We examine the sociodemographic and clinical characteristics of depressed primary care patients who receive at least minimal standards of evidence-based treatment, comparing those who remain depressed with those who recover; and their subsequent treatment patterns and other outcomes. We used observational data from a subset of 542 treated patients participating in a group-level randomized controlled trial of quality improvement interventions for depression conducted in six managed care organizations. Nonresponse to treatment was defined as having at least minimally appropriate treatment for at least two of three 6-month periods but continuing to have probable depression. Our definitions of depression and appropriate treatment are broader than those used in clinical trials, but relevant to primary care settings. Many of the factors predictive of treatment resistance in clinical trials predict nonresponse to guideline concordant care among diverse primary care, depressed patients. The main unique predictors of nonresponse to treatment include a clinical factor (suicide ideation) requiring clinician assessment and intervention, a social/economic factor (unemployment) usually not addressed by medical interventions, and medication nonadherence. Nonresponders used more adjunctive therapies and combination medications, suggesting clinicians and patients were searching for solutions. High rates of service use and poor outcomes emphasize the urgency of new research to find solutions for these patients.