RESUMO
BACKGROUND AND AIMS: Various models were devised for prediction of difficult intubation but have low positive predictive value, sensitivity and specificity. We aimed to predict difficult intubation from various airway predictive indices, in isolation and combination, and to formulate a multivariate model that can aid in accurate prediction of difficult intubation. MATERIAL AND METHODS: A prospective double blinded study was conducted on 500 adult patients scheduled for elective surgery under general anaesthesia. Preoperatively, they were assessed for airway screening tests. After standardized induction of anaesthesia, laryngoscopic view was classified according to the Modified Cormack and Lehane (MCL) classification. Variables' association with intubation findings was evaluated using Chi-square statistic. Stepwise logistic regression identified the multivariate independent predictors of difficult intubation and combinations were made using forward selection process. 8 models were formulated and a receiver-operating characteristic (ROC) curve worked out for them. Sensitivity and specificity analysis validated the final model. RESULTS: Age, sex, weight, BMI, snoring, obstructive sleep apnea (OSA), diabetes, hypertension, upper lip bite test (ULBT), Mallampati grade (MPS), thyromental distance (TMD), sternomental distance (SMD), neck movements (NM), neck circumference (NC) and inter-incisor gap (IIG) had significant correlation with difficult intubation. Based upon sensitivity and specificity analysis, model comprising of MPS, NM, NC and SMD was found to be most accurate. It had highest sensitivity 80%, specificity 87% and area under curve 0.90, thus validating the model. CONCLUSIONS: Our study found that a combination of MPS, SMD, NM and NC permits reliable, accurate and quick preoperative prediction of difficult intubation.