Related Experiment Video
Updated: Oct 19, 2025
Development of an Uncomplicated Mild Traumatic Brain Injury Model Modified by Weight-Drop Method and Evidenced by Magnetic Resonance Imaging
Published on: April 11, 2025
A Prediction Model for Selective Use of Facial Computed Tomography in Blunt Head Trauma Patients
Li-Kuo Huang1, Chih-Sheng Huang1, Hsi-Feng Tu1
1From the Departments of Radiology, Surgery, and Dentistry, National Yang-Ming University Hospital; School of Medicine and Department of Dentistry, Dental School, National Yang-Ming University; and Department of Trauma and Emergency Surgery, Chang Gung Memorial Hospital, Chang Gung University.
Background:
Head trauma patients may have concomitant facial fractures, which are usually underdetected by head computed tomography alone. This study aimed to identify the clinical indicators of facial fractures and to develop a risk-prediction model to guide the discriminative use of additional facial computed tomography in head trauma.
Methods:
The authors retrospectively reviewed head trauma patients undergoing simultaneous head and facial computed tomography at a Level II trauma center from 2015 to 2018. Multivariate logistic regression analysis was used to evaluate independent risk factors for concomitant facial fractures in head trauma patients using data collected from 2015 to 2017, and a risk-prediction model was created accordingly. Model performance was validated with data from 2018.
Results:
In total, 5045 blunt head trauma patients (development cohort, 3534 patients, 2015 to 2017; validation cohort, 1511 patients, 2018) were enrolled. Concomitant facial fractures occurred in 723 head trauma patients (14.3 percent). Ten clinical and head computed tomographic variables were identified as predictors, including age, male sex, falls from elevation, motorcycle collisions, Glasgow Coma Scale scores less than 14, epistaxis, tooth rupture, facial lesions, intracranial hemorrhage, and skull fracture. In the development cohort, the model showed good discrimination (area under the receiver operating characteristic curve = 0.891), calibration (Hosmer-Lemeshow C test, p = 0.691), and precision (Brier score = 0.066). In the validation cohort, the model demonstrated excellent discrimination (area under the receiver operating characteristic curve = 0.907), good calibration (Hosmer-Lemeshow C test, p = 0.652), and good precision (Brier score = 0.083). With this model, 77.1 percent of unnecessary facial computed tomography could be avoided.
Conclusion:
This model could guide the discriminative use of additional facial computed tomography to detect concomitant facial fractures in blunt head trauma.
Clinical Question/Level Of Evidence:
Risk, III.
