Reappraisal of criteria used to predict serious bacterial illness in febrile infants less than 8 weeks of age.

OBJECTIVE: To re-evaluate the Philadelphia protocol and the Rochester criteria for identifying infants at low risk for serious bacterial illness (SBI) in a new population.
METHODS: The authors prospectively enrolled infants 56 days of age or younger with rectal temperatures greater than 100.6 degrees F. Physicians assigned an overall impression of sepsis and objectively scored each infant using the Infant Observation Score. Following a history and physical examination, a complete sepsis evaluation was performed. Infants were considered to have SBI if their blood, urine, cerebrospinal fluid, or stool cultures grew pathogenic bacteria. Infants were assigned to high- and low-risk groups for SBI according to the Philadelphia protocol and the Rochester criteria by a single investigator blinded to the final culture results. The test performance parameters of the Philadelphia protocol and the Rochester criteria in this population were compared with those reported from previous validation studies.
RESULTS: One hundred eighty-one infants were assigned to risk groups using the Philadelphia protocol, and 259 infants using the Rochester criteria. In this population, the negative predictive value (NPV) of the Philadelphia protocol was 97.1% (95% confidence interval [95% CI] = 85.1% to 99.8%), compared with 99.7% in the original report, and the NPV of the Rochester criteria was 97.3% (95% CI = 90.5% to 99.2%), compared with a prior report of 98.9%.
CONCLUSIONS: The Philadelphia protocol and the Rochester criteria maintained their previously reported NPVs when applied to a new population of febrile infants. These data illustrate the usefulness of retesting clinical decision rules in new populations prior to their universal acceptance.