tional Medical Care Utilization and Expenditure Survey, :valuation, of Item Nonresponse / the National Medical Care - Utilization and Expenditure Survey “ , Series A, Methodological Report No. 3 us. DEPOSITURY OCT 26 1987 Health Care Financing Administration Office of Research and Demonstrations Public Health Service National Center for Health Statistics ''. “PUBLIC HEALTH LIBRARY tn teea pein of “BERKELEY +# |, LIBRARY "UNIVERSITY ‘OF + CALIFORNIA _/ National Medical Care Utilization and Expenditure Survey The National Medical Care Utilization and Expenditure Survey (NMCUES) is a unique source of detailed national estimates on the utilization of and expenditures for various types of medical care. NMCUES is designed to be directly responsive to the continuing need for statistical information on health care expenditures associated with health services utilization for the entire U.S. population. NMCUES will produce comparable estimates over time for evaluation of the impact of legislation and programs on health status, costs, utilization, and illness-related behavior in the medical care delivery system. In addition to national estimates for the civilian noninstitutionalized population, it will also provide separate estimates for the Medicaid-eligible populations in four States. The first cycle of NMCUES, which covers calendar year 1980, was designed and conducted as a collaborative effort between the National Center for Health Statistics, Public Health Service, and the Office of Research and Demonstrations, Health Care Financing Administration. Data were obtained from three survey components. The first was a national house- hold survey and the second was a survey of Medicaid enrollees in four States (California, Michigan, Texas, and New York). Both of these components involved five interviews over a period of 15 months to obtain information on medical care utilization and expenditures and other health-related informa tion. The third component was an administrative records surve\ that verified the eligibility status of respondents for the Medi- care and Medicaid programs and supplemented the household data with claims data for the Medicare and Medicaid populations. Data collection was accomplished by Research Triangle Institute, Research Triangle Park, N.C., and its subcontractors, the National Opinion Research Center of the University of Chicago, Ill., and SysteMetrics, Inc., Berkeley, Calif., under Contract No. 233-79-2032. Co-Project Officers for the Survey were Robert R. Fuchsberg of the National Center for Health Statistics (NCHS) and Allen Dobson of the Health Care Financing Administration (HCFA). Robert A. Wright of NCHS and Larry Corder of HCFA also had major responsibilities. Daniel G. Horvitz of Research Triangle Institute was the Project Director primarily responsible for data collection, along with Associate Project Directors Esther Fleishman of the National Opinion Research Center, Robert H. Thornton of Research Triangle Institute, and James S. Lubalin of SysteMetrics, Inc. Barbara Mose1 of Research Triangle Institute was the Project Director primar- ily responsible for data processing. For sale by the Superintendent of Documents, U.S. Government Printing Office Washington, D.C. 20402 ''Evaluation of Item Nonresponse in the National Medical Care Utilization and Expenditure Survey Series A, Methodological Report No. 3 | U.S. DEPARTMENT OF HEALTH AND HUMAN SERVICES Published by Public Health Service National Center for Health Statistics October 1987 ''SHo97se us PY Copyright Information All material appearing in this report is in the public domain and may be reproduced or copied without permission; citation as to source, however, is appreciated. Suggested Citation National Center for Health Statistics, S. J. Ingels: Evaluation of item nonresponse in the National Medical Care Utilization and Expenditure Survey. National Medical Care Utilization and Expenditure Survey. Series A, Methodological Report No. 3. DHHS Pub. No. 87-20003. Public Health Service. Washington. U.S. Government Printing Office, Oct. 1987. . Library of Congress Cataloging-in-Publication Data Ingels, Steven J. Evaluation of item nonresponse in the National Medical Care Utilization and Expenditure Survey. (National Medical Care Utilization and Expenditure Survey. (Series). Series A. Methodological report ; no. 3) (DHHS pub. ; no. 87—20003) Bibliography: p. 1. National Medical Care Utilization and Expenditure Survey (U.S.)—Evaluation. 2. Medical care surveys—United States—Response rate—Evaluation. |. National Medical Care Utilization and Expenditure Survey (U.S.) Il. National Center for Health Statistics (U.S.) Ill. DHHS publication ; no. (PHS) 87-20003. IV. Title. [DNLM: 1. Fees and Charges—statistics. 2. Fees, Medical—United States—statistics. 3. Hospitals— utilization—United States—statistics. 4. Length of Stay— United States—statistics. WX 157 146e] RA410.7.154 1987 362.1'0973'021 86-600378 ISBN 0-8406-—0359-2 ''ntents 4 pi SY J Loe 7 Z ¢ CPit7 A / PReUIVEUISIMNGRY ce hl ee ee a Gee ees oe eee RE ee te eee se oe eS woe eee Sie ta sas 1 POSUCHONE SS. Se a a ee we Se i ee Se See ea ee See he EEG 3 IG GUIVOY . co Ek a ee POS fel. ek alates bee cee 2 Se ee eee, Soe: 3 ANG ota BASS. fk wage Se a TE ee te eee oe VR RS 8 ce ot a ee ee 3 Purposes and. Strategy:of-the Evaluation’. 2... 2 2 6. wk we Se ere ee a ee Ee fee ee a Se 3 Focus and Organization of the Evaluation ..... 2.2... . . ee 3 edical Visit, Hospital Stay, and Flat Fee Files .. 2... 2. ee 5 Medical Visit rile © 4. ee a we ee ee ae se ee 5 Bospital olay Bile... sa. bE a eee eget eg eee ee lb eee lens 5 Paves File Ss ee eS. ak ke ie OR ae So SR ee ee 5 lem Wonresponse-in:NMCGUEG:.. ee ee a Oa ae ee BT PS 6 odes for Nonresbonse 2. Bl. ee Bbc ee ec ta ge ob eo SB weak eels 6 Calculation-of Response Rate... 6 Pe ee ie a ee ER Be oe one te ee ee a ees Be £ Pattoms of Nonresponse. ... 2 6. ce ee a a a eee a a, EN es t Siipetatiors RIODIEMS). <9 ee ace ae ee Re a ee es BS Sa 12 Analysis of Charge and Source-of-Payment Variables ..........0. 0.000. eee ee ee 14 Analysis.of a-Charge Variable:so:. 0... 6. OE a ea, SO a 14 Analysis of a Source-of-Payment Variable ..........-.-. 2... ee ee ee ee eee 16 Soepe of Nohresponise .. i. 0. de Se Sa ek a Se ee a PEE SE SA 20 Overall Leveliof:Nonresponse:. 22. 6... ok i eS le ts PR, Hs N.S 20 Monrosponse AcrOss FleSis 332 ea ee a ee a ee ee ee es 21 Demographic Characteristics of the Item Nonresponse Population... 2... ....0.0.000 0000 eee eee eee ee 22 PIGBPOHOONU AGG . 0 22 Peqponaent EQUCANONM,.-.°. 922s. ee ea Bee ee oe ee ee ate woe i 22 income-of Reporting Unit 2... i. a ee Le ee EE ee be be ee ee ee Ee A 23 Medicare-Medicaid:Status,-... 2 ck eo. Ek ee eee Pe 2 EO i 23 BOUIN se le ee Ne ae al ee oo ge eS Spe ee Ae eee 24 Recommendations: 2 i es ee oe ee. RR Sn ea. OND Soe ieee ce 25 INGISCNONCOS< 5. ee. ek iste RS Res SA Ag Sn te se ae 29 PO PONOIGSS ce a ie ee RE ere ee ee de be LE LS SR te Ge eae Be ae 30 1: = Definitions of Terms Used'in This Report’: 2... ek ee ek ee Sale wae ee 31 ll. -Mohresponse Variables. 2... ke ee be wn eee, bean ek Eos 6 ec ce ete ee 0 lees 34 iil. Selections From the Questionnaire. 2.2.5. 6 ee ee ee ed whe ge be eee ere 8 37 List of Text Tables A. Responses to select medical provider visit questions in the 12-month Medical Provider Visit file: “How much, all or none?” by. “How muchidid:you pay? =... ... aa. OE Se eae se gon. i SR ae 9 B. Responses to select hospital stay questions in the 12-month Hospital Stay file: “Family paid all or none” by “Amount TAMMY Paid: ea eee i ee A ete ee ce a ey Bay . eT ee eS 11 C. Number of responses and percent distribution as reported in the 12-month Hospital Stay file to questions “What was the total charge for hospital stay?” and “Why was there no charge?” 2... 2... eee 14 D. Responses to select hospital stay questions in the 12-month Hospital Stay file: “What was the total charge for hospital stay? by “Why was there: fio: charge?” we ee et eee be te ee te ee 15 E. Responses to select hospital stay questions in the 12-month Hospital Stay file: “First source amount” by “first source”. . 17 ili ''Number of responses and percent distribution as reported in the 12-month Hospital Stay file to questions of first SOCONG.caRG Hind. SOuree-OlepayilgnLamOunms 7. =. . ee a es es Eee G. Cumulative nonresponse for charge questions in the 12-month Medical Provider Visit, Flat Fee, and Hospital Stay files, by-Maugolttme ano-ty pe OF ROMIGSPONSe S- 0 s. e e ee , e H. Percent of “don’t know” responses reported to select charge and source amount questions in the 12-month Hospital Stay, Flat Fee, and Medical Provider Visit files, by three age groups .................. 002. ee eee J. Percent-of “don’t know” responses reported to select charge and source amount questions in the 12-month Hospital Stay, Flat Fee, and Medical Provider Visit files, by years of education of respondent. ................... K. Percent of “don’t know” responses reported to select charge and source amount questions in the 12-month Hospital Stay, Flat Fee, and Medical Provider Visit files, by income category of reporting unit. ................... L. Percent of “don’t know” responses reported to select charge questions in the 12-month Hospital Stay, Flat Fee, and Medical-Provider- Visit files; by: Medicare-Medicaid coverage’: -. =. ee eS Text Figure Nonresponse codes used in the National Medical Care Utilization and Expenditure Survey: 1980 ............... iv Symbols --- Data available for fewer than 75 cases Category not applicable - Quantity zero 0.0 Quantity more than zero but less than 0.05 - 10—20-percent nonresponse = 21—40-percent nonresponse *** 41-100-percent nonresponse ''aluation of Item Nonresponse the National Medical Care ilization and Expenditure irvey Steven J. Ingels, National Opinion Research Center ‘xecutive Summary The National Medical Care Utilization and Expendi- are Survey (NMCUES) was undertaken to obtain accu- ate person-level data on the use and charges of health ‘are services from the U.S. civilian noninstitutionalized yopulation during 1980. The survey had three compo- yents—the national household component, the State Medicaid component, and the administrative records component—and utilized several different survey instru- ments, including a core questionnaire, a summary of responses, and various supplements. The focus of this evaluation is the core questionnaire, and the analysis is restricted to the national household component of the survey. More specifically, this evalua- tion is an intensive examination of three data files con- structed from household survey responses to several sec- tions of the core questionnaire: the Hospital Stay file, the Medical Visit file, and the Flat Fee file. These three files are included in the 12-month NMCUES data base, an intermediate stage in the construction of the NMCUES Public Use Data Tape, and were chosen for evaluation partly because they are rich in charge and source-of- payment variables, two questionnaire items associated with high (10 percent or more) nonresponse. The strategy of the evaluation is to identify patterns of item nonresponse, to determine the causes of item nonresponse, and to recommend ways that such nonre- sponse could be reduced. Five patterns of nonresponse are identified: I. nonresponse as high “don’t know’s,” in which a “don’t know” response category has been provided, and over 10 percent of the responses are “don’t know’s”; II. blanks (98’s) as surrogates for “don’t know,” in which a “don’t know” response category has not been provided, and the hypothesis is that high levels of blanks are proxies for “don’t know”; III. nonresponse as legitimate internal skips, in which a high proportion NOTE: This report was prepared by NORC (formerly the National Opinion Research Center) under contract (Contract No. 282—84—2109) with the National Center for Health Statistics. Hyman Bern and Brenda Spencer of the NORC Center for Computing and Information Systems gave valuable programming assistance. Sophie Ravin, Susan Campbell, and Jeff Hackett of NORC’s editorial service assisted with the manuscript preparation. Several members of the National Center for Health Statistics staff also contributed to the production of this report. In particular, Robert Wright, Michele Chyba, P. Ellen Parsons, and Andrew White offered guidance in the technical review, and Jan Schweitzer and Mary Olmsted provided editorial help. of item nonresponse appears to be due to conventions for coding and to editing legitimate internal skips within a question (for example, no legitimate nonresponse code is available for legitimate skips); IV. nonresponse as an artifact of question disaggregation, in which item nonresponse is manufactured by breaking a question or subquestion into multiple variables; and V. nonresponse as legitimate lack of data, in which, though high nonre- sponse is legitimate, a question simply fails to elicit any appreciable amount of valid response (for example, it fails to apply to more than a minuscule proportion of the respondent population). It is important to realize that item nonresponse in the NMCUES files cannot be taken at face value. That is, Some item nonresponse is spurious, as in patterns III and IV, where the appearance of nonresponse reflects editing conventions or vagaries of the data file construc- tion process. In the other patterns of nonresponse, how- ever, item nonresponse is genuine, whether it is to be assigned to problems in questionnaire format or wording, or to a genuine lack of knowledge on the part of respondents. This evaluation shows that even where item nonre- sponse is largely spurious, it may yet have pernicious effects. For example, if illegitimate blanks and legitimate skips are combined, it may be impossible to identify the missing data for some values, and this may affect adversely such processes as imputation. In particular, the critical charge variables needlessly lose precision, in that some instances where a value should have been imputed will be missed. This evaluation suggests several strategies for reduc- ing certain categories of nonresponse in future surveys. Despite the overall high quality of the NMCUES core questionnaire, there are certain format problems; for ex- ample, in some cases subquestions seem to be packed too tightly and to involve skip patterns of undue complex- ity. Source-of-payment and condition variables should also be reformatted, so that the point at which a series ceases to be applicable to a respondent may be clearly marked. Because fourth condition items and third source- of-payment items elicit virtually no data, it is question- able whether they should continue to be response options. In addition, many coding and data cleaning proce- dures could be improved, particularly through the more precise and consistent use of coding conventions. A ''universal definition should be provided for each data file, to facilitate person-level analysis of the event files. In the case of genuine item nonresponse—a problem for hospital-stay charge variables, for all groups of re- spondents, and, in particular, for Medicare or Medicaid recipients—maximization of data can be achieved best through the supplemental consultation of medical provid- ers and linkage to administrative records. ''itroduction The Survey The National Medical Care Utilization and Expendi- ture Survey (NMCUES) was undertaken in order to ob- tain accurate information on the use of and charges for health care services in the United States. The survey had three main components: the national household component, the State Medicaid component, and the administrative (Medicare-Medicaid) records component. For purposes of this questionnaire evalua- tion, the NMCUES survey component utilized is the National Household Survey, which is further limited to those respondents (17,123) classified as “key persons” (that is, the residents of sample dwelling units enumer- ated during round | and followed during subsequent rounds). The initial interviewing of households occurred in early 1980, followed by three or four additional inter- views, scheduled at intervals of approximately 3 months, throughout 1980 and early 1981. Information about medi- cal care utilization and charges for all reporting unit members was obtained during each interview. Survey instruments included a control card, to track respondents and reporting units; a core questionnaire, administered during each interview; a summary, to up- date information reported in previous interviews; and supplements to the core questionnaire that were adminis- tered during the first, third, and fifth interviews. This evaluation is concerned with the core questionnaire. For further details about the survey methodology, see Bonham (1983). The Data Base There were several stages in data-base construction, which culminated in the Public Use Data Tape. The data base utilized for purposes of this analysis is the 12-Month Files. These files were the basis for the NMCUES National Household Survey Analytic Files and the subsequent Public Use Data Tape, after additional reformatting, editing, recoding, and imputing. A user of the Public Use Data Tape would find substantial differences between those files and the 12-month data base selected for this analysis. The 12-Month Files were chosen for this analysis precisely because they permit examination of patterns of nonresponse that may have been obscured by the substantial cleaning and imputation _done when preparing the Public Use Data Tape. The NMCUES data are organized into both person files and event files. The event files aggregate data for all five rounds. The 12-month data base consists of 10 files, 3 of which—Medical Visit, Hospital Stay, and Flat Fee—are the special focus of this report. Purposes and Strategy of the Evaluation Included in this report are findings of an analysis of item nonresponse in the NMCUES core questionnaire. These findings concern the identification and selection of high nonresponse items, identification of the logically distinct types of nonresponse, assessment of the mag- nitude and scope of true nonresponse, and comparison of nonresponse patterns both within files and across files. Findings concerning the characteristics of the nonre- sponse population are also reported, when germane to the issue of item nonresponse. Finally, specific recom- mendations aimed at the improvement of the question- naire and the coding, cleaning, and management of the data collected are made. Focus and Organization of the Evaluation In addition to the focus of the questionnaire evalua- tion on the three files (Medical Visit, Hospital Stay, and Flat Fee), special emphasis is on two kinds of vari- ables—charge and source of payment. This special em- phasis reflects priorities of the National Center for Health Statistics (NCHS). Several generalizations may be made about the consequences of a primary focus on these areas. As will be seen later, charge and source-of-pay- ment variables prove to be almost always problematic on the basis of a criterion of a minimum of 10-percent nonresponse. Moreover, these are also the variables that, though plagued by spurious nonresponse (that is, nonre- sponse that is simply an artifact of editing conventions or the vagaries of data-base construction), display most of the genuine nonresponse in the questionnaire. Not only, then, is the charge for any given medical service ''a centrally important datum, but it is also perhaps the most frequent source of genuine nonresponse. The charge and source-of-payment variables are well represented in the Medical Visit, Hospital Stay, and Flat Fee files. A systematic review of all 12-Month Files demonstrated that all standard patterns of nonre- sponse were represented in these three files, and that concentration on these three files could offer the sharpest possible focus, yet in no way impair the capacity to generalize about item nonresponse, its causes, and its remedies with regard to the NMCUES core questionnaire as a whole. The first section of this report is an overview of the three files that were intensively reviewed in the course of the questionnaire evaluation. The second section depicts the five distinct types of item nonresponse that typically appeared in the NMCUES 12-Month Files. This examination of nonre- sponse types permits a distinction between genuine and spurious nonresponse and an examination of how each operates in the NMCUES data base. This distinction is important because nonresponse as it appears in the data frequencies cannot be taken at face value, and an understanding of its true extent and magnitude requires that the types of nonresponse be thoroughly understood. In addition, knowledge of the types of nonresponse is necessary for an understanding of the causes of nonre- sponse, and in turn for the task of recommending ways that item nonresponse might be overcome. In the next section, the nonresponse typology used to analyze a typical charge question and source- payment question. This analysis tests the adequacy the entire analytic framework and permits testing the hypotheses about nonresponse that are posited | the nonresponse typology. The