Gerry Hendershot, Ph.D., National Center for Health Statistics
For many years before his death, my father owned and operated a service station, and, as part of his work, he made auto repairs. He was very good at it, but he always denied being an expert. He said that "I just do the best I can on whatever comes down the road."
I am in somewhat the same position. I work on a general purpose health survey. I am not an expert on long term care like many of you. I am here to tell you about the 1984 Supplement on Aging (SCA) to the National Health Interview Survey (NHIS).
There are two other people from our staff in the audience. Susan Jack was very much involved in preparing the specifications for the SOA, and is very familiar with the structure of the file. She is also co-authoring a report on functional limitations.
The other person is Joe Fitti, who was very much involved in the planning and development of the questionnaire and has also worked closely on the Longitudinal Study on Aging (LSOA).
I do want to acknowledge the important contribution that Dr. Mary Grace Kovar has made and continues to make to the SOA. She was the prime mover in the initiation, planning and implementation of the survey, and she continues to serve as the coordinator of analytic activities underway at the National Center for Health Statistics (NCHS). More informally she acts as the center of a network of researchers who are using the SOA data tapes.
We are part of the Office of the Assistant Secretary for Health (OASH). In June of this year that is going to change: we are going to become part of the Centers for Disease Control. We are not going to move geographically; we will remain for the time being in Hyattsville, although a move somewhere else in the Washington area is probable in the next couple of years.
NCHS has a complex organization. We do a lot of different things, and the NHIS is only one of those things.
The NHIS is conducted by the Division of Health Interview Statistics in NCHS. We have four branches. I head the Illness and Disability Statistics Branch. Joe Fitti is in the Survey Planning and Development Branch and Susan Jack is in my branch. We have a programming staff, and we have a fourth unit which is called the Utilization and Expenditures Statistics Branch. That branch now performs the function of doing follow-up surveys on the NHIS, such as the LSOA.
The NHIS has these characteristics. It is a continuous survey; it has been in existence since 1957, and except for a couple of brief interruptions has been in the field every week over that whole period to date.
Each week's interview assignments constitute a nationally representative sample of the civilian noninstitutionalized population of the United States. It is a multi-staged area probability sample of data collected by personal interviews in sample households.
The field work is done for us under an interagency agreement by the field division of the Bureau of the Census. We get approximately a 95 percent response rate, year after year. We are quite proud of that, and proud of the Census field staff for achieving that high level of response.
Currently around 46,500 households are in the sample, about 122,000 people are in those households, and we get information about all the persons in these households.
There are two basic data collection instruments which we use. One is the basic health and demographics questionnaire, which remains the same year after year; it is revised about every 10 years. It was most recently revised in 1982, and we are currently planning another revision which will be implemented in 1989 as our current target.
In addition to that questionnaire we have special questionnaires each year on selected health topics.
In the basic health and demographics questionnaire, we get responses from adults in the household if they are there at the time of the interview; we get proxy responses for children and people who are not there at the time of the interview. We spend an average of 25 minutes on this part of the questionnaire in each household.
A list of the topics which are covered are: household composition; limitation of activities, which is a measure of chronic conditions; restricted activity days; acute conditions; visits to the doctor in the 2 weeks before the interview; and various health indicators, one of the most important and most useful of which turns out to be respondent-assessed health status, categorized as excellent, very good, good, fair, or poor.
We have six different lists of chronic conditions, and one-sixth of the sample is asked each of those lists.
A very large number of conditions are covered by those six lists, and each list is organized around a body system, so we get information on the prevalence of chronic conditions in that way.
We get additional information on any condition which is reported anywhere in the questionnaire, which is used largely for the purpose of medically coding the conditions that are reported.
Finally, we get some demographic information.
Our emphasis in the NHIS has changed in recent years from an emphasis on the basic health and demographic questionnaire to the questionnaires on special health topics. The way we see the survey now is that we do a survey each year on selected health topics, in addition to which we collect the basic health and demographic questionnaire. Emphasis is on these special health topics, and usually it is a separate questionnaire, maybe more than one questionnaire. The 1985 questionnaire was on health promotion and disease prevention.
