National Invitational Conference on Long-Term Care Data Bases: Conference Proceedings. Overview of the 1982/1984 National Long Term Care Surveys


Kenneth Manton, Ph.D., Duke University

The 1982/1984 National Long Term Care Survey (NLTCS) is an extremely rich data base; not only was there a longitudinal follow-up of the 1982 sample in 1984, but we have complete cross- sections in both 1982 and 1984. There is also a linkage to administrative records. There is currently a proposal to replicate this survey with a longitudinal component in 1988.

We are looking approximately at the 10 year period of 1980-1989. The first survey was July 1982, then the replication 2 years later in July 1984, and now the proposed replication after 4 years.

You also have to think in terms of a continuous collection of Medicare Part A service records that are linked to all individuals in the various survey populations. We are currently discussing the possibility of us linking into Medicare Part B, but that is a tentative plan.

In 1982, there was the NLTCS of the community disabled. A total of 36,000 individuals were drawn in from the Health Insurance Master File (HIMF), there was a telephone screen and also a personal visit screen for those who could not be contacted by telephone.

This picked out individuals who either had or were expected to have either an activity of daily living (ADL) or an instrumental activity of daily living (IADL) limitation of 90 days or more duration. From that screen of 36,000 individuals, 6,393 individuals were identified as having a chronic disability. They were delivered a detailed community survey, which had a large number of questions in terms of their disability, the type of personal services and equipment used to meet these limitations, as well as questions on other types of service use.

There was no institutional survey in 1982. There was a Informal Caregivers Survey (ICS) of 1,925 individuals who were caregivers to people in the NLTCS. Interviewers talked to about 1,625 individuals for those who were in continuing care, and about 300 individuals who discontinued care, and asked them about the nature and burden of care, the number of hours that were spent, and--for those 300 who quit--the reasons for ceasing.

The 1982 round was the basis for a series of projections on the long term care population. You could also trace those through to the number of individuals in the national population who were responsible for delivering care to those individuals. You could project those numbers out as well as from the ICS to get some estimate of the number of hours per week that the informal caregivers were delivering care and the types of care.

In 1984, there was both a longitudinal component and an aging component. By an aging component, I mean individuals who in 1982 were age 63-64 and did not qualify were also sampled. In 1984 you have a complete cross-section as well as a longitudinal follow-up of about 20,000 individuals who were there in 1982.

For all individuals in both samples, you had the continuous Part A service use so that you could find out about their acute care use of hospitals, their use of Medicare-funded skilled nursing facilities (SNF's), and also the use of home health. The bill gives the dates of services, the number of episodes, and the reimbursement amounts.

In the 1982 survey, we have 25,541 individuals who are identified as non-disabled by the telephone or personal visit screen. Then you have a people who are followed through to 1984, decomposed by where they ended up in one of four statuses.

The disabled 65+ noninstitutionalized population were the target of the community survey. The institutional sample, 1,992 individuals, were both people who were identified as being in institutions on the April 1 freeze date for the sample population, plus people who became institutionalized between April 1 and the beginning of the survey operations.

The 1,992 individuals were not interviewed in 1982 so we do not have the detailed information on their institutional use, but we can identify that component of the population to identify transitions between 1982 and 1984.

There were 4,916 individuals aging in for 1984, who were 63-64 in 1982. We have only a 47.4 percent sample of people who were non-disabled in 1982 that were followed up and screened again in 1984. This produces the 14,130 individuals in that non-disabled block for the 1984 component.

The 6,182 individuals who received the detailed community instrument in 1984 were automatically brought into the survey population. We can took at the status of individuals reporting chronic disability in 1982, but improved in their functional and health status. There did seem to be a significant group of individuals who reported chronic disability but yet showed improvement in functional and health status on the long term basis over a 2 year period.

