Uncontrolled studies have suggested that Bacteroides fragilis group bacteremia has an attributable mortality rate between 13% and 33%. To determine the true attributable mortality rate and the mortality risk ratio associated with bacteremia due to the B. fragilis group, we conducted a matched-pair study in which cases were matched to controls for age, gender, year of admission, principal discharge diagnosis, and types of major surgery by an investigator blinded to survival status. Cases and controls were comparable in demographic and clinical characteristics. Cases had a significantly higher mortality rate (28% vs. 8.7%, P = .002, McNemar's test), with an attributable mortality rate of 19.3% (95% CI, 8%-30%; P = .003) and a mortality risk ratio of 3.2. In a multivariate analysis, three clinical factors were independently correlated with mortality: the presence of B. fragilis group bacteremia (RR: 4.9; 95% CI: 3.7-6.0; P = .009), congestive heart failure (RR: 8.0; 95% CI: 6.6-9.3; P = .003) or chronic liver disease (RR: 6.3; 95% CI: 4.8-7.7; P = .01). Cases also had a 16-day-longer stay in the hospital (P = .0007, Wilcoxon's signed rank test) compared with controls. Thus, B. fragilis group bacteremia contributes significantly to morbidity and mortality.