1328. Predicting the outcome of children with non-accidental traumatic brain injury: Utility of the GCS, GCS-P, and Rotterdam Score in children less than 3 years of age.
Award: Second Place Pediatrics Eposter Award
Authors: Edwin Kulubya, MD; Miriam Nuno, PhD; Marike Zwienenberg, MD; Laurie Ackerman, MD (Sacramento, CA)
Children with non-accidental trauma (NAT) are often reported to have worse outcomes than children with accidental injuries (AT). In many prior studies, patients with NAT have not been stratified by age and injury severity, which may impair our understanding of their functional outcome as compared to AT. In this study we seek to evaluate the discriminative accuracy of the Glasgow Coma Score (GSC), Glasgow Coma Pupil Score (GSC-P), and Rotterdam Score (RS) for survival and 6-month functional outcomes in age matched cohorts of children with NAT and AT.
Data of all patients with TBI (age < 3 years), collected at two institutions were merged for retrospective analysis. Patients were divided in two groups (NAT versus AT). Multiple predictors (GCS, GCS-P, RS) were correlated with the 6-month dichotomized Glasgow Outcome Score (DiGOS) and mortality. Univariate and bivariate descriptive analysis was utilized to summarize the cohort. Student T-tests and Chi-square were used compare NAT and AT cohorts. Multivariable logistic regression was used to develop prediction models. Estimated area under the receiver operating characteristic curve was used to assess performance of GCS and GCS-P.
1053 pediatric patients were analyzed: NAT 291 (27.6%), AT 762 (72.4%). NAT patients were younger (0.7 versus 1.3 years) and had more severe injuries (Mean GCS 11.8 versus 13.9). Functional outcomes were similar in NAT and AT after mild and moderate TBI, but significantly worse after severe TBI in NAT. AT was associated with higher odds of survival than NAT (OR 2.1, 95% CI: 0.04-0.32). GCS and GCS-P, but not RS had high discriminative accuracy in predicting the 6-month DiGOS.
Conclusion: The initial clinical severity of injury, not mechanism, remains the most important predictor of early outcome. Applying these prediction models may be useful for clinical decision-making and for managing expectations of patients and caregivers.