Difficult tracheal intubation remains a relatively constant and significant source of morbidity and mortality in anaesthetic practice [1,2]. Airway management must still remain a matter of great concern for the anaesthesiologist .
During the preanaesthesia visit, anaesthesiologists have to estimate the risk of difficult intubation, to finally anticipate a difficult airway management strategy including ‘alternative’ airway devices [4,5]. Preoperative detection of patients at risk for difficult intubation is, therefore, the first step in airway management. Many risk factors have already been identified, some models validated [6–9], and associated guidelines have been established [4,5]. Despite these recommendations, insufficient or lack of airway assessment in the preoperative period continues to be a major cause of unanticipated difficult intubation . Some explanations can be proposed preoperative airway evaluation is not always regarded as standard procedure [10,11], and the current predictive index is probably too complex to implement because the risk index is impossible to calculate without a pocket calculator [8,9]. Practice guidelines for management of the difficult airway reported by the American Society of Anaesthesiologists advise that ‘multiple airway features should be assessed’ . Nevertheless, the respective weight of each risk factor is unclear and has been rarely studied . So, even keeping to guideline recommendations, it is not always easy for a practitioner, not expert on airway management, to anticipate judiciously a strategy, namely: when must the practitioner go into the operating theatre with, or without, a portable storage unit for difficult airway management?
The present study, therefore, aimed to identify adjusted risk factors for difficult intubation in patients with a difficult airway, with the objective in mind to build a ‘weighted score’ according to the value of regression coefficients of these adjusted risk factors predicting difficult intubation.
Materials and methods
As this prospective observational study did not alter routine care, and according to our institutional ethics committee norms, approval was given for this study and signed informed consent by the patients was not required. However, during the preanaesthesia visit, patients were orally advised that they were participating in an observational study without change of routine care, with the right to refuse participation in this observational study, and that their personal information would be treated anonymously.
Eligible patients included those adult patients scheduled for orthopaedic, vascular, urological, abdominal, gynaecological, ophthalmic, or ear, nose or throat (ENT) surgery under general anaesthesia. Six senior anaesthesiologists were involved in this study, each committed to recruit eligible patients and fill out a specific evaluation form. We excluded all patients undergoing emergency surgery, or those scheduled for awake fibreoptic tracheal intubation (unstable cervical spine, previous impossible intubation) or airway control with a laryngeal mask.
We recorded both surgical procedure and patient characteristics [weight, height, BMI (calculated as weight in kilograms divided by the square of the height in metres), age and sex]. An airway examination was performed in every patient during the preoperative visit. Clinical predictive factors that have previously been reported as risk factors for difficult intubation were recorded (Table 1). All airway tests were carried out with the patient in the sitting position. Factors associated with difficult mask ventilation as previously described (age >55 years, BMI more than 26 kg m−2, lack of teeth, history of snoring, and presence of a beard) were also recorded [5,15].
In the operating theatre, the choice of the head position (simple head extension or sniffing position), the anaesthetic induction drugs and techniques including Macintosh laryngoscope blade size were left to the discretion of the attending anaesthesiologist (straight blades were not used during the first attempt). The use of a neuromuscular blocking agent was also recorded. Tracheal intubation was performed by a senior operator (a certified anaesthesiologist, nurse anaesthetist or resident anaesthetist with more than 2 years' experience), or by a junior operator (student nurse anaesthetist, or resident anaesthetist with less than 2 years' experience) with the anaesthesiologist in attendance. The operator's status was also recorded.
Assessment of a difficult airway
All data were collected by the attending anaesthesiologist on a standard form after anaesthesia induction. The operator was asked to rate difficulty of mask ventilation as defined previously . Intubation difficulty was also assessed, at the end of the procedure, according to the Intubation Difficulty Scale (IDS) (see the table below for details) . Difficult intubation was defined as IDS more than 5 .
