if they exist, would adversely affect the statistical end product to a significant degree. CONCLUSION The present study of vital signs present at birth was undertaken with two purposes In mind. One was to determine the feasibility of conducting studies of evidence of life at birth in an actual working situation in hospital delivery rooms. Although the study was largely success- ful in this regard, certain difficulties were encountered and have been described. The study was conducted in what was considered one of the most desirable settings in the country: within close range of a health department already actively engaged in the registration of all products of conception. The close relationships existing between the City of New York Depart- ment of Health and the obstetrical groups in the hospitals proved to be a distinct advantage. 16 That such studies are needed is evident. The relative scarcity of definitive information on vital signs present at birth and its effect on vital registration and vital statistics have resulted in confusion and indecision. Factual information is needed to form the basis of intelligent discus- sion. The study should be reproduced in other settings, thus adding to the local information available to vital statisticians and registrars re- garding their own systems. The studies do not require great expenditures for laboratory work or mechanical equipment—the greatest part of the cost is for observer time. The amount of unoccupied time of the observers adds to the cost. Therefore, hospitals of sufficient size should be chosen to minimize the added cost due to possible unproductive time of observers when too few births occur in a given hospital. The second purpose of the present study was to determine the quantitative effect of varying combinations of vital signs which may be used to define live birth. The study documented the vital signs which were observed at birth and demonstrated the weakness of relying only on respiratory effort in the first 2 minutes of life as a criterion for defining live birth. Pulsation of the umbilical cord and movement of voluntary muscles were also found to be less than ideal. Heartbeat as detected by palpation or ausculta- tion was most often present, having been absent in the first 2 minutes of life for only one of 2,565 live births. Because this sign was present in every instance but one, mortality rates based on data using this sign alone yielded virtually the same results as the WHO definition which includes any of the four vital signs. If heartbeat is omitted, the use of other signs to define a live birth and the resulting mortality rates were found to be less satisfactory. The use of perina- tal rates ameliorated, but did not overcome, the problems incurred by definitions of live birth based on different signs or combinations of vital signs. The study emphasizes the importance of basic information in vital registration and vital statis- tics. Its relevance extends beyond internal hos- pital procedures to problems of registration practices and the comparability of vital statis- tics. Not only are local, State, and national statistics affected by these basic definitional considerations, but international statistical com- parisons arc also influenced. In all areas of statistical investigation, the quality of the data are dependent on logical and specific definitions which are uniformly applied to the observations under consideration. The statistics are no better than the basic data, which in turn can be no better than the observations and their measurement. Demographic and epidemiologic studies of infant mortality will continue to depend heavily on such elementary considerations. In fact, the lower the rates, the more important it is to control the errors of measurement including those which are attribut- able to definitions. If uncontrolled, such errors may exceed real differences and obliterate or exaggerate statistical differences. With infant and perinatal mortality rates at their present levels in this country, it is be- coming increasingly important that errors of measurement be kept to a minimum. To this end, the present study contributes to the under- standing of vital signs at birth, definitions of live birth and fetal death, and the possible relation of these matters to the level of commonly used vital statistics rates. REFERENCES I National Center for Health Statistics: International compari- son of perinatal and infant mortality: The United States and six West European countries. Vital and Health Statistics. PHS Pub. No. 1000-Series 3-No. 6. Public Health Service. Washington. U.S. Government Printing Office, Mar. 1967. 2National Center for Health Statistics: Infant loss in the Netherlands. Vital and Health Statistics. PHS Pub. No. 1000-Series 3-No. 11. Public Health Service. Washington. U.S. Government Printing Office, Aug. 1968. National Center for Health Statistics: Infant and perinatal mortality in England and Wales. Vital and Health Statistics. PHS Pub. No. 1000-Series 3-No. 12. Public Health Service. Washing- ton. U.S. Government Printing Office, Aug. 1968. 4World Health Organization: Third World Health Assembly, resolutions and decisions. Official Records of the World Health Organization, No. 28. Geneva, (Dec.) 1950, pp. 16-17. Dudfield, R.: A critical examination of the methods of recording and publishing statistical data bearing on public health, Journal of the Royal Statistical Society 68 (Part I):10 (Mar.) 1905. 6Dudfield, R.: Still-births in relation to infantile mortality, Bulletin de L’Institute International de Statistique, 20 (Part 2):146, 1915. "League of Nations, Health Committee: Report of the Committee Studying the Definition of Dead-Birth. C.M. 1925, C.224, M. 80. App. 2, p. 78. 8 American Public Health Association, Committee to Consider the Proper Definition of Stillbirth: Definition of stillbirth, Am. J. Pub. Health 18:25-32, 1928. Pascua, M.: Diversity of stillbirth definitions and some statistical repercussions. WHO Epidemiological and Vital Statis- tics Report 1:210-222 (Mar.) 1948. Oworld Health Organization: Report on Definition of Still- birth and Abortion. WHO/HS/STDEF /4 (14 Feb.) 1950. United Nations: Handbook of Vital Statistics Methods, Studies in Methods. ST/STAT/Ser.F[7. New York, Statistical Office of the United Nations, (Apr.) 1955. Soop, E.: Svensk spadbarnsdodlighet och definitionen pa levande fodd (Swedish infant mortality and live birth defini tions). Svenska Lakartidningen. 55(16):1148-50 (18 Apr.) 1958, pp. 1-7. 13world Health Organization: Study of the Effect of Includ- ing in Vital Statistics Live-Born under 28 Weeks of Gestation and Dying before Registration. WHO/HS/Nat. Comm. 161. (24 Apr.) 1964. msmnte C3 Orem 17 18 12. APPENDIX | RECORDING FORM DIRECT OBSERVATION OF DELIVERIES CHILD'S NAME 2. CHILD'S CHART NO. Last MOTHER'S NAME __ 4, MOTHER'S CHART NO. Last First DATE OF DELIVERY 6. TIME OF DELIVERY 7 .SEX OF CHILD Month Day Type of Delivery 9. Presentation 10.Delivered by 11. Plurality Spontaneous [ ] Vertex [ 1] Attending { 1] Single birth Forceps [ ] Breech [ ] Resident 1f not single birth specify whether Cesarean [ 1 Transverse [ ] Intern [ 1 1st [ ] Twins Other,specify [ ] Other, specify [ ] Medical student [ ] 2nd of [ ] Triplets [ ] Nurse-Midwife [1 3rd [ ] Student Nurse- midwife Observer [ ] Other, specify DELIVERY ROOM OBSERVATIONS Observe neonate for presence or absence of vital signs within 120 seconds of the delivery of its entire body from the body of its mother. (if observation of any of the vital signs is not made explain below) Vital Signs (circle one on each line) 13. Respiratory Effort.............. wee.es...absent present 14. Pulsation of Umbilical Cord.......... +» absent present 15. Movement of Voluntary Muscles.......... absent present 16. Heart Beat....ceec.es wb RE BES Ew absent present 17. If live birth did infant die before leaving delivery room? Yes_ __ No_____ 18. If answer to 17 is Yes, give age at death Hours Mins . COMMENTS ELB-1 APPENDIX II COMPUTATION OF RATES (All rates are expressed per 1,000) Rate Numerator Denominator FETAL DEATH RATE Intermediate and late fetal deaths Late fetal deaths Fetal deaths with gestation periods of 20 completed weeks or more Fetal deaths with gestation periods of 28 completed weeks or more Live births, and fetal deaths with gestation periods of 20 completed weeks or more Live births, and fetal deaths with gestation periods of 28 completed weeks or more FETAL DEATH RATIO Intermediate and late fetal deaths Fetal deaths with gestation Live births periods of 20 completed weeks or more Late fetal deaths Fetal deaths with gestation Live births periods of 28 completed weeks or more INFANT MORTALITY RATE Deaths in the first year of life Live births NEONATAL MORTALITY RATE Deaths in the first four weeks Live births of life PERINATAL MORTALITY RATE! Fetal deaths with gestation periods of 28 completed weeks or more, and deaths in the first week of life Live births, and fetal deaths with gestation periods of 28 completed weeks or more PERINATAL MORTALITY RATIO! Fetal deaths with gestation periods of 28 completed weeks or more, and deaths in the first week of life Live births ! perinatal rates based on several combinations of gestation periods and age at death are found in the literature. These definitions refer to the data included in this report. * U. S, GOVERNMENT PRINTING OFFICE : 1972 482-003/4 19 Series 1. Series 2. Series 3. Series 4. Series 10. Series 11. Series 12. Series 13. Series 14. Series 20. Series 21, Series 22, VITAL AND HEALTH STATISTICS PUBLICATION SERIES Formerly Public Health Service Publication No. 1000 Programs and collection procedures.— Reports which describe the general programs of the National Center for Health Statistics and its offices and divisions, data collection methods used, definitions, and other material necessary for understanding the data. Data evaluation and methods research.—Studies of new statistical methodology including: experi- mental tests of new survey methods, studies of vital statistics collection methods, new analytical techniques, objective evaluations of reliability of collected data, contributions to statistical theory. Analytical studies.—Reports presenting analytical or interpretive studies based on vital and health statistics, carrying the analysis further than the expository types of reports in the other series. Documents and committee reports,—Final reports of major committees concerned with vital and health statistics, and documents such as recommended model vital registration laws and revised birth and death certificates. Data from the Health Interview Survey.—Statistics on illness, accidental injuries, disability, use of hospital, medical, dental, and other services, and other health-related topics, based on data collected in a continuing national household interview survey. Data from the Health Examination Survey.—Data from direct examination, testing, and measure- ment of national samples of the civilian, noninstitutional population provide the basis for two types of reports: (1) estimates of the medically defined prevalence of specific diseases in the United States and the distributions of the population with respect to physical, physiological, and psycho- logical characteristics; and (2) analysis of relationships among the various measurements without reference to an explicit finite universe of persons. Data from the Institutional Population Surveys — Statistics relating tothe health characteristics of persons in institutions, and their medical, nursing, and personal care received, based on national samples of establishments providing these services and samples of the residents or patients. Data from the Hospital Discharge Survey.—Statistics relating to discharged patients in short-stay hospitals, based on a sample of patient records in a national sample of hospitals. Data on health resources: manpower and facilities. —Statistics on the numbers, geographic distri- bution, and characteristics of health resources including physicians, dentists, nurses, other health occupations, hospitals, nursing homes, and outpatient facilities. Data on mortality.—Various statistics on mortality other than as included in regular annual or monthly reports—special analyses by cause of death, age, and other demographic variables, also geographic and time series analyses. Data on natality, marriage, and divorce.—Various statistics on natality, marriage, and divorce other than as included in regular annual or monthly reports—special analyses by demographic variables, also geographic and time series analyses, studies of fertility. Data from the National Natality and Mortality Surveys.— Statistics on characteristics of births and deaths not available from the vital records, based on sample surveys stemming from these records, including such topics as mortality by socioeconomic class, hospital experience in the last year of life, medical care during pregnancy, health insurance coverage, etc. For a list of titles of reports published in these series, write to: Office of Information National Center for Health Statistics Public Health Service, HSMHA Rockville, Md, 20852 Series 2-No. 46 DHEW Publication No. (HSM) 72-1043 U.S. DEPARTMENT OF HEALTH, EDUCATION, AND WELFARE POSTAGE AND FEES PAID Public Health Service U.S. DEPARTMENT OF H.E.W. HEALTH SERVICES AND MENTAL HEALTH ADMINISTRATION 5600 Fishers Lane Rockville, Md. 20852 OFFICIAL BUSINESS Penalty for Private Use, $300 Series 2 Number 47 VITAL and HEALTH STATISTICS DATA EVALUATION AND METHODS RESEARCH Subtest Estimates of the WISC Full Scale IQs RTH TTY U.S. DEPARTMENT OF HEALTH, EDUCATION, AND WELFARE Public Health Service Health Services and Mental Health Administration Vital and Health Statistics-Series 2-No. 47 For sale by the Superintendent of Documents, U.S. Government Printing Office, Washington, D.C., 20402 - Price 50 cents Data Evaluation and Methods Research Series 2 Number 47 Subtest Estimates of the WISC Full Scale 1Q’s for Children A research study of the use of Scaled Scores on the Vocabulary and Block Design subtests of the Wechsler Intelligence Scale for Children (WISC) for predicting Full Scale 1Q’s by socioeco- nomic, sex, and ethnic (Anglo, Negro, and Mexican-American) factors. The relative predictive power of these two subtests is compared with other subtest dyads among 11 of the 12 subtests of the WISC. Regression equations are provided for these two subtests with optimal prediction of Full Scale IQ by ethnic group and for the total sample of 1,310 children studied. VIRACAIT 2 NATRTARCAT | DOCUMENTS DEPARTMENT JUN L9 ] Lidar UNIVERSITY OF CA if ORNIA_ DHEW Publication No. (HSM) 72-1047 U.S. DEPARTMENT OF HEALTH, EDUCATION, AND WELFARE Public Health Service Health Services and Mental Health Administration National Center for Health Statistics Rockville, Md. March 1972 NATIONAL CENTER FOR HEALTH STATISTICS THEODORE D. WOOLSEY, Director PHILIP S. LAWRENCE, Sc.D., Associate Director OSWALD K. SAGEN, Ph.D. Assistant Director for Health Statistics Development WALT R. SIMMONS, M.A. Assistant Director for Research and Scientific Development JAMES E. KELLY, D.D.S., Dental Advisor EDWARD E. MINTY, Executive Officer ALICE HAYWOOD, Information Officer DIVISION OF HEALTH EXAMINATION STATISTICS ARTHUR J. McDOWELL, Director JAMES T. BAIRD, JR., Chief, Methodological Research Staff, HAROLD J. DUPUY, Ph.D., Psychological Advisor JEAN ROBERTS, Chief, Medical Statistics Branch LINCOLN I. OLIVER, Chief, Psychological Statistics Branch Vital and Health Statistics-Series 2-No. 47 DHEW Publication No. (HSM) 72-1047 Library of Congress Catalog Card Number 70-169281 ACKNOWLEDGMENTS We wish to thank Superintendent Raymond E. Berry; Former Superin- tendent Bruce Miller; Dr. Mabel Purl, Director of Research and Testing; Dr. Albert Marley, Director of Pupil Personnel Services, and all the other persons on the staff of the Riverside, California, Unified School District who made this testing program possible. In addition, we wish to thank the Socio- Behavioral Study Center for Mental Retardation at Pacific State Hospital, Pomona, California, and its Director, Dr. Richard Eyman, for use of the computer in making some of the calculations in this report, and Dr. A. B. Silverstein for his critical comments and assistance. We wish to thank Mr. Morton Saske who served as supervisor for the psychometric testing and the following psychometrists who administered tests for the project: F. L. Alvarez, R. C. Elliott, R. W. Ennis, N. M. Hansen, D. S. Harris, Jr., A. L. Hartry, M. F. Lane, E. M. Neumann, E. G. Simpson, and E. F. Williams. Dr. Lois Chatham, at that time Psychological Advisor of the Division of Health Examination Statistics, directed the psychometric training session which prepared the psychometrists for the fieldwork. FOREWORD The Health Examination Survey (IIES) is one of the major continuing programs of the Na- tional Center for Health Statistics, an agency authorized by Congress to provide statistical information on the amount, distribution, and effects of illness and disability in the United States. The collection, analysis, and publication of data obtainable only through direct examina- tion of people is the particular task of the HES. Examination programs for national samples of segments of our population began in November 1959 with a survey of adults between the ages of 18 and 79 (designated Cycle I). Mobile examina- tion centers with their teams of specialists began traveling throughout the United States, setting up in diverse locations to examine individuals selected in the national probability sample. The basic pattern of operation has continued through successive surveys and has included examinations of a sample of children 6-11 years of age (Cycle II) and of adolescents 12-17 years of age (Cycle III). While the initial effort in the adult examina- tion program was devoted primarily to obtaining information on several prevalent chronic dis- eases, when attention was directed toward younger age groups, the concern logically shifted to factors related to growth and development. At this point it became obvious that social and personal adjustment in the context of school and home is an integral part of healthy growth. Health problems of the developmental years are primarily those of retarded and disrupted growth, and the nature of personality develop- ment, as evidenced in acquisition of communica- tion skills, general mental abilities, and interper- sonal relationships, must be considered in assessment procedures. Because time and physical limitations must inevitably be imposed on a comprehensive health survey, no one health factor—whether dental, physiological, physical, or psychologi- cal —can be evaluated as thoroughly as it would be in a typical clinical or research setting. As a case in point, sound, widely accepted, brief tests of the psychological factors found to be impor- tant to the goals of the survey did not exist. To cover the necessarily broad arca, it was decided that the battery should be composed of either the briefest tests available or abbreviated and specially administered versions of widely used psychological instruments. The resulting battery, used in the children’s survey and continued into the adolescent’s survey, reflects the more fre- quent decision to use parts of longer tests and special administration procedures. Incumbent on the user of abbreviated tests is the need to conduct methodological studies to determine relationships between the new form and the original established instrument or other criterion measures or both. In the case of psychological data, the National Center for Health Statistics has attempted to fulfill this obligation primarily through contracts with several scientists. The study reported here is the result of one such contract. This report was written under contract with the National Center for Health Statistics, Public Health Service Grant #PH 43-67-756. The report does not deal with the issue of the validity of the WISC in measuring the intelligence of children from various socioeconomic levels and cthnic groups. The results from this study provide a means of estimating the Full Scale 1Q level of children aged 6-11 years examined in the Cycle II Health Examination Survey for the Wechsler Intelli- gence Scale for Children (WISC) based on the Vocabulary and Block Design subtests of the WISC which were used in Cycle 11 of the Health [.xamination Survey. Harold J. Dupuy, Ph.D. Psychological Advisor Division of Health Examination Statistics CONTENTS Acknowledgments . . . ... LL. LLL Foreword . . . . . . . Introduction . . . . . . . . LL. eee ee ee StUAY DESIGN. + + « » v0 5 2 + 5 5 50 8 9 5 5% 648 Bt wth. wy SAMPIE ov sv x vt kT as hE ae hd sree ns es rete wr Testing Procedures. . . . .............. Relationships of Age, Sex, Socioeconomic Status, and Ethnic Group 2-3 1 J I ET EL TLE EY EEE EE an AngloChildren . . . . . . LLL Age Le ee eee Sex Within Socioeconomic Status . . . . . ................. Intercorrelation of Scores for Children Aged 7and 10 . . . ........ Predicting Full Scale, Verbal, and Performance 1Q’s From Vocabulary and Block Design Scaled Scores . . . « + os 5s 4 6 sss x 62% vw 2a » The Percentage of Error in Predicting Subnormal IQ Using Tree Criteria « « « vv «vt 2 6 24 8 02 vw ve wwennensmusrns Optimal Prediction From Various Combinations of Subtests. . . . . . . . . Optimal Predictions Compared With Predictions Based on Vocabulary and Block Design « : « ov 4 5 vs vv si sins bam DISCUSSION « » o ¢ ww 2 4 © 3 45 20s wv 6u wu vn vs mmsnmsrsa Conclusions . . . . . .. Lee NegroCHlBIEn » + « « vs « s+ 5 x + 3 5 6 +8 5 54 5 8055 0raxemon J Sex Within Socioeconomic Status . . . . . . ................ Intercorrelation of Scores for Children Aged 7 and 10 . . . ........ Predicting Full Scale, Verbal, and Performance 1Q’s From Vocabulary and Block Design Scaled Scores . . . . . . ................ Percentage of Error in Predicting Subnormal 1Q Using Three Crlleria « + + us 35 5 50 5 5 2 4 «vv a vs TEE Optimal Prediction From Various Combinations of Subtests . . . . . . . . Optimal Predictions Compared With Predictions Based on Vocabulary and Block Design . . . . . ..... DISCUSSION . ow 0 vv aw v0 0 vs 2 55 +5 54 0 8404 80d 0 mroenma Conclusions , o « « « «0 0 0s 4 5 6 tv ns wnnmunronsme nas Page iil iv vi CONTENTS—Con. Mexican-American Children . . ©... . o.oo ooo L000 ATE & worm 0 vt 8 5 1a ¥ oe ew mre mom wm smo om amas SR Sex Within Socioeconomic S1atus . . « « + « x vc v5 5 «9% «vo » wx wu = Intercorrelations of Scores for Children Aged 7and 10 . . . . . . . .. .. Predicting Full Scale, Verbal, and Performance 1Q’s From Vocabulary and Block Design ScaledScores . . . . . . ................ Percentage of Error in Predicting Subnormal IQ Using Three Criteria . . . . : 2 2 ic 53 6 0 % 5 & 5 5 # 2 9+ 8 © 65% » 3 i» 3 Optimal Prediction From Various Combinations of Subtests . . . . . . .. Optimal Predictions Compared With Predictions Based on Vocabulary and Block Design. . . . . LL. LLL Discussion . oo +o « » 5 5 53 2 % 8 8 55 «2 #2 2 108 v1 282% Conclusions . . . » «v2 os +0 3 3 4% BW 5 5 55 % 5 9% 885 6% 353 0 General Summary and Conclusions... . . 000000 Bibliography . . . . + « vw cw 5 28 5 2 3 2 BE 5 « m8 26 55 3 5 8s 8 89s Appendix. AngloChildren . . . + 2 + «4 2 2 cv vs 0 wt tv we w wee Table IL. Intercorrelations of Tests in the WISC for Anglo Children Aged7 ....503 8853 58s Er AMEE EBL EEN DE Table II. Intercorrelations of Tests in the WISC for Anglo Children Aged lO oi ih hice mse er em ems aE RE Table III. Estimated Full Scale IQ from Block Design and Vocabulary Scaled Scores for Anglo Children Aged 6-11 . . . . . . .. .. Appendix Il. NegroChildren . . . + . + + 5 cts sso nm vs 2 25 1 5 8+ 2 Table IV. Intercorrelations of Tests in the WISC for Negro Children AEdT oe ee ee ae rea re hee EE Wa Table V. Intercorrelations of Tests in the WISC for Negro Children Aged 10 . oo vi sw vn rs ure mr eee ra ws Table VI. Estimated Full Scale IQ from Block Design and Vocabulary Scaled Scores for Negro Children Aged 6-11 . . . . . . .. .. Appendix 111. Mexican-American Children + . . 2 « + ov vv 5 vv sv 20 «vw Table VII. Intercorrelations of Tests in the WISC for Mexican-American Children Aged? . . vo vt havo ans @ ns mE md ®w sm Table VIII. Intercorrelations of Tests in the WISC for Mexican-American Children Aged 10 ; ; «2 + 5 5 + 2s 2 2 «2 v8 90 0 0 2 + a Table 1X. Estimated Full Scale IQ from Block Design and Vocabulary Scaled Scores for Mexican-American Children Aged 6-11 . . . . CONTENTS—Con. Page Appendix IV. All Sample Children . . . ................... 41 Table X. Estimated Full Scale IQ from Block Design and Vocabulary Scaled Scores for 1,310 Children Aged 6-11. . . . . ...... 41 Table XI. Estimated Full Scale IQ Using the Sum of the Scaled Scores for Block Design and Vocabulary for All 1,310 Children in the Sample and for Males and Females Separately . . ....... 42 Table XII. Percentage of Correct and Incorrect Predictions of Low Full Scale IQs Using Two Methods of Predicting from Block Design and Vocabulary Scaled Scores for 1,310 Children Aged 6-11 Using Three Different Criteria . . . . ............. 42 SYMBOLS Data not available-----ee-ennmooooemmeeeL ' Category not applicable---------eeeoeeeeee Quantity zero Quantity more than 0 but less than 0.05----- 0.0 Figure does not meet standards of reliability or precision-----------e-oeeeeee cena vii SUBTEST ESTIMATES OF THE WISC FULL SCALE IQ’S FOR CHILDREN Jane R. Mercer, Ph. D., and Joyce M. Smith? INTRODUCTION In Cycle II of the Health Examination Survey of noninstitutionalized children of the United States aged 6-11, the Vocabulary and Block Design subtests of the Wechsler Intelligence Scale for Children (WISC) were administered. The purpose of this report, written under contract with the National Center for Health Statistics, is to evaluate the use of these two subtests as the basis for estimating the Full Scale IQ’s of children aged 6-11 from various socio- economic levels and ethnic groups and to deter- mine the amount and direction of error likely to occur if these tests are used to estimate the rate of subnormal intelligence in these populations. The report does not deal with the validity of the WISC 1Q’s as a measure for children from socioculturally nonmodal backgrounds. There are various methods for estimating Full Scale 1Q’s from subtests: prorating the sum of the scaled scores, simple regression, and multiple regression. Silverstein (1967a) used all three methods to predict Full Scale IQ’s from a short form consisting of Vocabulary and Block De- sign. He found that while the error associated with proration always exceeded that of either simple or multiple regression, the improvement in prediction was relatively small (Silverstein, 1967d). Simple and multiple regression were the methods used in the present study. Dr, Mercer is Associate Professor, Sociology, and J. M. Smith is Research Psychologist, both at the University of Califor- nia, Riverside. STUDY DESIGN Sample The sample consisted of 1,310 children aged 6-11 attending public elementary schools in Riverside, California, during the school year 1967-68. These children were from three different ethnic groups: 505 were English-speaking Cau- casians (hereafter called ‘“Anglo”); 487 were of Mexican-American heritage; and 318 were Negro. The Mexican-American and Negro children included all the children aged 6-11 who attended three de facto segregated elementary schools prior to September 1966. The Anglo children were randomly selected from the student popu- lations of 11 elementary schools which were predominantly Anglo prior to comprehensive school desegregation which began in September 1966. Of the total sample, 1,270 were enrolled in regular classes and 40 were enrolled in classes for the educable mentally retarded. Table 1 presents the age, sex, and socio- economic status of the children. Socioeconomic status is based on the occupation of the head of the household in which the child was living. Occupations were categorized into three levels, using the Duncan Socioeconomic Index (Reiss, 1961). Low-status occupations are those rated 0-29 on the Index; middle-status occupations are those rated 30-69; and high-status occupations are those rated 70 or higher. Although the children in the sample were not Table 1. Distribution of the 1,310 sample schoolchildren, by sex, age, socioeconomic status, and ethnic group Socioeconomic status’ Sex and age . . : : : . Total Low Middle High Total Low Middle High Total Low Middle | High (0-29) | (30-69) | (70+) (0-29) | (30-69) | (70+) (0-29) | (30-69) | (70+) Both sexes Anglo Mexican-American Negro Total . . . . .. 505 a5 278 132 487 || 413 74 0 318 || 240 62 16 Percent each so- cial status . . 100 18.8 55.0 26.1 100 84.8 1541 0 100 75.5 195 5.0 Boys Total . uw « « + » 264 56 144 64 241 205 36 0 156 125 24 7 Gyears ...... 28 9 14 5 29 24 5 0 20 15 4 1 7 years . .. ... 62 13 36 13 53 43 10 0 31 20 8 3 8years . ..... 53 13 24 16 41 32 9 0 30 26 4 0 Ovyears ...... 37 7 19 1 44 39 5 0 26 24 0 2 10years : « : w » = 42 7 27 8 39 35 4 0 25 20 4 1 11vyears ...... 42 7 24 1 35 32 3 0 24 20 4 0 Girls Total . .. « 241 39 134 68 246 || 208 38 0 162 115 38 9 6years ...... 32 9 17 6 31 27 4 0 21 15 4 2 7vyears ...... 38 1 22 15 43 34 9 0 33 22 10 1 8vyears . ..... 47 5 2 15 34 31 3 0 28 24 3 1 9vyears . ..... 45 5 23 17 85 44 1 0 38 24 12 2 10years . . . . .. 48 10 30 8 49 42 7 0 23 19 3 1 11years . . .... 31 9 15 7 34 30 4 0 19 1 6 2 I Socioeconomic status was classified using the Duncan Socioeconomic Index. 0-29, low status; 30-69, middle status; and 70+, high status. a random selection from the general population of the city, the distribution of socioeconomic statuses within each ethnic group for the sample approximates the socioeconomic distribution for cach ethnic group in the city of Riverside. A household survey of a stratified area probability sample of 2,661 housing units conducted 3 years prior to the testing found that 25 percent of the Anglos lived in low-status families, 54 percent in middle-status families, and 22 percent in high- status families when the same classification categories on the Duncan Socioeconomic Index were used as those presented in table 1. Thus, there were relatively more Anglo children in the sample from high-status families and fewer from Occupation of the head of the household score: low-status families than there were in the River- side population. In the same survey, 78 percent of the Mexican-Americans in the city were low status, 20 percent were middle status, and 2 percent were high status, while 74 percent of the Negroes were low status, 17 percent were middle status, and 5 percent were high status. Thus, the sample percentages also approximate the per- centages of children from these ethnic groups in the three status levels in Riverside. Relatively few Mexican-American and Negro families in Riverside had a head of the household with an occupation rated 70 or higher on the Index, and there were no Mexican-American children and only 16 Negro children in the sample from such families. Therefore, no separate analyses of the performance of high-status Mexican-American and Negro children were made. Testing Procedures Each child was tested during the regular school day. Because some school buildings did not have testing rooms available that were quiet and undisturbed, trailers were rented and moved from campus to campus. Some trailers contained four testing rooms and others contained a single room, depending upon the number of children to be tested at a particular school. Field testing took place from February 15 to June 15, 1967. Psychometrists were recruited through profes- sional organizations, the University of California personnel office, and contacts with college campuses in the vicinity of Riverside. The 11 psychometrists who conducted the testing had all been trained to administer the WISC in regular college courses in psychometric testing. They were supervised by a school psychologist certified by the State of California. An intensive 3-day psychometric training session was con- ducted to assure that the Vocabulary and Block Design subtests would be administered and scored in a fashion identical to that used in Cycle II of the Health Examination Survey. Psychometrists were trained to use standard pronunciations of the words in the Vocabulary test following the “Pronunciation Guide” used in Cycle II, and test administration and scoring were monitored by the certified school psychol- ogist throughout the fieldwork in accordance with the guidelines developed for Cycle II. Information on the age, grade, and ethnic group of each child was secured from school records. Each child’s age was verified in an interview with his parents. Information about the occupation of the head of the household in which each child was living was also obtained during the parent interview. All data were keypunched, checked for internal consistency, and stored on magnetic tape. brhe parent interview was conducted as part of the Riverside Desegregation Study supported by the California State Depart- ment of Education, McAteer M6-14. Relationships of Age, Sex, Socioeconomic Status, and Ethnic Group to 1Q In order to determine which of the four characteristics being investigated was most highly correlated with IQ, three stepwise multi- ple regressions were run using Full Scale 1Q, Verbal 1Q, and Performance IQ as the dependent variables and age, sex, socioeconomic status, and ethnic group as independent variables. Ethnic group was dichotomized, Anglo vs. Mexican- American and Negro; socioeconomic status in- dex scores were dichotomized, 0-29 vs. 30+; and age was dichotomized, 6-8 vs. 9-11. Table 2 presents the results of these analyses. Ethnic group was the single best predictor of IQ. It correlated .428, .407, and .328 with Full Scale, Verbal, and Performance 1Q, respectively. Individual correlations between socioeconomic status and IQ) were almost as high: .377, .344, and .290. Together, ethnic group and socio- economic status correlated .457 with Full Scale IQ, .428 with Verbal IQ, and .350 with Perform- ance 1Q. All multiple correlations were statisti- cally significant beyond the .01 level of proba- bility. Knowledge of sex and age added very little to the accuracy of prediction of IQ after the effect of ethnic group and socioeconomic status were taken into account. The primary relationship of age with intellectual development was taken into account by the conversion of raw test scores to age-specific scaled scores at 4- month age intervals. It was concluded that the three ethnic groups should be analyzed sepa- rately, looking at socioeconomic levels within an ethnic group. ANGLO CHILDREN Age Although the sample of Anglo children was a random selection from the elementary school population of 11 elementary schools, the older children had a higher mean Full Scale and Verbal IQ than the younger children. This was not true for the Performance IQ. Table 3 pre- sents the mean scores, standard deviations, and F ratios. There were 34 children in the Anglo sample with a Full Scale IQ of 84 and below; 39 with a Table 2. Stepwise multiple regressions with WISC 1Q’s as the dependent variables and age, sex, ethnic group, and socioeconomic status as independent variables, 1,310 sample schoolchildren Full Scale 1Q Verbal 1Q Performance 1Q Independent variables rt R? rt R? rt R? Ehnic group » » + « + » 23 + 5 5 # 3.428 3.428 3.407 3.407 3.328 3.328 Socioeconomic status . . . . . . . . . 3 377 3 457 3.344 3428 3.290 3 350 BER «vv mwmom wn nm nm EWE 4.067 3.461 3.081 3.435 .034 3.351 Age . Lo. ee eee .002 3.462 .029 ¥ 437 .005 3.352 'r: Zero order product moment correlation coefficients. 2R: Multiple correlation coefficients. 3 Significant at .01 level. 4 Significant at .05 level. Table 3. Mean 1Q’s, standard deviations, and F ratios for Anglo children, by age Age in years Mean 1Q and F rati standard deviation 6 7 8 9 10 11 2 (N=60) (N=100) (N=100) (N=82) (N=90) (N=73) Full Scale 1Q MBAR . . v0: 5: sms. 101.9 104.3 105.1 104.9 109.8 107.5 2.57 BOD vv ov won wots ko 10.4 15.2 16.0 18.3 15.2 15.5 p <.02 Verbal 1Q Meal . . «+458 9s 58 82 + 98.4 102.4 103.5 102.5 108.3 105.2 3.71 BD hr mr mah s 48 EE 11.3 14.8 15.8 15.6 14.1 14.9 p <.01 Performance 1Q Mean wos ou ow oooa eons ww ow 104.7 106.1 106.3 106.7 110.7 108.7 1.56 21 1 10.5 15.7 16.3 15.1 16.3 16.1 NS Verbal 1Q of 84 and below; and 31 with a Performance 1Q of 84 and below. The children with low 1Q’s were evenly distributed through- out the six age groups. However, there were more children aged 10 and 11 with 1Q’s above 125 than there were at other ages. This accounts for the higher average IQ of the older children in the sample. Because of these age differences, subsequent analyses will study children’s per- formance by age group. Sex Within Socioeconomic Status The 1Q’s and subtest scores of Anglo girls were compared with those of Anglo boys within the three socioeconomic levels. Table 4 presents the mean scores for cach sex, and the values of ¢ when the means were compared. Four of the 42 sex comparisons were signifi- cant at the .05 level and four were significant at the .01 level, slightly more than would be expected by chance. Four of the significant differences were for low-status children, and in cach case, boys scored significantly higher than girls. There was no pattern to the other differ- ences except that middle- and high-status girls were significantly higher on Coding (p <.01). Differences across socioeconomic levels, how- ever, were large and consistent. The mean Full Scale, Verbal, and Performance 1Q’s for cach sex Table 4. Mean 1Q’s and subtest scores for Anglo children, by socioeconomic status and sex and mean 1Q’s across socioeconomic status Low status Middle status High status Gomparison across socioeconomic status Mean 1Q and subtest Boys Girls i Boys Girls 7 Boys Girls 7 as a (N=56) | (N=39) (N=144) | (N=134) (N=64) | (N=68) J B24) Mean 1Q FullScale IQ ., . vu ss 504 » 100.0 95.2 1.5 105.6 105.3 | 109.7 110.0 2 16.55 113.36 Verbal IQ . .......... 98.1 920 | 22.0 103.7 102.6 6 108.8 107.2 8 17.74 1156.36 Performance 1Q . ....... 102.1 99.5 8 106.6 107.2 3 108.7 111.4 1.2 23.12 17.75 Verbal subtests Information . ......... 9.3 8.3 1.7 10.8 10.2 1.4 11.7 11.3 7 18.35 112.57 Comprehension . . ...... 9.5 75 | '34 9.8 9.4 1.1 10.6 9.6 1.8 1.80 16.53 Arithmetic . . : ws cs a9 55 10.0 9.8 3 10.9 10.7 5 11.9 11.6 8 15.68 15.12 Similarities . . . ........ 10.1 9.7 6 11.2 11.7 1.2 1.7 1.7 JA 23.76 16.39 Vocabulary ©. . ......... 10.5 88 | 224 11.2 10.8 1a 12.8 11.7 22.2 17.26 110.37 Digitspan . . . ........ 8.8 8.4 6 9.4 9.7 7 9.5 108 | 225 1.05 18.42 Performance subtests Picture Completion . . . ... 10.7 8.6 | '3.3 10.8 10.2 15 10.4 10.3 2 32 24.43 Picture Arrangement . . . . . . 10.6 9.6 1.5 11.0 11.2 6 11.0 11.9 1.7 .40 '6.74 Block Design . . ........ 10.4 9.7 9 11.5 11.0 1.3 124 11.6 1.6 16.25 14.72 Object Assembly . . . . .... 10.2 10.9 1.2 11.1 11.0 3 11.9 11.7 5 15.49 1.40 Coding o win +s wm ss 0% 4 » 9.6 10.7 1.5 10.3 116 | '35 10.5 125] '4.4 1.56 24.06 ! Significant at .01 level. 2Significant at .05 level. rose consistently from low status to high status, and all increases were statistically significant. Four of the means of the verbal subtests increased significantly with socioeconomic level for boys and all six verbal subtests increased significantly for girls. The performance subtests were not as influenced by socioeconomic status as the verbal subtests. Only two of the increases were significant for boys, although four of the increases were significant for girls. It is clear that socioeconomic level and IQ are related. In general, scores on the WISC were positively correlated with socioeconomic level. This was especially true for verbal abilities. Therefore, we conclude that socioeconomic dif- ferences as well as ethnic and age differences should be held constant when analyzing the efficiency of using various combinations of subtests as short forms of the WISC for Anglo children. Intercorrelation of Scores for Children Aged 7 and 10 The WISC Manual contains tables presenting the intercorrelations of subtest scores with each other and with Full Scale, Verbal, and Perform- ance scaled scores for children aged 7% and 10% on whom the test was normed (Wechsler, 1949). Tables I and II in appendix I present the comparable intercorrelations for Anglo children aged 7 and 10 in the present study. When the z test of the significance of difference between r’s (Guilford, 1965) was used to test the difference between each correlation and its counterpart in the WISC Manual, there were four correlations in the 89 comparisons for children aged 7 in which the correlation in the Anglo sample was significantly lower (p <.05) than the compa- rable correlation for the 200 children on whom the test was normed. There was one instance in which the correlation for the Anglo sample was significantly higher than that for the children in the standardization sample. There was no pat- tern to the differences. When the correlations tor children at age 10% on whom the test was normed were compared with the correlations for the 10-year-old Anglo children in the present sample, there were many more significant differences than for the 7-year- old children. There were 26 correlations for the Anglo children in the present sample which were significantly lower (p <.05) than the compara- ble correlations in the standardization sample. Only two correlations were significantly higher in the Anglo sample than in the standardization sample. Wechsler did not publish the intercorre- lations for children at ages 6, 8, 9, and 11; therefore, no comparisons were possible. In general, it appears that the intercorrela- tions of subtests scores with each other and with Full Scale, Verbal and Performance 1Q’s are lower for the present Anglo sample than for the children on whom the test was originally normed. Predicting Full Scale, Verbal, and Performance IQ's from Vocabulary and Block Design Scaled Scores In Cycle IT of the Health Examination Survey, only the Block Design and Vocabulary subtests of the WISC were administered. Stepwise multi- ple regressions were performed with Full Scale, Verbal, and Performance 1Q’s as the dependent variables. The scaled scores for Block Design and Vocabulary were independent variables, within the three socioeconomic levels described earlier, yielding multiple correlations ranging from .642 to .883 (table 5). The multiple correlations with Full Scale 1Q were highest at all status levels. The multiple correlations with Verbal 1Q ranked second and those with Per- formance IQ ranked third. The multiple correla- CThe convention in many short-form studies is to use the sum of scaled scores on the short form to predict the sum of scaled scores on the Full Scale rather than predicting the Full Scale 1Q directly. Inasmuch as the sum of scaled scores on the Full Scale is perfectly correlated with the Full Scale IQ, corre- lations with the scaled scores of the short form are the same in either case. tions for low-status and middle-status children were of similar magnitude, but those for high- status children tended to be lower. As would be expected, standard errors were smallest when predicting Full Scale 1Q. They increased when predicting Verbal 1Q, and were largest when predicting Performance 1Q. If an investigator was primarily interested in the Verbal 1Q and Performance 1Q, he would use two verbal subtests and two performance subtests, respec- tively, as the basis for prediction. Multiple correlations are also presented using Block Design and Vocabulary as independent variables and Full Scale, Verbal, and Perform- ance 1Q’s as the dependent variables within three age groups: 6 and 7 years; 8 and 9 years; and 10 and 11 years. The multiple correlations were of approximately the same magnitude as those obtained within socioeconomic level. As in the case of multiple correlations within socioeconomic levels, correlations within age groups were slightly higher for Full Scale 1Q, followed by those for Verbal 1Q), and then those for Performance 1Q. Since R is inversely related to the standard error, standard errors were slightly smaller when predicting Full Scale 1Q than when predicting Verbal 1Q; they were highest when predicting Performance 1Q. When the entire group of Anglo children was analyzed without regard for age or socioeco- nomic status, the multiple correlation of the two subtests was .867 with Full Scale 1Q, .839 with Verbal IQ, and .798 with Performance 1Q. Standard errors were 7.50, 8.15, and 9.19 1Q points, respectively. Thus, prediction is not markedly improved by using subgroups of Anglo children based on social status or age. When the beta coefficients and constant terms presented in table 5 were inserted in the multi- ple regression equation with an individual child’s subtest scaled scores, the solution yielded the best prediction of IQ for that child. For ex- ample, to predict the Full Scale IQ of a low status child: Predicted 1Q = constant term, 49.82 + (2.62 X Vocabulary Scaled Score) + (2.22 X Block Design Scaled Score) Approximately two-thirds of the predicted 1Q’s will lic within one standard error, 7.50 1Q points, of the actual 1Q. Table C in appendix | Table 5. Full Scale, Verbal, and Performance 1Q’s from scaled scores for Anglo children, by socioeconomic status and age, and beta coefficients Full Scale 1Q Verbal 1Q Performance 1Q Status, subtest, and age r R Bes r R Bets r R Beta coefficient coefficient coefficient STATUS Low (N=95) Vocabulary . ........... 778 778 262 | .825 .825 3.21 .591 '(2).832 1.47 Block Design. . . ........ .700 883 2.22 | 519 .849 1.03 | .776 (1).776 3.15 Constant Term. . . . .. .... 49.82 53.64 54.76 Standard errorof Y . . . . .. .. 7.55 8.36 9.02 Middle (N=278) Vocabulary . ........... .787 .787 2.46 | .834 .834 3.09| 614 | '(2).819 1.29 Block Design. . . ........ 732 .875 2.19 | .580 .853 98 | .780 (1).780 3.1 Constant Term. . . . ...... 63.72 58.04 57.72 Standard errorof ¥Y . . ...... 7.56 7.87 9.02 High (N=132) Vocabulary . ........... .594 .594 2.00 | .707 .707 2.88 | .231 | '(2).642 .68 Block Design. . . . ....... 524 .749 1.76 | .227 721 63 | .622 (1).622 2.68 Constant Term . . ........ 64.27 65.09 69.53 Standard errorof Y . . ...... 7.12 8.39 9.56 AGE 6-11 years (N=505) Vocabulary oo. + 0 «50 5 wows + .765 .765 245 | .819 .819 3.14 | 558 | '(2).798 1.23 Block Design. . . . ....... .701 867 2.15 | .623 .839 97 | .758 (1).758 3.06 Constant Term . . ........ 53.77 57.22 58.56 Standard errorof ¥Y . . .. .... 7.50 8.15 9.19 6-7 years (N=160) Vocabulary . ........... 734 734 2.07 | .790 .790 2.70 | 522 | '(2).808 .99 Block Design. . . . ....... .698 .857 2.27 | .496 .814 1.02 | .774 (1).774 3.23 Constant Term. . .. ...... 56.77 60.97 60.06 Standard errorof ¥Y . . . . .... 7.08 8.02 8.25 8-9 years (N=182) Vocabulary . . ov v5 + wes ws .778 778 2.83 | .830 .830 3.43] .580 ( '(2).801 1.61 Block Design. . . ........ .640 .868 1.97 | 461 .848 88 | .726 (1).726 2.81 Constant Term. . . . ...... 50.42 53.98 55.84 Standard errorof Y . . . ... .. 7.82 8.34 9.28 10-11 years (N=163) Vocabulary . ........... 776 .776 2.35 | .833 .833 3.23 | 661 | '(2).796 .96 Block Design. . . ........ .768 880 2.30 | .602 .852 971 779 (1).779 3.33 Constant Term . . ........ 54.40 57.58 59.30 Standard errorof Y . . .. .. .. 7.25 7.80 9.76 ! The order of the subtests reverses for the Performance 1Q because Block Design, being a performance test, always has a higher linear correlation with Performance 1Q and independently accounts for more of the variance in Performance 1Q than does Vocabulary. presents predicted Full Scale 1Q’s for Anglo children with various combinations of Vocabu- lary and Block Design scaled scores. These multiple regressions and others pre- sented in the report were probably slightly inflated because correlations between scores on a short form and Full Scale 1Q’s obtained from the same administration of the test tended to be spuriously high. Such a correlation assumes that the subtests are perfectly reliable when, in fact, they are not. When Silverstein (1970a,b) used a corrected formula to calculate simple linear correlations between the sum of the scaled scores for Vocabulary and Block Design and Full Scale IQ, the standard error of estimate rose from 7.2 IQ points using the uncorrected esti- mate (Silverstein, 1967 a and ¢) to 7.8 1Q points using the corrected estimate (Silverstein, 1970 a). The correlation coefficient dropped from .879 to .856 (Silverstein, 1970 b). Percentage of Error in Predicting Subnormal 1Q Using Three Criteria The appropriate beta coefficients for predict- ing Full Scale IQ from scaled scores on Vocabu- lary and Block Design were inserted in the multiple regression equations for each age and socioeconomic group and used to predict the Full Scale IQ of each Anglo child in the sample. Each child’s actual Full Scale IQ was then compared with his predicted Full Scale 1Q. The percentage of correct and incorrect classifica- tions for subnormal IQ using three criteria (IQ below 85; below 80; and below 70) was calcu- lated for children in each socioeconomic group. There were no IQ’s below 85 in the high-status group; therefore, there are no estimates of the percentage of correct or incorrect predictions of low IQ for high-status children in table 6. If children with 1Q’s below 85 are regarded as intellectually subnormal, there were 10.5 per- cent of the 95 low-status Anglo children who were correctly identified as intellectually sub- normal. There were 5.3 percent who were identified as having low 1Q’s who actually had 1Q’s of 85 and above, and 3.2 percent who were identified as having 1Q’s of 85 and above who actually had I1Q’s below 85. Other percentages in table 6 indicate the percentage of children who were correctly or incorrectly classified using 79- and 69- as the criteria for subnormal 1Q. 8 The percentage of “false low” predictions tended to balance the percentage of “false high” predictions. Therefore, the aggregate prediction for low IQ’s was closer to the actual percentage in the population than the predictions for individuals. For example, 13.7 percent of the low-status children actually had IQ’s below 85 and 15.8 percent were predicted to have 1Q’s that low. Thus, 2.1 percent more children were predicted to have low 1Q’s than actually had low IQ’s. Some of the other differences were smaller. For middle-status children, the ‘false lows” equal the “false highs” at the 84- and 79- criteria. Table 6 also presents the percentage of individuals in each age group correctly and incorrectly identified as having low Full Scale IQ’s using the beta coefficients and constant terms appearing in table 5 as the basis for prediction. Most of the errors were underpredic- tions, ranging from -2.5 percent when predict- ing the percentage of children aged 6 and 7 with IQ’s below 85 to +.6 percent when predicting the percentage of children aged 10 and 11 with IQ’s below 85. Differences between age levels were negligible. When 1Q’s were predicted for all the Anglo children without regard for socioeconomic sta- tus or age, the differences between the actual percentage of children with low IQ’s and the estimated percentage of children with low 1Q’s was relatively low. The estimate was .2 percent high when the IQ 84- criteria was used and .4 percent and 1.8 percent low when the 79- and 69- criteria were used. However, errors in the placement of individual children were consider- ably higher—2.4 percent of the children were falsely predicted to have IQ’s below 85, and 2.2 percent were falsely predicted to have I1Q’s of 85 and above for a total misplacement of 4.6 per- cent of the individual children. The total mis- placement was 1.6 percent and 2.2 percent at the IQ 79- and 69- criteria, respectively. Although multiple correlation coefficients ranging from .622 to .883 (table 5) were rela- tively high, there was still considerable error in predicting Full Scale 1Q when only the Block Design and Vocabulary subtests were used. In predicting the aggregate percentage of children who will fall below the three criteria most commonly used by clinicians in diagnosis, the error in prediction ranged from -1.8 to +0.2 Table 6. Percent of correct and incorrect predictions of low Full Scale 1Q’s for Anglo children, by socioeconomic status’ and age using three different criteria Correctly Correctly Actual Predicted Status, 1Q, A ok False False A oo } hd BE identified low 1Q’s? | high 1Q's> identified | percent percent | Difference 9 g low 1Q’s 9 high 1Q's low 1Q’s low IQ's STATUS Percent Low (N=95) OBA . iv sms msm ims dk hu imsm 10.5 5.3 32 81.1 13.7 15.8 +2. 