The incidence of nosocomial infection in neonatal intensive care units (NICUs) is high compared with other wards. However, no definitions for hospital-acquired infection are available for NICUs. The aim of this study was to measure the incidence of such infections and to identify risk factors in the NICU of the VU University Medical Center, which serves as a level III regional NICU. For this purpose, a prospective surveillance was performed in 1998-2000. We designed definitions by adjusting the current definitions of the Centers for Disease Control and Prevention (CDC) for children <1 year of age. Birth weight was stratified into four categories and other baseline risk factors were dichotomized. Analysis of risk factors was performed by Cox regression with time-dependent variables. The relationship between the Clinical Risk Index for Babies (CRIB) and nosocomial infection was investigated. Furthermore, for a random sample of cases, we determined whether bloodstream infection and pneumonia would also have been identified with the CDC definitions. Seven hundred and forty-two neonates were included in the study. One hundred and ninety-one neonates developed 264 infections. Bloodstream infection (N=138, 14.9/1000 patient-days) and pneumonia (N=69, 7.5/1000 patient-days) were the most common infections. Of bloodstream infections, 59% were caused by coagulase-negative staphylococci; in 21% of neonates, blood cultures remained negative. In 25% of pneumonias, Enterobacteriaceae were the causative micro-organisms; 26% of cultures remained negative. Compared with the Nosocomial Infections Surveillance System (NNIS) of the CDC, our device utilization ratios and device-associated nosocomial infection rates were high. The main risk factors for bloodstream infection were birth weight [hazard ratio (HR) 1.79, 95% confidence interval (CI) 1.45-2.17] and parenteral feeding with hospital-pharmacy-produced, all-in-one mixture 'Minimix' (HR 3.69, 95%CI 2.03-6.69); administration of intravenous antibiotics (HR 0.39, 95%CI 0.26-0.56) was a protective risk factor. The main risk factors for pneumonia were low birth weight (HR 1.37, 95%CI 1.01-1.85) and mechanical ventilation (HR 9.69, 95%CI 4.60-20.4); intravenous antibiotics were protective (HR 0.37, 95%CI 0.21-0.64). In a subcohort of 232 very-low-birthweight neonates, the CRIB was not predictive for infection. With the CDC criteria, only 75% (21/28) of bloodstream infections and 87.5% of pneumonias (21/24) would have been identified. In conclusion, our local nosocomial infection rates are high compared with those of NICUs participating in the NNIS. This can be partially explained by: (1) the use of our definitions for nosocomial infection, which are more suitable for this patient category; and (2) the high device utilization ratios.