OBJECTIVE: To evaluate antimicrobial use and the influence of inadequate empiric antimicrobial therapy on the outcomes of nosocomial bloodstream infections (BSIs). DESIGN: Prospective cohort study with nested case-control analysis. SETTING: Neonatal intensive care unit (NICU). METHODS: All patients weighing 2,000 g or less were enrolled. Data collection included risk factors for nosocomial BSI, admission severity of illness, microbiology, antimicrobial therapy, and outcomes. Inadequate empiric antimicrobial therapy was defined as the use of antibiotics for more than 48 hours after the day that blood cultures were performed that did not cover the microorganisms causing the bacteremia or administration of antibiotics that failed to cover resistant microorganisms. RESULTS: Two hundred twenty-nine patients were enrolled. Forty-five developed nosocomial BSIs. The BSI rates were 11.2, 2.8, and 0 per 1,000 catheter-days for patients weighing 1,000 g or less, between 1,001 and 1,500 g, and between 1,501 and 2,000 g, respectively. After adjustment for severity of illness, the mortality in patients with nosocomial BSI receiving inadequate empiric antimicrobial therapy was higher than in those receiving adequate therapy (adjusted odds ratio [AOR], 5.3; 95% confidence interval [CI95], 1.2-23.2). By multivariate analysis, nosocomial BSI attributed to Candida species (AOR, 6.3; CI95, 1.4-28.0) and invasive procedure prior to onset of BSI (AOR, 6.4; CI95, 1.0-39.0) were associated with administration of inadequate empiric antimicrobial therapy. CONCLUSIONS: Administration of inadequate empiric antimicrobial therapy among NICU patients with nosocomial BSI was associated with higher mortality. Additional studies on the role of inadequate empiric antimicrobial therapy and the outcomes of BSIs among NICU patients are needed.