Extended-spectrum-β-lactamase (ESBL)-producing pathogens are associated with extensive morbidity and mortality and rising health care costs. Scant data exist on the impact of antimicrobial therapy on clinical outcomes in patients with ESBL bloodstream infections (BSI), and no large studies have examined the impact of cefepime therapy. A retrospective 3-year study was performed at the Detroit Medical Center on adult patients with BSI due to ESBL-producing Klebsiella pneumoniae or Escherichia coli. Data were collected from the medical records of study patients at five hospitals between January 2005 and December 2007. Multivariate analysis was performed using logistic regression. One hundred forty-five patients with BSI due to ESBL-producing pathogens, including K. pneumoniae (83%) and E. coli (16.5%), were studied. The mean age of the patients was 66 years. Fifty-one percent of the patients were female, and 79.3% were African-American. Fifty-three patients (37%) died in the hospital, and 92 survived to discharge. In bivariate analysis, the variables associated with mortality (P < 0.05) were presence of a rapidly fatal condition at the time of admission, use of gentamicin as a consolidative therapeutic agent, and presence of one or more of the following prior to culture date: mechanical ventilation, stay in the intensive care unit (ICU), and presence of a central venous catheter. In multivariate analysis, the predictors of in-hospital mortality included stay in the intensive care unit (odds ratio [OR], 2.17; 95% confidence interval [CI], 0.98 to 4.78), presence of a central-line catheter prior to positive culture (OR, 2.33; 95% CI, 0.77 to 7.03), presence of a rapidly fatal condition at the time of admission (OR, 5.13; 95% CI, 2.13 to 12.39), and recent prior hospitalization (OR, 1.92; 95% CI, 0.83 to 4.09). When carbapenems were added as empirical therapy to the predictor model, there was a trend between empirical carbapenem therapy and decreased mortality (OR, 0.61; 95% CI, 0.26 to 1.50). When added to the model, receipt of empirical cefepime alone (n = 43) was associated with increased mortality, although this association did not reach statistical significance (OR, 1.66; 95% CI, 0.71 to 3.87). The median length of hospital stay was shorter for patients receiving empirical cefepime than for those receiving empirical or consolidated carbapenem therapy. In multivariate analysis, empirical therapy with cefepime for BSI due to an ESBL-producing pathogen was associated with a trend toward an increased mortality risk and empirical carbapenem therapy was associated with a trend toward decreased mortality risk.