These are typically annual or they may be less than a year. Last year, we did a vitamin and mineral supplement for the Food and Drug Administration (FDA), which was only 6 months. It could be more than 1 year. In 1979-1980, we did a special supplement on home care for both years. There is usually a random selection of one person in the household, although other sampling plans have been used, and it is usually self-reported data, although in some cases we have accepted proxy reporting in these special health topics. Again, it takes an average of about 25 minutes for these special health topic questionnaires.
The special health topic portion of the NHIS are collaborative arrangements between NCHS and other federal agencies. So far, it has been other public health services agencies, but it is possible for us to collaborate with others as well. Collaboration means that we join with them in planning the survey, and we ask that they contribute toward funding the survey. We now support about one-third of the cost of the survey through these collaborative arrangements with other agencies.
In 1979 and 1980 we did a special supplement on home care, and in 1984 we did the SOA. In the same year, we did a supplement on health insurance, which is one that we do periodically, so there may be some interest in getting both the SOA and the health insurance file.
The sample design for the NHIS is a nationally representative sample. There are primary sampling units in the U.S. representing the civilian, noninstitutionalized population. The 1970 design was used through 1984; therefore, it would apply to the SOA.
The sample was redesigned in 1985, so the sample that was in use in 1984 had 376 primary sampling units, of which 156 were self-representing, and 220 non-self-representing. Of 50,000 dwelling units, 40,000 had completed interviews, and there were about 110,000 individuals in those households.
I want to say a word about the analysis of these data. The data from the basic health and demographic questionnaire come in five separate files. They can be obtained from the National Technical Information Service (NTIS), which is our agent for distributing public use data tapes from the basic health and demographic questionnaire.
Each of these files has weights on it. This is a probability sample. To get accurate estimates for the population, the data have to be weighted up to national estimates, and the weights are there on the tape. They are a little bit difficult for some people to understand, but there is a basic weight, and the weight that you would typically be using is a so-called annual weight. The average size of the annual weight is about 2,000, which means that each person in the sample, on average, represents about 2,000 persons in the population.
Since it is a stratified, clustered sample, you do not get the same precision from a sample like this that you would get from a simple, random sample of the same size, and a measure of the effect of the complexity of the sample on the precision of the estimates is the so-called design effect. To give you some idea of what the design effects are for the NHIS, some design effects for selected measures for 1975 and 1976 are around two. On the SOA, they are considerably less than that for most of the measures you would be using.
If you are interested in making estimates for the population, you must use the weights. If you are estimating statistics for fairly large population groups and you can be fairly confident that the precision is good, you might not worry about the complexity of the sample design. If you do not understand how weights are used and variances are estimated, and what effect they have on the kinds of answers you get from the data, my recommendation is that you find yourself a good mathematical statistician, and put yourself in his or her hands.
We include in the reports that we publish from NCHS, in a technical appendix, a description of how we estimate the sampling errors for statistics from the NHIS, and give you some examples of how to calculate an estimated sampling error for statistics which you might generate from the data files.
We disseminate the data in NCHS publications such as Advance Data, and there are five Advance Data reports published from the SOA. We also published data from the NHIS in an NCHS publication called Vital and Health Statistics, Series 10. Our most important product each year is Current Estimates from the National Health Interview Survey, which provides basic statistics from the whole range of data collected in the basic health and demographic questionnaire.
It also includes the technical appendices that I referred to earlier, and reproductions of the questionnaires that were used in that data year.
One place to look for the SOA questionnaire is in the appendix to the 1984 Current Estimates.
We also do a number of unpublished tabulations, which are available for your use simply by calling or writing us. We call these the social, economic and demographic characteristics (SED lists). All of the major health measures are cross tabulated by a large number of social, economic, and demographic characteristics.
We make public use data tapes available for the basic health and demographic questionnaire, and you get a package of documentation, which includes: a tape format, an Interviewer's Manual, the Medical Coding Manual, a copy of that data year's Current Estimates, and a publication which describes the basic methodology of the NHIS. These are available from NTIS.
The supplements we distribute directly from NCHS, so if you want the SOA, you order it from us.
I will give you a little flavor for the kind of data that come out of the basic health and demographic questionnaire. By combining a couple of years, we can make some estimates for some very small population subgroups, such as Hispanic groups. There are some fascinating differences among the different Hispanic origin groups with different measures of health. They probably have a lot to do with age structure of these populations, more than any underlying differences in their health conditions.