scope of nonresponse is then examined by takir the problematic variables from the three event files (Ho: pital Stay, Medical Visit, and Flat Fee) and classifyin these variables in terms of the type, magnitude, an extent of nonresponse attached to each of them. Thi, section also includes comparisons of like variables be tween or across files. The following section analyzes the demographic characteristics of the persons with a high “don’t know” frequency, possibly pinpointing identifiable and “‘target- able” groups for special data collection and supplementa- tion efforts. Finally, the insights reported earlier concerning the causes of item nonresponse and the characteristics of the nonresponse population are applied to the task of making specific recommendations for improvement of the NMCUES questionnaire and the data-base construc- tion process. Definitions of terms used in this report are given in Appendix I. Variables are classified by type of nonre- sponse in Appendix II. Relevant selections from the NMCUES questionnaire are reproduced in Appendix III. ''edical Visit, Hospital Stay, and at Fee Files This section is a description of the three files that were the sources of data for the questionnaire evaluation. The number and kinds of events contained in the Medical Visit, Hospital Stay, and Flat Fee files will be detailed. Medical Visit File The Medical Visit file is an event-level file in which a NMCUES respondent may have zero, one, or more records. The Medical Visit file data were collected in three sections of the core questionnaire: Medical Provider Visit, Emergency Room Visit, and Hospital Outpatient Department Visit. Information collected in the these three sections covered the following seven items: place of visit, type of medical personnel seen, medical services received, medical procedures performed, medicai cundi- tion, charges, and sources of payment. Because the Medical Visit file is a composite of three sets of data (or subfiles) from separate questionnaire sections, it had to be decided whether to analyze the files combined or separately. Typically, any given vari- able in the Medical Visit file either was represented only in one of the three subfiles or was represented in two or more but with precisely the same content. This design—variables being either unique to a subfile or wholly analogous in content across subfiles—was evidence in support of making the Medical Visit file the unit of analysis, rather than treating each subfile separately. However, one exception was made: In some cases, variables that appeared in more than one file were alike in content but different in format. In such cases, a separate and comparative analysis of subfiles was conducted to test whether some formats were more effective than others. The 12—Month File showed 88,981 events for the Medical Visit file when restricted to household question- naire records (17,123 persons). The file was checked for duplicate records. Where duplicate records appeared, the first was selected, reducing the number to 87,142 events. Hospital Stay File Like the Medical Visit file, the Hospital Stay file is an event-level file. It contains data about each hospitali- zation—specifically, reason for the hospital stay, opera- tions and procedures performed, length of stay, charge, and source of payment. The file includes both hospital visits not requiring an overnight stay and overnight stays. In addition, the file contains admission and discharge dates, and data on charges from physicians who provided care during the hospital stay but billed separately from the hospital. A file number was calculated for this evalua- tion by purging duplicate visits and restricting the file to key persons in the household survey who were rep- resented in the Hospital file. This yielded 3,132 events. Flat Fee File A flat fee refers to a single charge for medical serv- ices, supplies, or series of visits. This charge pertains oniy to one person. There were 10 types of health care specified in the flat fee section of the core questionnaire for which a flat fee may have been paid (dental care, hospital stay, eye examination, surgical care, prescribed medicine, counseling, and so on). It was anticipated that the majority of the identified flat fees would be applicable to one of the 10 categories; for the remaining inapplicable flat fees, an “other” category was provided. Reference was made to flat fees in the charge ques- tions of the various sections (for example, Hospital Stay, Medical Provider Visit, Dental Visit) of the core ques- tionnaire. If the respondent reported that the charge was a flat fee, the interviewer completed the flat fee section and assigned a flat fee letter (an alpha code was sequen- tially assigned to each flat fee reported within a reporting unit) to both the flat fee page and the sections in the questionnaire that applied to the flat fee charge. In the flat fee section of the questionnaire, respon- dents were asked to provide more detailed information, ranging from the nature of the services (for example, orthodontia, surgical care, counseling) to who actually paid (outside source or family). A four-part question about medical charges and payment was included to determine both the total charge and the sources of payment. In order to calculate a base event number for the Flat Fee file, duplicate records were purged and analysis was restricted to events tied to persons who were both key persons in the household file (N = 17,123) and in the Flat Fee file. The resulting base number that resulted from this process was 3,521. ''Item Nonresponse in NMCUES Several issues pertaining to item nonresponse in the NMCUES data are addressed in this section. First is an explanation of how the high nonresponse items were identified and selected. The starting point in this analysis was to define the nonresponse problem by designating as problematic all variables that had 10-percent or higher nonresponse when all 99 codes (legitimate skips) were excluded. Also included in this section is a statement of the formulas that were employed in calculating the response rate. Next, five patterns or types of nonresponse in NMCUES are identified and illustrated. Finally, the related matter of skip-pattern problems is addressed. By classifying NMCUES nonresponse into five basic patterns, generalizations can be drawn about the causes of item nonresponse in the 12-Month Files. This classifi- cation system will later be used as a framework for recommendations specific to those causes of nonre- sponse. It is also a useful device for distinguishing genuine nonresponse (which comes from respondents not knowing or not revealing the information that the question seeks) from spurious nonresponse (which is an artifact of the conventions employed in editing and presenting the data). This approach will make clear, then, how the conven- tions for presentation of the NMCUES data for the 12- Month Files and for the NMCUES codebook (Research Triangle Institute, 1982) account for a substantial amount of apparent item nonresponse. It will also reveal how these same conventions complicate the task of under- standing the nature and extent of nonresponse. That complexity will be apparent particularly when the analy- sis moves from questions as they appear in the question- naire, to these same questions as they appear in the codebook, and finally, to their appearance as response frequencies in the data files. The transition from question- naire to codebook to data files is seldom a simple one. For example, a single question in the questionnaire might appear as several separate variables in the codebook and data file. It is also true, however, that a single variable in the codebook and data files might reflect separate (though identical or similar in content and for- mat) questions from distinct sections of the questionnaire, the responses having been aggregated. Furthermore, it sometimes happens that frequencies appear in the data file for a response code that is not listed for that variable in the codebook. The implications of such data base construction conventions for the analysis of NMCUES item nonresponse will be made clearer in this, and the following, sections. 6 Codes for Nonresponse The figure presents the nonresponse codes (or ‘“‘con sistency codes,” as they are called in the NMCUES documentation) used in NMCUES. Figure Nonresponse codes used in the National Medical Care Utilization and Expenditure Survey: 1980 Numeric Code conversion Description NK 91 Never know. Respondent doesn’t know now and never will; therefore, no attempt will be made to update the data later. Mlegible. This code is used only for those questions in which the response could not be determined by the key operator and the task leader. Not applicable. Don't know. This code indicates a written response by the interviewer indicating that the respondent did not know the answer. Out-of-range response. This code is used when the response or transcription ex- ceeds the specified field width or allowable value range (for example, a month should not appear as “13”). Multiple response. This is used if the re- spondent gave several answers to a ques- tion when the directions called for only one, and the multiple response could not be resolved. Refusal. The respondent refused to answer an item either by written statement or in the interview. Blank, illegitimate blank, or nonresponse. This code represents nonresponse in all cases other than those identified as legiti- mate nonresponse (see below). Legitimate nonresponse. This code is used when the respondent should not have an- swered the question and did not; that is, he or she was routed around an item. The data entry operator NEVER ENTERS this value. NA 93 DK 94 BD 95 MR 96 RE 97 BL 98 LS 99 SOURCE: Allen, D.. and Moser, B.: Keying and Verification Manual for HH Interview Questionnaire and Supplement # 1. NACUES report by Research Triangle Institute. Revised Apr. 11, 1980. ''It should be noted‘that not all nonresponse codes int to genuine nonresponse problems. For example, ide 99 indicates legitimate nonresponse, such as a ques- yn the interviewer should have skipped and did. It so should be noted that response frequencies utilizing ese codes give a necessary but ultimately superficial icture of nonresponse. Analyses later in this report emonstrate that nonresponse as coded cannot always ie taken at face value. Calculation of Response Rate For purposes of determining rates of item nonre- sponse, questions containing subquestions were broken down into their constituent parts, and all questions and subquestions were given equal consideration. Thus (and in accordance with the conventions of the NMCUES codebook) a single question that posed three subquestions was treated as three (or more) variables. For example, the single question, “Do you expect any source to reim- burse or pay you back? (Yes/No)” produced seven vari- ables for the calculation of nonresponse because of its subquestions: “Who will reimburse or pay you back? ENTER BELOW,” “Anyone else?” (with three sources possible), and “How much will (EACH SOURCE) pay?” (with three amounts possible). Given this definition of relevant variables, response was calculated for each, successively utilizing two for- mulas. In the first instance, a response rate was calculated using formula 1A, which is number of valid responses Response rate = : number of applicable respondents for any file in which a person was the unit of analysis. For those files where an event rather than a person was the unit of analysis, formula 1B was used, which is number of valid responses Response rate = 3 number of applicable events. Any variable with a response rate of 90 percent or lower (that is, a 10-percent or higher rate of nonresponse) was automatically defined as problematic or appropriate for examination for this evaluation and subjected to a second response-nonresponse calculation based on the following formulas which incorporate legitimate nonre- sponses or skips into the numerator. Formula 2A, number of number of valid responses + legitimate skips Response rate = number of applicable respondents, was employed where the variable used the person as the unit of analysis. Where event was the analytical unit, formula 2B was utilized: number of number of valid responses + legitimate skips Response rate = 3 number of applicable events. Thus response was recalculated after the 99 or Legiti- mate-nonresponse code had been incorporated into the numerator. The same procedure was used on the 93 or not-applicable code; however, it was so rarely used, it was analytically inconsequential, and only those vari- ables that were problematic (again using a cutoff criterion of 10 percent) on the basis of calculations employing formulas 2A and 2B were used in the nonresponse analysis. Patterns of Nonresponse A major preliminary to any serious analysis of nonre- sponse is sorting the nonresponse type, so that cases of genuine nonresponse can be identified and addressed. This sorting process is also the precondition of setting forth recommendations designed to minimize or remove the causes of item nonresponse in the NMCUES question- naire. Examination of the problematic variables (that is, those with a nonresponse rate of 10 percent or higher, after exclusion of legitimate skips) revealed five logically distinct, though sometimes overlapping, types of nonresponse: (1) Nonresponse as high “don’t know’s,” where a “don’t know” response category has been provided, and over 10 percent of the responses are “don’t know’s”; (II) Blanks (98’ s) as surrogates for “don’t know,” where a “don’t know” response category has not been pro- vided formally but “don’t know’s” are anticipated, and 98’s (illegitimate blank) are over 10 percent of the total expected response; (III) Nonresponse as legitimate internal skips, where a high proportion of 98’s appear because of conven- tions for coding and editing legitimate internal skips within a question (for example, no legitimate skip code available for the variable); (IV) Nonresponse as an artifact of question disaggrega- tion, where illegitimate blanks (98’s) are, in effect, manufactured by breaking a question or subquestion into multiple variables; (V) Nonresponse as legitimate lack of data, where, though high nonresponse is legitimate, a question fails to elicit any appreciable amount of valid response. These five nonresponse patterns aid in the sorting of genuine and spurious nonresponse. Nonresponse is genuine in nonresponse types I and II. But in nonresponse types III, IV, and V the appearance of high nonresponse is simply an artifact of coding or editing conventions. (I) Nonresponse as high “don’t know’ s” —Probably the simplest nonresponse pattern is that of high respon- ''dent use of the “don’t know” option. In this pattern, the question provides “don’t know” as a response cate- gory, and this is chosen in 10 percent or more cases. This pattern implicates a variety of question types, but it is especially characteristic of the charge variables, any of which could serve as an example of this type of nonresponse. Question 18 from the Hospital Stay file will serve here: 18. How much was the total charge for (DOCTOR) including any amounts that may be paid by health insurance, Medicare, Medicaid, or other sources? ae a IEG) Included with other charges ........ 01 (FF_— (22)) MOM HOW . wc bw. ow ease ip 94 (19) The response frequencies from this question can be used to gauge the level of “don’t know” nonresponse. Note that the “don’t know” category appears in two places in the following frequencies, under “total charge” and “reason for no total charge”—variables AS18 and AS 18-1, respectively. (These designations are variable names as they appear in the NMCUES codebook (Re- search Triangle Institute, 1982). They appear (usually parenthetically) throughout the report for two reasons. First, they are provided as a courtesy to the reader who wishes to refer to the documentation from the study—spe- cifically the referenced codebook and the data files, both of which use these designations. Second, they are often the most accurate and efficient way to refer to the items under discussion. References to question num- bers, for example, would not be adequate because a single question may comprise several variables (see pre- vious discussion) and because some questions appear in more than one section of the questionnaire and have different numbers in each. These variables are presented in Appendix II.) Although the codebook gives no guid- ance on this point, both “don’t know’s” have the same referent: The respondent does not know the total charge. AS18 and AS18—1 Response Frequencies Question and code Number Total charge (AS18) SE. BG es ee eS 2,048 We I gc Se a Vi eh ee ee 999 Se ee as se Sees a. FS Sa 3 4 ee One en ae ee 108 OSS (on Ol mnpe) Gs. SS SSS 2 OOS: (OGM DINING 6 55 628. Sik ee eS 935 99’s (legitimate nonresponse or skip) ............ - Reason for no total charge (AS18—1) RE mee SS et eS SS 2,048 we A cae cs ec Se cs ee 179 Gos. (ONC eBIORNED S.-i: ee - re Oe a es a ee es we 641 OES europe) Os... Se SS ay SS Ss ~ ar SRBC es or ch AS ay a eR I Gee ce 1 el a ee eee 238 99’s (legitimate nonresponse or skip) ............ 989 +) The two “don’t know’s” have, however, a differe genesis: The number reported for total charge (ASI appears to represent those 94’s written by interviewe: into the margin of the questionnaire and later code into “don’t know,” and the 94’s for “reason for n total charge” (AS18—1) represent those responses tha invoked the “don’t know” response (final response cate gory, question 18) given in the questionnaire. Adding together the “don’t know’s” of AS18 an AS18-1 and determining their proportion relative to th« total expected response in the file (108 + 641/2,048 reveals that 36.6 is the percent of “don’t know” re- sponses. In all cases where this procedure produces a 10-percent or higher level of “don’t know’s,” the next step in the analysis is to investigate the cause of nonre- sponse. This investigation is accomplished by a systema- tic consideration of the clarity of the question, the format and logic of the questionnaire, and, if these are not problematic, the characteristics of the nonresponse population. (II) Illegitimate blanks (98's) as surrogates for “don’t know” —There are a few instances—for example, in the case of charge questions such as “amount family paid” (HS11, MVFFP—2)—where the printed question in the questionnaire does not offer a “don’t know” re- sponse category. Because charge information is some- thing that respondents often do not know, a fairly high level of “don’t know” response for such questions was anticipated. Therefore, the interviewers’ question-by- question specifications (NORC, 1979) indicate that a “don’t know” could be written in the margin of the questionnaire when such a category is not provided. Even with these write-in “don’t know’s,” a substantial (that is, over 10-percent) pool of illegitimate blanks remains, and these are hypothesized to be proxies for “don’t know.” It may be surmised, then, that the illegiti- mate blanks (98’s) would greatly diminish if an explicit “don’t know” response category was provided. A charge variable from the Medical Visit file (MVFFP-—2) will serve as an example of the second type of nonresponse: 10. How much of the (CHARGE) charge for the visit did or will you (or your family) pay? Pariah te se Se Total: Cheg@: cs hs Soeaes. F ee, 01 NenO@. = SS a ee 00 (C/BOX) Because this is a disaggregated question, the frequen- cies cannot be taken at their face value. One effect of splitting the question into multiple variables is that the number of 98’s is artificially inflated (a detailed look at nonresponse as a function of question disaggrega- tion appears later); therefore, the 98’s due to disaggrega- tion must be subtracted from the total 98’s and from the total expected response. Table A shows the distinction between genuine and spurious (artifact of disaggregation) 98's. Thus, of the 58,880 illegitimate blanks (98’s) in evidence, 51,562 are artifacts of disaggregation; and ''Table A Responses to select medical provider visit questions in the 12-month Medical Provider Visit file: “How much, all or none?” by “How much did you pay?” Response (with code) to question “How much, all or none?” Family paid Illegitimate Legitimate Response (with code) to question Percent None All Don't know Refusal blank skip “How much did you pay?” distribution Number (00) (01) (94) (97) (98) (99) Percent distribution AOle eS 3 RS PS. 100.0 17 42.2 0.0 0.0 11.6 29.1 Number of responses PUYRSOOHSSS «0. Ee SS ee eS beg 87,142 14,904 36,747 14 1 10,126 25,350 OR PS ee cg a ea. a 0.0 1 - - - - 1 - NONZerO aMOUNnt a. ee, a 9.1 7,885 22 46 - 1 - 7,816 Meover knew (91) 0 eo RA ee. 0.0 2 - - - - 2 - Bon Chow (94)s25° ses. doy oo SEP ee 3.3 2,835 15 “ 1 - 2,815 - Wout of. range (9S) =o diese ee cae 0.0 3 1 1 - - 1 - MISSAL (OT): goo. kk) nec hcccerels Ghee aa 0.0 2 - - - - 2 - Illegitimate blank (98). ................ 67.6 58,880 14,866 36,696 13 - 7,305 - Legitimate skip\(99). 