Enhancing Pediatric Trauma Survival Prediction: Integrating ICISS and Glasgow Coma Scale for Greater Accuracy.

Abstract

The mortality rate from injuries globally has declined but remains high in the Americas. Despite the proven efficacy of severity scores in improving injury outcomes, their use in Latin American children has been limited. We aimed to assess the predictive performance of the International Classification of Disease-based Injury Severity Score (ICISS) in children's survival and compare it with a novel predictive model. A retrospective cohort study was conducted at a trauma center in Cali, Colombia, including children (< 18 years) with trauma-related diagnoses from 2011 to 2019, utilizing electronic health records from the Colombian national health system. A logistic regression model was developed to predict 30-day survival. Its performance was assessed by discrimination with the area under the receiver operating characteristic curve (AUROC) and calibration and was statistically compared with a new model incorporating the Glasgow Coma Score (GCS). The new model was internally validated by bootstrap resampling. The study included 1047 children. The new model demonstrated superior discrimination (AUROC: 0.98, 95% CI: 0.97-0.99) compared to the ICISS-only model (AUROC: 0.94, 95% CI: 0.92-0.97), with a statistically significant difference (p = 0.01). Additionally, it exhibited better calibration (Hosmer-Lemeshow test: 6.7, p = 0.6 vs. 24.7, p < 0.01), indicating improved predictive accuracy. Following internal validation, the new model maintained excellent performance across all measures. Combining ICISS with the GCS significantly improved the accuracy of survival prediction in children.