There exists a substantial literature concerned generally with modeling nursing home use. Because it has recently been extensively reviewed elsewhere (Garber and MaCurdy, 1989; Greene and Ondrich, 1990; Greene, Lovely, and Ondrich, 1992; Wolinsky et al., 1992) we will not consider it further here. Most studies that explicitly consider the effects of CLTC services on nursing home use have been experimental or quasi-experimental evaluations of demonstrations which have measured the aggregate (average) effect of providing a package of supplemental community services to a treatment group. Because extensive and capable reviews of these demonstration findings have also appeared recently in the literature (Hughes, 1985; Kemper, Applebaum and Harrigan, 1987; Weissert, Cready and Pawelak, 1988), we will not consider them at length here except to note that in the more rigorously conducted studies, save for a demonstration integrated with nursing home admission screening, the demonstration services have not been found to offset sufficient nursing home use to cover their costs of operation. This was the finding also in the NLTCCD (Thornton, Dunstan, and Kemper, 1988).
The demonstration studies, while important, offer little guidance in certain key policy areas--particularly in matters of individual service targeting and allocational efficiency. This is because they consider aggregate, rather than individual level, service effects, and because they do not attempt to disaggregate the effects of the component services in the service "package." With respect to these evaluations, Kane (1988) has pointed out that "If all of these [CLTC] services continue to be bundled together as "alternatives" to nursing homes, we continue to postpone much needed research to look at the effective design and targeting of each kind of community service."
Only two empirical studies, to our knowledge, have appeared which address the issue of efficient or optimal allocation of CLTC services in reducing nursing home use. The most recent of these (Davidson, Moscovice, and McCaffrey, 1989) is not an optimization analysis as such, but an empirical test for a necessary efficiency condition that would hold under some conditions: assuming these conditions, allocations were found to be efficient. An earlier, and notably innovative, study was that by Miller (1987) which undertook a direct optimization of community service assignments in minimizing institutionalization. Each of these studies considers only a single composite "service," however, and so does not operationally address issues of individual service effectiveness or targeting.