To describe the prevalence and 9-month incidence of depressive symptoms among a cohort of elderly primary care patients and to determine whether different patterns of depression are associated with different patterns of health services use. Prospective study of depressive symptoms as measured by the Center for Epidemiologic Studies Depression (CES-D) scale and identification of patients' outpatient health services use through an electronic medical record system. An academic primary care group practice at an urban ambulatory care clinic. 1711 patients aged 60 and older who completed the CES-D at baseline and 9 months later; 935 of these patients also completed the CES-D at 6 months. The prevalence of significant symptoms of depression (CES-D > or = 16) was 17.1% at baseline and 18.8% at 9 months; 26.8% of patients exceeded the threshold on the CES-D either at baseline or 9 months, and the 9-month incidence was 11.7%. Among the patients re-interviewed at both 6 and 9 months, the 6-month incidence was 12%, and the incidence between the 6- and 9-month assessments was 10%. Of the 292 patients with depression at baseline, 140 (47.6%) remained depressed at the 9-month follow-up. Baseline and 6-month CES-D score, in addition to perceived health at 6 months, explained 45% of the variance in the 9-month CES-D score. Patients above the threshold on the CES-D at any time were more likely to rate their health as fair or poor (69.8% vs 43.7%) and more likely to have an emergency room visit (40.4% vs 29.4%). These patients also had 38% more outpatient visits (7.7 vs 5.6) and 61% higher total outpatient charges ($1209 vs $751) than patients who never exceeded the CES-D threshold over the 9-month window (all P values < 0.01). Depressive symptoms were frequent and often persistent in this patient population. We identified patterns of oscillating severity of symptoms within individuals but relatively stable incidence and prevalence rates over a 9-month period. Patients who exceeded the threshold on the CES-D at any time during the study had significantly greater health services use and poorer perceived health.