Answers are grounded in this paper and cite their sources.
Background Difficult airway is an important cause of serious perioperative adverse events. Conventional preoperative assessment is simple but partly operator dependent and has limited predictive accuracy. We developed machine learning models combining clinical measurements and airway ultrasound parameters to support preoperative difficult airway assessment. Methods Seven predictors were collected from 1,000 adults undergoing elective surgery under general anesthesia: body mass index, neck circumference, mouth opening, sternomental distance, skin-to-epiglottis distance, tongue thickness, and hyomental distance ratio. Data were split into training (n = 800) and test (n = 200) sets. Logistic Regression, Random Forest, Extreme Gradient Boosting (XGBoost), generalized additive model (GAM), and Naive Bayes were tuned by stratified 5-fold cross-validation using ROC-AUC. The principal metrics were the area under the receiver operating characteristic curve (ROC-AUC), average precision (AP), accuracy, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), balanced accuracy, and F1 score. Results Mean cross-validation ROC-AUC values were 0.818 for Random Forest, 0.797 for XGBoost, 0.790 for GAM, 0.786 for Logistic Regression, and 0.770 for Naive Bayes. The test set included 24 difficult airway cases. Test-set ROC-AUC values were 0.802 for GAM, 0.783 for XGBoost, 0.777 for Random Forest, and 0.771 for Logistic Regression. At the 0.5 threshold, Logistic Regression had the highest sensitivity (0.542) and a negative predictive value of 0.930; XGBoost had sensitivity 0.458, specificity 0.892, and F1 score 0.407. Conclusions Models combining clinical and airway ultrasound measurements showed some discrimination for difficult airway prediction. Given its sensitivity, negative predictive value, and interpretability, Logistic Regression is a candidate for calibration, threshold optimization, and external validation.
No takes yet. Share an insight, caveat, or question.
Sun et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: