A cohort of patients with diabetes mellitus hospitalised in Sweden from 1965 to 1983 was followed up until 1989, by linkages of population-based registers. Standardised mortality ratios (SMR), adjusted for confounding variables, and 95% confidence intervals (CIs) were calculated. After exclusion of the first year of follow-up (to reduce the effect of selection bias), the cohort consisted of 144,427 patients, of whom 92,248 patients died during follow-up. The SMR for all causes of death combined was 2.62 (95% CI 2.58-2.67) among men and 3.23 (95% CI 3.18-3.28) among women. The excess mortality was still evident 20 years after first hospitalisation, but became less marked with longer follow-up time. Patients with presumably insulin-dependent diabetes mellitus (IDDM) had the highest SMRs (10.2; CI 9.5-11.0); however, there was a significant (34%) improvement over time in their mortality risk. We conclude that excess mortality persisted throughout all calendar periods and at all ages, indicating the need for health care prevention measures.