We undertook a prospective, concurrent comparison of the prevalence, predictability, and outcome of bacteremia in children from 3 to 24 months of age with temperatures greater than or equal to 39.5 degrees C in three diverse clinical settings: primarily black lower-class children at an inner-city hospital (n = 532), primarily white middle-class children at a suburban hospital (n = 160), and primarily white middle-class children in offices of pediatricians in private practice (n = 94). The prevalence of bacteremia for the entire study sample (3.1% to 7.4%) and outpatients only (1.9% to 5.9%) was not statistically different among the three groups. There were no statistically significant differences among the three groups in identifying children with bacteremia (P greater than 0.05). There was no racial, geographic, or socioeconomic predilection for bacteremia in infants. At the first visit, antibiotics were prescribed (most commonly for otitis media) for 23 of the 25 bacteremic patients who were not initially hospitalized. One patient with otitis media developed meningitis. The others had uncomplicated courses and were well by 96 hours (most by 48 hours). In office settings, private practitioners were no better in predicting bacteremia in familiar patients than they were with first-time patients. Information from blood culture did not appear to alter patient management. We conclude, therefore, that routine blood cultures are unnecessary for all highly febrile infants given antibiotics.