OBJECTIVES: Policy makers and Medicaid administrators often attempt to replicate cost containment strategies that have worked in apparently similar states. Such measures result in varied degrees of success owing possibly to not-so-obvious differences between the states' characteristics and determinants of prescription drug expenditure. This study attempts to identify homogenous groups (or clusters) of state Medicaid programs based on variables that describe such characteristics, drug expenditures and their determinants. METHODS: Variables were identified following Andersen's Behavioral Model for Health Services Utilization. Latest available data (2002) were obtained for the above variables from public data sources for 48 fee-for-service state Medicaid programs. Hierarchical Cluster Analysis technique was employed to identify optimum number of homogeneous groups of states by minimizing within-group variation and maximizing between-group variation. Optimum number of clusters was identified based on the values of the R-squared (RSQ), cubic clustering criterion (CCC), and pseudo F (PSF) statistics. Following identification of optimum number of clusters, cluster memberships were assigned and variables that defined cluster characteristics were identified. RESULTS: The first step identified seven clusters to appropriately classify the 48 state Medicaid programs; RSQ = 68.6%, CCC = 2.0, and PSF = 14.9. The second step assigned cluster memberships ranging from 2 to 15 member states in a cluster. Key variables which described the differences in cluster characteristics were: poverty levels (p=0.024), access to primary care (p=0.000), access to hospitals (p=0.025), Federal matching for state drug expenditure (p=0.044), budgetary support for healthcare (p=0.000), drug expenditures (p=0.043), and medical services expenditures (p=0.000) CONCLUSIONS: Identified clusters and their membership were not restricted to the commonly considered criteria of geographic proximity, socioeconomic characteristics, or program expenditure levels. Hence, knowledge of these groups, their membership, and the key variables that characterize them can further inform decision-making resulting in implementation of more cost-effective policy measures for Medicaid.