1Q79- . . . oe eee 8.4 13 24 88.4 10.5 9.5 -1.0 MOBO=~ ov + vs ws ww awa em rm 5.3 0.0 21 92.6 7.4 53 -2.1 Middle (N=278) 1Q84— . . . eee 5.4 22 2.2 90.3 7.6 7.6 0.0 1IQ79- . «eee 3.6 1.2 1.1 94.2 4.7 4.7 0.0 1I069~ . . 5. ssn same wswsms ws 0.0 4 1.8 97.8 1.8 4 -1.4 AGE All ages, 6-11 years (N=505) 1084 . . + iv so vase mE Rs FEE 4.6 24 22 90.8 6.8 7.0 + 2 O79 vv ete mtn mE EEE 3.6 6 1.0 94.8 4.6 4.2 - 4 1Q69- . . . «ee 4 2 2.0 97.4 24 6 ~-1.8 6-7 years (N=160) 1Q84- . . . ee 38 1.3 3.8 21.3 7.6 5.1 -2.5 1Q79- . eee 25 1.3 13 95.0 3.8 3.8 0.0 VOBD= , 262 0 vs moms ms ms me mn 0.0 0.0 1.3 98.8 1.3 0.0 -1.3 8-9 years (N=182) 1Q84- . . . . eee 4.9 1.6 2.7 90.7 7.6 6.5 -1.1 1Q79- . oe ee ee 4.4 0.0 1.1 94.5 55 4.4 -1.1 BD= o . 0 5 55 2 5s 0 vw amsmsms 1.6 5 22 95.6 3.8 2.1 -1.7 10-11 years (N=163) IB ..... 0.08 imsm smn 4.3 1.2 6 93.9 4.9 5.5 + 6 1Q79- . «Le eee 25 0.0 1.8 95.7 4.3 25 -1.8 1Q69- . . . «oe ee eee 0.0 0.0 1.8 98.2 1.8 0.0 -1.8 ! There were no 1Q’s below 85 in the high-status group; therefore there were no estimates of the percentage of correct and incorrect predictions. 2 False low 1Q’'s were for children who were predicted to have an 1Q below the criteria but who actually had an IQ above the criteria. 3 False high 1Q’'s were for children who were predicted to have an IQ above the criteria but who actually had an 1Q below the criteria. percent for all children aged 6-11 in the sample with minimal error at the 1Q 84- and 1Q 79- cutoffs. Errors in the placement of individual children were even higher. Optimal Prediction From Various Combinations of Subtests Eleven subtests of the WISC were adminis- tered. The verbal subtests are Information, Comprehension, Arithmetic, Similarities, Vocab- ulary, and Digit Span. The performance subtests arc Picture Completion, Picture Arrangement, Block Design, Object Assembly, and Coding. In order to determine the optimal test battery for each socioeconomic level and age group, step- wise multiple regressions were run using Full Scale, Verbal, and Performance 1Q’s as the dependent variables and all 11 subtests as independent variables, Table 7 presents the multiple correlations between the subtests and the 1Q’s of Anglo children within socioeconomic levels and within age groups. The four subtests appearing first in the solution for cach socioeconomic group are listed for the Full Scale 1Q. The three subtests appearing first arc listed for the Verbal and Performance 1Q’s. For low- and high-status children, Vocabulary was the best single pre- dictor of Full Scale 1Q and Block Design appeared among the first four variables. Infor- mation replaced Vocabulary as the best single predictor for middle-status children, and Block Design did not appear for middle-status children when predicting ull Scale 1Q). Vocabulary was the best single predictor of Verbal 1Q for low- and middle-status children, but it did not appear in the first three subtests for high-status children. Block Design ranked second in predicting Performance 1Q for low- status children, but it did not appear for either middle- or high-status children. Standard errors ranged from 4.91 to 5.54 IQ points, but there was no pattern to the variation. Multiple regressions within age groups and for the total group yielded correlations and standard errors similar to those found within socioeco- nomic levels. All multiple correlations were above .90 when three variables were used to predict Verbal and Performance 1()’s and four variables were used to predict Full Scale IQ. Standard errors ranged between 4.56 and 5.58 IQ points. Vocabulary was the most important single variable predicting Full Scale 1Q for children aged 10 and 11 and was among the top three variables in predicting Verbal IQ for all age groups. Block Design ranked second in predict- ing Full Scale 1Q for children aged 6, 7, 10, and Il and ranked as the most important single variable in predicting Performance 1Q for the same two age groups. The particular subtests which appeared in the multiple regressions shown in table 7 varied considerably from one age group and socioeco- nomic level to another. Certain subtests were conspicuous by their absence. Digit Span did not appear in any analysis nor did Picture Arrange- ment. The other subtests appeared with about equal frequency. One consistent pattern was that a verbal subtest was the most highly correlated with Full Scale 1Q in cach instance— Vocabulary and Information shared the top rank equally. The second ranking subtest in all multiple correlations with Full Scale 1Q was consistently a performance subtest: Block De- sign appeared three times, Object Assembly twice, and Coding once. When Anglo children were treated as a single group, regardless of social status or age, Vocabu- lary and Block Design were the best two subtests for predicting Full Scale 1Q, Vocabulary was the best single predictor for Verbal 1Q, and Block Design was the best single predictor for Perform- ance 1Q). Optimal Predictions Compared With Predictions Based on Vocabulary and Block Design Table 8 compares the multiple correlations obtained using the two optimal subtest predic- tors with the multiple correlations obtained using Vocabulary and Block Design. For Full Scale 1Q, the optimal predictors were Vocabu- lary and Block Design for high-status children, for children aged 10-11, and for all Anglo children as a group. Differences were negligible for other groups. Differences between the optimal predictors and Vocabulary and Block Design were greater for Verbal IQ and Performance 1Q than for Full Scale 1Q. Although using both Vocabulary and Table 7. Optimal predictions of 1Q’s of Anglo children from various combinations of subtests, by socioeconomic status and age Full Scale 1Q Verbal 1Q Performance 1Q Status, subtest, Status, subtest, Status, subtest, a 308 Beta 81d BaC Beta and a Beta and'20 ¥ R coef- 2g r R coef- ge r R coef- ficient ficient ficient STATUS STATUS STATUS Low (N=95) Low (N=95) Low (N=95) Vocabulary . . . ..... 778 778 1.61 | Vocabulary . . ...... .825 .825 184: 1 Coding : cums smmy » .781 .781 221 Coding «i « vwn x www 719 899 1.59 | Arithmetic . ....... 778 917 1.67 | Block Design . . ..... .776 .900 2.21 Block Design. . . . ... .700 935 1.39 | Similarities . . ...... 814 951 1.64 | Picture Completion 615 949 1.61 Information . . . . .... 752 953 1.31 | Constant Term . . . . .. 44.66 Constant Term . . . . .. 40.40 Constant Term. . . . . . 40.48 Standard errorof Y . . . . 4.92 Standard errorof ¥ . . . . 5.10 Standard error of Y . . . . 4.91 Middle (N=278) Middle (N=278) Middle (N=278) Information . . . . .. .. .800 .800 1.81 Vocabulary oa: vow eos .834 .834 2.00 | Object Assembly . . . . . .781 .781 2.51 Object Assembly . . . . . 675 .886 1.75 | Arithmetic ........ 753 912 1.90 | Picture Completion .710 .882 1.97 Similarities . . . ..... .736 924 1.27 | Comprehension . . . . .. .703 .946 132 | Coding - - vu + vne 616 937 1.59 Arithmetic. . . ..... 712 942 1.20 | Constant Term . . . . .. 47.86 Constant Term. . . . . . 40.96 Constant Term. . . . . . 39.50 Standard errorof Y . . . . 4.92 Standard errorof Y¥ . . . . 5.51 Standard error of Y . . . . 5.28 High (N=132) High (N=132) High (N=132) Vocabulary . . coos 5 0 593 593 1.40 | Information . . . . .... 722 A22 1.87 | Object Assembly . . . .. 652 652 2.05 Block Design . . . .... .524 .749 1.45 | Similarities . . .. .... .686 .841 1.59 | Picture Completion .639 .815 1.72 Coding ous sum aww .436 .819 1.24 | Comprehension . . . . . . 633 .900 1.32 | Coding . ......... .584 .897 1.71 Information . . . . .. .. 572 874 1.28 | Constant Term . . . . .. 54.43 Constant Term. . . . . . 48.37 Constant Term. . . . .. 46.37 Standard errorof Y . . . . 5.31 Standard errorof ¥ . . . . 5.34 Standard errorof ¥ . . . . 5.26 AGE AGE AGE 6-11 years (N=505) 6-11 years (N=505) 6-11 years (N=505) Vocabulary .761 761 1.38 | Vocabulary . . ...... .819 819 1.87 | Block Design . . . . . .. .758 .758 2.26 Block Design .701 .867 1.60 | Arithmetic . ....... .750 .903 1.74 T Coding ws cw win vw ww .653 .861 1.82 Coding . ......... .602 907 1.21 Similarities . . . . . . .. 777 942 1.67 | Picture Completion 671 931 1.80 Information . . . ... .. .762 935 1.47 | Constant Term . . .. .. 45.68 Constant Term . . . . . . 42.74 Constant Term . . . . .. 43.35 Standard errorof Y . . . . 5.04 Standard error of ¥ . . . . 5.58 Standard errorof ¥ . . . . 5.34 6-7 years (N=160) 6-7 years (N=160) 6-7 years (N=160) Information . . . ... .. .758 .758 1.57 | Information . . . . .. .. .820 .820 1.65 | Block Design . . . .. .. 774 774 2.40 Block Design. . . . . .. .698 .884 1.70 | Vocabulary . . ...... .790 .897 1.47 | Picture Completion .569 .861 1.62 Similarities . . . . .. .. .649 918 133 | Arithmetic © . «55 5 «5 710 944 1.65 | Object Assembly . . . . . .706 921 1.77 Object Assembly . . . . . 611 939 1.06 | Constant Term . . . . .. 51.19 Constant Term. . . . .. 42.88 Constant Term . . . . .. 43.86 Standard errorof Y . . . . 4.56 Standard errorof Y . . . . 5.48 Standard error of Y . . . . 4.76 8-9 years (N=182) 8-9 years (N=182) 8-9 years (N=182) Information . . . . .... .805 .805 1.97 | Information . . . ..... .847 .847 2.11 | Object Assembly . . . .. .788 .788 2.50 Object Assembly . . . . . .695 .886 1.66 | Similarities . . .. .... .783 917 1.57 | Picture Completion 714 .894 2.01 Similarities . . . . .. .. 724 925 1.44 | Vocabulary . ....... .830 .947 165 | Coding . ......... .654 941 1.64 Picture Completion 631 947 1.16 | Constant Term . . . . .. 45.65 Constant Term. . . . . . 40.73 Constant Term . . . . .. 37.99 Standard errorof Y . . . . 5.08 Standard errorof Y . . . . 5.27 Standard error of Y . . . . 5.08 10-11 years (N=163) 10-11 years (N=163) 10-11 years (N=163) Vocabulary . . . ..... .776 776 1.80 | Vocabulary . ....... .833 .833 1.95 | Block Design . ...... 779 779 2.18 Block Design . . . .... .768 .880 1.52 | Arithmetic ........ 757 .909 1.25 | ‘Coding «ms « wun « an .686 .882 1.93 Coding. owms sews on 629 928 1.39 | Comprehension . . . . .. .750 .948 1.39 | Picture Completion .708 943 1.68 Picture Completion 661 950 99 | Constant Term . . .. .. 50.13 Constant Term . . . . . . 43.14 Constant Term . . . . .. 43.02 Standard errorof Y . . . . 4.79 Standard errorof Y . . . . 5.38 Standard errorof Y . . . . 4.80 1 Table 8. Comparison of Block Design and Vocabulary as predictors with the two optimal predictors for Full Scale, Verbal, and Performance 1Q’s for Anglo children, by socioeconomic status and age Full Scale 1Q Verbal |1Q Performance 1Q Status and age " R R Rk R Rh 2 9 Block Design Two Differ- | Block Design Two Differ- | Block Design Two Differ- and optimal ence and optimal ence and optimal ence Vocabulary predictors Vocabulary predictors Vocabulary predictors Status Low (N=9B} . ..... .883 899 016 .849 917 .069 .832 .900 .068 Middle (N=278) . . .. .875 .886 .011 .853 212 059 819 .882 .063 High (N=132) . . . . .. 749 749 .000 721 .841 120 642 B15 A373 Age Total, 6-11 years (N=505) . .. .867 .867 .000 .839 .903 .064 .798 .861 .063 6-7 years (N=160) . . . .857 .884 027 814 .897 .083 .808 .861 .053 8-9 years (N=182) . . . .868 .886 .018 .848 917 .069 .801 .894 .093 10-11 years (N=163) . . .880 .880 .000 .852 .909 .057 .796 .882 .086 Block Design to predict Verbal 1Q and Perform- ance 1Q was better than using either of them individually, the prediction was best if another verbal subtest was used with Vocabulary to predict Verbal IQ and another performance subtest was used with Block Design to predict Performance IQ). Since the Verbal 1Q was based on verbal subtest scores and the Performance 1Q was based on performance subtests, this result was to be expected. We conclude, therefore, that Vocabulary and Block Design were probably the two subtests which form the most efficient dyad for predict- ing the Full Scale IQ’s of Anglo children of various ages and sociocconomic levels. The correlations of Vocabulary and Block Design with Full Scale IQ closely approximated the correlations for the optimal combinations of any two subtests for the Anglo children. Discussion In the 20 years since the publication of the WISC, a number of studies have investigated the possibility of deriving selected short forms. Unfortunately, many of these studies have been conducted using special populations of children suffering from emotional and/or learning prob- lems; and most of them have studied various 12 combinations of three, four, or [ive subtests as predictors (Clements, 1965; Enburg, Rowley, and Stone, 1961; McKerracher and Watson, 1968; Nickols, J., and Nickols, M., 1963; and Thompson and Finley, 1963). Most correlations in these studies, expect for that of Yalowitz and Armstrong, were compara- ble in magnitude to those obtained in the present study when three or four subtests were used to predict Full Scale IQ. As in the present study, Nickols and Nickols found no major differences between subpopulations dichoto- mized on the basis of age, and Clements found no difference by sex in the validity of the subtest combinations he used. A short form composed of the two subtests (Vocabulary and Block Design) has been one of the most popular and widely used combinations. Simpson and Bridges (1959), using a sample of 120 children referred to the Division of Child Guidance of the Oklahoma City Public Schools, obtained a correlation of .874 between the Vocabulary-Block Design sum and the Full Scale IQ—a correlation comparable with that secured for the Anglo children (.867, table 7) in the Riverside sample. The chronological age of their subjects ranged from 65-192 months, with a mean of 124.23 months; Full Scale IQ ranged from 54-142, with a mean of 95.10. Silverstein (1967c¢), using the WISC standardi- zation sample, correlated the sum of scaled scores on Vocabulary and Block Design and the Full Scale IQ. The regressions were tested and found homogeneous for both tests. Conse- quently, the pooled within-groups regressions were used in the subsequent analysis. For the WISC, the pooled within-groups correlation of .878 corresponded to a standard error of esti- mate of about 7.2 IQ points. In a later study, Silverstein (1967b) investigated the possibility of using Vocabulary and Block Design as a short form of the Wechsler Preschool and Primary Scale of Intelligence (WPPSI), as well as of Wechsler’s other scales. Using the published tables of subtest intercorrelations, the correla- tion between the sum of scaled scores on Vocabulary and Block Design (V-BD) and the Full Scale (FS) score was determined for each of the six age agroups which comprised the WPPSI standardization sample. The regressions were tested and found homogeneous, and the method of deriving scaled scores made it possible to use a single regression equation for all groups: (FS = 3.57 V-BD + 28.53). The pooled within-groups correlation of .851 corresponded to a standard error of estimate of about 7.9 IQ points. These correlations and standard errors are very similar to those found in the present study (see table 8). Silverstein con- cluded that errors of this magnitude were not prohibitive for screening purposes, but no one would seriously advocate the use of a short form for a comprehensive assessment of intellectual functioning for individual case classification. In another study, Silverstein (1967d) used Doppelt’s formula to determine the correlation with the WISC Full Scale of all possible short forms of two, three, four, and five subtests. The range of correlations with the Full Scale of the 10 best short forms of each length at each age level were presented in table form. The range of correlations over age for the 10 best dyads was .807 to .906. In the present study, the range of correlations for the best dyads was .749 to .899 (table 7), very similar to those obtained by Silverstein. For obvious reasons, Silverstein’s correlations increased as the length of the short form increased. Although we found little varia- tion in correlations by age, Silverstein found a tendency for the correlations at age 10% to be higher than those at 13%, which in turn tended to be higher than those at 7%. Information- Picture Arrangement, Information-Block Design, and Vocabulary-Block Design were among the best dyads at all age levels. At ages 10% and 13%, the standard errors of estimate were approximately 7.0 IQ points for the best dyads. At age 7Y%, the standard errors were about 1.5 points higher for the best dyads. Silverstein also applied the Wherry-Doolittle method to the WISC standardization data. This method selects the best short form of each length but entails the differential weighting of subtest scores rather than their simple summation. Comparison of simple and multiple correlations revealed that the use of differential weighting did not result in appreciably higher validities. Silverstein agrees with Mumpower (1964) that a correlational measure of validity is not as meaningful as the agreement between Short Form IQ and Full Scale IQ in classifying individuals. Using Wechsler’s seven-category clas- sification system from Very Superior to Defec- tive, he estimated the agreement between the best short forms and the Full Scale IQ. The best dyads misclassified more than one individual out of every three, and even for the best pentads the corresponding figure was one out of every five. Thus, Silverstein concluded that data on agree- ment should be used to supplement correlational data in evaluating the validity of short forms. Mumpower (1964) noted that even with an as high as .90, only 81 percent of the variance in short-form IQ is attributable to the Full Scale IQ, leaving 19 percent, or nearly one-fifth, of the variance unaccounted for. His random sam- ple of 50 children referred for evaluation to the Special Education Department at the University of Southwestern Louisiana ranged in age from 7 years 2 months to 15 years 10 months, and included a variety of exceptional characteristics: slow learner, retarded educable, exceptionally able, emotionally disturbed, speech problem, vision problem, hearing problem, neurologically impaired, educationally retarded, and physically handicapped. The sum of Vocabulary and Block Design scaled scores was converted to a Short Form IQ. The Full Scale and Short Form IQ distributions were then correlated and yielded a 13 Pearson » or .95. The respective means were 85.64 and 83.52, and standard deviations were 19.93 and 18.33. Each Full Scale 1Q and Short Form IQ was classified into 10 categories: exceptionally able, superior, above average, high average, average, low average, slow learner, retarded educable, retarded trainable, or re- tarded custodial. Both IQ’s placed the child in the same category in 39 cases of the 50 comparisons made. The average difference be- tween Full Scale IQ and Short Form IQ for the remaining 11 cases was 9.5 IQ points. Thus, Mumpower concluded that 22 percent of the cases would have been misclassified on the basis of Short Form IQ and resulting recommenda- tions would in all probability have been inaccu- rate. The extent of classification errors using the multiple regression equations based on the Anglo children to identify children with sub- normal intelligence in the present sample is presented in table 6. Both Wechsler (1949) and Seashore et al. (1950) warned the user to take the fairly low reliabilities of some of the subtests into account in interpreting cither the absolute subtest scores or relations between them. Littell (1960) in his review of a decade of research on the WISC noted that at the age level of 7% years only Vocabulary, Picture Arrangement, Block Design, and Mazes had coefficients of internal consist- ency above .70, while Comprehension and Pic- ture Completion fell as low as .59. He reminded the reader that the reliability of the test tends to increase with age, so that at age level 13%, all subtests except Digit Span (.50) and Picture Completion (.68) were above .70. He suggested that age of subjects be studied as a variable in the construction of WISC short forms. However, age differences were negligible for the Anglo children in Riverside. Conclusions 1. Vocabulary and Block Design were the optimal dyad for predicting the Full Scale IQ for Anglo children in the Riverside sample. 2. Although there were socioeconomic status and age differences in the Full Scale, Verbal, and Performance 1Q’s of children in the Riverside sample, differences in multiple correlations be- tween Vocabulary and Block Design scaled scores and Full Scale 1Q’s by socioeconomic status and age were negligible. Therefore, we conclude that the regression equation based on all Anglo children was as efficient as the regression equation based on subgroups cate- gorized by age or socioeconomic level. 3. When Vocabulary and Block Design were used to predict Full Scale IQ (R =.867) of Anglo children, the regression equation was Predicted Full Scale 1Q = 563.77 + (2.45 X Vocabulary Scaled Score) + (2.15 X Block Design Scaled Score) The standard error of estimate was 7.50 IQ points. 4. When individual Anglo children were clas- sified as subnormal or normal using this equa- tion, 2.4 percent were falsely classified as having 1Q’s below 85; .6 percent were falsely classified as having 1Q’s below 80; and .2 percent were falsely classified as having 1Q’s below 70. On the other hand, 2.2 percent were falsely classified as having 1Q’s of 85 and above; 1.0 percent were falsely classified as having I1Q’s of 80 and above; and 2.0 percent were falsely classified as having IQ’s of 70 and above. The total error in classify- ing individuals as subnormal was, thus, 4.6 per- cent, 1.6 percent, and 2.2 percent for the three criteria. 5. When estimates were concerned with the percentage of Anglo children falling below the three criteria rather than the placement of individual children, the error was greatly re- duced. The predicted percentage of children with 1Q’s below 85 was .2 percent higher than the actual percentage; the predicted percentage with IQ’s below 80 was .4 percent less than the actual percentage; and the predicted percentage with I1Q’s below 70 was 1.8 percent less than the actual percentage. NEGRO CHILDREN Age Table 9 presents the mean IQ’s and standard deviations for Negro children by age. F ratios comparing mean scores over age indicate a significant age variation (p <.05) for Verbal 1Q but not for Full Scale or Performance 1Q’s. By Table 9. Mean IQ's, standard deviations, and F ratios for Negro children, by age Age in years Mean 1Q and Featio standard deviation 6 7 8 9 10 1 ra (N=41) (N=64) (N=58) (N=64) (N=48) (N=43) Full Scale 1Q Meal v . ws ss sms Hs os ® ¢ 8 92.1 93.7 88.2 93.4 94.9 91.9 2.19 2+ ir 11.8 11.3 124 11.5 109 12.3 NS Verbal 1Q Mean . .. ov woeaswssimsi 92.6 93.0 87.8 93.4 95.8 93.1 2.40 SD... eee 11.9 11.5 13.4 12.8 12.8 11.6 p <.05 Performance 1Q Mean . . . . . . «on 92.6 96.5 90.3 94.1 94.4 92.3 1.68 BD vw worm rm es soe ow ow 13.0 11.8 13.0 114 12.0 129 NS chance, 8-year-old children scored significantly lower. As with the Anglo sample, subsequent analyses will study the performance of Negro children by age group. There were 90 children in the sample with Full Scale IQ’s of 84 and below; 83 with Verbal I1Q’s of 84 and below; and 81 with Performance IQ’s of 84 and below. Sex Within Socioeconomic Status IQ’s and subtest scores for Negro boys were compared with those of Negro girls within two socioeconomic levels—low status and middle status. There was not a sufficient number of Negro children from high-status homes for anal- ysis. Table 10 presents the mean scores and the values of t when the means were compared. Low-status Negro boys did significantly better on verbal subtests than low-status girls. The boys had a significantly higher Verbal IQ (p <.05) and did significantly better than the girls on four of the six verbal subtests: Information, Compre- hension, Arithmetic, and Vocabulary. They also did better than the girls on Picture Completion, but the girls scored significantly higher on Coding. There were no significant sex differ- ences for middle-status children. Thus, the pat- tern of sex differences was similar to that found for Anglo children, although it was somewhat more accentuated for Negro children. In both ethnic groups, low-status boys had a signifi- cantly higher Verbal IQ than low-status girls, and boys in both ethnic groups did better in Comprehension, Vocabulary, and Picture Com- pletion than girls. Sex differences disappeared when middle-status boys were compared with middle-status girls. When Full Scale, Performance, and Verbal IQ’s of low-status and middle-status children were compared, there was no difference in Verbal 1Q for the boys, but all other compari- sons were statistically significant for both sexes. In every comparison, the middle-status children had higher IQ’s than low-status children. Socioeconomic differences on specific sub- tests were not as marked as they were for Anglo children. Among the verbal subtests, only Digit Span differentiated low-status boys from middle-status boys, while only Information and Vocabulary differentiated the girls. Both middle-status boys and girls did significantly better than low-status children on Object Assem- bly, and the middle-status boys did significantly better on Picture Arrangement. Although only six of the subtest comparisons were statistically significant, the mean scaled score of middle- status children was higher than that of low- status children on all subtests. 15 Table 10. Mean 1Q’s and subtest scores for Negro children, by socioeconomic status and sex and mean IQ's across socioeconomic status Comparison across Low status Middle status socioeconomic Mean 10 status and subtest ; Boys Girls ' Boys Girls ; al Ba (N=125) (N=115) (N=31) (N=47) 2 ; Mean 1Q POliScale JQ . os si ms ERS HEE ARs ES 91.6 89.4 1.4 97.4 94.9 4 12.6 12.6 Verbal 1Q ©... ...... 0... 92.7 89.1 29.2 97.2 93.8 1.1 1.8 29.0 Performance IQ . . .. .. ............ 92.0 91.7 2 98.0 97.0 '2.9 224 Verbal subtests Information . . . . .... 8.8 7.9 126 9.7 8.8 1.4 1.8 204 Comprehension . . . . . . . . . 8.8 8.0 225 9.7 8.8 1.4 1.7 1.5 Athmetic . . . + + ov «5 0 5s 5 5 wn om vom aw 9.3 8.4 22.5 9.4 8.5 1.3 2 2 Similarities «5 5 ss cvs a sms sms msn 8.9 9.1 -.5 9.3 94 -1 7 6 VOCEOUWBIY + + «wv rv ms ms 8 5 5 8 5 4 5d 9.0 7.7 '3.4 9.6 9.6 "| 1.3 '39 DigitSpan . . . . . . Looe 8.2 8.6 -1.1 9.6 89 9 22.0 5 Performance subtests Picture Completion . . . . . . . . . ....... 8.8 8.2 23. 9.8 8.8 1.5 1.8 1.7 Picture Arrangement . . . . . . . . . . ... .. 9.2 9.1 2 10.4 9.5 1.4 225 .6 Block Design. . . . . . . LL... 8.6 8.3 9 8.8 89 -2 3 1.4 Object Assembly . . . . . . . . ......... 8.6 8.3 1.0 9.9 9.7 4 '3.1 '29 Coding + « os 9 2 5 2 5 8 ms 8 sw Eww Bowe ow 8.9 10.2 1.35 9.7 109] -19 1.3 1.6 ! Significant at .01 level. 2Significant at .05 level. Intercorrelation of Scores for Children Aged 7 Only one of the intercorrelations between and 10 Tables IV and V in appendix II present the intercorrelations between subtest scores and the correlations between subtest scores and 1Q’s for Negro children 7 and 10 years of age. These tables were compared with those published by Wechsler for the standardization sample (Wechsler, 1949). There were three intercorrela- tions between subtest scores for 7-year-olds that were significantly smaller in the Negro sample than in the standardization sample and two correlations with Full Scale IQ that were signifi- cantly smaller. None of the calculations were larger in the Negro sample than in Wechsler’s sample. There were more differences which were significant for Negro children at age 10 than at age 7, just as there were for Anglo children. 16 Verbal subtests was significantly lower than in the standardization sample, but 14 of the 30 intercorrelations between Verbal subtests and Performance subtests were lower for Negro children at age 10 in the Riverside sample than for the children in Wechsler’s sample. Correla- tions were especially low between Comprehen- sion and Arithmetic scores and Performance subtest scores—four out of five comparisons were significantly different. Three out of five correlations with Information were also signifi- cantly lower. This, in turn, produced signifi- cantly lower correlations for Negro children between three Performance subtests (Picture Completion, Picture Arrangement, and Block Design) and Verbal 1Q, and very low correlations between Performance IQ and scores in Compre- hension and Arithmetic. Correlations with Full Scale IQ were, consequently, depressed for these five subtests. Thus, we conclude that correla- tions between Verbal and Performance scores were relatively low for Negro children at age 10. The number of children aged 10 in the Negro sample was small (only 48 children), and these relatively low intercorrelations must be inter- preted with caution. However, it does appear that Verbal and Performance tasks were less highly correlated in the Negro sample than in the Anglo sample on which the tests were originally normed. These differences tend ‘to justify our decision to treat the ethnic groups separately in analyzing the efficacy of various combinations of subtests as short forms of the WISC. Predicting Full Scale, Verbal, and Performance 1Q’s From Vocabulary and Block Design Scaled Scores Table 11 presents multiple regression coeffi- cients predicting Full Scale, Verbal, and Per- formance IQ’s from Vocabulary and Block De- sign scaled scores for Negro children by socioeco- nomic status and age. As with Anglo children, the multiple correlations were highest between the two subtests and Full Scale IQ, dropped slightly for the Verbal IQ, and dropped more markedly for Performance IQ. Correlations were slightly higher for low-status than for middle- status Negro children. Standard errors varied accordingly. The multiple correlations and standard errors of Verbal and Block Design scaled scores with Full Scale, Verbal, and Performance IQ’s for Negro children by age group were of approxi- mately the same magnitude as those found by socioeconomic level. In general, standard errors were lowest when predicting Full Scale IQ, increased slightly when predicting Verbal IQ, and were largest when predicting Performance 1Q. te multiple correlations and their related standard errors for all Negro children combined were about the same as those for the various subgroups by socioeconomic status and age, indicating that there was no appreciable im- provement in prediction when subgroup classifi- cations were used. Table VI (appendix II) shows the predicted Full Scale 1Q’s of Negro children with various combinations of Vocabulary and Block Design scaled scores based on the formula Estimated 1Q = (2.55 X Vocabulary Scaled Score) + (1.71 X Block Design Scaled Score) + 55.0 Percentage of Error in Predicting Subnormal 1Q Using Three Criteria The appropriate multiple regression equations were used to predict the Full Scale IQ of each Negro child by socioeconomic status and age. The percentages of children classified correctly and incorrectly as having subnormal IQ’s are shown in table 12, which also presents the nature and magnitude of the discrepancies. There was a relatively large percentage of individual children at the lower end of the IQ distribution who were misclassified, especially when 1Q 84- was used as the criterion. Among low-status Negro children, 6.3 percent of the children who were classified as having IQ’s below 85 in fact had 1Q’s of 85 and above, and 9.6 percent of those who were classified as having 1Q’s of 85 and above in fact had IQ’s below 85. This means that 15.9 percent of the individual low-status Negro children would be misclassified predicting IQ from Vocabulary and Block Design scaled scores using this criterion. The total misclassification at this level for middle-status children was even higher. Of the middle-status children, 7.7 percent were falsely classified as having 1Q’s below 85, and 14.1 percent were falsely classified as having 1Q’s of 85 and above, for a total of 21.8 percent misclassifications. The magnitude of error in classifying individuals is much less using criteria of IQ 79- and IQ 69-. Within the age levels, the percentage of ‘false lows” ranged from 0 to 11.5 percent, and the percentage of “false highs” ranged from 2.2 percent to 13.1 percent. To some extent, the “false lows” balanced the “false highs.” Thus, the overall error in predict- ing the percentage of subnormal IQ’s for the aggregate was lower than that for predicting individual IQ’s. In the sample of Negro children, all differences between the actual and predicted percentages of children having low IQ’s were underestimates. These ranged from 1.3 percent for middle-status children at the 69- criterion to 6.4 percent for middle-status children using the 84- and 79- criteria. Underestimates for low- Table 11. Full Scale, Verbal, and Performance IQ's from scaled scores for Negro children, by socioeconomic status and age, and beta coefficients Full Scale 1Q Verbal 1Q Performance 1Q Status, subtest, and age r R gue r R si r R Bes coefficient coefficient coefficient STATUS Low (N=240) Vocabulary . . . ..... .... .745 .745 249 | 816 .816 3.32 | .457 | '(2).748 1.11 Block'Design' , « + 2 55 v + = + a .583 835 1.64 | .339 .822 46 | 703 (1).703 2.68 Constant Term. . . . .. .... 55.78 59.36 59.79 Standard errorof Y . . . . . . .. 6.46 7.12 8.08 Middle (N=78) Vocabulary . . . . ........ .668 668 258 | .732 732 3.54 | 404 | '(2).729 1.08 Block Design. . . . . . . .... 578 .785 1.94 | .351 .748 .82 | .694 (1).694 2.81 Constant Term , . . «cs 6 2 4s 5 63.87 53.82 62.04 Standard errorof Y . . . . .. .. 7.64 9.01 8.32 AGE 6-11 years (N=318) Vocabulary . «vo » sms ws ws 4 .736 736 2.55 | .798 .798 3.35 | .464 | '(2).747 1.19 Block Design + : « s 5 5 5 3 « 5 + .581 .827 1.71 .346 .806 54 | .698 (1).698 2.71 Constant Term. . . . . . .... 55.07 58.30 59.61 Standard errorof ¥Y . . . . . . .. 6.78 7.60 8.21 6-7 years (N=105) Vocabulary ws a 5 5 5 » 5 8 5 3 .746 .746 268 | .777 777 3.12 | 547 | '(2).730 1.75 Block Design. . . . . .. .... 571 .830 1.69 | .399 .796 82 | .626 (1).626 2.36 Constant Term. . . . . . . ... 55.33 58.82 58.05 Standard errorof Y . . . . .. .. 6.56 7.25 8.41 8-9 years (N=122) Vocabulary . . . . ........ .704 704 2.30 | .805 .805 3.28 | .384 '(2).756 J4 Block Design. . . . . . . .... .607 821 1.91 374 .819 J3 1 7356 {1).735 2.90 Constant Term 5 . 4 + 5s 55 vw 5 54.78 56.61 61.15 Standard errorof Y . . . . . LL. 7.18 7.82 8.08 10-11 years (N=91) Vocabulary ous woz sow somos .786 .786 293 | .808 .808 3.83 | 516 | '(2).779 1.40 Block Design + vi x ¢ nos ow 8 ws .567 846 1.37 | .266 .809 -.08 | 727 (1).727 2.72 Constant Term. . . . . .. ... 54.04 58.76 857.25 Standard errorof Y . . . . . . .. 6.34 7.52 7.96 ! The order of the subtests reverses for the Performance 1Q because Block Design, being a performance test, always has a higher linear correlation with Performance 1Q and independently accounts for more of the variance in Performance IQ than does Vocab- ulary. Table 12. Percent of correct and incorrect predictions of low Full Scale 1Q’s for Negro children, by socioeconomic status and age using three different criteria Correctly Correctly Actual Predicted Status, 1Q, . — False False A ed 7 ana 308 identified low 1Q's’ high 1Q's? identified percent percent Difference 39 low 1Q’s INIEST high 1's | low 1Q’s | low 1Q's STATUS Low (N=240) Peregnt 14 § 7. CT IP 21.7 6.3 9.6 62.5 31.3 28.0 -3.3 1Q79— . oo oe eee eee 75 4.6 79 80.0 15.4 12.1 -3.3 1Q69— . . . «oe eee 4 4 29 96.3 3.3 .8 -25 Middle (N=78) OBA . vv wri FIRE EEE EE 5.1 7.7 14.1 73.4 19.2 12.8 -6.4 IQ79= + . + vr vt meme AE EE 3.8 0.0 6.4 89.7 10.2 3.8 -6.4 1Q69— . . oo eee eee 0.0 0.0 1.3 98.7 1.3 0.0 -1.3 AGE All ages, 6-11 years (N=318) 1Q84— . . . oo ieee 17.6 6.6 10.7 65.1 28.3 24.2 —4.1 JQ ZO + iw 2 « 5 s ws ws 3s 5 «mx mow 6.6 a5 7.5 824 14.1 10.1 -4.0 10BG . » + 45 sas a 85 8m s dsm» 3 3 25 96.9 2.8 6 -2.2 6-7 years (N=105) 1Q88= 2 i « «+ ws was sims ww ewe wo 16.2 4.8 10.5 68.6 26.7 21.0 -5.7 1079 vw vowsmsisw sia mn mms 5.7 3.8 6.7 83.8 124 9.5 -29 IQB9— . . ote he ee ee eee 0.0 0.0 29 97.1 29 0.0 -29 8-9 years (N=122) 088 . civ s in inpimimenss = 20.5 11.5 13.1 54.9 33.6 32.0 -1.6 QT ov wv wn ws FI FEE AEE PES 6.6 33 9.8 80.3 16.4 9.9 -6.5 1QB9— . oo he ee ee ee ee .8 .8 25 95.9 3.3 1.6 -1.7 10-11 years (N=91) 10 7 EE EE BE EE. 15.4 5.5 7.7 71.4 23.1 20.9 -2.2 HO 79 «vv vnwwwn v8 sab ® sd ii 7.7 3.3 55 83.5 13.2 11.0 -2.2 1Q69— . oo ee ee eee eee 0.0 0.0 2.2 97.8 22 0.0 -2.2 False low 1Q’s were for children who were predicted to have an IQ below the criteria but who actually had an 1Q above the criteria. 2 False high 1Q’s were for children who were predicted to have an 1Q above the criteria but who actually had an 1Q below the criteria. status children ranged from 2.5 percent to 3.3 percent. These discrepancies were much higher than the comparable discrepancies for Anglo children (table 6). When predicting by age group, errors in predicting rates for aggregates were again lower than errors in placing individuals, and all differ- ences were in the direction of underestimates. These ranged from a low of 1.6 percent when predicting the percentage of children at ages 8 and 9 with IQ’s below 85 to a high of 5.7 percent when predicting the percentage of chil- 19 dren at ages 6 and 7 with 1Q’s below 85. These underestimates are of approximately the same magnitude as those found by socioeconomic status. When predictions were made for all Negro children without regard for age or socioeco- nomic status, 4.1 percent fewer 1Q’s of 84 and below were predicted than were actually found, 4.0 percent fewer 1Q’s of 79 and below were predicted than found, and 2.2 percent fewer 1Q’s of 69 and below were predicted than were found. These discrepancies were larger than those found for Anglo children, indicating more error may be expected in assessing individual subnormality for Negro than for Anglo children using Vocabulary and Block Design scores as the predictors. Optimal Prediction From Various Combinations of Subtests Table 13 presents the results of a stepwise multiple regression analysis and shows the sub- tests that best predicted Full Scale, Verbal, and Performance 1Q’s for Negro children of various sociocconomic and age levels. The four tests predicting the most variance in Full Scale 1Q and the three tests predicting the most variance in Verbal and Performance 1Q are reported. Multiple correlations with Full Scale 1Q as the dependent variable ranged from .920 to .935, and standard errors ranged from 4.32 to 4.75 1Q points. These correlations were very similar to those found for Anglo children. Vocabulary was the best single predictor of Full Scale 1Q for low-status children and for age groups 6-7 and 10-11. Tt appeared among the top three tests in predicting Verbal IQs for children of both social statuses and in one age group. Block Design did not appear as a high-ranking subtest in any of the analyses by socioeconomic status or for the group as a whole. Digit Span, which did not appear at all for Anglo children, appeared as one of the three best predictors of Full Scale 1Q for low-status children and of Performance IQ for children at both socioeconomic levels and at all age levels. The size of the multiple correlations was similar for all subgroups of Negro children. Digit Span and Picture Arrangement, which did not appear in any of the solutions for Anglo 20 children, were highly correlated with Verbal 1Q and Performance IQ of Negro children. The pattern of subtests which was the best predictor of Full Scale 1Q for Negro children fluctuated. However, Vocabulary appeared most frequently as the Verbal subtest that was most highly correlated with Full Scale and Verbal 1Q. Object Assembly appeared most frequently as the Performance subtest having the highest correlation with Full Scale IQ and also was the best predictor in four of the six predictions of Performance 1Q. Picture Arrangement appeared among the top three variables in every predic- tion of Performance IQ. Block Design did not have the same consistent correlation with 1Q as cither of these subtests. A short form consisting of the Vocabulary and Object Assembly subtests would have given slightly better predictions for the group as a whole than the combination of Vocabulary and Block Design used in Cycle II of the Health Examination Survey. However, the difference in the multiple R is trivial (R = .835 vs. .827 or .008 difference). Optimal Predictions Compared With Predictions Based on Vocabulary and Block Design Table 14 compares the multiple correlations obtained using the two optimal subtest predic- tors with the multiple correlations obtained using Vocabulary and Block Design. Differences between optimal predictions and predictions of Full Scale IQ based on Vocabulary and Block Design were generally slightly larger for Negro than for Anglo children but were not as large as might be expected, considering that Object Assembly was a better predictor than Block Design for Negro children (table 8). For the entire Negro group, the difference between the two multiple correlations was only .008. Discrepancies in predicting Verbal and Per- formance 1Q were slightly higher than those found for Anglo children. Again, this may have been because Block Design was not as good a predictor for Negro children as for Anglo children. Discussion The number of studies of the WISC which have been published during the last two decades Table 13. Optimal predictions of 1Q’s of Negro children from various combinations of subtests, by socioeconomic status and age Full Scale 1Q Verbal 1Q Performance 1Q Status, subtest, Status, subtest, Status, subtest, 4 Beta 2nd age Beta an Beta and. age r R coef- 9 r R coef- ge r R coef- ficient ficient ficient STATUS STATUS STATUS Low (N=240) Low (N=240) Low (N=240) Vocabulary . ....... .745 .745 1.63 | Vocabulary . ....... .816 816 2.36 | Object Assembly . . . . . 737 737 2.35 Object Assembly. . . . . . .605 .840 1.38 | DigitSpan .. ...... 637 .903 1.35 | Picture Arrangement 671 .849 1.77 Arithmetic... ..... .653 .894 1.28 | Arithmetic . ....... 726 942 146 | Coding .......... 521 916 1.51 Picture Arrangement .625 921 1.04 | Constant Term . ..... 47.03 Constant Term. . . . . . 41.40 Constant Term. . . . .. 44.44 Standard errorof ¥ . . . . 4.21 Standard errorof Y . . . . 4.89 Standard errorof Y . . . . 4.59 Middle (N=78) Middle (N=78) Middle (N=78) Arithmetic. . . ..... .744 744 1.63 | Arithmetic ........ 815 .815 1.93 | Object Assembly . . . . . J12 711 2.35 Picture Completion 531 .855 1.62 | Vocabulary . .. .. 732 906 2.14 | Picture Completion 643 872 2.07 Object Assembly . . . . . .603 901 1.46 | DigitSpan . . .. . .704 947 1.27 | Picture Arrangement .627 922 1.27 Similarities . . ...... .655 935 1.33 | Constant Term . ..... 45.76 Constant Term. . . . .. 42.94 Constant Term . . . . . . 39.76 Standard errorof Y . . . . 4.39 Standard errorof Y . . . . 4.74 Standard errorof ¥ . . . . 4.43 AGE AGE AGE 6-11 years (N=338) 6-11 years (N=338) 6-11 years (N=338) Vocabulary . ....... .736 .736 1.68 | Vocabulary . ....... .798 .798 2.31 | Object Assembly . . . . . 742 .742 2.30 Object Assembly . . . . . .620 .835 1.59 | Arithmetic . ....... 744 .901 1.35 | Picture Arrangement .663 .846 1.79 Arithmetic. . . ..... .666 .894 1.15 | DigitSpan . ....... .661 944 158 | Coding .......... 516 913 1.52 DigitSpan . .. ..... .599 .920 98 | Constant Term . . . ... 46.40 Constant Term . . . . . . 41.51 Constant Term . . . . .. 44.81 Standard errorof Y . . . . 4.26 Standard errorof Y . . . . 5.05 Standard errorof ¥ . . . . 4.75 6-7 years (N=105) 6-7 years (N=105) 6-7 years (N=105) Vocabulary . ....... .746 746 1.83 | Vocabulary . ....... A717 777 2.22 | Picture Arrangement 697 697 1.96 DigitSpan . ....... 672 .848 1.19 | DigitSpan . ....... 729 .899 1.36 | Coding .......... 631 .862 1.90 Object Assembly . . . . . .608 .895 1.25 | Arithmetic . ....... .655 943 1.39 | Block Design . . ..... .626 932 1.77 Block Design . . . . . .. 571 921 1.06 | Constant Term . . .. .. 49.06 Constant Term . . . . . . 42.06 Constant Term . . . . .. 46.26 Standard errorof ¥ . . . . 4.01 Standard errorof Y . . . . 4.47 Standard errorof ¥ . . . . 4.63 8-9 years (N=122) 8-9 years (N=122) 8-9 years (N=122) Information . . . . .... 775 775 1.85 | Information . . ...... .829 .829 2.12 | Block Design . . ... .. .735 .735 2.40 Object Assembly . . . . . .630 .870 1.52 | DigitSpan . ....... 653 901 1.67 | Picture Arrangement .660 872 1.80 Arithmetic . . . ..... .749 912 1.52 | Vocabulary . ....... .805 946 171M | Coding .......... .456 929 1.52 Picture Completion .450 933 1.02 | Constant Term . . . . . . 44.48 Constant Term. . . . . . 40.31 Constant Term . . . . . . 40.26 Standard errorof ¥ . . . . 4.45 Standard errorof Y . . . . 4.59 Standard error of Y . . . . 4.56 10-11 years (N=91) 10-11 years (N=91) 10-11 years (N=91) Vocabulary . ....... .786 .786 1.47 | Information . . ...... 817 817 2.72 | Object Assembly . . . . . .806 .806 2.58 Object Assembly . . . . . 618 .859 1.61 | DigitSpan . ....... .594 .897 156 [Coding «x. iu. in .406 .872 1.48 DigitSpan . ....... 499 913 1.16 | Comprehension . . . . . . 670 942 1.64 | Picture Arrangement .635 930 1.72 Information . . . . .... .759 933 1.42 | Constant Term . . .. .. 41.38 Constant Term. . . . . . 39.40 Constant Term. . . . . . 42.76 Standard errorof ¥ . . . . 4.33 Standard errorof Y . . . . 4.70 Standard errorof Y . . . . 4.32 21 Table 14. Comparison of Block Design and Vocabulary as predictors with the two optimal predictors for Full Scale, Verbal, and Performance 1Q’s for Negro children, by socioeconomic status and age Full Scale 1Q Verbal 1Q Performance 1Q Stat nd age # # A 2 B A is Shang Block Design Two Differ- Block Design Two Differ- Block Design Two Differ- and optimal ence and optimal ence and optimal ence Vocabulary predictors Vocabulary predictors Vocabulary predictors Status Low (N=240) . . . . .. .835 .840 .005 822 903 .081 .748 .849 101 Middle (N=78) . . . .. 785 B55 .070 748 .906 .158 729 .872 143 Age Total, 6-11 years (N=318) . . . .827 835 .008 .806 901 .095 747 .846 .099 6-7 years (N=105) . . . .830 .848 .018 .796 .899 .103 .730 .862 A32 8-9 years (N=122) .821 .870 .049 .819 901 .082 756 B72 116 10-11 years (N=91) . . . .846 B59 013 .809 .897 .088 779 .872 .093 using samples of Negro children is very limited. In general, such studies have concluded that the Wechsler norms are not applicable to samples of rural, Southern Negro children (Young and Bright, 1954; Young and Pitts, 1951); that Performance 1Q’s should not be used as an alternative measure when cultural deprivation is suspected in Negro children (Teahan, 1962); and that geographic region is a significant variable influencing performance (Caldwell and Smith, 1968; Carson and Rabin, 1960). A study of children aged 7, 8, and 9 [ound that the subtest intercorrelation matrix was similar lor Negro and white children, except for lower intercor- relations involving the Arithmetic subtest for Negro children. However, a multivariate analysis of variance indicated that statistically dissimilar WISC structures existed for white and Negro children (Semler and Iscoe, 1966). A study of a selected sample of 84 Negro children aged 5 years 7 months to 12 years 6 months in five Southern States found Verbal 1Q significantly higher than Performance IQ for both sexes, all ages, and in all but one geographic location. Differences between Verbal IQ and Performance 1Q of groups from various geographic locations varied more than differences in Verbal and Performance 1Q’s by age or sex. The authors concluded that short forms of the WISC should not be used with Negro children, because every 22 factor analysis performed on this sample found over hall the total variance of the 12 subtests common and because the subtests were not correlated highly enough for was not shared in Negroes for any combination of them to be used as a short form of the test (Caldwell and Smith, 1968). Their conclusions are supported by a study of 177 children eligible for placement in classes for the educable mentally reterded (68 percent Negro and 32 percent white) conducted in Indianapolis. Only correlations based on pentads and hexads produced correlations of .90 or higher and thus qualified, according to the authors, for usc as short forms (Schwartz and Levitt, 1960). However, another study of a mixed 27 Negro and 56 white children found correlations comparable to those sccured in the present study—ec.g., a correlation of .91 between Full Scale IQ and a short form consisting of Vocabulary and Block Design (Wight and Sandry, 1962). These studies were all conducted using rela- tively small samples, special subpopulations such as cducable mental retardates, or mixed samples of Negro and white children. Therefore, it is difficult to interpret the meaning of conflicting results. The findings of the present study based on 318 Negro children, most of whom are in regular public school classes, showed that a short form consisting of Vocabulary and Block Design sample of correlated only slightly less well with Full Scale, Verbal, and Performance IQ’s of Negro children than it correlated with the IQ’s of Anglo children. Comparable correlations were .827, .806, and .747 for Negro children (table 14) and .867, .839, and .798 for Anglo children (table 8). Correlations using the two subtests yielding the optimal predictions were also very similar: .835, .901, and .846 for Negro children and .867, .903, and .861 for Anglo children. For the Riverside sample of Negro children, we must conclude that short forms of the WISC are just as feasible as they are for Anglos. In either case, however, a short form consisting of only two subtests produced predictions with relatively large standard errors. Using a short form consist- ing of Block Design and Vocabulary, the stand- ard errors were 6.78, 7.60, and 8.21 IQ points for Negro children when predicting Full Scale, Verbal, and Performance IQ, respectively, and 7.50, 8.15, 9.19 IQ points for Anglo children (see tables 5 and 11). Conclusions 1. Vocabulary and Object Assembly were the two subtests providing the best prediction of Full Scale IQ for Negro children in the Riverside sample. However, the prediction using this op- timal combination, R =.835 with a standard error of 6.63 IQ points, was only slightly better than that using Vocabulary and Block Design, R =.827 with a standard error of 6.78 IQ points. 2. Differences in multiple correlations be- tween Vocabulary and Block Design scaled scores and Full Scale IQ’s by socioeconomic status and age were negligible, in spite of the fact that there were sex differences in IQ for low-status children; age differences in Verbal IQ; and socioeconomic differences. The regression equation based on all Negro children was as good a predictor as the regression equations based on subgroups categorized by age or socioeconomic status. 