00. .5 So OPE 20.1 17,534 - - - - - 17,534 7,305 (10.5 percent of the total file after the legitimate skip (99’s)—respondents legitimately filtered out of the question—have been subtracted) are unexplained. When the 2,815 “don’t know’s” from the question “How much did you pay?” (MVFFP-2) are added to the 7,305 true illegitimate blanks, nonresponse for the question runs at 14.5 percent. In any event, it seems most plausible to suppose that the 7,305 true 98’s are largely to be explained as surrogates for “don’t know.” However, this example is a case where the “don’t know” option, though underused, and not a response category within the questionnaire, was at least mentioned in the question-by-question specifications to interview- ers. In other cases, there was no such mention in the interviewer instructions. As an example, after subtraction of the legitimate nonresponse (that is, those who an- swered “no” to the filter in FF6, which asked whether there were any visits covered by the charge before Janu- ary 1, 1980, and are thus classified as 99’s) for the question “How many visits did (person) have to the (doctor/dentist/medical provider) before January 1, 1980?” (FF6A in the Flat Fee file), response distribution is Percent Response Number _ distribution All responses... 2. ee 1,245 100.0 Valid responses ............. 477 38.3 DONT KNOW 4:25 ce Bd. eS is St 64 5.1 Out of range............... 2 0.0 BiG. he ee a ee SS 702 . 56.4 Although certainty is not possible in this matter, it seems plausible to make several suppositions here. The actual rate of “don’t know’s” (5 percent) is very low, suggesting that not all interviewers may have taken time to write in uninstructed and unprovided “don’t know” responses. Given the large number of unexplained illegitimate blanks, it would seem logical to hypothesize that, in many cases, these are surrogates for “don’t know.” Ultimately, something along the lines of a split-ballot experiment to test the effects of providing an explicit “don’t know” response category for the type II nonre- sponse variables is needed. It should be noted that there are items in the Flat Fee file that run counter to the expectations of this hypothesis. This observation is especially true for a ques- tion about total charges for flat fees (FF2 in the Flat Fee file), in which a “don’t know” category is provided: 2. | Whatwas the total amount of the charges, f21 8 including any amount that may be paid Don'tKnow 94 | |by health insurance, Medicare, Medicaid, or other sources? While 98 and “don’t know” (94) nonresponse codes are slightly under the 10-percent cutoff point for categori- zation of the variable as problematic, the proportion of blanks to “don’t know’s” is surprising. There are 3,187 source amounts, 24 “don’t know’s” (0.8 percent), and 310 98’s (8.8 percent). It is indeed unexpected that there should be so few “don’t know’s” and so many illegitimate blanks for an item that included an explicit “don’t know” response category. (III) Nonresponse as legitimate internal skips—The NMCUES machine-edit specifications reserved the code 99 for legitimate skips but applied it primarily to skips between questions. Yet, in many instances there are legitimate skips within a question, particularly where the question comprises subquestions. This means that the code 98 has been used in two radically different ways: (a) The item was left blank; that is, the respondent- was obliged to supply information, and did not do so; or (b) the item was properly skipped (for example, the respondent did not have a second source of payment or a third condition). When this pattern is operative, 9 ''there is an inflation of the level of nonresponse and a fusing of legitimate with illegitimate nonresponse. This pattern pervades the source-of-payment series and also the condition variables. A simple but typical example is the Surgical Operations question: 6. Were any operations performed on (PERSON) during this stay in the [hospital/nursing home]? Yes 01(A) No 02(7) A. What was the name of the operation? IF NAME OF OPERATION IS NOT KNOWN, DESCRIBE WHAT WAS DONE. Were there any other operations during this stay? Name: Lae] We ee EL Name: ue er eee | | | The number of responses for this question series is small (1,208) because many respondents skipped it appropri- ately. Those in the series reveal the following pattern: Operation and Response Number First operation (HS6A—A) Valid responses 1,125 illegitimate bianks (98's)... 56 Second operation (HS6A-1A) Valid responses 160 Megitimate-blanks (98's)... -. ge 1,021 Third operation (HS6A-—2A) Valid responses 25 Illegitimate blanks (98’s)............. 1,156 What is presumably happening is that each successive operation applies to a smaller portion of the respondent population. It may be that only 160 of the 1,125 persons with a first operation also had a second operation, and that the 965 98’s reported for “second operation” are justified. These 98’s should be recoded ultimately to appear as 99’s or some alternative legitimate skip code. What appears to be a pattern of nonresponse, coded as illegitimate blank, is in fact largely legitimate internal skip. Although the account in the previous paragraph is largely the case, it is not wholly or unqualifiedly so, and entering the qualification is instructive as to why the use of 98’s in the sense just described is especially pernicious: Not all of these 98’s are 99’s. In the variable “first operation,” for example, approximately 5 percent of the response was in the form of true illegitimate blanks—98’s that should be 98’s. If this pattern persisted, illegitimate blanks would accrue at the rate of 5 percent of the valid response for second dnd for third operations, and the cumulative total of illegitimate blanks for the third operation might be estimated at 65, with the number of 98’s that represent a legitimate skip equal to 1,091. 10 In other words, though 98’s as legitimate skips withii a linked series form the dominant pattern, they are mixec with a certain number of true 98’s. These true 98’: in some cases might be very roughly estimated, but there is no sure way of separating the legitimate from the illegitimate 98-coded nonresponse. Because it would make no practical difference if all these 98’s were recoded to 99, the implications for a condition or operations series are small. However, in the case of a variable subject to imputation for missing data, such as source-of-payment variables, it is desirable to have an effective means to distinguish proper skips from truly missing data. (IV) Nonresponse as an artifact of question disaggre- gation—Several of the examples of nonresponse that have been presented have been disaggregated questions, but the discussion has yet to focus on the effects of disaggregation in itself. Questions that appear as a single entity in the questionnaire but are made up of multiple response categories, subquestions, or both, typically are disaggregated in the codebook. This procedure some- times leads to a spurious increase in the illegitimate blank (98) category of nonresponse—when blanks are recorded for one part of the question, and the response is recorded in the other. A common example of nonre- sponse as an artifact of question disaggregation is pro- vided by responses to question 11 in the Hospital Stay file: 11. How much of the (CHARGE) charge for the stay did or will you (or your family) pay? Parte Gus Se % Tetst Ghamnge. 3 et 01 Re eS 6a, a ae 00 (C BOX) A crosstabulation of “amount family paid” (HS11) with “family paid all or none” (HS11—1) shows the spurious nonresponse generated by disaggregation (Table B). Question 11 may be answered with a partial amount, “family paid none,” or “family paid total.” When the question is split into HS11 and HS1 1-1, 98’s (illegitimate blanks) are used to mark the valid responses for “family paid none” in HS11, and for “family paid total” in HS11-1. Again, some of the 98’s are true 98’s: Note that 406 responses appear as 98’s for both HS11 and HS11-1. The fact that many times 98’s are simply ar- tifacts of question disaggregation is more support for the notion that 98’s in the response frequencies cannot be taken at face value. As always, however, it cannot be forgotten that entangled with the artificial 98’s are genuine ones. (V) Nonresponse as legitimate lack of data—There were some questions in NMCUES that applied to virtu- ally no one in the respondent population. In such in- stances, the question failed to elicit data, and legitimate nonresponse approached or reached 100 percent. Such questions typically appeared as the last or next-to-last elements in a series, such as third source of payment, ''Table B Responses to select hospital stay questions in the 12-month Hospital Stay file: “Family paid all or none” by “Amount family paid” Response (with code) to question “Family paid all or none?” Family Family Don't Illegitimate Legitimate Response (with code) to question of paid none ___spaid total know blank skip “Amount family paid” Total (00) (01) (94) (98) (99) PULTGSPONSOS ae SS ee 3,132 1,204 178 2 799 949 WEN NOONE Se a egos Po. ee es 590 2 1 - - 587 Dormiieiow:(94) 57 oo i. Se ee od PS eee oe 394 1 - 1 392 - Reuse ($7): 2c Paes es ES re See 1 - - - 1 - iNegitimate ‘blank (98): 2 sg... eyes ee ER eee eee 1,785 "4,201 "177 1 2406 - Gegitimate skip (99)... ns ee ee 362 - - - - 362 ‘Spurious nonresponse, an artifact of disaggregation. ?Genuine nonresponse, as a surrogate for “don’t know.” or third or fourth medical condition. This pattern may number of be illustrated by a series in the Flat Fee file (FF4B): number of legitimate skips valid responses + (99’s) FF4: How muchwill (SOURCE) reimburse or pay you back? Source Number First source (FFB) Vand responee: >... kPa ae eS. 289 Dent agow.. o>. ase Pe SSE ies 229 Second source (FFB-1) Valid response... ..............00.4 24 Dent know = 32. 2 Pes ee PR es 42 Third source (FFB—2) Valid responee ood A ww, - Dont KNOW 325s ose ct ee pes «ees 1 ‘Hetal-cnarge.pald= 22 Se. SS, 01 Pasial or none paid. ps5. Se ite 2 ES 02 (C BOX) By the final source of payment, no valid responses were recorded, although one valid response could have been expected. The problem then is deciding whether questions that are not used are worth retaining. However, before turning to this issue, it is worth investigating a little more deeply the problem of unused final data categories in source-of-payment and in condition variables. The series-linked variables (for example, first, sec- ond, and third source of payment) are subject to two problems. First, there is no way for the interviewer to indicate the terminus of the series when the alternatives cease to be applicable to a given respondent; hence, blanks are left (indicated by the illegitimate nonresponse code, 98) for what often are legitimate skips. Second, the last variable in such a series is usually unused or nearly empty of data; that is, it applies to a small propor- tion of the respondent population and consequently elicits virtually no information. The scope of the problem is outlined in the following discussion, which presents the adjusted proportions of data obtained from use of final data categories for series-linked variables, such as source of payment. The response rate calculation is based on formula 2B: Response rate = : number of applicable events Source-of-payment variables appear in all three of the files used in this evaluation (Medical Visit, Hospital Stay, and Flat Fee) and have a common structure across files. They are preceded by a “yes” or “no” filter question. Those who answer “yes” to the filter are routed to the first source of payment, and asked to name first the source, then the amount. They continue to a second source of payment, and then a third. Two distinct source- of-payment questions follow this structure. The first question is “Who will reimburse or pay you back?”; the followup question is “Who else paid or will pay any part of this charge?” Response frequencies for the final category for each of the two types of source-of-pay- ment question are as follows. The first type—“Who will reimburse or pay you back?”—appears in four sections: hospital stay; physician charge in the course of, but billed separately from, hospi- tal stay; medical provider visit; and flat fee. The percent of valid responses for the third source of payment are as follows: Percent of file Variable for the third source of payment: with valid responses “Who will reimburse or pay you back?” (excluding 99's) Hospital stay (HS12B-2) ........... 0.50 Physician charge (AS20B-2) ......... - Medical provider (MVSA-3).......... 0.04 Flat fee (FF4B-2) ...........004. For all four instances of source-of-payment variables of this form, the number of valid response data collected amounts to less than | percent of the total file number, after all 99’s have been deducted from the calculation. For the second type of source-of-payment question— “Who else paid or will pay any part of the charge?” —the valid responses for the third source of payment are out- lined as follows: ''Percent of file Variable for the third source of payment: with valid responses ‘Who else paid or will pay any part ofthe charge?” (excluding 99’s) Hospital stay (HS13B-2) ........... 0.54 Physician charge (AS21B-2) ......... 0.06 Medical provider (MVSC-3).......... 0.10 Reet SOC ian ee Se ee 0.14 Again, all third source-of-payment response categories contain less than 1 percent valid responses, measured against the file number with legitimate skips (99’s) excluded. In each of the two instances, the most data was collected for hospital stay, but even this is minuscule. In addition to source-of-payment variables, certain other variables are series-linked, specifically, the condi- tion variables in the Hospital Stay and Medical Visit files, and the operations variable in the Hospital Stay file. The structure of these variables differs somewhat from file to file. In the Hospital Stay section of the core questionnaire, respondents were asked to name the conditions (a maximum of four) for which they were hospitalized. Then, births were specially noted, and there was a filter for “normal delivery” (“yes” or “no”); those answering “no” were asked to specify up to four abnormal conditions. This question was followed by a second filter, “Was the baby normal at birth?,” with those answering negatively being routed to the same four condi- tion blanks. A “yes” or “no” filter question begins the condition series in the Medical Visit file: “Was this for any specific condition?” Those who answered “yes” to the filter were then asked about the conditions treated during the visit (a maximum of four conditions). However, these same four condition blanks are potentially affected by an im- mediately subsequent additional filter question, “Did pro- vider discover any condition?” An affirmative answer to this question led back to the condition blanks. Thus, the box with four condition blanks contains the sum of conditions that prompted a visit plus conditions dis- covered in the course of a medical visit. As with the source-of-payment variables, 99 is not an available re- sponse category, and the interviewer has no way to indicate the end of the series for a particular respondent. Again, 98’s mark blanks that usually are legitimate skips of the latter parts of a series. The valid responses collected by the final elements of these series are indicated as follows: Percent of file with valid responses Variable (excluding 99's) File name, fourth condition variable, and third operation variable Hospital stay—conditions (HS5-3A) ..... 0.06 Hospital stay—newborn, abnormal conditions (FeoO a ee SS - Medical provider—conditions (MVF-CN4) . . 0.42 7 Third-operation variable: “What was the name of the operation?” Hospital stay—name of operations (HGR) SS eS 2.00 12 Thus, the proportion of valid responses, as a percent of the total file number, with 99’s excluded, never ex- ceeds one-half of | percent for a fourth condition vari- able. For the “name of operations” variable, where the series ends with the third response option, the proportion of valid response rises to 2.0 percent. Although there is some variation in the form such questions can take and in the amount of data they elicit, in all cases both third sources of payment and fourth medical conditions elicit an extremely small number of valid responses—less than 1 percent of the adjusted file number in the Medical Visit, Hospital Stay, and Flat Fee files. Because an enormously high level of nonre- sponse would appear to be quite justified, given that third sources of payment and fourth conditions are rare within the respondent population, a serious reconsidera- tion of the inclusion of third sources of payment and fourth medical conditions may be in order for future NMCUES. Skip-Pattern Problems Of the legitimate skips coded as illegitimate blanks, one nonresponse pattern was identified as the spurious one. However, it should be remembered that not all skips are legitimate, and that illegitimate skips are often a source of high item nonresponse. Although it is not apparent that ambiguous or highly complex skip formats intrinsically constitute a pervasive pattern of high nonre- sponse (over 10 percent) in the NMCUES data, there is evidence that they may be a problem to some degree in the more complex and highly ramified questions. An example is question 5 in the Medical Visit section of the core questionnaire (see Appendix III). Included in this question is an instruction, “Code all that apply.” It consists of 15 separate variables, and each time some- thing applies, it can implicate from 2 to 8 of the variables. The question is designed efficiently and separated into a sensible series of subquestions. Nevertheless, it is still questionable as to whether a format of such compres- sion and complexity works well in practice. There is an indication from inconsistencies in the data that some interviewers find it difficult to follow the indicated pattern of skips. As an example, the number of valid responses for the first condition (69,768) actually exceeded the sum (68,068) of its three filters (MVFDIAG, 57,216; MVFSPEC Yes, 10,143; and MVFDISC Yes, 709). Crosstabulations show skip- pattern inconsistencies, though usually of a magnitude of less than 5 percent of the expected number of events for the variable. Occasionally, however, a discrepancy is more substantial. For example, those who answered “yes” to “general checkup” (MVF-GEN) were routed to “checkup for specific condition, yes or no” (MVFSPEC) (see question 5, Medical Visit; reproduced in Appendix III). Of the 9,877 responses to the former, 1,156 (or 11.7 percent) appeared as blanks for the latter. ''Questions of this general type might profit by simplifying the skip-pattern format. Another sort of format that may contribute to illegiti- mate skips is found in the Hospital Outpatient Department Visit section of the questionnaire, which was incorpo- rated into the Medical Visit file for the 12-month NMCUES data base. Difficulties in obtaining a visit-by- visit count of events have prevented quantification of the extent of this problem, but its nature may readily be described. The Hospital Outpatient Department Visit section of the questionnaire departed from the format of the Medical Visit section of the questionnaire, where each medical visit was self-contained. In the Outpatient sec- tion, questions regarding a series of visits (A-E) were spread over a page and a quarter of the questionnaire, and this format appeared three times before the end of the question series had been reached. Because a re- spondent could have reached the end of the relevant questions for visit A fairly early in the series, the entire subsequent pages could have been left empty. When filling in visit B, particularly if a respondent was return- ing to the section after having been routed out (as for the Flat Fee question), it would be easy for interviewers to lose their place and accidentally fill in visit B informa- tion on the empty visit A column. It is not clear how often such errors occurred; however, alternative formats that minimize or eliminate this kind of error should be considered. 1 ''Analysis of Charge and Source-of-Payment Variables This section presents an analysis of one typical charge question series and one typical source-of-payment series, and the structures and relationships to item nonresponse of each. These variables illustrate the four most common of the five nonresponse types: I (nonresponse as high “don’t know’s”), III (nonresponse as legitimate internal skips), IV (nonresponse as an artifact of question dis- aggregation), and V (nonresponse as legitimate lack of data). The variables selected for analysis are typical of charge and source-of-payment variables, the type of ques- tions most likely to be plagued with genuine nonresponse. The need to disentangle genuine from spurious nonre- sponse will become apparent. This section will also pre- sent a clearer picture of the relationship of the identified nonresponse problems to the question format and the structure of the data presentation. Analysis of a Charge Variable A series of questions on hospital charges (HS—10) appears in the hospital stay section of the questionnaire as a single complex question. It is an excellent example of the eliciting of a large number of possible skip patterns from an array of response categories, with some questions containing subquestions. Respondents who reported a dollar amount for the charge, and those who responded with “don’t know,” were routed to question 11. Those who indicated “no charge” were routed to question 10, subquestion A (that is, HS10—A). IF STILL IN HOSPITAL, GO TO NEXT HOSPITAL STAY OR NEXT SECTION 10. How much was the total [hospital/nursing home] charge for this Stay, including any amounts that may be paid by health insur- ance, Medicare, Medicaid or other source? (Include any charges for [X-rays/laboratory tests/diagnostic procedures], but do not include separate charges for doctors or surgeons. $ (11) No Charge G2" or ees (A) Included with other charges ...... 