Evaluation of Concomitant Facial Fracture in Traumatic Brain Injury Patients—Simplification and External Validation of a Prediction Model

Abstract

Background Patients with traumatic brain injuries (TBIs) often experience concurrent facial bone fractures. In 2021, a prediction model with 10 variables was published and precisely predicted concomitant facial fractures in TBI patients. Herein, external validation and simplification of this model was performed. Methods Traumatic brain injury patients treated at a major referral trauma center were retrospectively reviewed for 1 year. The original prediction model (published in 2021), which was developed from a rural level II trauma center, was applied for external validation. A new and simplified model from our level I trauma center was developed and backwardly validated by rural level II trauma center data. Results In total, 313 TBI patients were enrolled; 101 (32.3%) had concomitant facial fractures. When the previous prediction model was applied to the validation cohort, it achieved acceptable discrimination, with an area under the receiver operating characteristic curve (AUC) of 0.713 and good precision, with a Brier score of 0.083. A new and simplified model with 6 variables (age, tooth rupture, epistaxis, facial lesion, eye injury, and intracranial hemorrhage) was created with excellent discrimination (AUC = 0.836) and good precision (Brier score of 0.055). The backward validation of this new model also showed excellent discrimination in the cohort used to develop the original model (AUC = 0.875). Conclusion The original model provides an acceptable and reproducible prediction of concomitant facial fractures among TBI patients. A simplified model with fewer variables and the same accuracy could be applied in the emergency department and at higher- and lower-level trauma centers.