3. When Vocabulary and Block Design were used to predict Full Scale IQ (R =.827) of Negro children the regression equation was Predicted Full Scale 1Q = 55.07 + (2.55 X Vocabulary Scaled Score) + (1.71 X Block Design Scaled Score) The standard error of estimate was 6.78 1Q points. Table VI in appendix II presents the predicted Full Scale IQ for Negro children who had various combinations of Vocabulary and Block Design scaled scores. 4. When Negro children were classified as subnormal or normal using this equation, 6.6 percent were falsely classified as having IQ’s below 85; 3.5 percent were falsely classified as having IQ’s below 80; and .3 percent were falsely classified as having IQ’s below 70. On the other hand, 10.7 percent were falsely classified as having 1Q’s of 85 and above; 7.5 percent were falsely classified as having IQ’s of 80 and above; and 2.5 percent were falsely classified as having IQs of 70 and above. The total error in classifying individuals as subnormal was 17.3 percent, 11.0 percent, and 2.8 percent for the three criteria. 5. Error was reduced when estimates were concerned with the percentage of children fall- ing below the three criteria rather than the placement of individual children, because “false lows” tended to balance “false highs.” In every case, however, the error was in the direction of underestimation—e.g., 4.1 percent using the 84- criterion, 4.0 percent using the 79- criterion, and 2.2 percent using the 69- criterion. MEXICAN-AMERICAN CHILDREN Age There were no differences in mean IQ’s by age for Mexican-American children. However, table 15 does reveal a pattern which did not appear for either Anglo or Negro children. The Performance IQ for Mexican-American children was significantly higher at every age level than the Verbal 1Q. Mean Performance 1Q’s ranged between 95.7 and 99.4, while mean Verbal IQ’s ranged between 84.6 and 90.0, approximately 10 points lower. There were 137 Mexican-American children with a Full Scale IQ of 84 and below; 204 with a Verbal 1Q of 84 and below; and 66 with a Performance 1Q of 84 and below. The number of Mexican-American children with low Verbal IQ’s was 2.4 times greater than the number of children with low Performance IQ's. 23 Table 15. Mean IQ's, standard deviations, and F ratios for Mexican-American children, by age Age in years Mean 1Q and F ratio standard deviation 6 7 8 9 10 11 (N=60) (N=96) (N=75) (N=99) (N=88) (N=69) Full Scale 1Q Mean . ............... 93.9 92.7 91.3 91.3 91.6 88.9 1.23 8D viv svsmsmsmsws nn 12.8 12.3 13.3 12.2 1.1 13.3 NS Verbal 1Q Mean . . .............. 90.0 88.6 87.8 88.1 88.2 84.6 1.38 8D .s:insuswins misma 129 12.9 138 12.4 10.9 11.9 NS Performance 1Q Mean . . .............. 99.4 99.0 97.0 96.8 97.6 95.7 .83 SI sie ms sins we mind ws le wow 12.8 125 12.9 12.8 12.3 15.1 NS Sex Within Socioeconomic Status Table 16 presents sex differences by socioeco- nomic level. The most conspicuous pattern in this table was the consistent differences found for low-status children. Low-status Mexican- American girls had significantly lower Full Scale, Verbal, and Performance 1Q’s than boys. They scored lower on every subtest but Coding and were significantly lower on Information, Com- prehension, and Vocabulary among the Verbal subtests; and they scored lower on Picture Completion, Picture Arrangement, Block Design, and Object Assembly among the Performance subtests. The general pattern was the same for middle- status children—i.e., girls tended to score lower on most subtests. However, the differences were not statistically significant except for Picture Completion. This same pattern appeared for low-status Anglo and Negro children but was less pro- nounced. Low-status Anglo girls tended to score lower than low-status Anglo boys with four significant differences in 14 comparisons (ta- ble 4). Low-status Negro girls scored lower than low-status Negro boys with seven of 14 compari- sons being statistically significant (table 10). Differences tended to disappear or to become insignificant at higher status levels for Anglo and 24 Negro children, just as they did for Mexican- American children. We concluded that the lower performance on the WISC of low-status girls was a pervasive tendency in all three ethnic groups but was most marked among Mexican-American children. Socioeconomic differences were negligible for the boys but were significant for the girls. Middle-status girls had significantly higher Full Scale, Verbal, and Performance 1Q’s and scored significantly higher on five of the six Verbal subtests and one of the Performance subtests. For this sample, it appeared that a middle-status background did not materially improve the intellectual performance of boys, but that it did improve the intellectual performance of girls. A significant sex difference was found at the lower socioeconomic level, but it disappeared at the middle socioeconomic level. The girls’ perform- ance improved so that it matched that of the boys while the latter changed little with social status. For Anglos, high socioeconomic status also had slightly more impact on the girls’ scores than on the boys’ scores—13 of 14 socioeco- nomic comparisons for Anglo girls were signifi- cant as opposed to nine of 14 for the boys (see table 4). Among Negro children, there was a comparable improvement of scores with higher socioeconomic status. Table 16. Mean 1Q’s and subtest scores for Mexican-American children, by socioeconomic status and sex and mean 1Q’s across socio- economic status Low status Middle status Comparison across socioeconomic status Mean 1Q and subtest . Boys Girls Boys Girls Boys Girls (N-205) | (N=208) | © | (N=36) | (nen) | © | (N41) [IN-239) Mean 1Q FUNSCale IQ uu vie 5 © 8% 3 8 3 4 & 8 be 89.2 83.8 133 96.7 95.1 5 1.9 133 Verbal IQ ; 4 oi vi 5 6 5 8 Em mae 87.8 84.7 12.6 92.3 92.5 -1 1.8 13.3 Performance lQ . ................. 98.5 94.3 133 102.1 99.0 1.1 1.5 224 Verbal subtests Information . . u ¢ w sw s 4 5 & BF as ms nea 7.9 7.2 22.5 8.6 8.8 -.2 1.4 3.1 Comprehension . . . .. ............. 8.2 7.4 '3.4 8.7 7.9 1.0 .8 1.1 Arithmetic . . . . . .. .............. 9.0 8.5 1.9 10.0 9.8 2 1.7 12.8 Similarities . . .. ................ 8.1 7.8 9 8.3 93 | -14 4 '3.0 Vocabulary . . . . ................ 7.4 6.8 22.2 8.3 8.4 -1 1.7 i DigitSpan . . . .. .. .............. 7.7 7.2 8.7 8.6 1 22.3 22.0 Performance subtests Picture Completion . . .. ............ 9.4 8.4 '3.8 10.5 86 | '28 1.8 6 Picture Arrangement . . . . . .. ........ 9.6 9.0 220 10.3 9.9 .6 1.6 22.0 Block Design . .................. 9.6 8.7 '3.5 10.5 95 14 1.5 1.5 Object Assembly . . . . . ............ 10.3 9.1 '48 10.1 99 4 4 1.9 Coding . . . . ... 10.0 108 | '-2.7 10.1 113] -19 2 1.2 ! Differences are significant at .01 level. ? Differences are significant at .05 level. Intercorrelations of Scores for Children Aged 7 and 10 Tables VII and VIII in appendix III present the intercorrelations of the subtests of the WISC for Mexican-American children at ages 7 and 10. The correlations in these tables were compared with those published by Wechsler for children aged 7% and 10%2 on whom the test was normed. The z test of the significance of difference between two r’s was used. There were no Mexican-American children in the Wechsler sam- le. b There was a pattern of lower correlations between Coding and other subtests for 7-year- old children. There were no other significant differences in the correlations when 7-year-old Mexican-Americans were compared with chil- dren of similar age in Wechsler’s sample. For 10-year-old children, however, there were 34 correlations in the matrix which were signifi- cantly lower for Mexican-American children than for those in Wechsler’s sample. Correlations between individual subtests and Full Scale IQ were significantly lower for every subtest except Object Assembly and Coding, both Performance tests. Comprehension and Picture Arrangement also had low correlations with the other subtests and with Full Scale, Verbal, and Performance IQ’s. Ten of the 13 correlations for Comprehen- sion and nine of the 13 correlations for Picture Arrangement were significantly lower than for the sample on which the test was normed. Only the correlations of Comprehension with Arith- metic, Similarities and Picture Completion ap- proximated those of Wechsler’s sample. Correla- tions with Vocabulary were significantly lower on seven of 13 comparisons, including Full 25 Scale, Verbal, and Performance 1Q’s. The corre- lations of Block Design with the other subtests and with the 1Q’s were significantly lower on four of 18 comparisons. With such generally low intercorrelations, it was not surprising that the correlation between Verbal 1Q and Performance IQ was also significantly lower than that for the population on which the test was normed. These lower intercorrelations of the subtest scores with cach other and with Full Scale, Verbal, and Performance 1Q’s further justified the decision to treat ethnic groups separately in the evaluation of various short forms of the WISC. Predicting Full Scale, Verbal, and Performance 1Q’s From Vocabulary and Block Design Scaled Scores Table 17 presents the multiple correlation coefficients and standard errors obtained when Vocabulary and Block Design were used to predict the 1Q’s of Mexican-American children of various socioeconomic levels and ages. As with Anglo and Negro children, controlling for socioeconomic status and age did not appreci- ably improve the predictions. The multiple correlation for all the Mexican-American chil- dren for Full Scale 1Q was .846 with a standard error of 6.74 1Q points. The cocflicients for various socioeconomic and age categories ranged 832 to .873, and the standard errors ranged from 6.29 to 6.91. Predictions were most accurate for Full Scale IQ and least accurate for Performance 1Q. While the multiple correlation coefficients were ol similar magnitude, all the standard errors were consistently lower for Mexican-American children than Anglo children. They were about the same as those for Negro children. Table IX in appendix III presents data on the estimated Full Scale IQ of Mexican- American children, which was obtained based on various combinations of Vocabulary and Block Design scaled following formula: from scores using the Estimated 1Q = (2.64 X Vocabulary Scaled Score) + (1.98 x Block Design Scaled Score) + 53.49 26 The Percentage of Error in Predicting Subnormal 1Q Using Three Criteria The Full Scale 1Q for each Mexican-American child in the sample was estimated using the appropriate beta coefficients. The percentages of children classified correctly and incorrectly as having subnormal 1Q’s is shown in table 18. There were 30 percent of the low-status Mexican-American children and 17.6 percent of the middle-status Mexican-American children who had 1Q’s below 85, and 7.5 percent and 6.8 percent of these children, respectively, were [alsely predicted to have 1Q’s of 85 and above. Altogether, 14.8 percent and 8.2 percent of the individual children were misclassified using this criterion. Errors were somewhat reduced using the two lower criteria. However, “false highs” tended to be balanced by “false lows”; the predicted percentage of children with low 1Q’s was from .2 percent to 5.4 percent less than the actual percentage of children with low 1Q’s, a much smaller percentage of error than found with predictions for individual children. As with Negro children, differences between the actual and predicted percentage of low 1Q’s were all negative—i.c., in the direction of under- estimating the percentage of persons with sub- normal 1Q’s using the three different criteria. The magnitude of the errors was gencrally less than that found lor Negro children but greater than that found for Anglo children (tables 6 and 12). When the predicted Full Scale 1Q’s using the entire sample of Mexican-American children were compared with the actual IQs, 6.6 percent of the individual children were misclassified as having 1Q’s below 85 and 7.4 percent were falsely classified as having 1Q’s of 85 and above—a 14-percent error. The total error using the 79- criterion was 10.5 percent and 3.2 percent using the 69- criterion. Errors of this magnitude would seem to preclude using a short form consisting of two subtests for screening individual children. If the consistent bias toward underestimation is taken into account, however, such a short form could be used to obtain a rough estimate of the percentage of children in a given population having low 1Q’s using cach of the three criteria. This bias paralleled that found for Negro children (table 12). Table 17. Full Scale, Verbal, and Performance 1Q’s from scaled scores for Mexican-American children, by socioeconomic status and age, and beta coefficients Full Scale 1Q Verbal 1Q Performance 1Q Status, subtest, and age , AR Beta , R Beta 7 R Beta coefficient coefficient coefficient STATUS Low (N=413) Vocabulary . . . ......... .736 .736 2.68 | .806 .806 3.26 | 511 '(2).761 1.54 Block Design... ........ 612 .841 1.98 | 413 .825 .84 | 692 (1).692 2.87 Constant Term . . ........ 55.03 55.44 59.14 Standard errorof Y . . . . . . .. 6.72 6.91 8.46 Middle (N=74) Vocabulary . . .......... 721 721 2.36 | .834 .834 3.37 | 415 | '(2).792 81 Block Design. . . ........ .634 844 1.93 | .400 .850 75 | .768 (1).768 2.87 Constant Term. . . . ...... 56.92 56.73 65.11 Standard errorof Y . . . . .. .. 6.89 7:21 7.62 AGE All ages, 6-11 years (N=487) Vocabulary . . .......... .740 740 264 | .816 .816 3.30 | .504 | '(2).766 1.41 Block Design. . . .. ...... .620 846 1.98 | .420 .834 .83 | .706 (1).706 2.87 Constant Term. . ........ 53.49 55.40 60.06 Standard errorof Y . . . .. . .. 6.74 6.95 8.36 6-7 years (N=156) Vocabulary . . .......... .730 .730 2.44 | .802 .802 3.21 A485 | '(2).758 1.11 Block Design. . . ........ .634 .837 2.02 | .456 .825 99 | .713 (1).713 2.79 Constant Term. . . . ...... 55.79 56.13 63.70 Standard errorof Y . . .. . . .. 6.91 7.46 8.01 8-9 years (N=174) Vocabulary . . .......... 779 779 2.84 | .858 .858 3.56 | .533 | '(2).783 1.53 Block Design. . . ........ 612 .873 2.07 | .410 .872 .83 | .708 (1).708 3.05 Constant Term. . . . ...... 50.27 52.73 56.87 Standard errorof Y . . . . . . .. 6.29 6.38 8.08 10-11 years (N=157) Vocabulary . . .......... .729 729 2.69 | .800 .800 3.17 | .518 | '(2).761 1.66 Block Design. . . ........ 611 .832 1.76 | .389 813 59 | .693 (1).693 2.71 Constant Term . . «uo 5 5» = 4 « 54.63 57.82 59.63 Standard errorof Y . . . . .. .. 6.89 6.75 8.95 ' The order of the subtests reverses for the Performance 1Q because Block Design, being a performance test, always has a higher linear correlation with Performance 1Q and independently accounts for more of the variance in Performance IQ than does Vocabulary. 27 Table 18. Percent of correct and incorrect predictions for low Full Scale 1Q’s for Mexican-American children, by socioeconomic status and age using three different criteria Correctly Correctly Actual Predicted . False False . . Status, 1Q, and age identified low 1Q's"! high 10s? identified percent percent Difference low 1Q’s g high 1Q’s | low 1Q’s | low IQ's STATUS Percent Low (N=413) IQ BA «vv vv vv va hsm ewes 225 7.3 7.5 62.7 30.0 29.8 - 2 OTF s 53 s mos ews mememas nsw 1.1 34 8.2 77.2 19.3 14.5 -4.8 OBI , ohne snp HEHE MW 1.2 7 2.7 954 39 1.9 -2.0 Middle (N=74) BT 5 5 4 ws ws 2% #0 vu sme was mw 10.8 1.4 6.8 81.1 17.6 12.2 -5.4 179 ois v5 63 +9 + 9 #5 8 @s © ym 2.7 1.4 2.37 93.2 5.4 4.1 -1.3 [16 LR A EE LE 0.0 0.0 1.4 98.6 1.4 0.0 -1.4 AGE All ages, 6-11 years (N=487) 1Q84— . . . 20.5 6.6 7.4 65.5 279 27.1 - 8 1IQ79- . . eee 10.1 33 7.2 79.4 123 134 -3.9 1Q69- . . . 1.0 8 24 95.8 3.4 1.8 -1.6 6-7 years (N=156) HOB . . ov v2 msm bw B48 BE 16.0 4.5 8.3 71.2 24.3 20.5 -3.8 1Q79— . oo ee 6.4 1.9 7:7 84.0 14.1 8.3 -5.8 1Q69— . . oe ee 6 0.0 1.9 97.4 2.5 6 -1.9 8-9 years (N=174) 1084 , . 5 v5 55 ss 568 5B 538 19.5 75 9.2 63.8 28.7 27.0 -1.7 B70 \ vs ati Gm nm EEE BE EE 12.1 23 6.9 78.7 19.0 14.4 -4.6 1Q69- . . . eee 13 2.3 29 93.7 4.0 34 - 6 10-11 years (N=157) JQBA~ . . i: ss sm im Ew EET HEHE 236 6.4 7.6 62.4 31.2 30.0 1.2 1Q79- . «eee 11.8 5.1 7.0 76.4 18.5 16.6 -1.9 1Q69- . . . . eee 1.9 0.0 19 96.2 3.8 1.9 -19 False low 1Q's were tor children who were predicted to have an 1Q below the criterion but who actually had an 1Q above the criterion. 2 False high 1Q’s were for children who were predicted to have an 1Q above the criterion but who actually had an 1Q below the criterion. Optimal Prediction From Various Combinations of Subtests Table 19 presents the results of a stepwise multiple regression analysis in which cach of the scaled scores of the 11 subtests was used as an independent variable to predict Full Scale, Ver- bal, and Performance I1Q’s. Only the four best 28 predictors of Full Scale 1Q and the three best predictors of Verbal and Performance 1Q are shown. There was little difference in the size of the correlation coefficients or standard errors across socioeconomic status age. For the entire Mexican-American sample, R =.926 for Full Scale IQ using four subtests and ranged from 925 to 938 for various subgroups. The or Table 19. Optimal predictions of 1Q’s of Mexican-American children from various combinations of subtests, by socioeconomic status and age Full Scale 1Q Verbal 1Q Performance 1Q Status, subtest, Status, subtest, Status, subtest, id 358 Beta and Beta a Beta a g r R coef- 298 r R coef- and age r R coef- ficient ficient ficient STATUS STATUS STATUS Low (N=413) Low (N=413) Low (N=413) Vocabulary . . ...... .736 .736 1.73 | Vocabulary . ....... .806 .806 1.90 | Object Assembly . . . . . .709 .709 2.28 Block Design . . . .... 612 .841 1.32 | Arithmetic . ....... 725 .900 1.76 | Picture Arrangement .701 .850 2.03 Picture Arrangement .693 .897 1.28 | Similarities . . ...... .753 939 135 | Coding .......... .566 914 1.52 Arithmetic . . . ..... .688 925 1.30 | Constant Term . ..... 46.66 Constant Term . . . . .. 39.73 Constant Term . . . . .. 42.56 Standard errorof ¥ . . . . 4.19 Standard errorof ¥ . . . . 5.30 Standard errorof ¥ . . . . 4.72 Middle (N=74) Middle (N=74) Middle (N=74) Information . . . ..... 753 7563 1.561 | Vocabulary .834 834 1.81 | Block Design . . . .... .768 .768 2.34 Block Design . . . .... 634 .869 1.66 | Information .829 909 2.15 | Picture Completion 612 .8562 1.56 Similarities . . ...... .680 904 1.22 | Digit Span .5636 .945 146 | Coding . ......... 483 923 1.48 Picture Completion .682 927 97 | Constant Term . . .... 45.90 Constant Term . . . . .. 46.44 Constant Term . . . . .. 46.27 Standard errorof VY . . . . 4.50 Standard errorof Y . . . | 4.84 Standard errorof Y . . . . 4.88 AGE AGE AGE 6-11 years (N=487) 6-11 years (N=487) 6-11 years (N=487) Vocabulary . ....... .740 .740 1.77 | Vocabulary . ....... 816 .816 1.92 | Block Design . . ..... .706 .706 2.21 Block Design . . . .... .620 .846 1.32 | Arithmetic . .... .725 903 1.71 | Picture Arrangement .693 .845 2.01 Picture Arrangement 676 .898 1.26 | Similarities . . . . . .761 952 139 | Coding .......... 665 913 1.54 Arithmetic . . . ..... 691 926 1.26 | Constant Term + . 46.67 Constant Term . . . . . . 41.49 Constant Term. . . . .. 42.85 Standard errorof Y . . . . 4.22 Standard errorof Y . . . . 5.32 Standard errorof Y . . . . 4.78 6-7 years (N=156) 6-7 years (N=156) 6-7 years (N=156) Information . . . ..... .745 .745 146 | Vocabulary . . ...... .802 .802 1.88 | Picture Arrangement 728 .728 2.10 Picture Arrangement 722 853 1.34 | Information . . .. .... .789 .895 203 Coding +s von www a .540 .858 1.48 Vocabulary . ....... .730 902 1.35 | Digit Span .647 940 1.60 | Block Design . . . . ... 713 925 1.78 Block Design . . . .. .. 634 934 1.29 | ‘Constant Term . 46.40 Constant Term . . . . . . 49.93 Constant Term . . . . . . 46.30 Standard errorof ¥ . . . . 4.51 Standard errorof ¥ . . . . 4.68 Standard errorof Y . . . . 4.53 8-9 years (N=174) 8-9 years (N=174) 8-9 years (N=174) Vocabulary . ....... 779 779 2.13 | Vocabulary . ....... .858 .858 2.09 | Block Design . . ..... .708 .708 2.36 Block Design . . ..... 612 .873 153 | Arithmetic ........ 731 919 1.65 | Picture Arrangement .706 .845 1.92 Coding .......... .563 913 1.02 | Similarities . . ...... .783 951 1.36 [| Coding . . «ovo .687 .909 1.47 Picture Arrangement .681 938 1.20 | Constant Term v 45.53 Constant Term . . . . . . 41.95 Constant Term . . . . .. 39.11 Standard errorof ¥ . . . . 4.06 Standard errorof Y¥ . . . . 5.43 Standard errorof ¥ . . . . 4.50 10-11 years (N=157) 10-11 years (N=157) 10-11 years (N=157) Vocabulary . ....... 729 729 1.66 | Vocabulary .800 .800 2.33 | Object Assembly Yea # .808 .808 2.42 Object Assembly . . . . . 713 852 1.76 | Arithmetic 737 906 1.71 Picture Completion .686 .884 1.64 Arithmetic . . ...... 647 908 1.40 | Digit Span 540 .950 1.49 | Picture Arrangement .650 934 1.61 DigitSpan . ....... 503 931 1.17 | Constant Term . . .... 42.60 Constant Term. . . . .. 43.88 Constant Term . . .. .. 40.09 Standard errorof ¥ . . . . 3.46 Standard errorof Y . . . . 493 Standard errorof Y . . . . 4.57 standard error for the total sample was 4.78 1Q) points and ranged from 4.50 to 4.88 for various subgroups. Differences between multiple correla- tion coefficients for the total sample and those for subgroups were negligible for Verbal 1Q and Performance IQ). Vocabulary and Block Design were the two subtests which, together, accounted for the most variance in Full Scale IQ for the total sample of Mexican-Americans. Vocabulary consistently ap- peared as onc of the better predictors for the entire sample and for different subgroups based on socioeconomic status and age. The pattern for Block Design was less clear in the subgroups, but it emerged as the best performance test when all children were treated as a single group. Optimal Predictions Compared With Predictions Based on Vocabulary and Block Design Table 20 compares the multiple correlation coefficients based on Vocabulary and Block Design with the coefficients based on the two subtests yielding the highest correlation. In three comparisons, the optimal combination of two subtests was Block Design and Vocabulary, and in the three other comparisons differences ranged from .016 to .025. As in the case of Anglo children, we concluded that the two optimal tests for a short form of the WISC for Mexican-American children are Vocabulary and Block Design, the tests used in Cycle II. Discussion A view of the literature revealed no studies specifically investigating the use of short forms of the WISC in evaluating a sample of Mexican- American children, although there are investi- gators who have studied the possibility of using the Performance 1Q as a more valid indicator of intelligence in Mexican-American children than Full Scale IQ. Other investigators have found differences between Verbal and Performance 1Q similar to those found in the present study. For example, Altus compared the WISC scores of a group of bilingual children of Mexican descent with unilingual children of non-Mexican descent. All these children were being screened for special education classes. She found significant differences in favor of the unilingual children on Verbal IQ but no difference in Performance 1Q (Altus, 1953). Thompson found discrepancies in verbal performance to be significantly higher for a bilingual group of 60 children than for a control group of English-speaking children equated for age and Performance 1Q (Thomp- son, 1951). Table 20. Comparison of Block Design and Vocabulary as predictors with the two optimal predictors for Full Scale, Verbal, and Performance 1Q’s for Mexican-American children, by socioeconomic status and age Full Scale 1Q Verbal 1Q Performance 1Q R R R R R R Stat d Bisandiane Block Design Two Differ- | Block Design Two Differ- | Block Design Two Differ- and optimal ence and optimal ence and optimal ence Vocabulary predictors Vocabulary predictors Vocabulary predictors Status Low (N=413) . . . . .. .841 .841 .000 .825 .900 075 .761 .850 .089 Middle (N=74) . . . . . .844 869 .025 .850 909 .059 S92 .852 .060 Age Total, 6-11 years (N=487) . .. .846 .846 .000 .834 803 .069 .766 .845 .079 6-7 years (N=156) . . . .837 .853 .016 .825 .895 .070 .758 .858 .100 89 years (N=174 . . . . .873 .873 .000 .872 919 .047 .783 .845 .062 10-11 years (N=157) . . .832 .852 .020 B13 .906 .093 .761 .884 123 30 Holland conducted a study in which he administered the WISC bilingually to 36 Spanish-speaking children in first through fifth grades. This testing yielded an English Verbal IQ and a Bilingual Verbal IQ. He interpreted the difference in the two IQ’s as a measure of the language barrier. The mean English Verbal 1Q was 80.6, with a range of 45 to 118, while the mean Bilingual Verbal IQ was 85.2, with a range of 48 to 118, an average language barrier of 4.6 IQ points. In the same study, Holland found the mean Performance 1Q was 10.2 points higher than the mean English Verbal IQ (» <.01) and 5.6 points higher than the mean Bilingual Verbal IQ (p <.01) (Holland, 1960). The difference between Verbal and Performance 1Q’s is almost identical to that found in the Riverside sample. The difference between the mean Verbal IQ and mean Performance IQ was 9.4 for 6-year-old Mexican-American children, 10.4 for 7- year-olds, 9.2 for 8-year-olds, 8.7 for 9-year- olds, 9.4 for 10-year-olds, and 11.1 for 11- year-olds (table 20). The feasibility of using the Performance IQ as a short form was further explored by Cate, who administered the Performance subtests of the WISC and WAIS, depending upon the child’s age, to a sample of students enrolled in classes teaching English as a second language. He compared these Performance 1Q’s with IQ’s derived from Tests of General Ability admin- istered to grades 6 through 9; Raven’s Progres- sive Matrices; the Cattell Culture Fair Intelli- gence Test, Scale 2; and a verbal test in Spanish, the Test Rapido Barranguilla. There were no marked differences in the mean IQ’s obtained nor any significant difference between the mean scaled scores for Picture Completion, Picture Arrangement, Block Design, and Object Assem- bly. There was a considerable difference be- tween the mean for Coding and all other performance tests. Intercorrelations between subtests and the Performance scaled score ran a little higher than intercorrelations shown in the WISC Manual (Wechsler, 1949) for children age 13%, and the ranking of correlations for each subtest were similar (Cate, 1967). We found that the multiple correlation coef- ficients predicting Full Scale IQ from the scaled scores on Vocabulary and Block Design were .846 for Mexican-American children (with a standard error of 6.74 IQ points) as compared with R = .827 (with a standard error of 6.78 1Q points) for Negro children and R = .867 (with a standard error of 7.50 IQ points) for Anglo children (see tables 5, 11, and 17). Differences between the groups were inconsequential. Conclusions 1. Vocabulary and Block Design were the two subtests which provided the best prediction of Full Scale IQ for Mexican-American children in the Riverside sample. 2. Differences in multiple correlations be- tween Vocabulary and Block Design scaled scores and Full Scale IQ by age and socioeco- nomic status were minimal although there were sex differences in mean IQ for low-status Mexican-American children, there were socio- economic differences in mean IQ, and the mean Performance 1Q was consistently higher than the mean Verbal IQ for each age group. The regression equation based on all Mexican- American children was as efficient for predicting Full Scale IQ as the regression equations for subgroups categorized by age or socioeconomic level. 3. When Vocabulary and Block Design were used to predict Full Scale IQ of Mexican- American children, the regression equation was Predicted Full Scale 1Q = 53.49 + (2.64 X Vocabulary Scaled Score) + (1.98 X Block Design Scaled Score) with R = .846. Table IX in appendix III can be used to facilitate the calculation of estimated 1Q’s for individual children using scaled scores on Vocabulary and Block Design. 4. When Mexican-American children were classified as intellectually “normal” or “sub- normal” using IQ’s estimated from their scaled scores on Vocabulary and Block Design, 14 percent of the individual children were misclas- sified when 84- was the criterion, 10.5 percent were misclassified when 79- was the criterion, and 3.2 percent were misclassified when 69- was the criterion. Errors of this magnitude seem to indicate that two subtests were not sufficiently correlated with Full Scale IQ for identifying individual subnormals. 31 5. When estimates were concerned with the percentage of children falling below each of the three criteria, the error was reduced. As with Negro children, the error was always in the direction of underestimation—e.g., .8 percent at the 84- criterion; 3.9 percent at the 79- criterion; and 1.6 percent at the 69- criterion. GENERAL SUMMARY AND CONCLUSIONS The purpose of this report was to evaluate the use of two subtests of the WISC, Vocabulary and Block Design, as a short form for estimating the Full Scale IQ’s of children 6-11 years of age, with specific attention to their efficiency in differentiating children with subnormal IQs. Because of significant correlations of Full Scale, Verbal, and Performance 1Q’s with ethnic group and socioeconomic status, it was decided to conduct the analysis separately for Anglo, Negro, and Mexican-American children and to examine relationships by socioeconomic status and age for each group. The first finding was that girls from homes of lower socioeconomic status in all three ethnic groups did less well than low-status boys. Low- status boys had significantly higher Verbal 1Q’s in all ethnic groups. Differences were especially marked for Negro and Mexican-American chil- dren. Low-status Negro boys were significantly higher than low-status Negro girls on six of the 11 subtests, and Mexican-American boys were significantly higher than Mexican-American girls on seven of the 11 subtests. Sex differences tended to disappear for middle-status children (tables 4, 10, and 16). All scores tended to increase with social status for both Anglo boys and girls. The same pattern held for Negro and Mexican-American children, except that relatively few of the increases for boys were statistically significant whereas many of the increases for girls were significant. It appears that the disappearance of sex differences for middle-status Negro and Mexican-American children resulted primarily from the relatively greater improvement in the scores of girls with the increase in status. The intercorrelations of subtest scaled scores with each other and with Full Scale, Verbal, and Performance 1Q’s for the three ethnic groups were compared with comparable intercorrelation 32 matrices reported by Wechsler for children at ages 7% and 10% on whom the test was normed. The intercorrelations for 7-year-olds in the Riverside sample approximated the matrix pre- sented by Wechsler, but many of the intercorre- lations for the 10-year-olds in the sample were significantly lower—e.g., 26 of 89 for Anglo 10-year-olds; 27 of 89 for Negro 10-year-olds; and 33 of 89 for Mexican-American 10- year-olds. This indicates that the intercorrela- tions lor the Riverside sample, in general, may not have been as high as for Wechsler’s sample, which would influence the amount of error in predicting IQ from Vocabulary and Block De- sign in the Riverside sample. When Vocabulary and Block Design were used to predict Full Scale, Verbal, and Performance 1Q’s, it was found that estimates were not materially improved when each ethnic group was categorized by socioeconomic status or by age. Multiple correlation coefficients were approxi- mately the same for all three ethnic groups. Therefore, we concluded that Vocabulary and Block Design provide an essentially equivalent short form for predicting the 1Q’s for all three groups. However, when we focused specifically on the lower end of the distribution to determine the number of children correctly and incorrectly categorized as “subnormal” using three different criteria (IQ 84-, 79-, and 69-), more errors were made in categorizing individual Negro and Mexican-American children than in classifying individual Anglo children. Proportionately, about three times as many Negro and Mexican- American children were incorrectly classified using the 84~ criterion than Anglo children; six times as many using the 79- criterion; but less than twice as many using the 69- criterion. These differences were undoubtedly related to the fact that more Mexican-American and Negro children than Anglo children had 1Q’s below 85. It is doubtful that a short form of the WISC consisting of Vocabulary and Block Design should be used for anything but a crude initial individual screening, and then only at the traditional level for defining mental retardation (i.e, IQ 69-). While it might be used for Anglo children at the 79- criterion, the error is large for Negro and Mexican-American children, and the short form should probably not be used at that criterion for these groups. The 84- criterion had a relatively large percentage of classification errors for all three groups (tables 6, 12, and 18). However, false high IQ’s tended to balance false low IQ’s, so that predictions of rates are feasible for groups but not for individuals. Predictions for groups tended to underestimate rates of low IQ’s for both Negro and Mexican-American children. Multiple correlations were used to identify the optimal subtests for predicting Full Scale, Verbal, and Performance IQ’s for each ethnic group and for subclasses within ethnic groups based on socioeconomic status and age. Vocabu- lary and Block Design proved to be the two optimal subtests for Anglo and Mexican- American children. Vocabulary and Object As- sembly were the optimal dyad for Negro chil- dren; however, the difference in predictability when compared with Vocabulary and Block Design was minimal (tables 7, 13, and 19). When the correlation between Full Scale IQ and the optimal dyad was compared with the correlation between Full Scale IQ and the dyad consisting of Vocabulary and Block Design, Vocabulary and Block Design were the optimal dyad in three of seven categories, and differ- ences in multiple correlation coefficients ranged from .011 to .027 in the other four comparisons for Anglos (table 8). Vocabulary and Block Design were not the optimal dyad for predicting Full Scale IQ for any of the groups of Negro children, but differences in the multiple correla- tion coefficients between the optimal dyad and Vocabulary and Block Design were relatively small, ranging from .005 to .070 (table 14). For Mexican-American children Vocabulary and Block Design were the optimal dyad in three of the six categories of children, and differences ranged from .016 to .025 in the other three groups (table 20). When the empirical and theo- retical distributions for IQ were compared, the two matched exactly at the mean for Anglo and Mexican-American children—i.e., predicted IQ for scaled scores of 10 on Vocabulary and Block Design = 100, but predicted IQ for Negro children was one point lower, 99. We concluded that the choice of Vocabulary and Block Design as the two subtests to use in Cycle II of the Health Examination Survey was justified. No other dyad would have produced better overall predictions for all three ethnic groups in the Riverside sample. However, the multiple correlation coefficients when this dyad was correlated with Full Scale IQ were only .867, .827, and .846, with standard errors of 7.50, 6.78, and 6.74 1Q points for the three ethnic groups. This means that only approxi- mately 72 percent of the variance in Full Scale IQ was accounted for by these two subtests, leaving 28 percent of the variance unaccounted for. Prediction of group rates based on this dyad is feasible, but classification of individuals based on this or any other dyad of subtests should be done with discretion. Table X in appendix IV provides estimated full scale IQ’s predicted from the scaled scores of the Vocabulary and Block Design subtests using the multiple regression equation based on the entire sample of 1,310 children without regard for ethnic group. The multiple correlation coefficient of .880 was higher than for any individual ethnic group. The regression equation for the estimated Full Scale IQ was (2.57 X Vocabulary Scaled Score) + (2.05 X Block Design Scaled Score) + 53.15. The standard error of ¥Y was 7.09. Table XI in appendix IV presents estimated full scale IQ’s predicted from the sum of the scaled scores on Block Design and Vocabulary for the entire 1,310 children. A correlation coefficient of .880 was obtained and the regres- sion equation was: Estimated Full Scale 1Q = 52.51 + (2.3 X Sum Scaled Score on Block Design and Vocabulary) and the standard error of Y was 7.14. Thus, the Full Scale IQ for the entire sample can be predicted with about the same accuracy from the unweighted sum of the scaled scores as from the weighted scaled scores. Table XI also pre- sents separate predictions for males and females. The correlation for females of .889 is slightly higher than that of .871 for males. The female distribution exactly matched the theoretical distribution at the mean—i.e., predicted IQ for a sum of the scaled scores of Block Design and Vocabulary of 20 was 100. Table XII in appendix IV presents the percent- age of correct and incorrect predictions of Low Full Scale 1Q’s comparing the two methods of 33 predicting from Block Design and Vocabulary scaled scores. Using the 1Q 69- criterion, the than the weighted scaled scores (1.2 percent vs. 1.7 percent). When the 1Q 79- and 1Q 84- criteria are used, there was slightly less error in the weighted scaled scores based on the multiple regression equation. BIBLIOGRAPHY Altus, G. T.: WISC patterns of a selective sample of bilingual school children. J. Gen. Psychol. 83:241-248, 1953. Caldwell, M. B., and Smith, T. A.: Intellectual structure of southern Negro children. Psychol. Rep. 23:63-71, 1968. Carson, A. S., and Rabin, A. lL: communication in Negro and white children. J. Psychol. 51:47-51, 1960. Cate, C. C.: Test behavior of ESL students. California J. Educ. Res. 18:184-187, 1967. Clements, G. R.: An abbreviated form of the Wechsler Intelligence Scale for Children. J. Consul. Psych. 29:92, 1965. National Center for Health Statistics: Children’s Health Exami- nation Survey (Cycle II) Field Psychologist’s Manual. Public Health Service. Washington. U.S. Office. Enburg, R., Rowley, V. N., and Stone, B.: Short forms of the WISC for use with emotionally disturbed children. J. Clin. Psychol. 17:280-284, 1961. Guilford, J. P.: Fundamental Statistics in Psychology and Education. (4th ed), New York. McGraw-Hill, 1965. Holland, W. R.: Language barrier as an educational problem of Spanish-speaking children. Excep. Child. 27:42-50, 1960. Littell, W. M.: The Wechsler Intelligence Scale for Children: Review of a decade of research. Psychol. Bull. 57:132-156, 1960. McKerracher, D. W., and Watson, R. A.: Validation of a short form WISC with clinic children. Brit. J. of Educ. Psychol. 38:205-208, 1968. McNemar, Quinn: Psychological Statistics. New York. John Wiley & Sons, 1949. Mercer, Jane R: Sociological perspectives on mild mental Verbal comprehension and Educ. Government Printing retardation, in H.C. Hayward, ed., Socio-cultural Assets of Mental Retardation: Process of the Peabody-NIMH Confer- ence. New York. 1970, pp. 378-391. Mercer, Jane R.: Institutionalized Anglocentrism; Labeling Mental Retardates in the Public Schools, in Orleans, P., and Ellis, W. A., Jr., eds., Race, Change, and Urban Society, Vol. 5, Urban Affairs Annual Reviews. Beverly Hills, Calif. Sage Publications, Inc., 1971. Mercer, Jane R.: Who is normal? Two perspectives on mild mental retardation, in Jaco, E. Gartley, ed., Patients, Physi- cians and Illnesses, revised edition. New York. The Free Press. In press. Mercer, Jane R.: The Eligibles and the Labeled. In preparation. Appleton-Century-Crofts, Mumpower, D. L.: The fallacy of the short form. J. Clin. Psychol. 20:111-113, 1964. Nickols, J., and Nickols, M.: Brief forms of the WISC for research. J. Clin, Psvehol, 19:425, 1963. 34 Reiss, A. J., Jr.: Occupations and Social Status. Glencoe. The Free Press, 1961. Schwartz, L., and Levitt, E. E.: Short forms of the Wechsler Intelligence Scale for Children in the educable, noninstitutionalized mentally retarded. J. Educ. Psychol. 51:187-190, 1960. Seashore, H., Wesman, A., and Doppolt, J.: The standardization of the Wechsler Intelligence Scale for Children. J. Consult. Psychol. 14:99-110, 1950. Semler, 1. J., and Iscoe, L.: Structure of intelligence in Negro and white children. J. Educ. Psychol. 57:326-336, 1966. Silverstein, A. B.: Estimating Full Scale IQ's from WISC short forms. Psychol. Rep. 20:1264, 1967(a). Silverstein, A. B.: A short form of Wechsler’s scales for screening purposes. Psychol. Rep. 21:842, 1967(b). Silverstein, A. B.: A short form of the WISC and WAIS for screening purposes. Psychol. Rep. 20:682, 1967(c). Silverstein, A. B.: Validity of WISC short forms at three age levels. J. Consult. Psychol. 31:635-636, 1967(d). Silverstein, A. B.: Reappraisal of the validity of a short form of Wechsler’s Scales. Psychol. Rep. 26:559-561, 1970a. Silverstein, A. B.: Reappraisal of the validity of WAIS, WISC, and WPPSI short forms. J. Consult. and Clin. Psychol. 34:12-14, 1970b. Simpson, W. IL, and Bridges, C. C., Jr.: A short form of the Wechsler Intelligence Scale for Children. J. Clin. Psychol. 15:424, 1959. Teahan, J. E.: A comparison of Northern and Southern Negro children on the WISC. J. Consult. Psychol. 26:292, 1962. Thompson, G. M.: WISC patterns of a selective sample of dull bilingual children. Amer. Psychol. 493-494, 1951. Thompson, J. M., and Finley, C. J.: An abbreviated WISC for use with gifted elementary school children. California J. Educ. Res. 14:167-177, 1963. Wechsler, D.: WISC Manual, Wechsler Intelligence Scale for Children. New York. The Psychological Corporation, 1949. Wight, B. W., and Sandry, M.: A short form of the Wechsler Intelligence Scale for Children. J. Clin. Psychol. 18:166, 1962. Yalowitz, J. M., and Armstrong, R. G.: Validity of short forms of the Wechsler Intelligence Scale for Children (WISC). J. Clin. Psychol. 11:275-277, 1955. Young, F. M., and Bright, II. A.: Results of testing 81 Negro rural juveniles with the Wechsler Intelligence Scale for Children. J. Soc. Psychol. 39:219-226, 1954. Young, I. M., and Pitts, V. A.: The performance of congenital syphilitics on the Wechsler Intelligence Scale for Children. J. Consult. Psychol. 15:239-242, 1951. APPENDIX | ANGLO CHILDREN Table I. Intercorrelations of tests in the WISC for Anglo children aged 7: 62 boys and 38 girls Yast Infor- Compre: Arith- Similar- Vocab- Digit Hue Fond Block Object Coding Verbal Perform: ay mation hension metic ities ulary Span Hon Hiei Design | Assembly a a a Comprehension . . . ...... 50 Arithmetic... ........ 59 37 Similarities . . . ........ .54 34 51 Vocabulary . .......... 61 51 45 50 DigitBpan . .« vwvrwrwe vm 29 10 47 .35 214 Picture Completion . . . . . .. 42 2.09 .29 29 42 *.05 Picture Arrangement . . . . . . .35 18 .32 42 27 .26 24 Block Design. . ........ 42 33 A8 .39 43 .38 .28 .49 Object Assembly . . . . . ... 44 A? .46 33 41 .36 .35 .45 .56 Coding ws anemens rvs 41 15 42 24 .36 .34 a7 19 47 .34 Verbal lO ; .cimewis vive .74 52 67 62 .64 35 .38 41 .56 .50 .45 Performance IQ . . . ...... .59 27 57 49 55 .40 41 54 in .66 47 66 Full ScalelQ . ......... 66 233 .62 55 70 34 35 43 .62 .56 42 - - - Mean iQ ............ 102.5 106.1 | 104.3 Standard deviation . . . . . .. 14.8 15.7 15.2 All r's were corrected for spuriousness whenever a single test was correlated with a composite of which it is a contributing member. The same correction was used as that used by Wechsler (1949, p. 9), i.e., the correction suggested by McNemar (1949). 2 The correlation in the Anglo sample is significantly lower (p < .05) than that in the standardization sample. 3The correlation in the Anglo sample is significantly higher (p < .05) than that in the standardization sample. Table 11. Intercorrelations of tests in the WISC for Anglo children’ aged 10: 42 boys and 48 girls 757 Infor- | Compre: | Arith- | Similar- | Vocab- | Digit anne has Block | Object | (| Verbal Perform: ou mation hension metic ities ulary Span tion mer Design | Assembly a a a Comprehension . . . . . . ... 245 ARAIRHE «vise w vow 1.45 27 Similarities . . . ........ 2.51 47 243 Vocabulary . sv ws ws 5 win 67 2.60 49 64 DigtSpan wus v5 va ss 16 2 2.19 20 an Picture Completion . . . . . . . 49 .28 28 39 42 31 Picture Arrangement . . . . . . 2.28 2.00 22 2.06 227| 2.0 29 Block Design . . . ....... .50 32 .50 62 61 21 51 229 Object Assembly . . . . . ... 32 2.06 18 26 28 14 .39 42 .49 Coding . ............ 37 27 .26 40 .38 39 37 31 .39 43 Verbal IQ . . .......... ?.66 .58 252 66 .78 97 .50 fut 3 .66 29 48 Performance lQ . . . . ..... .56 2.28 a1 50 .56 32 .60 48 64 .64 3.56 61 Full Scale dQ .. . « wvowv ws » 2.62 2.36 2.44 .56 2.68 2.25 53 2.28 .68 44 .50 Mean ..wu wmws ma 108.28 110.67 | 109.83 Standard deviation . . . . . . . 14.13 16.29 | 15.20 All r's were corrected for spuriousness whenever a single test was correlated with a composite of which it is a contributing member. The same correction was used as that used by Wechsler (1949, p. 9), i.e., the correction suggested by McNemar (1949). 2 The correlation in the Anglo sample is significantly lower (p < .05) than that in the standardization sample. 3 The correlation in the Anglo sample is significantly higher (p < .05) than that in the standardization sample. 35 Table Ill. Estimated Full Scale 1Q from Block Design and Vocabulary scaled scores for Anglo children aged 6-11 Vocabulary Block Design 0 1 2 3 4 5 6 2 8 9 10 11 12 13 14 15 16 17 18 19 | 20 Oc rvmn t namn se #0 54 | 56 59 61 64 66 68 7m 73 76 78 81 83 86 88 91 93 85 98 | 100 | 103 Tovwme ve nmin vam 56 | 58 61 63 66 68 71 73 76 78 80 83 85 88 90 93 95 98 | 100 | 103 | 105 he a bhi n Bd: Bu 58 | 61 63 65 68 70 73 75 78 80 83 85 87 90 92 95 97 | 100 | 102 | 105 | 107 Biv omue cma vow 60 | 63 65 68 70 72 75 77 80 82 85 87 90 92 95 97 99 | 102 | 104 | 107 | 109 4 62 | 65 67 70 72 75 77 80 82 84 87 89 92 94 97 99 | 102 | 104 | 107 | 109 | 111 Bi cwma so smn ov wa 65 | 67 69 72 74 77 79 82 84 87 89 92 94 96 99 | 101 | 104 | 106 | 109 [ 111 | 114 Bo vnmn rome ron 67 | 69 72 74 77 79 81 84 86 89 91 94 96 99 | 101 103 | 106 | 108 | 111 113 | 116 Vismme awnmias vow 69 | 71 74 76 79 81 84 86 88 91 93 96 98 | 101 | 103 | 106 | 108 | 111 | 113 | 115 | 118 8... 71 | 73 76 78 81 83 86 88 9 93 96 98 | 100 | 103 | 105 | 108 | 110 | 113 | 115 | 118 | 120 Dn vin ow sme vo 73 | 76 78 81 83 85 88 90 a3 95 98 | 100 | 103 | 105 | 107 | 110 | 112 | 115 | 117 | 120 | 122 1 sins Gomi 4 £0 75 | 78 80 83 85 88 90 92 95 97 | 100 | 102 | 105 | 107 | 110 | 112 | 115 | 117 | 119 | 122 | 124 1 suns smn us ww 77 | 80 82 85 87 90 92 95 97 | 100 | 102 | 104 | 107 | 109 | 112 | 114 | 117 | 119 | 122 | 124 | 127 12 0. 80 | 82 85 87 89 92 94 97 99 | 102 | 104 | 107 | 109 | 111 114 | 116 | 119 | 121 | 124 | 126 | 129 13 vwws vas smn 82 | 84 87 89 92 94 96 99 | 101 104 | 106 | 109 | 111 [ 114 [116 | 119 | 121 123 | 126 | 128 | 131 M4 84 | 86 89 91 94 96 99 | 101 104 | 106 | 108 | 111 113 | 116 | 118 | 121 123 | 126 | 128 | 131 | 133 IB coin v wen x wma 86 | 89 og 23 96 98 | 101 | 103 | 106 | 108 | 111 113 | 116 | 118 | 120 | 123 | 125 | 128 | 130 | 133 | 135 18 cua sas £45 88 | 91 93 96 98 ( 101 | 103 | 105 | 108 | 110 | 113 | 115 | 118 | 120 | 123 | 125 | 127 | 130 | 132 | 135 | 137 VT ois 5 cms wx ems 90 | 93 95 98 | 100 | 103 | 105 | 108 | 110 | 112 | 115 | 117 [| 120 | 122 | 125 | 127 | 130 | 132 | 135 | 137 | 139 18 wis s 08 3 00 93 | 95 97 | 100 | 102 | 105 | 107 | 110 | 112 | 115 | 117 | 120 | 122 | 124 | 127 | 129 | 132 | 134 | 137 | 139 | 142 19 0. 95 | 97 | 100 | 102 | 105 | 107 | 109 | 112 | 114 | 117 | 119 | 122 | 124 | 127 | 129 | 131 134 | 136 | 139 | 141 | 144 20 pws po rmgp ow mos 97 | 99 | 102 | 104 | 107 | 109 ( 112 | 114 | 116 | 119 | 121 124 | 126 | 129 | 131 134 | 136 | 139 | 141 143 | 146 NOTE: Estimated 1Q = (2.45 X Vocabulary scaled score) + (2.15 X Block Design scaled score) + 53.77. Standard error = 7.50. 36 APPENDIX II Table IV. Intercorrelations of tests in the WISC for Negro children’ aged 7: NEGRO CHILDREN 31 boys and 33 girls ro re | Goma | to | Str | ot | 91 | opt: | are: | Sk || Oo | og | Venn | "| Co tion ment Q a Comprehension . . . ...... 27 Arithmetic . .......... .38 18 Similarities... 005 00 28 31 27 Vocabulary . . ......... 57 49 .30 55 Digit Span : ; iu: ra sm ies 39 32 .38 .39 43 Picture Completion . . . . . . . 14 30 | *-.06 18 .30 .18 Picture Arrangement . . . . . . .256 34 39 34 51 34 .24 Block Design . . ........ .23 2-02 .27 19 14 .23 24 .29 Object Assembly . . . ..... .28 .16 .16 27 .35 .23 23 35 a2 Coding . ............ 37 14 .05 27 .38 23 35 .10 72 .38 Verbal IQ . ........... 55 43 43 51 .70 54 .25 52 .25 .34 .30 Performance 1Q . . . ...... 34 .30 .28 41 65 40 48 47 37 49 43 54 Full ScalelQ .......... 42 33 231 45 .65 43 31 47 224 .38 .30 MeanilO oh ves was 93.056 96.48 | 93.76 Standard deviation . . ..... 11.47 11.84 | 11.28 "All r's were corrected for spuriousness whenever a single test was correlated with a composite of which it is a contributing member. The same correction was used as that used by Wechsler, (1949, p. 9), i.e., the correction suggested by McNemar (1949). The correlation in the Negro sample is significantly lower (p < .05) than that in the standardization sample. Table V. Intercorrelations of tests in the WISC for Negro children' aged 10: 25 boys and 23 girls re mn | mi | | S| a | S| com | ar | 2 | 20% | cag | Vig | Tce” | ce tion ment 1Q 1Q Comprehension . . . . ..... 59 Arithmetic . . . ........ 2.47 53 Similarities . . ......... 53 54 .50 Vocabulary . . ios «0500 81 62 51 49 DigitSpan . .......... 39 .30 47 48 34 Picture Completion . . . . .. . 2.00 1-14 2-19 2.08 202] -.03 Picture Arrangement . . . . . . 2.23 210 ?-.03 a7 .36 14 26 Block Design . . ........ %12 2-14 2-1 15 ?.26 .02 2.10 fo 1 4 Object Assembly . . . ..... 19 02 2.01 14 .30 14 44 40 58 CONG «cv vosmn omen 48 .38 .65 24 51 34 -.08 2-.06 .08 .07 Verbal 1Q . . . vows views I 65 62 64 .70 47 2-10 2.20 2.056 A7 54 Performance dQ ... . «savin .35 2.08 2.09 29 50 21 .35 41 50 .70 .08 2.30 FullScale IQ « «ws &mv ms 61 237 2.36 .46 J2 .35 2.01 2.25 2.18 .36 .33 Mean dQ wu ows swsma 95.83 94.38 | 94.90 Standard deviation . . ..... 1277 12.02 | 10.95 ' All r's were corrected for spuriousness whenever a single test was correlated with a composite of which it is a contributing member. The same correction was used as that used by Wechsler, (1949, p. 9), i.e., the correction suggested by McNemar (1949). 2The correlation in the Negro sample is significantly lower (p < .05) than that in the standardization sample. 