03 (FF_(14)__ FOR NEWBORNS ONLY: Included in mother’s bill (Parean ess: Od ee (15) Don’t know Sex, -. .. (11) A. Why was there no charge for this hospital stay? Welfare/MedicaidPaid 01 ... . 25. i) Included with other charges ...... 02 (FF) Free from provider OFS ee (13) Other source(s) will pay ........ 04 (13A) FOR NEWBORNS ONLY: Included in mother’s Ol {Parmon Fa ets es 06 (15) 14 The codes are HS10 Total charge (in dollars) HS10-1 Reason for no total charge 02 No charge 03 Included with other charges 04 For newborns only: included in mother’s bill 94 Don’t know HS10A Reason for no charge 01 Welfare or Medicaid paid 02 Included with other charges 03 Free from provider 04 Other sources will pay 06 For newborns only: included in mother’s bill As can be seen from the frequencies (Table C), and from the excerpt from the codebook, question 10 is a disaggregated question; that is, it is disaggregated into two variables: HS10—“How much was the total hospital charge?” = dollar amount; and HS10—1—‘Reason for no total charge” = “No charge,” “Included with other charges,” “Included in mother’s bill,” “Don’t know.” Each of these is independent and sums to the total file Table C Number of responses and percent distribution as reported in the 12-month Hospital Stay file to questions “What was the total charge for hospital stay?” and “Why was there no charge?” Percent Number _ distribution Responses (with codes) to “total charge” and to “no charge” questions Total charge (question HS10) All réspofiees=—. 3,132 100.0 Validstesnonse= os 2 os Gece 1,218 38.9 Adjusted total nonresponse? ......... 1,912 61.0 Legitimate nonresponse (99). ....... - - Hegme (92) | Se eee es - - Don EKNOWS (94) os ee 56 1.8 Out Of anda (95) Sass es - - Multiple response (96) ........... - . Metusalieng: So ee - - lilegitimate blank (98) ........... 1,856 59.3 Not applicable (93)... 2 ee 2 0.0 No total charge (question HS10—1) Ailg@espoliaes \.25 3s 2 en 2 6. ange 100.0 Weliaresponee- a ae 537 47.1 Adjusted total nonresponse ......... 1,433 27 Legitimate nonresponse (99). ....... 1,162 37.1 WONT MNOW (04). 28 ee Ce 1,094 34.9 Illegitimate blank (98) ........... 339 10.8 ‘Excludes adjusted total ncnresponse. Excludes legitimate nonresponse (93, 99). ''event number (3,132). In turn, a third element appears in the series (HSI0A), a subquestion to which respond- ents are routed who chose “no charge” (response category 02 of HS10-1). It may be expected, then, that HS10 and HS10A may be reaggregated for purposes of pictur- ing the overall nonresponse pattern for question 10. De- termination of the level and pattern of nonresponse is in fact contingent on such reaggregation for two reasons. First, some of the “illegitimate nonresponse” (coded 98) is merely an artifact of the disaggregation process. (To sum to the total number in the file, each of the two disaggregates must count, under some nonresponse code, the responses or coded nonresponses belonging to the other disaggregate.) Secondly, the principal nonresponse category (“don’t know”) in HS10-1 has as its proper referent the variable HS10. The frequencies, however, display puzzling inconsis- tencies, thus resisting such reaggregation. The response categories for question 10 (“How much was the total hospital charge?”) are structured to accept only one re- sponse, not multiple responses. Thus, all who answer with a “charge” amount (HS10) are precluded from answering “no charge” (02), “included with other charges” (03), “in mother’s bill” (04), or “don’t know” (94) (HS10—1). When the. question is broken into two, then, the 1,218 recorded charges in HS10 should appear as 1,218 legitimate skips in HS10-1. For the second half only of this type of disaggregated question, the machine-edit instructions for NNMCUES provided a code of 99 (legitimate skip). (For the first half, 98 is used as the disaggregation marker code.) However, note that there are 1,162 legitimate skips in HS10-1, not 1,218. Instead of the expected match, where valid responses in HS10 precisely equal 99’s in HS10~-1, there is a discrepancy of 56. Because of such inconsistencies in the data, it is not possible to know just what is happening in the question unless the constituent variables are crosstabulated. When reaggregation of Hospital Stay question 10 is attempted, another puzzling feature appears. The code for “don’t know” (94) appears only once in the question as a response category. Yet the “don’t know” category appears twice in the question 10 data, separately, and with strikingly different response frequencies, for each disaggregate (HS10 and HS10~1). In the codebook, the “don’t know” response category is offered as a choice for respondents only once, assigned to HS10-1. The codebook presentation of categories seems to support the interpretation that 1,094 “don’t know reason for no total charge” responses were given for HS10—1, utiliz- ing the response category in the questionnaire; and 56 “don’t know’s” were given for HS10, presumably written into the margins of the questionnaire by interviewers. However, it is unlikely that charge figures (HS10) elicit such a low level of “don’t know’s,” and that “reason for no total charge” (HS10-1) responses elicit such a high level. The real answer to this puzzle is that there is only one type of “don’t know” response, namely, “don’t know total charge.” The small number of 94’s (don’t know) in HS10 support this, as do the large number of “don’t know’s” in HS10-1, which really belong in HS10. The true level of nonresponse for the “total charge” item (HS10) is the sum of the “don’t know’s” from HS10 and HS10—1. Thus, the nonresponse levels indicated by the disaggregates HS10 and HS10~—1 are wholly mis- leading. The “don’t know” rate for HS10 is not 1.8 percent, but 36.7 percent. In summary, understanding nonresponse requires reaggregating the disaggregated questions. The disaggre- gations can be confusing and even misleading, and that confusion is magnified by inconsistencies in the data that can best be resolved or understood through crosstabu- lation of the disaggregates. Table D presents a crosstabu- lation of HS10 (total charge for hospital stay) with HS10-1 (reason for no charge). The results of this crosstabulation show that although most valid HS10 total charge responses became 99’s in HS10-1, a handful have a second substantive answer (one 03, one 04, four 94’s), and 50 appeared as 98’s (illegitimate blank). In other words, the legitimate re- sponses of HS10 fail to match with the legitimate skip responses of HS10-1, primarily because some of the data appears appropriately as 99’s, whereas another por- tion of the data appears, inappropriately, under the “il- legitimate blank” code 98. Table D Responses to select hospital stay questions in the 12-month Hospital Stay file: “What was the total charge for hospital stay?” by “Why was there no charge?” Response (with code) to question “What was the total charge for hospital stay?” All Valid Not applicable Don’t know Illegitimate Response (with code) to question “Why was there no charge?” _responses response (93) (94) blank (98) All reSpONnsests.. asc SS yet ape Biel 3,132 1,218 2 56 1,856 No-chame (02)... te ee. Pe ee Ae 308 - - - 308 Included with other (03) ..............2.-2226- 130 1 - - 129 in mother's bIIE(O4). i ce ee ee 99 1 - - 98 DOT Ow (G4) es a, OR Ss oe. 1,094 4 - - 1,090 llisgitimate blank (96). Fs Sy eon. 339 50 2 56 231 Legitimate'skip (99) - ct. ies ce i ee 1,162 1,162 - - - 15 ''On the other hand, the use of the “don’t know” (94) category appears to be largely internally consistent. All 56 cases of HS10 “don’t know’s” appear as 98’s - for HS10-1. Of the 1,094 cases of “don’t know” for HS 10-1, all but 4—apparently the 4 valid responses—ap- pear as 98’s in HS10. These findings can be related to the typology of nonresponse. HS10 and HS10—1 exemplify two of the nonresponse patterns. First, there is an inflation of puta- tively illegitimate nonresponse (98’s), which is a simple artifact of question disaggregation (nonresponse type IV). This nonresponse disappears when the question is reaggregated. In addition, there is a very high level of “don’t know” (94) nonresponse (nonresponse type I), and this nonresponse is genuine and in need of further investigation. The final part of the HS10 question series—the sub- question HS10A—sought a reason for no charge from those who indicated in HS10—1 that they had no hospital- stay charge. “No charge” response to question 10 (re- sponse category 02) routes respondents to HS10A. Thus, the 308 “no charge” responses, and only those responses, for HS10—1 should be found in the legitimate response categories 01 through 06 in HSI0A. None of the 02 cases from HS10-1 should appear as a 99 in HS10A, whereas all of the responses to 03 and 04 should appear. The frequencies for HS10A show an apparently high level of nonresponse: Response Number All TSSPONSOS 2". EE cee 3,132 Legitimate: régpense.. ..ca2agie SO POS 320 ont hnew (04) Ss eS 5 Dine they ee SS 443 Legitimate nonresponse (99) .......... 2,364 In addition, it appears that there is more legitimate re- sponse than is licensed by the filter variable (that is, in Table D the 308 “no charge” responses in HS10—1-02, and only those responses, were routed to a legitimate response category of HS10A). The crosstabulation of HS10—1 with HS10A provides an explanation for the large number of legitimate re- sponses. A very small number of respondents had not given the “no charge” (02) answer to HS10-1 yet did give a reason for no charge in HS10A (2 “other source(s) will pay” 04’s, 10 “don’t know’s,” 13 blanks). This inconsistency in the data, therefore, is attributable to interviewer error. As for the high level of nonresponse, 330 of the HS10A 98’s are simply carryovers from HS10-1; that is, if HS10—-1 had been left blank, HS10A would have been left blank as well. Of the remaining 98’s, 100 gave a valid response tg 03 or 04 in HS10-1 and therefore should have been coded as legitimate skips (99) in HS10A. As in the preceding example of HS10 crosstabulated with HS10-1, many, but by no means all, of the legitimate skips are marked with a 99, the balance with a 98. Of those who did in fact answer “no charge” in HS10-1 and were thus routed to HS10A, 16 none answered “don’t know,” and only 12 (4 percent) are blank (98). Thus what appears, from the response frequencies, to be a high nonresponse item, is actually a case where a high percent of those who could be expected to answer did so. The nonresponse is largely spurious, and this variable also displays some amount of inconsistency in the data. Analysis of a Source-of-Payment Variable This section examines a problematic source-of-pay- ment variable taken from the Hospital Stay file. The HS12 series represents one complex question (12) in the questionnaire: 12. Doyou expect any source to reimburse or pay you back? Yes . . . 01 (A) No .. .02 (C BOX) A B Who will reimburse or pay you back? ENTER BELOW. Anyone else? How much will (EACH SOURCE) reimburse or pay you back? SOURCE AMOUNT $ % C |CODE ONE: BOX |TOTAL CHARGE PAID INQ. 11 ......... 01(14) PARTIAL OR NONE PAID INQ. 11 ....... 02(13) The codes are HS-12 Expect reimbursement 01 Yes 02 No HS12A First source code HS12B First source amount in dollars HS12B-P _ First source percent indicator HS12A-1 Second source code HS12B-1 Second source amount in dollars HS12B1P Second source percent indicator HS12A-2 Third source code HS12B-2 Third source amountin dollars HS12B-2P Third source percent indicator nou In HS12, respondents were asked whether they expected reimbursement (“‘yes” or ‘“‘no”); those who answered “yes” were routed through a maximum of three sources: a first—HS12A, a second—HS12A-1, and a third— HS12A-—2. For each source, a source amount was re- quested, expressed in dollars or percent; the amount then could have been split into two variables, dollars and percent. In structure, question 12 has two defining features: It is a disaggregated source-of-payment question; and it contains series-linked subquestions (for example, first source, second source, and third source, where having ''a third source implies having had a second and first, and having a second source implies having had a first). The earlier analysis of disaggregated questions dem- onstrated that their parts must be reaggregated by cross- tabulation to allow the identification of the true pattern and magnitude of nonresponse. This procedure was fol- lowed for the HS12 series. In addition, a consistency check was performed on the series-linked variables, to ascertain whether data had followed the logically pre- scribed path. First, the source-of-payment series was related to its filter, HS12, which asked whether reim- bursement was expected. Subsequently the consistency of the series was checked both horizontally (first source code versus first source amount versus first source per- cent) and vertically (first source versus second source versus third source). How do the responses in the source-of-payment series relate to their filter, HS12 (“Expect reimbursement, yes or no?”)? Among other things, crosstabulation of HS12 with HS12B (“How much will each source pay you back?”) show what happened to the 196 “expect reim- bursement, yes” responses. These are distributed over valid amounts (84 responses), “don’t know” (103 re- sponses), and blank or 98 (9 responses). Thus, HS12B exhibits a high level of genuine nonresponse: Over half of the “expect reimbursement” responses match “don’t know” when asked to specify an amount. Crosstabulation also shows what happened in HS12B to the 1,164 “expect reimbursement, no” responses of HS12. Only one ap- pears with an amount in HS12B; the balance appear, appropriately, as legitimate skips (99’s), where they are added to the legitimate skips (1,572) that appeared for HS12. Having shown the relationship of the first source-of- payment responses to their preceding filter, the next area for examination is the consistency of the source-of- payment series itself, both horizontally and vertically. When HS12B (first source amount) was crosstabulated with HS12B-P (first source percent indicator in dollars or percent), the patterns were largely consistent, with the exception of the disaggregation effect of the HS12B “don’t know’s.” These “don’t know’s” are marked in HS12B-P by 98’s. However, comparison of first source code with first source amount is both more informative and more alarming from the perspective of an analysis of nonresponse. Crosstabulation of HS12A (first source-of-payment code) with HS12B (first source amount) leads to a largely consistent result but an enormous disparity in the number of “don’t know’s” (Table E). For the 193 valid first source codes reported in HS12A, there are 84 valid amounts given, 102 “don’t know’s,” and 7 blanks. Over half of the cases for which there is a first source, then, show “don’t know” for first source amount. This pattern is repeated for second and third sources and also occurs in the other source-of-payment variables. For the second source of payment (HS12A-1 versus HS12B-1), 50 respondents identified a source of payment and 1 gave a “don’t know”; of these 51 responses, 21 show a second source amount in HS12B-1, 28 are “don’t know’s,” and 2 are blank. This pattern holds when the third source is considered. The third source code (HS12A-—2) crosstabulated with the third source amount (HS12B—2) yields nine responses to the third source amount: three appear with valid amounts, five appear as “don’t know’s,” and one is a blank (98). It seems, then, that respondents are much more likely to know the source of payment than to know the source- of-payment amount. As would be expected, the amount variable is plagued by high “don’t know’s” (over 50 percent in this case). To understand the complexities involved in nonre- sponse, it is also illuminating to follow the responses through the three parts of the source-of-payment series, from first source amount to third. Again, in the filter (HS12), the 196 respondents who indicated that reim- bursement was expected became the relevant response population for HS12A and HS12B. Following the HS12B series (see Table F) from first source (HS12B) to second source (HS12B-1) to third source (HS12B-2), reveals a pattern of increasing illegitimate blanks and decreasing data. The response frequencies show only 84 valid re- sponses (with 103 “don’t know’s”) for the first source amount, dropping to 21 valid responses for the second source amount (28 “don’t know’s”), then to 3 valid responses (with 5 “don’t know’s’) for the third source amount. While the 94’s represent genuine nonresponse, the increasing levels of illegitimate nonresponse (98's) appear to be an artifact of the machine-edit convention for this series, which, as with other linked series, did not assign 99 or another legitimate nonresponse or skip code. The lack of provision of a legitimate internal skip Table E Responses to select hospital stay questions in the 12-month Hospital Stay file: “First source amount” by “first source” Response (with code) to question of first source amount Illegitimate Response (with code) to question of first source All responses Amount Don’t know (94) _ blank (98) All responses.) . 2. cE eo ce ee ee es 636 84 103 449 Soufce-of payment’ ee os 193 84 102 7 Qaetorranges95) - Pe eee ee bee 1 - - 1 liiégkiniate blank (98)... 50. 65. SS ee AS PO eT aes 442 - 1 441 17 ''Table F Number of responses and percent distribution as reported in the 12-month Hospital Stay file to questions of first, second, and third source-of-payment amounts Percent Number distribution Responses, (with codes) to first, second, and third source-of-payment questions First source-of-payment amount (question HS12B) Alptesponses.. =... | Gu os 3,132 100.0 NaNO SES DOGS foe, cs he eS 84 ar Adjusted total nonresponse? .......... 552 86.8 Not appicanie (G3) 0 so er ee - - Legitimate nonresponse (99) .......... 2,496 79.7 Wegitie (92) Aye rr Se eR - - Dont knew tee) os eee pe 103 3.3 Gut chiarae (Saye oh SF le - - Multiple response (96) ............. - - Merusaoy ra er eee - - iliegitimate:- bianic ($8)... 5s SS 449 14.3 Second source-of-payment amount (question HS12B-1) Alltesponses! = <->. 3,132 100.0 Valid feptenGe st a eS 21 0.7 Adjusted total nonresponse* .......... 615 96.7 NOt applemvigtgeay =... Se - . Legitimate nonresponse (99) .......... 2,496 79.7 MegibIetO2) 55 oo So, - ~ DONT ROOW eye oi 5. Ge td Ge 28 0.9 Out oF ranggeo) sc - - Multiple response (96) ............. - - RelGomiterys steer, te et - - illegitimate Clank(88)s 8S eg 587 18.7 Third source-of-payment amount (question HS12B-2) All'vasponses?) 3,132 100.0 Valid response tt. Goa oat aa 3 0.1 Adjusted total nonresponse* .......... 632 99.4 Notapricanie (Saj=- ee - - Legitimate nonresponse (99).......... 2,496 79.7 Negins: (92) rs a5 ge - - DOW KNOW (94) oe es 5 0.2 Outofrange-(95). oe a. ee - - Multiple: response (96)... - - ated) Pa ee Se - - illegitimate blank (96)-:.... 55 Sa es 628 20.0 ‘Excludes adjusted total nonresponse. Excludes legitimate nonresponse (99). code for the series implies that all who expect reimburse- ment must have three repayment sources. Yet the more realistic expectation would be that with each successive source of payment and source-of-payment amount, the proportion of the respondent population that would have such a source would decrease drastically. Thus the 98’s in the HS12B series provide an example of type III nonresponse, nonresponse as legitimate internal skips, as well as high genuine “dgn’t know” nonresponse (type I). But they also provide, in the virtually empty third source-of-payment amount category, an example of type V nonresponse, in which nonresponse is the result of a legitimate lack of data. 18 The same pattern would have been obtained from examination of the source-of-payment code variable or the percent indicator. For example, the crosstabulation of the first (HS12A), second (HS12A-1), and third source-of-payment codes (HS12A-2) reveals a largely consistent result. All second source-of-payment re- sponses also show a first source; all but one (a 98) of the third source instances show a second source. The nonresponse patterns for this series are the same two seen in the discussion of HS12B: a high number of 98’s increasing with each successive source-of- payment category that are representative of legitimate internal skips (type III nonresponse) and a virtually empty final data category (type V). Reviewing nonresponse in the light of this informa- tion reveals that the source-of-payment (HS12) series presents four of the five patterns of nonresponse: 1. In such cases as first, second, and third source amount (HS12B, HS12B-1, and HS12B-2) there is high genuine “don’t know” nonresponse (type I). In the latter two, the high proportion of “don’t know’s” is obscured by the avalanche of 98’s that arise from using that code for what amounts to a legitimate internal skip (nonresponse type III; see item 3 below). 