Performance of the Marshall and Rotterdam scales as predictors of mortality in children with severe traumatic brain injury

Objective: Traumatic brain injury (TBI) is a public health problem and the Glasgow Coma Scale (GCS) is recognized as the gold standard tool for grading TBI severity, but it may underestimate severity when applied in the pediatric population. The combined use of the GCS with criteria established in brain CT scan images enhances the ability to predict prognosis in this group of patients. The Marshall and Rotterdam scales are the most commonly used for this purpose. This study aims to evaluate and compare the performances of these scales as predictors of in-hospital mortality in children with severe TBI. Methods: This was a retrospective study using data from consecutive medical records of children and adolescents (<17 years old) with severe TBI (GCS score <= 8) hospitalized between January 2016 and December 2023 at a Brazilian trauma reference center. Brain CT scan images were retrospectively analyzed for calculation of scores in the Marshall and Rotterdam scales. The cutoff score of 4 established as a predictor of prognosis was used to dichotomize the group. We calculated the performance and accuracy of both scales as predictors of prognosis. Results: The area under the receiver operating characteristic curve (AUC) of the 37 children with severe TBI included in the study was 0.65 in the Marshall scale and 0.73 in the Rotterdam scale, considering in-hospital mortality as the outcome. In both scales the mortality rate was higher in children with scores >= 4, being 42% (42% vs. 12%; p = 0.05; OR 5.23; 95% CI 0.9910-27.6872) in the Marshall scale, and 50% (50% vs. 11%; p = 0.01; OR 8; 95% CI 1.4247-44.9212) in the Rotterdam scale. Conclusion: The Rotterdam scale performed better as a predictor of in-hospital mortality (AUC 0.73) when compared to the Marshall scale (AUC 0.65) in the group of children with severe TBI analyzed. The cutoff point of 4 in the Marshall and Rotterdam scales can be used in clinical practice as a predictor of mortality with accuracy of 73% and 78% respectively.