To determine optimal clinical laboratory techniques for detecting pediatric bacteremia, we studied 7,768 consecutive blood cultures in a 1-year period. Blood was inoculated into one vented 50-ml bottle of brucella broth with 0.05% sodium polyanetholsulfonate and one unvented 50-ml bottle of Columbia broth with 0.05% sodium polyanetholsulfonate and 0.05% cysteine. Bottles were visually examined for growth on days 1 through 7 and blindly subcultured aerobically and anaerobically on days 1, 2, and 7. There were 724 (9.3%) positive cultures, and 484 (6.2%) were clinically significant. The most frequent isolates from bacteremic patients were Haemophilus influenzae (24%) and Streptococcus pneumoniae (17%). Growth was noted in only one bottle in 25% of clinically significant isolates. Bottles inoculated with greater than or equal to 1 ml of blood became positive earlier than bottles inoculated with less than 1 ml. After 1 day of incubation, 48% of the clinically significant cultures showed growth on visual examination, whereas 85% showed growth on subculture. Only 19% of Haemophilus isolates were detected visually on day 1, whereas 88% were recovered on subculture. By day 7, 3.5% of all isolates (including 18% of pneumococcal isolates and 1% of Haemophilus isolates) could no longer be recovered on subculture. We conclude that a two-bottle blood culture system and blind subculture within 24 h will optimize detection of pediatric bacteremia.