Abstract
BACKGROUND/OBJECTIVES: Children receiving chemotherapy are highly susceptible to infection, and bloodstream infection (BSI) is a major cause of morbidity in febrile neutropenia. Post-transfusion fever represents a specific diagnostic dilemma, where febrile non-hemolytic transfusion reactions may be clinically indistinguishable from early BSI. We aimed to develop and internally validate a simple bedside score to predict BSI in children presenting to the ED with post-transfusion fever. METHODS: We performed a retrospective, single-center diagnostic prediction study of consecutive ED encounters between 2015 and 2024 in a tertiary children's hospital. Eligible encounters involved patients ≤ 18 years with an underlying malignancy receiving systemic chemotherapy who presented with fever within 24 h of red blood cell or platelet transfusion, had neutropenia, and with at least one blood culture obtained. BSI was defined as growth of a clinically significant pathogen within 48 h; episodes with only contaminants or colonizing flora were classified as non-BSI. Multivariable logistic regression with four prespecified predictors-transfusion-to-ED arrival interval, body temperature, absolute neutrophil count (ANC), and C-reactive protein (CRP)-was used to develop the model and derive a 0-5 point bedside score. Performance was assessed using AUC, diagnostic indices at prespecified cut-offs, calibration, and bootstrap internal validation. RESULTS: Of 507 screened encounters, 287 met inclusion criteria; 39 (13.6%) were adjudicated as BSI. The full model showed good discrimination (AUC 0.82). The derived score (2 points for ANC = 0/µL; 1 point each for temperature ≥ 38.5 °C, CRP ≥ 2.5 mg/dL, and transfusion-to-ED interval ≥ 7 h) achieved an AUC of 0.84. At a cut-off ≥2, sensitivity was 97.4% and negative predictive value 98.8%, misclassifying 1 of 39 BSIs as low risk; at ≥3, specificity was 59.7% with sensitivity 89.7%. Bootstrap-corrected AUC was 0.83. CONCLUSIONS: In children receiving chemotherapy who present with post-transfusion fever, a simple 0-5 point bedside score based on temperature, ANC, CRP, and transfusion-to-ED interval provided useful early stratification of BSI risk in this single-center cohort. Prospective multicenter validation is needed before clinical implementation.