Related Experiment Video
Updated: Jun 4, 2026
Controlled Cortical Impact Model for Traumatic Brain Injury
Published on: August 5, 2014
Machine Learning Insights Into Social Determinants Driving Child Abuse in Pediatric Traumatic Brain Injury
Foad Kazemi1, Melanie Alfonzo Horowitz, David Lee
1Department of Neurosurgery, Johns Hopkins University School of Medicine, Baltimore , Maryland , USA.
Background And Objectives:
Social determinants of health (SDOH) are key drivers of health inequities, shaping disparities in patient outcomes that must be addressed. This study examines the association between SDOH and suspected child abuse (SCA) in pediatric patients sustaining traumatic brain injury (TBI), leveraging newly proposed Centers for Disease Control and Prevention (CDC)/PLACES measures to identify the most contributing measure to SCA.
Methods:
A retrospective review of our institutional database (2016-2023) identified pediatric TBI cases (18 years and younger) using International Classification of Diseases, 10th Revision codes based on a modified CDC framework. Cases of SCA investigated by a multidisciplinary child protection team were identified through electronic medical records review. Publicly available CDC/PLACES data were merged with the authors' data population to derive nine measures of SDOH. A composite measure of the Social Deprivation Index (SDI), with higher scores indicating worse neighborhood disadvantage, was obtained. Adjusted multivariate regression analysis examined correlation between SDOH measures and SCA. Extreme Gradient Boosting (XGBoost) machine learning and SHapley Additive exPlanations identified the key contributing factors to SCA.
Results:
This study included 2945 TBI patients with a median (IQR) age of 6 (1-12) (36.6% female, 40.3% White, 7.3% Hispanic). Among the overall cohort of TBI cases, 13.3% were SCA. In multivariate regression models adjusted for age, sex, race, and ethnicity, 7 out of 9 SDOH measures were independently and significantly associated with SCA (all P -values <.01), and the SDI was also significantly correlated with SCA ( P < .001). SHapley Additive exPlanations analysis identified "Unemployment among people 16 years and older in the labor force" as the most influential factor contributing to SCA, followed by "Crowding among housing units" and "Persons of racial or ethnic minority status," outperforming the SDI score.
Conclusion:
SDOH, particularly unemployment, crowding, and persons of racial or ethnic minority status are strongly associated with SCA in pediatric TBI patients, highlighting the need for targeted interventions to mitigate these disparities.
