006. Predicting Outcomes in Pediatric Traumatic Brain Injury: Utility of IMPACT Prognostic Model and Glasgow Coma Score

Award: Top Poster

Top Poster

Authors: Edwin Kulubya, MD

Traumatic brain injury?remains?a leading cause of a death and disability in children.?Few models exist to prognosticate outcomes after injury, and most include only adult cohorts. We aim to evaluate the discriminative accuracy of the?IMPACT,?Glasgow Coma (GCS), and GCS-Pupils?Score (GCS-P) to predict mortality and functional outcome in children.Methods: We analyzed single institution data (2008-2020) of children (0-18 years) who suffered moderate-severe traumatic brain injury (GCS 3-13) and underwent Glasgow Outcome Score Extended (GOS-E) assessment after six months. Logistic?models were developed for?mortality and 6-month?dichotomized Glasgow Outcome Score (DiGOS), utilizing GCS, GCS-P and IMPACT?score variables (Core/Extended/Lab) upon admission. The predictive accuracy of each model was determined by the Area Under the Receiver Operating Curve (AUROC). Variables were added sequentially to the models to assess predictive benefit.
Of the 515 patients analyzed, median age was 8 years and 43% had severe TBI. Mortality and unfavorable outcome rates were 11% and 29%, respectively.? The models were better at predicting mortality than unfavorable outcome. The stand-alone GCS (motor score) had a high predictive accuracy for mortality (AUROC>0.92). The addition of CT findings alone did not improve accuracy, whereas the addition of pupil reactivity score, hypoxia, hypotension,?hemoglobin, and glucose?values improved accuracy?for mortality (>0.95). Models were less robust predicting unfavorable outcome (AUROC 0.86-0.88), with GCS, hypoxia and hypotension identified as the strongest predictors.
Initial clinical severity based on GCS remains a strong predictor of mortality. Certain IMPACT variables improved predictive accuracy of outcomes in this pediatric cohort indicating IMPACT score calculation may be useful in pediatric TBI. Defining new TBI characteristics and biomarkers have the potential to further enhance our predictive models. Ultimately, these models can help with clinical decision making and managing expectations of patients and caregivers.