OBJECTIVES: To identify which patient baseline characteristics and which types of early changes in patients' clinical status are most predictive of switching antipsychotics in the long-term treatment of schizophrenia. METHODS: This post-hoc analysis used data from a randomized, open-label, multi-site, one-year cost-effectiveness trial of olanzapine, risperidone, and typical antipsychotics in the treatment of schizophrenia. Study protocol permitted switching of antipsychotics when clinically warranted. Baseline characteristics were assessed using standard psychiatric measures and systematic review of medical records. In addition to baseline socio-demographics, co-morbid medical and psychiatric conditions, body weight, clinical, and functional characteristics, the prediction model included change scores on clinical measures (PANSS, five PANSS factor subscales, Barnes Akathisia Scale, Simpson Angus Scale) during the first two weeks of treatment. Cox proportional hazards model was used to identify the best predictors of switching from patients¢ initial randomized antipsychotic. RESULTS: About one-third (29.5%, 190/644) switched antipsychotics before the end of the one-year trial. Five variables were identified as best predictors of switching during the 1-year trial (p<0.05): absence of antipsychotic use in the prior year, pre-existing depression, lack of lifetime substance use disorder, less improvement or worsening following two weeks of treatment on either clinician-rated akathisia (Barnes Akathisia Scale), and/or anxiety/depression symptoms (PANSS). A strong trend was observed for female gender (p=0.058). CONCLUSIONS: Switching of antipsychotics appears to be prevalent in the naturalistic treatment of schizophrenia, and can be predicted by a small and distinct set of measures. Interestingly, pre-existing depressive symptomatology and less improvement or worsening of anxiety and depressive symptoms following two weeks of treatment were among the more robust predictors of future switching of antipsychotics in this one-year study.