George Carcagno, Mathematica Policy Research
I am going to give you an overview of the National Long Term Care Channeling Demonstration so you will have a better appreciation of the factors that shape the data base.
The Channeling data base was constructed with a focus on supporting the evaluation, and so it is very different from the other data bases we will be talking about.
A recurring theme that you will be hearing is that there is a tradeoff; when you look at Channeling data, you find a very comprehensive detailed data base. In exchange for that detail and comprehensiveness, we do not have a statistically representative sample. That is a shortcoming that you will need to be aware of when you use it.
Channeling was designed to test comprehensive case management interventions that were designed to provide community services to frail elderly individuals, and where an appropriate substitute for nursing home placements.
Two models were tested, the basic case management model and the financial control model. They differed with respect to their capacity to purchase services.
The case managers did not have control over medical or nursing home expenditures, so that the Channeling case management was a particular variant of a case management approach. It is not the only way you could do case management, and that is something that you need to be sensitive to.
Case management is not a well-defined concept. Everybody has something else in mind when they talk about it, so Channeling tested a particular version of case management.
The two models shared a set of common core functions related to case management: outreach, screening activity to determine eligibility of people who were applying, in-person assessment, care planning, the arrangement of services, monitoring the receipt of services, and periodic reassessment.
The basic model relied primarily on existing services that were available in the community. They had limited gap-filling dollars to actually purchase services directly. Essentially they relied on what already was out there in the existing community care system in the way of services.
In effect, what we were testing with the basic model was the notion that the problem with the long term care system was that it was difficult for people to find their way through it; that with better information and more coordination of the existing service providers, we could provide services that would substitute for nursing home placements.
The financial control model was much more ambitious with respect to our services. There were substantial expansion of funds that were available to purchase community services. These services could be provided without any regard to eligibility for particular programs other than Medicare, which was not really a constraint. One did not have to be eligible for Medicaid, for example to receive the services.
The case managers could authorize the amount, duration and scope of community services. There were some cost controls built into the model. For the caseloads as a whole, the total community care expenditures for the entire Channeling caseload in each site was limited to, and could not exceed, 60 percent of the nursing home rate in each community. In fact, they came in well below that, around 40 percent.
There was also a limit on the costs for any individual care plan. That was set at 85 percent of the nursing home rate. It could not be exceeded without special approval.
We tested two models of the case management approach; one essentially just case management, and the other was case management plus greatly expanded access to community services.
We sought participants in this demonstration who were frail, elderly, and at risk of institutionalization. We wanted to get people who had multiple deficits, in activities of daily living (ADL) and instrumental activities of daily living (IADL) measures.
The demonstration was tested in ten sites, five for each model.
The demonstration began in early 1982, and we continued to collect data through the middle of 1984. We were interested in testing the effects of the demonstration on formal service use; that is, hospital and ambulatory care services, long term care services, and community care services. The impacts on individuals with respect to physical functioning, mortality, and well-being. We also looked at expenditures incurred by both, individuals and public programs.
Finally, we looked at the impacts on informal caregivers, family and friends that provided care to the elderly sample members. Here we looked at the amount of care that was provided, caregivers stress, well-being, and the amount of financial contributions they made to the elderly sample members.
A randomized experimental design was used, which meant there was random assignment to the treatment or control group in each model so that in the data base there are, the treatment group that actually received the Channeling services and control group that was simply interviewed, and other data sources were used for that group as well. In fact, the control group represents everybody else in the community who did not participate, and Channeling then was used as the basis for the comparison as to the effects of Channeling's impacts.
Several data sources were used. There were interviews with elderly sample members and their caregivers; a baseline interview and multiple follow-ups for the elderly sample members as a baseline, follow-ups at 6 and 12 months for everybody, and an 18 month follow-up for half the sample; and for the informal caregivers we interviewed a subset of caregivers for only a portion of the sample.
For the caregiver group, there was a baseline interview and two follow-ups at 6 and 12 months.
The data base is quite complex. There are numerous data sources and also it is longitudinal, so you need to pay some attention. It is not a simple cross-sectional survey.
The sample consists of people who applied to Channeling, and met the eligibility criteria. This was a highly disabled group; average age was about 80.
The Channeling project agencies, the sites that were responsible for conducting the demonstration itself cast a wide net to get referrals to the program, and so there were referrals from a number of health service providers. From hospitals, home health agencies, social service agencies, and a number of referrals by family and friends.
There was not a concentration of referrals from particular type or source. For example, referrals only from nursing home discharges or only from hospital discharges.