The SOA was done in January 1984 through the first couple of weeks of January 1985. It represents the population 55 years of age and older. We took all persons in the NHIS sample households who were 65 years and older, and a systematic 50 percent sample of persons in those households who were 55-64 years of age; so we have a half sample for the younger persons, 5564, and a full sample for those 65 and older.
There was actually about 16,148 cases, I believe. It was a personal interview wherever we could get personal interviews, and that was in over 90 percent of the sample cases. Proxies and telephone interviews were allowed when the data could not be obtained in any other way. The information that is available on the data tape, which was released in December 1986, includes not all the core information. There are, however, five records, one of which we call the "person record." The person record has all the demographics, and summary measures of some of the health characteristics, from the basic health and demographic questionnaire. They are included on the SOA file, along with all the information from the SOA questionnaire itself.
I just want to discuss one item by broad age groups from each of those major topical areas covered in the SOA. In the topical area on family, for instance, we have some measures of whether or not they have children, and if they are in contact with children.
In the area on community, we have some information about contact with friends and neighbors, and other kinds of community services.
In the section on retirement, we ask a question about whether they retired for health reasons, and some other information about the circumstances of retirement.
In the section on impairments, we get information, for instance, on vision and hearing problems. In the section on functional limitations, we look at the instrumental activities of daily living (IADL's) and activities of daily living (ADL's).
We also ask some questions about nursing home stays.
In the section on health opinions, we asked a question about how much they worried about their health.
Five Advance Data reports have been published. They were all based on preliminary data for the first 6 months of the data collection period. Those analyses were repeated using the full year's data, and were presented at a meeting of the Gerontological Society of America last fall.
There was a report by Dr. Kovar on the basic and health demographic measures. It shows, for instance, living arrangements cross-classified by marital status in the population.
There was another report, an Advance Data, by Dr. Kovar on the characteristics and health of elderly persons living alone. This shows, for instance, where children are relative to those who are living alone and to those who are not living alone. A lot of those who live alone had children living near enough to provide some kind of support and care for them.
Dr. Robyn Stone did a paper on use of community services. One of the findings was that the elderly do not make a lot of use of community services. We do see that those living alone are much more likely to use those services than those who are living with others.
There is a report by Dr. Tamara Harris on incontinence, and this, for instance, shows that people who have more severe problems of incontinence are less likely to socialize.
There is a report by Dr. Richard Havlik on vision and hearing problems. This report shows the relationship between visual impairment and selected ADL problems.
In addition to the five Advance Data reports which have been published, we have already mentioned that there is a report on the methodology, which we expect to be published this summer, by Joe Fitti and Mary Grace Kovar. There is also a report in preparation on functional limitations which will be published by NCHS as a Series 10 report, co-authored by Susan Jack and John Fulton and Sidney Katz of Brown University.
I mentioned a longitudinal study. This is a follow-up to the 1984 SOA. This takes advantage of a new capability of the NHIS, which we now call targeted follow-up surveys. Basically, we now have the capability after the NHIS is done, to select persons from that sample who have already been interviewed, according to any kind of characteristics that we may be interested in, such as ADL's, for instance, or Hispanic origin, and go back to those same people to get additional information, either of the same kind or to expand the range of information obtained.
This can be done by personal interview, mail questionnaire, or telephone. Typically, it is done by telephone. It is also possible now to match NHIS interviews, or follow-up interviews with the National Death Index (NDI), which is a mechanism for determining if a death record has been filed for the person, and that is being done as part of the LSOA. For the LSOA, we are also matching to Medicare records.
The interviewing was done for the first follow-up in the LSOA in 1986. A public use data tape is expected to be available in July of this year, and I am told by Richard Suzman of the National Institute on Aging (NIA) that NIA is encouraging applications for research grants to do analyses of the SOA and the LSOA.
The LSOA was intended to identify a representative sample of those people who were living in the community in 1984, and to trace them into nursing homes, see if they died, see if they stayed in the community, and what happened to them over time. Those data for the first follow-up will be available on the LSOA tape, which will be released in July of this year.
The functional limitations questions, the ADL's and the IADL's, were also asked in 1986 in the NHIS. That public use data tape is expected to be available about December of this year.