Study design
This was a diagnostic, retrospective cohort study using an exploratory model. It was conducted at the ED of Ramathibodi Hospital, a university-affiliated super-tertiary care hospital in Bangkok, Thailand. The data were collected from the Ramathibodi Hospital database and emergency medical record system from June 2018 to July 2020.
Sample size
The data were collected from June 2018 to July 2020 and analyzed by STATA version 16.0 (StataCorp, College Station, TX, USA). A pilot study was performed to calculate the sample size by employing a two-sample comparison of difficult intubation and non-difficult intubation. Using the formula N = Zα / 22p(1 − p) / d2, standard normal variate (Zα/2) at 1%, probability of the expected value (p) = 0.12, and a two-sided test, the minimum sample size was determined to be 281 patients.
Participants
Patients were required to be >15 years of age and to have undergone emergency intubation in the ED. We excluded patients who had been intubated by a non-EP (general practitioner) and those who had developed cardiac arrest. From the minimum sample, 281 patients who underwent emergency intubation in the ED were enrolled in this study. The study period was 2 years (from June 2018 to July 2020).
Definition of difficult intubation (reference standard)
The definition of difficult intubation was the same as that of difficult laryngoscopy, according to the structures that can be visualized and identified by laryngoscopy. Using the four-grade classification of the laryngoscopic view defined by Cormack and Lehane,22 intubation was defined as easy (grade I or II) or difficult (grade III or IV). The laryngoscopic view was assessed and graded according to the final intubation after each intubation was finished.
Definitions of predictors
Intubation method
An “intubation method” was defined as one set of medication or devices, such as rapid-sequence intubation with direct laryngoscopy.
Intubation attempt
An “intubation attempt” was defined as one effort to place an airway. Each attempt could be performed using one or more methods, and each method could have one or more attempts. After each intubation was finished, the clinician entered all data in the medical record form.
Operator level
We classified the operator’s level of training into three groups: low experience (general practitioners), moderate experience (first-year residents in emergency medicine and first-year residents in general medicine), and high experience (second- to third-year residents in emergency medicine and emergency attending staff).
Failed intubation
We defined “failed intubation” as multiple efforts to place an airway (more than one effort).
Indicators of difficult intubation
The difficult intubation assessment tool “LEMON” and difficult ventilation assessment tool “MOANS” (Mask seal, Obesity, Age (elderly), No teeth, Stiffness) were used to evaluate the patients undergoing endotracheal intubation.8,18 For the “Look externally” criterion of “LEMON,” we assessed any significant facial injury, large incisors, significant beard or mustache, and large tongue. For the “Evaluate the 3-3-2 rule” criterion, we assessed a mouth opening of less than three finger breadths, a hypomental distance of less than three finger breadths, and a thyrohyoid distance of less than two finger breadths. For the “Obstruction” criterion, we record all conditions that may make the laryngoscopic view more difficult, such as a mass, hematoma, or massive bleeding. For the “Neck mobility” criterion, we assessed limited neck mobility, such as collar immobilization. With respect to the “Mallampati score” criterion of the “LEMON” assessment tool, however, we did not record the Mallampati score because it requires patient cooperation, which is limited in critically ill patients in the ED. Indicators of difficult intubation using the “MOANS” assessment tool were also noted, but we did not consider them as indicators of difficult intubation in this study.
Data collection and study variables
We collected data regarding the patients’ characteristics, including sex, age, Glasgow coma scale score, main indication for intubation, initial method of intubation, dosage of any medications used, operator level of training, number of attempts, success or failure, and structures identified by direct and indirect laryngoscopy. The patients were divided into two groups: the difficult intubation group and the non-difficult intubation group. We then used a multivariable regression model to identify significant predictors of difficult intubation. The data were collected from June 2018 to July 2020, and after excluding patients who did not meet the study criteria, the total sample size was 617 patients. The protocol is illustrated in Figure 1.
Outcome of interest
The outcome of interest was a positive difficult anatomical laryngoscopic view according to the final intubation after each set of intubation attempts was finished. Patients in the difficult intubation group had a Cormack–Lehane grade III or IV laryngoscopic view, and those in the non-difficult intubation group had a Cormack–Lehane grade I or II laryngoscopic view. Finally, we used the patients’ data to develop a risk score with which to predict difficult intubation in the ED.
Statistical analysis
The data were analyzed using STATA version 16.0 (StataCorp). All variables were compared between the difficult intubation and non-difficult intubation groups by descriptive statistics. The predictive power of each variable for a positive difficult laryngoscopic view was calculated using univariable logistic regression and is presented as the area under the receiver operating characteristic curve (AuROC) with the 95% confidence interval (CI). A multivariable stepwise backward logistic regression model was used to develop the predictive model. Predictors with a cut-off P-value of 0.05 after the univariable analysis were included in the model and eliminated with a significance level of 0.001. Regression coefficients for each level of each clinical predictor were divided by the smallest coefficient of the model and rounded to the nearest 0.5, resulting in an item risk score. According to this score, the coefficients were then changed into item scores and added together into a single score, and the patients were thus classified into low-, moderate-, and high-probability categories. Discrimination of the airway assessment scores is presented as the AuROC and 95% CI for the clinical risk score of difficult intubations. Calibration of the prediction was performed using the Hosmer–Lemeshow goodness-of-fit test. The score-predicted risk of difficult intubation and the observed risk were then compared in a graph. The number of reports and percentages of each group are presented with the positive likelihood ratio, 95% CIs, and P-values.