Prehospital Risk Stratification for Early Transfusion in Pediatric Trauma: A National Trauma Data Bank Analysis

OBJECTIVES: Pediatric criteria remain limited for identifying injured children who may benefit from early blood transfusion for hemorrhagic shock. This study evaluated whether emergency medical services (EMS)-observable variables in a national trauma registry could identify children at risk for early hospital blood transfusion. METHODS: We conducted an observational retrospective cohort study using the National Trauma Data Bank (2017-2020). Among 391,126 records for patients aged 1-15 years, we included children transported by EMS directly from the scene to a participating trauma center after excluding burn injuries, interfacility transfers, and missing pre-identified EMS variables (systolic blood pressure, heart rate, or Glasgow Coma Scale [GCS]). The primary outcome was any blood product administered within 4 hours of emergency department arrival. Five prespecified EMS-observable predictors composed the model: age-adjusted hypotension, age-adjusted tachycardia, GCS ≤14, penetrating mechanism, and prehospital traumatic cardiac arrest. We fit a multivariable logistic regression and a classification and regression tree (CART) using an 80/20 stratified train-validation split. Missing data sensitivity analyses included multiple imputation and CART refitting in the broader eligible cohort. RESULTS: Among 79,764 children, 2,634 (3.3%) received early transfusion in the hospital. All five predictors were independently associated with early transfusion, led by GCS ≤14 (adjusted odds ratio [aOR] 8.38; 95% confidence interval [CI], 7.61-9.22), penetrating mechanism (aOR 5.87; 95% CI, 5.28-6.51), and age-adjusted hypotension (aOR 5.39; 95% CI, 4.74-6.13). In the validation cohort, CART analysis stratified patients into low-, intermediate-, and high-risk tiers comprising 77.0%, 19.7%, and 3.3% of patients, respectively. Early-transfusion rates were 1.2%, 6.1%, and 36.7% across these tiers. The high-risk tier accounted for 36.8% of all early transfusion events. Using the high-risk tier as a decision threshold, sensitivity was 36.8%, specificity 97.8%, positive predictive value 36.7%, and negative predictive value 97.8%. CONCLUSIONS: Five EMS-observable features, including age-adjusted hypotension, age-adjusted tachycardia, GCS ≤14, penetrating mechanism, and prehospital traumatic cardiac arrest, identified a small subgroup of pediatric trauma patients with substantially elevated risk of early hospital transfusion. This parsimonious framework may inform prospective validation of prehospital transfusion criteria.