BACKGROUND: Nosocomial infections are common in intensive care units (ICU). The objectives of this study were to determine risk factors of ICU-acquired infections, and potential mortality attributable to such infections. METHODS: An observational study was performed in a 10-bed multidisciplinary ICU. For a period of 27 months, all patients admitted for >or=48 h were included. Infections were diagnosed according to Centers for Disease Control and Prevention definitions. Airway colonization was explored by molecular typing. Risk factors for infection were determined by multivariable logistic regression. Survival was analyzed with time-varying proportional hazards regression. RESULTS: Of 278 patients, 81 (29%) were infected: urinary tract infections in 39 patients (14%), primary bloodstream infections in 25 (9%), surgical site infections in 22 (8%) and pneumonia in 21 (8%). Of the total of 147 episodes, Gram-negative bacilli were isolated in 90, Gram-positive cocci in 49 and Candida sp. in 25. Risk factors for pneumonia were mechanical ventilation [odds ratio (OR=7.9, CI 1.8-35), lack of enteral nutriment (OR=8.0, CI 1.4-45) and length of time at risk (OR=1.8, CI 1.2-2.8), while gastric acid inhibitors did not affect the risk (OR=0.99, CI 0.32-3.0). Transmission of bacteria from the stomach to the airway was not confirmed. The risk of death was increased as patients were infected with pneumonia [hazard ratio (HR)=3.6; CI: 1.6-8.1], or primary bloodstream infection (HR=2.5; CI: 1.2-5.4), independent of age and disease severity. CONCLUSIONS: Mortality was increased by ICU-acquired pneumonia and primary bloodstream infections. Our findings did not support the gastro-pulmonary hypothesis of ICU-acquired pneumonia. The proposition that blood transfusions increase the risk of ICU-acquired nosocomial infections was not supported.