Intubation Difficulty Scale 
The primary endpoint of the present study was to identify risk factors of difficult intubation using bivariate analysis. The potential predictive risk factors were classified into preoperative and intraoperative risk factors (operator's status, use of neuromuscular blocking agent, difficult mask ventilation). Then, we determined adjusted risk factors for difficult tracheal intubation using multivariable analysis with stepwise logistic regression, including only preoperative risk factors. The P value for the selection of variables for the model was 0.15. We selected the adjusted risk factors of difficult intubation with a stepwise selection procedure, only allowing the introduction of additional variables if P remained less than 0.05. Regression coefficients of each adjusted risk factor in the final analysis represent their respective weight to predict difficult intubation. These values were then used to build a ‘weighted score’. As the values of these regression coefficients were unsuitable for clinical purposes, we assumed a ‘simplified score’ by multiplying regression coefficients by nine and rounding the latter to the closest half-ten. We named this new ‘simplified score’ the Simplified Predictive Intubation Difficulty Score (SPIDS). The respective performance of the SPIDS and the ‘weighted score’ were assessed by comparing their areas under the receiver operating characteristic curve (AUC) .
Therefore, the optimal predictive level of the SPIDS was determined using ROC curve analysis. The threshold was chosen to optimize sensitivity and specificity, defined as the point on the curve nearest the ideal point (sensitivity = specificity = 100%; i.e. the left superior angle) maximizing the curve area.
Bootstrapping technique was then used to obtain an estimate of the bias in the predictive accuracy of the SPIDS . The bootstrapping technique is a general tool for assessing statistical distribution of some parameters. It is a method of randomly resampling from a given experimental sample to simulate the effect of drawing multiple samples from the same population. A total of 1000 samples were drawn at random with replacement. On each bootstrap sample, the area under the ROC curve, sensitivity and specificity for the chosen threshold were calculated. With these 1000 simulations, average values and exact 95% confidence interval (CI) were estimated for the area under the ROC curve, sensitivity and specificity.
Moreover, because using the same dataset leads to overly optimistic values, we used cross-validation to obtain more realistic estimates. We randomly split the data into 10 equal-sized parts, we fitted the model to nine parts (calibration set) and calculated sensitivity and specificity of the fitted model on the last part (validation set). We restarted this procedure 1000 times, and we evaluated the model by the mean sensitivity and specificity. Thus, we obtained a prediction of the real performance of the model.
Finally, to evaluate the clinical relevance of the SPIDS, we compared its accuracy: first, with the corresponding ‘nonweighted score’ obtained by pooling together the predictive factors identified in the multivariable analysis, without taking into account the values of their regression coefficients (addition of the presence or the absence of each of the components of the SPIDS); and second, with a reference method. As a reference method, we decided to use the risk index reported by Arné et al. showing high sensitivity [93% (95% CI 80–99)] and specificity values [93% (95% CI 91–95)]. For this purpose, we compared their AUC .
Statistical analysis was performed using Statistical Analysis Software (version 8.2; SAS Institute Inc., Cary, North Carolina, USA) and S-PLUS 6 for Windows (2001; Insightful Corporation; Seattle, Washington, USA). Data are expressed as mean ± SD or as percentages.
We included 1024 consecutive patients in this study during a 24-month period. Patients' characteristics and type of surgical procedure are reported in Table 2. An IDS more than 5 was reported in 61 patients (6%). Neuromuscular blocking agents were used to perform tracheal intubation in 353 patients (34% of cases), and no tracheal intubation failure was observed.
Bivariate and multivariable analysis
The risk factors for IDS more than 5 identified from the bivariate analysis are listed in Table 3. Multivariable analysis by stepwise logistic regression identified five adjusted risk factors associated with IDS more than 5 and are detailed in Table 4. Points for the ‘weighted score’ and the SPIDS are expressed in Table 5.
Receiver operating characteristic analysis, optimal predictive level and internal validation
ROC analysis of the ‘weighted score’ and the SPIDS provided a comparable discriminating power to predict difficult intubation: AUCs were, respectively, 0.78 (95% CI 0.73–0.84) and 0.78 (95% CI 0.72–0.84) (P = 0.52). The threshold for an optimal predictive level of the SPIDS was above 10. The sensitivity and specificity of the SPIDS above 10 for predicting difficult intubation were, respectively, 65% (95% CI 55–85) and 76% (95% CI 55–83) (assessment of these CIs by bootstrapping ). We found sensitivity and specificity predictions of 64% and 76%, respectively, given by the cross-validation method. PPV and NPV of the SPIDS were, respectively, 14% and 97%.