We examined several dimensions to see how our sample compared to a more representative sample of the frail elderly. One of the things we did was, to use the National Long Term Care Survey (NLTCS) to simulate the eligibility process and identity a sample within that nationally representative sample, identity the sample of people who would be eligible for Channeling, and then use that to compare the characteristics of that simulated sample to the characteristics of people who actually participated in Channeling.
We estimate on the basis of this simulation, almost 5 percent of the noninstitutionalized population over 65 would have been eligible for Channeling in 1982. If you look at the simulated national sample and Channeling, there is similarity as you might expect, given that we selected the sample on the ADL/IADL measures, there is a great deal of comparability in these measures.
With respect to the use of informal care, there is also a great deal of similarity across the two samples. Looking at things like percent who live alone, percent married, and income, we see that the Channeling sample tended to live alone. Fewer of them were married, and they had lower incomes.
The biggest differences were in formal service use. Keep in mind, we are talking about characteristics at baseline. That is, four people who actually participated in Channeling received Channeling services. This was the actual interview before the start of services.
If you look at formal service use, you see that the Channeling sample was about twice as likely to be receiving formal in-home care. More than twice as likely to have had a hospital stay in the last 2 months; almost six times more likely to have had a nursing home admission within the last couple of months; and about four-five times more likely to be on a nursing home waiting list.
Probably some triggering event led people to apply for Channeling in the first place, e.g., an acute episode. That event differentiated the people who actually came forward to participate in Channeling from people who had similar ADL/IADL disabilities and impairments, who were not participating in Channeling.
We also looked more generally at the socioeconomic characteristics of the aged and the sites that actually were in the demonstration. There we found that taken as a whole, the demonstration sites were probably similar to the national data. The only difference worth noting was that we had so many more Hispanic persons in the sample, and that was because of the Miami site's participation in the demonstration.
On most measures of income, age, and sex, the aged population in the Channeling site was quite comparable to the national aged population.
When you get down to looking at individual sites, the sample sizes are small, and you get much more variability at that level than you do when you use the data base as a whole.
Clearly, another thing we had to look at with respect to the issue of how representative this sample looks, even though it was not statistically designed to be a representative sample, it was designed to look at the service environments of the ten sites.
We looked at nursing home bed supply data. Beds per 1,000 people over 65 for the nation and for the Channeling counties, and we saw the bed supply in the Channeling sites was somewhat lower than the national nursing home bed supply; about 50 beds per thousand in the basic model compared to 57 nationally, and about 43 in the financial control sites. If you take Miami out of the financial control numbers, the financial control model is much more like the basic model.
On the basis of these bed supply measures, bed supply seems somewhat lower in sites than in the nation as a whole, although bed supply is the best you can do, but you have this problem of whether you are looking at demand factors, supply factors or what.
We also obtained information from people in the sites about waiting times for nursing home admission; and on the basis of that found that wait times were relatively short, except for Medicaid recipients. We had concluded that at least as far as the outcome of the evaluation itself was concerned, that nursing home bed supply was probably not a factor that affected the outcomes of the demonstration, whether or not it affects your research is something you need to keep in mind in your analysis.
We also tried to look at the availability of community services in our sites compared to the rest of the country. We found that there were not data available that one could readily find to compare the richness or the poorness of our service environments in the sites with communities around the country.
We do know looking at control group data at baselines, that there was substantial use among the controls of formal community services; 10-20 percent of the control group used case management services; and 60-69 percent used at least one formal community service at baseline. These variations depend on the model you look at.
There was clearly some substantial service use among control groups. We also know that sites applied to participate in Channeling; there was a competitive selection process. One could assume that the Channeling sites might well be more highly developed in the community service systems than other areas in the country, although I suspect that in large urban areas there is not a lot of difference.
That is a word of caution when you use these data to keep in mind the differences in environment and what effects they may have on your results.
To sum up, I think the Channeling data base can be very useful. You need to be aware of its strengths and weaknesses, however. It requires caution and good judgment when you use it. I think the leaps of faith are the kind we are all used to making when you do research in this area, when you combine the Channeling data base, for example, with other data bases. You are following in a long tradition of the way people have used demonstration experimental data from past studies, to do some estimates of new programs, new benefits, and so forth.
One strong advantage of the Channeling data base is that it incorporates a behavioral response to an intervention, and to the extent you are interested in similar interventions it can give you some idea of the response that you might expect.
In using the data for that purpose you are probably well advised to try and find other demonstration data to see whether you get similar results, put some boundaries around the results and maybe do some sensitivity tests. I think you would be well served by using the data base which is exceedingly rich and comprehensive.