Abstract INTRODUCTION There is paucity of research addressing how health care decisions are made in cases of severe traumatic brain injury. This study aims to identify the demographic and clinical factors associated with withdrawal care in the setting of severe traumatic brain injury. METHODS This is a retrospective study using the Trauma Quality Improvement Program database from 2013 to 2015. Patients with severe traumatic brain injury (GCS 3-8, AIS 2-5) were included. Univariate and multivariate analyses with adjusted P-values were performed for descriptive statistics. A logistic regression analysis was used to identify demographic and clinical factors associated with withdrawal of care. RESULTS The analysis included 7869 (21%) patients who withdrew and 30 080 (79%) who did not. More than half of those withdrawn were at a University hospital. Patients undergoing withdrawal of care were older (57.6 ± 20.6 vs 42.9 ± 18.8 yr) and 69% were male, 82% were white, 43% were from the Southern United States and 31% had private insurance. Black or other races were less likely to undergo withdrawal of care as compared to white (odds ratio OR 0.7, 95% confidence interval CI 0.6-0.7 and 0.8, 95% CI 0.8-0.9, respectively). Injury severity score (ISS) was significantly different between those who underwent withdrawal and those who did not, 28.2 ± 12.3 vs 26.1 ± 12.1. The presence of epidural or subdural hematoma was also significantly different between the 2 groups (78% vs 72%). Those who did not have a craniotomy were less likely to undergo withdrawal and those who did not have a hematoma were less likely to undergo withdrawal (OR 0.7, 95% CI 0.7-0.8 and 0.8, 95% CI 0.7-0.9, respectively). The average ICU length of stay and ventilation days were shorter for patients who underwent withdrawal. CONCLUSION Age, ISS, craniotomy, race (white), and type of insurance (Medicare) were each positive predictors for withdrawal of care. By contrast, region (Southern United States), race (black), and being functionally independent were negative predictors.