Although new strategies for managing depression in primary care (e.g., nurse telephone calls, collaborative care) have been shown to be effective, no models are available for their systematic implementation in the "real world." To test whether a continuous quality improvement (CQI) intervention could be used to implement systems in primary care clinics to improve the care and outcomes for patients diagnosed with depression. Before-after study with concurrent controls. A multidisciplinary team from the three intervention clinics developed and implemented a graded set of five care management options, ranging from watchful waiting (nurse telephone call in 4 to 6 weeks) to mental health management, which clinicians could order for their patients with depression. 9 primary care clinics in greater Minneapolis-St. Paul, Minnesota. Outpatients 18 years of age and older whose primary care clinic visit included an International Classification of Diseases, 9th revision, code for depression and who completed baseline and 3-month follow-up surveys before and after the intervention. Measures of process of care (follow-up depression visits to physician, mental health visits, follow-up telephone calls) and outcomes of care (improved depression symptoms over 3 months, satisfaction with care). Although the CQI team appeared to function well, only 30 of the 257 patients identified from depression-coded visits for this study were referred to the new system during the 3-month evaluation period. In both the intervention and control clinics, follow-up visits, mental health referrals, and follow-up telephone calls did not improve significantly from the preintervention levels of about 0.5 for a primary care visit, 0.4 for a mental health visit, or 0.1 for a follow-up phone call per person. The same was true of patient outcomes: The proportion of patients in the intervention and control clinics who had improved depression symptoms and those who were very satisfied with their depression care did not change significantly from the preintervention levels of 43% and 26%, respectively. Our attempt to improve the primary care management of depression failed because physicians used the new order system so infrequently. Whether a greater leadership commitment to change or a different improvement process would alter our findings is an open question.