37 Table VI. Estimated Full Scale |1Q from Block Design and Vocabulary scaled scores for Negro children aged 6-11 Vocabulary Block Design 0 1 2 3 4 5 6 7 8 9 10 1 12 13 14 15 16 17 18 19 | 20 Bis cs amin os wma» sin v 65 | 68 | 60 | 63 | 65 68 70 73 75 78 81 83 86 88 91 93 96 98 | 101 | 104 | 106 To 57 | 59 | 62 | 64 | 67 70 72 75 77 80 82 85 87 90 92 95 98 | 100 | 103 | 105 | 108 Disp nme BEEPS BEE 8 58 | 61 | 64 | 66 | 69 71 74 76 79 81 84 87 89 92 94 97 99 ( 102 | 104 ( 107 | 110 Smee rma vn 60 | 63 | 65 | 68 | 70 73 76 78 81 83 86 88 91 93 96 98 | 101 | 104 | 106 | 109 | 111 EF 62 | 64) 67 ( 70 | 72 75 77 80 82 85 87 90 a3 95 98 | 100 ( 103 | 105 | 108 | 110 | 113 Biv vo nim vo mm on ows 64 | 66 | 69 | 71 | 74 76 79 81 84 87 89 92 94 97 99 | 102 | 104 | 107 | 110| 112 | 115 GB: ins Ema mae 65 | 68) 70 | 73 | 76 78 81 83 86 88 91 93 96 98 | 101 | 104 | 106 | 109 | 111 | 114 | 116 Tv vcmn nonin oy wins 67 | 70 | 72 | 75 | 77 80 82 85 87 90 93 95 98 100 | 103 105 108 110 | 113 116 | 118 Bc LEE AMEE REE a 69 | 71 74 | 76 | 79 82 84 87 89 92 94 97 99 102 104 107 el 12) Ns 117 120 G': wows ox wma ow pawn 70 | 73 | 76 | 78 | 81 83 86 88 91 93 96 99 101 104 | 106 109 111 114 | 116 119 121 TO ; ums 2s ews 2a 2 72 | 75| 77 | 80 | 82 85 87 90 93 95 98 100 103 105 108 110 113 | 116 | 118 121 123 TY iis mmm oh man 74 | 76 | 79 | 82 | 84 87 89 92 94 97 99 | 102 | 105 | 107 | 110 | 112 | 115 | 117 | 120 | 122 | 125 T2 wos s twa cans 76 | 78 | 81 83 | 86 88 91 93 96 99 101 104 106 109 111 114 | 116 119 122 124 127 13 0 77 | 80| 82 | 85 | 88 90 93 95 98 | 100 | 103 | 105 108 110 | 113 | 116 118 121 123 126 | 128 4 wm ow mms kx para 79 | 82 | 84 | 87 | 89 92 94 97 99 | 102 | 105 | 107 | 110 | 112 | 115 117 | 120 | 122 | 1256 | 127 | 130 15 ©. oL 81 83 | 86 | 88 | 91 93 96 99 101 104 106 109 11 114 116 119 122 124 127 129 132 T6 ve swmn » sma « 82 | 85| 88 | 90 | 93 25 98 100 103 | 105 108 1m 113 116 118 121 123 | 126 128 131 133 17 snmp ams 2 ime 3 84 ( 87) 89 | 92 | 94 97 99 | 102 | 105 | 107 | 110 | 112 | 115 | 117 | 120 | 122 | 125 | 128 | 130 | 133 | 135 18 86 | 88 | 91 94 | 96 99 101 104 106 109 111 114 116 119 122 | 124 127 129 132 134 137 18 www s goes 505 3 88 | 90| 93 | 95 | 98 | 100 103 | 105 108 111 113 116 | 118 121 123 126 128 131 133 136 | 139 20 LL 89 | 92| 94 | 97 | 99 | 102 105 107 110 | 112 115 17 120 122 125 128 130 | 133 | 135 138 | 140 NOTE: Estimated 1Q = (2.55 X Vocabulary scaled score) + (1.71 X Block Design scaled score) + 55.07. Standard error = 6.78. R = .827. 38 APPENDIX IIIT MEXICAN-AMERICAN CHILDREN Table VII. Intercorrelations of tests in the WISC for Mexican-American children’ aged 7: 53 boys and 43 girls Tos ms | re | | Se | oc | 25 Gomi: | ava | ik | 90k cong | Vo | "ine | nie tion ment a 1Q Comprehension . . . ...... 48 Arithmetic © vo. wv vu iowa 51 20 Similarities . . ......... .69 .35 .34 Vocabulary . .......... 64 53 41 54 DigitSpan . .......... 37 27 52 40 .37 Picture Completion . . . . .. . 241 28 .36 29 44 16 Picture Arrangement . . . . . . A3 35 34 50 35 43 .30 Block Design . ......... .33 27 40 31 .35 45 23 .50 Object Assembly . . ...... .35 29 .38 28 .38 2.46 22 .40 .50 Coding ............. 16 ’-.09 3.08 -.01 02 | 2-01 a3 3.04 15 .09 Verbal dQ . oo: wu swiswn vn 73 49 .50 .60 .70 51 44 .55 48 48 2.03 Performance IQ . . . . . . . .. 52 34 50 43 48 48 42 52 63 65 .20 62 Full ScalelQ. . ......... 64 37 49 48 .59 49 .40 52 .50 .50 4-02 - - Mean dQ. ...w5 cwsws wn 88.6 98.9 92.7 Standard deviation . . .. ... 12.9 12.5 12.3 "All r's were corrected for spuriousness whenever a single test was correlated with a composite of which it is a contributing member. The same correction was used as that used by Wechsler, (1949, p. 9), i.e., the correction suggested by McNemar (1949). ? The correlation in the Mexican-American sample is significantly higher (p < .05) than that in the standardization sample. ® The correlation in the Mexican-American sample is significantly lower (p < .05) than that in the standardization sample. * The correlation in the Mexican-American sample is significantly lower (p < .01) than that in the standardization sample. Table VIII. Intercorrelation of tests in the WISC for Mexican-American children’ aged 10: 39 boys and 49 girls ros a, | Some | | Sar | oc | 288 | om: | are | St | 9001 | cong | Ve | oe” | cme tion ment Q [[e} Comprehension . . . ...... 235 Arithmetic. . . ........ 353 31 Similarities . . ......... 3.50 40 3.39 Vocabulary . . ......... .68 a 46 62 DigitSpan . . ......... 41 a3 23 a3 3.15 Picture Completion . . . . . . . 2 a7 12 16 233 -.05 Picture Arrangement . . . . . . 3.22 2.16 324 7 2.34 14 27 Block Design . . ........ .36 ci 27 .20 2.29 .28 .26 2.28 Object Assembly . . . ..... 39 *.10 .36 24 40 18 .34 .40 61 Coding « «sais cwmsma sn 41 3-05 .37 14 24 11 .06 210 .10 .25 Verbal ld «vc ws smsws un 73 2.41 .55 .60 an 29 .24 252 .37 42 .30 Performance IQ . . . . ..... 49 2.16 41 29 350 .20 41 46 .55 *.70 22 251 Full'ScalelQ un: cnvwamy 263 2.22 247 2.39 2.62 *.20 2.26 334 2.43 55 22 - - MeanilQ .. ..::ioomsas 82.2 97.6 91.6 Standard deviation . . .. ... 109 12.4 na "All r's were corrected for spuriousness whenever a single test was correlated with a composite of which it is a contributing member. The same correction was used as that used by Wechsler (1949, p. 9), i.e., the correction suggested by McNemar (1949). 2 The correlation in the Mexican-American sample is significantly lower (p < .01) than that in the standardization sample. The correlation in the Mexican-American sample is significantly lower (p < .05) than that in the standardization sample. * The correlation in the Mexican-American sample is significantly higher (op < .05) than that in the standardization sample. 39 Table |X. Estimated Full Scale 1Q from Block Design and Vocabulary scaled scores for Mexican-American children aged 6-11 Vocabulary Block Design 0 1 2 3 4 5 6 7 8 9 10 1" 12 13 14 15 16 17 18 19 | 20 0 wns smms 3 mn 53 | 56 | 59 61 64 67 69 72 75 7 80 82 85 88 90 93 96 98 | 101 104 | 106 Tsim ween sme 65 | 58 | 61 63 66 69 71 74 77 29 82 84 87 90 92 95 98 | 100 | 103 | 106 | 108 2 swws paws g awe 57 | 60 | 63 65 68 7 73 76 79 81 84 86 89 92 94 97 | 100 | 102 | 105 | 108 | 110 3 69 | 62 | 65 67 70 73 75 78 81 83 86 88 91 94 96 99 | 102 | 104 | 107 | 109 | 112 A cnmw seme 3 Fes 61 | 64 | 67 69 72 75 77 80 82 85 88 90 93 96 98 | 101 | 104 | 106 | 109 | 111 | 114 Lp 63 | 66 | 69 7 74 77 79 82 84 87 90 92 95 98 | 100 | 103 | 106 | 108 | 111 113 | 116 8 Lassies seas 65 | 68 | 71 73 76 79 81 84 86 89 92 94 97 | 100 | 102 | 105 | 108 | 110 | 113 | 115 | 118 Town vs vmwin wwma 67 | 70 | 73 75 78 80 83 86 88 91 94 96 99 102 104 107 109 112 115 117 120 B iuiussmmastans 69 | 72 | 75 77 80 82 85 88 90 93 96 98 | 101 | 104 | 106 | 109 | 111 | 114 | 117 | 119 | 122 O sis c nmin vim an 7M 74 | 77 79 82 84 87 90 92 95 98 100 | 103 106 108 11 113 | 116 | 119 121 124 VO ww 2 smn s swims 73176 | 79 81 84 86 89 a2 94 97 | 100 | 102 | 105 | 108 | 110 | 113 | 115 | 118 | 121 123 | 126 Mm. 75 | 78 | 80 83 86 88 91 94 96 99 | 102 | 104 | 107 | 109 | 112 | 115 | 117 | 120 | 123 | 125 | 128 V2 sin 5 pimm se wman 77 | 80 | 82 85 88 90 93 96 98 | 101 104 | 106 | 109 | 111 | 114 | 117 | 119 | 122 | 125 | 127 | 130 IB sms 5 slum ik smd 79 | 82 | 84 87 90 92 95 98 100 | 103 | 106 108 | 111 113 | 116 19 | 121 124 | 127 129 132 18 vv 2 cma wv mwa 81 | 84 | 86 89 92 94 97 | 100 | 102 | 105 | 108 | 110 | 113 | 115 | 118 | 121 | 123 | 126 | 129 | 131 | 134 i Fr FER EERE 83 | 86 | 88 91 94 96 99 102 104 107 109 112 115 117 120 123 125 128 | 131 133 136 16... oo... 85 | 88 | 90 93 96 98 101 104 106 109 11 114 117 119 122 125 127 130 | 133 135 138 17 svn 5 sews sama 87 | 90 | 92 95 98 | 100 | 103 | 106 | 108 | 111 | 113 | 116 | 119 | 121 | 124 | 127 | 129 | 132 | 135 | 137 | 140 8 viv 2 winin wow 89 | 92 | 94 97 100 | 102 105 108 7110 | 113 115 | 118 121 123 126 129 | 131 134 | 136 138 142 19 m0 5s yws 2B WE Nn 94 | 96 99 102 | 104 107 109 MZ 118 17 120 123 125 128 131 133 | 136 | 138 141 144 20... 93 | 96 | 98 | 101 | 104 | 106 | 109 | 111 | 114 | 117 | 119 | 122 | 125 | 127 | 130 | 133 | 135 | 138 | 140 | 143 | 146 NOTE: Estimated 1Q = (2.64 X Vocabulary scaled score) + (1.98 X Block Design scaled score) + 53.49. Standard error = 6.74. R = 846. 40 APPENDIX IV ALL SAMPLE CHILDREN Table X. Estimated Full Scale 1Q from Block Design and Vocabulary scaled scores for 1,310 children aged 6-11 Vocabulary Block Design 0 1 2 3 4 5 6 7 8 9 10 1 12 13 14 15 16 17 18 19 | 20 O imme ems vu 53 | 66 | 58 61 63 66 69 n 74 76 79 81 84 87 89 92 94 97 99 | 102 | 104 1 o covms & Blind BF 65 | 68 | 60 63 65 68 Al 73 76 78 81 83 86 89 91 94 96 99 | 101 | 104 | 107 Z ccaumwnr ams vo 57 | 60 | 62 65 68 70 73 75 78 80 83 85 88 fn 93 96 98 | 101 | 103 | 106 | 109 Bl wrmins Fin bw 59 | 62 | 64 67 70 72 75 77 80 82 85 88 90 93 95 98 | 100 | 103 | 105 | 108 | 111 Bown ems 61 | 64 | 66 69 72 74 7 79 82 84 87 90 92 95 97 | 100 | 102 | 105 | 108 | 110 | 113 8 scowis emus sw 63 | 66 | 69 IA 74 76 79 81 84 86 89 92 94 97 99 | 102 | 104 | 107 | 110 [ 112 | 115 B in mm ow 65 | 68 | 71 73 76 78 81 83 86 89 91 94 96 99 | 101 | 104 | 107 | 109 | 112 | 114 | 117 T scams s smms 50 67 | 70 | 73 75 78 80 83 85 88 91 93 96 98 | 101 | 103 | 106 | 109 | 111 | 114 | 116 | 119 B iwmev swmn ww 70 [72 | 75 77 80 82 85 88 90 93 95 98 | 100 | 103 | 105 | 108 | 111 | 113 | 116 | 118 | 121 .: EE I FI: 72 | 74 | 77 79 82 84 87 90 92 95 97 | 100 | 102 | 105 | 108 | 110 | 113 | 115 | 118 | 120 | 123 10: civwn v comma ow 74 | 76 | 79 81 84 86 89 92 94 97 99 | 102 | 104 | 107 | 110 [ 112 | 115 | 117 | 120 | 122 | 125 MM aiwmns tmms v4 76 | 78 | 81 83 86 89 91 94 96 99 | 101 | 104 | 106 | 109 | 112 | 114 | 117 | 119 | 122 | 124 | 127 122 swwn v« vem vw 78 | 80 | 83 85 88 91 93 96 98 | 101 | 103 | 106 | 109 [ 111 | 114 | 116 | 119 | 121 | 124 | 126 | 129 13... iii 80 | 82 | 85 87 90 93 95 98 | 100 | 103 | 105 | 108 | 111 | 113 | 116 | 118 | 121 | 123 | 126 | 129 | 131 18: sump v 2mm vu 82 | 84 | 87 90 92 95 97 | 100 | 102 | 105 | 107 [ 110 | 113 | 115 | 118 | 120 | 123 | 126 | 128 | 131 | 133 15... 0 84 | 86 | 89 92 94 97 99 | 102 | 104 | 107 | 110 | 112 | 115 | 117 | 120 | 122 | 125 | 128 | 130 | 133 | 135 Vy smmn 2 swms o 86 | 88 | 91 94 96 99 | 101 | 104 | 106 | 109 | 112 | 114 | 117 | 119 | 122 | 124 | 127 | 130 | 132 | 135 | 137 17 oo 88 [91 | 93 96 98 | 101 | 103 | 106 | 109 | 111 | 114 | 116 | 119 | 121 | 124 | 126 | 129 | 132 | 134 | 137 | 139 1B: 3 vn s smms 24 90 | 93 | 95 98 | 100 | 103 | 105 | 108 [ 111 | 113 | 116 | 118 | 121 | 123 | 126 | 129 | 131 | 134 | 136 | 139 | 141 19 + sven « www wi 92 [95 | 97 | 100 | 102 | 105 [ 107 | 110 | 113 | 115 | 118 | 120 | 123 | 125 | 128 | 131 | 133 | 136 | 138 | 141 | 143 200: ann a sms vo 94 | 97 [99 | 102 | 104 | 107 | 110 | 112 | 115 | 117 | 120 | 122 | 126 | 127 | 130 | 133 | 135 | 138 | 140 | 143 | 145 NOTE: Estimated 1Q = (2.57 X Vocabulary scaled score) + (2.05 X Block Design scaled score) + 53.15. Standard error = 7.09. R = .880. 41 Table XI. Estimated Full Scale 1Q using the sum of the scaled scores for Block Design and Vocabulary for all 1,310 children in the sample and for males and females separately Estimated Full Scale 1Q Estimated Full Scale 1Q Sum of scaled scores—Block Sum of scaled scores—Block Design and Vocabulary Total Males Females Design and Vocabulary Total Males | Females sample sample 0 sms sms aus m Es BEE 53 53 BZ 121 0 io 230m 95 5845 # 101 101 102 1 ee 55 55 B54 | 22 . i. te toms meena 104 103 105 TF nl dm adie Bmw mm 57 57 56 | 23 ... 106 105 107 3 59 60 B89 128 ws ws ms mr wr ome ews 108 108 109 Hows umm mS ES EER 62 62 61 28 Lin im RIB EEE EB 11 110 112 B isms Ei mime 64 64 B48 | 28 sims ms 3 a8 45.45 5 113 112 114 6 66 67 66 | 27 ... 115 115 117 TZ in we mew ow wm 69 69 69 [28 .... 118 117 119 8 ee 71 & 71 20 Lia mies me we ows 120 119 121 EE EE EE RL TE 73 73 B30 wiwswimswsmn ips 122 121 124 ; i 1 HEE EE EE EE 76 76 78 181 sina sssmime $i 125 124 126 TE ed mH EB osm tower 78 78 IB 182 .isiimemnmiwn nme 127 126 129 FZ 2 fie in nln do aon war sa Bm 80 80 81 BB rm mame mn ome mow 129 128 131 13 83 83 83 |3 ..... 132 131 133 TE ihm wm ww 85 85 BB [3B .wiwiwimswenms ows 134 133 136 MB Lvs wsmewswrmasmsas 87 87 B8 [3 .::n:wviuswims ww 136 135 138 18 ov 25 swims mv sms a 90 89 90 [37 ...0:¢0:ms@2m5 8 m3 139 137 141 7 ws sms me Ss Himes 4B 92 92 93 [38 .v:s:iuisimsms ens 141 140 143 18 is sas BI EE mw bm 6 94 94 OB 1 89 :u:oitomimsds ome 143 142 145 FO oles oe wlan Bt a how ales a 97 96 97 {40 .... sma oes 146 144 148 20 Lee 99 99 100 NOTE: Estimated Full Scale 1Q for all children = 52.51 + (2.33 X sum scaled scores in Block Design and Vocabulary); standard error of Y =7.14; r = 880. Estimated Full Scale 1Q for males = 52.92 + (2.28 X sum scaled scores on Block Design and Vocabulary); standard error of Y =7.34; r=.871. Estimated Full Scale 1Q for females = 51.68 + (2.40 X sum scaled scores in Block Design and Vocabulary); standard error of Y = 6.90; r = .889. Table X11. Percent of correct and incorrect predictions of low Full Scale 1Q’s using two methods of predicting from Block Design and Vocabulary scaled scores for 1,310 children, aged 6-11 using three different criteria Correctly Fase False Correctly Actual Predicted Method and criteria identified | Qs! high 1Q’s? identified percent percent Difference low 1Q’'s ow 9 high 1Q’s low 1Q’'s low IQ's Based on Multiple Percent regression equation 1Q84- . . 12.8 4.4 7.1 75.7 19.9 17.2 -2.7 IQ79- . . . eee 6.8 23 4.8 86.1 11.6 9.1 -2.5 1Q69- . . 7 .5 22 96.6 2.9 1.2 -1.7 Based on the regression on sum of scaled scores 1Q84- . . . 12.9 4.0 7.0 76.1 19.9 16.9 -3.0 1Q79- . . eee 6.3 2.72 5.3 86.2 11.6 8.5 -3.1 QB . ws v5 sw swvmsmsms ns a 1.0 7 1.9 96.4 29 1.7 -1.2 ! False low 1Q’s were for children who were predicted to have an 1Q below the criteria but who actually had an 1Q above the criteria. 2 False high 1Q’s were for children who were predicted to have an 1Q above the criteria but who actually had an 1Q below the criteria. 42 ¥ U. S. GOVERNMENT PRINTING OFFICE : 1972 482-005/17 Series 1. Series 2. Series 3. Series 4. Series 10. Series 11. Series 12. Series 13. Series 14. Series 20. Series 21. Series 22, VITAL AND HEALTH STATISTICS PUBLICATION SERIES Formerly Public Health Service Publication No. 1000 Programs and collection procedures.— Reports which describe the general programs of the National Center for Health Statistics and its offices and divisions, data collection methods used, definitions, and other material necessary for understanding the data. Data evaluation and methods reseavch.— Studies of new statistical methodology including: experi- mental tests of new survey methods, studies of vital statistics collection methods, new analytical techniques, objective evaluations of reliability of collected data, contributions to statistical theory. Analytical studies. —Reports presenting analytical or interpretive studies basedon vital and health statistics, carrying the analysis further than the expository types of reports in the other series, Documents and committee veports.—Final reports of major committees concerned with vital and health statistics, and documents such as recommended model vital registration laws and revised birth and death certificates, Data from the Health Interview Survev.— Statistics on illness, accidental injuries, disability, use of hospital, medical, dental, and other services, and other health-related topics, based on data collected in a continuing national household interview survey. Data from the Health Examination Survey.—Data from direct examination, testing, and measure- ment of national samples of the civilian, noninstitutional population provide the basis for two types of reports: (1) estimates of the medically defined prevalence of specific diseases in the United States and the distributions of the population with respect to physical, physiological, and psycho- logical characteristics; and (2) analysis of relationships among the various measurements without reference to an explicit finite universe of persons. Data from the Institutional Population Surveys. — Statistics relating tothe health characteristics of persons in institutions, and their medical, nursing, and personal care received, based on national samples of establishments providing these services and samples of the residents or patients, Data from the Hospital Discharge Survey.— Statistics relating to discharged patients in short-stay hospitals, based on a sample of patient records in a national sample of hospitals. Data on health resources: manpower and facilities, —Statistics on the numbers, geographic disici- bution, and characteristics of health resources including physicians, dentists, nurses, other health occupations, hospitals, nursing homes, and outpatient facilities. Data on mortality.—Various statistics on mortality other than as included ir regular annual or monthly reports—special analyses by cause of death, age, and other demographic variables, also geographic and time series analyses. Data on natality, marriage, and divorce.—Various statistics on natality, marriage, and divorce other than as included in regular annual or monthly reports—special analyses by demographic variables, also geographic and time series analyses, studies of fertility. Data from the National Natality and Mortality Surveys.— Statistics on characteristics of births and deaths not available from the vital records, based on sample surveys stemming from these records, including such topics as mortality by socioeconomic class, hospital experience in the last year of life, medical care during pregnancy, health insurance coverage, etc. For a list of titles of reports published in these series, write to: Office of Information National Center for Health Statistics Public Health Service, HSMHA Rockville, Md, 20852 DHEW Publication No. (HSM) 72-1047 ~~ Series 2-No. 47 CYC Number 48 VITAL and HEALTH STATISTICS DATA EVALUATION AND METHODS RESEARCH ae RS) (A rd Le) NCHS EA © (CR CLT (EH GOT In the Health Interview Survey U.S. DEPARTMENT OF HEALTH, EDUCATION, AND WELFARE Public Health Service Health Services and Mental Health Administration Vital and Health Statistics-Series 2-No. 48 For sale by the Superintendent of Documents, U.S. Government Printing Office, Washington, D.C. 20402 - Price $1.00 Data Evaluation and Methods Research Series 2 Number 48 Interviewing Methods In the Health Interview Survey A comparison of the person and condition approaches in the collection of condition data in a household interview. —— Pea NAM IMENT i DAT . DUCUMENTS ULPARTMENT JIJN 1 qQ 1 ) - YJ N i J 13(2 LiGininy UNIVERSITY OF CALIFORNIA — DHEW Publication No. (HSM)-72-1048 U.S. DEPARTMENT OF HEALTH, EDUCATION, AND WELFARE Public Health Service Health Services and Mental Health Administration National Center for Health Statistics Rockville, Md. April 1972 NATIONAL CENTER FOR HEALTH STATISTICS THEODORE D. WOOLSEY, Director PHILIP S. LAWRENCE, Sc.D., Associate Director OSWALD K. SAGEN, Ph.D., Assistant Director for Health Statistics Development WALT R. SIMMONS, M.A., Assistant Director for Research and Scientific Development JAMES E. KELLY, D.D.S., Dental Advisor EDWARD E. MINTY, Executive Officer ALICE HAYWOOD, Information Officer DIVISION OF HEALTH INTERVIEW STATISTICS ELIJAH L. WHITE, Director ROBERT R. FUCHSBERG, Deputy Director RONALD W. WILSON, Chief, Analysis and Reports Branch KENNETH W. HAASE, Chief, Survey Methods Branch COOPERATION OF THE BUREAU OF THE CENSUS Under the legislation establishing the National Health Survey, the Public Health Service is authorized to use, insofar as possible, the services or facilities of other Federal, State, or private agencies. In accordance with specifications established by the Health Inter- view Survey, the Bureau of the Census, under a contractual arrangement, participates in most aspects of survey planning, sclects the sample, collects the data, and carries out certain parts of the statistical processing. Vital and Health Statistics-Series 2-No. 48 DHEW Publication No. (HSM) 72-1048 Library of Congress Catalog Card Number 74-169282 CONTENTS Introduction . . . . . . LL... Background of the Health Interview Survey . . . . ............. Initiation of the Interview Survey . . . . . ................ Concepts Used in the Survey . . . . . .................. Data Collection in the Health Interview Survey . . . . . .......... Format of the First Questionnaire . . . . . ................ The Condition Approach . . ......... Boo BML Woe ow om koe Questionnaire Revisions . . . . . ..................... Reevaluation of the Survey . . . . . LL LLL. Evaluative Studies . . . . . . ... LLL The Person Approach . . . . ...... Collection of Data on Chronic Conditions . . . . . .. .......... The Split-Sample Procedure . . . . . ................... Comparison of the Two Half-Samples Used in the Health Interview Survey, July 1967-June 1968 . . . . . . . LLL Population Composition . . . . .................... Collection of Prevalence Data . . . . . . ................ Limitation of Activity . . . . . . ........ Questionnaire Revisions Effective in January 1968 . . . . . .... . .. Comparative Estimates of Activity Limitation . . . ........... Questionnaire Changes in 1969 and 1970 . . . . .. ........... Chronic Conditions Causing Activity Limitation . . . . .. .... ... Incidence of Acute Conditions . . . . . ......... ee Collection of Data on Acute Conditions . . . . . . . . . . . ....... Sources of Differences in Estimates . . . . . . . . ............ Questionnare Changes . . . . . ...................... Disability DAYS + « « « 2 + 5 0 2.4 5 28 5 5 4 2 0 vv mo en vsnmessas Days of Person Disability . . . . . .................... Days of Disability Due to Acute and Chronic Conditions . . . . . . . . . Days of Disability Due to Acute Conditions . . . . . .. ......... Days of Disability Due to Chronic Conditions . . . . . .......... Physician VISE . « + s # « + 2 6 64 58 52 so vr awrmucan™ ames Volume of Physician Visits. . . . . ................... Persons Seen by a Physician During the Past Year . . . . .. ....... Page QO = BB 00 Wo OO 0a 10 11 12 14 15 15 16 16 18 18 18 19 20 20 20 21 21 22 22 22 CONTENTS —Con. Page Hospitalization . . . . . « «vv vs x vss as ss es ens saves onan 23 Conclusions «+ a ws + vv = x 2 2 « © + co vo mw so wb so 88 8 4 2 8 5» 23 References . . . . ot i i i i i i ee ee ee eee eee eee 24 List of Detailed Tables . . . . . . «o.oo nee 25 Appendix I. Technical Notes on Methods . . . . . ............. 45 Backgroundof ThisReport . . . . . . . «tv vv ttt veo v vv va 45 Statistical Design of the Health Interview Survey . . . . . . ........ 45 General Qualifications . « « + « + + 4s + vs vce suv anv wv mos wns 47 Guide to Use of Standard Error Charts. . . . . .............. 50 Appendix II. Definitions of Certain Terms Used in This Report. . . . . .. 54 Terms RelatingtoConditions . . . . . . . . «vv vv vv vv vv 54 Terms Relating to Disability . . . . . . o.oo 56 Terms Relating to Persons Injured . . . . . . o.oo. o.oo. 58 Terms RelatingtoClassof Accident . . . . . .......... cc... 58 Terms Relating to Hospitalization . . . . . . . 59 Terms Relating to Dental Visits . . . . . . . ovo. 60 Terms Relating to Physician Visits. . . . . . o.oo 60 Demographic Terms . . . . « vo «4 so vs cv sass avs sa mrimnsss 60 Appendix III. Questionnaire—Condition Approach . . . . . ......... 63 Appendix IV. Questionnaire—Person Approach . . . . . ........... 73 SYMBOLS Data not available------=esemeesmmeeee eee een . Category not applicable--------meeerereemeoooe eee Quantity zero ws Quantity more than 0 but less than 0.05----- 0.0 Figure does not meet standards of reliability or precision (more than 30 percent relative standard error)----------- INTERVIEWING METHODS IN THE HEALTH INTERVIEW SURVEY Geraldine A. Gleeson? INTRODUCTION On July 1, 1967, the Health Interview Survey (HIS) had completed its first 10-year cycle of data collection. In accordance with a long-range plan set up during the early years of the survey, a general evaluation of the design and format of the survey was undertaken. As a result of this evaluation, a major revision of the survey proce- dure was made and took effect on July 1, 1967. However, because of the experimental nature of the revision, the survey sample was divided into two half-samples, both of which were ‘intended to measure the health status of the civilian, noninstitutional population of the United States. In one of the half-samples, the proce- dures used during the preceding 10 years were continued; in the other, the revised procedures were introduced. This split-sample technique was continued during the 12-month period ending in June 1968. In short, the collection system used from the beginning of the survey, called the “condition approach,” consisted of eliciting all kinds of information about illnesses and injuries through a series of illness-recall (probe) questions and then collecting information on disability and medical care associated with the conditions. In the other collection method, known as the “person approach,” information about short- and long-term disability and/or the receipt of medical care was collected first, and information about the condition(s) responsible for the dis- ability or the need for care was then collected. 3Mrs. Gleeson was formerly Special Assistant to the Director of the Division of Health Interview Statistics, now retired. Although both approaches were intended to produce the same general kinds of information, the method of collection would be expected to have an effect on the estimates produced. Table A summarizes the estimates made that were based on data collected by the two approaches. This report presents more detailed information on both approaches, describes rea- sons for differences in estimates, and lists some of the measures that have been taken to improve data produced from the newly adopted proce- dure, the person approach. BACKGROUND OF THE HEALTH INTERVIEW SURVEY Initiation of the Interview Survey In July 1956 legislation was enacted author- izing the U.S. Public Health Service to conduct continuing surveys of illness and disability in the United States. This legislation—the National Health Survey Act (Public Law 652, 84th Congress)—specifically provided for the estab- lishment of an organization within the Public Health Service to plan, organize, and publish the survey findings. The Health Interview Survey, the first of a series of data collection systems organized to implement the legislation, was planned during the fall and winter of 1956. Through contractual arrangement, the services of the Bureau of the Census were obtained to prepare a sample design! for the conduct of the survey and to carry out the collection and data processing phases of the program. The procedures were pretested in February 1957 in the city and Table A. Summary of comparative estimates derived from the half-samples used in the Health Interview Survey during July 1967-June 1968 Type of approach Questionnaire topic Condition Person Limitation of activity: Percent unable to carry on major activity . . ............ i... 2.2 3.0 Percent limited in amount or kind of major BCUVITY oo is 3 Siow 6 55 SAME FFs 8 HEME 8 80 Gomme sxx wwe vn 6.6 6.0 Percent limited in other than major BCUVITIBE os cme rer mR AE EEE HAE TI NMEA CES REE EE 33 1.6 Acute conditions: Total conditions per 100 Persons . . . .. .. coi i iii 194.5 184.6 Activity restricting: With no medical attention . . .. .. .... ee 67.3 75.8 With medical attention . : weua ss sss sap cist mu mm ese xs mmmms vu 70.2 74.9 Medically attended only . .. coos seis ammas ss mE@@s sss mmm osu 56.9 34.0 Disability days per person per year: Restricted activity . . . oo iii tt ee eee eee 15.7 14.8 Bed disability . . oot ee 59 6.0 Work-loss {currently employed) . . css ass cess smsmsr es vmmmwnvse an 5.4 49 School-loss (B-18 years) .......::eceamssssmpumssssommmensss => 4.1 49 Disability days due to acute conditions per person per year: Restrictar aChIVIEY: .. . ¢ +» wv wis + 6 5 3 Ra 85 £4 5 BETH €€ 833 nEw vss mmm 13 8.4 Bed diSBBIY «vv vv stm sess pmb ss mmm Fs ana pre mn 3.1 3.7 Work-loss (currently employed) . ..... 3.4 3.4 School-10ss (6-16 Years) . . . . ou ii iit 3.6 4.4 Disability days due to chronic conditions per person per year: Restricted @CHIVILY . « vv vom vc vr mmm ch sa ama vs ss Ramm 5 #8 80 5s 13.0 11.6 BOA QIODITY onus 1 sms vrs mmm as sn mmm si RIERA ELF EERE 4.2 43 Work-loss (currently employed) . ....... 3.0 2.4 School-10ss (6-16 Years) . . . ov i iii te ee et eee 0.8 0.7 Physician visits: Physician visits per person Per year . . .. ...... co. oe enue 4.3 4.1 Percent of persons with visits during past 771 REE Serpe prgv ge ET TL LE 69.0 70.9 Hospitalization: Hospital discharges per 100 persons per YVEBE vr vs 83 SHER TIF ERE EE Ey ee ee 12.2 11.7 Length of hospital stay per discharge indays . .................... 8.8 9.0 suburbs of Charlotte, N. C. Revisions and modi- fications were made in view of the findings of the Charlotte pretest, and a national “dress rehearsal” of the survey was conducted during May and June 1957. Data collection for the Health Interview Survey started officially on July 1, 1957, and has continued without inter- ruption since then. Concepts Used in the Survey In the development of the concepts of mor- bidity, disability, and medical care and in the construction of the basic questionnaire used in the collection of health information, Health Interview Survey personnel drew heavily on the experiences of researchers who had previously used the interview method for the collection of health data. Early in the 1920’s, certain populations in Hagerstown, Md., were selected for epidemio- logical studies and the determination of some estimates of the levels of selected health meas- ures in a local population. These studies con- tinued throughout the next several decades. During 1935-36, a major nationwide health survey was conducted in which 737,000 urban households were visited by interviewers. Both of these studies contributed to the information about the basic kinds of health data that can be collected by household interview. Since 1936, with the development and refine- ment of sampling techniques and procedures, the interview method has been used as a means of data collection in a number of local studies of morbidity. Foremost among these are surveys conducted in Baltimore, Md.; Pittsburgh, Pa.; Hunterdon County, N.J.; Kansas City, Mo.; New York City; and the State of California. Morbidity is defined in the Health Interview Survey as a departure from a state of physical or mental well-being. This departure is a result of a disease or injury of which the affected individual is aware. Morbidity includes not only active or progressive disease but also impairments that are static in nature resulting from disease, injury, or congenital malformation. In the survey the concept of a condition that causes morbidity is usually limited by specifying that it includes only conditions for which the person has taken one or more actions. Such actions include restriction of usual activities and seeking medi- cal attention. Using action as the criterion for inclusion of a condition in the class of morbidity condition is justified on the basis of logic and practical utility. If the condition is so unimpor- tant that the individual takes no action of any kind, it is usually of little health significance. In addition, earlier surveys revealed that there is a marked increase in response error associated with the reporting of minor illnesses that in- volved neither disability nor medical attention. However, in the area of chronic disease reported in health interviews where conditions are of long duration and medically related actions within the interview recall period might not have occurred, prevalence estimates are produced without regard to recent restricted activity or medical attention. The concept of morbidity is modified to some extent in the process of constructing an opera- tional procedure, a questionnaire and the in- structions for its administration in order to achieve objectivity and simplicity in the inter- view method. Actually, the whole structure of the survey becomes the working definition of morbidity. Therefore, in the construction of questions which are used to elicit information about conditions from respondents, it is neces- sary to cover all the aspects of morbidity that the original concept requires. DATA COLLECTION IN THE HEALTH INTERVIEW SURVEY Format of the First Questionnaire During the planning phase of the interview survey, two general questionnaire formats were considered. The one referred to as “alternative B” was designed to elicit information about conditions through the reporting of actions a person might have to take as a result of chronic disease. For example, a respondent would be asked if he had to: 1. Cut out or reduce all or part of his activities regularly or from time to time. 2. Change his activities. 8. Change his diet. 4. Take medicine or treatment over a long period of time. 5. Wear or use some special device. After a positive response to any of the above statements, the respondent would be asked, “What was the matter?” While the original intent of this proposed format was to clicit information about chronic conditions, the same general approach was applicable to both chronic and acute conditions. The other format, the one actually used during the first 10 years of the Health Interview Survey (July 1957-June 1967), provided for the reporting of all kinds of morbidity conditions through a series of direct questions designed to encourage the reporting of illnesses and injuries. In contrast to alternative B, no attempt was made to determine at the time the condition was initially reported if some action had been taken by the person because of the condition. This format was used to maximize the number of conditions reported regardless ol their impact or severity and to apply the criteria of medical attention, restricted activity, or limitation of activity during the coding and transcribing of the collected data. The Condition Approach The selection of this questionnaire format, which is usually identified as the condition approach, was influenced by its general accept- ance in earlier health surveys. Ilness-recall ques- tions which had been formulated and used successfully in the collection of health data in carlicr surveys served as a prototype for the first questionnaire used in the Health Interview Survey. Using a tested collection procedure made it possible to begin the interviewing phase of the survey much earlier than would have been the case if a completely untested procedure such as alternative B had been adopted. The illness-recall (probe) questions used during the first year of the interview are shown in figure I. The wording of the introduction to these questions--“We are interested in all kinds of illness, whether serious or not” —indicates the comprehensive nature of this section of the questionnaire. ‘I'hese questions were structured to elicit information about any departure from a state ol physical or mental well-being resulting from disease or injury, i.c., a morbidity condi- tion. The questions which were limited to occur- rences during the last week or the week before were designed primarily to aid in the reporting of acute conditions. Question 15 and the checklists of chronic conditions and impairments were used to assist the respondents in recalling and reporting conditions of a more lasting nature. All reported conditions were recorded regardless of which type of question had prompted the reply. Whether these conditions were chronic or not was established later in the interview on the basis of a series of questions relating to the nature of the discase and its duration. The section of the questionnaire designated “Ta- ble I" (figure 2) provided this kind of informa- tion for each reported condition. The questions in Table I also ascertained whether a doctor had been seen and the days of disability associated with cach of the conditions so that the severity criteria—restriction of activ- ities and/or medical attention—could be applied in the selection of acute conditions for inclusion in the count of conditions basic to the deriva- tion ol incidence estimates. For persons who reported chronic conditions, additional informa- tion was obtained from Table I on the presence (or absence) of chronic limitation of activity or mobility. Questionnaire Revisions During the succeeding years of the interview survey, the section of the questionnaire dealing with acute and chronic illness underwent certain changes. Progressive experience in survey col- lection procedures on the part of the Health Interview Survey staff and the findings produced from continuing studies on survey methodology led to periodic changes,” which in tum led to some improvement in the reporting of illness by the respondent. These changes included varia- tions in the order in which illness-recall ques- tions were asked, introduction of a small calen- dar outlining the recall period for the bQuestionnaire format changes from 1958-1964 have been described carlier. Illness-Recall Questions We are interested in all kinds of illness, whether serious or not -- [3 Yes CJ No 11. Were you sick at any time LAST WEEX OR THE WEEX BEFORE? (a) What was the matter? (b) Anything else? 12. Last week or the week before did you have any accidents or injuries, either at home or away from home? 0 yes OJ Ne (a) What were they? (b) Anything else? 13. Last week or the week before did you feel any ill effects from an earlier accident or injury? OC yes Ol No (=) What were these effects? (b) Anything else? . 14. Last week or the week before did you take any medicine or treatment for any 03 ves CI No condition (besides ...which you told me about)? (a) For what conditions? (b) Anything else? 15. AT THE PRESENT TIME do you have any ailments or conditions that have con- OJ Yes CI No tinued for a long time? (If “No) Even though they don't bother you all the time? (8) What are they? (b) Anything else? 16. Has anyone in the family - you, your--, etc. - had any of these conditions DURING THE OO yes OC ne PAST 12 MONTHS? (Read Card A, condition by condition; record any conditions mentioned in the column for the pe ) 1. Asthma 16. Kidney stones or other 2. Any allergy kidney trouble 3. Tuberculosis 17. Arthritis or rheumatism 4. Chronic bronchitis 18. Prostate trouble 5. Repeated attacks of sinus trouble 19. Diabetes 6. Rheumatic fever 20. Thyroid trouble or 7. Hardening of the .arteries goiter 8. High blood pressure 21. Epilepsy or convulsions 9. Heart trouble of any kind 10. Stroke 22. Mental or nervous 11. Trouble with varicose veins trouble 12. Hemorrhoids or piles 23. Repeated trouble with 13. Gallbladder or liver trouble back or spine 14. Stomach ulcer 24. Tumor or cancer 15. Any other chronic 25. Chronic skin trouble stomach trouble 26. Hernia or rupture 17. Does myone in the family have any of these conditions? (Read Card B, condition by condition; record any conditions 3 ves no mentioned in the column for the person) 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 foor or leg, fingers, arm, or back. Figure 1. lliness-recall (probe) questions used in the first year of the Health Interview Survey. Table I - ILLNESSES, IMPAIRMENTS AND ACCIDENTS LAST WEEX Did What did the doctor say it If an impairment or symptom, ask: What kind of ...trouble What part of the body OR THE you was? -- did he use any is it? was affected? WEEK Bo. YEE | medical terms? What wus the cause of (1t eye ORE Ha trouble of nu 808 | 0s doctor not alii to « she? any kind | (If kind of trouble (1? part of body can be |You to does | In col, (C)~ Petord and B years | 8lready entered in col. | determined from entries |cut down tor respondent's description (If cause 1s already old or (d-1), circle “X*” with- in cols. (4-1) through on your a Sage ) entered in (d-1) circle over, ask): | Out asking the question) | (d-4), circle “X* without WAL A ¢ col. wee t ~ “X'* without asking the asking the question) vitier Jor lod | thier st arti | cetion) for a 2 of [tim e Ay Om you gmich a © |per-|No. For an accident or injury (If accident of injury el a day a occurring durin t 2 * ordinary A= oo a auring. Pas {iLL Table A) nn heck one print with No Yes What part of the body was glasses? hurt? What kind of ipjury {0s (mmr was (t? Anything else? Cot (Also, fill Table A) (x) ®]| ®»] «© @-1 @-2) @-3) (d-4) (4-3) ole Cl yes 1 Clves x ¥ 1 One CI ne BN Table I - ILLNESSES, IMPAIRMENTS AND ACCIDENTS 1 6 yours wig Did you first notice ... To Inter-| Did you first | When did | Do you About how or over, ask: DURING THE PAST 3 MONTHS viewer: notice ... op 1881 still Atay days . DURING THE talk to & | teke any ring the Last week| 1f “yes | or before that time? It Col. | PAST 12 MONTHS ctor medicine [past 12 or the [in col. chek Did ... start | (k) is or before that | about...? | or treat- |months, has Jos ty Check One | siring the past | checked | time? ment oo. Kept fare How many 2 or the (Month and| the doctor |you in bed Pave been 48Y8 did [ae fore| Dur ing condition| (rf guring past | 7¢SF Year| prescribed | for all or » ne n ’ 3 1s on Yi months, anki: only if for ...? most of the|Then tell this 40 4 oe BoORLhs) Bonthy either one| * ‘| prior to otal day? 88 Which sand and of Cards 1956) r, follow 8 e- te. me or buai- 50 S12 Yirink nase Aor B, Wich month? any advice ment fits| which of it “ to * © | continue; he gave? you best. | these school) cat. Which week, last| Othermise, (Show Stte- except (n)) week or the STOP Sancy 0% es s for ...” week before? F. as os Dy appro- iby priate) | (Show Card a) (®) hy 1) x | (3) (an) (n) (0) () @ (n 8) —_ pues CT) Yes. CO Last Before Mo. — | mo. fen C0 bs = ro— |r. | Days | [] None CI No Oo ook Jpetore Cpirth | (Ing pr. | C2Ne Br. | CJ None 1 Card C Card D Card E For: Fors Workers and other persons except Children from 6 to 16 years old and Pp ; Housewives and Children For: Housewife others going to school 1. Cannot work at all at present, 1. Cannot keep house at all at 1. Cannot go to school at all at present. present time. 2. Can work but limited in amount or kind of work. 2. Can keep house but limited in anount BF Ring GF HOUSEWORK. 2. Can go to school but limited to 3. Can work but limited in kind or certain types of schools or in 5 Begley ¥ school attendance. amount of outside activities. 3. Can keep house but limited in outside activities. Co 4, Not limited in any of these ways. 3. Can go to school but limited in we 2 other activities, 4. Not limited in any of these ways. 4. Not limited in any of these ways. play with ot 2. Can play with limited 4. Not limited her other children but children. in amount or in any of these ways. Card F Card 8 For: Childred under 6 years old 1. Cannot take part at all in ordinary | 1. Confined to the house all the time, except kind of play 2. Can go outside but of another person around outside. 3. Can go outside alone but have trouble 4, Not limited in emergencies. in getting around freely. in any of these ways. need the help in getting Figure 2. Table | used in the first year of the Health Interview Survey. convenience of the respondent, restructuring of the checklists of chronic diseases and impair- ments, the identification of the condition(s) causing either limitation of activity or limitation of mobility, and format changes to accommo- date revised data processing procedures. Despite these changes in the questionnaire, certain kinds of health related information con- tinued to be underreported in the survey, although to a lesser extent than in the first years of the survey when, on the basis of research studies comparing interview data with medical records, it had been established that chronic conditions were not completely reported in the interview.3:4 For example, the prevalence of selected chronic conditions has increased with changes in the questionnaire formats. REEVALUATION OF THE SURVEY Early in 1963, after 6 years of data collection and in accordance with a long-range plan set up during the early years of the survey, a general evaluation of the design and format of the survey was undertaken. A timetable was pre- pared which provided for considering proposed changes, for deciding whether to accept, reject, or modify the proposed changes, and for pretest- ing and evaluating the approved changes. A target date of July 1, 1967, was established for the completion of the evaluation and for the introduction of any new procedures in the collection phase of the survey. During the 4-year evaluative period, 1963-67, the ongoing survey continued in line with collection procedures developed during the early years of the survey. Evaluative Studies While the procedures used in the Health Interview Survey were being evaluated, consider- ation was given to items in three general areas: (1) sample design and collection proce- dures, (2) general objectives and types of analy- sis, and (3) questionnaire content and format. Studies relating to the first general area were primarily concerned with the adequacy of the reporting of chronic conditions. One of the studies was an experimental project in which 20 percent of the households interviewed during July-September 1963 were again interviewed during the same quarter in 1964. This study, which was an attempt to improve the reporting of chronic illness, simulated the longitudinal method of data collection. Other studies exam- ined (1) the effectiveness of preinterview mate- rials (pamphlets, letters, and calendars) in moti- vating household respondents, (2) the productivity of intensive and serial interviews, and (3) the advisability of collecting data on only one type of chronic illness during a specific collection period. Questions pertaining to general objectives of the survey led to a study designed to find out about data needs and how published material is used by consumers and to an investigation of the relative importance of measuring levels, time changes, and relationships as objectives of the survey. New types of analysis that evolved from the evaluative considerations were the synthetic estimates for individual States® and the exper- imental estimates of health characteristics on a family basis. Evaluation of the survey in terms of question- naire content and format led to the major changes that were introduced in July 1967. The result of these changes is the person approach questionnaire. Because of the experimental nature of the new format, it was decided for methodological purposes to split the Health Interview Survey sample during July 1967-June 1968 into two separate samples, each of which would provide measures of the health character- istics in the portion of the population covered by the respective sample. The questionnaire and procedures used during July 1966-June 1967, which were essentially those used during the first 10 years of the survey, were continued in one of the half-samples, and the new question- naire and procedures were applied to the other.¢ Some information on the reliability of the data collected by the two approaches, including sampling error charts, is presented in appendix I. Appendix II contains the definitions of terms used in this report. All questionnaires used for the collection of material during July 1967-June 1968 are shown in their entirety in appen- dixes III and IV. “Minor revisions involving the wording and order of items on the person approach questionnaire were introduced in January 1968. In addition to the split-sample procedure, a methodological project was carried out for the purpose of measuring differences introduced in the transcription of data collected by the condi- tion and person approaches. This study was conducted in accordance with a variation of a split-half matching technique. First, data from a random selection of 215 questionnaires from the person approach were (transferred to the condi- tion approach questionnaire. Both sets of ques- tionnaires were then processed separately, ac- cording to a common editing program, with procedures unique to each document format to produce two sets of computer tapes. The con- tent of these tapes was then matched item for item for cach pair ofl records derived from matching questionnaires. In the second matching procedure, the proc- ess was reversed. Data originally collected by the condition approach were transferred to per- son approach questionnaires for an equivalent number of household interviews. As in the first operation, both sets of questionnaires were processed separately, and the prepared computer tapes were matched. From this project it was found that no appreciable amount of variation was introduced during the transcription and processing phases of health data collected by the person and condi- tion approaches. Where substantial differences were found during the matching procedure, they were attributable for the most part to defini- tional differences in the two approaches. The Person Approach The new questionnaire introduced as a data collection instrument in July 1967 resembled the approach suggested, by the alternative B method of data collection considered at the beginning of the interview survey. The illness- recall questions, with a 2-week relerence period, were replaced with probe questions pertaining to health related actions during the period—e.g., cutting down on usual activities, spending days in bed, losing time from work or school, or secking medical attention (figure 3). Informa- tion about conditions responsible [or such ac- tions was obtained from persons with positive response to the health related action probe , questions. This format provided for the reporting of (1) acute conditions that had caused restricted activity or received medical attention during the 2-week reference period, (2) disability days (re- stricted activity, bed disability, or work- or school-loss) occurring during the reference period due to acute or chronic conditions, and (3) visits to physicians and hospitalization as estimates of medical care utilization indepen- dent of the conditions for which the care was received. In addition to the variety of informa- tion obtained through the probe questions, the person approach resulted in a reduction in interviewing time because of the automatic exclusion of acute conditions that did not result in cither restriction of activity or medical attention. In summary, the adoption of a collection procedure that stresses health related behavior such as visiting a physician or losing days from work, with the causative condition playing a secondary role, has several advantages. On the basis of our own experience and that of other researchers, it is known that the interview method produces more reliable information on disability and medical care than on illness per se, particularly illness by diagnosis. For this reason, it scems reasonable to obtain information on disability and care experienced by the sample individual first and on the causative condition(s) later, rather than depend on the initial reporting of the condition to elicit information about the presence of disability and care. Furthermore, a person’s degree of illness may be measured and compared with the experience of others by using data that will summarize the unduplicated count of disability days for all conditions affecting the person. When these data were obtained for each condition separately (in the condition approach) and then unduplicated and summed for the individual, the unduplicating process was awkward and imprecise. The person approach obtains the person and condition disability information directly during the inter- view. This procedure permits the respondent to report the total days of disability experienced during the 2-week reference period and later in the interview to report the number of days attributable to each condition. By focusing the interview on the collection of conditions that have some impact on persons in If related persons 19 years old or over are listed in addition to the respondent, say: We would like to have all adults who are at home take part in the interview. Is your — —, your — —, etc., at home now? 1[JAt home o[ Under 19 years H If other eligible respondents are at home, ask: 2[ Not at home Would you please ask — —, — —, etc., to join us? i HAND CALENDAR TO RESPONDENT i 5a. During the past two weeks (the 2 weeks outlined in red on that calendar) did — — stay in bed all or most of the day because of Sa. [Yes — Asks any illness or injury? 