The analysis presented below is based on the integration of two distinct datasets. These are the:
- 1999 Private LTC Insured Panel;
- 1994 National Long-Term Care Survey (1994 NLTCS).
The Insured Panel is comprised of disabled individuals who are currently receiving benefits under their LTC insurance policy and disabled policyholders that have not yet accessed benefits. The subset of the 1994 NLTCS that we use is comprised of individuals who are disabled in the community.1
The Insured Panel
Eight LTC insurers provided a qualified sample of 849 claimants who were receiving benefits under their LTC policies in the community. Of these, 693 participated in the study -- 82% of those contacted. These individuals had policies covering both nursing home care and home health care and were age 65 and over. A trained nurse or social worker interviewed each in their home.2
In addition, we interviewed by telephone 670 policyholders (randomly selected from non-claimants) to identify the proportion of individual policyholders who may be disabled (i.e. have two or more limitations in activities of daily living (ADLs) or are cognitively impaired) but not accessing insurance benefits. Only four non-claimants were identified as disabled.3 Based on this exercise, we concluded that claimants were not a sub-set of disabled policyholders. We did add the four disabled policyholders to the 693 claimants to get an Insured Panel based on 697 individuals.
Why are disabled non-claimants added to the sample of disabled claimants? This is because the key research questions relate to the impact of LTC insurance policy ownership on the use of formal and informal services. Suppose the insured sample was comprised only of claimants, who by definition are in the sample because they have a propensity to use services. The sampling basis for the non-insured sample -- the 1994 NLTCS -- is the presence of ADL or IADL limitations, that is, disability status. Individuals in that sample may or may not be using formal services. Thus, the non-insured sample includes disabled persons with and without a propensity to consume formal services. If we compared service utilization between the 1994 NLTCS and a sample comprised wholly of LTC insurance claimants, we would not be able to rule out the competing interpretation that any statistically significant differences in service use result from insured individuals' propensity to consume services rather than from the presence of insurance. Even if we were able to identify or match a sub-set of individuals from the non-insured sample who "looked like LTC insurance purchasers" the problem would not be solved: this is because by definition, a claimant sample would have been comprised of individuals with a propensity to consume services.
To make the samples more comparable, that is, "drawn" on a similar basis, we include in the claimant sample individual policyholders who are disabled (by the insurance policy definitions) but who do not make claims on their insurance. In this way, it is disability status and not service use that represents the underlying sampling basis for the Insured Panel, and this can be compared directly to the 1994 NLTCS Panel.
The 1994 NLTCS Panel
The 1994 National Long-Term Care Survey (NLTCS) was designed to collect data about the health and function of the Medicare beneficiary population age 65 and older. The questions were meant to identify persons who had certain disabilities or health problems lasting three months or longer. In general, individuals who completed a comprehensive community survey were those who had at least one chronic limitation in either an instrumental activity of daily living (IADL) or in an activity of daily living (ADL). The total 1994 NLTCS sample consisted of 4,167 individuals.
Only a sub-set of the Insured Panel and the NLTCS sample are used in the analyses that follow. This is because the criterion for inclusion in the 1994 NLTCS, a limitation in one or more IADLs or ADLs, is typically more liberal than the criteria for inclusion in the Insured Panel sample -- disability defined as having two or more ADL limitations or cognitive impairment. Thus, unless disability level was controlled for, we would not be able to make direct meaningful comparisons of the two populations.
To make the sampling basis of the two populations more similar, we applied the same minimum disability threshold to both samples. The criteria that are applied to individuals in each sample include being:
- disabled in at least two out of six ADLs that require some level of human assistance;4or
- cognitively impaired as measured by the Short Portable Mental Status Questionnaire (SPMSQ);5or
- diagnosed with dementia or Alzheimer's disease .
Once these minimum disability selection criteria are applied, the samples can then be compared one to the other on all dimensions related to formal and informal service use. All subsequent analyses are based on sub-samples that reflect this underlying minimum disability threshold.
The NLTCS sub-sample is comprised of 1,357 individuals (out of 4,167) and the claimant sub-sample is comprised of 581 individuals (out of 697).6