OBJECTIVES: The primary objective of this study was to compare variance in cost of pharmaceuticals explained by demographic variables with variance explained by plan-characteristics within various therapeutic categories using prescription claims data. The secondary objective was to examine differences in utilization among demographic variables after controlling for covariates. METHODOLOGY: Data were obtained from 1996 prescription claims information for the commercial population administered by a Rhode Island-based PBM. There were 29,211 subscribers representing 64,815 enrollees (mean age 31.5) eligible during 1996. Subscribers were mainly employed in health (39.8%), sales and services (17.7%) and manufacturing (10.3%). Six therapeutic categories were analyzed. Statistical analyses utilizing multiple regression and analysis-of-covariance were performed. RESULTS: Plan-characteristics out-performed demographic variables sixteen-fold for all drug categories combined in explaining variance in cost of pharmaceuticals. Significant differences (p<0.0001) in utilization were found among demographic variables after controlling for average wholesale price and days supply. CONCLUSIONS: The results obtained in this study have practical significance in the determination of capitation rates when utilization history of prospective members is unavailable. In this situation, PBMs may have to set capitation rates based solely on eligibility data. PBMs contract with commercial clients to provide pharmacy benefits to their employees irrespective of their occupation. Significant differences in utilization among the members based on place of employment suggest that benefit managers should consider differentiating capitation rates according to their clients’ businesses. Finally, this study indicated that commercial members residing in Tennessee had the lowest level of drug utilization among all states evaluated. The fact that one PBM manages over 80% of the TennCare prescription program along with a significant commercial client base suggests that “spillover-effect” may exist.