OBJECTIVES: To profile trends in the levels of treatment persistence over time across several typical and atypical antipsychotic agents among patients with schizophrenia in the Veterans Health Administration (VA). METHODS: Using VA fiscal years 2000-2004, we defined initiation of the target agent as 6-month "clean" period of no target drugs prior to initiation and reserved one year following the initiation to calculate treatment persistence, or time to discontinuation, as defined by a gap of > 15 days (a sensitivity analysis was also conducted using a gap of > 30 days). Using a floating data approach, we created four time periods for which analyses were conducted for each period. We calculated hazard ratio using Cox proportional method and used a multi-stage regression model, in which one set of covariates was introduced into the model sequentially. RESULTS: Compared to patients who initiated typical antipsychotics, those who initiated atypical antipsychotics tended to have better treatment persistence as reflected in longer stay on the medication within one year between initiation and the first gap of > 15 or > 30 days (99 vs. 141 days on average; p <0.001). However, between 10/1/1999 and 3/31/2005, treatment persistence with typical antipsychotics remained the same, whereas treatment persistence with atypical antipsychotics decreased from 149 to 135 days. Among individual typical antipsychotics, treatment persistence with chlorpromazine decreased from 110 to102 days, treatment persistence with haloperidol remained the same, whereas treatment persistence with perphenazine increased from 116 to 128 days. CONCLUSION: Despite being efficacious in reducing symptoms of schizophrenia, the likelihood of sustaining control of schizophrenia may depend on treatment persistence. However, poor treatment persistence with antipsychotics is a common problem among patients with schizophrenia. Future research needs to explore factors associated with different levels of treatment persistence across different antipsychotics.