New compounds have been introduced to treat schizophrenic patients in the last 4 years. Their safety and efficacy with respect to conventional treatments are well documented, but their economic impact in a naturalistic context is still pending. OBJECTIVE: The purpose of the study is to evaluate new antipsychotics in usual practice with a pragmatic Markov Model of patients compliance under treatment. METHODS: The model is based on a 6-month Markov cycle tree divided into four sub-trees: M1, M2, M3, and M4. (1) M1 identifies the drug strategies in schizophrenia: sertindole versus olanzapine versus haloperidol. (2) M2 enumerates the care structures. Five care management categories are defined in function of the place where patients live (hospital, residential, or private home) and the intensity of care (intensive or mild residential care, intensive or mild own-home personal care). Care management depends on the clinical status (relapse and non-relapse). Long stay hospitalization is integrated as a separate branch. (3) M3 identifies the clinical events. Each of the treatments has side effects determining the compliance or noncompliance and the frequency of relapses. Adverse events are extrapyramidal symptoms, somnolence, weight gain, and sexual disturbance. Toxicity rates are estimated from three integrated summaries of safety. Compliance and relapse rates are obtained by a metaanalysis of the literature. They are differentiated according to the patients' settings. (4) M4 shows the patients' paths in the health care system. The corresponding transitional probabilities are derived from two French cohorts (2,747 pts), a German cohort (459 pts), and a British cross-sectional study (1,051 pts). The model is based on 18 health states and 19,000 nodes. RESULTS: The relative risk of relapse of haloperidol and olanzapine with respect to sertindole is respectively equal to 1.4 and 1.2. Not only is sertindol self-financing because of saved hospital admissions (−$12,000 compared with haloperidol and −$6,500 compared with olanzapine), but it produces a slight net savings compared with the two components. Olanzapine and haloperidol are dominated strategy with a lower effectiveness and a greater cost. A sensitivity analysis carried out on toxicity, compliance, relapse, and drop-out rates confirms the robustness of the results. CONCLUSION: In schizophrenia, sertindole brings a benefit of 5 months without relapse compared with olanzapine and 13.5 months with respect to haloperidol. In terms of cost-effectiveness, the study clearly shows the interest of sertindole.