2. Variables such as the percent indicator for first source amount (HS12B-—P) are disaggregated questions. Some of these 98’s are actually 94’s in the other disaggregate of question 12, first source amount (HS12B). This kind of code 98 nonresponse (type IV) is an artifact of the mode in which the data are presented. In this instance, 98 is a surrogate for another kind of nonresponse (“don’t know’), just as was the case for the disaggregated total charge question (HS10). 3. Type III nonresponse occurs in the case of the series- linked variables, such as first, second, and third source codes (HS12—A, HS12—A-1, and HS12-A-2, respectively) and first, second, and third source amounts (HS12—B, HS12-B-1, and HS12-B-2, re- spectively) in which 98’s are used for legitimate skips internal to a question. As this type of nonre- sponse is essentially spurious rather than real, there is no need to investigate the characteristics of the nonresponse population. It is extremely important to note, however, that even though this kind of nonresponse is ultimately spurious, it is still capable of imposing a pernicious lack of clarity on the data. Because the source-of-payment and condition ques- tions do not allow the interviewer to indicate when the series ceases to apply to a given respondent, there is no way to know precisely to what extent data are truly missing. If a respondent has three sources of payment but can name only one, 98’s are generated for the second and third sources. If a respondent has only one source of payment and properly names it, 98’s are again generated for the second and third sources. Inability to distinguish ''between these two cases is tantamount to inability to distinguish between cases for which source-of- payment data are missing and those for which they are not. It is true that two kinds of checks are possible in such cases, and that these might in some measure differentiate illegitimate blanks and legitimate skips. An example of a consistency check relating the parts of the series would originate from the assumption that a respondent who had a second source would necessarily have had a first source, and therefore failure to name a first-source would make the first- source blank properly illegitimate (98). Assuming that such series skips are not interviewer error, this would allow for a count of one kind of true 98. A true illegitimate blank, however, is more likely to appear at the end of a series, rather than within a series, so that such a check is likely to identify the lesser amount of the true missing data. As an additional check, the instance where a source-of- payment code is indicated, but the corresponding amount is blank, would constitute a true 98. Apart from such internal inconsistencies that implicate only a small proportion of the data, however, there is no way to distinguish with certainty which of the 98 codes should in fact be legitimate skips. On the basis of the evidence presented, it nevertheless can be presumed that most of the alleged 98’s are in fact legitimate skips. However, certainty in the indi- vidual case, as well as an overall indicator of the extent and nature of nonresponse, is needed if the variable is to be subject to imputation procedures. Finally, in such multiple source-of-payment series as second and third source (HS12A—1 and HS12A-2, respectively) and second and third source amount (HS12B-—1 and HS12B-2, respectively), there is al- most complete legitimate nonresponse (type V). The third source-of-payment code (HS12A-—2), for exam- ple, is virtually empty of data, with a valid response representing less than one-tenth of | percent of the file universe. The problem here (apart from the machine-edit convention of using 98’s for legitimate internal skips in linked series) is not illegitimacy of nonresponse, but a question that asks for informa- tion about something (a third source of payment) that applies to almost none of the respondents. 19 ''Scope of Nonresponse This section attempts to convey some sense of the extent and magnitude of the item nonresponse problem in the three critical event files (Medical Visit, Hospital Stay, Flat Fee) that are used in this analysis. Presented here are the proportion of variables that are problematic and a classification of these problematic variables by the type or types of nonresponse that affect them. Also included is a comparison of levels of nonresponse for parallel charge and payment variables across files. Again, the formula for computing response and non- response (formula 2B) excludes all legitimate skip consis- tency codes (99) from nonresponse: number of number of valid responses + legitimate skips Response rate = : number of applicable events This formula gives the truest picture of the overall response rate, but a further distinction, that between genuine and spurious nonresponse, must be included to distinguish between cases in which nonresponse is a genuine problem of missing information and cases in which nonresponse is an artifact of the conventions for coding, editing, or presenting the data. For some purposes, then, an additional equation, formula 3, is used for determining the rate of response and attendant nonresponse. Formula 3 is: number of valid responses + number of legitimate skips + number of spurious nonresponse Response rate = 3 number of applicable events Overall Level of Nonresponse The following series of questions and variables from the Hospital Stay, Medical Visit,-and Flat Fee files are classified as problematic on the basis of the 10-percent criterion (after legitimate skips, code 99, have been excluded). Hospital Stay file HSS series (variables HS5—1A, HS5—2A, HS5-3A; HS5C-1A, HSSC-2A,HS5C-3A) 20 HS6 series (variables HS6A—1A and HS6A-2A) HS10 series (variables HS 10, HS10-1) HS11 series (variables HS11, HS11—1) HS12 series (variables HS 12, HS12A—1 and HS12B-1, HS12A—2 and HS12B-2) HS13 series (variables HS13, HS13A—1 and HS13B-—1, HS13A—2 and HS13B-2) AS 18 series (variables AS 18-1) AS19 series (variables AS 19-1) AS20 series (variables AS20A-1, AS20B-1, AS20A-—2, AS20B-2) AS2] series (variables AS21A—1, AS21B-1, AS21A—2, AS21B-2) Medical Visit File MVF-DIAG MVF-GEN MVF-EYE MVF-IMM MVF-FP MVF-OTHR MVF-TC series (variables MVF-TC, MVF-NTC, MVF-NC) MVF-CN series (variables MVF—CN2, MVF-CN3 MVF-CN4) MVF-SOP series (variables MVF—SOP2, MVF-SOP3) MVF-SA series (variables MVF-SA2, MVF-SA3) MVF-SC series (variables MVF-SC1, MVF-SC2, MVF-SC3) MVF-AP series (variables MVF—AP1, MVF—AP2, MVF-—AP3) Flat Fee file FF3 series FF4 series (variables FF4B, FF4B—1, FF4B-2) FF5 series (variables FFSA, FFSB—1, FF5B—2) FF6 series (variables FF6, FF6A, FF6B) > ''Perhaps the simplest way to quantify nonresponse is to look at the proportion of variables for each file that are problematic according to the 10-percent criterion. Given that restriction, over half the variables in the three files are problematic. However, this is a misleading way to state the extent of the nonresponse problem be- cause, as has been shown, the 98 code cannot be taken at face value, and a great deal of this nonresponse is spurious. A more realistic look at the extent of nonresponse might be gained by classifying the problematic variables under the five nonresponse types. This approach offers the advantage of differentiating between genuine and spurious nonresponse, and, because different magnitudes of nonresponse are assignable to different nonresponse patterns, it allows generalizations about the level of non- response within a given category of problematic vari- ables. Appendix II assigns the problematic variables to the five types of nonresponse. The magnitude of nonre- sponse is indicated on a scale of comparatively low (10-20 percent) to moderate (21-40 percent) to high (41-100 percent). Some variables are attributed to more than one category because they are affected by more than one kind of nonresponse. Several generalizations can be drawn from the appendix; a summary is presented in Table G. Of the 113 variables considered, 17 (15 percent) fall into the genuine nonresponse categories (types I and II), and 96 (85 percent) fall into the spurious nonre- sponse categories. Generally, the high magnitude nonre- sponse belongs to the cases where the illegitimate blank (98) code is used to mark legitimate nonresponse data (types III, IV, and V). The variables for which there is a high level of genuine nonresponse are charge or payment items. In some cases, the level of genuine nonresponse can be disturbingly high (as in HS10 and HS10-1, where the “don’t know” rate on the critical hospital charge item exceeds 36 percent). Still, the great- er part, both in extent and magnitude, of the face-value nonresponse in these data files is spurious, a function of editing and data presentation conventions rather than of true item nonresponse. Nonresponse Across Files Nonresponse to problematic charge items can be compared from file to file. For example, in deciphering whether the level of genuine nonresponse was the same or different on parallel items in the Medical Visit and Hospital Stay files, several key items were selected based on type I (high “don’t know”) nonresponse. All of the items were alike in content, wording, and format, except that in one case, the item dealt with a hospital visit, and in another, with a medical visit. The first point of comparison is the total charge item (HS10, HS10-1, MVF-TC); the second, the amount-family-paid item (HS11, MVFFP-2); the third is “expect reimbursement, first source amount” (HS12B, MVF-SA1); and the last is “Will anyone else pay?” first source amount (HS13B, MVF-AP1). Percents of the “don’t know” nonresponse are as follows: Hospital Medical Amount Stay visit Percent FotmL ChARGe 5 eee a ee ne 8 37 15 Amount family paid“... ee Ps. oe. 14 12 First source reimbursement amount ... . 29 20 First source anyone-else-pay amount .. . 27 1§ Thus, while charge and source-of-payment questions are consistently a problem, it would seem that they may be more of a problem for certain kinds of events than for others. Specifically, the level of genuine nonre- sponse is higher in the Hospital Stay file than in the Medical Visit file when questions rigidly parallel in con- tent and format are compared. This result is not unex- pected because patients are less likely to see their bills or pay directly for their hospital stays than for their medical visits. Greater efforts, perhaps greater recourse to provider medical records, may be required to improve charge data in such especially high nonresponse categories as hospital stay. Table G Cumulative nonresponse for charge questions in the 12-month Medical Provider Visit, Flat Fee, and Hospital Stay files, by magnitude and type of nonresponse Type of nonresponse’ Magnitude 21—40 percent Total 10-20 percent 41-100 percent Potala gs oe eT. oe Be SS I. Nonresponse as high “don't know’s” ............ ll. Blanks (98's) as surrogates for “don’t know” ........ il. Nonresponse as legitimate internal skips .......... IV. Nonresponse as an artifact of question disaggregation . . . V. Nonresponse as legitimate lack of data........... «tates 113 11 7 95 aca 13 8 5 - Peed 4 3 1 A $3. 58 : - 58 ae 12 - 1 11 in 26 - : 26 ‘Includes data from the Medical Visit, Hospital Stay, and Flat Fee files combined. 21 ''Demographic Characteristics of the Item Nonresponse Population Such genuine high nonresponse as has been identified throughout this report usually occurred with charge and source-of-payment amount questions, where “don’t know” rates of over 10 percent are the rule. Because these questions do not appear to be plagued by lack of. clarity in question wording or problems in formatting, it may be assumed that respondents are not as familiar with medical charges as with other medical care and utilization items. The question then arises as to whether those who make up the “don’t know” population for the charge and payment variables have any common characteristics. If so, particular populations could be targeted for special efforts designed to obtain better charge information about them, perhaps from medical care providers. In order to identify common characteristics, the “don’t know” response in the high “don’t know” variables was crosstabulated with such variables as sex, race, region, round (1-5), income of reporting unit, age of respondent, education of respondent, Medicare-Medicaid status, and report status (all self-reported, partially self- reported, or proxy). These crosstabulations produced in- formation about the percent of each given group that answered “don’t know.” For example, on typical charge variables, if 10 percent of those with 12 years or more of education answered “don’t know,” compared with 30 percent of those with 0-7 years of schooling, this would mark a revealing difference between groups. (A few problematic question variables with a high proportion of nonresponse but a low absolute number of events were excluded from the analysis. ) The problematic charge and payment variables were a problem for all groups, whether young or old, rich or poor, educated or uneducated. Nonetheless, systematic differences between groups were observed on several of the demographic variables, including respondent age, respondent education, income of reporting unit, and Med- icare-Medicaid status. In the case of other distinctions (sex, race, region, report status, and round), only small or inconsistent differences were observed across groups of “don’t know” respondents. The direction and mag- nitude of the systematic differences for the age, educa- tion, income, and Medicare-Medicaid status variables are summarized in Tables H, J, K, and L, respectively. All events coded 93 (not applicable) and 99 (legitimate nonresponse) have been excluded from the analysis. 22 Table H Percent of “don’t know” responses reported to select charge and source amount questions in the 12-month Hospital Stay, Flat Fee, and Medical Provider Visit files, by three age groups Age Charge and source amount questions 18-35 36-59 60 years with codebook names years years or over Hospital stay Percent Total charge (HS10-1) .......... 59 52 65 Amount family paid (HS11). 2.2... ~. 10 13 21 First source amount (HS13B) ..... 19 24 3 Second source amount (HS13B-1) . . 4 9 31 Doctor in hospital Total physician charge (AS18) ..... 50 62 68 Amount family paid (AS19)....... 12 18 25 First source amount (AS21B) ..... 20 31 44 Flat fee How much paid; first source (FF5B) . . 16 TF 20 Medical visit No total charge (MVF-NTC) ...... 26 32 37 First source amount (MVF-SA1) .. . . 12 18 34 First source amount (MVF-AP1) .. . . 9 13 26 Respondent Age x There is a marked tendency for the oldest age group / / ¢ ‘ to incur the highest percent of “don’t know’s” in the selected charge and payment variables. Table H depicts _this phenomenon. With the exception of “no total charge” (MVF- NTC), the charge variables share in the pattern of the 60-years-and-over age group having a higher percent, sometimes substantially higher, of their responses failing in the “don’t know” category. The same pattern holds if participant age is employed as the age variable rather than respondent age. (“Participants” are all persons in a reporting unit who were assigned a Participant Identifi- cation (PID) number; “respondents” are persons who reported for themselves and other participants.) Respondent Education Years of education were divided into three categories (Table J): 0-7 years, 8-12 years, and 12 years or more. ''The tendency in this case is for the least educated group \ to show a higher percent of “don’t know” responses. The exception is the “no total charge” category. Table J Percent of “don’t know” responses reported to select charge — and source amount questions in the 12-month Hospital Stay, Flat Fee, and Medical Provider Visit, files, by years of education of respondent Charge and source amount Youre of etlucation Income of Reporting Unit Four income categories were used in this analysis: lowest (less than $10,000), low-middle ($10,000- $19,999), high-middle ($20,000-$34,999), and highest ($35,000 or more). Although it is difficult to generalize about the other groups, the lowest income group consis- tently had the highest reporting of “don’t know’s,” with the exception of the HS10~-1 (‘reason for no total charge”) variable, under which both the highest and the lowest income groups showed proportionally higher use of “don’t know” (the reporting of the highest income questions with codebook names 0-7 8-12 12 or more : : group, as a matter of fact, was slightly higher than Hospital stay Percent that of the lowest group—Table K). However, these Total charge (HS10-1) ....... 65 57 57 “don’t know’s” should have been assigned to HS10 Amount family paid (HS11). . . . 19 15 10 (“total charge for hospital stay”); thus, the exception First source amount (HS13B) .. . 37 27 27 is on that variable. Second source amount (HS13B-1)..........-.. 21 13 11 Doctor in hospital Medicare-Medicaid Status Total physician charge (AS18—1) . . 74 58 51 Amount family paid (AS19) . . . . . 26 18 15 A Medicare-Medicaid variable was used to determine eee = 25 whether the responses with respect to knowledge of Flat fee charges and payments of those who received Medicare How much paid: first source or Medicaid benefits were different from those of the (FPSO... ed, aS 23 16 19 respondents who did not receive benefits. Those with Soe Medicare or Medicaid coverage consistently showed a Medical visit : “ > % “ higher tendency toward “don’t know” response. “No oe ke = 7 total charge” (MVF-NTC) again appears as an exception First source amount (MVF-SA1) . . 31 20 20 : ; 3s First source amount (MVF-AP1).. 20 15 12 (Table L). Interestingly, the percent of “don’t know responses in the category “amount family paid” (HS11) is the same for those with and without Medicare- Medicaid coverage. Table K Percent of “don’t know” responses reported to select charge and source amount questions in the 12-month Hospital Stay, Flat Fee, and Medical Provider Visit files, by income category of reporting unit Charge and source amount questions Income category with codebook names Lowest Low-middle High-middle Highest Hospital stay Percent Total charge (HS10-1) 2... 0... i ee ee ee 57 53 55 62 Amount family paid (HS11).. 2... 2.2.2... 2.2... ee ee ee ee ee 17 14 11 12 First source amount (HS13B) .............02. 2000022 eee 32 26 22 21 Second source amount (HS138B-1) ............2.2.0 2000022 ee 20 11 8 10 Doctor in hospital Total physician charge (AS18-1). 2.2... .......2.-.2.02. 2.200005 68 56 54 61 Amount family paid (AS19).. 2... 2... ee ee ee ee ee 20 20 14 16 First source amount (AS21B) ...........0 000 eee eee ee eee 37 31 27 26 Flat fee How much paid, first source (FF5B) ...............-.-222-00- 19 17 16 14 Medical visit No total charge (MVF-NTC) ..........2.2.0 00 ee eee eee eee 34 29 30 28 First source amount (MVF-SA1) .. 2... ..0..0.2..0 00.2 e eee ee eee 22 20 16 21 First source amount (MFV-AP1).............02.0 0200200005 20 14 13 8 a ''Table L Percent of “don’t know” responses reported to select charge questions in the 12-month Hospital Stay, Flat Fee, and Medical Provider Visit files, by Medicare-Medicaid coverage Charge and source amount Medicare-Medicaid coverage questions with codebook names Yes No Hospital stay Percent Total charge (HS10-1) .... 2 ae 71 61 Amount family paid (HS11)......... 60 60 First source amount (HS13B) ....... 41 25 Second source amount (HS13B—1) 31 6 Doctor in hospital Amount family paid (AS19)......... 23 21 First source amount (AS21B) ....... 43 33 Flat fee How much paid, first source (FF5B) ... . 43 16 Medical visit No total charge (MVF-NTC) ........ 33 38 First source amount (MVF-SA1)...... 32 26 First source amount (MVF-AP1)...... 