Preoperative Risk Stratification for Major Perioperative Airway Events in Adult Tonsillectomy: A Retrospective Cohort Study.

IntroductionAlthough adult tonsillectomy is common, it has a relatively high risk of perioperative airway adverse events. To identify and validate preoperative predictors of major airway adverse events in tonsillectomy and evaluate a predictive model.MethodsThis single-center retrospective study included 218 adults undergoing tonsillectomy (Jan 2023 to Sep 2025). Patients were grouped by the occurrence of a composite airway event, which included difficult intubation, severe hypoxemia, postoperative airway intervention, or unplanned ICU admission.Univariate and multivariate logistic regression identified independent predictors. A nomogram-based model was developed. Data were split into training and validation sets. Performance was assessed using ROC curves, calibration plots, bootstrap resampling, and decision curve analysis (DCA).ResultsMultivariate analysis identified body weight (OR=1.11), modified Mallampati classification (OR=2.55), and tonsil size grade (OR=3.74) as independent predictors. The model showed excellent discrimination with AUCs of 0.904 (training) and 0.924 (validation). At a cutoff of 0.244, the model showed 83.8% sensitivity and 84.8% specificity in the training set, and 88.9% sensitivity and 91.7% specificity in the validation set, with negative predictive values of 94.1% and 98.2%, respectively. High-risk patients had a 7.09-fold increased risk (95% CI: 2.35-27.04, P<0.001). The model was well-calibrated, and DCA indicated a net clinical benefit for threshold probabilities between 0.05 and 0.50.ConclusionsAA model incorporating body weight, modified Mallampati classification, and tonsil size demonstrated good discriminative performance and may serve as a useful preoperative screening tool for perioperative airway risk stratification in adults undergoing tonsillectomy. Its high negative predictive value may be particularly useful for identifying patients at relatively low risk, while high-risk predictions should be interpreted in conjunction with clinical assessment.