Multivariable Analysis-Based Risk Prediction Model for Intracranial Hematoma Expansion in Traumatic Brain Injury...

To develop and validate a clinical prediction model for hematoma expansion (HE) in traumatic brain contusion (TBC) patients, providing a quantitative tool for early identification of high-risk patients.

A single-center retrospective cohort study was conducted, collecting clinical data from 263 TBC patients admitted to the Department of Neurosurgery at Lianyungang First People's Hospital between July 2022 and December 2024. Patients were randomly divided into training (n = 184) and validation (n = 79) cohorts at a 7:3 ratio, with an additional 88 patients serving as a supplementary validation cohort. Demographic characteristics, clinical presentations, laboratory parameters, and imaging features were collected. HE was defined as >33% increase in hematoma volume or an absolute increase of >6 mL on follow-up cranial computed tomography within 24 hours postinjury, or the emergence of new hemorrhagic lesions. Univariate and multivariate logistic regression analyses were performed to identify independent predictors. Model performance was evaluated using receiver operating characteristic curves, calibration plots, and decision curve analysis.

Among 263 TBC patients, HE occurred in 60% (157/263). Univariate analysis identified 9 factors significantly associated with HE. Multivariate analysis determined five independent predictors: subdural hematoma (odds ratio [OR] = 7.71, 95% confidence interval [CI]: 4.06-14.65), fibrin degradation products >30 mg/L (OR = 3.46, 95% CI: 1.80-6.63), subarachnoid hemorrhage (OR = 3.04, 95% CI: 1.50-6.17), frontal lobe injury (OR = 2.52, 95% CI: 1.30-4.89), and Glasgow Coma Scale <13 (OR = 2.50, 95% CI: 1.11-5.60) (all P < 0.05). The prediction model achieved area under the curves of 0.937 (95% CI: 0.95-0.99) and 0.888 (95% CI: 0.82-0.95) in training and validation cohorts, respectively. Hosmer-Lemeshow tests demonstrated good calibration (training cohort P = 0.702, validation cohort P = 0.944). Decision curve analysis confirmed favorable clinical net benefit. Risk stratification classified patients into low-risk (0-2 points), intermediate-risk (3-4 points), and high-risk (5-6 points) groups, with HE rates of 11%, 48%, and 87%, respectively, in the supplementary validation cohort (P < 0.001).

The developed prediction model for HE in TBC patients demonstrates excellent discrimination and calibration, providing quantitative evidence for individualized monitoring and treatment strategies that may improve patient outcomes. This scoring system is simple, practical, and holds promising clinical application potential.