Objective: Given the widespread use of antidepressants in primary care and specialty populations, we sought to examine whether provider specialty and patient demographic and clinical characteristics were associated with nonadherence to antidepressant therapy. Method: We conducted an observational cohort study of 11,878 patients enrolled in Harvard Pilgrim Health Care who were newly treated with antidepressants between May 2002 and May 2004. Using generalized estimating equations, we examined predictors of 2 types of antidepressant nonadherence: (1) immediate nonadherence: never refilling an antidepressant prescription; and (2) 6-month nonadherence: refilling an antidepressant prescription at least once, but not satisfactorily completing a 6-month treatment episode. Results: Compared with patients treated by primary care physicians (PCP), being treated by a psychiatrist was associated with significantly lower odds of immediate nonadherence (PCP 18% vs. psychiatrist 13%). Being treated by another type of specialist was associated with significantly higher odds of both immediate (other specialist 23%) and 6-month nonadherence (PCP 53%, psychiatrist 49%, other specialist 62%). Treatment by multiple providers was associated with lower odds of nonadherence than being treated by only I provider. Younger patient age and use of pain medication were associated with greater nonadherence. Conclusion: Rates of both immediate and 6-month nonadherence are high, and clinicians should emphasize the importance of continuing antidepressant treatment for a sufficient duration. Patients whose depression treatment is initiated by nonpsychiatric specialists may benefit from collaborative care models. These strategies may enable providers to better manage the long-term disability associated with their patients' depression.