OBJECTIVES: To estimate costs and outcomes associated with atypical antipsychotic treatment and switching patterns in patients with schizophrenia. METHODS: We developed a Markov model that includes five health states defined by scores on the Positive And Negative Syndrome Scale (PANSS): acute episode, persistent negative symptoms, response state, fourth-line therapy (clozapine), and death. Following two unsuccessful medication switches, patients transitioned to the clozapine health state. Utility weights for each health state were determined from a published utility assessment based on PANSS. The model accounts for adherence and discontinuation, relapse of symptoms, and adverse events, and uses the Framingham risk equation to account for metabolic and cardiovascular effects. Unit healthcare costs were determined from standard US data sources. Key model outputs included inpatient and outpatient resource use and costs, costs related to cardiovascular and metabolic complications, and relapse rates and associated costs. We compared time and cost in each health state for two antipsychotic regimens: 1) initiating therapy with a first-generation atypical antipsychotic and switching as needed to a second-generation agent, and 2) initiating therapy with a second-generation agent and switching as needed to a first-generation agent. RESULTS: Initiating therapy with a second-generation atypical antipsychotic (vs the opposite strategy) reduced time in the acute episode state (12.2 vs 12.8 weeks) and increased time in the response state (22.1 vs 19.8 weeks) at one year. This scenario was associated with lower annual total nondrug medical costs in the acute health state ($32,185 vs $33,193) and lower total nondrug costs in all states ($43,061 vs $44,280). CONCLUSION: Initiating therapy with a second-generation atypical antipsychotic was associated with more time without symptoms than was seen when therapy was initiated with a first-generation agent, and lower nondrug costs during the first year of treatment.