The hospital-wide attack rate for Clostridium difficile-associated diarrhea at our tertiary-care university hospital was 0.02 per 100 patient discharges (0.02%) in 1982, but 0.41% and 1.47% in 1986 and 1987, respectively, with a peak incidence of 2.25% in the fourth quarter of 1987. Hospital antibiotic usage patterns showed concurrent increased use of third-generation cephalosporins, and intravenous vancomycin and metronidazole. Thirty-seven cases selected for study were older than 37 control patients, more likely to have an underlying malignancy and less likely hospitalized on the obstetrics/gynecology service. Their mean duration of hospitalization prior to diagnosis was 21 days, versus a mean total length of stay of eight days for controls. All cases received antibiotics, compared to 24 of the controls. Cases were given more antibiotics for longer periods, and more often received clindamycin, third-generation cephalosporins, aminoglycosides and vancomycin. Gender, race, duration of hospitalization, prior surgery and antiulcer therapy were not significant by logistic regression analysis. Epidemiologic variables with significantly different adjusted odds ratios (95% confidence intervals) were age greater than 65 years (14.1, 1.4-141), intensive care unit residence (39.2, 2.2-713), gastrointestinal procedure (23.2, 2.1-255) and more than ten antibiotic days (summation of days of each antibiotic administered) (16.1, 2.2-117). Control measures included encouraging earlier isolation and treatment of suspected cases and formulary restriction of clindamycin, with use of metronidazole for therapy of anaerobic infections. By the second half of 1988, the attack rate had dropped progressively to 0.74%.