Between 1977 and 1981, there were 4.9 episodes of community-acquired bacteremia and 5.1 episodes of hospital-acquired bacteremia per 1,000 patients in the four major hospitals of one metropolitan area. Case fatality rates were 30.1 per cent based on deaths due to all causes and 14.7 per cent based on deaths attributed specifically to bacteremia. Patients who experienced bacteremia had a 12-fold excess in mortality compared with other patients. Bacteremia occurred more frequently and was associated with greater case fatality rates at university-affiliated teaching hospitals compared with nonteaching community hospitals. At the nonteaching community hospitals, the odds of mortality for patients with bacteremia were lower even after adjustment for age, sex, severity of underlying medical problems, and severity of infection. Patients on private services at a teaching municipal hospital experienced greater odds of mortality compared with private patients at two nonteaching community hospitals. These latter observations may reflect, at least in part, limitations in the standard parameters used for determining severity of underlying medical problems and severity of infection.