00[ JNo — Ask ¢ mem rm mE mm Em mm me mm mm mm me — — — — — — — — — — — — — — —— —— — — —————————— re Le ———— b. During that two week period, how many days did — — have to stay in bed all or most of the day? b. stance tc ti etre ee Se, eS ee J ie, eT 38 Says Amh te. c. During that two week period, did he have to cut down on the things he usually does because of illness S or injury? [JYes-Ask d [CINo—Go to 6a mm mm mm ee mm fo ro dee tv. ov: vn po d. Did — — have to cut down for as much as a day? AT CTYes AK e Snpmm— OA s — Ask f or | If under 6 yrs. — Go to 6a TT Tl 17yearsoldoroverask: CTT TooToo Tmommooom om mmmm momen 1 Tool None TTT ] f. How many days did illness or injury keep — — from work during these two weeks? soi oe EOL IEMBIES pid Not counting work Sound slo BOUND... coe omen 0 ii io 0 ein i 8 6 = me 4818.60 fo 68. | If 6-16 years old ask: 9-{00[_]None g- How many days did illness or injury keep — — from school during those two weeks? days — Go to 6a “ dis] If 1+ days recorded in Q. 5e, ask: ba. 6a. What condition caused — — to cut down on the things he usually does during the past two weeks? — Enter condition in C above [CINo cut down days Go to next person b. During the past two weeks, did any other condition cause him to cut down on the things he usually does? b./ []Yes—Reask a and b [JNo—Go to next person 7. During the past 2 weeks (the 2 weeks outlined in red on that calendar) how mony times has — — seen a doctor 7. either at home or at a doctor's office, or clinic? [None Number of visits 8a. (Besides those visits) During that 2-week period has anyone in the family been to a doctor's office or 8a. clinic for shots, x-rays, tests, or examinations? [JYes — Ask b and c [INo — Go to 9 b. Who was this? —Mark ‘“Yes’” in person's column b.[ []Yes Doctor's visits c. Anyone else? [Yes _ Reask b and c For each ‘Yes’ marked, ask: d. How many times did— — visit the doctor during that period? -Exclude visits made on ‘‘mass’’ basis [JYes — Ask b and c| 9a. [INo — Go to 10 9a. During that period, did anyone in the family get any medical advice from a doctor over the telephone? b. If *‘Yes’} ask: Who was the phone call about?—Mark ‘“Yes’’ in person's column. c. Any calls about anyone else? [C]Yes — Reask b and c [JNo —Go to d For each ‘“Yes’’ marked, ask: d. How many telephone calls were made to get medical advice about — —? d| Number of calls [INo 2-week visits—Ask 11 If doctor was seen or talked to during the past two weeks, ask: 10a. 10a. For what condition did — — see or talk to a doctor during the past two weeks? — Enter condition here and in c above b. During that period, did — — see or talk to a doctor for any other condition? b. If pregnancy reported ask: During the past 2 weeks was — — sick because of her pregnancy? If ‘Yes’ ask: What was the matter? []Yes—Reask 10a [INo—Go to next person Figure 3. Questions relating to short-term disability and physician visits that also served as illness-recall questions in the person approach questionnaire in the Health Interview Survey during July-December 1967. terms of seeking medical care, cutting down on the things they usually do, or limiting their activities, interviewing time is appreciably re- duced and proper emphasis is placed on illnesses that represent actual health problems. Collection of Data On Chronic Conditions Since the initiation of data collection in the Health Interview Survey in 1957, a continuous effort has been made to improve the quality of data on the prevalence of chronic conditions. Methodological studies have shown that chronic conditions are generally underreported in inter- view surveys. They have also indicated that the expansion of the Checklist of Chronic Condi- tions to include as many descriptive titles as possible for a particular condition will increase the probability that a respondent will report the condition, assuming, of course, that he is aware of its existence. This finding led to the decision to restrict the collection of prevalence data on chronic condi- tions to specific types of conditions during a given collection year. This change in collection procedure is completely apart from the defined person approach; however, since both proce- dural changes were experimental during the collection year July 1967-June 1968, they were tested on the new questionnaire introduced in the field. Although the person approach and the Checklist of Chronic Conditions involving a specific body system are both integral parts of the HIS collection procedure at the present time, either could be discarded at some future time without reducing the effectiveness of the other. Concentrating on a group of chronic condi- tions involving a specific system of the body (e.g., those affecting the digestive system) rather than on the entire spectrum of chronic condi- tions not only improves the quality of response but also permits the collection of more detailed diagnostic information related to that body system. The survey plan calls for the collection of different types of conditions each year, so that within 5 or 6 years after the initiation of this plan, information on the prevalence of virtually all chronic conditions will have been obtained. Once the decision had been made to modify the collection procedure for chronic conditions by emphasizing a specific type of condition during a given year, it was necessary to develop at the same time procedures that would provide comparable data for other measures of morbid- ity that had been derived previously from data collected on all types of chronic conditions. One of these measures, the number of persons with limitation of activity (long-term disability), had previously been generated by consolidating the data on activity limitation attributable to specif- ic chronic conditions reported by an individual to represent the activity limitation status of that individual. The most obvious alternative to this consolidation was to build a person-data foundation in terms of the degree of activity limitation and then ascertain the conditions responsible for the activity limitation status of the individual. The Split-Sample Procedure The questionnaire used in the collection of interview data is a very sensitive instrument. Supposedly minor changes in the wording or order of questions can result in subsequent changes in the levels of estimates for health related items. It was anticipated that the intro- duction of a major procedural innovation such as the person approach in the collection of health data would undoubtedly produce new levels and possibly new relationships. For this reason a split-sample was used to provide some information on the effect of a new collection procedure on the estimates established during the first 10 years of the survey. The following sections are devoted to a comparative evaluation of the health data collected on the question- naires using the condition and person ap- proaches during July 1967-June 1968. This evaluation revealed certain shortcomings in each of the approaches. Revisions in either the format or content of the questionnaire were made in January 1969 and January 1970 to improve the collection of data by the person approach. These revisions are described at the end of the section dealing with the measure of health to which they apply. Additional changes will undoubtedly be made in future question- naires to improve data collection. COMPARISON OF THE TWO HALF-SAMPLES USED IN THE HEALTH INTERVIEW SURVEY, JULY 1967-JUNE 1968 Population Composition Direct supervision of the interviewing phase of the Health Interview Survey is cxercised through the 12 data collection centers of the Bureau of the Census. Before the two half samples were taken, half of the interviewers in each data collection center were trained to administer the questionnaire with the condition approach, and the other half were trained to administer the person approach. This procedure was carried out by pairing interviewers and by randomly selecting from each pair an interviewer to use the questionnaire employing the person approach. The other interviewer used the condition approach ques- tionnaire. Within each data collection center, interviewers were paired on the basis of approxi- mate equality in expected quarterly workload and similarity in degree of urbanization of workload area. Two interviewers working in the same enumeration area were automatically paired together. In the case of an odd number of interviewers in a given data collection center, the “odd” interviewer was paired with an “odd” interviewer from another center. If an inter viewer was unable to carry out an assignment, an interviewer trained in the same approach was normally recruited to complete the assignment. The selection process equalized as much as possible the total workload between the new and old procedures and also achieved for each half-sample a representative subset of the total sample. During the year, the condition approach was used to interview persons composing 21,215 households, and the person approach, to inter- view persons composing 22,361 households. When sample persons were weighted to represent the total civilian, noninstitutional population, the average total annual population of 194,461,000 persons was composed of a condi- tion approach average population of 94,557,000 persons and a person approach average popula- tion of 99,904,000 persons. The composition of the populations in the two half-samples was quite similar for a number of population characteristics (table 1). There was a sizable difference in the population distribution by geographic region. This variation is explained by the fact that the boundaries of 12 census regions, the areas on which the interviewer pairings were based, do not coincide in some instances with the limits of the four major geographic areas used in the survey analysis. Furthermore, the pairing of an “odd” interviewer in a given regional office with an “odd” interviewer in a different regional office could lead to an unbalanced sample within a given geographic region. Collection of Prevalence Data The new questionnaire, introduced in the ongoing survey in July 1967, was constructed to produce total prevalence estimates for chronic conditions involving the digestive system only. As indicated earlier, all the digestive conditions of a chronic nature, regardless of the question that prompted the response, were included in the prevalence estimate. However, approxi- mately 90 percent of the reported conditions included in the estimate were first reported in response to the reading of the checklist (fig- ure 4). This high proportion was anticipated since the checklist was preceded on the ques- tionnaire only by probe questions involving an action such as receiving medical attention or cutting down on one’s activities during the 2-week period prior to interview. Unless the person had received medical attention or experi- enced disability days because of a specific chronic digestive condition, his first opportunity to report the presence of such a condition would be in response to the reading of the checklist. In comparison, the general Checklist of Chronic Conditions used in the half-sample to which the condition approach questionnaire was administered revealed only about 60 percent of the chronic digestive conditions reported. The lower proportion of digestive conditions re- ported in response to the general checklist was primarily the result of the comprehensive nature of the condition probes preceding the checklist, e.g., “Was sick anytime last week or the week before? Did have anything else during that 2-week period? Did__take any medicine or treatment for any condition? Did ever have an accident or injury that still bothers him in any way?” This greater dependence on the specialized checklist for the reporting of prevalence data in the person approach indicates that the list should be as inclusive and descriptive as possible in terms of specific conditions which might involve the body system for which data will be collected. Since checklists of conditions are included on questionnaires for the purpose of reminding and encouraging respondents to report conditions which might otherwise be missed, the names and titles of conditions or disease categories used are ones that would be familiar and understandable to the respondent. Once the condition is re- ported, more information about it is obtained so that it can be properly classified in the appro- priate disease category in accordance with the International Classification of Diseases (ICD). This additional information (see Condition page, appendix III, p. 66) is ob- 1 Condition approach (Checklists included all types of chronic conditions and impairments) Person approach (Checklists included only chronic conditions of the digestive system) these conditions? 1. Tuberculosis? 2. Emphysema? 5. Hardening of the +. High 1 (Exclu 6. leart trouble? essure” Card A Al A-2 Now I'm going 10 read a lis of conditions — Have you, your — —, etc., had ony of these Please tell me if you, your — -, efc., have conditions DURING THE PAST 12 MONTHS? had ony of these conditions DURING THE PAST 12 MONTHS? 1. Asthma? 12. Thyroid trouble or goer? 2. CHRONIC bronchitis? 13. Any allergy? REPEATED attacks of sinus trouble” 14. CHRONIC nervous trouble? + TROUBLE with varicose veins” 15. CHRONIC skin trouble? 5. Hemarrhods or piles? 16. Palsy? 6. Hay fever? 17. Paralysis of any kind” Tumor, cyst, or growth?” 18. REPEATED trouble with back or spine” B. CHRONIC gallbladder or Liver trouble? 19. Cleft palate? 4. Stomach alee? 20. Any speech defect? 10. Any other CHRONIC stomach trouble? 21. Hemia or rupture? 11. Kidney stones or CHRONIC kidney 22. Prostate trouble? trouble? Card B B-1 B-2 Hove you, your - -, etc., EVER had any of Do you, yout - -, etc, HAVE any of these conditions? with one or both ears? arteri Missing fingers, hand or arm — during pregnancy) foot or leg? 5. Club foot” 1. Deafness or SERIOUS trouble hearing 2. SERIOUS trouble seeing with one or < both eyes even when wearing glasses? Missing lung or kidney (or breast)? 6. PERMANENT stiffness or any deformity of foot, Leg, fingers, arm or back? Now I'm.going to read a list of conditions: 120. During the past 12 months, hos anyene in the fomily (you, your — —, etc.) had any of the following conditions — If “Yes,” ask b and ¢ Yes| No 1. Gallstones? 2 Any other gallbladder trouble? 3. Hemorrhoids or piles? 4. Cirrhosis of the liver? b. Who wos this? 5. Paty liver? c. During the past 12 months has anyone else had . . . 6. Hepatitis? 7. Yellow jaundice? 8. Any other liver trouble? o. During the past 12 months, hos anyone in the family had — If “Yes,” ask b and c Yes| No 9. A disease of the pancreas? 10. A disease of the esophagus? 11. Any other disease thot offects swallowing? b. Who wos this? c. During the past 12 months has anyone else had . . . 12. Peptic ulcer? 13. Duodenal ulcer? 14. Stomach or gastric ulcer? 15 Any other ulcer? a. During the past 12 months, has anyone in the family hod — If “Yes,” ask b and c Yes| No 16. Hiatal hemia? 17. Umbilical hernia? 18. Any other hernia or rupture? 19. Gastritis? b. Who was this? 20. Frequent indigestion? €. During the past 12 months has anyone else had . . . 21. Concer of the stomach? 22. Any other stomach trouble? 23. Enteritis? 24. Diverticulitis? ©. During the past 12 months, has anyone in the fomily had — If “Yes,” ask b and ¢ Yes| No 25. Colitis? 26. Constipation or other bowl trouble? 27. Spastic colon? b. Who was this? 28. Cancer of the colon or rectum? c. During the past 12 months hos anyone else had . . . 29. Any other concer of the digestive system? | 30. Any other intestinal trouble? 31. Any other condition of the digestive system? Figure 4. The checklists of chronic conditions on the questionnaires used in the Health Interview Survey during July 1967-June 1968. tained from responses to questions such as Comparative Prevalence Estimates “What did the doctor say it was? What is the The prevalence estimate of chronic digestive cause of the condition? What part of the body is ~~ conditions derived from the half-sample inter- affected?” The reclassification of reported con- ~~ viewed during July 1967-June 1968 by the ditions into ICD categories means that preva- person approach was 115.7 conditions per 1,000 lence estimates are not always derived for the population, while the condition approach half- separate disease and condition categories shown sample produced a comparable estimate of 65.6 on the qu 12 estionnaire. conditions (table 2). For conditions which were Major activity Condition approach’ (July 1967-June 1968) Person approach? July-December 1967 Preschool (Under 6 years) . NOT ABLE TO TAKE PART AT ALL IN ER ORDINARY PLAY WITH OTH CHILDREN. . ABLE TO PLAY WITH OTHER CHILDREN BUT LIMITED IN KIND OR AMOUNT OF Y. . NOT LIMITED IN ANY OF THE ABOVE WAYS. 15a. In terms of health, is — — able to take part at all in ordinary ploy with other children? b. Is he limited in the kind or amount of play because of his health? Going to school (6-16 years) . NOT ABLE TO GO TO SCHOOL AT ALL. ABLE TO GO TO SCHOOL BUT LIMITED TO CERTAIN TYPES OF SCHOOLS OR IN SCHOOL ATTENDANCE. E TO GO TO SCHOOL BUT LIMITED ne eR ACTIVITIES. . NOT LIMITED IN ANY OF THE ABOVE WAYS. 200. Does (would) he have to go to a certain type of school because of his health? Is — ~ limited in the kind or amount of other activities because of his health? Usually working (17+ years) - NOT ABLE TO WORK AT ALL. ABLE TO WORK BUT LIMITED IN KIND OR AMOUNT OF WORK. . ABLE TO WORK BUT LIMITED IN KIND OR AMOUNT OF OTHER ACTIVITIES. . NOT Lwin IN ANY OF THE ABOVE Is — — able to work at all? 18a. Is — — limited in the kind or amount of work he con do becouse of his health? c. Is — = limited in the | kind or amount of other activities because of his health? Keeping house (17+ years) ~ - NOT ABLE TO KEEP HOUSE AT ALL. . ABLE TO KEEP HOUSE BUT LIMITED IN KIND OR AMOUNT OF HOUSEWORK . ABLE FO KEEP HOUSE BUT LIMITED IN KIND OR AMOUNT OF OTHER ACTIVITIES. . NOT LIMITED IN ANY OF THE ABOVE WAYS. c. Is = — limited in the kind or amount of other activit 18a. Is — — limited in the kind or amount of work he can do becouse of his health? s because of his health? Retired (45+ years) NOT ABLE TO WORK AT ALL. ABLE TO WORK BUT LIMITED IN KIND OR AMOUNT OF WORK. ABLE TO WORK BUT LIMITED IN KIND OR AMOUNT OF OTHER ACTIVITIES. . NOT LIMITED IN ANY OF THE WAYS. ABOVE 17a. In terms of health, is — — able to work? b. Is*= ~ limited in the kind or amount of work he could do because of his health? January-June 1968 Preschool (Under 6 years) | NOT ABLE TO TAKE PART AT ALL IN ORDINARY PLAY WITH OTHER CHILDREN, ABLE TO PLAY WITH OTHER CHILDREN BUT LIMITED IN KIND OR AMOUNT OF NOT LIMITED IN ANY OF THE ABOVE WAYS. c. Is he limited in the amount of play because of his health? Going to school (6-16 years) NOT ABLE TO GO TO SCHOOL AT ALL. . ABLE TO GO TO SCHOOL BUT LIMITED TO CERTAIN TYPES OF SCHOOLS OR IN SCHOOL ATTENDANCE. . ABLE TO GO TO SCHOOL BUT LIMITED IN OTHER ACTIVITIES. NOT LIMITED IN ANY OF THE ABOVE WAYS, 240. Does (would) —~ have to go to a certain type of school because of his health? c. Is he limited in the kind or amount of other activities because of his health? Usually working (17+ years) NOT ABLE TO WORK AT ALL. ABLE TO WORK BUT LIMITED IN KIND OR AMOUNT OF WORK ABLE TO WORK BUT LIMITED IN KIND OR AMOUNT OF OTHER ACTIVITIES. . NOT Heo IN ANY OF THE ABOVE W 22a. In terms of health, is —— PRESENTLY able to work at all? d. Is he limited in the kind or amount of other activities because of his health? Keeping house (17+ years) "” - NOT ABLE TO KEEP HOUSE AT ALL. ABLE TO KEEP HOUSE BUT LIMITED IN KIND OR AMOUNT OF HOUSEWORK. ABLE TO KEEP HOUSE BUT LIMITED IN KIND OR AMOUNT OF OTHER ACTIVITIES. NOT LIMITED IN ANY OF THE ABOVE WAYS. 22a. In terms of health, is —— PRESENTLY able to work at all? d. Is he limited in the kind or amount of other activities because of his health? Retired (45+ years) NOT ABLE TO WORK AT ALL. ABLE TO WORK BUT LIMITED IN KIND OR AMOUNT OF WORK. ABLE TO WORK BUT LIMITED IN KIND OR AMOUNT OF OTHER ACTIVITIES. NOT Lv 1e0 IN ANY OF THE ABOVE W, 21a. Does —~ health keep him from working? d. Is he limited in the kind or amount of other activities because of his health? Figure 5. The different questionnaire formats used for the collection of data on limitation of activity in the Health Interview Survey during July 1967-June 1968. 13 mentioned in both lists, such as stomach ulcer, hernia, and conditions of the liver and the gallbladder, the estimates derived from the half-samples were quite similar. The higher prevalence rates produced from the question- naire with the person approach were attributable to the increased reporting of conditions that were not included in the list of chronic condi- tions on the condition approach questionnaire. The conditions on the checklist that contributed to increases in specific categories (shown in table 2) are: Checklist Condition Specific Categary hooks bendwien (table 2) 1. Gastritis . . . . ....... Gastritis and duodenitis Functional and symptomatic upper gastrointestinal disorders 2. Frequent indigestion 3. Diverticulitis Enteritis Colitis (ulcerative) Chronic enteritis and ulcerative colitis 4. Constipation Constipation Other functional disorders of the intestines 5. Colitis (mucous) Spastic colon 6. Any other bowel trouble Any other intestinal trouble Intestinal and bowel trouble 7. Any disease of the PANCIEBS . . wo; wos ow x 8 Other diseases of the digestive system This increased prevalence produced by the more detailed checklist of conditions is further confirmation that more complete reporting of the prevalence of a specific condition is accom- plished by the interviewer’s referring directly to the condition than by the respondent’s volun- teering information about the condition in answer to a more general probe question. If, on the other hand, interest lies in the prevalence of digestive conditions that have had some impact on the life of the individuals involved—e.g., inability to carry on the usual activity for their age or sex group or limitation in amount or kind of usual or other activities— then the collection procedures are equally effec- tive. It is obvious from table 3 that there are no marked differences because of the collection approach in the diagnostic categories that con- tribute to the approximate rate of 8 digestive conditions per 1,000 population that cause 14 limitation of activity. This finding substantiates the premise that the interview method of collec- tion is most appropriate for the collection of data on chronic conditions that have had some impact on the population. Furthermore, it is reassuring to find that the expansion of the collection procedure providing total prevalence estimates does not unduly affect the estimates restricted to the narrower concept of prevalence of conditions causing activity limitation. Since the split-sample procedure was in operation only during the period when total prevalence data were being collected for chronic digestive condi- tions, the above findings are based on the rates produced for such conditions. Their applica- bility to other types of chronic conditions may be influenced to some extent by such factors as the absence of medical attention or embarrass- ment associated with the reporting of certain kinds of conditions. LIMITATION OF ACTIVITY In the condition approach used in the survey during July 1957-June 1967 and in the half- sample during July 1967-June 1968, informa- tion about chronic limitation of the activity was obtained only for sample persons who reported one chronic condition or more. In accordance with the major activity of the person (working, keeping house, going to school, or retired), the respondent was handed a card (figure 5) by the interviewer and asked to select the limitation category which best suited the persons in terms of health. For those with any degree of activity limitation, the causative condition(s) were assigned from those that had been reported previously. In the person approach used in the July 1967-June 1968 half-sample, the restriction re- garding the presence of chronic condition(s) was removed, and questions regarding the degree of activity limitation were asked for all sample persons. Chronic conditions causing the limita- tion, which may or may not have been pre- viously reported in response to other probe questions, were then ascertained. If only acute conditions such as injuries were reported as dRetired persons were asked to rate themselves in terms of their ability to work. causing limitation, the person was considered to have no chronic limitation of activity. There were two major changes in the collec- tion of data on activity limitation in the person approach: (1) all persons were asked questions pertaining to chronic activity limitation regard- less of whether a chronic condition had been reported, and (2) instead of asking the respond- ent to select the appropriate activity limitation status from a printed card, the interviewer read the options to the respondent (figure 5). Questionnaire Revisions Effective in January 1968 Interviewing problems encountered during the 6-month period July-December 1967 led to extensive changes in the activity limitation section of the person approach questionnaire used for data collection during January-June 1968. As shown in figure 5, persons who had been working, keeping house, or going to school during most of the 12 months prior to interview in the earlier version were first asked about limitation in amount or kind of major activity. In the revised version, the question regarding ability to carry on major activity was asked first, followed by separate questions designed to obtain information on limitation in the amount and in the kind of major activity. For retired persons, inability to work was determined from responses to the question “In terms of health, is able to work?” in the earlier version. In the revised version, this question was worded “Does health keep him from working?” The order in which the questions were asked and the division of the question regarding limitation in amout or kind of major activity had little effect on the estimates derived from the two versions of the person approach ques- tionnaire. However, the revised wording of the question on inability to work produced a major change in the percent of retired persons in this category (table 4). It is difficult to obtain reliable data on limitation of activity. Unlike other health re- lated items such as disability days or physician visits that are objective by nature, limitation of activity represents, in a sense, an opinion or attitude on the part of the respondent in the interview. This is particularly true in the case of a retired person who is, in effect, placed in a hypothetical situation involving work and asked to describe his ability to function as a worker. Because the questions relating to activity limita- tion require a subjective judgment on the part of the respondent, even minor changes in wording may cause marked variations in response. Despite the disparity in the estimates ob- tained from the two versions of the person approach questionnaire, data collected in the two versions have been combined, and estimates have been derived for comparison with those produced from the condition approach. Comparative Estimates of Activity Limitation From the data in table 5, it is apparent that the estimates of the percent of persons with activity limitation derived from the condition approach are higher than those obtained from the person approach. However, when the per- cent of those unable to carry on their major activity was combined with the percent who were limited in amount or kind of major activity, the two approaches produced very similar estimates. The deficiency in the person approach, if it is a true one, appears to be the failure to differentiate those who are limited in “other than major activity” from those with no limitation of activity. On the other hand, it is possible that the condition approach tends to overestimate the percentage of persons with limitation in other than their major activity. With the exception of persons under 17 years, the percent of persons with activity limitation in each of the sex and age groups is significantly higher when derived from the condition ap- proach than when obtained from the person approach. While the difference in the two approaches in terms of percentage points was greatest among persons 65 years and over, the relative difference in terms of percentages is greatest in the 17-44 age group (table 5). Since the greatest disparity in the measure- ment of activity limitation by the person and condition approaches occurred among persons 17 years and over, the various degrees of limitation were examined by major activity status (table 6). The general patterns in limitation of activity as measured by the two approaches for the 15 various age groups were also present by major activity status. As mentioned earlier, when the two types of limitation related to the person’s activity (inability to carry on major activity, and limited in amount or kind of major activity) are combined, the differences in the two approaches are negligible. However, the estimate of the percent of persons with limitation in “other than major activity” derived from the person approach was appreciably lower than that based on the condition approach, particularly for the major activity limitation groups comprised of persons 17 years and over. This discrepancy between the two methods, which was consist- ently present for males and females and for the several age groups shown in table 5, may be due to confusion in the person approach arising from the wording of the question regarding activity limitation in activities other than the major one (figure 3). For example, the question relating to this degree of limitation (Q21c, Q22¢, and Q24c in the January-June 1968 version) might have been more definitive if they had been prefaced with “even though is able to (work, keep house, go to school), ....” With this introduc- tory clause, which is used in part on the questionnaire with the condition approach, the meaning of “other activities” becomes more specific. As previously mentioned, the estimates of the percent of persons with chronic limitation affecting their major activity were approxi- mately the same, regardless of the questionnaire approach used. However, the person approach consistently produced higher estimates of the percent of persons who were unable to carry on a major activity, while the condition approach generally produces higher estimates of the per- centage of persons limited in amount or kind of major activity (table 5). The compensating nature of these percentages cannot be explained by specific changes in questionnaire format. It seems to be most characteristic of persons 65 years and over, a population segment in which the measurement of activity limitation is admit- tedly vague and ill defined. Questionnaire Changes in 1969 and 1970 The general format of the section of the 1969 questionnaire dealing with activity limitation 16 was essentially the same as that used in the person approach during July 1967-December 1968. However, a question was added that was intended to improve the estimates of persons with limitation in amount or kind of major or other activities. Each person whose responses to direct questions indicated that he was able to carry on his major activity and was not limited in the amount or kind of major or other activities was asked the additional questions “Is limited in ANY WAY because of a disability of health?” If a positive reply was obtained, the interviewer then asked “In what way is he limited?” Since the interviewer was instructed to record verbatim responses to this question, it then became the responsibility of the coder to reclassify responses, where appro- priate, to the proper limitation category, which would usually be limitation in amount or kind of either major or other activities. Another question relating to the duration of the present degree of limitation was added to the question- naire. This question, however, was added not to improve data collection but to provide for more meaningful analysis of the data. In the 1970 questionnaire, the questions added to the 1969 questionnaire were retained. In addition, the activity limitation questions for persons who had been working most of the past 12 months were prefaced with an introductory question, “Does now have a job?” This question eliminated the need for asking the question “Is he able to work at all?” for persons who were obviously in good health and working at the time of the interview. Division of the retired population into those who had retired for health reasons and those who had retired for other reasons was a change made for the purpose of providing more meaningful data for analysis. Chronic Conditions Causing Activity Limitation The estimated average number of chronic conditions reported as causing a limitation of activity was slightly higher when using the con- dition approach responses than when using the person approach responses, 1.4 compared with 1.3 conditions per limited person (table 7). The greatest difference in the average number of limiting conditions obtained from the two approaches was among persons who were com- pletely unable to carry on the major activity for their age and sex group (work, keep house, or go to school). The amount of difference in the approaches was much greater by degree of limitation than by age of limited persons. The difference in procedure by which information on limiting conditions was collected in the two approaches was probably responsible for the greater frequency of multiple conditions in the condition approach. Thus, in the questionnaire employing this approach, information about all types of illness for a given person was obtained by illness-recall (probe) questions; more detailed information was obtained for each of these conditions by questions on the condition page. After the interviewer had filled out a condition page for all chronic conditions, the respondent was asked to select from a card the activity limitation status which best described him. He was then asked to name the conditions, about which he had already given information, that were responsible for limiting his activity. In this type of interview situation, a respondent who has several limiting conditions is inclined to ascribe responsibility to all of them or at least to a goodly number of those that he has reported to the interviewer. In the person approach, the respondent was asked questions about his activity-limitation status. Responses to these questions determined whether activity limitation was chronic. If it was, the respondent was asked about the condi- tions causing the limitation. Since the person in this situation was asked to name conditions causing his limitations, with no previous assist- ance from the questions on the condition pages, he would usually name fewer conditions. Ob- viously, since activity limitation is less frequent- ly attributed to multiple chronic conditions in the person approach, a smaller estimate of activity-limiting conditions will be obtained from the person approach than from the condi- tion approach. From the distribution of condi- tions causing limitation shown in table 8, it is apparent that heart conditions and arthritis (or rheumatism), the two leading causes of limita- tion, accounted for about the same amount of activity limitation as was seen earlier,® regardless of the approach used or the degree of limitation considered. Even though the two principal causes of limitation were reported at about the same rate, regardless of the approach used, other causes were reported less frequently in the person approach. There are other contributing factors for this besides the lower reporting of multiple limiting conditions in the person approach. Conditions that the respondent is hesistant to report to the interviewer because he feels they are embarrass- ing, such as mental and nervous disorders, are known to be underreported in a household survey. If such conditions are included in a checklist read by the interviewer, there is less reluctance on the part of the respondent to report them, but if he must volunteer the condition—e.g., as a cause of activity limitation in the person approach—the underreporting will be greater. The percent of persons limited in activity and reporting mental illness as a cause of limitation in the person approach was essentially half that derived from the condition approach (table 8). In general, impairments were reported as causes of activity limitation at a much lower rate in the person approach than in the condition approach. It is reasonable to assume that re- spondents will volunteer an impairment as a cause of limitation if it is the only cause. On the other hand, if the impairment is so longstanding that the person has learned to live with it and another, more recently acquired condition is present, in many cases the impairment would not be named as a contributing cause of limita- tion. In addition to these factors, there is the added possibility that conditions involving the body system for which information is collected in a given year may tend to be overreported as causes of limitation. It is likely that a new question on the person approach questionnaire about the main condi tion causing limitation instead of about multiple limiting conditions will be used as the basis for future analytical studies on activity limitation prepared from survey data. This measure would be less subject to response error and would provide a higher degree of year-to-year con- sistency in the data. 17 INCIDENCE OF ACUTE CONDITIONS Collection of Data on Acute Conditions As indicated earlier, one of the basic differ- ences between the (wo questionnaires used during July 1967-June 1968 was related to the method of obtaining information about acute conditions from which incidence estimates were derived. In the condition approach, all conditions that were reported in response to the general illness- recall questions were recorded by the inter viewer regardless of the degree of severity. Because methodological studies have shown that health related events are more completely re- ported if they have had some impact on the individual, severity criteria were applied at the coding level to exclude minor conditions, which are usually poorly reported. Conditions which met two criteria—(1) the presence of medical attention or one day or more of restricted activity or both and (2) conditions which had their onset during the 2-week period prior to the week of interview—were used as a basis for the derivation of estimates of the incidence of acute conditions. In the person approach, the initial questions regarding health pertained to days of disability and physician visits experienced by household members during the 2-week period prior to the interview (figure 1). When disability days or physician visits were reported, then information about conditions causing these health related actions was obtained. The severity criteria of restricted activity or medical attention or both of these were thereby applied automatically at the interview level. Thus, the interviewer would complete condition pages only for acute conditions that caused disability or required medical care during the 2-week period prior to interview, and only these acute conditions would reach the coding level. Sources of Differences in Estimates With the exception of persons 65 years and over, the incidence of all acute conditions per 100 persons derived from data collected by the condition approach was higher by sex and age (table 9) than that produced from the person approach data. It is obvious from table 9 that the difference was primarily the result of fewer conditions that received medical attention but caused no days of restricted activity being reported in the person approach questionnaire. Estimates of physician visits derived from the two approaches were compared to determine if the lower rate of medically attended acute conditions from the person approach might be duc to the failure of respondents to report some of their physician visits (table 10). Among chil- dren under 6 years, where the incidence of medically attended acute conditions derived from the person approach was markedly lower than that produced by the condition approach, the number of physician visits per child per year was likewise lower for the data collected by the person approach (table 10). However, in other age intervals, in which the incidence of medi- cally attended conditions derived from the person approach was also quite low, there is no appreciable difference in the rate of physician visits produced from the two approaches. This would indicate that additional factors may be contributing to the differences in the acute condition incidence rates. Further investigation of the estimates of the incidence of acute conditions revealed that the differences in the rates derived from the two approaches were contributed for the most part by upper respiratory conditions and injuries (table 11). More specifically, the rates for the common cold and for open wounds and lacera- tions were markedly different (table 12). Because this difference exists only for condi- tions that were reported as medically attended with no activity restriction, it seems reasonable that respondent error about medical attention may be responsible for the noted variations. The information about medical attention for a cer- tain condition is obtained, in both approaches, from the question “Did ever at any time talk to a doctor about his...?” For acute conditions that are basic to incidence esti- mates—l.e., those that have their onset during the 2-week period prior to week of interview— the medical attention, if received, should have been received during the same period and should have been reported as a physician visit in response to the probe questions about physician visits. In the condition approach, however, it is possible that the respondent interpreted the phrase “at any time” in the question under discussion to mean before the 2-week reference period, possibly about an earlier condition of the same type, such as a cold, or even more recent medical visits during the interview week itself. On the other hand, such a situation would not normally arise in the person approach because information about the condition would be obtained either because medical attention had been sought or activity had been restricted during the reference period. If restricted activity was responsible for the reporting of the condi- tion and medical attention occurring outside the reference period was mistakenly reported on the condition page, the condition would be classi- fied as one both medically attended and activity restricting. Another reporting irregularity which may occur in the condition approach and is partic- ularly applicable to injuries is the reporting of the present effects of an injury as medically attended when actually the treatment was re- ceived for the original injury, which happened prior to the 2-week reference period. For exam- ple, an infection at the site of an injury, developing within the 2-week reference period and for which no medical treatment was re- ceived during that period, should not be re- ported as medically attended if the physician was seen only for the treatment of the original injury that happened a month before the inter- view. This discussion leads to the conclusion that differences in the estimates of medically attended acute conditions derived from the two approaches may be attributable to the combined effects of underreporting of physician visits in the person approach and overreporting of medi- cal attention for a specific condition in the condition approach. Questionnaire Changes Even before the collection year July 1967- June 1968 was over, it was apparent from preliminary tabulations that the incidence of medically attended acute conditions represented a reporting problem. Because it could not be determined at this point which approach was at fault and because the person approach was the experimental one, efforts were made to increase the number of conditions reported in this area. On the second version of the person approach questionnaire, used during January-June 1968, additional illness-recall (probe) questions relat- ing to dental visits and to hospitalizations during the 2-week reference period were used. These questions were added to insure that information about acute or chronic conditions ordinarily treated by a dentist and about conditions such as injuries for which a person was hospitalized immediately would be obtained by the health related action probe questions used in the per- son approach. Another change in the second version of the person approach questionnaire was the elimina- tion of the place of the physician visit in the initial probe for physician visits occurring during the 2 weeks prior to interview. It was felt that an inquiry about visits “either at home, at a doctor’s office, or at a clinic’ might have a limiting effect on the visits reported. For exam- ple, there was a possibility that physician visits to first aid stations, industrial health units, or similar places were not being reported, and consequently conditions treated in these places were also not reported. The illness-recall questions on the 1969 ques- tionnaire were quite similar to those in the second version of the person approach question- naire used during January-June 1968 and con- tinuing through December 1968. However, by the time the 1970 questionnaire was being planned, it was possible to make use of the findings from a comparative analysis of the two approaches. As a result, three major revisions were made: (1) the probe question relating to hospitalizations during the 2-week reference period was eliminated because information about the number of conditions obtained by this question (which would not be reported in response to other probe questions) was negli- gible; (2) a specific probe question for injuries or accidents occurring in the 2-week reference period was added on a trial basis to determine if those that result in medical attention or activity restriction fail to be reported in response to the health related action probe questions; and (3) the question on the condition page relating to medical attention “Did ever at any time talk to a doctor about his. ..?” was changed to 19 “When did last see or talk to a doctor about his. ..?” This last change definitely estab- lishes whether the occurrence of medical atten- tion for the specific condition was during the 2-week reference period. DISABILITY DAYS In the condition approach, for each acute or chronic condition, discovered by the illness- recall questions, further questions were asked on the condition page about the occurrence of restricted activity, bed disability, or work- or school-loss days because of the specific condi- tion. Estimates of the number of disability days were later derived for each condition. When these estimates were summed to obtain a total for all conditions, duplication did result to the extent that days were attributed to more than one condition. Interviewers were instructed to footnote instances in which this type of dupli- cation occurred. From this information it was possible for coders to unduplicate the count of disability days reported and to code the disabil- ity days experienced on a person basis. This procedure was reversed in the person approach. From the health related-action probe questions relating to disability days, it was possible to obtain a direct unduplicated count of days of person disability. To obtain the number of condition days of disability, it was necessary to ask the disability questions on the condition page for each specific condition reported. Days of Person Disability The measures of days of person disability estimated from data collected by the condition and person approaches are conceptually the same. Even though they would be expected to differ to some extent because of the change in the collection procedure, the number of disabil- ity days per person per year derived from the two approaches (table 13) was quite similar. The condition approach yielded higher rates of dis- ability days among persons 17-64 years than did the person approach, but the person approach produced higher rates for those under 17 years and those 65 years and over. For the most part, however, differences in comparable rates by age 20 and by sex derived from the two approaches were within the range of sampling error. Days of Disability Due to Acute and Chronic Conditions The relationship of disability days associated with persons and those associated with condi- tions is shown in table 14. Restricted activity is, by definition, the most inclusive measure of disability used in the survey. All days of bed disability, work loss, and school loss are also days of restricted activity. Days of work loss and school loss are included as days of bed disability if the condition caused the person to be in bed all or most of the day when the time was lost from work or school. Restricted activity and bed disability refer to the total survey population. Work-loss days are reported only for the cur- rently employed population 17 years and older, and school-loss days are reported only for persons 6-16 years of age. As explained earlier, the number of condition days per person per year is always greater than the number of person days. The excess days, when expressed as Condition days per person 100 Person days per person - 100 represent the percent of person days that were associated with more than one condition (ta- ble 14). The proportion of restricted activity and bed disability days attributable to more than one condition is considerably higher than for work- or school-loss days, regardless of the type of approach used in the interview. The high frequency of multiple conditions, particularly chronic conditions, in the older population, a segment of the population included in the measurement of restricted activity and bed disability, is responsible for this difference. It has been found from evaluative studies that a single episode of illness during the 2-week reference period is more likely to involve two chronic conditions or a chronic and an acute condition than to involve two acute conditions. Since acute conditions occur more frequently than chronic conditions among the school and employed populations, the probability of two conditions occurring concurrently in persons in these populations is much lower than that for the total population. The higher percentage of person days of restricted activity and bed disability days due to two conditions or more, in estimates derived from the person approach may be the result of the collection procedures. In the person ap- proach, the estimates of person and condition days are derived from separate areas on the questionnaire, with the information in both cases obtained directly from the respondent. In the condition approach, the person days are derived from the total number of condition days, and the estimate of the person days associated with more than one condition is entirely dependent on the footnoting of dupli- cated days by the interviewer and the transcrip- tion of this material by the coder. Lapses on the part of either the interviewer or the coder would decrease the number of duplicated days and thus increase the number of days of person disability. In general, the person approach is more productive of days attributable to acute condi- tions, while the condition approach produces generally higher rates of disability days asso- ciated with chronic conditions (table 14). This phenomenon may be due to reporting differ- ences inherent in the particular approach used. If early in the interview before any reference was made to particular types of conditions a respondent was questioned about days of dis- ability occurring during the 2-week reference period, he would probably associate them with the most immediate cause of disability, e.g., a cold, influenza, or a current injury. Even though the respondent is further questioned in the person approach about any other condition(s) associated with the disability days, he does not always relate a continuing underlying condition to current disability. On the other hand, when a respondent is questioned first about the pres- ence of conditions, both acute and chronic, as in the condition approach, and then, in connec- tion with each condition, about associated dis- ability days, he is more likely to report days of disability attributable to chronic conditions. Days of Disability Due to Acute Conditions It is apparent from table 9 that the incidence of acute conditions causing restriction of activ- ity, with or without medical attention, is higher when derived from the person approach than from the condition approach. It would be expected, therefore, that the rate of disability days associated with acute conditions would also be higher when derived from the person ap- proach (table 15). The higher rates of restricted activity days due to acute conditions derived from the person approach are attributable to respiratory condi- tions, digestive conditions, and the residual group “other acute conditions.” In general, the same types of conditions together with injuries are responsible for the slightly higher rate of bed disability days produced by the person ap- proach. Rates of work-loss days due to acute conditions per person were practically the same for each of the specific condition groups regard- less of the approach used in the collection of the data. Among children 6-16 years, the rates of school loss due to upper respiratory conditions and influenza were substantially higher when derived from the sample in which the person approach was used. The same general pattern of higher rates of disability days due to acute conditions from the person approach was apparent when rates were based on the number of conditions (number of days per condition) rather than on the popula- tion (table 16). Disability days associated with injuries (by specific type) were markedly higher when derived from the person approach. This difference is, of course, due to the greater number of injuries with “medical attention only” reported in response to the condition approach. When rates are computed, the condi- tions which contribute no days to the numerator inflate the denominator and cause a lower rate of days per condition. Days of Disability Due to Chronic Conditions Rates of restricted activity and work-loss days associated with chronic conditions derived from data collected by the condition approach were somewhat higher than those obtained from the person approach. On the other hand, the two approaches yielded rates of bed disability days and school-loss days of about the same magni- tude (table 17). 