Recognition of Serious Bacterial Infections (Sepsis, Meningitis, and Urinary Tract Infection) in Febrile Infants Presenting to the Emergency Department: A Systematic Review of Clinical Prediction Rules

Background: Fever in infants aged 90 days or younger often leads to lumbar puncture, hospital admission and empirical antibiotics, although only a minority of these infants have a serious bacterial infection (SBI). Clinical prediction rules were developed to separate low-risk infants from those who need a full evaluation. Objective: To synthesize evidence on how well clinical prediction rules identify SBI and invasive bacterial infection (IBI) in febrile infants presenting to the emergency department. Methods: This systematic review followed the PRISMA 2020 statement. Two authors independently searched PubMed, Scopus, Web of Science and the Cochrane Library from inception to 31 May 2026. They screened records, extracted data and assessed risk of bias with QUADAS-2. Eligible studies derived or validated a prediction rule in febrile infants aged ≤90 days and reported accuracy against culture-confirmed SBI or IBI. Narrative synthesis was used because the studies were heterogeneous. Results: Ten studies met the inclusion criteria. Together they covered more than 17,000 infant evaluations, with partial overlap between cohorts. Rules combining urinalysis, absolute neutrophil count and procalcitonin gave the best balance of sensitivity and specificity: sensitivity was 92.0-97.7% and negative predictive value was above 99%. The updated PECARN rule kept 94.2% sensitivity in neonates and missed no case of meningitis. Used without procalcitonin, the American Academy of Pediatrics algorithm reached sensitivity of 88.9-100%, with specificity of 0-40.4%. Point-based rules and rules without procalcitonin showed lower specificity. Conclusion: Rules based on procalcitonin reliably identify febrile infants at low risk. External validation in neonates and in resource-limited settings is still limited.