External validation of a decision rule for bacteremia vs contaminants in pediatric blood cultures
Discuss this preprint
Start a discussion What are Sciety discussions?Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
BACKGROUND
Half of positive blood cultures in pediatric emergency departments (PEDs) represent contaminants, driving unnecessary hospitalization, antibiotic exposure, and repeat visits. A clinical decision rule derived at CHU Sainte-Justine showed 99% sensitivity and 60% specificity for distinguishing bacteremia from contaminants but had not been externally validated. We sought to validate this rule in an independent pediatric cohort.
METHODS
This retrospective diagnostic study spanned from January 2015 to May 2025 at a tertiary PED in Switzerland, using positive blood cultures from patients younger than 16 years. The four predictors (Gram-negative organisms or Gram-positive cocci in pairs or chains; time to positivity <17 hours; indwelling device; suspected osteoarticular infection) classified each case as low, moderate, or high risk. The primary outcome was bacteremia, adjudicated by two independent reviewers, based on organism identity and infectious disease specialist’s assessment. Diagnostic accuracy was assessed with 95% CIs.
RESULTS
Of 130 children enrolled (median age 3.8 years [IQR 0.9-9.9]; 61.5% male), 78 (60.0%) had true bacteremia. The rule yielded a sensitivity of 97.4% (95% CI, 91.0-99.7), specificity of 69.2% (95% CI, 54.9-81.3), positive predictive value of 82.6% (95% CI, 73.3-89.7), and negative predictive value of 94.7% (95% CI, 82.3-99.4). Both false-negatives were immunocompetent children with methicillin-susceptible Staphylococcus aureus bacteremia without indwelling devices. Among contaminants, 71% received antibiotics under usual care versus 31% classified as moderate or high risk by the rule.
CONCLUSIONS
This first external validation supports the Sainte-Justine rule in a distinct pediatric population, preserving sensitivity with higher specificity. Multicenter validation is warranted before adoption.
