Evaluating long-term care demonstrations in real time with study design and plan performance interactions.

The evaluation of long-term care demonstrations has to deal with complex organizational entities, with large, heterogeneous client populations that, during the course of study, may have to change features of their organization or operation. The implications of such "real-time" changes are discussed for analyses of the operation and performance of Social/Health Maintenance Organizations over a 5-year period (2 years of start-up and enrollment and 3 years of follow-up). Analyses conducted of the plans in the context of realtime changes have to be based on different statistical models than for classic experimental study designs, where treatment factors are fixed rather than dynamic. A number of issues that may arise are identified, and possible approaches to their solutions described.