Abstract
Objective:
To develop and internally validate a clinically interpretable prediction model for 30-day postoperative infectious complications after emergency craniofacial trauma surgery.
Methods:
This retrospective cohort-style prediction model study included 720 emergency craniofacial trauma surgery records. Patients were randomly assigned to a training set (n=504) and an internal validation set (n=216). Candidate predictors included host factors, injury severity, fracture and wound characteristics, airway intervention, antibiotic timing, operative duration, transfusion, and adequacy of initial debridement. The primary outcome was any 30-day infectious complication, including surgical site infection, fixation-related infection, pneumonia, deep orbital infection, or intracranial infection. Model performance was assessed with area under the receiver operating characteristic curve (AUC), calibration, Brier score, risk stratification, and decision curve analysis.
Results:
The cohort included 720 patients, of whom 103 (14.3%) developed infectious complications. Important predictors included open fracture, oral wound communication, moderate or severe contamination, higher Injury Severity Score, longer operative duration, current smoking, delayed antibiotic administration, transfusion, airway intervention, and inadequate debridement. The model showed an AUC of 0.836 (0.783–0.888) in the training set and 0.738 (0.623–0.833) in the validation set. The validation Brier score was 0.093. Infection rates in low-, medium-, and high-risk strata were 5.6%, 9.7%, and 22.2%, respectively.
Conclusions:
A model based on routinely available emergency and perioperative variables demonstrated acceptable internal validity and clinical interpretability. The findings support risk-stratified infection surveillance and provide a structured framework for subsequent external validation.