OBJECTIVE: To compare the mental health resource use associated with risperidone, olanzapine, and quetiapine for treatment of bipolar disorder in a real-world setting. METHODS: This was a retrospective, comparative study based on claims data compiled from several US health plans from 1999 to early 2002. Antipsychotic treatment episodes were constructed to more accurately identify mental health resources associated with risperidone, olanzapine, and quetiapine. Selection bias was reduced by focusing only on episodes involving antipsychotic monotherapy and for which the patient did not switch from a prior antipsychotic. The primary measure analyzed was non-antipsychotic mental health care charges per patient per month (PPM), defined as total mental health care charges excluding antipsychotic drug charges during treatment episodes with risperidone, olanzapine, or quetiapine. To control for differences in patient characteristics, regression models combining risperidone, olanzapine, and quetiapine treatment episodes were estimated to determine their effects on non-antipsychotic mental health resource use. RESULTS: Regression estimates showed that quetiapine was associated with the lowest non-antipsychotic mental health care charges PPM. These charges averaged about $14 PPM (P=0.069) and $9 PPM (P=0.231) lower than those of risperidone and olanzapine, respectively, a savings of 2%-3% based on a $526 PPM mean charge. Differences in non-antipsychotic mental health care charges PPM among bipolar patients treated with risperidone, olanzapine, or quetiapine were largely explained by differing patient and treatment characteristics rather than the antipsychotic used. After dosage levels were standardized, however, olanzapine was associated with significantly higher drug acquisition costs than those of risperidone and quetiapine (57% and 49% respectively; P<0.01). Although the difference in drug charges between risperidone and quetiapine was not statistically significant, quetiapine was associated with lower PPM resource utilization. CONCLUSION: Quetiapine appears to be associated with modestly lower non-antipsychotic mental health resource use compared to risperidone and olanzapine for treatment of bipolar disorder.