OBJECTIVE A retrospective study was conducted to evaluate the time to discontinuation (TTD) of the first- (FGAs) and second-generation antipsychotics (SGAs). METHODS In total, 918 treatment episodes of patients with schizophrenia, initiated on one of the investigated drugs on an outpatient basis during 2004-2006, were entered into the study. The primary outcome was the duration of the investigated treatment episode. Discontinuation was defined when either patients were admitted or the investigated drug had been stopped for more than 28 days. We used the Cox proportional hazard model to compare hazards of discontinuations among 8 SGAs versus 2 FGAs (haloperidol and sulpiride). The follow-up period was up to 18 months. RESULTS During the follow-up period, clozapine had the highest rate of continuous treatment in the primary analysis: clozapine, 40.6%; olanzapine, 23.4%; aripiprazole, 22.9%; amisulpride, 21.9%; zotepine, 21.3%; sulpiride, 17.0%; risperidone, 12.8%; quetiapine, 12.5%; haloperidol, 10.6%; and ziprasidone, 10.4%. Compared with haloperidol, 5 SGAs had significantly longer TTD (adjusted hazard ratios and 95% confidence intervals): clozapine (0.403, 0.267-0.607), olanzapine (0.611, 0.439-0.849), aripiprazole (0.570, 0.407-0.795), amisulpride (0.680, 0.487-0.947), and zotepine (0.687, 0.497-0.948), but only clozapine had significantly longer TTD compared with sulpiride (0.519, 0.342-0.786). The sensitivity analysis showed similar results. CONCLUSIONS The current findings suggested that SGAs or FGAs are not homogeneous groups. Clozapine has the highest rate of continuous treatment among SGAs, and haloperidol is not the representative drug for all FGAs. Furthermore, antipsychotics dropout rate is high in naturalistic situation. A good service model needs to be constructed to enhance antipsychotic treatment adherence of people with schizophrenia.