Clinical relevance of the Simplified Predictive Intubation Difficulty Score
The predictive accuracy of the SPIDS was statistically better than the ‘nonweighted score’, with an AUC of 0.78 (95% CI 0.72–0.84) and 0.69 (95% CI 0.64–0.73) (P < 0.001), respectively (Fig. 1). The AUC of the risk index validated by Arné et al. was 0.76 (95% CI 0.70–0.82) (Fig. 1). ROC analysis of the SPIDS and the risk index validated by Arné et al. provided a comparable discriminating power for predicting difficult intubation (P = 0.19).
This study with a 6% incidence of difficult intubation in a general surgical population is in agreement with previous studies [20,21]. We identified only five adjusted risk factors, associated with an IDS more than 5 [pathological conditions associated with difficult intubation (Table 1), mouth opening <3.5 cm, head and neck movement <80°, the ratio of the patient's height to the thyromental distance (RHTMD) ≥25, and Mallampati's modified test ≥2] (Table 5). When considering the new ‘simplified score’, SPIDS, built with the values of regression coefficients of these five adjusted risk factors, the cut-off value of SPIDS strictly above 10 produced the best compromise between sensitivity [65% (95% CI 55–85)] and specificity [76% (95% CI 55–83)]. Finally, the predictive accuracy of the SPIDS was comparable to those obtained with the chosen reference method  and statistically better than the corresponding ‘nonweighted score’.
The first step required when developing a predictive test is to define clearly the outcome . Most screening test or other models previously described to predict difficult intubation used difficult laryngoscopy (Cormack–Lehane grade >2) as the definition of difficult intubation, despite the fact that it was never designed for this purpose [20,23]. Moreover, the relationship between the grade of direct laryngoscopy and difficult intubation has been questioned [24,25]. More than two laryngoscopy attempts, and/or the use of an alternative device or technique, following failed intubation with direct laryngoscopy are other definitions of difficult intubation in guidelines [5,26]. These various definitions did not mention other difficulties observed when performing a tracheal intubation such as the necessity of changing the operator, requirement of any unusual force applied during laryngoscopy, the necessity to perform pressure on the larynx allowing better glottic exposure, and the type or the number of alternative techniques used. Therefore, the IDS was proposed in 1997 to characterize and standardize the complexity of endotracheal intubation , and with the objective in mind to ‘provide a uniform approach to compare studies related to difficult intubation, and with the aim of determining the relative values of risk factors of intubation difficulty’ . Since then, IDS more than 5 has been used as the definition of difficult intubation in different populations, in particular by Combes et al. to determine predictive factors of a difficult airway in the prehospital setting, by Amathieu et al. to assess risk factors of difficult intubation in thyroid surgery, and recently by Gonzalez et al. to evaluate risk factors of difficult intubation in obese patients. However, our study is the first that effectively assessed the risk factors of IDS more than 5 in a scheduled general surgery setting.
To generalize the culture of difficult airway assessment in preoperative period, screening tools have to be simple and easy to use. We think that the major strength of SPIDS is its ease and quick assessment. The airway physical examination requires two steps, with two airway tests at each step, and with the corresponding ‘points’ for each airway test very easy to memorize:
- The first step is carried out standing in front of the patient to assess both mouth opening (10 points scored if <3.5 cm) and Mallampati's modified test (10, 15 or 25 points scored if class II, III or IV, respectively).
- The second step is an airway physical examination with the patient in profile to assess both the maximum range of head and neck movement (five points scored if <80°) and thryromental distance (five points scored if calculated RHTMD more than 25, assuming that height is available in preanaesthesia questioning).
Considering the predictive accuracy of our new simplified score, the SPIDS, our findings are in accordance with those reported in the meta-analysis by Toshiya et al.. Consequently, some could argue that attempting to predict difficult intubation using the SPIDS is unlikely to be useful. The main concern of a difficult airway is being able to finally manage it safely and successfully. Nevertheless, the role of preoperative airway assessment should not be underestimated. The preoperative airway assessment is not necessarily to give the exact risk of a difficult airway, but to estimate this risk to transform an unanticipated (potentially repeated laryngoscopic attempts, associated with an increased incidence of airway and haemodynamic complications ) into an anticipated difficult airway in most of the cases. The advantages in attempting to predict difficult intubation are: first, to force the anaesthetists at least to think about airway difficulty, and second to define a preoperative strategy anticipating a difficult airway algorithm including the anaesthetic technique (awake intubation versus intubation after induction of general anaesthesia) and the airway control device with primary (plan A) and alternative strategies (plan B, C and others). Moreover, despite the moderate accuracy of the SPIDS, the proportion of well classified patients was 74.9%, about three patients out of four (proportion calculated from the values of sensitivity and specificity). The high NPV of the SPIDS (97%) implies that, when a score of 10 or less is found, a difficult intubation can be excluded, with a risk of false prediction of 3%. Consequently, a preoperative airway assessment performed by using the SPIDS will identify most of the patients who ultimately will have either a difficult airway or not, indicating that use of the SPIDS could be recommended on these bases.
