OBJECTIVES: State Medicaid programs vary drastically in terms of their prescription drug expenditure per recipient. Prior research has attempted to explain these variations by identifying potential determinants of drug expenditure. However, analyses in the past have been restricted to a few variables and several other potential determinants and their interactions have not been investigated. Objectives of this study were: a) To identify potential determinants of Medicaid per capita drug expenditures based on an established comprehensive model for health services utilization; and b) to test impacts of the identified determinants on per capita drug expenditures. METHODS: This study employs Andersen's Behavior Model of Health Services Utilization to identify potential determinants of pharmaceutical expenditures in state Medicaid programs using publicly available data. A structure equation model was built to test relationships among the latent constructs of policy, access, predisposing characteristics, enabling resources, and need for healthcare, and their influence on drug utilization. RESULTS: ‘Predisposing characteristics' were found to impact ‘enabling resources' which, in turn, was found to significantly impact drug utilization. Among the observed variables, ‘access to hospitals' and ‘access to primary care physicians' significantly described ‘healthcare resources'; ‘risk of diseases' described ‘need for healthcare'; and ‘poverty' described the latent construct of ‘enabling resources'. The ‘policy' construct was not described adequately by the indicator variables. CONCLUSION: Based on the study results, we conclude that Medicaid policy and program interventions, as described in this model, do not influence drug costs significantly. Population characteristics like predispositions and enabling resources determine drug costs in the state Medicaid programs.