Previous studies of the value of the complete blood count (CBC) in distinguishing viral from bacterial infection in young febrile children have failed to exclude children with clinically evident bacterial infection and thus have inflated the positive predictive value of the test for occult focal infection. We prospectively studied 2492 children 3-24 months of age who presented to a children's hospital emergency department between March 1989 and August 1990 with fever (> or = 38.0 degrees C) of acute (< or = 4 days) onset but no evident bacterial focus of infection, 433 (17.4%) of whom received a CBC. We also carried out an 8-year retrospective analysis to estimate prior, or pre-test, probabilities (prevalences) and examine CBC results for rare occult bacterial infections (meningitis, osteomyelitis, and septic arthritis). Estimated prior probabilities for the four most common categories of infection that can be diagnosed at the initial visit were: non-pneumonitic viral infection, 88.6% in boys and 86.0% in girls; pneumonia, 8.5% in both sexes; urinary tract infection (UTI), 3.0% in boys and 5.5% in girls; and bacterial meningitis, 0.0066% in both sexes. The likelihood (sensitivity) of a total white blood cell (WBC) count > or = 15,000/mm3 was 25.5, 64.5, 62.5, and 50.0% for viral infection, pneumonia, UTI, and meningitis, respectively. Among children with a high total white blood cell count, neither a total polymorphonuclear count > or = 10,000/mm3 nor a band count > or = 500/mm3 was associated with significantly elevated likelihoods for occult pneumonia or UTI, a finding confirmed by multiple logistic regression analysis.(ABSTRACT TRUNCATED AT 250 WORDS)