- Joseph E. Fitti, National Center for Health Statistics
- Mary Grace Kovar, Centers for Disease Control
Concerns among a number of public health agencies and individuals about the increasing proportion of older people in the United States population led, as early as 1980, to recommendations that the National Health Interview Survey (NHIS) address this special subgroup. Issues dealing with the health and functional status of older people and the need for alternatives to institutionalization as the mode for providing care were identified at this early point by professionals in the field of aging. Information about these and related characteristics of the older population was needed.
Statements of the need for this information were made by the Department of Health and Human Services in the 1980 National Long Term Care Plan of the Division of Long Term Care Policy, Office of the Assistant Secretary for Planning and Evaluation; by the Office of Management and Budget (OMB) in its 1980 report of the Interagency Statistical Committee on Long Term Care of the Elderly; and by the 1981 White House Conference on Aging.
It was postulated that information about the health conditions that were most prevalent, about living arrangements, family and social support availability, retirement income and financial obligations, functional status and limitations, and attitudes and opinions about their own health and abilities would help in assessing the future needs of the elderly.
In addition to responding to the topic recommendation of the NHIS's Technical Consultants Panel that these informational needs about the elderly could be addressed through the NHIS, a special Supplement on Aging (SOA) in 1984 was particularly timely because the National Center for Health Statistics (NCHS) planned to conduct the National Nursing Home Survey (NNHS) in 1984 (later postponed to 1985). An NHIS survey on the elderly would provide data on the noninstitutionalized population which would complement the NNHS data on residents of nursing homes and would provide, for the first time, comprehensive data on almost the total elderly population.
The development of a supplement to help provide some of this information from a national survey of elderly people themselves began in 1982 and resulted in the 1984 NHIS SOA. The SOA was sponsored and funded by NCHS, which was responsible for survey design, data processing, and preliminary analysis. Data collection was conducted by the Bureau of the Census under an interagency agreement.
The first step in development was to determine the topics to be included. Topic suggestions were received from a variety of sources inside and outside NCHS. Suggestions from outside NCHS came in response to the topic solicitation from the Division of Health Interview Statistics and from notification to interested agencies and persons, including the National Institute on Aging, the Administration on Aging, the U.S. Senate Select Committee on Aging, the Social Security Administration, voluntary and nonprofit organizations, and experts in the field of aging.
Selection of topics from those suggested was guided by three considerations: feasibility of collecting the data in a household interview survey; comparability to data collected in other NCHS surveys, especially the planned NNHS; and the creation of a baseline data source for a longitudinal study of aging. Activities undertaken in the topic selection process included literature reviews, reviews of earlier surveys, consultation with other agencies and individuals, and participation in conferences on the aging.
During this process it was decided that the information being sought would usually be most reliably reported by the sample person himself or herself; therefore, a rule was established that the sample person would respond for himself or herself except in cases where the sample person was physically or mentally unable to respond. In these cases, an adult, preferably living in the household, would be accepted as proxy.
The end result of this process was a pretest version of a questionnaire. Two pretests were conducted in the development of the SOA. Approval of OMB was requested in two submissions, the first covering the first pretest, and the second covering the second pretest and the full 1984 SOA. The first pretest (n=256) was conducted in June 1983 in Bradenton, Florida, which was selected because it is a popular location for retirement. The questionnaire was revised extensively on the basis of the Bradenton experience, and a second pretest was conducted in September 1983 in Wilmington, Delaware (n=234); subsequently, further revisions were made, resulting in the final form of the SOA questionnaire.
The NHIS sample is designed to produce national estimates for the civilian noninstitutionalized population residing in the United States. The United States is first divided into geographically defined primary sampling units, which cover the 50 states and the District of Columbia. These units are grouped in strata with similar characteristics, and in 1984 one unit was selected from each stratum. Within each selected unit, compact clusters of housing units are then selected for the sample. Because the sample is clustered, it is not a "simple random sample."
In 1984, 41,471 eligible households were in the NHIS sample, and interviews were conducted in 39,996 (96.4 percent), including 105,290 persons of all ages residing in the households at the time of interview. Details of the sample design for the 1984 NHIS are published in Vital and Health Statistics, Series 1, No. 18 and Series 10, No. 156.
Because the SOA intended to obtain information on both those who were currently elderly (over age 65), and to provide baseline information for longitudinal studies of persons who would become elderly in the next ten years, it was decided to include persons 55 years of age and older. However, because there are large numbers of persons 55-64, it was not necessary to include all NHIS sample persons in that age range to achieve the desired level of statistical precision. Therefore, the sample for the SOA consisted of all NHIS sample persons 65 years of age and over, and a systematic sample of 50 percent of NHIS sample persons 55-64 years of age.
A total of 16,347 persons were selected for the SOA, for whom 16,148 interviews were completed, for a 96.7 percent response rate; in more than 90 percent of the completed interviews the sample person was the respondent. The effective overall response rate for the NHIS-SOA (the product of the response rate on the NHIS and the SOA) was 93.2 percent.
The NHIS is designed to make national estimates of statistics; therefore, the data must be weighted to inflate the sample numbers to national estimates. The weights indicate the estimated number of persons in the population represented by each sample person, and depend (most importantly) on the probability of selection, the response rate in the sample segment, and an adjustment to independent estimates (from the Current Population Survey) of 60 age-sex-race population groups. In addition to these NHIS adjustments, the SOA had additional adjustments for nonresponse on the SOA and for the 50 percent subsampling rate for persons 55-64 years of age. These weights are included on the public use sample tapes, and should be used in estimating population statistics from the sample.
