Abstract
OBJECTIVE: To derive and internally validate a clinical prediction rule to identify febrile infants aged 61-90 days at low risk of invasive bacterial infections (IBIs). METHODS: Using data from 17 Pediatric Emergency Care Applied Research Network Registry (PECARN) emergency departments, we included noncritically ill, previously healthy infants aged 61-90 days with temperatures greater than or equal to 38°C and urinalyses and blood cultures obtained between January 1, 2012, and April 30, 2024. Our outcome was IBI, defined as growth of pathogenic bacteria from blood or cerebrospinal fluid culture. Using recursive partitioning with 10-fold cross-validation, we derived and internally validated a prediction rule using age, temperature, urinalysis (negative/positive), and absolute neutrophil count (ANC) as candidate predictors. Limiting the analysis to infants with procalcitonin (PCT) and ANC results, we evaluated PCT as an additional predictor. RESULTS: Of 4952 infants included, 100 (2.0%) had IBIs, including 95 (1.9%) with bacteremia without meningitis and 5 (0.1%) with bacterial meningitis. The optimal prediction rule identified low-risk infants as those with negative urinalyses and highest temperatures less than or equal to 38.9°C, yielding a sensitivity of 86.0% (95% CI, 77.6-92.1) and specificity of 58.9% (95% CI, 57.5-60.3). In the subset of 1207 infants with PCT and ANC measurements, including 27 (2.2%) with IBIs (2 [0.2%] with bacterial meningitis), we identified PCT of 0.24 ng/mL or less and ANC of 10 710 cells/mm3 or less as low-risk predictors. This PCT-based rule demonstrated sensitivity of 100.0% (95% CI, 87.2-100.0) and specificity of 65.8% (95% CI, 63.0-68.5). CONCLUSIONS: We derived 2 accurate clinical prediction rules to identify febrile infants aged 61-90 days at low risk of IBIs when urine and blood testing are obtained. Prospective validation is needed.