The study aims to establish the predictive value of a diagnosis of depression among elderly according to the 10th revision of the International Statistical Classification of Diseases (ICD-10) by measuring morbidity, medication usage, health service utilization and mortality during an 8-year follow-up of depressed elderly inpatients (n=76) and community-living depressed patients (n=38) compared with controls (n=116). The data were taken from GPs' medical records and health statistics registers. At baseline, no significant differences were observed between the two cohorts of depressed patients and the controls in terms of prevalence of cardiovascular, respiratory or cerebrovascular morbidity. During follow-up, both cohorts of depressed patients had significantly increased rates of recurrent depressions, consumption of antidepressants, psychiatric in- and outpatient admissions, and home visits; inpatients used more psychiatric hospital days. Health service utilization in somatic hospitals and somatic diagnoses was not significantly increased. Inpatients used significantly fewer GP office-hour services but more out-of-hours services than the control group. Community-living depressed patients experienced no significant increase in use of GP services. Survival was unaffected in both cohorts. In agreement with other studies, especially inpatient depression predicted increased rates of recurrent depressions and increased use of psychiatric hospital services, indicating poor long-term outcome. Inpatients consumed fewer GP office-hour services but more out-of-hours services, possibly due to less office-hour contact. Contrasting with other studies, ICD-10 depression among elderly predicted no increase in the use of somatic hospital facilities.