Although there have been previously published algorithms for the treatment of depression, few have been rigorously studied in terms of outcome. We present the Duke Somatic Treatment Algorithm for Geriatric Depression (STAGED) approach, developed by clinicians in the Duke Mood Disorders Program. The algorithm consists of five stages based on treatment history. In developing the approach, we enrolled 228 depressed elderly patients in a National Institute of Mental Health-funded longitudinal study of depression. Patients were treated by geriatric psychiatrists using the algorithm as a guideline for treatment. Initially, 174 patients were treated with antidepressant medications, and 54 patients were treated with electroconvulsive therapy. Over the course of 18 months of treatment, 88.6% of patients responded, and 65.35% of patients were able to achieve full remission of symptoms. The STAGED approach appears to be a clinically useful method of successfully treating elderly depressed patients.