Developing a predictive nomogram model for incidence of adverse outcomes in oral and maxillofacial tumor patients undergoing free flap reconstruction.

[BACKGROUND] Surgical treatment of oral and maxillofacial tumors is complex, and the factors contributing to adverse outcomes remain incompletely understood. This study aimed to identify independent risk factors and develop an efficient nomogram for predicting such outcomes.

[METHODS] Potential risk factors were identified using univariate logistic regression, followed by multivariate logistic regression to determine independent predictors. Three predictive models were constructed using different variable selection criteria. The Akaike Information Criterion (AIC) was used to select the optimal model. A nomogram was built based on the final selected model and validated via Receiver Operating Characteristic (ROC) curve analysis, calibration curves, and decision curve analysis (DCA).

[RESULTS] Among 915 patients, 117 (12.8%) experienced adverse outcomes. Six independent risk factors were identified: age, male sex, history of cerebral infarction, presence of comorbidities, surgery starting between 0:00 and 8:00, and elevated preoperative blood urea nitrogen (BUN) levels. A stepwise model including five of these factors yielded the lowest AIC (634.69) and was selected as the final model. This model demonstrated robust performance with an AUC of 0.744 (95% CI: 0.699-0.789), a Brier score of 0.103, and a positive net benefit across threshold probabilities from 0.00 to 0.49.

[CONCLUSION] We developed and internally validated a nomogram to predict adverse outcomes following oral and maxillofacial tumor resection with free flap reconstruction. This tool may assist clinicians in preoperative risk stratification and postoperative management.