OBJECTIVE: This study examined the association between a Medicaid prior-authorization policy for second-generation antipsychotic and anticonvulsant agents and medication discontinuation and health service use by patients with bipolar disorder. METHODS: A pre-post design with a historical comparison group was used to analyze Maine Medicaid and Medicare claims data. A total of 946 newly treated patients were identified during the eight-month policy (July 2003-February 2004), and a comparison group of 1,014 was identified from the prepolicy period (July 2002-February 2003). Patients were stratified by number of visits to community mental health centers (CMHCs) before medication initiation (proxy for illness severity): CMHC attenders, at least two visits; nonattenders, fewer than two. Changes in rates of medication discontinuation and outpatient, emergency room, and hospital visits were estimated. RESULTS: Compared with nonattenders, at baseline CMHC attenders had substantially higher rates of comorbid mental disorders and use of medications and health services. The policy was associated with increased medication discontinuation among attenders and nonattenders, reductions in mental health visits after discontinuation among attenders (-.64 per patient per month; p<.05), and increases in emergency room visits after discontinuation among nonattenders (.16 per patient per month; p<.05). During the eight-month policy period, the policy had no detectable impact on hospitalization risk. CONCLUSIONS: The prior-authorization policy was associated with increased medication discontinuation and subsequent changes in health service use. Although small, these unintended effects raise concerns about quality of care for a group of vulnerable patients. Long-term consequences of prior-authorization policies on patient outcomes warrant further investigation.