In the present study, an adaptive social systems model is developed and utilized to examine the dynamic properties of a health services delivery system. The model employs a Markov chain as a structure for analyzing the utilization of physician services both inside and outside of a prepaid plan. Four system states are defined, namely, no use, inside use, outside use, and both inside and outside use of physician services. Subscribers are considered to be in one of these states on each episode of illness. Transition probabilities are estimated from data collected from a sample of subscribers to a university health plan using a regression approach that permits a heterogeneous population to be examined within a Markovian framework.This framework is used to examine factors that account for differences in utilization patterns among individual subscribers and to examine overtime changes in the transition probabilities resulting from population shifts on specific demographic and health-related attributes. Based on the assumption that individual subscribers rather than the population as a whole follow a Markov process, the k-step transition matrix is constructed for the heterogeneous subscriber population and is used to examine the long-run effects of particular utilization patterns on the functioning of the prepaid plan.