PROJECT SUMMARY Of the ~200,000 infants S90 days old evaluated for fever annually in U.S. emergency departments (EDs), ~2% will harbor invasive bacterial infections (I Bis), defined by bacteremia and/or bacterial meningitis. To reduce unnecessary invasive procedures and hospitalizations, clinicians frequently rely on prediction rules to identify febrile infants at low risk of I Bis who may be safely discharged from the ED without lumbar punctures (LPs) or empirical antibiotics. Newer clinical prediction rules such as the Step-by-Step approach, Pediatric Emergency Care Applied Research Network (PECARN) prediction rule, and IBI score have high sensitivity for IBI in febrile infants and potentially obviate the need for LPs. However, these rules have been derived, validated, and/or studied exclusively in febrile infants S60 days old or in cohorts with few infants 61 to 90 days old with IBI. The recently published American Academy of Pediatrics clinical practice guideline for the evaluation and management of febrile infants only addresses infants 8 to 60 days old due to the lack of evidence regarding febrile infants 61 to 90 days old. Consequently, there is wide variation in the management of these infants. In addition, although bacterial meningitis is rare in this age group, ~18% of febrile infants 61 to 90 days old undergo LPs, and 16% are hospitalized; at some hospitals, the proportion approaches two-thirds. Nevertheless, I Bis occur in approximately 1 % to 1.5% of febrile infants 61 to 90 days old, which underscores the need for risk stratifying these infants to identify those with non-negligible risk of IBI who require LPs and hospitalization. The long-term goal of this research is to improve the management of febrile infants 61 to 90 days old by providing clinicians with an evidence-based risk stratification tool. The objective of this proposal is to derive and internally validate such a clinical prediction rule using objective demographic, clinical, and laboratory data to identify infants at low versus non-negligible risk of IBI. To accomplish this objective, we will use the PECARN Registry, which includes data from >5.5 million visits from 12 EDs in 9 health systems, including nearly 14,000 visits for febrile infants 61 to 90 days old. Recursive partitioning analysis will be used to derive the clinical prediction rule with 10-fold cross internal validation. Collectively, the investigative team is well-positioned to conduct this research with expertise in febrile infants, clinical prediction modeling, and natural language processing. The principal investigator, Paul L. Aronson, MD, MHS, has studied febrile infants for a decade, including the derivation and internal validation of a prediction rule for febrile infants S60 days old. After the derivation and internal validation of a highly sensitive and specific prediction rule for febrile infants 61 to 90 days old at risk for IBI, the rule will be externally validated in a multicenter prospective ...