For identification of risk factors for bloodstream infection (BSI) among neonatal intensive care unit patients, prospective 6-month studies in three neonatal intensive care units were conducted. BSI was diagnosed in 42 of 376 (11.2%) enrolled infants. Pathogens included coagulase-negative staphylococci,Candidasp., Group B streptococci and Gram-negative species. Patients with BSIs were more liley to die during their neonatal intensive care unit stay than were patients who did not acquire BSIs (6 to 42vs.11 of 334,P= 0.007). BSI rate was highest in infants with birth weight < 1500 g (relative risk (RR) = 6.8,P< 0.001), these treated with H-2 blockers (RR = 4.2,P< 0.001) or theophylline (RR = 2.8,P< 0.001) and those with admission diagnoses referable to the respiratory tract (RR = 3.7,P< 0.001). Infants who developed BSI were more severely ill on admission than other infants (Median physiologic stability index 13vs.10 (P< 0.001) and were of lower gestational age (28vs.35 weeks,P< 0.001). In logistic regression analysis, risk of BSI was independently associated only with very low birth weight, respiratory admission diagnoses and receipt of H-2 blockers. Risk of isolation of a pathogen from blood culture was independently associated with Broviac, umbilical vein or peripheral venous catheterization > 10, 7 or 3 days, respectively, at one insertion site. Rate of isolation of a pathogen was higher (9 of 59 (15%)) within 48 hours of a measurable serum interleukin 6 concentration than an interleukin 6 level of 0 pg/ml (10 of 159 (6%),P= 0.04). Conversely < 1 day of exposure to gentamicin or ampicillin before the sepsis evalutation was associated with lower BSI risk in infants with intravascular catheters (20 of 127 (16%)vs.9 of 16 (56%),P= 0.06). These findings indicate that very low birth weight, respiratory diagnoses, H-2 blocker use and prolonged intravascular cath-eterization at one insertion site are associated with elevated risk of BSI. Clinical trials of interventions addressing these risk factors are warranted.