Perhaps no facet of long-term care (LTC) research has received as much attention as the evaluation of community-based LTC demonstrations. Recent thinking has called into question the validity and policy relevance of the findings of these studies, yet their efforts will continue to receive strong financial backing. The time has come for LTC analysts to rethink the place of evaluation research as a tool for LTC policy development and to strengthen the methods for such development. Following the introductory analyses of LTC evaluation research, this paper describes a quasi-experimental research design, cohort analysis, used to evaluate the Hospital Discharge Demonstration Project (HDDP) funded by the State of Illinois, which provided short term, in-home assistance to impaired persons, sixty and older, upon discharge from Chicago hospitals. The benefits of the quasi-experimental cohort analysis, when an experimental design cannot be used, are explored; specific topics discussed include sample comparability (in background and characteristics), sampling techniques, and the effects of political and organizational realities on the study design. A final section compares the benefits of traditional, comparative cost and effectiveness evaluations of community-based and institutional LTC and the quasi-experimental cohort analysis design. (MCF)