Moreover, considering the five risk factors independently associated with an IDS more than 5 (Table 5), our findings are in accordance with the recommendations of Canadian and French guidelines for difficult airway management [5,26]. So, performing the SPIDS in daily clinical practice should not be more time-consuming than current guidelines. Furthermore, the best accuracy of the SPIDS in comparison with the corresponding nonweighted scores is another argument in favour of using the SPIDS in daily practice during preoperative difficult airway assessment. So by weighting risk factors of difficult intubation, SPIDS could help anaesthesiologists to judiciously decide (SPIDS >10), or not (SPIDS ≤10), to begin the anaesthetic induction with the portable storage unit for difficult airway management in the operating theatre.
Our study does, however, have some limitations. Some methodological bias could explain any potential overestimation in the calculation of the IDS: the use of muscle relaxant and the experience of the initial operator performing the tracheal intubation were neither controlled nor standardized. Consequently, a two-point addition to the IDS was in theory possible: one point for the vocal cords in adduction and one point for a second operator. Nevertheless, the incidence of difficult intubation (6%) in the current study is identical to that observed in the literature [20,21]. Our population seems to be representative of daily clinical practice, indicating that these biases had no major impact on the results of the current study. Another limitation is the validation of the SPIDS from a general surgical population. Other prospective studies are needed to determine whether the SPIDS can be applied to specific populations such as ENT surgical or obstetric populations with this accuracy.
In conclusion, we have demonstrated that, in a general surgical population, pathological conditions associated with difficult intubation (see Table 1), mouth opening less than 3.5 cm, head and neck movement less than 80°, the RHTMD 25 at least, and Mallampati's modified test 2 at least, are five adjusted risk factors of difficult intubation defined by an IDS more than 5. Taking into account the respective weight of these five risk factors, we propose a new simplified weighted score to predict difficult intubation, the SPIDS.
Some guidelines have already advised the preoperative assessment of risk factors included in the SPIDS, and the predictive accuracy of the SPIDS is statistically better than the corresponding nonweighted score. Therefore, performing the SPIDS in daily practice, first, should not be more time-consuming, and, second, should slightly improve the predictive accuracy of preoperative airway assessment. The SPIDS could help anaesthesiologists to define difficult airway management strategy: a value of SPIDS strictly above 10 could encourage the anaesthesiologists to plan for the beginning of the anaesthetic induction with ‘alternative’ airway devices ready in the operating theatre.
1 Cheney F, Posner K, Lee L, et al
. Trends in anesthesia-related death and brain damage. A closed claims analysis. Anesthesiology 2006; 105:1081–1086.
2 Lienhart A, Auroy Y, Péquignot F, et al
. Survey of anesthesia-related mortality in France. Anesthesiology 2006; 105:1087–1097.
3 Peterson GN, Domino KB, Caplan RA, et al
. Management of the difficult airway
: a closed claims analysis. Anesthesiology 2005; 103:33–39.
4 Practice Guidelines for Management of the Difficult Airway
. An updated report by the American Society of Anesthesiologists Task Force on management of the difficult airway
5 Diemunsch P, Langeron O, Richard M, Lenfant F. Prediction and definition of difficult mask ventilation and difficult intubation
: question 1. Société Française d'Anesthésie et de Réanimation. Ann Fr Anesth Reanim 2008; 27:3–14.
6 Wilson ME, Spiegelhalter D, Robertson JA, Lesser P. Predicting difficult intubation
. Br J Anaesth 1988; 61:211–216.
