Based on five years of surveillance in a neonatal intensive care unit (NICU), host and therapeutic risk factors for nosocomial infection were determined and the impact of staffing and environment on the rate of nosocomial infection was evaluated. From January 1974 to February 1977, infants occupied a crowded, hectic NICU that lacked basic infection control features, and 5.2% of the infants had at least one major nosocomial infection. The risk of nosocomial infection was associated with low birth weight, patent ductus arteriosus, surgery, and multiple supportive measures. After a new NICU opened in February 1977, 0.9% of the patients had major nosocomial infections (relative risk [old nursery/new nursery] = 5.06; P less than 10(-5); 95% confidence interval, 2.62-9.73). Improvements included 50% more nurses, increased space per infant, convenient sinks, and isolation facilities. Host and therapeutic risk factors for nosocomial infections were comparable in the old and new nurseries. The decrease in the rate of nosocomial infections therefore appeared to be due to improved staffing and environment.