OBJECTIVES: There is accumulating evidence of sub-therapeutic second-generation antipsychotic (SGA) dosing for patients diagnosed with bipolar disorder, leading to suboptimal control of disease and higher overall treatment costs.  The objectives of this study were to identify Medicaid bipolar patients receiving clinically effective doses of SGAs and compare their medical costs.  METHODS: Patients with bipolar disorder taking an oral SGA (aripiprazole, olanzapine, quetiapine, risperidone or ziprasidone) were identified in Medicaid claims databases (2005 - 2008) from 8 US states.  Patients were followed for 18 months (6-month pre-index period during which patients did not receive an SGA, followed by a 12-month post-index utilization period to determine total costs). For patients on recommended dosing, costs were compared using a generalized linear model with a gamma distribution and log-link function. Baseline covariates (age, gender, race, pre-index costs, Charlson co-morbidity score, and specific psychiatric co-morbidities) were adjusted for. Ziprasidone-treated patients comprised the reference group. RESULTS: A total of 2446 patients met inclusion criteria, with 45% (N=1102) taking clinically effective doses by day 61 of their follow-up period. Patients on quetiapine had the lowest percentage of effective dosing at 26% (N=280/1072). Other results were aripiprazole 77% (N=336/448), olanzapine 52% (N=118/226), risperidone 50% (N=238/474), and ziprasidone 58% (N=130/226). Regression analyses indicated that mental health-related prescription costs (p<0.01) and all-prescription costs (p<0.01) were statistically significantly lower for the risperidone group compared to the ziprasidone group. There were no significant differences between the groups for total mental health-related costs or total all-cause costs (includes prescription and medical services). CONCLUSIONS: Less than half of the patients in this sample were prescribed clinically recommended doses 2 months after their initial start.  Among patients using recommended doses, while those on risperidone had lower prescription costs, there were no significant differences for total costs compared to patients taking ziprasidone.