OBJECTIVES: Depression is the most prevalent major mental health disorder, affecting between eight and ten percent of the U.S. population. The U.S. Medicaid programs spent approximately $2 billion on antidepressant drugs in 2005, across three categories of antidepressants including selective serotonin reuptake inhibitors, tricyclic antidepressants, and other antidepressants. Predicting future prices and utilization of antidepressants would facilitate Medicaid's planning process. Our objective is to build forecasting models that can be used to improve Medicaid's budgeting efforts. METHODS: We gather quarterly data (1991-2004, Centers for Medicare & Medicaid Services) on Medicaid national antidepressant expenditure. We use Box-Jenkins forecasting techniques on expenditure and utilization time series for specific antidepressants including Paxil, Prozac, Wellbutrin and Zoloft. Intervention analysis is used to determine the effects of patent expiration, new branded-drug entry, and new indication approval. Forecasts are computed and compared to a holdout sample, comprised of the 2005 data, to assess the performance of the models. RESULTS: The Box-Jenkins ARIMA algorithms for fitting and diagnosing the expenditure and utilization data resulted in four distinct non-stationary forecasting models. Final models were selected using standard information-based criteria. The Paxil and Prozac models incorporated an intervention term corresponding to patent expiration: a step function for Paxil and an exponential decay for Prozac. The best fitting model for Wellbutrin is a third order moving average in the first differences; the Zoloft model is a Random Walk. Maximum likelihood was used for estimations. Usual checks on the residuals proved to be satisfactory. CONCLUSION: ARIMA modeling can be used to capture the time series of individual antidepressant drugs purchased by Medicaid. Moreover, intervention analysis can be used to demonstrate the effect that generic entry has on the utilization of or expenditure on a branded medication. We find that the drugs studied are affected differently by this type of event.