OBJECTIVE To evaluate the risk factors for nosocomial sepsis among infants hospitalized in 23 neonatal intensive care units. METHODS Risk factors for nosocomial sepsis among 5760 admissions are analyzed by birth weight groups, <1 kg, 1 to 1.5 kg, and >1.5 kg. A Cox hazard regression model was used to evaluate further detail in the two lower weight groups. RESULTS Use of corticosteroids had no effect on the incidence of nosocomial sepsis in the two lower weight categories although it was significant among the >1.5 kg infants. In a simple Cox model, significant risk factors included lowest birth weight category, ventilatory support, and presence of a central venous catheter. The complex Cox model revealed that an increase in total days of presence of central arterial catheter, use of antibiotics, and ventilatory support were significant but that total days of presence of a central venous catheter was not. A model for Candida sepsis revealed as a risk factor an increase in total days of use of antibiotics before infection. CONCLUSIONS The risk for infection associated with presence of a central venous catheter is the same for each day of exposure (i.e., the same risk on day 5 of presence of the line as on day 30), but the risk associated with ventilatory support increases over time. Candida sepsis is associated with prolonged antibiotic use before the first episode of nosocomial sepsis and not with birth weight group.