Abstract
BACKGROUND: Predicting difficult intubation (DI) is a key challenge, as no single clinical predictor is sufficiently valid to predict the outcome. We evaluated the effectiveness of four upper airway ultrasonographic parameters in predicting DI. The validity of the models using combinations of ultrasonography-based parameters was also investigated. METHODS: This prospective, observational, double-blinded cohort trial enrolled 1,043 surgical patients classified as American Society of Anesthesiologists physical status I-III without anticipated difficult airway. Preoperatively, their tongue thickness (TT), invisibility of hyoid bone (VH), and anterior neck soft tissue thickness from the skin to thyrohyoid membrane (ST) and hyoid bone (SH) were measured by sublingual and submandibular ultrasonography. The logistic regression, Youden index, and receiver operator characteristic analysis results were reported. RESULTS: Overall, 58 (5.6%) patients were classified as DI. The TT, SH, ST, and VH had accuracies of 78.4%, 85.0%, 84.7%, and 84.9%, respectively. The optimal values of TT, SH, and ST for predicting DI were > 5.8 cm (sensitivity: 84.5%, specificity: 78.1%, AUC: 0.880), > 1.4 cm (sensitivity: 81%, specificity: 85.2%, AUC: 0.898) and > 2.4 cm (sensitivity: 75.9%, specificity: 85.2%, AUC: 0.885) respectively. VH had a sensitivity and specificity of 72.4% and 85.6% (AUC: 0.790). The AUC values of the five models (with combinations of three or four parameters) ranged from 0.975-0.992. ST and VH had a significant impact on the individual models. CONCLUSIONS: SH had the best accuracy. Individual parameters showed limited validity. The model including all four parameters offered the best diagnostic value.