Predictive performance of clinical, physiological, and trauma scores for life-saving interventions in pediatric trauma: a retrospective cohort study

Abstract Background Early identification of pediatric trauma patients requiring life-saving interventions (LSIs) is fundamental for optimizing emergency triage and allocation planning. This study aimed to evaluate and compare the predictive effectiveness of the Abbreviated Injury Scale (AIS), Injury Severity Score (ISS), Pediatric Trauma Score (PTS), Shock Index (SI), and the BIG score for determining LSI requirements in a pediatric emergency trauma population. Methods A retrospective observational study was conducted at the Pediatric Emergency Department of Hacettepe University, a Level I pediatric trauma center, between January 2019 and December 2022. Patients aged 0–18 years presenting within one hour of injury and observed for ≥ 8 h were enrolled. Demographic, clinical, laboratory, and injury scoring data were systematically collected. Multivariable logistic regression analysis was performed to identify independent predictors of LSI requirement. Results A total of 559 patients were included (median age 91.4 months [IQR 35.1–170.2]; 64.6% male), of whom 32.0% ( n = 179) required a life-saving intervention. Because several of the evaluated scores are structurally interdependent, a parsimonious multivariable model was constructed from one representative variable per score family (ISS, PTS, and SI). Higher ISS (adjusted OR 1.145, 95% CI 1.091–1.202) and lower PTS (adjusted OR 0.552, 95% CI 0.474–0.642) were independent predictors of LSI requirement (both p < 0.001), whereas the shock index was not ( p = 0.134). This combined model (ISS + PTS + SI) showed excellent discrimination (AUC 0.961, 95% CI 0.943–0.978) and significantly outperformed the best individual score, the PTS (AUC 0.947; DeLong p = 0.004). Among individual scores, discrimination was highest for the PTS, followed by the GCS (0.920), ISS (0.918), the BIG score (0.897), and AIS (0.878), with the SI (0.601) the lowest. After adjustment for injury severity and neurological status, base deficit—rather than lactate—remained an independent metabolic predictor. Conclusion In pediatric trauma patients, the PTS, GCS, and ISS were the strongest individual predictors of LSI requirement, and a parsimonious model integrating anatomical, integrated physiological, and hemodynamic domains significantly outperformed the best-performing individual score, the PTS (DeLong p = 0.004). In a separate adjusted analysis of the subgroup with complete laboratory data, base deficit—but not lactate—remained an independent metabolic predictor; these metabolic markers were not components of the combined ISS + PTS + SI model, in which the SI itself demonstrated limited discrimination. A multiparametric approach—integrating anatomical, physiological, and hemodynamic domains in the combined model, with metabolic markers evaluated in a separate analysis—may provide a more robust framework for early in-hospital risk stratification after initial diagnostic evaluation.