OBJECTIVE: To determine the excess length of stay, extra costs, and mortality attributable to nosocomial bloodstream infection in critically ill patients. DESIGN: Pairwise-matched (1:1) case-control study. SETTING: Surgical intensive care unit (SICU) in a tertiary health care institution. PATIENTS: All patients admitted in the SICU between July 1, 1988, and June 30, 1990, were eligible. Cases were defined as patients with nosocomial bloodstream infection; controls were selected according to matching variables in a stepwise fashion. METHODS: Matching variables were primary diagnosis for admission, age, sex, length of stay before the day of infection in cases, and total number of discharge diagnoses. Matching was successful for 89% of the cohort; 86 matched case-control pairs were studied. MAIN OUTCOME MEASURES: Crude and attributable mortality, excess length of hospital and SICU stay, and overall costs. RESULTS: Nosocomial bloodstream infection complicated 2.67 per 100 admissions to the SICU during the study period. The crude mortality rates from cases and controls were 50% and 15%, respectively (P <.01); thus, the estimated attributable mortality rate was 35% (95% confidence interval, 25% to 45%). The median length of hospital stay significantly differed between cases and controls (40 vs 26 days, respectively; P <.01). When only matched pairs who survived bloodstream infection were considered (n = 41), cases stayed in the hospital a median of 54 days vs 30 days for controls (P <.01), and cases stayed in the SICU a median of 15 days vs 7 days for controls (P <.01). Thus, extra hospital and SICU length of stay attributable to bloodstream infection was 24 and 8 days, respectively. Extra costs attributable to the infection averaged