20 15 24 Summary In summary, although knowledge of charge and pay- ment information is problematic for all groups, the data indicate that it is more of a problem for some groups than for others. In particular, four overlapping groups— the elderly, the least educated, those with lowest in- comes, and those receiving Medicare and Medicaid—are disproportionately more likely to give “don’t know” re- sponses to charge questions. For these overlapping popu- lations (which, for most practical purposes, may be primarily thought of as the Medicare-Medicaid group), charge information is likely to be lacking and may need to be sought from medical provider or other administra- tive records. ''Recommendations In this discussion, recommendations are made ad- dressing the specific issues raised by each of the five types of item nonresponse identified in the NNCUES 12—month data files. This discussion will be followed by anumber of more general recommendations. ! Two of the five types of item nonresponse are in regard to a misleading presentation of data in the data file, in which nonresponse is only apparent. Two other types have to do with genuine nonresponse, either “don’t know” rates of 10 percent or higher, or response categories virtually empty of data. The final type involves genuine but legitimate nonresponse, categories empty of data that misleadingly appear coded as illegitimate nonresponses (98’s). The issues and recommendations concerning spurious nonresponse will be addressed first, followed by a discussion of genuine nonresponse. The two types of spurious nonresponse are nonre- sponse as legitimate internal skips (type III), when the illegitimate nonresponse code (98) was used to mark legitimate internal skips within a linked series; and nonre- sponse as an artifact of question disaggregation (type IV), when the illegitimate nonresponse code was used as a marker for what was, in fact, legitimate response appearing in the disaggregated part of the question. As noted, in linked-series variables such as source of payment and condition, no legitimate internal skip code had been assigned. Nevertheless, the most realistic assumption would be that each successive source of payment or condition would have been applicable to less and less of the respondent population. Given that third and fourth sources of payment, condition, and so on, were virtually devoid of valid response, and given that crosstabulations showed a high degree of logical consistency between the successive parts of linked-series variables, it would seem that the overwhelming number of illegitimate blanks recorded in such series would con- stitute legitimate internal skips. The problem with this kind of nonresponse is that it is misleading; it dramati- cally inflates the number of illegitimate blanks (98), and it fuses the large number of legitimate skips with 'These recommendations are directed toward a future NMCUES. However, it might be prudent to reexamine the current Public Use version of the 1980 NMCUES data (see Public Use Data Tape Documentation: NMCUES, 1980) in light of the 12-month file findings, to determine to what extent difficulties identified for the intermediate stage of the data base construction process have implications for the final stage. a small number of illegitimate blanks, so that they are indistinguishable. In other words, in the example of source-of-payment information, there is no way to distin- guish between a respondent who had only one source of payment and therefore had no second or third source information to provide, and a different respondent who had three sources of payment but could name only one and was, in effect, answering “don’t know” to the second and third. If a respondent lacked a second, third, or fourth condition or source of payment, the interviewer had no option other than to leave the question blank. A solution to the type III nonresponse situation would be to provide a response category that would permit the interviewer to record the place where a series would cease to be applicable to the respondent. This could be done by breaking down the highly compact linked series into discrete subquestions and by incorporating into the subquestion series a place to mark closure of the series. This recommendation will be elaborated on later in this section. Another example of misleading presentation of the data may be seen in the “disaggregated” questions, such as the total charge variables. In these instances, a com- plex question is broken into two parts; the sum of the parts totals the file number. Thus, a valid response in one part of the question calls for a nonresponse code in the other part. Typically, one response was coded 99 and the other 98. Though such variables often were plagued with high genuine nonresponse, they also represented a distinct, and spurious, nonresponse type, nonresponse as an ar- tifact of question disaggregation (type IV). Thus, as seen in such cases as Hospital Stay file question 10 (disaggregated into HS10 and HS10-1), some 98’s (il- legitimate blanks) were simply markers for a valid re- sponse in the other part of the disaggregate. Three basic problems of ambiguity and lack of clarity resulted from disaggregated questions. 1. When a 98 was used as a marker for a legitimate response in the other disaggregate of a question, this variety of 98 and the true illegitimate blank 98 were classified as being the same. While cross- tabulation successfully distinguishes between the two cases, it should not be necessary to go to such lengths to achieve clarity of data. = ''2. Because of inconsistencies in the data, the two halves of a disaggregate could not readily be reaggregated. For example, in HS10 (“total charge”) and HS10-1 (“reason for no total charge”), the valid responses in HS10 should have been marked by a 99 in HS10-1, but the set of relevant numbers failed to match. 3. The nonresponse categories were often presented in an ambiguous and arbitrary way. In the case of HS10 (‘total charge”) and HS10-1 (“reason for no total charge”), the codebook assigned the “don’t know” category for question 10 to HS10-1 (“reason for no total charge”) and not to HS10 (“total charge”). While there is one “don’t know” response category provided for in the question, the frequencies give two “don’t know’s,” one for HS10 (barely utilized) and one for HS10-1 (heavily utilized). Although only one “don’t know” response category appears in the question, two conceptually distinct sorts of “don’t know” are possible: (a) The respondent does not know the total charge, or (b) the respondent does not know why there was no total charge. The “don’t know” frequency assigned to HS10-1 is pre- sumably translated as “lack of a reason for no total: charge,” whereas the frequency that appears for HS10 presumably would have been written in by the interviewer as “don’t know total charge.” The results of crosstabulations are wholly consistent with this interpretation, but the frequencies run counter to intuition: More people would be expected to lack information about the amount of their hospital-stay charge than would lack information as to why there was no charge. In fact, the “don’t know” frequencies appearing under both HS10-1 and HS10 refer to HS10 (“total charge”). Again, a general lack of clar- ity in the meaning of the data, not entirely a function of question disaggregation, but certainly com- pounded by it, is the result. There are several ways in which the presentation of data in disaggregated questions could be made clearer. In the case of the first example just presented, more consistent use of codes 98 and 99 would largely solve the problem. In the second case, better data cleaning would clearly be an appropriate solution. In the third case, adding the two “don’t know” frequencies together (for they are alike in kind and differ only in genesis, the smaller number having been written into the question- naire margin beside the “total charge” blank) and provid- ing a clearer account in the codebook would suffice. With respect to the formatting of the questionnaire, if the “don’t know” in a “total charge” question really is intended to refer to “total charge” and only to “total charge,” then the “don’t know” category should be put in physical proximity with the “total charge” category to which it refers. Thus, instead of letting “don’t know” trail as the final element, a less ambiguous arrangement would be as follows: 26 $ (11) Dont know-totel charge 2. es a 94 (11) RR eae, er oes, ee es 02 (A) Included with other charges ........... 03 (FF__[14]) FOR NEWBORNS ONLY: Included in mother’s bill (Person # ee ae en tee 04 (15). Another possible approach to the problems of ques- tion disaggregation might be to break down the highly compressed questions of the NMCUES questionnaire. In the case of the “total charge” questions, for example, the fact that in 2 percent of cases interviewers wrote in “don’t know” rather than using the available “don’t know” response category suggests that there may have been some confusion on the part of the interviewers as to the reference of the “don’t know” response category; it is also possible that there may have been some misuse of it (for example, if there was no total charge but the reason was not known, that response could have ended up in the “don’t know” response category). Thus, there might be an argument for separating in one box the “total charge” amount and its “don’t know” category, and in another the alternative response options of “reason for no total charge,” In some cases, disaggregation of the questions them- selves also might be appropriate for condition and source- of-payment variables. For example, instead of the box used in the questionnaire (which does not allow a way to differentiate between a respondent who had only one source of payment and another who had three but was able to name only one), sources could be disaggregated. Thus, instead of the present structure, Do you expect any source to reimburse or pay you back? Vase iS, Oe ee ae ee (A) a eee See eS ee (C) A B SOURCE AMOUNT die Ghia oi cts the following series could be utilized: Do you expect any source to reimburse or pay you back? WO a Ot os SO OR ne ee eee a ee (A) SOURCE Did you have a second source of payment? Vee... SS see (A_1) Ad B4 SOURCE AMOUNT This format could be extended to a third source, if a third source category is deemed desirable. ''The remaining types of nonresponse are all genuine. One type, which involves genuine but legitimate nonre- sponse and occurs in close conjunction with such cases of nonresponse as 98 codes for legitimate internal skips, is that of categories empty of data (type V). Multiple condition and source-of-payment series, in their final and sometimes next-to-final categories, were often virtu- ally empty of data. Specifically, in the Medical Visit, Hospital Stay, and Flat Fee files, the quantity of valid response data elicited is less than | percent of the file number for all fourth conditions and for all third sources of payment. In such cases, it should be reassessed care- fully whether this small amount of data is worth collect- ing. If not, the format of the questionnaire might be altered so that third sources of payment and fourth condi- tions would be dropped. The other two patterns of genuine nonresponse in- volve “don’t know’s,” explicit on the one hand (type I) and hidden on the other (type II). For some variables there is simply a high (10 percent or more) rate of “don’t know” response. For other variables, there is reason to assume that many respondents would lack requested information, but no “don’t know” category has been provided. This assumption, concomitant with a high ratio of 98’s, is supportive of the hypothesis that a proportion of illegitimate blanks are surrogates for “don’t know.” In general, there appears to be a substantial, though not unreasonable or unexpected, amount of charge and payment nonresponse. Because this nonresponse is ex- pected and shows no evidence of resulting from question wording or questionnaire format, the strategy of this analysis for dealing with genuine nonresponse has been to examine the character of the nonresponse population to see whether some groups were disproportionately rep- resented. Certain groups (the elderly, the less educated, the poor, and those covered by Medicare and Medicaid) do tend to have higher rates of genuine nonresponse. Therefore, it may be worthwhile to consult medical pro- vider records to supplement respondent information for these groups. (Also, high response was harder to achieve for certain kinds of events than for others; for example, it was more difficult to achieve high response for hospital stays than for medical visits. It also might be profitable to pursue charge data for certain kinds of events from medical provider sources.) Although special measures to elicit additional medi- cal utilization data for high nonresponse groups would seem the most sensible approach to the problem of high levels of “don’t know” response, the NMCUES data present the additional obstacle that sometimes “don’t know’s” are masked and appear under other nonresponse codes such as illegitimate blanks. This problem raises the question of what could be done to enforce more consistent use of the “don’t know” category. Earlier, three distinct relationships applicable to any given ques- tion with regard to the “don’t know” response category were identified: 1. “Don’tknow” category provided. 2. No “don’t know” category provided, but interviewers instructed to write in “don’t know.” 3. No “don’t know” category provided; no write-in op- tion instructed. There is no likelihood that additional legitimate data would have been collected had a “don’t know” response category been consistently provided; however, nonre- sponse, insofar as this is of interest for its own sake, would have been more precisely interpretable. The possi- ble negative aspect of providing a “don’t know” category would be that, because it is the “easiest” option for both interviewer and respondent, such a category might encourage nonresponse, especially for questions that are particularly sensitive or demanding. It would therefore be appropriate to conduct an experiment in which some respondents and not others are provided with a “don’t know” option. The variables that would be used would be those for which there is no “don’t know” response category and for which there had been high use of 98’s. As a pendant to the five types of nonresponse iden- tified in this analysis, possible problems of illegitimate skips in highly compressed questions were examined. It is recommended that questions such as Medical Visit question 5, which has 15 constituent variables, multiple complex skips, and an instruction to “Code all that apply,” be reformatted into several discrete questions. It also is recommended that such “open” formats for recording multiple visits as those that appear in the Hospi- tal Outpatient Department Visit section (core question- naire pages OPD-26 through OPD-—31) be reformatted to reflect the “self-contained” and comparatively error- proof visit-recording structure that characterizes the pre- cisely parallel questions in the Medical Provider Visit section (section MV of the questionnaire). In addition to recommendations specific to the five kinds of item nonresponse and illegitimate skip problems found in the NMCUES 12-month data files, a number of general recommendations may be made, primarily about the data-cleaning and quality-control processes. One such recommendation concerns the need for a universal definition for each file. This is especially important if the event-level files are to be used for person- level analyses. Thus, the universal definition for the file should state specifically how many persons are in the file as well as how many events. Another recommendation concerns the need to have an effective quality-control plan from the earliest stages of the project to provide feedback that will improve the final form of the data collection and reduction effort. Evaluation and quality control are, of course, ongoing processes that have a vital role in each round; however, their utility may be maximized by putting special em- phasis on round 1. An extremely careful field edit to identify problem interviewers and a very close machine edit to identify problems in the questionnaire will be 2T ''vital to the goal of achieving the highest quality NMCUES data. The recommendations discussed in this section of the report may be summarized as the following: 1 28 Divide complex questions into subquestions in the questionnaire itself, not just in the codebook. Use this process to reduce complexity and ambiguity; for example, where there is one “don’t know” re- sponse category, included with multiple response categories to which it might refer, include the “don’t know” with the germane alternative. Reformat into multiple discrete questions each ex- tremely tightly packed question with complex skip patterns. Avoid formats that maximize the likelihood of illegitimate interviewer skips. Use nonresponse codes more consistently (especially codes 98 and 99) in data-reduction tasks. Avoid assigning data with an identical question refer- ence to separate variables. Combine such frequencies in one place. Drop categories that elicit 99-percent or higher legiti- mate nonresponse. Supply a universal definition for each data file. Such a definition should be designed to facilitate person- level analysis in the event files by providing a person number in addition to an event number. Conduct a split-ballot experiment to test the effects of providing an explicit “don’t know” response cate- gory for the type II (illegitimate blanks as surrogates for “don’t know”) nonresponse variables, where there were high 98’s but no “don’t know” response option. Emphasize evaluation and quality control in round 1. ''References Allen, D., and Moser, B.: Keying and Verification Manual for HH Interview Questionnaire and Supplement #1. NMCUES Report. Research Triangle Institute, revised Apr. 1980. Bonham, Gordon S.: Procedures and Questionnaires of the National Medical Care Utilization and Expenditure Survey. National Medical Care Utilization and Expenditure Survey. Series A, Methodological Report No. 1. DHHS Pub. No. 83-2001. National Center for Health Statistics, Public Health Service. Washington. U.S. Government Printing Office, Mar. 1983. National Medical Care Utilization and Expenditure Survey Core Questionnaire. NORC. National Medical Care Utilization and Expenditure Survey: Question-by-Question Specifications. 1979. Public Use Data Tape Documentation: National Medical Care Utili- zation and Expenditure Survey, 1980. National Center for Health Statistics, Public Health Service. Hyattsville, Md., Apr. 1984. Research Triangle Institute: National Medical Care Utilization and Expenditure Survey: 12-Month Data Base Codebook. Feb. 1982. 29 ''Appendixes I." DEHAIDOns OF =LoIms Wood IT TiissOpon os ee ee ae 31 li oNGhrosmonce Vanebieed 2" oe ae ee a SS ee ee ee 34 Type laNoOnreaponse-aaihigh Dent Knows’. oho kg eS ae ee 34 ‘Type:li: Blanks (08's) as Surrogates for“Don't Know™. kg 4. PS RR ee 34 Type tl Nonresponse as Legitimate internal’ Skips... oe ee a 34 Type IV: Nonresponse as an Artifact of Question Disaggregation ................-.00 0. eee eee 35 ‘hype -V sNenresponge asogitimare Lackol Dalai a 35 It; Selections Fromm dio Questionndlie. =. 