We describe a secondary analysis of data from a randomized trial conducted at seven primary care clinics of a Seattle area HMO. Adults with major depression (n=290) beginning antidepressant treatment completed structured interviews at baseline, 1, 3, 6, 9, 12, 18, and 24 months. Interviews examined clinical outcomes (Hamilton Depression Rating Scale and depression module of the Structured Clinical Interview for DSM-IIIR), employment status, and work days missed due to illness. Medical comorbidity was assessed using computerized pharmacy data, and medical costs were assessed using the HMO's computerized accounting data. Using data from the 12-month assessment, patients were classified as remitted (41%), improved but not remitted (47%), and persistently depressed (12%). After adjustment for depression severity and medical comorbidity at baseline, patients with greater clinical improvement were more likely to maintain paid employment (P=.007) and reported fewer days missed from work due to illness (P<.001). Patients with better 12-month clinical outcomes had marginally lower health care costs during the second year of follow-up (P=.06). We conclude that recovery from depression is associated with significant reductions in work disability and possible reductions in health care costs. Although observational data cannot definitively prove any causal relationships, these longitudinal results strengthen previous findings regarding the economic burden of depression on employers and health insurers.