Demographic information, risk factors, therapy, and outcome for all patients who had candidemia at Barnes Hospital, St. Louis, between 1 September 1988 and 1 September 1989 were retrospectively reviewed. One hundred six candidemic patients were identified, representing 0.5% of all medical and surgical discharges and 0.33% of total patient discharges. These percentages represent a 20-fold increase in the incidence of candidemia at our hospital in comparison with that during 1976-1979. Candida albicans was the most frequently isolated species (63%), followed by Candida tropicalis (17%), Candida glabrata (13%), Candida parapsilosis (6.5%), and Candida krusei (0.9%). Overall mortality was 57%, and 14 (23%) of 60 deaths occurred within 48 hours of the detection of candidemia. Mortality was associated with higher APACHE II scores (25 for nonsurvivors vs. 16 for survivors; P = .0001), the presence of a rapidly fatal underlying illness (P = .0009), and sustained positivity of blood cultures (P = .02). In cases of sustained candidemia, the isolation of non-albicans Candida species also correlated with increased mortality (8 of 8 vs. 10 of 21; P = .005). Thirty candidemic patients (28%) did not receive any antifungal therapy, and 19 (63%) of these untreated patients died. Eleven untreated patients (37%) survived without sequelae. There has been a marked increase in the incidence of candidemia in our institution that is associated with a high overall mortality. Candidemia lasting less than 24 hours was associated with a lower mortality than was that of longer duration. Severity of illness and duration of candidemia should be used as stratifying factors in prospective studies to determine optimum therapy.