6 a ee eee 37 30 ''Appendix | Definitions of Terms Used in This Report Age—tThe age of the person as of January 1, 1980. Babies born during the survey period were included in the category “under 5 years.” Condition—Any entry on the questionnaire that de- scribes a departure from a state of physical or mental well-being. It is any illness, injury, complaint, impair- ment, or problem perceived by the respondent as inhibit- ing usual activities or requirir.g medical treatment. Preg- nancy, vasectomy, and tubal ligation were not considered to be conditions; however, related medical care was recorded as if they were conditions. Neoplasms were classified without regard to site. Conditions, except im- pairments, are classified by type according to the Ninth Revision of the International Classification of Diseases (World Health Organization, 1977) as modified by the National Health Interview Survey Medical Coding Man- ual (NCHS, 1979); these modifications make the code more suitable for a household interview survey. Impair- ments are chronic or permanent defects, usually static in nature, that result 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. Impair- ments are classified by using a supplementary code specified in the coding manual. In the supplementary code, impairments are grouped according to type of functional impairment and etiology. Control card—A_ computer-generated instrument providing administrative control of the samples, informa- tion to help the interviewer to locate and identify sample persons, procedures for determining reporting unit com- position, and places to record information required across rounds of interviewing. Core questionnaire—The basic interview instrument used during each interview to obtain data about health, health care, charges for health care, sources of payment, and health insurance coverage. Disaggregation, disaggregated question—A single question in the core questionnaire is often related to several discrete items in the response frequencies. These items are labeled separately, and their response frequen- cies sum to 100 percent. In such instances, the question as it appeared in the questionnaire is said to have been disaggregated in the response frequencies. NOTE: For the purposes of this report, the definitions included in this appen- dix may differ from the definitions presented in other NMCUES series reports. Education of individual (respondent)—The years of school completed for people 17 years of age and over. Only years completed in regular schools, where persons are given a formal education, were included. A “regular” school is one that advances a person toward an elementary or high school diploma or a college, university, or profes- sional school degree. Thus, education in vocational, trade, or business schools outside the regular school system was not counted in determining the highest grade of school completed. Emergency room visit—The emergency room visit section of the NMCUES questionnaire contained ques- tions about the conditions requiring treatment and the reason that the person visited the emergency room rather than some other source of care. General questions on procedures, the charge, and source of payment were included in this section. An emergency room visit and a hospital stay were recorded if a person was admitted to the hospital as a result of the visit. Event file—Most NMCUES files were organized with events (for example hospital stays, medical visits, and hospital outpatient department visits) as the unit of analysis. In an event file, any given respondent might have zero, one, or more records (events), depending on that person’s response to the provider probe questions. Family—A group of people living together and re- lated to each other by blood, marriage, adoption, or foster care status. An unmarried student 17-22 years of age living away from home was also considered part of the family, even though his or her residence was in a different location during the school year. Filter—A question asked to determine which sub- sequent question or questions, if any, should be asked. Flat fee—A single lump sum charge for a variety of services or supplies or a series of visits. The single charge was paid in one lump sum or by installments, but in a way that could not be related to individual events of health care. A flat fee was associated with only one person. If a hospitalization was involved, the total flat fee was assigned to the hospitalization, and a zero charge was assigned to all visits. Otherwise, the flat fee was equally distributed among all the as- sociated visits. High nonresponse—For purposes of this report, item nonresponse was termed “high” if it was 10 percent or more of the expected number of responses. To deter- 31 ''mine the expected number of responses for a given item, the number of legitimate skips of that item was subtracted from the total number of events in the file. Hospital admission—This is the formal acceptance by a hospital of a patient who is provided room, board, and regular nursing care in a unit of the hospital. A patient admitted to the hospital and discharged on the same day is considered to have had a hospital admission. Also included is a hospital stay resulting from an emergency department visit. Hospital stay—The Hospital inpatient section of the NMCUES questionnaire collected information about each admission to a hospital, including admissions that did not require an overnight stay, such as in-and-out surgery. Hospital stays resulting from an emergency room visit are also included. Hospital outpatient department—A hospital-based ambulatory care facility organized to provide non- emergency medical services. Persons receiving services do not receive inpatient nursing care. Examples of outpa- tient departments or clinics are pediatric, obstetrics and gynecology, eye, and psychiatric. Hospital outpatient department visit—A face-to-face encounter between an ambulatory patient and a medical person. The patient comes to a hospital-based ambulatory care facility to receive services and departs on the same day. If more than one department or clinic is visited on a single trip, each department or clinic visited is counted as a separate visit. Household—Occupants of group quarters or of a housing unit that was included in the sample constitute a household. A household can comprise one person, a family of related people, a number of unrelated people, or a combination of related and unrelated people. Housing unit—A group of rooms or a single room occupied or intended for occupancy as separate living quarters is a housing unit if the occupants do not live and eat with any other persons in the structure, and if there was either direct access from the outside or through a common hall, or there were complete kitchen facilities for the use of the occupants only. Illegitimate blanks (illegitimate nonresponses)—The NMCUES consistency code 98 that was reserved for all instances in which an item that should have been answered was left blank. Income—Although income data were collected in both rounds 1 and 5, the income figures in this report reflect the income reported during the round | data collec- tion. In round 1, respondents were asked to categorize the income level of their families or themselves during the preceding 12 months. Thus, the annual income fig- ures collected for each reporting unit primarily reflect income received in 1979, with minor overlap into early 1980. Key person—A key person was (1) an occupant of a national household sample housing unit or group quarters at the time of the first interview; (2) a person related to and living with a State Medicaid household 32 case member at the time of the first interview; (3) an unmarried student 17-22 years of age living away from home and related to a person in one of the first two groups; (4) a related person who had lived with a person in the first two groups between January 1, 1980, and the round | interview, but was deceased or had been institutionalized; (5) a baby born to a key person during 1980; or (6) a person who was living outside the United States, was in the Armed Forces, or was in an institution “at the time of the round | interview but who had joined a related key person. Medical provider—Any medical person who pro- vided medical (nondental) services. Bonham (1983) notes: “As used in this survey, the term ‘medical pro- vider’ referred to all persons engaged in the prevention, diagnosis, and treatment of physical or mental health problems whether or not they had medical degrees. This definition included persons such as chiropractors, speech therapists, faith healers, psychologists, and nurses, as well as medical and osteopathic doctors.” In the Medical Provider Visit section of the NMCUES questionnaire, provision was made for recording eight specific kinds of medical persons, in addition to “Other (SPECIFY)”: medical doctor, chiropractor, podiatrist, optometrist, psychologist, social worker, nurse, and_ physical therapist. Medical visit—An ambulatory visit to a medical pro- vider (e.g., doctor, nurse, physical therapist, laboratory technician). The number of visits is based on responses to the medical provider, emergency room, and hospital Outpatient department visit sections of the core questionnaire. National household component—The component of NMCUES that consisted of multiple household inter- views with an area-probability sample of people in the civilian noninstitutionalized population of the United States in 1980. Nonkey person—A person related to a key person who joined him or her after the round 1 interview but was part of the civilian noninstitutionalized population of the United States at the date of the first interview is considered nonkey. Person file—In the person round-specific file of the 12-month data base and in the person file of the Public Use files, the person is the unit of analysis. There is one record per participant in the NMCUES person files. PID #—Participant identification number, a unique number assigned to a person for the duration of the survey. Proxy respondent—As used in this survey, a proxy respondent was a person who provided information for people in the reporting unit but who was not a member of the reporting unit. A proxy respondent was used only when no member of the reporting unit could supply the information because of physical or mental incapacity. Public Use Data Tapes—The Public Use data base consists of six files, constituting the final, cleaned, refor- matted, and imputed version of the NNMCUES data. The ''six files are the Person file, the Medical Visit file, the Hospital Stay file, the Dental Visit file, the Condition file, and the Prescribed Medicine and Other Medical Expense file. Reporting unit—The basic unit for reporting data in the household components of NMCUES. A reporting unit consisted of all related people residing in the same housing unit or group quarters. One person could give information for all members of the reporting unit. Round—tThe administrative term used to designate all interviews that occurred within a given period of time, and that used the same instruments and procedures. Skip—Where appropriate, skip instructions are writ- ten into the NMCUES questionnaire for the interviewer’s use, to determine, on the basis of the question just asked, what question to ask next. In the case of a legiti- mate skip, the interviewer skips items that are not relevant or do not apply. If the interviewer skips an item that the respondent should have answered, then this consti- tutes an illegitimate skip. Source-of-payment variablee—NMCUES attempted to ascertain the source of payment for the total charge for medical services and supplies. An initial question dealt with the family as the source of payment, with the family defined as those persons in the reporting unit. The next two questions were about payments by sources other than the family. In addition to the family, three separate sources could be recorded. Total charge—For each service provided and supply obtained information was asked about the total charge. The total charge included the charges for every procedure performed during the visit or the charges for all supplies of the same type. The total charge was the amount billed, not necessarily the actual amount paid or accepted as payment by the provider of the care. 12-month data base files—The NMCUES data went through several stages of data processing before a final version was produced as the Public Use Data Tape. The 12-month files are an intermediate stage, containing 10 data files, both event-level files and person-level files for each round. The 12-month files were cleaned minimally and were not subject to imputation of missing data. 33 ''Appendix Il. Nonresponse Variables In this section, problematic variables and their corre- sponding codebook variable names from the Medical Visit, Hospital Stay, and Flat Fee files are classified by type of nonresponse. Because more than one type of nonresponse may be present in a variable, some vari- ables appear more than once. The level of nonresponse also is indicated for each variable, in accordance with the scheme of classification. * 10—20-percent nonresponse 21—40-percent nonresponse 41-100-percent nonresponse ak kkk Type I: Nonresponse as High “Don’t Know’s” For some variables, there is a high level of “don’t know” (consistency code 94) response. A “high level” is defined as 10 percent or more of the expected valid response (that is, the total file N minus the legitimate skips (99’s)). High “don’t know” variables are listed, by file, as follows. (Note that all the high “don’t know” variables are charge questions. ) Medical Visit MVF-NTC “Reason for no total charge”; but 94’s refer to MVF-TC, “total charge”’* First source-of-payment amount* First source-of-payment amount* MVF-SA1 MVF-AP1 Hospital Inpatient Stay HS10-1 “Reason for no total charge”; but 94’s refer to HS 10, “total charge”** HS11 Amount family paid* HS12B First source-of-payment amount* HS13B First source-of-payment amount** HS13B-1 —_ Second source-of-payment amount* AS18-1 “Reason for no total charge”; contains 94’s for AS-18, “total charge [physician]”** AS19 How much of the charge for the doctor did or will you or your family pay?* AS20B How much will each source reimburse or pay you back?** AS21B How much did or will [first source] pay?** Flat fee FF—5B How much did or will [first source] pay?* 34 Type II: Blanks (98’s) as Surrogates for “Don’t Know” In some instances, there was no provision for a “don’t know” response category, yet a high number of unexplained 98’s was recorded. In these cases, it is hypothesized that many of the 98’s are in fact proxies for “don’t know.” Medical Visit MVFFP-2 Amount family paid* (Some 98’s were ar- tifacts of disaggregation; however, 12 per- cent remained after these were subtracted.) Hospital Inpatient Stay HS11 Amount Family Paid* (Some 98’s were artifacts of disaggregation; however, 14 percent remained after these were subtracted. ) Flat Fee FF6A How many visits did [person] have to the [doctor, dentist, or other] before January 1, 19807*** FF6B Was hospital stay prior to January 1, 1980, covered?* Type III: Nonresponse as Legitimate Internal Skips In this type of nonresponse, legitimate skips in a question series, as when an additional medical condition or source of payment ceases to apply to the respondent, have been misleadingly assigned the illegitimate blank code, 98. Variables that are characterized by this kind of nonresponse are listed, by file, as follows: Medical Visit Condition variables: MVF-CN2_ Condition 2*** MVF-CN3_ Condition 3*** MVF-CN4 Condition 4*** Charge and source-of-payment variables: MVF-SOP2 Second source code*** MVF-SA2 __ Second source amount*** ''MVF-SP2 MVF-SOP3 MVF-SA3 MVF-SP3 MVF-SC1 MVF-AP1 MVF-SP1 MVF-SC2 MVF-AP2 MVF-SP2 MVF-SC3 MVF-AP3 MVF-SP3 Second source percent indicator*** Third source code*** Third source amount*** Third source percent indicator*** First source code*** First source amount*** First source percent indicator*** Second source code*** Second source amount*** Second source percent indicator*** Third source code*** Third source amount*** Third source percent indicator*** Hospital Inpatient Stay Condition variables: HS5-1A HS5-2A HS5-3A HSSC-1A HSSC-2A HSSC-3A HS6A-1A HS6A-2A Second condition*** Third condition*** Fourth condition*** Second abnormal condition*** Third abnormal condition*** Fourth abnormal condition*** Second operation*** Third operation*** Charge and source-of-payment variables: HS12A-1 HS12B-1 HS 12B-1P HS12A-2 HS12B-2 HS 12B-2P HS13A-1 HS13B-1 HS13B1P HS 13A-2 HS13B-2 HS13B2P AS20A-1 AS20B-1 AS20B-1P AS20A-2 AS20B-2 AS20B2P AS21A-1 AS21B-1 Second source code*** Second source amount*** Second source percent indicator*** Third source code*** Third source amount*** Third source percent indicator*** Second source code*** Second source amount*** Second source percent indicator*** Third source code*** Third source amount*** Third source percent indicator*** Second source code*** Second source amount*** Second source percent indicator*** Third source code*** Third source amount*** Third source percent indicator*** Second source code*** Second source amount*** AS21B-—1P Second source percent indicator*** AS21A-2 Third source code*** AS21B-2 Third source amount*** AS21B2P _ Third source percent indicator*** Flat Fee Charge and source-of-payment variables: FF—4B1 and FF-4B1P Second source amount and percent indicator*** FF—4B2 and FF-4B2P Third source amount and per- cent indicator*** FF—5B1 and FF—5B1P Second source amount and percent indicator*** FF—5B2 and FF-SB2P Third source amount and per- cent indicator*** Type IV: Nonresponse as an Artifact of Question Disaggregation In this type of nonresponse, an illegitimate blank code (98) was used as the marker in one part of a disaggregated question for a legitimate response in the other part. The “nonresponse” disappears when the ques- tion is reaggregated and assumes its original form. Medical Visit Charge and source-of-payment variables: MVF-TC Total charge*** MVFFP-2 Partial charge amount, family pays*** Other variables: MVFDIAG _ Diagnosis or treatment** MVF-GEN _ General checkup*** MVF-EYE _ Eye examination*** MVF-IMM — Immunization*** MVF-FP Family planning*** MVF-OTHR Other (specify)*** Hospital Inpatient Stay HS10 Total charge*** HS11 Charge, how much family pays*** AS18 Total charge, physician*** Flat Fee FF3 Partial charge*** Type V: Nonresponse as Legitimate Lack of Data For some response categories, a very small amount of data was collected. For example, the third source-of- payment and fourth condition variables listed below con- tain (with one exception—HS6A-2A) a level of valid 35 ''response of less than 1 percent of the expected number (that is, the total file number, after 99’s—legitimate skips—have been deducted from the calculation). Medical Visit Condition variables: MVF-CN4_ Fourth condition*** Charge and source-of-payment variables: MVFSOP _ Third source code*** MVF-SA3_ Third source amount*** MVF-SP3 _ Third source percent indicator*** MVF-SC3 _ Third source code*** MVF-AP3 Third source amount*** MVF-SP3 _ Third source percent indicator** Hospital Inpatient Stay Charge and source-of-payment variables: HS12A-2 Third source code*** HS12B-2 Hospital charge amount, third source*** HS12B2P Third source percent indicator*** HS13A-2 Third source code*** HS13B-2 Hospital charge amount, third source*** HS13B2P Third source percent indicator*** 36 AS20A-2 Third source code*** AS20B-2 __ Physician’s charge amount, third source*** AS20B2P Third source percent indicator*** AS20A-2 Third source code*** AS20B-2 _ Physician’s charge amount, third source*** AS20B2P Third source percent indicator*** AS21A-2 Third source code*** AS21B-—2 _ Physician’s charge amount, third source*** AS21B2P Third source percent indicator*** Condition variables: HS5-3A _ Fourth entry condition*** HS5C-3A Fourth abnormal condition*** HS6A-2A Third operation*** Flat Fee Source-of-payment variables: FF-4B-2 and FF-4B2P Third source amount and third source percent indicator*** FF-5B-2 and FF-5B2P Third source amount and third source percent indicator*** ''Le HOSPITAL OUTPATIENT DEPARTMENT VISIT VISIT A = PERSON # © 5. Why did (PERSON) visit the (CLINIC NAME) on (DATE)? CODE ALL THAT APPLY S$’) Diag. or Treat. 8! 64s . GQ) © General Checkup . . ... . 02(A) ¢ Eye Exam (glasses). . . . . 03(6) O Immunization. . ..... . 04(6) ” Family Planning ..... . 05(6) ome Other (SPECIFY) 3° 06(A) = A. Was this for any specific condition? BA STOR 0 8 eke 6 tee 6a es OL) = Meds Ses. we. 6 ee 2 = B. What was the condition? Any other condition? B Condition Cond. # @ & D ce {6) cc (6) c¢ (6) Cc (6) C. Did (PROVIDER) discover any condition? CH) Nese a ebeee 6 ee wes ete OL®@) No, . 2. 2 6 2 2 ee © 2 © © 02(6) D. What was it? Any other condition? RECORD IN B ABOVE Yes No 6. Were any X-rays taken during this visit to (NAME OF CLINIC) on (DATE)? 6 01 02 T. Were any laboratory tests taken such as a blood test, urinalysis, culture, or other kind of test done? 7 01 02 8. Was an EKG, EEG, (a pap smear) or any other diagnostic procedure done? 8 ol 02 9. How much was the total charge for this visit on (DATE), including any amounts that may be paid by 9] $ (10) health insurance, Medicare, Medicaid or other sources? (Include any separate charges for $3.00 or less ....--- ory X-rays/laboratory tests/diagnostic procedures].) No charge .. +++ ++ - . = ” Included with other charges 03(FF_ (RV)) Dont know. .. . <6 os - SECO) A. Why was there [no/such a small) charge for this visit? A | Welfare/Medicaid paid . . . 01(RV) . Included with other charges 02(FF_ (RV)) Free from provider. . . - - 03(12) Other source(s) will pay. . 04(12A) Standard HMO/PHP/Health Center charge . ...- - - 05(RV) Other = . 2. co -c.c e #2 22, OF7G0) OPD-26 Ill xipueddy WO1-] SUO}IE|ES ''8t VISIT B VISIT C VISIT D VISIT E PERSON # | PERSON # PERSON 800 os PERSOM ie a # Diag. or Treat. 26. sy OCR) Dimas OF TERRE. 6: Vcc eo OLER) 5 Diag. or Treat. . . .. . . 01(B)|Diag. or Treat. . .... . 01(B) General Checkup . . .. . . 02(A)|General Checkup .... . . 02(A) General Checkup . . .. . . 02(A) |General Checkup... . .. . 02(A) Eye Exam (glasses). . . . . 03(6)|Eye Exam (glasses). . . . . 03(6) Eye Exam (glasses). . . . . 03(6) |Eye Exam (glasses). . . . . 03(6) Impminizvations .. s ..«; .s 6) O46) piemumigation. 2.6.06 208. 2’ 0406) Immunization. . ... . . . 04(6) |Immunization. ...... . 04(6) Family Planning... . . . 05(6)|Family Planning ..... . 05(6) Family Planning . ... . . 05(6) |Family Planning ..... . 05(6) Other (SPECIFY) Other (SPECIFY) Other (SPECIFY) Other (SPECIFY) 06(A) 06(A) 06(A) 06(A) NOD og ie ete’ eo naiiee iaiia Ts RE OR 65'§ 6) 'Gi esis s( ela) «wae A FOS 0s 68 6 0 ee 6. 6: He ORB) NOM oe & oR 6 + eee 2 OLED NO eee wie lee a eee cite CURRIES 6 1ek es Meh ees Oe eisete TORUE) NOs cis cogs 6 heels Oe OREO INOS bce" vi tar oe ere c's eee) Condition Cond. # Condit ton Cond. # B Condition Cond. # Condition Cond. # & Cc (6) Cc (6)} p co (6) cq (6) cc (6) cd (6) ce (6) ca (6) cc (6) cq (6) cq__(6) cq (6) c¢ (6) Cc (6) cc (6) Cc (6) WER sss eiic 6 ei we we Ww hOLM INGE 6 Dole se 6 6 6 6 6 6 oe OLD Cc MOS oes eis 6. sw ete Se A PCOR, % ee) eters rw. as Oe Oe 9. aise is, seen lie: \6% 6 a) GL OMDENO’ OF 6) iy a aUEO Sie Je uae ah eee Da ee en ae ee ee Yes No Yes No Yes No Yes No 01 02 01 02 6 01 02 01 02 01 02 01 02 7 01 02 01 02 01 02 01 02 8 01 02 01 02 $ (10) $ (10) $ (10) 9 $3.00 or less . . . . .01(A) #3.00 or less . . .. .O1(A) $3.00 or less . . . . .O1(A) $3.00 or less . . . . .01(A) No charge... ~~ ~ .02(A) 0; CUBE BO. 75 sw, 3. 