Background: Conventional practice involves obtaining a blood culture during or immediately after a fever to increase diagnostic yield. There are no data to support this practice in children.
Methods: Retrospective single-center case-control study of children (0-18 years) who had blood cultures performed as part of routine care. Cases had an a priori defined pathogen isolated from blood culture (n = 410) and were age-matched with contemporaneous controls with a sterile blood culture (n = 410). The predictive value of fever (before and after blood culture), C-reactive protein and hematologic indices were analyzed by multivariate regression and area under the receiver operating characteristic curves (AUCs) in neonatal, general pediatric and pediatric oncology patients.
Results: One thousand one hundred seventy-two (6.7%) of 17,607 blood cultures were positive, of which 410 (35%) cultured pathogen(s). Three hundred and twenty four (79%) cases and 275 (67.1%) controls had a fever (≥37.5°C) during the 12 hours pre- or post-collection. Fever 2-6 hours before a blood culture was neither sensitive nor specific for predicting bacteremia in neonatal or pediatric patients and marginally predictive in oncology patients (AUC 0.59-0.63). Cultures obtained 2-6 hours before fever were nonpredictive in neonates (AUC 0.56-0.59), marginally predictive in pediatric patients (AUC 0.64-0.67) and moderately predictive in oncology patients (AUC 0.70). C-reactive protein was marginally predictive in neonates (AUC 0.60). Hematologic indices were nonpredictive in all groups.
Conclusions: Fever before obtaining blood culture was neither sensitive nor specific for culture positivity; timing of pediatric blood cultures relative to fever is unimportant. Bacteremia precedes a fever, but this is of limited clinical applicability.