21 The difference in days of restricted activity derived from the two approaches is largely the result of higher rates by the condition approach for respiratory conditions (other than asthma or hay fever), mental and nervous conditions, and orthopedic impairments. For work loss, how- ever, the difference in days per person in the population per year in the two approaches was not more than one-tenth of a day for any specific chronic condition. A more meaningful comparison of the dis- ability days attributable to certain chronic con- ditions could be made by computing the number of disability days per condition for specific types of chronic conditions. However, this type of comparison on all conditions was not possible because the person approach provides total prevalence data only for conditions involving the body system for which the collection of infor- mation is scheduled during the year. This type of comparison is shown in table 18 for days of restricted activity and bed disability involving selected conditions of the digestive system. Even though the number of restricted-activity days associated with chronic digestive conditions per person in the population is higher from data collected by the person approach (table 17), the condition approach produces a higher estimate of restricted-activity days per chronic digestive condition (table 18). This reversal would indi- cate that more serious conditions with many disability days are discovered by the condition approach, while the person approach, in which the more detailed checklist of conditions is used, results in the reporting of more conditions as shown in table 2. Disability days associated with the less severe conditions discovered by the person approach would be fewer per condition but would accumulate to the degree that a higher rate of days would be obtained when the entire population is used as a base figure. PHYSICIAN VISITS In the person approach questionnaire, which was introduced in the collection phase of the survey in July 1967, the probe questions relating to physician visits occurring during the 2-week period prior to week of interview serve a dual purpose. They are used to discover conditions, principally acute, for which a physician was seen 22 and to provide information basic to the deriva- tion of estimates of the volume of physician visits in a given year. For this reason, the comparative estimates of physician visits per person per year were discussed to some extent in the section of this report dealing with the incidence of acute conditions. In addition to the material presented there, comparative estimates from the two approaches are presented in table 19 by selected demographic characteristics. Volume of Physician Visits The lower estimate of physician visits per person per year among small children derived from the person approach seems to indicate a response or reporting problem which hopefully will be corrected by revisions made in the 1969 and 1970 questionnaires. For other age groups and for segments of the population classified by sex, race, education, and geographic region, the differences in the rates of physician visits by the two approaches were consistent and only slight- ly lower when derived from data collected by the person approach. Persons Seen by a Physician During the Past Year In contrast with the visits per person per year, the percent of the population that saw a physician at least once during the past year was higher for all population groups when based on data collected by the person approach (table 19). Because of this consistency of higher percentages of persons with recent medical attention based on the person approach, inequi- ties in the distribution of the populations in the two half-samples shown in table 1 were not responsible for the differences in the two ap- proaches. Even though the specific questions from which the information on interval since the last physician visit was obtained are identical on the two questionnaires, there are other factors which may have had some eftect on the re- sponses. In the condition approach, the question about the interval since the last physician visit was followed by a question relating to number of such visits in the past year. This question was not included in the person approach. Persons who had reported their most recent visit about a year prior to the interview might decide that the reported visit had actually occurred during the previous year when asked about the number of visits in the past year. Since there was no check of this kind on the person approach question- naire, it is possible that overreporting of visits during the year may have occurred. Another factor which may have contributed to the higher percentage in the person approach was the addition of a probe question in the version of the questionnaire used during January-June 1968. This question, which pre- ceded the section on physician visits, related to hospitalization occurring during the 2-week period prior to interview. This question might remind respondents of recent hospitalizations during which a doctor was seen. The slightly higher percentages yielded by the second version of the person approach questionnaire in com- parison with the first version support this possibility. HOSPITALIZATION With the exception of areas of the question- naire dealing with personal characteristics of household members, the sections dealing with hospitalization during the 12 months prior to interview were the only ones that remained the same in format, purpose, and content on the questionnaires employing the condition and person approaches. Use of the questions on physician visits in the person approach as probes to discover conditions as well as to obtain information on the use of medical services changed the purpose and location of these questions in the questionnaire even though their format and content were the same in the two approaches. The fact that the wording of the hospital questions on the two ques- tionnaires was identical provided an oppor- tunity to compare the estimates derived from the two half-samples for a health related item unaffected by major differences in the data collection procedures. The similarity of the rates shown in table 20 for short-stay discharges per 100 persons and for average length of hospital stay attests to the reproducibility of estimates by the interview method in populations of similar age and sex composition (table 1). All of the differences in comparable estimates shown in table 20 are within the limits of sampling error. CONCLUSIONS Investigation of the estimates derived from data collected by the person and condition approaches used in the Health Interview Survey during July 1967-June 1968 leads to the conclu- sion that the person approach is feasible as an interviewing technique. This investigation also indicated that no drastic changes in levels and relationships of health measures have resulted from the adoption of the person approach in the collection phase of the ongoing survey. The decision to gather data in two half- samples using alternative collection methods was justified by the following: 1. Areas in which questions in the person approach needed revision were identified by the magnitude of differences in com- parative rates derived from the two ap- proaches. 2. Imperfections in the condition approach which had persisted throughout the first 10 years of the survey because no comparative data had been available were detected by the evaluation of the data in the half- samples. 3. This experiment has reemphasized some of the problems in the collection of data on activity limitation in the Health Interview Survey. Neither of the approaches used during July 1967-June 1968 produced esti- mates that could be described as objective measures of chronic limitation of activity. Response bias that was introduced by differences in attitudes, judgment, and in- telligence of interview respondents is prob- ably greater in this area of the survey than in any other. More research is needed to develop questions that relate specifically to a person’s ability to function within his current area of activity before less subjec- tive measures of activity limitation can be applied to all segments of the population. 23 4. Restriction of the collection of information on chronic conditions to specific types of conditions during a given data collection year results in a larger number of reported conditions affecting the system under con- sideration. This restriction, which permits the use of a list of conditions that are relevant to a particular body system, pro- duces adequate prevalence data without distorting the distribution of chronic condi- tions causing activity limitation. REFERENCES 1y.s. National Health Survey: The statistical design of the Health Household-Interview Survey. Health Statistics. PHS Pub. No. 584-A2. Public Health Service. Washington. U.S. Govern- ment Printing Office, July 1958. ?National Center for Health Statistics: Health survey proce- dure. Vital and Health Statistics. PHS Pub. No. 1000-Series 1-No. 2. Public Health Service. Washington. U.S. Government Printing Office, May 1964. “National Center for Health Statistics: Health interview responses compared with medical records. Vital and Health Statistics. PHS Pub. No. 1000-Series 2-No. 7. Public Health Service. Washington. U.S. Government Printing Office, July 1965. 4National Center for Health Statistics: Interview data on chronic conditions compared with information derived from medical records. Vital and Health Statistics. PHS Pub. No. 1000-Series 2-No. 23. Public Health Service. Washington. U.S. Government Printing Office, May 1967. 5National Center for Health Statistics: Synthetic State Esti- mates of Disability Derived From the National Health Survey. PHS *Pub. No. 1759. Public Health Service. Washington. U.S. Government Printing Office, 1968. 6National Center for Health Statistics: Chronic Conditions Causing Activity Limitation, United States, July 1963-June 24 1965. Vital and Health Statistics. PHS Pub. No. 1000-Series 10- No. 51. Public Health Service. Washington. U.S. Government Printing Office, Feb. 1969. TNational Center for Health Statistics: Estimation and sam- pling variance in the Health Interview Survey. Vital and Health Statistics. PHS Pub. No. 1000-Series 2-No. 38. Public Health Service. Washington. U.S. Government Printing Office, June 1970. 8National Center for Health Statistics: Reporting of hospi- talization in the Health Interview Survey. Vital and Health Statistics. PHS Pub. No. 1000-Series 2-No. 6. Public Health Service. Washington. U.S. Government Printing Office, July 1965. National Center for Health Statistics: Comparison of hospi- talization reporting in three survey procedures. Vital and Health Statistics. PHS Pub. No. 1000-Series 2-No. 8. Public Health Service. Washington. U.S. Government Printing Office, July 1965. 10National Center for Health Statistics: The influence of interview and respondent psychological and behavioral variables on the reporting in household interviews. Vital and Health Statistics. PHS Pub. No. 1000-Series 2-No. 26. Public Health Service. Washington. U.S. Government Printing Office, Mar. 1968. Table 1. 10. 11. 12. 13. LIST OF DETAILED TABLES Number and percent distribution of the population in the two half-samples used in the Health Interview Survey by selected characteristics, according to type of questionnaire, July 1967-June 1968 . . . . . .. .......... Comparative prevalence estimates of chronic conditions of the digestive system, by type of condition and questionnaire obtained from the two half-samples used in the Health Interview Survey, July 1967-June 1968 Comparative prevalence estimates of chronic conditions of the digestive system that cause activity limitation, by type of condition and questionnaire obtained from the two half-samples used in the Health Interview Survey, July V0BZ-JUNE TOBE ; uv oi 5 % & i & MC 1 5 @ 5 5 505 Bin ER HW sw we ww ww ew he wen da 8 WEEE WR Estimates of the percent of persons with limitation of major activity, by type of limitation and time period derived from the two versions of the person approach questionnaire, July-December 1967 and January-June 1968 Percent of population with limitation of activity, by degree of limitation, type of questionnaire, sex, and age obtained from the two half-samples used in the Health Interview Survey, July 1967-June 1968 . . . . . . . . .. Percent of population with limitation of activity, by degree of limitation, type of questionnaire, and major activity obtained from the two half-samples used in the Health Interview Survey, July 1967-June 1968 . . . . . . . . .. Average number of limiting conditions for persons with activity limitation, by degree of limitation, age, and type of questionnaire obtained from the two half-samples used in the Health Interview Survey, July 1967-June 1968 Percent distribution of conditions causing activity limitation by type of condition, according to degree of limitation and type of questionnaire obtained from the two half-samples used in the Health Interview Survey, July TOBZ- JUNC 1OBB . . oo 0 oi av 6s % + 51 5 5 5 § 8 9 8 % 5 3 8 4 8 ® % # 0% bv Hem ww meme Incidence of acute conditions, by severity criteria, type of questionnaire, sex, and age obtained from the two half-samples used in the Health Interview Survey, July 1967-June 1968 . . . . . . . .. . .. .......... Incidence of medically attended acute conditions, with and without activity restriction, and physician visits, by age and type of questionnaire derived from the two half-samples used in the Health Interview Survey, July 1967-June TOBE . oo visi ES EF EES USE NH WW EEN EE EW EW Em wn wom ww www we ks am dom Bod BR Incidence of acute conditions, by condition category, type of questionnaire, sex, and age obtained from the two half-samples used in the Health Interview Survey, July 1967-June 1968 . . . . . . . ... ............ Incidence of common colds and wounds and lacerations, by severity criteria, type of questionnaire, sex, and age obtained from the two half-samples used in the Health Interview Survey, July 1967-June 1968 . . . . ... . .. Number of disability days per person per year, by type of disability, type of questionnaire, sex, and age obtained from the two half-samples used in the Health Interview Survey, July 1967-June 1968 . . . . ........... Page 27 28 30 31 32 33 33. 35 36 37 25 Table 14. 26 15. 16. 17. 19. 20. LIST OF DETAILED TABLES—Con. Number of person and condition days of disability and percent of person disability days associated with more than one condition, by type of disability day and type of questionnaire obtained from the two half-samples used in the Health Interview Survey, July 1967-June 1968 . . . . . . . . . . . Number of disability days per person per year associated with acute conditions, by condition category, type of disability day, and type of questionnaire obtained from the two half-samples used in the Health Interview Survey, July 1967-dune 1988 . . . . LL LL ee ee ee Number of disability days associated with acute conditions, by condition category, type of disability day, and type of questionnaire obtained from the two half-samples used in the Health Interview Survey, July 1967-June 1968 Number of disability days associated with chronic conditions, by condition category, type of disability day, and type of questionnaire obtained from the two half-samples used in the Health Interview Survey, July 1967-June 1968 Number of disability days associated with selected chronic digestive conditions, by type of condition, type of disability day, and type of questionnaire obtained from the two half-samples used in the Health Interview Survey, July 1967-June 1968 . . . LL. LLL Number of physician visits per person per year and percent of persons seen by a physician in past year, by selected characteristics and type of questionnaire obtained from the two half-samples used in the Health Interview Survey, JUIYIOBT-JURB TOBE. .. , 4 5 5 4 5 2 5 5 6 5 6 5 5 5 5 5 5 0 & wm ¢ a» 0 a soe om om m om wey esa Number of discharges from short-stay hospitals and average length of stay, by type of questionnaire, sex, and age obtained from the two half-samples used in the Health Interview Survey, July 1967-June 1968 . . . . . .. . .. Page 37 38 40 a1 42 43 Table 1. Number and percent distribution of the population in the two half-samples used in the Health Interview Survey by selected characteristics, according to type of questionnaire, July 1967-June 1968 Population Condition Person Condition Person characteristics approach approach approach approach Number of persons in thousands Percent distribution Total population’ . . ............ 94,657 99,904 100.0 100.0 sex NIBIE x in mls on srlmoe wom a2 6 0 vm 0 3 ml 45,643 48,038 48.3 48.1 Female . . . . i vi ie ie eee 48914 51,867 51.7 51.9 Age Under 17vyears . . . . . . «vv vv vv vu vu 32,541 34,632 34.4 34.6 17-44 years . . . . vv vv ieee 33,5619 35,813 35.4 35.8 AEB YEBIS + « + 5 sw vi PE TR WE MLE ew 19,511 20,358 20.6 20.4 OBYsars ant OVer « ou piv wis # 2 3s 5 ¢ © 3 4s @ 8,986 9,201 9.5 9.2 Color WAIE! 53 ws lm ptm 2 3 gles mo or 3 wr malas 83,061 87,806 87.8 87.9 ANOhEr . . u + ov sis sis ws m8 ms 88 @® sw vo 11,496 12,008 12.2 124 Family income Under$3000 . .+:vsmemimsmsmepn va 12,200 12,659 12.9 12.7 $300083999 . . vi 0iw ims ows ow sw sie we 11,689 3,322 12.4 13.3 $5,000-$9,999 . . . . Lo... eee ee 38,079 12,133 40.3 42.2 $10,000$14,999 . . . .. . . 17,668 18,095 18.7 18.1 $15,000and OVer . . . . LL. ee ee eee 9,201 8,702 9.7 8.7 Education of family head UnderQyears . . . . . . . «vv vv vv van 24,246 27,194 25.6 27.2 O-12VEArS + vv vt ee ee ee eee ee 46,487 48,561 49.2 48.6 1B3yearsand Over , « « «+ vs ps @ 3 8 ws 5 5 + 22,553 22,118 23.9 22.1 Region Northeast . . . . . . « «vv vv i ee eee 21,5630 26,343 22.8 26.4 NorthCentral + . + 4 +0 cs swe 5m we 29,368 25,713 31.1 25.7 South © vv 5 s 5s wa sm sw sms ww ws ws 29,370 29,426 31.1 29.5 WESE uous md RIBS E IE PE RE HE GEES 14,288 18,422 15.1 18.4 Residence SMSA'S ©. i ee ee 61,999 64,063 65.6 64.1 Outside SMSA's: NOMEBIIN. = 5 ws wv 0 3 ss Ba wews®s 27,820 30,868 29.4 30.9 FAI o : oo 4 5 4 5 sa en ERs Wm Ems & 4 4,737 4,974 5.0 5.0 ! Includes persons of unknown income or education. 27 Table 2. Comparative prevalence estimates of chronic conditions of the digestive system, by type of condition and questionnaire ob- tained from the two halfsamples used in the Health Interview Survey, July 1967-June 1968 Condition Person Condition Person Condition Person Condition approach approach approach approach approach approach Number of conditions Rate per Percent seen by a (in thousands) 1,000 population physician at any time All chronic digestive conditions . . . . . . . 6,200 11,560 65.6 115.7 94.1 84.7 Diseases of teeth and supporting structures 115 56 1.2 0.6 67.8 73.2 Diseases of buccal cavity and esophagus . . . . . . 53 114 0.6 1.1 94.3 86.0 Ulcer of stomach and duodenum . . . . . . . . . 1,911 1,887 20.2 18.9 98.5 98.0 Gastritis and duodenitis. . . . . .. ....... 70 780 0.7 7.8 100.0 84.1 Functional and symptomatic upper gastro- intestinal disorders. . . . . .......... 249 1,441 2.6 14.4 83.5 69.1 Appendicitis, all forms . . . . . * * % * * * Hernia (abdominal cavity) . . . . ... ...... 1,536 1,783 16.2 17.8 93.0 93.9 Chronic enteritis and ulcerative colitis . . . . . . . 216 872 2.3 8.7 97.7 93.9 Constipation . . . « + + + ovo 2 om memes 213 2,500 2.3 25.0 82.2 65.7 Other functional disorders of intestines . . . . . . 66 235 0.7 2.4 100.0 94.0 Intestinal or bowel trouble, NOS . . . . . .. .. 71 150 0.8 1.5 95.8 91.3 Liver trouble, NOS . . . . . . . .......... 98 64 1.0 0.6 90.8 92.2 Other specified diseases of liver, NEC . . . . . . . 67 27 0.7 0.8 100.0 100.0 Specified diseases of gallbladder . . . . . . .. .. 418 581 4.4 5.8 99.0 99.1 Gallbladder trouble, NOS. . . . . .. ...... 509 441 5.4 4.4 95.7 971 Stomach trouble, NOS . . . . . . . .. ...... 408 243 4.3 2.4 84.6 82.3 Other diseases of the digestive system . . . . . . . 173 314 1.8 3.1 99.4 95.9 28 Table 3. Comparative prevalence estimates of chronic conditions of the digestive system that cause activity limitation, by type of con- dition and questionnaire obtained from the two half-samples used in the Health Interview Survey, July 1967-June 1968 ch, Condition Person Condition Person Condition approach approach approach approach Number of conditions causing activity limitation Rats per > (in thousands) 1000 population All chronic digestive conditions . . . . . . 803 827 8.5 8.3 Diseases of teeth and supporting structures (6) (4) (0.1) (0.0) Diseases of buccal cavity and esophagus . . . . . . (6) (17) (0.1) (0.2) Ulcer of stomach and duodenum . . . ...... 212 175 2:2 1.8 Gastritis and duodenitis . . . ... ........ (3) (30) (0.0) (0.3) Functional and symptomatic upper gastro- intestinal disorders . . . . ........... (31) (24) (0.3) (0.2) Appendicitis, all forms . . . . ........... (2) (1) (0.0) (0.0) Hernia (abdominal cavity) . . ........... 241 247 2.5 2.5 Chronic enteritis and ulcerative colitis . . . . . . . (30) 60 (0.3) 0.6 Constipation . . . . «uae eee eee (7) (20) (0.1) (0.2) Other functional disorders of intestines . . . . . . (4) (14) (0.0) (0.1) Intestinal or bowel trouble, NOS . .. ...... (13) (12) (0.1) (0.1) Liver trouble, NOS . . . . . . . .. ........ (18) (5) (0.2) (0.1) Other specified diseases of liver, NEC . . . . . .. (20) (20) (0.2) (0.2) Specified diseases of gallbladder . . . .. ..... (40) 59 (0.4) 0.6 Gallbladder trouble, NOS . . ........... 53 (20) 0.6 (0.2) Stomach trouble, NOS . . . . . .......... 73 (37) 0.8 (0.4) Other diseases of the digestive system . . . . . . . (45) 80 (0.5) 0.8 NOTE: Estimates shown in parentheses in this table are not statistically reliable when considered individually. They have been retained so that the material can be related to the data shown in table 2. Table 4. Estimates of the percent of persons with limitation of major activity, by type of limitation and time period derived from the two versions of the person approach questionnaire, July-December 1967 and January-June 1968 Preschool Limitation status All and Usually Keeping Retired and time period activities’ school working house (under 17 years) Unable to carry on major activity: July-December 1967 . . . . . .. .. ... 2.8 0.3 0.4 0.6 47.3 January-June 1968 . . ........... 3.1 0.2 0.8 14 38.8 Limited in amount or kind of major activity: July-December 1967 . . . .. ....... 5.9 6.0 13.4 13.8 January-June 1968 . . . .......... 6.1 6.1 14.1 12.9 ! Includes other and unknown activity group. 29 Table 5. Percent of population with limitation of activity, by degree of limitation, type of questionnaire, sex, and age obtained from the two half-samples used in the Health Interview Survey, July 1967-June 1968 Sex Age Degree of limitation and All type of questionnaire persons Male Female Under 17 17-44 45-64 65 years years years years and over Percent All limited persons: . Condition approach . . . . . ...... 12.2 12.8 11.6 2.0 8.1 19.6 48.4 Person approach . . . . ......... 10.5 11.1 9.9 23 6.5 17.2 41.5 Unable to carry on major activity: Condition approach . . . ........ 2.2 3.2 1.3 0.2 0.7 2.8 13.9 Person approach . . . . ......... 3.0 4.4 WS 0.2 0.9 4.4 18.4 Limited in amount or kind of major activity: Condition approach . . . . ....... 6.6 6.6 6.6 0.7 4.3 11.1 26.9 Person approach . . . . . ........ 6.0 5.3 6.7 1.2 4.2 10.8 20.5 Limited in other than major activity: Condition approach . . . .. .. .... 3.3 29 37 1.1 3.0 8.7 1.7 Person approach . . . . ..... .... 1.5 1.4 1.6 1.5 2.1 2.6 30 Table 6. Percent of population with limitation of activity, by degree of limitation, type of questionnaire, and major activity obtained from the two half-samples used in the Health Interview Survey, July 1967-June 1968 a . i Other Degree of limitation All Preschool Going to Usually Keeping Retired | (17 years and type of questionnaire activities school working house and over) Percent All limited persons: Condition approach . . . . . ........ 12.2 1.1 25 10.4 19.8 64.5 18.5 Person approach . . . . ........... 10.5 1.4 2.8 8.2 17.1 57.1 18.9 Unable to carry on major activity: Condition approach . . . . ......... 2.2 0.1 0.2 0.3 0.8 32.5 7.8 Personapproach . . . . ........... 3.0 0.4 0.1 0.6 1.0 42.9 12.7 Limited in amount or kind of major activity: Condition approach . . . . . ........ 6.6 0.8 0.7 6.3 12.8 26.7 7.8 Person approach . . . . . .......... 6.0 09 1.3 6.1 13.8 13.3 3.8 Otherwise limited: Condition approach . . . . . ........ 3.3 0.1 1.6 3.7 6.2 53 29 Person approach . . . . . .......... 1.56 * 1.4 1.5 2.3 0.8 24 Not limited: Condition approach . . . . . .... .... 87.8 98.9 97.5 89.6 80.2 35.5 81.5 Personapproach . . . . . .......... 89.5 98.6 97.2 91.8 82.9 42.9 81.1 31 Table 7. Average number of limiting conditions for persons with activity limitation, by degree of limitation, age, and type of question- naire obtained from the two half-samples used in the Health Interview Survey, July 1967-June 1968 Degree of limitation and age Condition approach Person approach All limited persons: T7-44 years . . oo oe ABBA YBAIS . . cove von» sms x x hx ews ees EE kes Fes Yu YEE EEE EE BE Years BND OVEY ; cu: so 5s si 5s 5 IRS SAME SF i FEE GME s Ambon Emm nm sms Limited in amount or kind of major activity: A BOBS wns oss 0s ims FERS FIDE SENT FE Vas § HERS Gomis mmm nmr bom LINDEBE 17 YBBIS . . ovo some vine smms mmmn vinnie smn owns main wan we we TTA years . . o.oo ABBA YOBE «wns cnn smn sips Lamps sms $86 sn my SRBC ins IE NL 2EST FH%0 BE years and OVer ; ov sn vs ME § HB 4 SR TE s CME FMEE SRY AMES CHEE EOE Sk a Limited in amount or kind of other activities: ABIES o.. oor nots v nore mans soning vmws Sans GREY RE SW RHEE © GEE TER UNDer TT Years . .....o: ems inms sms sss 656 5 403s rio ims Favs dues o 17D VCBYS msn 2 13 03 3 SH mE FEBS E55 SB HS Fowind Som a sommes nmmn oomsmmn xm BBB YOAIS . vo vo ons svnin vmmn mn in ae e eA eeae Ee we ee BS Years ANdiOVBY ». . wos » son su ws SUME SR WE SWE SAME § FONE EMME EWE CREE LE Average number of limiting conditions per limited person 1.2 1.4 32 Table 8. Percent distribution of conditions causing activity limitation by type of condition, according to degree of limitation and type of questionnaire obtained from the two half-samples used in the Health Interview Survey, July 1967-June 1968 Unable to carry on Limited in amount or Limited in amount or Conditions causing major activity kind of major activity kind of other activity activity dimination Condition Person Condition Person Condition Person approach approach approach approach approach approach Percent distribution Persons limited in activity . . . . ... 100.0 100.0 100.0 100.0 100.0 100.0 Asthma or hay fever . . . . . ........ 3.7 31 4.4 4.7 7.7 8.1 Other respiratory conditions . . . . . . . .. 9.6 7.7 6.2 4.3 5.5 4.9 Mental and nervous conditions . . . . . . . . 10.7 54 7.6 4.5 7.0 x Heart conditions . . . . . . . ........ 24.3 22.2 15.0 15.8 11.8 10.0 Hypertension (no heart conditions volved) . . . i sw sms ms me mo 5.1 4.0 59 5.3 4.8 4.8 Other circulatory conditions . . . . . . . .. 79 6.4 5.9 5.4 6.4 5.3 Digestive conditions. . . . . ........ 9.7 9.8 7.4 8.3 4.9 3.7 Arthritis and rheumatism . . . . .. ..... 16.0 16.2 16.0 15.0 124 114 Visual impairments . . . . . . ........ 11.3 7.6 4.8 3.4 32 4.2 Orthopedic impairments. . . . . . .. ... 14.8 10.9 21.2 14.7 21.5 18.4 Other impairments . . . . . . . .. ..... 22.6 15.6 10.0 8.2 6.6 7.5 All other conditions . . . . . ........ 46.8 45.2 26.7 321 25.6 32.6 Table 9. Incidence of acute conditions, by severity criteria, type of questionnaire, sex, and age obtained from the two used in the Health Interview Survey, July 1967-June 1968 half-samples Sex Age Severity criteria All and type of questionnaire persons Male Female Under 17 17-44 45-64 65 years years years years and over Number of conditions per 100 persons per year All acute conditions: Condition approach. . . .. ...... 194.5 188.3 200.2 272.7 180.4 133.9 95.2 Person approach . . . .. ........ 184.6 177.0 191.7 258.4 173.6 117.4 99.5 Medically attended only: Condition approach . . . . . ...... 56.9 59.4 54.7 80.1 48.3 42.8 36.0 Personapproach . . « cs oa ¢ 5 +» « » 34.0 34.2 33.8 43.1 32.1 26.2 24.3 Medically attended and activity restricting: Condition approach . . . . . ...... 70.2 66.7 73.5 98.8 68.0 45.3 29.4 Person approach . . . .......... 749 70.8 78.7 105.3 69.1 47.0 45.0 Activity restricting only: Condition approach . . . .. ...... 67.3 62.3 72.0 93.8 64.1 45.8 29.8 Person approach . . . .. .. ...... 75.8 721 79.2 110.0 724 44.2 30.1 33 Table 10. Incidence of medically attended acute conditions, with and without activity restriction, and physician visits, by age and type of questionnaire derived from the two half-samples used in the Health Interview Survey, July 1967-June 1968 Age Medically attended acute conditions Physician visits With no activity With activity Total a, fa restriction restriction Condition Person Condition Person Condition Person approach approach approach approach approach approach Person approach Condition approach All ages Under 6 years 6-16 years 17-24 years 25-44 years 45-64 years 65 years and over Number of conditions per 100 persons per year Visits per person per year 127.2 108.9 56.9 34.0 70.2 74.9 4.3 4.1 281.4 212.8 144.5 15.7 136.9 137.1 5.7 4.8 126.2 114.1 47.1 25.8 79.2 88.4 2.6 2.5 131.1 112.4 53.5 34.9 77.7 71.5 4.3 4.2 108.8 95.2 45.7 30.6 63.2 64.6 4.4 4.0 88.1 73.2 42.8 26.2 45.3 47.0 4.7 4.7 65.4 69.3 36.0 24.3 29.4 45.0 5.4 5.8 Table 11. Incidence of acute conditions, by condition category, type of questionnaire, sex, and age obtained from the two half-samples used in the Health Interview Survey, July 1967-June 1968 Sex Age Condition category and All type of questionnaire persons Male Female Under 17 1744 45-64 | 65 years years years years and over Number of conditions per 100 persons per year All acute conditions: Condition approach . . . ........ 194.5 188.3 200.2 272.7 180.4 133.9 95.2 Person approach . . ........... 184.6 177.0 191.7 258.4 173.6 117.4 99.5 Infective and parasitic conditions: Condition approach . . . ........ 23.7 24.5 229 441 16.8 10.4 * Personapproach . ............ 20.2 18.8 21.6 356.3 16.1 9.1 * Upper respiratory conditions: Condition approach. . . .. ...... 64.6 60.5 68.4 101.1 52.0 39.1 36.1 Personapproach . ............ 57.6 54.7 60.3 93.6 46.3 30.1 27.5 Influenza: Condition approach. . . ........ 40.6 37.9 43.2 47.9 43.2 33.4 204 Person approach . . . .......... 41.9 41.0 42.7 51.3 43.7 31.7 22.2 Other respiratory conditions: Condition approach. . . ........ 28 2.7 29 5.1 * * * Personapproach . . ........... 4.9 4.8 5.0 6.5 4.2 * * Digestive conditions: Condition approach. . . .... .... 7.6 7.6 7.6 8.3 8.1 6.8 * Person approach . . ........... 10.2 9.4 10.9 14.1 8.6 71 8.5 Injuries: Condition approach . . . ........ 32.6 384 27.1 39.2 31.6 29.8 17.7 Person approach . . . .......... 25.5 30.7 20.7 30.1 26.4 20.7 16.7 All other acute conditions: Condition approach . . . ........ 22.6 16.7 28.2 27.0 27.0 13.6 9.9 Person approach . . . .......... 24.3 17.6 30.6 27.6 28.4 156.6 16.56 35 Table 12. Incidence of common colds and wounds and lacerations, by severity criteria, type of questionnaire, sex, and age obtained from the two half-samples used in the Health Interview Survey, July 1967-June 1968 Sex Age Condition, severity criteria, All and type of questionnaire persons als Famale Under 17 17-44 45-64 65 years years years years and over Common cold Number of conditions per 100 persons per year Medically attended only: Condition approach . . . . . . . .. ... 16.3 18.5 17.0 24.4 11.6 12.5 13.1 Personapproach’ , « . o : + i 6 s 5 5 uw 5 6.5 8.3 6.7 8.6 4.6 6.9 * Medically attended and activity restricting: Condition approach . . . . . . . ... . . 10.3 8.6 12.0 17.3 7.8 5.0 * Personapproach . ............. 11.5 109 12.0 17.4 9.3 7.6 * Activity restricting only: Condition approach . . . . . . . .. ... 24.2 23.1 25.2 33.2 22.8 17.2 11.6 Person approach . . . . . .. .. ..... 24.1 23.2 249 37.8 21.3 12.1 9.7 Wounds and lacerations Medically attended only: Condition approach . . . . . . . ..... 6.6 8.9 45 10.6 53 4.2 * Person approach . . . . . .. ....... 3.8 5.1 2.6 6.1 3.4 * * Medically attended and activity restricting: Condition approach . . . . . . . .. ... 2.2 2.8 1.7 3.4 * ¥ % Person approach . . . . . ......... 2.4 2.8 24 39 2.1 * # Activity restricting only: Condition approach . . . . . . ...... 0.8 x * * * * x Person approach . . . . . . ........ * Ld * * a * 36 Table 13. Number of disability days per person per year, by type of disability, type of questionnaire, sex, and age obtained from the two half-samples used in the Health Interview Survey, July 1967-June 1968 Sex Age Type of disability and All type of questionnaire persons Male Female Under 17 17-44 45-64 65 years years years years and over Number of disability days per person per year Restricted activity: Condition approach . . . . . ....... 15.7 14.0 173 9.0 13.0 21.8 36.4 Person approach . . .. .......... 14.8 13.7 15.8 9.7 11.2 19.6 37.2 Bed disability: Conditionapproach . . . ......... 5.9 5.0 6.7 4.0 5.0 13 12.7 Personapproach . . ............ 6.0 5.3 6.7 4.3 4.7 74 15.6 Work-loss' : Conditionapproach . . . ......... 5.4 5.0 6.2 vid 5.0 6.1 5.6 Person approach . . ............ 49 4.8 5.3 Ao 4.4 5.6 7.0 School-loss?: Condition approach . . . . . ....... 4.1 39 43 4.1 Person approach . . .. .......... 49 4.6 5.2 49 ! Rates based on currently employed population 17 years and over. 2 Rates based on persons 6-16 years of age. Table 14. Number of person and condition days of disability and percent of person disability days associated with more than one con- dition, by type of disability day and type of questionnaire obtained from the two half-samples used in the Health Interview Survey, July 1967-June 1968 P f Days of disability due to: BjosAL Of La person disability Type of disability day Days of person 3 ; : Co ; days associated and type of questionnaire disability All Acute Chronic . bat tat LL. with more than conditions conditions conditions rue 1 condition Days per person per year Restricted activity: Condition approach . . . . . . 15.7 20.3 7.3 13.0 29.3 Person approach . . . . . .. 14.8 20.0 8.4 11.6 35.1 Bed disability: Condition approach . . . . . . 5.9 7.3 3a 4.2 23.7 Person approach . . . .. .. 6.0 8.0 37 4.3 33.3 Work-loss' : Condition approach . . . . . . 5.4 6.4 3.4 3.0 18.5 Person approach . . . .. .. 4.9 5.8 3.4 2.4 18.4 School-loss? : . Condition approach . . . . . . 4.1 4.4 3.6 0.8 73 Person approach . . . . . .. 49 5.1 4.4 0.7 4.1 1 Based on currently employed population 17 years and over. 2 Based on population 6-16 years of age. 37 Table 15. Number of disability days per person per year associated with acute conditions, by condition category, type of disability day, and type of questionnaire obtained from the two half-samples used in the Health Interview Survey, July 1967-June 1968 Condition category Restricted Bed hi . resi Lo Work-loss School-loss and type of questionnaire activity disability Number of disability days per person per year All acute conditions: Condition approach . . . . . ... .... .. 7.3 3.1 3.4 3.6 Person approach . . . . . ........... 8.4 3.7 3.4 4.4 Infective and parasitic conditions: Condition approach. . . . . . ... . .. .. 0.8 0.5 0.2 0.8 Person approach . . . ........... . . 0.8 0.4 0.3 0.8 Upper respiratory conditions: Condition approach . . . . . . . . .. .... 1.6 0.7 0.5 Td Person approach . . . . . ......... .. 1.7 0.7 0.5 1.4 Influenza: Condition approach. . . . . .. .. ... .. 1.6 0.9 0.8 1.0 Person approach . . . . . .. ......... 1.8 1.0 0.8 1.4 Other respiratory conditions: Condition approach. . . . . . . . .. .. .. 0.2 0. 0.1 0.7 Person approach . . . . ........... . 04 0.2 0.2 0.1 Digestive conditions: Condition approach. . . . . . . . . . .... 0.3 0.1 0.3 0.1 Person approach . . . . ..... ..... . . 0.5 0.2 0.2 0.2 Injuries: Condition approach . . . . . .. ....... 1.7 0.3 1] 0.2 Person approach . . . . . .. ........ . 1.7 0.5 1.0 0.3 All other acute conditions: Condition approach. . . . . . . . .. .. .. 1.0 0.4 03 0.3 Person approach . . . .... ......... 1.4 0.5 0.5 0.4 38 Table 16. Number of disability days associated with acute conditions, by condition category, type of disability day, and type of ques- tionnaire obtained from the two half-samples used in the Health Interview Survey, July 1967-June 1968 Condition category and Restricted Bed Work-los: School-l0ss type of questionnaire activity disability Number of disability days per condition All acute conditions: Condition approach . . . . . .. .. ..... 3.8 1.86 2.2 1.6 Person approach . . . . . . . . . ....... 4.5 2.0 2.3 1.9 Infective and parasitic conditions: Condition approach. . . . . .. ....... 3.5 1.9 1.5 241 Person approach . . . . . .. .. ....... 4.2 2.1 2.3 2.3 Upper respiratory conditions: Condition approach. . . . . . . ....... 24 1.0 1.2 1.6 Person approach . . . . . ........... 3.0 1.2 1.2 1.7 Influenza: Condition approach . . . . . ......... 4.1 2.3 23 20 Person approach . . . . . . . . . ....... 4.4 2.4 21 25 Other respiratory conditions: Condition approach. . . . . . . . . ..... 8.6 4.8 10.7 25 Person approach . . . . . . . .. ....... 8.4 4.5 5.3 2.7 Digestive conditions: Condition approach. . . . . . . ....... 4.5 1.9 4.0 1.3 Personapproach . . . . ............ 4.4 24 2.1 1.2 Fractures and dislocations: Condition approach . . . . . . . . ...... 10.3 1.9 5.1 0.8 Person approach . . . . ............ 111 3.1 5.0 1.9 Open wounds: Condition approach . . . . . .. ....... 25 0.3 1.2 0.2 Person approach . . . . . ........... 3.9 0.9 2.3 05 Contusions: Condition approach . . . . . .. . ...... 39 0.8 2.1 0.3 Personapproach . . . .. . a : 6 «oso 4 vs 49 1.7 25 0.6 All other injuries: Condition approach. . . . . . . ....... 4.2 1.2 3.5 1.3 Personapproach . . . .. .. .. cess... 5.3 2.2 49 0.7 All other acute conditions: Condition approach . . . . . .. ....... 4.2 1.7 19 1.3 Person approach . . . . . .. ......... 5.7 22 2.4 1.8 Table 17. Number of disability days associated with chronic conditions, by condition category, type of disability day, and type of questionnaire obtained from the two half-samples used in the Health Interview Survey, July 1967-June 1968 Condition category Restricted Bed . , Lo Co Work-loss School-loss and type of questionnaire activity disability Number of disability days per person per year All chronic conditions: Condition approach. . . . . ......... 130 4.2 3.0 0.8 Person approach . . . ............. 11.6 4.3 24 0.7 Asthma or hay fever: Condition approach . . . . . ......... 0.5 0.2 1 0.1 Person approach . . . .. ........... 0.3 0.1 0.1 0.1 Other respiratory conditions: Condition approach. . . . . ......... 1a 0.4 0.2 0.3 Person approach . . . . ............ 0.3 0.1 0.1 Mental and nervous conditions: Condition approach. . . .. ....... .. 09 0.3 0.2 0.0 Person approach . . . . .......... .. 0.5 0.2 0.1 0.0 Heart conditions: Condition approach . . . . . ......... 1.2 0.5 0.3 0.0 Person approach . . . .. ........... 14 0.6 0.2 0.0 Hypertension (no heart conditions involved): Condition approach. . . . . ......... 0.4 0.1 0.1 * Person approach . . . . ............ 0.3 0.1 0.1 * Other circulatory conditions: Condition approach. . . . . . ........ 0.6 0.2 0.2 * Person approach . . . . ............ 0.5 0.2 0.1 0.0 Digestive conditions: Condition approach. . . . . ......... 04 0.4 0.0 Person approach . . . ............. 1 0.5 0.4 0.1 Arthritis and rheumatism: Condition approach . . . . . . .. ...... 1 0.3 0.2 ¥ Person approach . . . .. ........... 1.1 04 0.1 0.0 Visual impairments: Condition approach. . . . . . .. .... .. 0.3 0.0 0.0 * Person approach . . .............. 0.2 0.1 0.0 * Orthopedic impairments: Condition approach. . . . . ......... 1.4 0.3 0.3 0.0 Person approach . . . ............. 09 0.2 0.2 0.0 Other impairments: Condition approach. . . . . ......... 0.7 0.3 0.1 0.0 Person approach . . . ............. 0.7 0.3 0.1 0.3 All other chronic conditions: Condition approach . . . . . ......... 3.6 1.3 0.9 0.3 Person approach . . . . ............ 35 1.56 0.8 0.2 40 Table 18. Number of disability days associated with selected chronic digestive conditions, by type of condition, type of disability day, and type of questionnaire obtained from the two half-samples used in the Health Interview Survey, July 1967-June 1968 Chronic digestive condition All chronic digestive conditions Ulcer of stomach or duodenum . . Hernia (abdominal cavity) . . . .. Chronic enteritis and ulcerative colitis Livertrouble . . . : « « « ww vw» » Other specified diseases of the liver Specified diseases of the gallbladder Gallbladder trouble . . . . . . . .. Stomach trouble . . . . . ..... Other diseases of the digestive system Restricted activity Bed disability Condition Person Condition Person approach approach approach approach Number of disability days per condition per year 17.1 11.4 5.7 4.1 19.7 17.3 6.2 5.8 14.9 1.7 4.9 45 241 11.0 4.7 4.2 17.8 28.7 * * 48.3 54.0 28.0 20.0 19.7 19.4 7.0 8.4 12.9 13.7 2.7 5.3 14.6 23.3 8.8 9.3 14.0 7.4 4.1 23 41 Table 19. Number of physician visits per person per year and percent of persons seen by.a physician in past year, by selected charac- teristics and type of questionnaire obtained from the two half-samples used in the Health Interview Survey, July 1967-June 1968 Physician visits Persons seen by a physician in past year Population characteristics Condition Person Condition Person approach approach approach approach Visits per person per year Percent Allpersons . . . . .............. 4.3 4.1 69.0 70.9 Sex Mas... ois cms BE Rem ne mas. 3.8 3.6 66.2 68.1 Female . . . .................. .. 4.8 4.6 71.6 73.6 Age UnderByears . . ................. B.7 4.8 80.9 82.6 6-16years . . . 26 25 62.4 63.7 17-24 years . . . . . . 4.3 4.2 73.2 75.2 25-44 years . . . 4.4 4.0 68.9 70.6 45-64 years . . . ..... 4,7 4.7 66.2 69.0 65 yearsandover . . . ............... 5.4 5.8 71.1 72.9 Color White . . .... 4.4 4.2 70.0 71.5 Allother . . . . 3.5 3.2 61.3 66.7 Education of family head UnderQvyears . . ................. 3.8 3.8 62.0 64.2 912vears Lo... 4.2 4.0 69.1 71.4 13vyearsandover . . . . .............. 5.0 4.7 76.6 78.6 Region NOFEREAST + . » «wv 8 5 fF 5 ve ms mn ve ww 4.2 4.0 721 72.5 North Central . . . . ............... 3.9 3.7 66.8 69.7 BOU 4 4 ¢ 5 6 iv vn sme mens wm Es EE 4.2 3.9 67.9 69.3 West . ................. 5.3 5.1 70.6 72.9 42 Table 20. Number of discharges from short-stay hospitals and average length of stay, by type of questionnaire, sex, and age obtained from the two half-samples used in the Health Interview Survey, July 1967-June 1968 Sex Age Hospital discharges, average length All of stay, and type of questionnaire persons Male Female Under 17 17-44 45-64 65 years years years years and over Number of discharges per 100 persons per year Hospital discharges: Condition approach . . . ......... 12.2 9.9 14.3 5.8 16.0 13.8 21.4 Person approach . . .. .......... 11.7 9.6 13.7 6.1 13.8 13.7 20.3 Average length of stay: Conditionapproach . . .......... 8.8 10.1 8.0 5.4 6.3 11.2 15.3 Parson approaCh « « ow « = s 5 +0 55 5 2 0 9.0 11.0 7.7 5.6 6.6 12.8 13.5 43 APPENDIX | TECHNICAL NOTES ON METHODS Background of This Report This report is one of a series of statistical reports prepared by the National Center for Health Statistics (NCHS). It is based on informa- tion collected in a continuing nationwide sample of households in the Health Interview Survey (HIS). The Health Interview Survey utilizes a ques- tionnaire which, in addition to personal and demographic characteristics, obtains information on illnesses, injuries, impairments, chronic con- ditions, and other health topics. As data relating to each of these various topics are tabulated and analyzed, separate reports are issued which cover one or more of the specific topics. The present report is based on data collected in household interviews during 1968. The population covered by the sample for the Health Interview Survey is the civilian, noninsti- tutional population of the United States living at the time of the interview. The sample does not include members of the Armed Forces or U.S. nationals living in foreign countries. It should also be noted that the estimates shown do not represent a complete measure of any given topic during the specified calendar period since data are not collected in the interview for persons who died during the reference period. For many types of statistics collected in the survey, the reference period covers the 2 weeks prior to the interview week. For such a short period, the contribution by decedents to a total inventory of conditions or services should be very small. However, the contribution by decedents during a long reference period (e.g., 1 year) might be sizable, especially for older persons. Statistical Design of the Health Interview Survey General Plan.—The sampling plan of the sur- vey follows a multistage probability design which permits a continuous sampling of the civilian, noninstitutional population of the United States. The sample is designed in such a way that the sample of households interviewed each week is representative of the target popula- tion and that weekly samples are additive over time. This feature of the design permits both continuous measurement of characteristics of samples, more detailed analysis of less common characteristics, and smaller categories of health related items. The continuous collection has administrative and operational advantages as well as technical assets since it permits fieldwork to be handled with an experienced, stable staff. The overall sample was designed in such a fashion that tabulations can be provided for each of the four major geographic regions and for urban and rural sectors of the United States. The first stage of the sample design consists of drawing a sample of 357 primary sampling units (PSU’s) from approximately 1,900 geograph- ically defined PSU’s. A PSU consists of a county, a small group of contiguous counties, or a standard metropolitan statistical area (SMSA). The PSU’s collectively cover the 50 States and the District of Columbia. With no loss in general understanding, the remaining stages can be combined and treated in this discussion as an ultimate stage. Within PSU’s, then, ultimate stage units called segments are defined in such a manner that each segment contains an expected six households. (Prior to July 1, 1968, the expected segment size was 45 nine households). Three general types of seg- ments are used: Area segments which are defined geographi- cally. List segments, using 1960 census registers as the frame. Permit segments, using updated lists of build- ing permits issued in sample PSU’s since 1960. Census address listings were used for all areas of the country where addresses were well defined and could be used to locate housing units. In general, the list frame included the larger urban areas of the United States from which about two-thirds of the HIS sample was selected. The total HIS sample of approximately 8,000 segments yields a probability sample of about 134,000 persons in 42,000 interviewed house- holds in a year. Descriptive material on data collection, field procedures, and questionnaire development in the HIS has been published? as well as a detailed description of the sample design! and a report on the estimation procedure and the method used to calculate sampling errors of estimates derived from the survey.” Collection of Data.—Field operations for the survey are performed by the U.S. Bureau of the Census under specifications established by the National Center for Health Statistics. In accord- ance with these specifications, the Bureau of the Census participates in survey planning, selects the sample, and conducts the field interviewing as an agent of NCHS. The data are coded, edited, and tabulated by NCHS. Estimating Procedures. —Since the design of the HIS is a complex multistage probability sample, it is necessary to usc complex proce- dures in the derivation of estimates. Four basic operations are involved. 1. Inflation by the reciprocal of the probabil- ity of selection.—The probability of selec- tion is the product of the probabilities of selection from each step of selection in the design: PSU, segment, and household. 2. Nonresponse adjustment.—The estimates are inflated by a multiplication factor that NOTE: The list of references follows the text. 46 has as its numerator the number of sample households in a given segment and as its denominator the number of households interviewed in that segment. 