Positive blood cultures in very low birthweight or preterm infants usually reflect bacteraemia, septicaemia, or failure of asepsis during sampling and lead to increased costs and length of stay. Rates of nosocomial, or hospital acquired, bacteraemia may therefore be important indicators of neonatal unit performance, if comparisons are adjusted for differences in initial risk. In a preliminary study the risk of nosocomial bacteraemia was related to initial clinical risk and illness severity measured by the clinical risk index for babies (CRIB). Nosocomial bacteraemia was defined as clinically suspected infection with culture of bacteria in blood more than 48 hours after birth. One or more episodes of nosocomial bacteraemia were identified retrospectively in 36 of 143 (25%) infants in a regional neonatal unit between 1992 and 1994. Biologically plausible models were developed using regression analysis techniques. After correcting for period at risk, nosocomial bacteraemia was independently associated with gestation at birth and CRIB. Death was independently associated with CRIB, but not with nosocomial bacteraemia. CRIB may contribute, with other explanatory variables, to more comprehensive predictive models of death and nosocomial infection. These may facilitate future risk adjusted comparative studies between groups of neonatal units.