OBJECTIVE: To estimate the costs and effects of long-acting risperidone versus a conventional depot and an oral atypical over a five-year period in Germany. METHODS: A discrete-event model was developed comparing three scenarios. In scenario 1, patients start on haloperidol depot and may be switched to olanzapine followed by clozapine. In scenario 2 patients start on long-acting risperidone instead of haloperidol depot. In scenario 3, patients start on olanzapine, and may be switched to risperidone (oral) followed by clozapine. The model simulates individual patient histories accounting for age, gender, type, severity of disease and side effects. Based on these patient characteristics, the model simulates visits, psychotic episodes, symptom-scores, treatment, compliance and location. Outcomes are expressed in terms of number and duration of psychotic episodes, symptom-score and costs. Costs of psychiatrist visits, medication and location were included. Information was derived from literature and an expert panel. RESULTS: Over a 5-year time horizon and per patient, starting with long-acting risperidone was estimated to avoid 0.21 and 0.32 relapses and to save €131 and €2192 compared to a conventional depot (scenario 1) and to an oral atypical (scenario 3) respectively. In subgroup analysis of only high-risk non-compliant patients starting with long-acting risperidone was estimated to avoid 0.22 and 0.46 relapses and to save €1442 and €9082. Sensitivity analyses showed that the results are robust and that they are mainly related to estimates about compliance (when compared to oral atypical agents) and to the effects of atypical and conventional formulations on the symptom-score. CONCLUSION: Long-acting risperidone is cost-effective, as it has higher effectiveness and is cost-neutral compared to a conventional depot and cost-saving compared to an oral atypical. Further benefits may be expected, when treatment is aimed at patients with a high-risk of being non-compliant and at patients in whom further deterioration is expected.