3. First-stage ratio adjustment. —Sampling the- ory indicates that the use of auxiliary information that is highly correlated with the variables being estimated improves the reliability of the estimates. To reduce the variability between PSU’s within a region, the estimates are ratio adjusted to 1960 populations within six color-residence classes. 4. Poststratification by age-sex-color.—The estimates are ratio adjusted within each of 60 age-sex-color cells to an independent estimate of the population of each cell for the survey period. These independent esti- mates are prepared by the Bureau of the Census. Both the first-stage and poststrati- fied ratio adjustments take the form of multiplication factors applied to the weight of cach elementary unit (person, house- hold, condition, and hospitalization). The effect of the ratio-estimating process is to make the sample more closely representative of the civilian, noninstitutional population by age, sex, color, and residence, thus reducing sampling variance. As noted, each week’s sample represents the population living during that week and charac- teristics of the population. Consolidation of samples over a time period, e.g, a calendar quarter, produces estimates of average charac- teristics of the U.S. population for the calendar quarter. Similarly, population data for a year are averages of the four quarterly figures. For prevalence statistics, such as number of persons with speech impairments or number of persons classified by time interval since last physician visit, figures are first calculated for cach calendar quarter by averaging estimates for all weeks of interviewing in the quarter. Preva- lence data for a year are then obtained by averaging the four quarterly figures. For other types of statistics—namely those measuring the number of occurrences during a specified time period—such as incidence of acute conditions, number of disability days, or num- ber of visits to a doctor or dentist, a similar computational procedure is used, but the statis- tics are interpreted differently. For these items, the questionnaire asks for the respondent’s experience over the 2 calendar weeks prior to the week of interview. In such instances the estimated quarterly total for the statistic is 6.5 times the average 2-week estimate produced by the 13 successive samples taken during the period. The annual total is the sum of the four quarters. Thus the experience of persons inter- viewed during a year—experience which actually occurred for each person in a 2-calendar-week interval prior to week of interview—is treated as though it measured the total of such experience during the year. Such interpretation leads to no significant bias. General Qualifications Nonresponse.—Data were adjusted for non- response by a procedure which imputes to persons in a household which was not inter- viewed the characteristics of persons in house- holds in the same segment which were inter- viewed. The total noninterview rate was about 5 percent—1 percent was refusal, and the remain- der was primarily due to the failure to find an eligible respondent at home after repeated calls. The interview process.—The statistics pre- sented in this report are based on replies obtained in interviews of persens in the sampled households. Each person 19 years of age and over present at the time of interview was interviewed individually. For children and for adults not present in the home at the time of the interview, the information was obtained from a related household member such as a spouse or the mother of a child. There are limitations to the accuracy of diagnostic and other information collected in household interviews. For diagnostic informa- tion, the household respondent can usually pass on to the interviewer only the information the physician has given to the family. For conditions not medically attended, diagnostic information is often no more than a description of symp- toms. However, other facts, such as the number of disability days caused by the condition, can be obtained more accurately from household members than from any ather source since only the persons concerned are in a position to report this information. Rounding of numbers.—The original tabula- tions on which the data in this report are based show all estimates to the nearest whole unit. All consolidations were made from the original tabulations using the estimates to the nearest unit. In the final published tables, the figures are rounded to the nearest thousand, although these are not necessarily accurate to that detail. Devised statistics, such as rates and percent distributions, are computed after the estimates on which these are based have been rounded to the nearest thousand. Population figures.—Some of the published tables include population figures for specified categories. Except for certain overall totals by age, sex, and color, which are adjusted to independent estimates, these figures are based on the sample of households in the HIS. These are given primarily to provide denominators for rate computation, and for this purpose are more appropriate for use with the accompanying measures of health characteristics than other population data that may be available. With the exception of the overall totals by age, sex, and color mentioned above, the population figures differ from corresponding figures (which are derived from different sources) published in reports of the Bureau of the Census. (For population data for general use, see the official estimates presented in Bureau of the Census reports in the P-20, P-25, and P-60 series.) Reliability of estimates.—Since the statistics presented in this report are based on a sample, they will differ somewhat from the figures that would have been obtained if a complete census had been taken using the same schedules, in- structions, and interviewing personnel and pro- cedures. As in any survey, the results are also subject to reporting and processing errors and errors due to nonresponse. To the extent possible, these types of errors were kept to a minimum by methods built into survey procedures. Although it is very difficult to measure the extent of bias in the Health Interview Survey, a number of studies have been conducted to study this problem. The results have been published in several reports.3.4,8-10 NOTE: The list of references follows the text. 47 The standard error is primarily a measure of sampling variability, that is, the variations that might occur by chance because only a sample of the population is surveyed. As calculated for this report, the standard error also reflects part of the variation which arises in the measurement process. It does not include estimates of any biases which might lie in the data. The chances arc about 68 out of 100 that an estimate from the sample would differ from a complete census by less than the standard error. The chances are about 95 out of 100 that the difference would be less than twice the standard error and about 99 out of 100 that it would be less than 2% times as large. The relative standard error of an estimate is obtained by dividing the standard error of the estimate by the estimate itself and is expressed as a percentage of the estimate. For this report, asterisks are shown for any cell with more than a 30-percent relative standard error. Included in this appendix are charts from which the relative standard errors can be determined for estimates shown in the report. In order to derive relative errors which would be applicable to a wide variety of health statistics and which could be prepared at a moderate cost, a number of approximations were required. As a result, the charts provide an estimate ol the approximate relative standard error rather than the precise error for any specific aggregate or percentage. Three classes of statistics for the health survey are identified for purposes of estimating vari- ances: Narrow range. —This class consists of (1) statistics which estimate a population at- tribute, c.g., the number of persons in a particular income group, and (2) statistics for which the measure for a single individual during the reference period used in data collection is usually either 0 or 1, or on occasion may take on the value 2, or very rarely 3. Medium range. —This class consists of other statistics for which the measure for a single individual during the reference period used in data collection will rarely lie outside the range 0tob. 48 Wide-range.—This class consists of statistics for which the measure for a single individual during the reference period used in data collection can range from 0 to a number in excess of 5, e.g., the number of days of bed disability. In addition to classifying variables according to whether they are narrow-, medium-, or wide-range, statistics in the survey are further defined as: Type A. Statistics on prevalence and inci- dence data for which the period of reference in the questionnaire is 12 months. Type B.—Incidence-type statistics for which the period of reference in the questionnaire is 2 weeks. Type C.—Statistics for which the reference period is 6 months. Only the charts on sampling error applicable to data contained in this report are presented. General rules for determining relative sam- pling errors.—The “‘guide” on page 50, together with the following rules, will enable the reader to determine approximate relative standard errors from the charts for estimates presented in this report. Rule 1. Estimates of aggregates: Approxi- mate relative standard errors for esti- mates of aggregates such as the number of persons with a given char- acteristic are obtained from appropri- ate curves on pages 51 and 52. The number of persons in the total U.S. population or in an age-sex-color class of the total population is ad- justed to official Bureau of the Cen- sus figures and is not subject to sampling error, Rule 2. Estimates of percentages in a percent distribution: Relative standard errors for percentages in a percentage distri- bution of a total are obtained from appropriate curves on page 53. For values which do not fall on one of the curves presented in the chart, Rule 3. Rule 4. visual interpolation will provide a satisfactory approximation. Estimates of rates where the numera- tor is a subclass of the denomina- tor: This rule applies for prevalence rates or where a unit of the numera- tor occurs, with few exceptions, only once in the year for any one unit in the denominator. For example, in computing the rate of visual impair- ments per 1,000 population, the numerator consisting of persons with the impairment is a subclass of the denominator which includes all per- sons in the population. Such rates if converted to rates per 100 may be treated as though they were percent- ages and the relative standard errors obtained from the chart P4AN-M. Rates per 1,000, or on any other base, must first be converted to rates per 100 then the percentage chart will provide the relative standard error per 100. Estimates of rates where the numera- tor is not a subclass of the denomina- tor: This rule applies where a unit of the numerator often occurs more than once for any one unit in the denominator. For example, in the computation of the number of per- sons injured per 100 currently em- ployed persons per year, it is possible that a person in the denominator could have sustained more than one of the injuries included in the numer- ator. Approximate relative standard errors for rates of this kind may be computed as follows: (a) Where the denominator is the total U.S. population or in- cludes all persons in one or more of the age, sex, and color categories of the total popula- tion, the relative error of the rate is equivalent to the rela- tive error of the numerator, which can be obtained directly from the appropriate chart. (b) In other cases the relative stand- ard error of the numerator and of the denominator can be obtained from the appropriate curve. Square each of these relative errors, add the result- ing values, and extract the square root of the sum. This procedure will result in an upper bound on the standard error and often will overstate the error. Rule 5. Estimates of difference between two statistics (mean, rate, total, etc.): The standard error of a differ- ence is approximately the square root of the sum of the square of each standard error considered separately. A formula for the standard error of a differenced = X| - X,, is O04 = X Va, + X, Vig) where X, is the estimate for class 1, X, is the estimate for class 2, and V,, and V, are the relative errors of X, and X,, respectively. This formula will represent the actual standard error quite accurately for the difference between separate and uncorrelated characteristics although it is only a rough approximation in most other cases. The relative stand- ard error of each estimate involved in such a difference can be deter- mined by one of the four rules above, whichever is appropriate. 49 GUIDE TO USE OF RELATIVE STANDARD ERROR CHARTS The code shown below identifies the appro- priate curve to be used in estimating the relative standard error of the statistic described. The four components of each code describe the statistic as follows: (1) A=aggregate, P=percent- age; (2) the number of calendar quarters of data collection; (3) the type of the statistic as desc- ribed on page 48 and (4) the range of the statistic as described on page 48. Use: Statistic Rule Code on page Number of Persons in a population group... ......... nu... 1 A4AN 51 Chronic conditions by specified type .................. 1 A4AN 51 Chronic conditions causing activity limitation eee 1 A4AN 51 Chronic conditions per 1,000 persons... ..........uuuuuu... 3 P4AN-M 53 a Numer: A4BN 51 Acute conditions per 100 persons Per year . .. ...........o.o.... 4(b) Denom: A4AN Co Numer: A4BW Disability days per person per year ..............ouuuueenn. 4(b) Denom: A4AN 51 Numer: A4BM ici iSItS Per Person Per Year . ......... u're. 51 Physician visits per person per year ....... 4(b) Denom: A4AN Bos discharges per 100 persons per ” Naren: A4CN a VHBE 3 v5.0 7.5 7 3 ii hoi fon vn moma x vw wren 4 Ae es wi Denom: A4AN Percentage distribution of Persons with activity limitation ...................... 2 P4AN-M 53 Persons by interval since last physician VISE ee 2 P4AN-M 53 NOTE: Estimates of the populations derived from the samples interviewed by the condition and person approaches are shown in table 1. These figures are appropriate for the derivation of aggregates of health items to be used in the approximation of relative sampling errors. 50 Relative standard error (%) Relative standard errors for aggregates based on four quarters of data collection for data of all types and ranges Code: A4BW - Type B data, Wide range variable A4BM - Type B data, Medium range variable A4BN - Type B data, Narrow range variable A4AW - Type A data, Wide range variable A4AM - Type A data, Medium range variable A4AN - Type A data, Narrow range variable 1, 10,000 Size of estimate (in thousands) Example of use of chart: An aggregate of 2,000,000 (on scale at bottom of chart) for a Narrow range Type A statistic (code: A4AN) has a relative standard error of 3.6 percent, (read from scale at left side of chart), or a standard error of 72,000 (3.6 percent of 2,000,000). For a Wide range Type B statistic (code: A4BW), an aggregate of 6,000,000 has a relative error of 16.0 percent or a standard error of 960,000 (16 pexcent of 6,000,000). 100,000 10 vd. 51 Relative standard error (%) 52 Relative standard errors for aggregates based on four quarters of data collection for type C, Narrow range, and type C, Wide range data 100 100 60 50 40 30 20 1 8 3 4 5 6 7 8 00 2 3 4 5 6 7 1,000 2 1 ,000 . 2 3 4 5 Size of estimate (in thousands) 100,00 Example of use of chart: An aggregate of 1,000,000 (on scale at bottom of chart) for a Narrow range type C statistic (code: A4CN) has a relative standard error of 7.1 percent, read from scale at left side of chart, or a standard error of 71,000 (7.1 percent of 1,000,000). Relative standard error (%) Relative standard errors for percentages based on four quarters of data collection for type A data, Narrow and Medium range (Base of percentage shown on curves in millions) 100 30 20 a NN ® © o Code: P4AN-M Type A data Narrow and medium range variables 9 e8 LE i ®6 e5 3 2 Estimated percentage Example of use of chart: An estimate of 20 percent (on scale at bottom of chart) based on an estimate of 10,000,000 has a relative standard error of 3.2 percent (read fram the scale at the left side of the chart),the point at which the curve for a base of 10,000,000 intersects the vertical line for 20 percent. The standard error in percentage points is equal to 20 percent X 3.2 percent or 0.64 percentage points. 53 APPENDIX II DEFINITIONS OF CERTAIN TERMS USED IN THIS REPORT Terms Relating to Conditions Condition.—A morbidity condition, or simply a condition, is any entry on the questionnaire which describes a departure from a state of physical or mental well-being. It results from a positive response to one of a series of “medical disability impact” or “illness-recall” questions. In the coding and tabulating process conditions arc selected or classified according to a number of different criteria such as whether they were medically attended, whether they resulted in disability, or whether they were acute or chronic; or according to the type of disease, injury, impairment, or symptom reported. For the purposes of cach published report or set of tables, only those conditions recorded on the questionnaire which satisfy certain stated cri- teria are included. Conditions are classified by type according to the International Classification of Diseases, 1955 Revision, with certain modifications adopted to make the code more suitable for a household interview survey. Acute condition.—An acute condition is de- fined as a condition which has lasted less than 3 months and which has involved either medical attention or restricted activity. Because of the procedures used to estimate incidence, the acute conditions included in this report are the condi- tions which had their onset during the 2 weeks prior to the interview week and which involved either medical attention or restricted activity during the 2-week period. However, it excludes the following conditions which are always classi- fied as chronic even though the onset occurred within 3 months prior to week of interview: Tuberculosis Chronic bronchitis Asthma Hay fever chrenie-stomaelhr trouble Kidney stones or chronic kidney Repeated attacks of sinus trouble Rheumatic fever Hardening of the arteries trouble High blood pressure Arthritis or rheuma- Heart trouble tism Stroke Mental illness Trouble with varicose Diabetes veins Thyroid trouble or Hemorrhoids or piles goiter Tumor, cyst, or growth Any allergy Chronic gallbladder or Epilepsy liver trouble Chronic nervous Deafness or serious trouble trouble with hearing Cancer Chronic skin trouble Hernia or rupture Prostate trouble Paralysis of any kind Repeated trouble with back or spine Club foot Permanent stiffness or deformity of the foot, leg, fingers, arm, or back Condition present since birth Serious trouble with seeing, even when wearing glasses Cleft palate Any speech defect Missing fingers, hand, or arm—toes, foot, or leg Palsy Stomach ulcer Any other chronic stomach trouble Kidney stones or Acute condition groups.—In this report all tables which have data classified by type of condition employ a 5-category regrouping plus several selected subgroups. The International Classification code numbers included in each category are as follows: Condition Groups Infective and parasitic diseases . . . . . . Common childhood diseases . . . . . . . Theirs, NOS. os 5s ssw 5 0% w am Other infective and parasitic diseases . . . Respiratory conditions . . . . . ... .. Upper respiratory conditions . . . . . . . Commoncold . . ............ Other acute upper respiratory conttioOns . , ou 5 5 sw 5 in os & » « wo Influenza . : » + ov 5s ws 0 sw 2 8 08 ws Influenza with digestive manifestations . Other influenza . . . . ......... Other respiratory conditions . . . . . . . Pneumonia . oo + vw 5 ¢ a0 ws = 2 ow x Bronghltis + « av soo #2 w 2 9 + ® & = & Other acute respiratory conditions Digestive system conditions . . . . . . .. Dental conditions . . . ......... Functional and symptomatic upper gastrointestinal disorders, N.E.C Other digestive system conditions INJUPIBS & . 5 5 v5 2 5 wo 8 8 8m sw Pew Fractures, dislocations, sprains, and Sang . . « «5 + v3 ws wa www ow Fractures and dislocations . . . . . . . Sprainsand strains . . . . . . . .... Open wounds and lacerations . . . . . . Contusions and superficial injuries . . . . Other current injuries . . . . . . . . .. All other acute conditions . . . . . . . .. Diseases of theear . . . ......... Headaches ., , vs +s 54 » 5 ws os vw 5 = Genitourinary disorders. . . . . . . .. Deliveries and disorders of pregnancy and the puerperium . . . . . . ..... Diseases of theskin . . . . ....... Diseases of the musculoskeletal system All other acute conditions . . . . . . .. Chronic condition.—A condition is con- sidered to be chronic if (1) the condition is described by the respondent as having been first noticed more than 3 months before the week of the interview or (2) it is one of the conditions always classified as chronic regardless of the onset. (See list under the definition of an acute condition.) International Classification Code Number (excluding chronic inclusions) 021-138 056, 085-087, 089 021-055, 057-084, 088, 090-138 470-501, 511, 517-525, 527, 783 470-475, 511, 517 470 471-475, 511,517 480-483 482 480, 481, 483 490-501, 518-525, 527, 783 490-493 500, 501 518-525, 5627, 763 530-539, 543-553, 570, 571, 573- 587, 784, 785 530-535 544, 784 536-539, 543, 545-563, 670, 571, 573-5687, 785 N800-N885, N990-N995, N900-N994, N996-N999 N800-N848 N800-N839 N840-N848 N870-N885, N890-N895, N900-N908 N910-N929 N850-N869, N930-N994, N996-N999 All other acute code numbers 390-396, 781.3 791 590-637, 786, 789 640-689 690-716 726-743, 787 Other acute code numbers Impairments.—Impairments are chronic or permanent defects, usually static in nature, resulting from disease, injury, or congenital malformation. They represent decrease or loss of ability to perform various functions, particularly those of the musculoskeletal system and the sense organs. All impairments are classified by means of a special supplementary code for 55 impairments. Hence code numbers for impair- ments in the International Classification of Diseases are not used. In the Supplementary Code, impairments are grouped according to type of functional impairment and etiology. Prevalence of Conditions. —In general, preva- lence of conditions is the estimated number of conditions of a specified type existing at a specified time or the average number existing during a specified interval of time. The preva- lence of chronic conditions is defined as the number of chronic cases reported to be present or assumed to be present at the time of the interview; those assumed to be present at the time of the interview are cases described by the respondent in terms of one of the chronic diseases on the Checklist of Chronic Conditions and reported to have been present at some time during the 12-month period prior to the inter- view. Onset of condition.—A condition is consid- ered to have had its onset when it was first noticed. This could be the time the person first felt sick or became injured, or it could be the time when the person or his family was first told by a physician that he had a condition of which he was previously unaware. Incidence of conditions. —The incidence of conditions is the estimated number of condi- tions having their onset in a specified time period. As previously mentioned, minor acute conditions involving neither restricted activity nor medical attention are excluded from the statistics. The incidence data shown in some reports are further limited to various subclasses of conditions, such as “incidence of conditions involving bed disability.” Activity-restricting condition.—An activity- restricting condition is one which had its onset in the past 2 weeks and which caused at least 1 day of restricted activity during the 2 calendar weeks before the interview week. (See definition of “Restricted-activity day.”) Bed-disabling condition.—A condition with onset in the past 2 weeks involving at least 1 day of bed disability is called a bed-disabling condi- tion. (See definition of “Bed-disability day.”) Medically attended condition.—A condition with onset in the past 2 weeks is considered medically attended if a physician has been consulted about it either at its onset or at any time thereafter. Medical attention includes con- sultation cither in person or by telephone for treatment or advice. Advice from the physician transmitted to the patient through the nurse is counted as well as visits to physicians in clinics or hospitals. If during the course of a single visit the physician is consulted about more than one condition for each of several patients, each condition of each patient is counted as medi- cally attended. Discussions of a child’s condition by the physician and a responsible member of the household are considered as medical attention even if the child was not seen at that time. For the purpose of this definition, the term “physician” includes doctors of medicine and osteopathic physicians. Terms Relating to Disability Disability. —Disability is the general term used to describe any temporary or long-term reduc- tion of a person’s activity as a result of an acute or chronic condition. Chronic activity [mitation.—Persons with chronic conditions are classified into four cate- gories according to the extent to which their activities are limited at present as a result of these conditions. Since the usual activities of preschool children, school-age children, house- wives, and workers and other persons differ, a different set of criteria is used for each group. There is a general similarity between them, however, as will be seen in the descriptions of the four categories below. L. Persons unable to carry on major activity for their group (major activity refers to ability to work, keep house, or go to school) Preschool children: inability to take part in ordinary play with other children. School-age children: inability to go to school. Housewives: igability to do any housework. Workers and all other persons: inability to work at a job or business. 2. Persons limited in the amount or kind of major activity performed (major activity refers to ability to work, keep house, or go to school) Preschool children: School-age children: Housewives: Workers and all other persons: limited in the amount or kind of play with other children, e.g., need special rest peri- ods, cannot play strenuous games, can- not play for long peri- ods at a time. limited to certain types of schools or in school attendance, e.g., need special schools or special teaching, cannot go to school full time or for long periods of time. limited in amount or kind of housework, i.e., cannot lift chil- dren, wash or iron, or do housework for long periods at a time. limited in amount or kind of work, e.g., need special working aids or special rest pe- riods at work, cannot work full time or for long periods at a time, cannot do stren- uous work. 3. Persons not limited in major activity but otherwise limited (major activity refers to ability to work, keep house, or go to school) Preschool children: School-age children: not classified in this category. not limited in going to school but limited in participation in athletics or other ex- tracurricular activi- ties. not limited in house- work but limited in other activities such as church, clubs, hob- bies, civic projects, or shopping. Housewives: Workers and all other persons: not limited in regular work activities but limited in other activ- ities such as church, clubs, hobbies, civic projects, sports, or games. 4. Persons not limited in activities. Includes persons with chronic conditions whose activities are not limited in any of the ways described above. Disability day.—Short-term disability days are classified according to whether they are days of restricted activity, bed-days, or work-loss days. All days of bed disability are, by definition, days of restricted activity. The converse form of this statement is, of course, not true. Days lost from work are also days of restricted activity for the working population. Hence “days of restricted activity” is the most inclusive term used in describing disability days. Condition-day.—Condition-days of restricted activity, bed disability, and work loss are days of the various forms of disability associated with any one condition. Since any particular day of disability may be associated with more than one condition, the sum of days for conditions may add to more than the total number of person- days. Restricted-activity day.—A day of restricted activity is one on which a person substantially reduces the amount of activity normal for that day because of a specific illness or injury. The type of reduction varies with the age and occupation of the individual as well as with the day of the week or season of the year. Re- stricted activity covers the range from substan- 57 tial reduction to complete inactivity for the entire day. Bed-disability day.—A day of bed disability is one on which a person stays in bed for all or most of the day because of a specific illness or injury. All or most of the day is defined as more than half of the daylight hours. All hospital days for inpatients are considered to be days of bed disability even if the patient was not actually in bed at the hospital. Work-loss day.—A day is counted as lost from work if the person would have been going to work at a job or business that day but instead lost the entire workday because of an illness or an injury. If the person’s regular workday is less than a whole day and the entire workday was lost, it would be counted as a whole workday lost. Work-loss days are determined only for currently employed persons 17 years of age and over. (See “Currently employed persons” under “Demographic Terms.”) Person-days.—Person-days of restricted activ- ity, bed disability, and work loss are days of the various forms of disability experienced by any one person. The sum of days for all persons in a group represent an unduplicated count of all days of disability for the group. Terms Relating to Persons Injured Injury condition.—An injury condition, or simply an injury, is a condition of the type that is classified according to the nature of injury code numbers (N800-N999) in the International Classification of Diseases. In addition to frac- tures, lacerations, contusions, burns, and so forth, which are commonly thought of as injuries, this group of codes includes effects of exposure, such as sunburn; adverse reactions to immunization and other medical procedures; and poisonings. Unless otherwise specified, the term injury is used to cover all of these. Since a person may sustain more than one injury in a single accident, e.g., a broken leg and laceration of the scalp, the number of injury conditions may exceed the number of persons injured. Statistics of acute injury conditions include only those injuries which involved at least 1 full day of restricted activity or medical attendance. 58 Person injured. —A person injured is one who has sustained one or more injuries in an accident or in some type of nonaccidental violence. (See definition of “Injury condition” above.) Each time a person is involved in an accident or in nonaccidental violence causing injury that re- sults in at least 1 full day of restricted activity or medical attention, he is included in the statistics as a separate “person injured”; hence one person may be included more than once. The number of persons injured is not equiva- lent to the number of “accidents” for several reasons: (1) the term “accident” as commonly used may not involve injury at all; (2) more than one injured person may be involved in a single accident so that the number of accidents result- ing in injury would be less than the number of persons injured in accidents; and (3) the term “accident” ordinarily implies an accidental ori- gin, whereas “persons injured” as used in the Health Interview Survey includes persons whose injury resulted from certain nonaccidental vio- lence. The number of persons injured in a specified time interval is always equal to or less than the incidence of injury conditions since one person may incur more than one injury in a single accident. Terms Relating to Class of Accident Class of accident. — Injuries, injured persons, and resulting days of disability may be grouped according to class of accident. This is a broad classification of the types of event which re- sulted in personal injuries. Most of these events are accidents in the usual sense of the word, but some are other kinds of mishap, such as overex- posure to the sun or adverse reactions to medical procedures, and others are nonaccidental vio- lence, such as attempted suicide. The classes of accidents are: (1) moving motor vehicle acci- dents, (2) accidents occurring while at work, (3) home accidents, and (4) other accidents. These categories are not mutually exclusive. For example, a person may be injured in a moving motor-vehicle accident which occurred while the person was at home or at work. The accident class “motor vehicle” includes ‘“home-motor vehicle” and “while at work-motor vehicle.” Similarly, the classes while at work and home include duplicated counts, e.g., motor vehicle- while at work is included under “while at work.” Motor vehicle accident. —The class of accident is “motor vehicle” if a motor vehicle was involved in any way. Thus it is not restricted to moving motor vehicles or to persons riding in motor vehicles. A motor vehicle is any mechani- cally or electrically powered device, not oper- ated on rails, upon which or by which any person or property may be transported or drawn upon a land highway. Any object, such as a trailer, coaster, sled, or wagon, being towed by a motor vehicle is considered a part of the motor vehicle. Devices used solely for moving persons or materials within the confines of a. building and its premises are not counted as motor vehicles. Moving motor vehicle.—The accident is classi- fied as “moving motor vehicle” if at least one of the motor vehicles involved in the accident was moving at the time of the accident. This category is subdivided into “traffic” and “non- traffic.” Moving motor vehicle traffic accident.—The accident is classified as “traffic” if it occurred on a public highway. It is considered to have occurred on the highway if it occurred wholly on the highway, if it originated on the highway, if it terminated on the highway, or if it involved a vehicle partially on the highway. A public highway is the entire width between boundary lines of every way or place of which any part is open to the use of the public for the purposes of vehicular traffic as a matter of right or custom. Moving motor vehicle nontraffic accident.— The accident is classified as ‘“‘nontraffic” if it occurred entirely in any place other than a public highway. Nonmoving motor vehicle.—The accident is classified as ‘“nonmoving motor vehicle” if the motor vehicle was not moving at the time of the accident. Accident while at work.—The class of acci- dent is “while at work” if the injured person was 17 years of age or over and was at work at a job or a business at the time the accident happened. Home accident.—The class of accident is “home” if the injury occurred either inside or outside the house. “Outside the house” refers to the yard, buildings, and sidewalks on the prop- erty. “Home” includes not only the person’s own house but also any other house in which he might have been when he was injured. Other.—The class of accident is “other” if the occurrence of injury cannot be classified in one or more of the first three class-of-accident categories (i.e., moving motor vehicle, while at work, or home). This category therefore in- cludes persons injured in public places (e.g., tripping and falling in a store or on a public sidewalk) and also nonaccidental injuries such as homicidal and suicidal attempts. The survey does not cover the military population, but current disability of various types resulting from prior injury occurring while the person was in the Armed Forces is covered and is included in this class. The class also includes mishaps for which the class of accident could not be ascertained. Terms Relating to Hospitalization Hospital discharge.—A hospital discharge is the completion of any continuous period of stay of one or more nights in a hospital as an inpatient except the period of stay of a well newborn infant. A hospital discharge is recorded whenever a present member of the household is reported to have been discharged from a hospital in the 12-month period prior to the interview week. (Estimates were based on discharges which occurred during the 6-month period prior to the interview.) Hospital episode.—A hospital episode is any continuous period of stay of one night or more in a hospital as an inpatient except the period of stay of a well newborn infant. A hospital episode is recorded for a family member when- ever any part of his hospital stay is included in the 12-month period prior to the interview week. Hospital.—For this survey a hospital is defined as any institution meeting one of the following criteria: (1) named in the listing of hospitals in the current Guide Issue of Hospitals, the Journal of the American Hospital Association; (2) named in the listing of hospitals in the Directories of the American Osteopathic Hospi- tal Association; or (3) named in the annual inventory of hospitals and related facilities submitted by the States to the Division of 59 Hospital and Medical Facilities of the U.S. Public Health Service in conjunction with the Hill-Burton program. Hospital ownership.—Hospital ownership is a classification of hospitals according to the type of organization that controls and operates the hospital. The category to which an individual hospital is assigned and the definition of these categories follows the usage of the American Hospital Association. Type of hospital service.—Type of hospital service is a classification of hospitals according to the predominant type of cases for which they provide care. The category to which an individ- ual hospital is assigned and the definition of these categories follows the usage of the Ameri- can Hospital Association. Short-stay hospitals. —A short-stay hospital is one for which the type of service provided by the hospital is general; maternity; eye, ear, nose, and throat; children’s; or osteopathic; or it may be the hospital department of an institution. Hospital day.—A hospital day is a day on which a person is confined to a hospital. The day is counted as a hospital day only if the patient stays overnight. Thus a patient who enters the hospital on Monday afternoon and leaves Wednesday noon is considered to have had 2 hospital days. Hospital days during the yecar.—The number of hospital days during the year is the total number for all hospital episodes in the 12-month period prior to the interview week. For the purposes of this estimate, episodes overlapping the beginning or end of the 12-month period are subdivided so that only those days falling within the period are included. Terms Relating to Dental Visits Dental visit. —A dental visit is defined as any visit to a dentist’s office for treatment or advice, including services by a technician or hygienist acting under a dentist’s supervision. Interval since last dental visit.—The interval since the last dental visit is the length of time prior to the week of interview since a dentist or dental hygienist was last visited for treatment or advice of any type. Terms Relating to Physician Visits Physician visit. —A physician visit is defined as consultation with a physician, in person or by 60 telephone, for examination, diagnosis, treat- ment, or advice. The visit is considered to be a physician visit if the service is provided directly by the physician or by a nurse or other person acting under a physician’s supervision. For the purpose of this definition, “physician” includes doctors of medicine and osteopathic physicians. The term “doctor” is used in the interview rather than “physician” because of the need to keep to popular usage. However, the concept toward which all instructions are directed is that which is described here. Physician visits for services provided on a mass basis are not included in the tabulations. A service received on a mass basis is defined as any service involving only a single test (e.g., test for diabetes) or a single procedure (e.g., smallpox vaccination) when this single service was admin- istered identically to all persons who were at the place for this purpose. Hence obtaining a chest X-ray in a tuberculosis chest X-ray trailer is not included as a physician visit. However, a special chest X-ray given in a physician’s office or in an outpatient clinic is considered a physician visit. Physician visits to hospital inpatients are not included. If a physician is called to a house to see more than one person, the call is considered a separate physician visit for cach person about whom the physician was consulted. A physician visit is associated with the person about whom the advice was sought, even if that person did not actually see or consult the physician. For example, if a mother consults a physician about one of her children, the physi- cian visit is ascribed to the child. Interval since last physician visit.—The inter- val since the last physician visit is the length of time prior to the week of interview since a physician was last consulted in person or by telephone for treatment or advice of any type whatever. A physician visit to a hospital in- patient may be counted as the last time a physician was scen. Demographic Terms Age.—The age recorded for each person is the age at last birthday. Age is recorded in single years and grouped in a variety of distributions depending on the purpose of the table. Currently employed persons.—Currently em- ployed persons are all persons 17 years of age and over who reported that at any time during the 2-week period covered by the interview they either worked at or had a job or business. Current employment includes paid work as an employee of someone else; self-employment in business, farming, or professional practice; and unpaid work in a family business or farm. Persons who were temporarily absent from a job or business because of a temporary illness, vacation, strike, or bad weather are considered as currently employed if they expected to work as soon as the particular event causing the absence no longer existed. Free-lance workers are considered as currently employed if they had a definite arrangement with one employer or more to work for pay according to a weekly or monthly schedule, either full time or part time. Excluded from the currently employed popu- lation are persons who have no definite employ- ment schedule but work only when their services are needed. Also excluded from the currently employed population are (1) persons receiving revenue from an enterprise but not participating in its operation, (2) persons doing housework or charity work for which they receive no pay, (8) seasonal workers during the portion of the year they were not working, and (4) persons who were not working, even though having a job or business, but were on layoff or looking for work. The number of currently employed persons estimated from the Health Interview Survey (HIS) will differ from the estimates prepared from the Current Population Survey (CPS) of the U.S. Bureau of the Census for several reasons. In addition to sampling variability they include three primary conceptual differences, namely: (1) HIS estimates are for persons 17 years of age and over; CPS estimates are for persons 16 years of age and over. (2) HIS uses a 2-week reference period, while CPS uses a 1-week reference period. (3) HIS is a continuing survey with separate samples taken weekly; CPS is a monthly sample taken for the survey week which includes the 12th of the month. 61 APPENDIX Ill QUESTIONNAIRE—CONDITION APPROACH NOTICE - All information which would permit identification of the individual will beheld in strict confidence, will be used only by persons engaged inand for the purposes of the survey, and will not be disclosed or released to others for ony purposes. Form NHS-HIS-1 (FY67) U.S. DEPARTMENT OF COMMERCE —BUREAU OF THE CENSUS 1. REVISED 9-30-66 Budget Bureau No. 68-R1600 ACTING AS COLLECTING AGENT FOR THE U.S. PUBLIC HEALTH SERVICE Approval Expires 3-31-68 U.S. HEALTH INTERVIEW SURVEY 23.3:1 Book. of Books 2a. STREET ADDRESS House No., Street, Apt. No. or other ident. ! FOR AREA 2b. MAILING ADDRESS If different from 2a [_] Some os 20 ' SEGMENTS, ENTER: I . I Sheet City State Zip Code I No. : City ) [Store [Zip Cade Line 2c. SPECIAL DWELLING PLACE - Nome ond Sample Number ! | | No. Name "Sample No. 3. [J Ask — WHEN WAS THIS STRUCTURE ORIGINALLY BUILT? Og ig 1 [Before 4-160 Continue interview 42. SAMPLE Circle One B38 B-39 B40 BAI B42 B43 i fom3 | [J After 4.1.60-Go to Q. J0c, ask if required, and end interview. | qb. PSU rma 3 COMPLETE ITEMS 10-16 AT THE END OF THE INTERVIEW Rrite in and mark [1] i 10. Ta.[J4sk: ARE THERE ANY OCCUPIED OR VACANT LIVING + [3 Do Not! QUARTERS BESIDES YOUR OWN IN THIS BUILDING? 5a. SEGMENT NUMBER i ak : [3 Yes-Fill Table x Ono Write in and mark n tem 10-, Lt EE a ee at ——— pL ! GoTo |b. [J4sk: ARE THERE ANY OCCUPIED OR VACANT LIVING b. SEG. TYPE Circle—=A B P LSDP ? QUARTERS BESIDES YOUR OWN ON THIS FLOOR 5 SERIAL NUMBER i [3 Yes-Fill Table X Ono Write in and mark [1] “¢.[J4sk: IS THERE ANY OTHER BUILDING ON THIS PROPERTY FOR i PEOPLE TO LIVE IN - EITHER OCCUPIED OR VACANT? | 7. SPECIAL DWELLING PLACE -Type and Code Mark type code nu O Yes-Fill Table x ONe Ture Cole | renl | Ol Rwol=tskiems 11end iz [ Alloter (= Goro 13 Forney ow 11. DO YOU OWN OR RENT THIS PLACE? Type A If other” is marked . [J Own — tsk 12a [OJ Rent —dsk 126 [J Rent Free — sk 12a y isl hs Jovtudie _— we vs we oa om |p [eT space. ype 12a. DOES THIS PLACE HAVE 10 OR MORE ACRES? ............ | [J Yes: dsk 12¢ ” nS ene b. DOES THE PLACE YOU RENT HAVE 10 OR MORE ACRES”. .\ [J No-4sk 12d Om Ms ESS 4140 OTH c. DURING THE PAST 12 MONTHS DID SALES OF CROPS. | Tyre C LIVESTOCK, AND OTHER FARM PRODUCTS FROM @ 3. TYPE OF LIVING QUARTERS ccring Une Cer Unit THIS PLACE AMOUNT TO $50 OR MORE? .. ‘Orne @ PEGF LIVING fan Chen d. DURING THE PAST 12 MONTHS DID SALES OF CROPS, (7 ves (3) nu LIVESTOCK, AND OTHER FARM PRODUCTS FROM \ 12. LAND USAGE . THIS PLACE AMOUNT TO $250 OR MORE? . Ono) Seskicode: fiom Irom Lod Sergrt20 I [13 HOW MANY ROOMS ARE IN THIS — = (CIT) Tremere COUNT THE KITCHEN BUT NOT THE BATHROOM. Tete! Roms 3 1 [14 HOW MANY BEDROOMS ARE IN THIS — — ((A/7)7 tei etmet If *None® describe in footnotes n'a No. of Bedrooms nm 15. WHAT IS THE TELEPHONE NUMBER HERE? irene = Ne On 16. INTERVIEWER CHECK ITEM: Check questions 22a-22d & 23c on pages 4 & § [) Yes-Fill Home Care Supplement Is a llome Care Supplement required? [J No - Leave Thank You Letter and depart 17. RECORD OF CALLS AT HOUSEHOLD ITEMS 18-23 ARE TO BE FILLED AFTER THE INTERVIEW nm DATE AND | 1. 18. NUMBER OF CALLS AT HOUSEHOLD 2 frem TIME OF 19. DATE OF COMPLETION nO Aw O Wy O Gr MW | CALL Tire Enter from item 17 Month Feb May Ap O Nov O Sept ( Nec LENGTH OF . . oo INTERVIEW Minutes Day a 200.NAME OF OBSERVER /f 206 marked Yes" 1200. WAS THIS INTERVIEW OBSERVED? Yo 210. INTERVIEWER NAME Rrite-in T21b. INTERVIEWER NUMBER FOOTNOTES 22. IDENTIFICATION CODE NO. =m Wark from tab of Segment folder 23. REGIONAL OF FICE NUMBER WASHINGTON USE Book Number See item 1 m i Total Number of Conditions this HH. . Total Number of Hospitalizations this H.H. i Total Number of Doctor Visits this H.H. un 4 i Totol Number of Persons this H.H. ! Total Persons Requiring Home Care o i this Household ‘| a [eo | 64 First Nome on First Nome 02 la. WHAT IS THE NAME OF THE HEAD OF THIS HOUSEHOLD? b. WHAT ARE THE NAMES OF ALL OTHER PERSONS WHO LIVE HERE? Lis all Yes No CI HAVE LISTED read names IS THERE ANYONE ELSE STAYING HERE Now? (1) Cf | ast Name Last Name d. HAVE | MISSED ANYONE WHO USUALLY LIVES wore Bane ul | HERE BUT IS NOW AWAY FROM HOME? Prenton rates 3 0 e. DO ANY OF THE PEOPLE IN THIS HOUSEHOLD Relationship Tage | Relationship Age HAVE A HOME ANYWHERE ELSE? ag | f. ARE ANY OF THE PERSONS IN THIS HOUSEHOLD ON Yes No HEAD FULL - TIME ACTIVE DUTY IN THE ARMED FORCES? If “ses”, deere (J 2. HOW IS — — RELATED TO (head of household)? ee] {= eee ane eeee — I. = ni 2 2 3. PERSON NUMBER First column should have person 01. second column person 02, ete = 2 . | d B 1 1 | 42. HOW OLD WAS ~ — ON HIS LAST BIRTHDAY #ritr in next to “relationship” and mark | % 3 le Fame we Fame $b SEX Mark without ashing unless sex is not obtious from name C. RACE Mark without ashing White, Negr Crher | White Near Other i 5. 1S — — NOW MARRIED, WIDOWED. DIVORCED SEPARATED OR NEVER MARRIED’ IE meal r— I I 6. WHAT WAS — — DOING MOST OF THE PAST 12 MONTHS — wo» so mr (for males) WORKING OR DOING SOMETHING ELSE? | (for females) KEEPING HOUSE, WORKING OR DOING SOME THING ELSE? | Hf “SE” marked in (6 ard prison iv £3 vears old or otek. dst | Yo 3 | nN 7. 1S = = RETIRED? | | 7 If related persons 19 vears old eure Diged on addin tasthe veg. Ne — Th Bh oo { WEWOULD LIKE TO HAVE ALL ADULTS WHO ARE AT HOME TAKE PART IN THE ‘ Etre * n we INTERVIEW. IS YOUR - = ETC.. AT HOME NOW? (WOULD YOU PLEASE ASK - — | ETC. TO JOIN US?) | - . lL oo THIS SURVEY COVERS ALL KINDS OF ILLNESSES THESE FIRST QUESTIONS REFER TO [J Yes One| [J Yes Ore LAST WEEK AND THE WEEK BEFORE. THAT IS, THE 2-WEEK PERIOD OUTLINED IN | RED ON THIS CALENDAR Hand calendar to respondent and ask 8a. = 8a. WAS — — SICK AT ANY TIME LAST WEEK OR THE WEEK BEFORE (THE 2 WEEKS | SHOWN ON THAT CALENDAR)? | Bb. WHAT WAS THE MATTER? | c.DID - — HAVE ANYTHING ELSE DURING THAT 2-WEEK PERIOD’ J | 93 LAST WEEK OR THE WEEK BEFORE. DID - - TAKE ANY MEDICINE OR TREATMENT | [J Yes One | gg Oves Ono FOR ANY CONDITION (BESIDES . .. WHICH YOU TOLD ME ABOUT)? |b. FOR WHAT CONDITION? | c. DID ~~ TAKE ANY MEDICINE FOR ANY OTHER CONDITION? | : 10a. LAST WEEK OR THE WEEK BEFORE, DID — — HAVE ANY ACCIDENTS OR INJURIES? | [J Yeo One | pg 0 ves One b. WHAT WERE THEY? [ mc DID — - HAVE ANY OTHER ACCIDENTS OR INJURIES DURING THAT 2-WEEK PERIOD?! | 4 11a. DID — — EVER HAVE AN (ANY OTHER) ACCIDENT OR INJURY THAT STILL BOTHERS| [] ves One| [ves One HIM OR AFFECTS HIM IN ANY WAY? nm b. IN WHAT WAY DOES IT BOTHER HIM? Kecord present ffir | 12. Open your Flashcard booklet to Card A and read both sides of Card A (A-1, 4-2) | [J Yes ONo| [Yes Ono condition by condition; record in his column any conditions mentioned 3 for the person. : um Le E— ee BE oo ~ — 13. Turn 1 Cord B and rec beth sides of Card ROBT, 8-2), condition by condivion; | Yes [J No 10 ves Ono record in his column any conditions mentioned for the person. y nn i 142. DOES — ~ HAVE ANY OTHER AILMENTS. CONDITIONS. OR PROBLEMS WITH [J ves [ONo 10 Yes No | HIS HEALTH? |b. WHAT IS THE CONDITION? Kecord condition itself if sull present. otheruise record < | present effects } il | n Sc. ANY OTHER PROBLEMS WITH HIS HEALTH? | 31 : Cre | MES 3 B | Barmresons 18 yours ullioroner, Yow whe responded for iors morn duremeiiet 1), Responded tor selfwmirely [J Responded for selfentiely ¢ R ashing of) 4). 8-11. 