RH Pantell, DA Bergman, JI Takayama, TB Newman, J Bernzweig, M Spitalny, S Finch, RC Wasserman. Pediatrics, University of California, San Francisco, San Francisco, CA; Pediatrics, Stanford University, Stanford, CA; and Pediatric Research in Office Settings (PROS), Center for Child Health Research, American Academy of Pediatrics, Elk Grove Village, IL. 

Presented as a presidential plenary at the 2000 Pediatric Academic Societies Annual Meeting.

Background: The evaluation of febrile infants is controversial. While clinical guidelines exist, their usefulness in primary care practice has not been tested.

Objective: To determine the management and clinical outcomes of febrile (>=38 C) infants (<=3 months) seen in primary care settings in order to develop an optimal clinical strategy to detect bacteremia and bacterial meningitis (B/BM).

Design/Methods: Prospective cohort study from February 1995 to April 1998. Data on 3066 infants were collected in 220 practice locations by 577 members of the AAP PROS Network in the US, Canada and Puerto Rico. Clinicians managed infants according to their usual standard of care.

Results: Much of the serious disease occurs in the first month:

Clinicians did not follow established guidelines: only 42.7% of infants <=1 month were hospitalized and treated; 31.8% of sick appearing infants 1-3 months were hospitalized/treated; 44.4% of well appearing infants 1-3 months had a WBC and UA. Nevertheless, pediatricians initially treated, with antibiotics, 98.4% of infants with B/BM; the specificity of their management was 48.2%. This clinical practice resulted in greater treatment of infants with B/BM as well as 32% fewer hospitalizations and less antibiotic use than would have occurred if practitioners followed current guidelines (hospitalize and treat infants <1 month, obtain WBC and UA on infants 1-3 months) which would have resulted in antibiotic treatment of 93.5% of infants with B/BM and specificity of 34.3%. Using extensive modeling we were unable to develop a clinical prediction scheme with superior performance to current practice.

Conclusion: This study provides data on the incidence of and risks for serious bacterial illness in community practice. It is of interest that clinicians' performance exceeded any existing or new model we could develop. This suggests that (1) guidelines for a complex clinical problem, such as fever in infancy, should not be applied rigorously but allow for variability based on clinician judgment (2) guidelines useful in emergency rooms may be less applicable in primary care.