How expensive is antipsychotic polypharmacy? Experience from five US state Medicaid programs

Objective: To characterize healthcare costs associated with antipsychotic polypharmacy and to investigate predictors of high-cost patients. Methods: A retrospective cohort study using Medicaid claims data from California, Nebraska, Oregon, Utah, and Wyoming evaluated 55 383 fee-for-service patients with antipsychotic prescriptions between 1998 and 2002. Polypharmacy was defined as initiating multiple antipsychotic drugs or concomitant antipsychotic therapy ( >= 60 days). Healthcare costs (drug and nondrug) were summed for 365 days following index antipsychotic claim. Adjusted mean costs were compared to antipsychotic monotherapy. Logistic regression was performed to identify predictors of high-cost patients (top quintile) with regard to patient age, gender, race/ethnicity, mental disorders, hospitalization, index antipsychotic, concomitant psychotropic drugs, and polypharmacy. Results: The average annual prevalence of antipsychotic polypharmacy was 6%. 70-80% of total healthcare expenditures for polypharmacy patients were drug-related. Polypharmacy was associated with significantly higher drug expenditures ($1716-2079) in the year following drug initiation than monotherapy even after adjusting for case mix and index antipsychotic (p < 0.05). Differences in non-drug expenditures versus monotherapy were smaller and varied by state ranging from a $77 increase in California (p < 0.001) to a $211 savings in Utah (p = 0.02). In California, polypharmacy alone (OR = 2.69; 95% CI: 2.30-3.16) or in combination with concomitant psychotropics (OR = 6.26; 95% CI: 5.51-7.11) was associated with greater likelihood of being a high-cost patient than monotherapy. Conclusions: Cost savings from limiting antipsychotic polypharmacy could be significant. Caution must be taken in ensuring reductions in polypharmacy do not lead to unintended consequences or shift care to more costly alternatives. Study limitations, including the known shortcomings of claims data and differences across state Medicaid programs, should be considered when interpreting the results of this or any multi-state study.