7 El-Ganzouri AR, McCarthy RJ, Tuman KJ, et al
. Preoperative airway assessment
: predictive value of a multivariate risk index. Anesth Analg 1996; 82:1197–1204.
8 Karkouti K, Rose DK, Wigglesworth D, Cohen MM. Predicting difficult intubation
: a multivariable analysis. Can J Anesth 2000; 47:730–739.
9 Arné J, Descoins P, Fusciardi J, et al
. Preoperative assessment
for difficult intubation
, in general
and ENT surgery: predictive value of a clinical multivariate risk index. Br J Anaesth 1998; 80:140–146.
10 Rosenstock C, Hansen EG, Kristensen MS, et al
. Qualitative analysis of unanticipated difficult airway
management. Acta Anaesthesiol Scand 2006; 50:290–297.
11 Auroy Y, Benhamou D, Péquignot F, et al
. Mortality related to anaesthesia
in France: analysis of death related to airway complications. Anaesthesia
12 Sansoon GLT, Young JRB. Difficult tracheal intubation
: a retrospective study. Anaesthesia
13 Lewis M, Keramati S, Benumof JL, Berry C. What is the best way to determine oropharyngeal classification and mandibular space length to predict difficult laryngoscopy? Anesthesiology 1994; 81:69–75.
14 Mallampati SR, Gatt SP, Gugino LD, et al
. A clinical sign to predict difficult tracheal intubation
: a prospective study. Can Anaesth Soc J 1985; 32:429–434.
15 Langeron O, Masso E, Huraux C, et al
. Prediction of difficult mask ventilation. Anesthesiology 2000; 92:1229–1236.
16 Adnet F, Borron SW, Racine SX, et al
. The Intubation
Difficulty Scale (IDS): proposal and evaluation of a new score characterizing the complexity of endotracheal intubation
. Anesthesiology 1997; 87:1290–1297.
17 Delong ER, Delong DM, Clarke-Pearson DL. Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics 1988; 44:837–845.
18 Efron B, Tibshirani R. Estimates of bias. In: Cox DR, Hinkley DV, Reid N, Rubin DB, Silverman DW, editors. An introduction to the bootstrap. New York: Chapman & Hall; 1993. pp. 124–140.
19 Juvin P, Lavaut E, Dupont H, et al
. Difficult tracheal intubation
is more common in obese than in lean patients. Anesth Analg 2003; 97:595–600.
20 Toshiya S, Zen'ichiro W, Tetsuo I, Atsuhiro S. Predicting difficult intubation
in apparently normal patients: a meta-analysis of bedside screening test performance. Anesthesiology 2005; 103:429–437.
21 Molliex S, Berset JC, Billard V, et al
. Airway management in adult anesthesia with the exception of difficult intubation
. Recommendations of the jury. Ann Fr Anesth Reanim 2003; 22:745–749.
22 Yentis SM. Predicting difficult intubation
– worthwhile exercise or pointless ritual? [Editorial]. Anaesthesia
23 Cormack RS, Lehane J. Difficult tracheal intubation
in obstetrics. Anaesthesia
24 Cattano D, Panicucci E, Paolicchi A, et al
. Risk factors assessment
of the difficult airway
. Anesth Analg 2004; 99:1774–1779.
25 Garcia-Guiral M, Garcia-Amigueti F, Ortells-Polo MA, et al
. Relationship between laryngoscopy degree and intubation
difficulty. Rev Esp Anesthesiol Reanim 1997; 44:93–97.
26 Crosby ET, Cooper RM, Douglas MJ, et al
. The unanticipated difficult airway
with recommendations for management. Can J Anaesth 1998; 45:757–776.
27 Combes X, Jabre P, Jbeili C, et al
. Prehospital standardization of medical airway management: incidence and risk factors of difficult airway
. Acad Emerg Med 2006; 13:828–834.
28 Amathieu R, Smail N, Catineau J, et al
. Difficult intubation
in thyroid surgery: myth or reality? Anesth Analg 2006; 103:965–968.
29 Gonzalez H, Minville V, Delanoue K, et al
. The importance of increased neck circumference to intubation
difficulties in obese patients. Anesth Analg 2008; 106:1132–1136.
30 Mort TC. Emergency tracheal intubation
: complications associated with repeated laryngoscopic attempts. Anesth Analg 2004; 99:607–613.