5 OZ CA) No charge . ... . . .02(A) io charge <.3. . . «' «<62@@) Included with other cluded with other Included with other Included with other charges + + + + + + -03(FF_(RV)} charges... . . . .03(FF r_(RV)) charges . . .. . . .03(FF_(RV)} charges. .... . .03(FF_(RV)) Don't know. . . . . . -94(10) ‘ OO OD De i a Oe Riess ess « misice sae» 9 we OPD-29 ''HOSPITAL OUTPATIENT DEPARTMENT VISIT 15. Of those (ANSWER TO Q.14) visits, how many cost the identical amount as the visit we just LS talked about? VISIT A PERSON eee Visits (16) pe Visits included in same FF_ (17) NORG@: ie es « iets « OCS BOX) 16. Of those (ANSWER TO Q. 15) visits, how many were paid for in the same way as the visit you 16 ET beteits (17) just told me about? None... . . s - » « «+ %: + 00S BOX) 17. How many of the (ANSWER TO PREVIOUS QUESTION) visits did not include any X-rays, lab tests 17 ‘ Visits(18) or diagnostic procedures? None ...-... . se « OO(S BOX) 18. Not counting the visit on (DATE) you just told me about, what were the dates of the other 18 1) 6) / (ANSWER TO Q. 17) visits? Month / Date Month / Date 2) 7) Month / Date 3) l Month / Date 4) / Month / Date f Month / Date 8) Month / Date 9) / Month / Date 5) / 10) / Month Date Month Date s CODE ONE: BOX HHS Sample ...... . O1(NV) SMHS Sample. . . . . . . 02(19) 19 Street: 19. What is the complete address of the hospital clinic or City: outpatient department? y: State: Zip: % FA - ' IF MEDICAL DOCTOR SEEN (SEE Q.4), ASK Q's. 20 & 21. 20 Name® 20. What is the name of the doctor (PERSON) saw? Don't know. ...... . 94(NV) 21. Does (DOCTOR) have an office outside the hospital? 21 Yeseet 2. Ne eee ee | OLGA) NO oe «+. uh 6. do 4 a. 0. aida... OLEND. Don't know. ...... . 94(NV) A. What is the complete address of the doctor's office? A Place: Street: City: State: Zip: NEXT VISIT OPD-30 Iv ''cy VISIT B VISIT C VISIT D VISIT E PERSON i PERSON # PERSON # PERSON # PS Visits (16) 4 Visits (16) 15 Pd Visits (16) fa] Visits (16) 1 Visits included in 4 Visits included in a Visits included in eg Visits included in same FF_ (17) same FF_ (17) same FF_ (17) same FF_ (17) Woe. 34 es a OOS BON) home aoe ce 6 ee. 0Q(S BOX) NONE. S058) isis wives MOUS BOX TMoOne 0 oo ea cin ea 6 GO(S BOX) Pe} Visits (17) P<] Visits (17) 16 Dat Visits (17) ft Visits (17) BONE 6. ee. ORES BOX) Phone 2400 8 6. aie. 6 <', OOCR RE) Nene ry yy » « «OOS BOX) None hes 6c 25 kes OOS, BOX) [ | visits (18) [| visits (18) 17 Visits (18) ee Visits (18) ONO ie LT ie ie a MOUS BOM) A ees A (er 00 (BOR) NONE 6 oie 5) cg 8h wd OOLS: BOR NoneLi is Je. < cease «. - 00(8 BOX) 1) L 6) 1) / 6) 18 }1) l 6) l 1) / 6) l Month / Date Month / Date Month / Date Month / Date Month / Date Month / Date Month / Date Month / Date 2) 1 7) L 2) 1 7) l 2) L 7) f 2) J 7) If Month / Date Month / Date Month / Date Month / Date Month / Date Month / Date Month / Date Month / Date 3) 8) 1 3) 1 8) J 3) 8) 3) / 8) Month / Date Month / Date Month / Date Month / Date Month / Date Month 7 Date Month / Date Month / Date 4) / 9) i. 4) l 9) L 4) l 9) 4) / 9) / Month / Date Month / Date Month / Date Month / Date Month / Date Month / Date Month / Date Month / Date 5) 10) Vi 5) l 10) J 5) Sop 10) 7 5) 10) l Month / Date Month / Date Month / Date Month / Date Month / Date Month / Date Month / Date Month / Date Street: Street: 19 Street: Street: City: City: City: City: State: Zip: State: Zip: State: Zip: State: Zip: Name: Name: 20 |Name: Name: Don't know. . oie «ie: 9 SMUD Dont RnOW 66 6 Sie - . 94(NV) Don't know. . - « « 94(NV) [Don't know. ...... ~ 94(NV) WEBS ie iia ee aie 0 ate. Whe? RD AVOR esse. Co ma i, oe ORCAS Bh Re ia en ei. (526 ORR) TOMS oa as Singing ier Gigi 6 OLOD DO See a ee oie e. ‘sike CRUD HIM eS ee ee GOR EMY) OY 6 gee ee sic 6 it ORD, TOUR eule, oie “se. een: 6 ODEN Don't know. . se ee « SACAV) [Don’t know. .... .'. .s SERV) Don't know . . » « » « 94(NV) [Don’t know... ..... 94(NV) Place: Place: A Place: Place: Street: Street: Street: Street: City: t City: City: City: State: Zip: State: Zip: State: Zip: State: Zip: OPD-31 ''eV HOSPITAL STAY (INPATIENT) Person Name: # Hospital Stay # You told me that (PERSON) was a patient in a [hospital/nursing home] (NUMBER) times since (REF. DATE). 1. When did (PERSON) enter the [hospital/nursing home] (the [first/ next] time)? Month / Date ! Year A. When did (PERSON) leave the [hospital/nursing home] (that time)? i Month / Date / Year Still there °<6 +.) 6g) 6 6s. ~ 0103) IF COMPLETE DATES GIVEN IN QUESTIONS 1 & 1A SKIP TO QUESTION 3 2. How many nights was (PERSON) in the [hospital/nursing home]? nights None....s “s&s 0 © « » - . 00(4) 3. Were these days in the [hospital/nursing home] includea in the number of days (PERSON) spent in bed that you told me about earlier in the interview? WOR ee ce A ei ee Le, 6% OE Neg 0: 6 cece. oo apne pices. 0 Fae Ul CADD THESE DAYS TO Q. 1A IN DISABILITY 5. For what condition did (PERSON) enter the [hospital/nursing home]? Was there any other condition? CONDITION COND. # (6) (6) (6) (6) Delivery.’ . 8% s%ei0 *« O1CA) Newborn baby... . . 02(B) Other. oo 6 45 te 6 ORB) A. FOR DELIVERY, ASK: Was this a normal delivery? MO 6s 6s oteey~es 5 GEOG) NG ees ee oe - 02(C) B. FOR NEWBORN, ASK: Was the baby normal at birth? Yes. 6 5 seed - 01(6) No . 6 6 © © © © «© © O2(C) C. What was the matter? cc cc DAYS SECTION AND REASK ALL QUESTIONS IN DISABILITY SECTION) 4. What is name and address of this [hospital/nursing home]? Name: Street: City: State: Zip: HS-32 6. Were any operations performed on (PERSON) during this stay in the [hospital/nursing home]? VOW oi gtd os oes x hee A) NOess oases 3. tee s OZ) A. What was the name of the operation? IF NAME OF OPERATION IS NOT KNOWN, DESCRIBE WHAT WAS DONE. Were there any other operations during this stay? Name: Name: Name: '' 12. Do you expect any source to reimburse or pay you back? OA ie tis 4 * ie) 0 ie eR NO ac eee Or Ron) A B Who will reimburse or pay you How much will (EACH SOURCE) back? ENTER BELOW. reimburse or pay you back? Anyone else? . SOURCE AMOUNT $ z $ zi x C | CODE ONE: pox | TOTAL CHARGE PAID INQ. ll. ..... . 01(14) PARTIAL OR NONE PAID INQ. 11... . . 02(13) 13. Did or will anyone else pay for this hospital stay? WS osha rig ert a Sig -01(A) INOS we! cs ea ee 6 eee ee ED A B Who else paid or will Pay any How much did or will (EACH part of the charge for this SOURCE) pay? stay? ENTER BELOW. Anyone else? SOURCE AMOUNT $ z $ % $ % aa HOSPITAL STAY (INPATIENT) = Yes No 7. Were any X-rays taken during this [hospital/nursing home] stay? 01 02 8. Were any laboratory tests such as a blood test, urinalysis, culture or other kind of test done? 01 02 9. Was an EKG, EEG, (a pap smear) or any other diagnostic proce- dure done? 01 02 td IF STILL IN HOSPITAL, GO TO NEXT HOSPITAL STAY OR NEXT SECTION. 10. How much was the total (hospital/nursing home] charge for this stay, including any amounts that may be paid by health insurance, Medicare, Medicaid or other source? (Include any charges for [X-rays/laboratory tests/diagnostic procedures], but) do not include separate charges for doctors or surgeons. $ (11) No*€haree. 20 53s. -02(A) Included with other charges. .03(FF (14)) FOR NEWBORNS ONLY: Included in mother's bill (Person # Pierue -04(15) Dont: knoe eae -94(11) A. Why was there no charge for this hospital stay? Welfare/Medicaid paid. -01(14) Included with other charges. .02(FF (14)) Free from provider . . -03(13) Other source(s) will pay . -04(13A) FOR NEWBORNS ONLY: Included in mother's bill (Person # Ps seeits lune -06(15) 11. How much of the (CHARGE) charge for the stay did or will you (or your family) pay? Partial $ z TOGaE Charge ee a MODE cia: Fb et av ania ie -00(C BOX) CODE ONE: "YES" WAS ANSWERED IN Q. 7, 35 QR 9 . ss O2G4) "NO" WAS ANSWERED IN Q. 7, 8, ANDO... -02(15) 14. How much were the charges for the [X-rays/laboratory tests/ diagnostic procedures] ? $ (15) Don't know or no separate CHALREs 0. seas ee CSL HS-33 ''Sv HOSPITAL STAY (INPATIENT) DOCTOR A 15. Were there any doctors or surgeons who treated (PERSON) and from whom there was a separate Name or Type charge? MOS 6S ee 8s 01(A) No. 2 2 es ots O2EB) A. What are the names of all the doctors or surgeons who treated (PERSON) and from whom 15 there was a separate bill? ENTER EACH NAME IN SEPARATE DR. COLUMN. IF MORE THAN ONE A DOCTOR IS INCLUDED IN A SINGLE CHARGE, LIST ON SEPARATE LINES IN ONE DOCTOR COLUMN. & Cc B. Were there any other doctors who treated (PERSON) such as anesthesiologists, path- ologists, radiologists, or psychiatrists from whom there was a separate charge? Yes . . - - « - O1(CC) Nog se ate 3 O2 CLG) Cc. Who was that? ENTER NAME OR TYPE OF DOCTOR IN NEXT AVAILABLE DR. COLUMN(S). 16. CODE ONE: DOCTOR(S) REPORTED INQ. 15 .......2..-.-.-. - - O1(17-22 FOR EACH DOCTOR) NO DOCTOR(S) REPORTED IN Q. 15. . . .....-.+ +... ~ © O2(NEXT STAY OR NEXT SECTION) CODE OR ASK: 17. What type of doctor is (NAME OR TYPE)? 17 | General Practitioner . 01 Anesthesiologist . . . 02 Cardiologist .... . 03 Internist. ..... . 04 OB/GYN 60 hs Sas eS OS Ophthalmologist. . . . 06 Orthopedist. . ... . 07 Pathologist. . ... . 08 Pediatrician .... . 09 Psychiatrist .... . 10 Radiologist. ..... 11 Other (SPECIFY). . . . 12 18. How -much was the total charge for (DOCTOR) including any amounts that may be paid by health 18] $ (19) insurance, Medicare, Medicaid, or other sources? Included with other charges. . . . . - ~ OL(FF_ (22)) Don't know... . . . 94(19) 19. How much of the (CHARGE) for the doctor did or will you (or your family) pay? 19} Partial $ z HS-34 Total charge ..... Ol None s. 686 P's ees OZ BOX) '' DOCTOR B DOCTOR C DOCTOR D DOCTOR E Name or type Name or type Name or type Name or type iD A & c General Practitioner . 01 General Practitioner . 01 7 General Practitioner . 01 General Practitioner . 01 Anesthesiologist . 02 Anesthesiologist . . . 02 Anesthesiologist . . . 02 Anesthesiologist . . . 02 Cardigiogist . 66%, «108 Cardiologist .. «303 Cardiologist... . . 03 Cardiologist ..... 03 TORGRMESE snc 0). 8) 6 Oe THEORIES 65s - 04 TREGROE Ris 6) 0. 6 her 6 OR gntermist. . ...°. . . 06 ark atin! in) 5! OD DBP GUN, oct is ecu” se GRIGIN ee a is 6 OS PGI on er 6 Oe Ophthalmologist. . . . 06 Ophthalmologist. . .. 06 Ophthalmologist. . . . 06 Ophthalmologist. . . . 06 Orthopediat. 3.04.6) 4 6: OF Ortnopeiier. 2.6... < OF OFrthopedtets) oi 65... OF Orthopedist. 2. 2+ . OF Pathologist. ... +, ae Fecholpetet.. 2.5.5... «ee Pathologist... .. . 08 Pathologist. .. . . . 08 Pediatrician... . oF wedsatrician . «) . 4's. 09 Pediatrician +... 5 . OF Pecsatrician . . .)«'..°09 Psychiatrist . . oh weyeutacrist 6.) 6s 6: 20 Peycnsetriet 2°. .-s . 30 Psychiatrist . . . . < 10 MACTOLORL GE iG i gig) ne PACLOLORI OE. oO ee EE MBOLOLOUSOL. (5-6 4%. EE Radiologist. ..)...° .%. il Other (SPECIFY). 12 Other-(SPECIFY)... 6.6/6 42 Other (SPECIFY): .°. .°22 Other (SPECIFY). . .. 12 $ (19) $ (19) 1g |$ (19) $ . (19) Included with other Included with other Included with other Included with other charges. . . .. . . O1(FF_(22))] charges. . «/ SOL(FF (22) charges. , . . . - . OL(FF_(22)] charges. . .... . O1(FF_ (22) Don" t know. 6) 945 526°. 946019) Don't know... . 94(19) Don’t know... .. . 94(19F7 Don't know... . . . 94(19) Partial $ a Partial $ an 19 |Partial $ z Partial $ z Total charge’, << °°s OL Total charge... .01 Total Charee so. 4. OL Total charge... . . O1 NOME sk ae c's . . O2EC BOX) NOG 05, % s+ 8 ce O2Ce BORD HS-35 MONG ea se oe 6 Ole BOD WONG 6 sais sicerie 6 ¢ C2KO BOX) ''Lv HOSPITAL STAY (INPATIENT) 20. Do you expect any source to reimburse or pay you back? 20 DOCTOR A Mes Goes 6? eeWargere eer 01(A) Dies eg ER 02(C BOX) A. Who will reimburse or pay you back? ENTER UNDER "SOURCE". Anyone else? A & SOURCE [ AMOUNT B. How much will (EACH SOURCE) reimburse or pay you back? ENTER UNDER AMOUNT. B s $ $ Zz CODE ONE: Cc GODE ONE: pox | JOTAL CHARGE PAID IN Q. 19. os Seeiat 5: Ge “4. ie PARTIAL OR NONE PAID IN Q. 19. " ee 21. Did or will anyone else pay for this doctor's charge? 21 }XOB es 2 6 ue « 0: ee © O1(A) BHO: etic: nig ler ren ging ag wighny: ORCS) BOR) A. Who else paid or will pay? ENTER UNDER "SOURCE". Anyone else? A SOURCE | AMOUNT & $ z B. How much did or will (EACH SOURCE) pay? ls % Is % s HHS SAMPLE. ...... . O1(NEXT DR.) BOX | suis SAMPLE... .. . . . 02(D BOX) D CODE ONE: BOX INDICATE IF DOCTOR'S NAME IS KNOWN. | Name known . . fas OUD BY DOCTOR'S NAME NOT KNOWN. Name not known... . .02(NEXT DR.) 22. Does (DOCTOR) have an office outside of the hospital? OM i ee Oe eee 0 0 8 0 let ss « OZCNERT DR.) A. What is the complete address of (DOCTOR'S) office? A [Place: treet: ity: tate: Zip: GO TO NEXT DOCTOR AFTER ASKING FOR ALL DOCTORS, GO TO NEXT STAY. IF NO OTHER STAYS, GO TO NEXT SECTION. HS-36 ''8V DOCTOR B DOCTOR C DOCTOR D DOCTOR E Nés 2. . 01(A) eg gg Lv . 01(A) 20 Wee ua get On En Nee eae ee oe OLS Ne, 4 - 02(C BOX) NO er sikelele iat a: Ce pore OR. BORD . IND oe mire dine sso OeCCu BOX) NO 6 eae so ce eRe Bom) SOURCE [AMOUNT SOURCE AMOUNT 7 & SOURCE AMOUNT SOURCE | AMOUNT B Is % $ $ % ls 2% ‘ys 3 $ oe bs § eo) 4 2 Total Charge Paid . . 01(S BOX) Total Charge Paid . . 01(S BOX) C Total Charge Paid - . 01(S BOX) Total Charge Paid . . 01(S BOX) Partial or None Partial or None BOX |Partial or None Partial or None Paid. - 02(21) Paid...» » 02 (21) Paid. Cares. s . ORT MA a ro gi eae 02(21) Yes . - 01(A) NR a seig ig, > ORK) Dae CANORA CLG! age elt a ex GOON) OR a tye, «be 6. 6 OLR No. - 02(S BOX) WN ecg eww at aie gt a RE ROM) ms Wee ea ee PUOEAS BOR) NG: iy aitrlite | ok ee ORES BOX) SOURCE AMOUNT SOURCE AMOUNT A SOURCE AMOUNT SOURCE AMOUNT $ % $ & $ % $ z : $ $ % $ Zz $ B $ $ Zz Name known. . ... . 01(22) (Name known. . ... . 01(22) Name known. . ... - 01(22) Name known. . ... . 01(22) Name not known. . . . 02(NEXT DR.)|Name not known. . . . 02(NEXT DR.) Name not known. . . . 02(NEXT DR.) Name not known. . . . % (EXT DR.) a cn ie i, a . 01(A) os Mee 30 PT OR Me aces ss hs ea wee 02 (NEXT a be a a eee <).6 6 OZ CRERT DR.) Ms wee ew xe SG - - O2(NEXT Bie i oe «6: e:¢ « « O2(NEXT DR.) Place: Place: A |Place: Place: Street: Street: Street: Street: City: City: City: City: State: Zip: Zip: State: Zip: ee Zip: HS-37 ''4 MEDICAL PROVIDER VISIT Person Name # [Besides the visits we already talked about/You told me that (PERSON) had seen a medical person (NUMBER) times since (REF. DATE) .] 1. On what date did (PERSON) [first/next] see a medical person? MONTH / DATE 3. 2. Where did (PERSON) see the medical person on (DATE), at what type of place -- was it a clinic, hospital, doctor's office, or some other place? IF CLINIC, ASK: Was it a hospital outpatient clinic, a company clinic, or some other kind of clinic? Doctor's office or group practice.01 DEGtOr Ss Clinics. oie 2 0 0 re 202 Neighborhood/Family Health Center.03 Company clinic ......... .04 School clinic. ......... .05 IF SOME OTHER PLACE, ASK: Other clinic .......... .06 Where was this? MMO 2 oe Se yk hee be ee ee ce LabOratorys oo. eh 6 ee ee eo «08 Hospital outpatient clinic, hospital inpatient, emergency room. . . - 2 e © e © «© «© «© «© © © OFC INSTRUC- TION BOX) Other (SPECIFY) .........410 MAKE SURE A HOSPITAL STAY, EMERGENCY ROOM OR HOSPITAL H'NSTRUCTION | OrpaTIENT VISIT HAS BEEN COMPLETED FOR THIS DATE. BOX INVALIDATE THIS PAGE AND GO TO NEXT VISIT. A. What is the name of the medical‘person (PERSON) saw on (DATE)? Provider's Name What is the name of the medical place (PERSON) went to on (DATE)? In what city and state is it located? Place Name City / State MV-38 Did (PERSON) see a medical doctor on that visit? NOGith 4 etki 6! alte! oi: OH. ig ORCA) DO Pe Wee wee 8 et wt 6 cate ORCC) Don't know... . «4... ee se -94(5) Is the doctor a general practitioner or a specialist? General practitioner ..... . .01(5) Specialist ......... . . .02(B) Don't know os gees 8 6 8 ew 8 nw 9465) What is the doctor's specialty? Cardiologist. . . .01(5) Tntéemiist> << oss). .02(5) OB/GYNG 3. we S03.) Ophthalmologist . .04(5) Orthopedist. . .05(5) Pediatrician . .06(5) Psychiatrist . .07(5) Other (SPECIFY) .08(5) What type of medical person did (PERSON) see? Chiropractor. . . .01(5) Podiatrist. . .. .02(5) Optometrist . .. .03(5) Psychologist. . . .04(5) Social Worker. .05(5) Nutse. 0. <06@) Phy. Therapist .07(D) Other (SPECIFY) .08(D) Does (MEDICAL PERSON) work for or with a doctor? VO6. ie. a ce whoa ee 0 0 OL NO oak A os ee eee ts os ee Don't:-know ss. Gi ee oe oe JOH ''os MEDICAL PROVIDER VISIT 5. Why did (PERSON) visit (PROVIDER) on (DATE)? CODE ALL THAT APPLY. /9. How much was the total charge for this visit on (DATE), including any amounts that may be paid by health insurance, Medicare, Medicaid, Diag. or treatment.01(B) Immunization . .04(6) or other sources? (Include any separate bill for [X-rays/laboratory General checkup . .02(A) Family Planning.05(6) tests/diagnostic procedures ].) Eye examination Other (SPECIFY) .06(A) for glasses , . . .03(6) $ (10) SOCMO OT LOB Be eis Serie igs sk er MAY A. Was this for any specific condition? ‘No CRALEB His os ee ie eis ee ORCA) Yes 01(B) Included with other charges. . . 03(FF (RV) ) WO 66 io3 64 baw Sr ar te of Mpls ORKO) Don't know... . 1... ~ 94(10) B. For what condition did (PERSON) visit (PROVIDER) on (DATE)? A. Why was there [no/such a small] charge for this visit? Any other condition? Welfare/Medicaid paid. . .. . . 01(RV) COND. # Included with other charges. . . 02(FF (RV) ) Free from provider ...... . 03(12) Other source(s) will pay . .. . 04(12A) Standard HMO/PHP/Health Center CHETEO 6) iitie) 6) .6 Se ie 6 6. es OO URV) C. Did (PROVIDER) discover any condition? DERE som peeled ae gee OTM @ ca, le hanes ieiiCg (e's mahi: we ae ae a sees 10. How much of the (CHARGE) charge for the visit did or will you (or adie ach eiugh ergy nee we faelids saat — — D. What was it? RECORD IN B ABOVE. Any other condition? Partial $ z NOtel (ChAwSe 632005 ee eine» veg HOE Yes No 6. Were any X-rays taken during 7 NONE 2 Sikes 9 wie. 94.6.6) «. «6. OOK: BOX) this visit on (DATE)? 01 02 11. Do you expect any source to Yes . . . 01(A) Ms . lace ae sd pag reimburse or pay you back? No. . . . 02(C BOX) > > culture, or any other kind A. B. one cont ™“ - Who will reimburse or pay How much will (EACH 8. Was an EKG, EEG, (a pap smear) oe Sees oe a SOURCE) reimburse or any other diagnostic procedure Anyone elset or pay you back? done? 01 02 SOURCE AMOUNT Mv-39 ''I¢ MEDICAL PROVIDER VISIT 16. Of those (ANSWER TO Q. 15) visits, how many were paid for in the same way as the visit you just told me about? Cc CODE ONE: pox | TOTAL CHARGE PAID INQ. 10. . . . . 01(RV) visits (17) PARTIAL OR NONE PAID IN Q. 10. . . 02(12) vo i None. . . . .00(S BOX) 12. Did or will anyone else pay for this visit? Nea. we's « «: « OL) L7. How many of the (ANSWER TO PREVIOUS QUESTION) visits did not DOs sias eee eerie ree I ZERY) include any X-rays, lab tests, or diagnostic procedures? A. B. visits (18) Who else paid or will pay How much did or will oe ee ey any part of the charge? (EACH SOURCE) pay? None. . . . .00(S BOX) ENTER BELOW. Anyone else? SOURCE AMOUNT 18. Not counting the visit on (DATE) you just told me about, what $ z were the dates of the other (ANSWER TO Q. 17) visits? $ x 1) f 6) / 11) Month / Date Month / Date Month / Date IF PERSON HAS FEWER THAN 5 ADDITIONAL VISITS TO A MEDICAL 2) [ 7?) [ 12) f PROVIDER, GO TO S BOX. Month / Date Month / Date Month f Date pv | IF PERSON HAD 5 OR MORE ADDITIONAL VISITS TO MEDICAL PROVIDER, 3) { 8) L 13) __f CHECK Q's. 6, 7 & 8, CODE BELOW. Month / Date Month / Date Month / Date "YES" WAS ANSWERED IN Q. 6, OR 7 OR 8. -01(S BOX) / f) / "NO" WAS ANSWERED TO ALL QUESTIONS , . .02(13) 4) 9) 14) Month / Date Month / Date Month / Date You mentioned that (PERSON) had (NUMBER) medical visits. J / / 5) 10) 15) 13. We have already talked about (NUMBER) of those visits. How many Month ?¢ Date Month / Date Month / Date of the remaining (REMAINING NUMBER) were also to (PROVIDER/PLACE)? visits(14) ‘CODE ONE: ROM: ++ 24 2 + 2OOCS BOR) ae Ws SAMPLER. ww ww ok oe woe LOY) SMHS SAMPLE .......-..-.- -02(19) 14. Of those (ANSWER TO Q. 13) visits, how many were also for (CONDITIONS) ? 19. What is the complete address of (PROVIDER/PLACE)? visits(15) Place: NODC 6) 284. eis - - .00(S BOX) Street: 15. Of those (ANSWER TO Q. 14) visits, how many cost the identical amount as the visit you just told me about? City: visits (16) iia State: $ ____sviisits included in FF___(17) pec P MOGB.. 5 ates 6 82s 00(S BOX) NEXT VISIT Mv-40 ''cs FLAT FEE SECTION IF A FF HAS PREVIOUSLY BEFN REPORTED FOR RU, ASK Q. 1. OTHERWISE, ENTER "A" IN COLUMN, CODE "FF" SECTION, AND CONTINUE. ee Is this [visit/hospital stay/service] included in a charge you already told me about, (either 1 {Flat Fee Letter: in a previous interview or) today? Yes. 2. 6 « «© © © «© © «© « © « « Which FF was that? (ENTER FF LETTER AT QUESTION WHERE FF Person # WAS REPORTED. DO NOT RECORD ON THIS FF PAGE.) NOs ee ater iy er POW Re ee (ENTER FLAT FEE LETTER AND PERSON NAME AND # AND CONTINUE.) CODE TYPE OF VISITS/SERVICES COVERED BY FLAT FEE. PROBE, IF NECESSARY, TO DETERMINE MOST FF |Orthodontia......... .02 SPFEOPRLASE DESCRIPTION. : Other dental care)... suck. 02 BUTKICRL CaRe teri yi ete S, 6 6 OS Physical;.therapy ss... ..... - 506 Prescribed medicines. . .. . .05 Tests/diag. procedures. . .. .06 Pre/poet natal care. <)3s.