Because the SOA sample is large and its design is efficient, its estimates of population statistics are generally very precise. Vital and Health Statistics, Series 10, No. 156 includes instructions and examples for estimating sampling errors for estimates from the 1984 NHIS. Those procedures may be applied directly to estimates from the SOA for persons 65 and over; for persons 55-64, an additional step is necessary to account for the 50 percent subsample-multiplying variance estimates by 1.4. On average, the precision of estimates from the SOA is nearly as good as from a simple random sample of the same size: standard errors of estimates from the SOA would average less than 25 percent greater than from a single random sample of the same size.
IV. DATA SOURCES
Data were collected in personal interviews. Almost always the respondent was the sample person, and almost always the interviews were face-to-face; however, proxy respondents and telephone interviews were permitted in certain circumstances when necessary to complete interviews.
V. DATA COLLECTION PROCEDURES
Interviewing for the 1984 SOA was conducted by the Bureau of the Census, Field Division, in the standard face-to-face interviewing procedure for conducting the NHIS. The interviews period for the 1984 NHIS and SOA was January 9, 1984 through January 6, 1985, with interviewing conducted weekly throughout the year.
NHIS interviewer training is conducted by the Bureau of the Census and consists of two types: initial training for new NHIS interviewers; and annual training for experienced interviewers on the current year's special procedures and special topics questionnaires. The initial training for NHIS interviewers takes about a full week and includes about one and one-half days for any supplement to be covered; this was modified to give additional time for training new interviewers on the SOA. Training on the SOA for experienced interviewers consisted of one and one-half days of classroom sessions in January, and portions of a later one day training session and five hours of home study materials.
A number of measures are taken in the NHIS to control the quality of data collection. All inter- viewers are observed in the field by supervisory personnel at least twice each year, with new interviewers and interviewers with problems being observed more frequently. Interviewers are required to edit all completed interview schedules before submitting them to their supervisors in the Census Regional Offices, where additional edits are done; this editing is more intensive in the early part of a data collection year and for new interviewers. Approximately 5 percent of all interviews are designated for re-interview. The re-interview serves as a check on interviewer performance and as a measure of the reliability and accuracy of the NHIS and SOA data. For each household designated for re-interview, a subset of questions is asked by telephone by the interviewing supervisor.
VI. DATA PROCESSING
Quality control measures also are taken during the processing of data by NCHS staff. Quality control of the coding of questionnaire information consists of recoding 10 percent of all questionnaires by two independent coders. The quality of machine keying is maintained by a 100 percent independent key verification of all items in the questionnaires. After the data are on tape, computer edits are performed in the preparation of final data tapes. The computer edit checks for inconsistencies and invalid responses, provides algorithms for imputation, and generates recodes. The specifications for computer edits for the SOA included more than 350 decision logic tables.
VII. PUBLIC USE FILES
The SOA data tapes are in EBCDIC format and contain the SOA interview information with the following record structure: (1) a file of person records containing, for each person for whom an interview was completed, all items in the NHIS basic questionnaire that are on the person file, weights, all items in the SOA questionnaire (except the items used to permit matching to the National Death Index), special recodes, and selected condition and utilization information; and (2) a file of condition records, containing all conditions mentioned in the SOA interview plus any condition for the individual that is related to a "limited activities" status from the basic NHIS questions. To protect confidentiality, no items are included on the data tape which would uniquely identify an individual respondent. The detail of the content, coding, and structures of these two SOA data record types is contained in the public use tape documentation.
VIII. PHYSICAL TAPE SPECIFICATION
For the person tape DSN=HISY1984.AGINGPER, LRECL=1001, BLKSIZE=31031, and COUNT=16148. For the condition tape, DSN=HISY1984.AGINGCON, LRECL=380, BLKSIZE=31920 and COUNT=46320. Both tapes are standard label and are available in either 1600 BPI or 6250 BPI.
Preliminary reports based on data collected in the first six months of 1984 have been published in the NCHS series ADVANCE DATA, Numbers 115, 116, 121, 124, and 125. They focus on demographic and general health measures (115), persons living alone (116), urinary problems (121), community services (124), and vision and hearing impairments (125). Comparable analyses on the same topics were done using data for the full 12 months and presented in a special session at the 1986 meetings of the Gerontological Society of America. Additional ADVANCE DATA reports, based on the full year of data, which are soon to be published focus on functional limitations (activities of daily living and instrumental activities of daily living), and on ability to perform work related activities. A detailed analysis of the SOA data on functional limitations is in preparation for publication in Vital and Health Statistics, Series 10; and a full report on the SOA methodology entitled "The Supplement on Aging to the 1984 National Health Interview Survey," authored by Joseph E. Fitti and Mary Grace Kovar, will soon be published in Vital and Health Statistics, Series 1, No. 21. Additional analyses by staff of NCHS and other Federal agencies on a variety of topics are underway or planned. More than 30 public use data tapes have been sold by NCHS to researchers in universities and research centers, and the National Institute on Aging is encouraging applications for grants to conduct research using the SOA data.
Public use data tapes from the SOA may be purchased form NCHS for $275, which includes two data tapes (described above) and all documentation. Orders should be mailed to the Division of Health Interview Statistics, National Center for Health Statistics, Center Building, Room 2-44, 3700 East-West Highway, Hyattsville, Maryland 20782, with a check made payable to the U.S. Department of Health and Human Services for Statistical Studies. The order should include an assurance that the data will be used solely for statistical research or reporting purposes. The density desired, 1600 BPI or 6250 BPI, should be specified. Order forms are also available from the same address.