1f persons responded for self. shou whether entirely or party. | [J Responded for self-portly [J Responded for selt-partly 3 Cg |For persons under 19 show who responded for thon If eligible respondent 1s " o | “at home™ but ded nor respond for self, enter the reason in a footnote LAR wi responders) Parser 1 ‘| 1 A I } Make 50 mark in this margin Make no mark in this margin 15a. HAS — — BEEN IN A HOSPITAL AT ANY TIME SINCE A YEAR AGO? | [J Yes 0 No [J Yes Ono If “Yes”, ask: b. HOW MANY TIMES WAS — — IN A HOSPITAL DURING THAT PERIOD? Times Times 16a. HAS ANYONE IN THE FAMILY BEEN IN A NURSING HOME, CONVALESCENT HOME, | [7] ves Ono [7] ves OJ No REST HOME OR SIMILAR PLACE SINCE A YEAR AGO? If “Yes,” ask: b. WHO? For each person reported in 16b ask: c. HOW MANY TIMES WAS — — IN A NURSING HOME OR SIMILAR PLACE DURING % un } THAT PERIOD? See Times Examine ages in question I for babies | year old or under. For cach child Month iDay Year Month "Day [Year 1 year old or under, ask 17a. | | | | 17a. WHEN WAS — = BORN? If on or after the date stamped in 15a, ask 17b. : ! i b. WAS — — BORN IN A HOSPITAL? If “Yes” and no hospitalizations entered in his | [J Yes (J No 0 Yes O No column, enter “I” in 15. If “Yes” and a hospitalization is reported for the | mother and baby ask 17c. c. IS THIS HOSPITALIZATION INCLUDED IN THE NUMBER YOU GAVE ME FOR- =? | [] Yes No O Yes O No If “No,” correct entry for mother and baby. THESE NEXT QUESTIONS ARE ABOUT RECENT VISITS TO OR FROM A MEDICAL DOCTOR. [J Nore [J None 18. DURING THE PAST 2 WEEKS (THE 2 WEEKS OUTLINED IN RED ON THAT CALENDAR) un HOW MANY TIMES HAS — — SEEN A DOCTOR EITHER AT HOME OR AT A Co DOCTOR'S OFFICE OR CLINIC? Cr.Vists CeVisws___ 19a. (BESIDES THOSE VISITS) DURING THAT 2 WEEK PERIOD HAS ANYONE IN THE J Yes 0 No OJ Yes Ono FAMILY BEEN TO A DOCTOR'S OFFICE OR CLINIC FOR SHOTS, X-RAYS, TESTS, OR EXAMINATIONS? If “Yes,” ask: b, WHO YAS THiS? | Mark “Yes,” in person’s colu " c. ANYONE ELSE? BRU Demons vatinar [ For cach “Yes” marked, ask: TT i I Nn d. HOW MANY TIMES DID — — VISIT THE DOCTOR? EXCLUDE visits made on “mass” basis. Vises Visits 20a. DURING THAT PERIOD, DID ANYONE IN THE FAMILY GET ANY MEDICAL ADVICE |] Yes Ono |W OJ Yes 0 No FROM A DOCTOR OVER THE TELEPHONE? If “Yes” ask: b. WHO WAS THE PHONE CALL ABOUT? | I c. ANY CALLS ABOUT ANYONE ELSE? | © % “ef fn persons cotumn. For each “Yes” marked, ask: mn d. HOW MANY TELEPHONE CALLS WERE MADE TO GET MEDICAL ADVICE ABOUT = =| oj tone Teles callstoDr. | calls to Dr. Visits reported in questions 18-20 for this person. Mark here —_— Viste efit hips Y Miah rapt in 1920 ) If no visits reported in questions 18-20 Ask: During past 2 weeks /not previously reported o During past 2 weeks/not previously reported O 21a. ABOUT HOW LONG HAS IT BEEN SINCE - - SAW OR TALKED TO A DOCTOR? thwiasonn 5 | ll les eens 0 Estimate is acceptable. If less than 1 year, mark appropriate circle; if more than I year, || T mark number of whale years. 2) 2 DK Never [51.4 Never ) Oo o If the last visit was within the past 12 months ask: De Nee pe Hew b. IN TOTAL, ABOUT HOW MANY TIMES HAS — — SEEN OR TALKED TO A DOCTOR 3. 2 Da DURING THE PAST 12 MONTHS? f) 3 ) If person is 55 years old or over, ask: THE FOLLOWING QUESTIONS REFER TO DIFFERENT KINDS OF adm. © | Sn. £ PERSONAL CARE SOME PEOPLE NEED AT HOME: ...................ccovieeeenn, Shr BSA. © Soot drt 7e CO 22a. DOES - ~ NEED ANY HELP IN BATHING, DRESSING OR PUTTING ON HIS SHOES? ...| 0 a0 0 we 0 ox © Lo mB YO 000] b. DOES — — NEED ANY HELP AT HOME WITH INJECTIONS, SHOTS OR OTHER TREATMENTS? ..........oo iio oii Yor Sp O Ne O DK Ves Sp O No O DK c. DOES - — NEED ANY ONE'S HELP WHEN WALKING UP STAIRS } } i OR GETTING FROM ROOM TO ROOM? . . ovo Yor Sp O Na O OK O Yes Sup O Na O DK d. DOES- — NEED ANY HELP AT ALL IN CARING FORHIMSELF? .................... Yos Sp O No O DK C Ye: Sp O No O OK 23a. DURING THE PAST 12 MONTHS, HAS — — RECEIVED ANY CARE | tna: © 0 RO " ier 6 oo Of G AT HOME FROM A NURSE? o.oo eee ’ b. DURING THIS 12 MONTH PERIOD, ABOUT HOW MANY VISITS \ \ DID A NURSEMAKE TO CARE FOR = =7 00 oto eee ee | / c. WERE ANY OF THESE VISITS DURING THE PAST 2WEEKS? ..._...................... Yu 0 MO DX Yes O No OOK | 66 Make no mark in this margin Wake no mark in this margin Condition Page Person number CONDITION NO. 1 1. Person number Write in and mark Enter person number and “name of condition” ! Name of and ask question 2. | condition Ask for all conditions 12. DID -- EVER AT ANY TIME TALK TO A DOCTOR ABOUT HIS. ..? Yes No v | ; 5d ; ! WASHINGTON USE Examine “Name of condition” entry in Item 1 | Aceld d Neith i and mark ane bor. ———e NCE OG Ces 0 Coma) Somer 27 P10 neo If “Doctor talked to”, ask: —— 3a. WHAT DID THE DOCTOR SAY IT WAS? DIDHEGIVEITA i If *Doctor not talked to record MEDICAL NAME? Cond. adequate description of EE | condition or illness. ! | “No. of this | 3b. WHAT WAS THE CAUSE OF. . .? condition a A fons or injury Wei tnd Chronic Acute : - Total If the entry in 3a or 3b includes the words: 3c. WHAT KIND OF ... IS IT? ote Asthma ~~ “Ailment” “Disease” / ng Yes fe Cyst “Attack” “Disorder” code vr to Growth “Condition” “Trouble” | Ask: | Required } - " hospitalization Measles Defect ! | TM Cth - Tumor : Cther Acc. For ALLERGY OR STROKE, Ask: ——> 3d. HOW DOES THE ALLERGY (STROKE) AFFECT HIM? ' IC or dum | code. "For conditions on Card B-2 and for any entry | 38. WHAT PART OF THE BODY IS AFFECTED? Person days of disability v that includes the words: ! ] RA. Abscess Cyst Paralysis Ache (except ~~ Growth Sore ¥ headache) Hemorrhage Sereness SHOW THE FOLLOWING DETAIL: Wks. | BD. | Bleeding Infection Tumor Ear or eye. .one or both Blood clot Inflammation Ulcer ) Ask: | Head... skull, scalp, face Under 6 v Boil Neuralgia Weak ’ Back ....... upper, middle, lower Tu! a Cancer Neuritis Weakness | Arm... shoulder, upper, elbow, lower, wrist, ! Cramps (except Pain | hand; one or both v menstrual) Palsy / | Leg......... hip, upper, knee, lower, ankle, foot; : 12 o one or both Months tk FILL QUESTIONS 4--8 FOR ALL ACCIDENTS OR INJURIES 4a. DID THE ACCIDENT HAPPEN DURING THE [J During past 2 years-Ask 45 | 6a. WAS A CAR, TRUCK, BUS, OR OTHER PAST 2 YEARS OR BEFORE THAT TIME? (J Before 2 years-Go to Sa MOTOR VEHICLE eh IN THE Wes: So Zotul? v - - - = - ( 4b. WHEN DID THE ACCIDENT HAPPEN? Enter month and year; mark one box rr EI A NY geome semen oo [OD Lost week b. WAS MORE THAN ONE VEHICLE Yes No Month Year [[] Week before INVOLVED? o 2 ks.» 3 th EA LIE TE Ts ist em L SS aan c. WAS IT (EITHER ONE) MOVING AT Ye to v 1 -2 years THE TIME? o 0 : Ask for all accidents or injuries: 7. WHERE DID THE ACCIDENT HAPPEN? 5a. AT THE TIME OF THE ACCIDENT WHAT PART OF THE BODY WAS HURT? At hore finside house) ov... o WHAT KIND OF INJURY WAS IT? ANYTHING ELSE? Specify place At hore (adjacent premises) . . ; Port(s) of body Kind of injury(injuries) Street and highway (includes roadway). . . . oO re errr? I Industrial ploce (includes premises) . . School includes school premises)... ..... - - - Place of recreation ond sports {net schaol) . —— Cher (specify place where accident happened) (1 = ~ v 8. WAS - — AT WORK AT HIS JOB OR Urine 4 Wille th Yes oor time Armed Forces If accident happened BEFORE 3 months, ask: a EN THE ACOINENT i * a ' 2 , : 5b. WHAT PART OF THE BODY IS AFFECTED NOW? : HOW IS HIS —— AFFECTED? Footnotes Part(s) of body ~~ Present effects IL | ‘m Make ne mark in this margin Nake no mark in this margin Condition Page—Con. CONDITION (Con'd.) REFER RESPONDENT TO TWO-WEEK CALENDAR FOR QUESTIONS 9-14 Ask question 9a for all conditions. 9a. LAST WEEK OR THE WEEK BEFORE DID HIS . . . CAUSE HIM TO CUT DOWN ON } Yes Ne = Go to I4a v THE THINGS HE USUALLY DOES? | ee — ER ! Yes No - Go to I4a v | b. DID HE HAVE TO CUT DOWN FOR AS MUCH AS A DAY? i > ) Ask questions 10 and 11 if “Yes” 10. HOW MANY DAYS DID HE HAVE TO CUT DOWN i v marked in question 9b. : DURING THAT TWO WEEK PERIOD? Write in ] Doys | ! and mark YS) IT DURING THAT TWO WEEK PERIOD, HOW MANY DAYS ! Nore v } DID HIS... KEEP HIM IN BED ALL OR MOST Write in [1] Do < | : OF THE DAY? and mark % Ask question 12 if person is ) 12. HOW MANY DAYS DID HIS . . . KEEP HIM FROM | Under 6 None v 6-16 years old. , SCHOOL DURING THAT TWO WEEK PERIOD? Write in [1] D ! A and mark ys Ask question 13 if person is 113. HOW MANY DAYS DID HIS . . . KEEP HIM FROM WORK None 17 years old or over. ! DURING THAT TWO WEEK PERIOD? (For females add) Write in | NOT COUNTING WORK AROUND THE HOUSE? and mark Ask question 14 for all conditions. ' 14a. WHEN DID HE FIRST NOTICE HIS . . .? WAS IT DURING THE PAST 3 MONTHS OR BEFORE THAT TIME? i b. DID HE FIRST NOTICE IT DURING THE PAST TWO WEEKS OR BEFORE THAT TIME? During 3 mos. Before 3 mos.-Gote IS V Past 2 wks. Before 2 wks.-Goto 16 VV Lost week Week before v | c. WHICH WEEK, LAST WEEK OR THE WEEK BEFORE? | : jie va Ask question 15 only if condition | 3-12 mo Before 12 mos. v was first noticed “Before 115. DID - — FIRST NOTICE IT DURING THE PAST 12 MONTHS OR BEFORE THAT TIME? ) 3 months.” ! \ Ask for person 6 years old or [J Not an eye condition [J Not first eye condition [J Under 6 \ over for whom an eye condition | | Yes - Ask 166 No -Omic 168, ¢ or vision problem (including | 16a. CAN - — SEE WELL ENOUGH TO READ ORDINARY NEWSPAPER PRINT WITH GLASSES? | tt a fa m5 oP sl i tm nt: fe 3 a 1 ae to 1 so " - - cotaracts and glaucoma) hes [y " CAN _ _ SEE WELL ENOUGH TO RECOGNIZE A FRIEND WALKING ON THE OTHER DT Yeomiise | Notniie been reported. : SIDE OF THE STREET? | sm sit ms a 4 Sr eR ev i Cee mm crm tt fo - RA | c. HOW MUCH TROUBLE WOULD YOU SAY THAT — — HAS IN SEEING: A GREAT DEAL, | Hrd | SOME, OR HARDLY ANY AT ALL? : Gren deol Some or none AA: IF THIS IS A CONDITION ON CARD A OR B, OR STARTED “BEFORE 3 MONTHS,” ASK Q. 17; OTHERWISE GO TO ITEM BB. Ask question 17b if “1” or more | 17a. ABOUT HOW MANY DAYS DURING THE PAST | Noe Goro88 V days in question 17a and | 12 MONTHS HAS HIS . . . KEPT HIM IN BED re | question 11 is blank or ALL OR MOST OF THE DAY? en [1] Doys{ marked “None.” 4 ! * b. WERE ANY OF THESE — — DAYS DURING LAST ST TTT Ye NeGewss Write in minal] oor] BB: Is this the LAST condition for this person? 0 Yes — Ask 18-21 if person has “I” or more conditions past AA [J No — Go to next condition Show Card D, E, F, or G, as appropriate based on activity status or age. 18. PLEASE LOOK AT EACH STATEMENT ON THIS CARD (CARD D, E, F, G). THEN TELL ME WHICH STATEMENT FITS — — BEST IN TERMS OF HEALTH. Mark statement number —»— 1'2 3 4.Cowo20 If 1, 2, or 3 marked in 18 ask: —>— 19. If 4 marked in 18 go to 20. | | | 1 | | | | | | | | | | I | IS THIS BECAUSE OF ANY OF THE CONDITIONS YOU HAVE TOLD ME ABOUT? On WHAT DOES CAUSE © THIS LIMITATION? Enter cause WASHINGTON USE OK 20. PLEASE LOOK AT THE BLUE CARD, CARD H. WHICH ONE OF THOSE STATEMENTS FITS — — BEST IN TERMS OF HEALTH? Mark statement number —»— 123 4 5 6si0p If1, 2, 3, 4, or 5 marked in 20, ask: If 6 marked, omit 21 and } 80 to next person. 21. IS THIS BECAUSE OF ANY OF THE CONDITIONS YOU HAVE TOLD ME ABOUT? WHAT DOES CAUSE OI No—> "This LIMITATION? Tri 2aani ‘| WASHINGTON USE 67 Wake no mark in this margin Make no mark in this margen 68 | _ Person number HOSPITAL PAGE 1. Person number Welt. thank Enter month, day, years if the | YOU SAID THAT — — WAS IN THE (HOSPITAL/NURSING [yoo] | WASHINGTON USE | exact date is not known, | HOME) DURING THE PAST YEAR: obtain the best estimate. ! Jan A uly Cet 2. WHEN DID — — ENTER THE (HOSPITAL/NURSING eed | ls Mam ! HOME) (THE LAST TIME)? oy Mar June Sept Dec | Write in @ USE YOUR CALENDAR ; — Da Year Y Make sure the YEAR is correct. ——>— - . 4 = -, Year Do not include any nights in | 3. HOW MANY NIGHTS WAS — — IN THE i Sr a oo _ interview week. If the exact (HOSPITAL/NURSING HOME)? Total nights in hospital — 1] number is not known, accept nursing home | B® the best estimate. ! Nights Complete question 4 from 4a. HOW MANY OF THESE — — NIGHTS — oo] entries in questions 2 and 3; | WERE IN THE PAST 12 MONTHS? Nights past 12 months if not clear, ask the ! Q. No. 15 16 17 Hop. Other questions. | aes — | BH Do not include any nights IB, HOW MANY OF THESE - - NIGHTS = | in interview week. WERE LAST WEEK OR THE Nights past 2 weeks Dey. WEEK BEFORE? | c.WAS — — STILL IN THE (HOSPITAL/NURSING Dioanos: TT] USE YOUR CALENDAR ~*~ HOME) LAST SUNDAY NIGHT FOR ad = THIS HOSPITALIZATION (STAY)? OvYes [No sli es Se B "5. FOR WHAT CONDITION DID — — ENTER THE (HOSPITAL/NURSING HOME) — pen mottadl nons noc inom, DO YOU KNOW THE MEDICAL NAME? - enter an adequate 5 . 9 If “No® ask: ne For delivery ask: WAS THIS A NORMAL DELIVERY? | J CREE 0 hreer | Comarion For newborn, ask: WAS THE BABY NORMAL AT BIRTH? V Record in *Condition” box a |] ! [Cordiiten E———— | Operation 3 | Entry must show CAUSE, KIND, | Cause To SESS and PART OF BODY in same . detail as required for the servic | Condition page. ! Kind . | | ! ~ | | Cornershi | Part of body | _ IC or dum. — - a ——— de | BH 6a. WERE ANY OPERATIONS PERFORMED ON - - DURING | | | THISSTAY AT THE (HOSPITAL/NURSING HOME.)? [J Yes [JNo-Goro7| : | If name of operation is not | b.WHAT WAS THE NAME OF THE OPERATION? Footnates: [ known, describe what ! was done. [Operation - c. ANY OTHER OPERATIONS ? [1 Yes - Describe above O No | | 7. WHAT IS THE NAME AND ADDRESS OF THE (HOSPITAL/NURSING HOME)? ‘m Enter the full name of the hospital or nursing home; | the street or highway on | Name of Hospital which it is located, and the [ city and State; if the city is | | not known, enter the county. | Street | | | City (or county) "State : —>— CONTINUED ON NEXT PAGE —<— HN | ® HOSPITAL PAGE (CONT'D) ASK QUESTIONS 8-10 FOR ALL COMPLETED HOSPITALIZATIONS Mark one “Yes” inQ. 4c Coto 16 O Make no mark in this margin Nake no mark in this margin circle “No inQ. dc -Ask 8-10 © Ask if “No” marked in question dc: Dollars | Cents Tok ay aSkiNGTON USE 8. WHAT WAS THE TOTAL AMOUNT OF THE (HOSPITAL/NURSING HOME) BILL FOR THIS STAY? ! DO NOT INCLUDE DOCTORS' OR SURGEONS’ BILLS. "9a.DID (WILL) HEALTH INSURANCE PAY ANY PART OF Se THISBILL? [7 Yes [ONo-Goto 10 Name of Insurance Plan Dollars | Cents b. WHAT IS THE NAME OF THE INSURANCE PLAN? —>— | TE ! 10. Source c.DID (WILL) ANY OTHER HEALTH INSURANCE PLAN PAY |_ ER STI. SR oe ABCDE FGHID PART OF THIS (HOSPITAL/NURSING HOME) BILL? | 00000 00000 | Amount BL DK If “Yes” Reask 9 |. | ° For each Health Insurance Plan named, ask: ! d. WHAT WAS (WILL-BE) THE AMOUNT | PAID BY (Name of Plan)? Enter total amount paid by health insurance in line A ~ Dollars | Cents | Foo ___ Enter ANY amount paid by Social Security Medicare in line B All plans-exclude ' 10. Source 2 A [J Health insurance- Hevicare | ABCDE FGHIDK 10a.WHO PAID (WILL PAY) THE (REMAINDER OF THE) Fy RR HOSPITAL BILL? Mark each category mentioned 8 O Social Security Medicare ! oO ev er a wl b.DID ANY OTHER PERSON OR AGENCY PAY ANY : OTHER PART OF THE HOSPITAL BILL? C [J Self and/or Family [J Yes-Ask 10c [J No-Go to 10d a am — ray REE ER ag mm em RE A Sn SA mm ABCDE FGHID L SNOWING Widahsngymndond ne € [OJ Friend rams ioe d.WHAT WAS THE AMOUNT PAID BY - =? } ? Enter amount paid opposite appropriate category. F [J Kerr Mills or other Fed. Plans | "INTERVIEWER: TTT 16 [OD Armed Forces Medicare Add amounts entered (include any amount paid by health 70. Se Co rT coo insurance) and enter in TOTAL box, then mark one of the source following boxes. H [J State or Local Welfare Agency ABCDE FGHID [J Toto! amount paid (to be paid) agrees with ) ! PR Soe hb a amount of hospital bill - Go to Q. 11 I [0 Other Specify | 00 [J Total amount paid (to be paid) does NOT agree TCTAL OF ABCVE - include amount —— | with amount of hospital bill - Resolve difference paid by health insurance ! with respondent. t Under 55 - Go to 14 5S of over- Ask la ASK QUESTIONS 11 - 13 IF PERSON IS 55 YEARS OLD OR OVER Mark one circle ————>— - 11a. WHEN — - LEFT (Name of hospital /nursing home), [J Home = Go to Question 12 WASHINGTON USE DID HE RETURN HOME OR GO SOME OTHER PLACE? [[] Some other place — Ask Question 116 b.WHAT KIND OF PLACE DID - = GO TO? Specify —>— INTERVIEWER: If the “Place” in 11b is a Hospital, Nursing i i} Home or a similar place, was a Hospital _— 0 Hospital poge not filled-Fill Hosp. page for unreported stay. Page filled for that stay? Mark one box. [[] Hospitel page filled-Stop Blonk (and 55) C ger SEC Home O Some corer place O 12. AFTER LEAVING THE (HOSPITAL/NURSING HOME,) HOW MANY DAYS DID — - HAVE TO REMAIN IN BED ALL OR MOST OF THE DAY? Mark entry Still in bed - Go to I4 None 13. (ALTOGETHER) HOW MANY DAYS WAS — — CONFINED TO RETURNING HOME FROM THE (HOSPITAL/NURSING HOME.)? Mark entry THE HOUSE AFTER Still confined to house © None OK 0 14. NOTE TO INTERVIEWER: If the condition in question 5 or 6 is on Card A (A-1, A-2) or B (B-1, B-2) or there is “I” or more nights in question 4b, the condition must have a completed Condition page. If the condition does not have a Condition page, fill one after completing all required Hospital pages. 69 Make no mark in this margin Person number DOCTOR Visits: PAGE (1) i 1. Person number Write in and mark See questions 18-21a on Pages 4 and 5 | Record each date on which a Doctor was | EARLIER YOU TOLD ME THAT — — HAD SEEN OR TALKED TO A To we Wo mo visited in a separate Question 2a of the | DOCTOR DURING THE PAST 2 WEEKS. Write in and mark Month Feb Moy © Aug Nov Doctor Visits Questions. ! | Mor © Joe O Seg Dee | 2a. ON WHAT DATES DURING THAT 2-WEEK LW wB i PERIOD DID - = VISIT OR TALK TO A DOCTOR? the last set of Doctor Visits Questions Ask and record the answer to Question 2b on | b. WERE THERE ANY OTHER DOCTOR VISITS FOR - — DURING THAT PERIOD? Moke no mark in this margin No. Visits for each person. [J Yes-Reask Q. 2a [J No-Ask Q. 35 for each visit ltem D: Interviewer Check Item 3. WHERE DID - - SEE THE DOCTOR ON THE (Date)? Mark one circle orn Talwphors . Enter the number of Doctor Visits reported in CHlicy: ws usmss ( for each person in question 18-21a on | Pre-pcid Insurance Growp «© pages 4 and 5. If “None” reported for all || Hospital Emergency Room persons, check here | Hospital Cut-patient Clinic ( [0 None -eport=d Go to Person pages Heolth Department . . . . . . Componyoe industeg ... +. ( T T essen emitimtmsmmsants ISTHE SHRI cacemmmonsn Person) 102 103 104 | 05 | 06 ther Specily . | | . WASHINGTON USE Fill one Doctor Visit section for each visit | . Co orcall reported including additional visits| 4. HOW MUCH WAS THE DOCTOR'S BILL FOR THAT VISIT (CALL)? or calls reported in question 2b. ’ Dollars If bill not received, ask: [Dollars Cents] FOOTNOTES: | HOW MUCH DO YOU EXPECT THE DOCTOR'S | BILL TO BE FOR THAT VISIT (CALL)? : Cents /5. IS THE DOCTOR A GENERAL PRACTITIONER OR A SPECIALIST? Z z i [J General Practitioner [J Specialist wm Sa} \ First es No "If “Specialist” ask: WHAT KIND OF SPECIALIST IS HE? — Visit? ao ry Kind of Spec. i Person number DOCTOR VISITS PAGE (2) i 1. Person number Write in and mark Record each date on which a Doctor was H EARLIER YOU TOLD ME THAT —- — HAD SEEN OR TALKED TO A { Jn O Ap O ly Oct visited in a separate Question 2a of the | DOCTOR DURING THE PAST 2 WEEKS. Write in and mark Month Feb O Moy O Aw CO Nov O Doctor Visits Questions. ! | Mor O June Sept © Dec C '2a. ON WHAT DATES DURING THAT 2-WEEK wwe PERIOD DID - — VISIT OR TALK TO A DOCTOR? Day! 0 Ask and record the answer to Question 2b on | b. WERE THERE ANY OTHER DOCTOR VISITS FOR — — DURING THAT PERIOD? the last set of Doctor Visits Questions for each person. | [] Yes-Reask Q. 2a 0 No-Ask Q. 3-5 for each visit | FOOTNOTES: 13. WHERE DID - — SEE THE DOCTOR ON THE (Date)? Mark one circle Pre-pcid Insurance Growp . O Hospital Emergency Room Hospital Cut-patient Clinic © Heolth Department 0 Cormrpany or Indust af Cthar Specify ........ | WASHINGTON USE "4. HOW MUCH WAS THE DOCTOR'S BILL FOR THAT VISIT (CALL)? py | ollars U6 not receiund, ask: Dollars Certs || | HOW MUCH DO YOU EXPECT THE DOCTOR'S | BILL TO BE FOR THAT VISIT (CALL)? ! Cents : ! 5. IS THE DOCTOR A GENERAL PRACTITIONER OR A SPECIALIST? Br ! O General Practitioner 0 Specialist eT } If “Specialist” ask: WHAT KIND OF SPECIALIST IS HE? 7 iE 0 | ! 3 Kind of ! Spec. ’ | Ronis Be None =Co to 25a 1 Make no mark in this margin Make no mark in this margin Ask for all persons 17 years old or over. Elementary |E | El 24a. WHAT IS THE HIGHEST GRADE (YEAR)- ~ATTENDED IN SCHOOL? High school [Hi Under 17 Hi Under 17 College Co Co o oo TTT oo | vo No b. DID- FINISH THE-~GRADE (YEAR)? 7 S o Ask for all persons 17 years old or over. Yes No Yes No 25a. DID-—WORK AT ANY TIME LAST WEEK OR THE WEEK BEFORE? Co to 260 Askborh band ¢ Coto26a Ask both band c For females add: NOT COUNTING WORK AROUND THE HOUSE? © ° | S o m "“b. EVEN THOUGH--DID NOT WORK DURING THOSE 2 WEEKS, DOES HE HAVE | vee wo | ve Ne A JOB OR BUSINESS? o 0 o o c. WAS HE LOOKING FOR WORK OR ON LAYOFF FROM A JOB? Hor Ont Yau~trky Har mird Oo oO Oo d. WHICH - LOOKING FOR WORK OR ON LAYOFF FROM A JOB? pee El pmo en If *Yes® in 25¢ only I Ask for all persons with a “Yes” in 25a, 25b, or 25c. Employer Employer questions 26a "| 26a. WHO DOES (DID)-~WORK FOR? through 26d apply | to this person’s TTT TTTmmmm mmm a | vir ¢ 5 LAST full-time |b. WHAT KIND OF BUSINESS OR INDUSTRY IS THIS? Industiy deity civilian job. un | un c. WHAT KIND OF WORK IS (WAS)--DOING? Qecupation Srsipotion "Fill 26d from entries in 26a-26c; if not clear, ask. | ( Poomid GorFed GorCer| Pusoid GoviFed. GovrCher d. CLASS OF WORKER ) Bn 0 0 o | Own Non-oid ~~ Nev-Warked Own Non-poid ~~ Nev-Worked | Oo ) ( Oo oO o Ask for all males 17 years old or over. Yes No-Goto28 Yes No- Go to 28 27a. DID-—EVER SERVE IN THE ARMED FORCES OF THE UNITED STATES? 0 0 o o Ce sw Ne OK Ves-sop No I ok b. WAS ANY OF HIS SERVICE DURING A WAR? o 5 0 "| o o o un © If *No” or “DK” in 2b ask: TT Yes-sp No OK Yos-Sp No © x c. WAS ANY OF HIS SERVICE BETWEEN JUNE 27, 1950, AND JANUARY 31, 1955? 3 C o o o = If “No” or “DK” in 27¢ ask: - oo Yes TTT No o OK Yes ) No Ni ok o d. WAS ANY OF HIS SERVICE AFTER JANUARY 31, 19557 - z > 0 28. WHICH OF THESE INCOME GROUPS REPRESENTS YOUR TOTAL COMBINED FAMILY un " AB CDEFG HI J Vv AB CDEFG HI J Vv INCOME FOR THE PAST 12 MONTHS - THAT IS, YOURS, YOUR —'S, ETC.? | ~ SHOW CARD I. INCLUDE INCOME FROM ALL SOURCES SUCH AS WAGES, - - SALARIES, SOCIAL SECURITY OR RETIREMENT BENEFITS, HELP FROM RELATIVES, RENTS FROM PROPERTY, AND SO FORTH. Mark income group in each related persons column. FOCTNCTES WASHINGTON USE WASHINGTON USE BM WASHINGTON USE =m *Transcribe codes for Respondent Item R (Respondent) | 0 — Self-entirely | Age of respondent 1 — Self-partly ! SF SF J PL SE PF mmm me @ Pl SI PF me ee nu 2 — Spouse ! | Family relationship 3 — Mother | Heod | Heod 2+ Wife Child Cth. relative Heod | Heod 2+ Wife Child Cth. relative 4 — Father i - : c 2 ' Und. 17 None Und. 17 Nene 5 — Other female family | Egucation of head o 3 - makes | m n 6 — Other male family member 7 — Other | Industry | AB CODEFG HJKLM AB CDEFG HJKLM | > 00000 00000 00 00000 00000 oom ee Ween m | Occupation NP QRSTU VWXYZ NP QRSTU VWXYZ 00 00000 ) O OX O00 00000 OO 0C ‘; nm Card A Card D Card F Card H A--1 Now I'm going to read a list of | A--2 Have you, your , etc., had | For: For: For: Mobility conditions--Please tell me if any of these conditions DUR- Workers and other persons except Children from 6 through 16 years old you, your, efc., have had ING THE PAST 12 MONTHS? | Housewives and Children any of these conditions DUR- ING THE PAST 12 MONTHS? 1. Asthma? 12. Thyroid trouble or goiter? 1. Not able to work at all. 1. Not able to go to school at all. 1. Must stay in hed all or most of a he time. 2. CHRONIC bronchitis? 13. Any allergy? 2. ‘Able vo work but limited in amount of 2. Able to go to school but limited to fhe ume. 3 REPEATED attacks of sinus 14. CHRONIC nervous trouble? work or kind of work. certain types of schools or in 2. Must stay in the house all or trouble? 15. CHRONIC skin trouble? 3. ‘Able to work but limited iv Kind ot school attendance most of the time. 4. TROUBLE with varicose veins? 16. Palsy? amount of other activities 3. Able to go to school but limited in 3. Need the help of another person in : : . . Lo . other activities. getting around inside or outside the 3. Hemouhols or pilus? 17 Paralysis of any kind? 4. Not limited in any of the above ways. i he abo Poise 6. Hay fever? 18. REPEATED trouble with i Noulmited mvanyotithe above ways. . Hay fever? Sock or spine? 4. Need the help of some special aid, 7. Tumor, cyst, or growth? 1. one > such as a cane or wheelchair, in 8. CHRONIC gallbladder or liver s Cléltpaluts getting around inside or outside trouble? 20. Any speech defect? the house. 9. Stomach ulcer? 21. Hemia or rupture? S. Does not need the help of another 10. Any other CHRONIC stomach 22. Prostate trouble? person or a special aid but has trouble? trouble in getting around freely. 11. Kidney stones or CHRONIC 6. Not limited in any of the above ways kidney trouble? Cord B Cord E Card G Card | B--1 Have you, your, etc., B--2 Do you, your , etc., HAVE | For: Housewife For: Children under 6 years old Which of the following income groups EVER had any of these any of these conditions? represents your total combined fomily conditions? income for the past 12 months? In- clude income from all sources such 1. Tuberculosis? 1. Deafness or SERIOUS trouble | 1. Not able to keep house at all. L. Movable uw silaiphrvianallduondin ary ys saclal security 2. Emphysema? hearing with one or both ears? | 5 Able to keep house but limited in play with other children. or retirement benefits, help from rela- 3. Hardening of the arteries? 2. SERIOUS trouble seeing with amount or kind of housework. 2. Able to play with other children but tives, rents from property, and so forth. 4. High blood pressure? one or both eyes even when 3. Able to keep house but limited in kind limited inumopnene kindiotiptuy 5. Cancer? wearing glasses? o amount of other activities. 4. Not limited in any of the above ways. : 3. Missing fingers, hand or arm -- Under $500. . . . . . . Group A 6. Heart trouble? foo 4. Not limited in any of the above ways. frase woup 7. Stroke? 4 Mizsing ung or kidney tor $500-- $999... .. ..... Group B 8. Rheumatic fever? breast)? $1,000-- $1,999. ......... Group C 9. Arthritis or theumatism? S. Club foot? 2,000-- wae RR ¥ 10. Mental illness? 6. PERMANENT stiffness or any $2,000. $2,959 GrougiD 11. Diabetes? deformity 2 foot, leg, fingers, $3,000 $3,999. ......... Group E Epi back? 12. Epilepsy? RENEHEA0 $4,000-- $4,999... ....... Group F $5,000-- $6,999... ....... Group G $7,000-- $9,999... ....... Group H $10,000--$14,999 . . . . . . Group 1 $15,000 and over . . . . . . . Group J 72 APPENDIX IV QUESTIONNAIRE—PERSON APPROACH NOTICE - All information which would permit identification of the individual will be held in strict confidence, will be used only by persons engaged in and for the purposes of the survey, and will not be disclosed or released to others for any purposes. BUDGET BUREAU NO. 68-56701.6 APPROVAL EXPIRES JAN. 1, 1968 FORM MHS-HIS-) X (1968) 7) U.S. DEPARTMENT OF COMMERCE BUR (oe UREA OF THE CENSUS ACTING AS COLLECTING AGENT FOR THE U.S. PUBLIC HEALTH SERVICE Book of Books U.S. HEALTH INTERVIEW SURVEY 2. Street Address — House No., Street, Apt. No. or other ident. | FOR AREA [a. ] (If marked **Ask’ do so before the interview) | SEGMENTS \ rim er mate rere peer OSE | ENTER: | ! % [JAsk ——>| When was this structure originally built? eet * ee What was the matter? Kid 1 1 | 1 1 I and PART OF BODY in same | For newborn, ask: Was the baby normal at birth? detail as required for the 4 R eres, | Part of body | T 1 | + I 1 Record in “Condition” bo Condition page. Ask for all conditions except deliveries and births 6. Was this the first time — — was hospitalized for... ? 1[_]Yes 2[ No 7a. Were any operations performed on — — during this stay at the (hospital/ nursing home.)? [JYes 2[ |No—Go to 8 | If name of operation is not ) mmm mm mm mm mm mm mm me ee me ee mm em mm em mm em mm mm mm mmo pr Rasa, describe what was |b. What was the name of the operation? Operation c. Any other operations? []Yes—D. ib: CINo 8. What is the name and address of the (hospital /nursing home)? 1 1 I 1 1 Enter the full name of the I hospital or nursing home; the : L 1 I 1 I i street or highway on which it Name of Hospital: is located, and the city and Street City (or county) State State; if the city is not known, enter the county. Ask questions 9-18 for all completed hospitalizations — Mark one box [ _]**Yes'" in Qdc — Go to Item 18 Hospital Bill [“]*No” in Q4c — Ask Q. 9 Dollars I Cents 1 9. What was the total amount of the hospital (nursing home) bill for this stay? Do not include any doctor's or surgeon's bills. ] 100. Did (will) health insurance pay any part of the hospital bill? | (es — Ask NO — Gi CJyss—aers CNe ~ Go to 12 Name of Insurance Plan Dollars _ 1 Cents 1 i c. Did (will) any other health insurance plan pay part of [JYes — Reask b ST TT “== mm this hospital (nursing home) bill? [No —Go tod i ~ 7 “For each Health lnsurance Plan named, ask: ~~ ~~ ~~~ ~~ ~~~" "777777 [Ee rE re Se er semper em d. What was (will be) the amount paid by (Nome of plan)? | Fnter total amount paid by health insurance in line A Source of Payment Dollars Cents Enter any amount paid by Social Security Medicare in line B 11a. Who paid (will pay) the (remainder of the) hospital bill? A. 1[JHealth Insurance—All plans excl. Medicare tre tree ion Si Sm He, Sh Sn TSAR} Te Piet en ST gp i mks sess: Sel smcohe mt Stn Sono a wm bi Did ony offier person or. agenay pay sny wiles gait ohths feapirel illy LIYEASke Ly moon sovaiity Medios 1 [[INo—Go tod - : 1 TTEWhewesthis? TTT CT TT TT TT TTT Tomo oom oom ee TT rn i nn etn Si, on em fe C. 3[1Self and Family ! d. What was the amount paid by — bajo sper) [TTC 1 we | | Interviewer: After totaling all sources of payment for the hospital bill, check one of the following boxes: Total of above—include amount 7 paid by health insurance ————— | [Total amount paid (to be paid) agrees with amount of the hospital bill = Go to Q. 12 1 [J Total amount paid (to be paid) does not agree with amount of the hospital bill-Resolve difference with respondent 80 HOSPITAL PAGE (Cont'd) Doctor/ 5 irgeon Dollars Cents 120. What was the amount of the doctor's and surgeon's bill for this stay? b. Is the § for the doctor's and surgeon's bill included in the $ amount you gave me for the hospital bill? [Yes (In a footnote, indicate the actual amount of the hospital bill after deducting the 2[JNo—Go to 13 doctors and surgeons bills, also indicate any changes in the amounts paid by health insurance or other sources if the entries in Qs. 10 and 11 include payments for expenses other than the hospital bill). Enter any amount paid by Social Security Medicare in line B 14a. Who paid (will pay) the (remainder of the) doctor's and surgeons bill? A. 1[ Health Insurance—All plans excl. Medicare I 13a. Did (will) health insurance pay any part of the doctor's and surgeon's bill? [Yes [JNo—Go to 14 Name of Insurance Plan Dollars | Cents b. What is the nome of the Insurance Plan? ] I ATER Ee ET Se SUSE ei SC I ol oo lt od Sh CA th lt i lesa wo mmm — c. Did (will) any other health insurance plan pay part of the doctor's and surgeon's bill? r [ Yes—Reask b { No—Askd bem eee mmm L ——— = i. = - EE mmm E 1 For cach Health Insurance Plan named, ask: ra mc on Tn wm ne i d. What was (will be) the amount paid by (Name of plan)? | | Source of P t Dollars | C Fnter total amount paid by health insurance in line A wenn ant ory : Ry 1 | ! b. Did any other person or agency pay any other part of the doctor's and surgeon's bill? Yes—Ask c [| No-Go tod c. Who was this? Lwin Sm D. 4[ ]Other—Specity d. What was the amount paid by — Total of above~include ; amount paid by health [~]Total amount paid (to be paid) agrees with amount of doctor's bill-Go to Q. 15 insurance —> Interviewer: After totaling all sources of payment for the doctor's and surgeon's bill, check one of the following boxes: [| Total amount paid (to be paid) does not agree with amount of the doctor's hill-Resolve difference with respondent Mark one box (Under 55-Go to 18 [155 and over—Ask 15a 150. When — — left (name of hospital ‘nursing home) did he return home or go some other place? Home-Ga to 16 | ]Some other place—Ask 15b b. What kind of place did — — go to? Specify Interviewer: If the place in 15h is a hospital, nursing home or similar place, was a hospital page filled for that stay? iiospital page filled-Stop [Hospital poge not filled—Fill hospital page for unreported stay 16. After leaving the hospital (nursing home) how many days did — — hove to remain in bed all or most of the day? ooo[ _|None xx x[_]Still in bed LC] days 17. ALTOGETHER how many doys was — — confined to the house after returning home from the hospital (nursing home)? 000 None xxx{__|Still confined to or] days 18. NOTE TO If the condition in Q.5 or 7 is listed in item AA on the Condition Page or any part of this hospitalization was during the past 2 weeks the condition must have a completed Condition Page. If the condition does not have a Condition page, fill one after completing all INTERVIEWER required Hospital pages. DOCTOR VISITS (1) C1] - 1. Person number | | | I Record each date on which a Doctor | was visited in a separate question 2a | of the Doctor Visits questions. | | 5 Ask and record the answer to question | 2b on the last set of Doctor Visits | questions for each person. | Earlier you told me that — — had seen or talked to a doctor during the past 2 weeks. 2a. On what dates during that 2-week period did — — visit or talk to a doctor? Month Day = Sromerrrmes b. Were there any other doctor visits for — — during that period? [JYes—Reask Q. 2a [CINo—Ask Q. 3-5 for each visit FOOTNOTES: 3. Where did — — see the doctor on the (Date)? Mark one o01[_JHome 20 [_]Doctor’s Office 40 [_] Hospital Out-patient Clinic ~~ 70[_]Health Department 10 [_] Telephone 30 [Pre-paid Insurance Group 50 [| Hospital Emergency Room 8o[__]Company or Industry 90[]Other- Specify xx[__|While inpatient in hospital ~ Stop 4. How much was the doctor's bill for that visit (call)? Dollars | If bill not received, ask: | 1 1 How much do you expect the doctor's bill to be for that visit (call)? 5. Is the doctor a general practitioner or a specialist? [JGeneral Practitioner [Speci If ‘Specialist’ ask: What kind of specialist is he? — st DOCTOR VISITS (2) 1. Person number CL] Record each date on which a Doctor was visited in a separate question 2a of the Doctor Visits questions. 2a. On what dates during that 2-week period did — — visit or talk to a doctor? Month | ] Day b. Were there any other doctor visits for — — during that period? [CJYes—Reask Q. 2a [JNo—Ask Q. 3-5 for each visit FOOTNOTES: 3. Where did — — see the doctor on the (Date)? Mark one o1 [Home 20 [J Doctor’s Office 40 [_JHospital Out-patient Clinic ~~ 70 [|Health Department 10 [Telephone 30 [] Pre-paid Insurance Group so [JHospital Emergency Room so [_] Company or Industry 90 []Other—Specify XX[__]While inpatient in hospital —Stop 4. How much was the doctor's bill for that visit (call)? 1 If bill not received, ask: BS — | | How much do you expect the doctor's bill to be for that visit (call)? 5. Is the doctor a general practitioner or a specialist? [JGeneral Practitioner [Specialist If “‘Specialist’* ask: What kind of specialist is Hel 81 82 Person No. Control HOME CARE PAGE For cach “*Yes "answer to la, Ask: Earlier in the interview you mentioned that —— needed help of some [prmnee TT ! 2 Not Yes kind here at home. | am going to read a list of different kinds of 1 Tb. Who helps ——? ! Does anyone else help — personal care some people need in the home. Please tell me if —— . needs help in any of the following ways. | I lo. Does —— need help - | in walki-g up stairs or getting from room to room? . . . } [7] Ne in dressing or putting on shoes?. . ............. : No T Does —— need help — \ with bathing (shaving) or other toilet activities? . . . . Co ! [J Ne in eating or having meals served in bed? . . . . . . 1 [| No ' Does —— need help — bs \ with chonging bandoges? «. + + vos vas Ce | = I LS | in receiving injections? . ... Lu... LL... oo ! [J Ne with other treatments? . . ................... I If “Yes,” ask: What kinds of treatment? ! i Specify _ | ] No LL] Bows =< need help y . in changing bed positions? .. .... | | No [= in exercising or physical therapy? . ............ 1 [1] Ne BN T in cutting toenails? ........... Creer [| No Yi Does —— get any OTHER help or care here at home?.. . .. ... ! If “Yes,” ask: What kinds of other help or care? | | I SPER cmc ——— i a. yt 1 No a the situation in the footnote space below. PERSON IS NOT RECEIVING CARE (All **No's'" ta question la), reconcile differences between answers in Q. 25 or 26¢ and Q. la above or describe 2. For what condition(s) does —— receive this help or care?—— Specify condition(s) 1 3. How long has —~ received help or care ot home? Mark one box: 0] 1 month oi less 1] Over 1 to 6 months 2] Over 6 to 12 months 3[ | Over 110 3 years 4[ Over 3105 years 5 [] Over 5 years 4. Because of ——’s health, must someone be in the house with him all of the time, part of the time, or only when providing the [LAI of the time needed help or care? ; 2[ | Part of the time Only when providing the needed help or care F P For cach person, other than a nurse, listed in 1b, ask: Sa.ls —— a nurse, a physical therapist, or some other kind of health worker? If “Nurse reported in Q. 1b or Sa, ask: 5b. Is the nurse that cores for —— a registered nurse, a practical nurse, or some other kind of nurse? (Determine the type(s) of person(s) providing the care in question | and mark appropriate box in cqlumn (1) of Table H.) FOOTNOTES: Page 34 Home Care Page—Con. TABLE H During the past two weeks About how many hours a day does — — receive help Is (relative, nur on about how many days or care from (relative, nurse, etc.)? paid for the: did = — receive help or care Type of persons providing care from (relative, nurse, ere}? (2) (3) (4) (1) Days xx Don’t know Hours 00 Less than 1 hour xX Don’t know 1 Yes ‘2 No NON-HEALTH . WORKERS A. 8] Related household members B. 1 [_] Related persons not in household C. 2[] Friend or neighbor D. 3[_] Other Specify HEALTH WORKERS E. a[ | Nurse — Registered F. 5] Nurse — Practical or other G. 6] Physical therapist H. 7] Other Specify INTERVIEWER: Mark the [] Person 65+ and “Yes” in column (4). Ask Q’s 6, 7, and 8. appropriate box before going to Q's 6-8. > [] Person 55-64 and “Yes"" in column (4). Ask Q's 7 and 8. [CJ All ““No’s’" in column (4) or only ‘A’ checked in column (1) of Table H. Skip to question 8. 6. Are any of these services paid for by Medicare? 1] Yes 2[ No X [] Don’t know 7a. Who pays (the remainder of the bill) for these services? b. Anyone else? 1] Self or family 2[_] Other relative or friend 3[] Health insurance a[_] Agency or organization (Visiting Nurses Association, etc.) 5] Welfare 6] Other — Specify [] Yes — Ask 8b 000 [_] No — Stop 80. During the past 12 months, has — — received any care of home from a nurse? BB BB a Ee aT. = ember al vimss b. During the past 12 months, ABOUT how many visits did a nurse make to core for — — ? FOOTNOTES: WASHINGTON USE Page 35 83 Ask for all persons 14 years of age and older: 28alxv [7] Under 13 years} Go to next 28a. Hos — — driven a motor vehicle during the past 12 months? xx [No person [IYes — Ask 28b b. How mony yeors hos ~ — been dri 1.1 001 Less than | year — Number of years R For persons 19 years old or aver, show who responded for {or was present during the asking of) Q. 5-28 of Responded tor sell aire] If persons responded for self, show whether entirely or partly. For persons under 19 show who responded R 1[] Responded for sel-pantly @'5-20 for them. If eligible respondent is “tot home™ buy did not reapond far self, enter the reasan ine footore. , Person ______ was respondent These next questions are about health insurance. We are interested in all kinds of health insurance which pays for MOST KINDS of illness. However, we do not want to include insurance which pays ONLY for accidents. 290. Is anyone in the family covered by a health insurance plan which pays all or part of a hospital bill? [Yes Ask b and ¢ INo—Go to 30a b. What is the name of the plan? — Record in Table H. I. c. Is-anyone in the family covered by any other health insurance plan which pays all or part of a hospital bill? [Yes Reask band c [INo—Complete Table H.1. for each plan reported 300. (Besides the — — plan you told me about) is anyone in the family covered by a health insurance plan which pays all or part of a surgeon's bill? [Yes—Ask b and ¢ _INo—Go to 31a b. What is the name of the plan? — Record in Table H.1. c. Is anyone in the family covered by ony other health insurance plan which pays all or part of a surgeon's bill? [ ]Yes—Reask b and ¢ [ |No—Complete Table H.I. for each plan reported 3lo. (Besides the — — plan you told me about) is anyone in the family covered by a health insurance plan which pays all or part of a doctor's bill for home calls or office visits? {Yes—Ask band c [{No—Go to 32a b. What is the name of the plan? c. Is anyone in the family covered by any other health insurance which pays all or part of a doctor's bill for home calls or office visits? r »s—Reask b and c ["INo—Complete Table H.I. for each plan reported 320. (Besides the — — plan you told me about) is anyone in the family covered by a deductible health insurance plan which pays some part of a bill for doctor visits or for hospiral or surgical care, after o certain amount has been paid by the family? [)Yes—Ask b and « { INo—Go to 33a b. What is the name of the plan? c. Is anyone in the family covered by any other deductible heolth insurance plan which pays some part of a bill for doctor visits or for hospital or surgical care ofter a certain amount has been paid by the family? [Yes--Reask b and ¢ [INo—Complete Table H.I. for each plan reported INTERVIEWER CH o Mark one box for each person Und. 65-Go to next person _— [7165 or over—Ask 33a 330. Is — — covered by that part of Social Security Medicare which pays for doctor visits; that is the Medicore plan for which he or | IYeudok some agency must poy $3.00 a month? No-Go to next person If person is covered by any insurance plan in Table H.1. ask for EACH plan: Tne Be Is this the (nome of plan) you teld me obeut before? x or No Go to next person FOOTNOTES: WASH. USE ONLY Type | Number | Coverage 7 2 . Plan Plans Head H s D Health Insurance TABLE H. I. Does this |Does this Does this | Does this plan [ Which members of the | If 2 or more memberd For each person 65+ covered by plan pay all |plan pay all [plan pay all |pay any part of | family are covered by [of family covered | this plan ask: orpartof a [or partof a [or part of a |a doctor's bill | (name of plan)? by this plan ask: hospital bill?| surgeon's doctor's bill | for home calls Is this (name of plan) which Name of Plan bill? for home or office visits| Circle column numbers | Are all of these covers — — a Social Security P calls or after a certain persons covered | Medicare plan? Zz office visits? amount has by the same policy? 8 been paid by a the family? - m © ® @ © ©) @ ® Covered: Pers. Pers. Pers. [Yes [Yes [I¥es—Go | [Yes 104s fy No. No. No. wel — | |= Fi A © Not covered: Cori line [Yes | [Yes | [Yes [Ne [CINe [CINe [No 1 23456 for each policy | [No [No [No Covered: [Yes Pers. Pers. Pers. [Yes [Yes [)Yes—Go | [Yes No. No. No. to6 1.23 3.6 [CINo—Filt B separate line | []Yes | [_]Yes | [Yes [Ne [Ne [CIN [CINe for each policy | [F]No [No [No Pers. Pers. Pers. [Yes [Yes [Yes [Yes Go [Yes 123 656 | Dne pun No. No. No. c Not covered separate line | [_JYes | [Yes | [Yes [No [No [No [No 1 2 3 456 for sack policy | [No [No [No Covered: [Yes Pers. Pers. Pers. [Yes [Yes [JYes—Go | (Yes 1 2445 % No. No. No. we] | R-f n3 Boe [INo—Fint D Not covered: separate tine | [Yes |[JYes | [Yes [No [No [No [No 1285 48% for each policy | []No [No [No Covered: Pers. Pers. Pers. [Yes [Yes [IYes—Go | [Yes 12% 4 Ed ClYes No. No. No. E to 6 Not covered: ~~~ =| CINE | EYes | [Yes [Yes [No [No [No [CINo 123456 for each policy | []No [No [No Covered: [Yes Pers. Pers. Pers. [Yes [Yes ¥erys [Yes 1 2 3 45 6 CNo-Filt No. No. No. BE" pPe2 28 i. F Not caverear separate line [dyes |[JYes |[[JYes [No [No [CIN [CINo 1 2 3456 foreach policy | [No [ [No | [No Covered: [Yes Pers. Pers. Pers. [Yes [Yes []Yes—Go | [Yes 123456 Ee No. No. No. to6| 000 |mmmm— mmm — = A . G Sool separate line | [Yes |[JYes |[JYes [Ne [Ne [No [CIN 1 23456 for each palley' | Finn [No [No Covered: [Yes Pers. Pers. Pers. Yes Yes Yes—Go es 1 23456 0. 0. o. ¥ Ever N N N oe TE Ee eR Ra i " °6 Not covered: separate line | [Yes |[]Yes |[[]Yes [No Ne [No [Ne NE for each policy | [No = ®@ []Und. 65—Go to next person [65 or over—Ask 33a ® []Und. 65-Go to next person []65 or over—Ask 33a ® [165 or over—Ask 33a [JUnd. 65—Go to next person ® [165 or over—Ask 33a []Und. 65—Go to next person © [JUnd. 65-Go to next person [165 or over—Ask 33a {_IYes—Askb [Yes—Ask b []Yes—Ask b [JYes—Ask b [INo—Go to next person [_INo—Go to next person | CINo—Go to next person [_INo—Go to next person Se Fel pyre a ir spy yrs ome SE Rem mem emg me [oe Te fr om fo pt. ed] Line | Line 1 Line Line | Line ' Line Line | Line | Line Line | Line Line No. 1 No. ! No. No. I No. No. No. 1 No. | No. No. | No. 1 [des | [Ives | [Yes [Ne ![Ne | [No Go to next person | 4 [Yes ! [Yes I [No | [Ne Go to next person [Yes ves ) CINe + [Ne 1 Go to next person [Yes Ne [Yes : [Yes [No 1 [Ne Go to next person | + [No [Yes—ask b [_JNo—Go to next person | 1 Line | Line | Line No. | No. | No. [Yes | Yes ! Caves [Ne 1[Ne 1 [Ne Go to next person WASH, USE ONLY WASH. USE ONLY WASH. USE ONL WASH. USE ONLY " WAS! Type Number | Coverage Coverage Type Number [Coverage Type Number [Coverage |! Type Number {Coverage I 0 ol of iT 0! of =. 0 of ol 0! o of Plan Plans Head Head Plan Plans Head Plan Plans Head Plan Plans Head Hi ws Ss s : S s DJ | D « Dj D 85 86 1f 17 years old or over, ask: b. Did — - finish the — — grade (year)? 1Und. 17 yrs.-next person Ask for all males 17 years old or over. 35a. Did — - ever serve in the Armed Forces of the United States? d. Was any of his service after January 31, 1955? (Yes (INo | Ask Stop [IDK 35¢ [C1¥es [No } Ask IDK J 35d Ask for all persons 17 years old or over. 360. Did — — work at any time last week or the week before?—For females add: Not counting work around the house. b. Even though — — did not wok during those 2 weeks, does he have a job or business? c. Wes he looking for work or on Jayoff from a job? d. Which — looking for work or on layoff from a job? 1[ JLooking dd z[ Layoff 3{_|Both | | full-time civilian job. | | | | r 7 Yo th Yi Ask for all persons with a **Yes' in 36a, 36b, or 36c. Fmployer only, questions 37a |_370. Who does (did) — - work fo. Braff through 37d apply to Industry this person's LAST b. What kind of business or industry is this? b. of Pvt pd. 3[ Own 1||Gov. Fed. a |Non-pd. dl 2[ Gov. Oth. s[ Nev. worked If person is under 17 years, or not in Labor Force (Q. 37 a-d blank) check *“Not in Labor Force." If in Labor Force (Q. 37 filled) refer to Question Se and make appropriate entry. af Not in Labor Force Corl 7 No work -loss days-in LF Go to next person | “1Work-loss days Go to 38a Earlier you said thot — — lost — — days from work during the past 2 weeks — (If self-employed, ask b; for other workers, ask a) 380. Was — — paid any wages by his employer for the days that he lost? 80. 10 |Yes—Ask [INo-Ask b JhaBelore | zl ater 1[ Before zl Ader yours, your = —'s etc.? (Show Card I) Include income from all sources such as wages, salaries, social security or retirement benefits, help from relatives, rents from property, and so forth. h. Did — — receive this income for these doys through a sick leave plan, loss-of-pay insurance, or some other way? h 1 [Sick leave plan 2[ Loss-of-pay insurance 3[ Other - Specity 39. Which of these income groups represents your total combined family income for the past 12 months — that is, 39. | Group 1A LB SE sJ6C e[F © U. S. GOVERNMENT PRINTING OFFICE : 1972 482-005/20 7 DAY Tre” Series 1. Series 2, VITAL AND HEALTH STATISTICS PUBLICATION SERIES Formerly Public Health Service Publication No. 1000 Programs and collection procedures.— Reports which describe the general programs of the National Center for Health Statistics and its offices and divisions, data collection methods used, definitions, and other material necessary for understanding the data. Data evaluation and methods research.— Studies of new statistical methodology including: experi- mental tests of new survey methods, studies of vital statistics collection methods, new analytical techniques, objective evaluations of reliability of collected data, contributions to statistical theory. Series 3, Analytical studies.—Reports presenting analytical or interpretive studies basedon vital and health Series 4. Series 10. Series 11. Series 12. Series 13. Series 14, Series 20. Series 21. Series 22, statistics, carrying the analysis further than the expository types of reports in the other series, Documents and committee veports.—Final reports of major committees concerned with vital anc health statistics, and documents such as recommended model vital registration laws and revised birth and death certificates. Data from the Health Interview Survev.— Statistics on illness, accidental injuries, disability, use of hospital, medical, dental, and other services, and other health-related topics, based on data collected in a continuing national household interview survey. Data from the Health Examination Survey.—Data from direct exarnination, testing, and measure- ment of national samples of the civilian, noninstitutional population provide the basis for two types of reports: (1) estimates of the medically defined prevalence of specific diseases in the United States and the distributions of fie population with respect to physical, physiological, and psycho- logical characteristics; and (2) lysis of relationships among the various measurements without reference to an explicit finite universe of persons. Data from the Institutional Population Surveys — Statistics relating tothe health characteristics of persons in institutions, and their medical, nursing, and personal care received, based on national samples of establishments providing these services and samples of the residents or patients, Data from the Hospital Discharge Survey.—Statistics relating to discharged patients in short-stay hospitals, based on a sample of patient records in a national sample of hospitals. Data on health resources: manpower and facilities. —Statistics on the numbers, geographic distri- bution, and characteristics of health resources including physicians, dentists, nurses, other health occupations, hospitals, nursing homes, and outpatient facilities. Data on mortality.—Various statistics on mortality other than as included in regular sunual or monthly reports—special analyses by cause of death, age, and other demographic variables, also geographic and time series analyses. Data on natality, marriage, and divorce,—Various statistics on natality, marriage, and divorce other than as included in regular annual or monthly reports—special analyses by demographic variables, also geographic and time series analyses, studies of fertility. Data from the National Natality and Mortality Surveys,— Statistics on characteristics of births and deaths not available from the vital records, based on sample surveys stemming from these records, including such topics as mortality by socioeconomic class, hospital experience in the last year of life, medical care during pregnancy, health insurance coverage, etc. Tor a list of titles of reports published in these series, write to: Office of Information National Center for Health Statistics Public Health Service, HSMHA Rockville, Md. 20852 2h de] Rr Add Eid © J o oN N (=) 3 7) I A ] 4 OQ c i I ho I - 8 4 11] 4 a] U.C. BERKELEY LIBRARIES MAA (pD21206052