Development and Validation of a Predictive Model for Difficult Tracheal Intubation in Thai Obese Patients

Abstract

The most concern of anesthetists in anesthesia for obese patients is airway management. Such bedside tests information is nowadays not capable of discriminate obese patients who have no outstanding features of problematic patients from non difficult conditions. Existing predictive models were developed for general, not specific to obese group. We conducted a multi-center to set up a practical new predictive model and to validate it using a separate set of patients. This observational study was conducted in 280 obese patients who were expected to not use alternative tools for first-line management in four hospitals. Difficult intubation was defined using an intubation difficulty scale (IDS) score > 5. Clinical assessment, including malformation of central teeth in the upper jaw, a modified Mallampati test, an upper-lip bite test, the range of motion of the neck (flexion and extension), the inter-incisor gap, the hyomental distance, the thyromental distance, the sternomental distance, the neck circumference, and the length of the neck, were examined in all patients. The inter-observer reliability of raters was > 0.7 before initiation of the study. Overall, only three patients experienced difficult intubation during conventional endotracheal intubation. Logistic regression model for troublesome intubation (IDS > 0) was then developed based on 200 patients from Siriraj Hospital. The simplified final model comprised only three independent variables. It revealed that patients with bigger neck circumference had a higher risk of troublesome intubation with adjusted OR of 1.15 for one centimeter increment in NC. Decreased in one centimeter of inter-incisor gap increased the risk of troublesome intubation to 1.59 (95% CI: 1.03, 2.50). Regarding modified Mallampati test, class II and III had a higher risk of troublesome intubation compared to class I (adjusted OR of 2.20 and 3.68 respectively). The final logistic regression fit the data quite well with p-value from Hosmer-Lemeshow test of 0.254. To validate the final logistic regression model, this model was applied to another set of 80 patients. Probability of troublesome intubation for each subject was then calculated. Cut point of 0.45 resulted in the sensitivity of 70.0%, specificity of 45.0% and accuracy of 57.5% respectively.