The treatment of enterococcal bacteremia not associated with endocarditis has been controversial. We retrospectively reviewed 81 episodes of enterococcal bacteremia and categorized them as to their clinical significance, using a strict case definition. Of the 81 episodes, 41 met our criteria for clinical significance. Mortality was 51% among the 41 patients with clinically significant bacteremia and 50% among the 40 patients with bacteremia of uncertain clinical significance. Despite these equivalent overall mortality figures, antibiotic therapy specific for Enterococcus species was associated with reduced in-hospital mortality among patients with clinically significant infections (relative risk [RR] = 0.46, 95% confidence interval [CI] = 0.27-0.77); mortality was also reduced in the first 7 days after the detection of bacteremia, when death was relatively likely to be directly due to the bacteremic episode (RR = 0.17, CI = 0.04-0.74). The association between appropriate antibiotic therapy and reduced mortality remained statistically significant when adjustments were made for a number of other factors related to mortality, including age, underlying conditions, prior use of antibiotics, nosocomial acquisition, polymicrobial etiology, prior surgery, and source of infection. Thus enterococcal isolates from the blood, even when of doubtful clinical significance, are poor prognostic markers associated with high mortality. However, when the clinical significance of bacteremia is defined by strict criteria, specific therapy against Enterococcus species is associated with improved outcome.