Comparative performance of CT scales for outcome prediction at discharge in pediatric traumatic brain injury: a retrospective cohort study

OBJECTIVE: The aim of this study was to compare the performance of four CT-based classification systems-Marshall, Rotterdam, Helsinki, and Stockholm-in predicting outcomes at discharge in pediatric patients with traumatic brain injury (TBI). METHODS: This is a retrospective cohort study of patients in a pediatric intensive care unit of a tertiary referral hospital. Children admitted with TBI were included. Patients with incomplete clinical or imaging records were excluded. Collected data included baseline clinical characteristics, CT scores, discharge destinations, and outcomes (measured using the extended Glasgow Outcome Scale Pediatric Revision [GOS-E Peds]). Univariable and multivariable penalized logistic regression models were fitted to assess the associations between CT scales and an unfavorable outcome, defined as a GOS-E Peds score ≤ 6 at discharge. Interactions between CT scales and age groups were explored. Discriminative performance was evaluated using the area under the receiver operating characteristic curve (AUROC) and classification metrics. RESULTS: Of the 295 patients analyzed, 73.9% were male, and the most frequent mechanisms of injury were vehicle related (47.1%) and falls (37.9%). Most patients were classified in the lower-risk categories across the CT scales, and 84.4% attained upper good recovery at discharge. In univariable analyses, all four CT scales were significantly associated with outcome. In the multivariable model including all scales simultaneously, all scales except the Rotterdam classification remained independent predictors. The Stockholm score demonstrated the highest discriminative ability (AUROC 0.889), followed by the Marshall (AUROC 0.829) and Helsinki (AUROC 0.761) classifications, whereas the Rotterdam score showed lower performance (AUROC 0.685). CONCLUSIONS: In pediatric TBI, the Marshall, Helsinki, and Stockholm CT scoring systems independently predict outcome at discharge. Among these classifications, the Stockholm score demonstrated the strongest discriminative ability, suggesting that weighted, multidimensional CT classifications may provide superior prognostic information in children. Multicenter validation and long-term follow-up are needed to confirm these findings and refine pediatric prognostic tools.