In 1984, there was a short institutional questionnaire, which got somewhat of a retrospective history of nursing home service use, and there are about 1,870 individuals. There was a deceased questionnaire, a next-of-kin survey describing something of the characteristics of health service use for those individuals who died over the 2 year period. Of the 3,219 individuals who died over the 2 year period, 970 individuals died out of the non-disabled group and 1,383 died out of the disabled and 65+ noninstitutionalized group.

Potentially, with the 1988 round, you have a time-frame of recovering a large number of changes and reimbursement of policies and analyses looking at some service substitution of facts over at least the 1982-1984 changes.

One of the missing components is Medicaid institutional service use. The instrumentation in the survey itself is similar to the instrumentation in a number of the major demonstration projects. Here you have a nationally representative data file, and potentially you have the health and functional measures, and certain outcome measures to relate it to the outcomes of various demonstration data sets where you might either have interventions or more intensive Medicaid service use records. By crosswalking between the nationally representative data and the slack population in the demonstration data sets, you can tell something about how well the demonstration results might translate to a national level.

Surveys looking at particular target populations where you can specialize the instrumentation and focus your samples are cost-effective. Think of three major surveys, the NLTCS linked to its Medicare service use, the Longitudinal Study on Aging (LSOA) with the Supplement on Aging (SOA) in 1984 and again in 1986, and the National Nursing Home Survey (NNHS).

There is potential coordination of these surveys that can reflect on different aspects of the elderly population. In the sample design one of those gaps in the NLTCS is the non-chronically disabled, community dwelling elderly; the SOA will give you information on that group.

The institutional samples about 1,800 individuals in 1984 and in 1988 we project to be 1,600. That gives you transitions or changes from 1982-1984, but it is nowhere near as detailed a survey as, for example, the NNHS.

One can think of a coordinated analytic strategy with the NLTCS being one component fitting in with these other two surveys, and then also think intentionally about the various demonstration studies that might be coordinated to analysis of the surveys.

For the highly disabled, very elderly population, there were over 2,000 individuals age 85+ in each of the 1982 and 1984 survey dates. The diagnosis of senility was made on the basis of proxy response. There is a high proportion of full individual surveys but, as you would expect, with disability levels, the number of proxy respondents increases.

There is the information under Medicare service use linked in. In 1984, you have a next-of-kin report on health service use among the deceased, an institutional instrument for the institutionalized person, and as of the 1984 survey date the detailed community service use.

You can follow people who were disabled by various levels here, e.g., those individuals who only had IADL impairments, and those who had different levels of ADL impairment.

You can now identify an institutionalized component from 1982, and look at their status in the 1984 survey. In 1984, you can look at the deceased and at individuals who report chronic disability of 90 days duration but who even at very high levels of disability were community dwelling. There is a reasonably sizable proportion of these individuals who are improving somewhat in functional status, about 25 percent. It is interesting, as well, that the mortality of this particular group is nearly as high as the institutionalized group, 37 percent.

Of survivors to 1984, about 35 percent got better to some degree on the long term basis, suggesting that, at least, at that level of functional disability, you might have two sub- populations--one with an acute problem, perhaps hip fractures or something on this order, but also one which has a 2 year long term improvement even at very profound levels of disability.

We have constructed simple life table measures out of the 12 month service use window about the 1982 and 1984 survey dates. On this life table are survival curves, where down the left-hand side you have the proportion of all people. On the hospital service use table, if you look at 100 percent of people entering a hospital during a 12 month window associated with the 1982 service date, and then look at a 12 month window of eligibility in the 1984 service date, for those episodes you can see how rapidly people were discharged.

Associated with each of these curves are statistics like mean length of stay or median length of stay for the sub-populations. Obviously, from the survey itself you could start looking for the community disabled population, the discharge and hospital service uses, a function of disability level, and a number of other co-variates. This allows you to perform some very detailed analyses of the interrelationship of chronic functional Impairment and acute service use, and allows you to start looking at the systemic effects or the interaction.

View full report


"87cfproc.pdf" (pdf, 2.52Mb)

Note: Documents in PDF format require the Adobe Acrobat Reader®. If you experience problems with PDF documents, please download the latest version of the Reader®