Development and internal validation of a nomogram for predicting prolonged ICU length of stay after oral free flap reconstruction without prophylactic tracheostomy

To develop and validate a nomogram for predicting prolonged ICU length of stay (LOS) in patients undergoing oral free flap reconstruction without prophylactic tracheostomy. This single-center retrospective study included 419 consecutive patients treated between January 2019 and July 2025. Patients were randomly assigned to a training set ( n = 294) and a validation set ( n = 125) at a 7:3 ratio. LASSO regression was used for variable selection in the training set, followed by multivariable logistic regression to build the prediction model, which was visualized as a nomogram. Model performance was assessed using the area under the curve (AUC), calibration plots, and decision curve analysis (DCA). Seven predictive factors were ultimately included in the model: primary site (Location), flap source, smoking history, intraoperative blood loss (per 10 mL increment), postoperative blood glucose, postoperative procalcitonin (PCT), and the International Normalized Ratio (INR). The AUC was 0.722 (95% CI: 0.652–0.793) in the training set and 0.769 (95% CI: 0.667–0.872) in the validation set. The model demonstrated good calibration and significant net clinical benefit. The developed nomogram model shows good discrimination and calibration. Based on routine perioperative indicators, it can be used for rapid bedside identification of high-risk patients, providing a quantitative basis for ICU resource allocation and individualized enhanced recovery pathways.