Modified Mallampati Airway Grade Interpreter
Modified Mallampati Airway Grade Interpreter
Four classes of pharyngeal view, graded in about three seconds and agreed on rather less reliably than that suggests — published inter-rater kappas for this scale run from 0.10 to 0.80, and the technique used to obtain the view is part of the reason.
Modified Mallampati grade
View and technique → grade and what it predictsSitting upright, mouth maximally open, tongue maximally protruded, not phonating; soft palate and the base of the uvula visible
The classes, the technique, and the agreement between raters
I all four structures · II pillars hidden · III soft palate and uvular base only · IV soft palate not visible
- inter-rater reliability, which is the honest measure
- the published kappas span more than sevenfold and are the reason this page exists. Between specialists, 0.80; between residents, 0.41; between dental students and a dentist, 0.54; between gastroenterologists and anaesthetists, 0.103 — no better than chance; between one gastroenterologist and another, 0.120; and for the same gastroenterologist examining the same patient twice, 0.420. A three-way Fleiss kappa across patient, physician assistant and anaesthetist was 0.195, with patient and anaesthetist agreeing exactly half the time. That the same rater agrees with themselves better than with a colleague, and still only moderately, says the grade is partly a property of the rater
- 1985 and 1987 are two scales
- Mallampati’s original had three classes, derived in a 210-patient prospective series. Samsoon and Young split the original class 2 by uvular visibility and added a class IV in which the soft palate is not visible at all. A chart recording “class 3” without saying which scale is ambiguous between a second-worst grade on one and a third-of-four on the other, and the two performed very differently in meta-analysis: summary ROC area 0.83 for the modified test against 0.58 for the original against difficult intubation
- accuracy, from two meta-analyses
- Lee and colleagues, 42 studies and 34,513 patients: summary ROC areas of 0.89 and 0.78 for difficult laryngoscopy and 0.83 (modified) against 0.58 (original) for difficult intubation, with both poor for difficult mask ventilation, and the conclusion that they are not useful as standalone screening tests. The larger review of 133 studies and 844,206 participants: modified Mallampati sensitivity 0.53 and specificity 0.80 for difficult laryngoscopy, 0.51 and 0.87 for difficult intubation, 0.17 and 0.90 for difficult mask ventilation — while still being the best single predictor of difficult intubation among the bedside tests examined
- why the predictive value collapses
- Mallampati’s 210-patient series reported specificity 0.995 and positive predictive value 0.933. In 1,518 unselected adults with 48 difficult intubations — a prevalence of 3.2 per cent — sensitivity was 0.646, specificity 0.824, negative predictive value 0.986 and positive predictive value 0.107. The test did not get worse; the prevalence fell, and a test applied to a low-prevalence population generates mostly false positives whatever its specificity
- what it is not
- not a diagnosis, not a plan, and not a test that should be read alone — which is the explicit conclusion of the larger of the two meta-analyses. It is one component of a structured airway assessment alongside mouth opening, thyromental distance, neck movement, mandibular protrusion, dentition and a history of previous difficulty, and this page grades the one component
Worked example
Sitting upright, mouth maximally open, tongue maximally protruded, not phonating; soft palate and the base of the uvula visible
The technique was as published, so the grade is comparable with the literature
Soft palate and uvular base only is class III on Samsoon and Young's four-class modification
In the 42-study, 34,513-patient meta-analysis the modified test reached a summary ROC area of 0.83 for difficult intubation — and that review concluded the test is “not useful as a standalone screening test”
In an unselected clinic the positive predictive value of a high grade is about 0.107: roughly one high grade in nine marks an airway that actually proves difficult, because difficult intubation runs at about 3 per cent
Change the technique answer to no and the result becomes Not comparable. Phonation lifts the soft palate and this exact view is the one that most often becomes a class II when the patient says “ahh”
Published kappas for this grade run from 0.10 between specialties to 0.80 between specialists. A second assessor may well record something else, and that is a property of the scale rather than of either assessor
The four classes, and the two renderings in circulation
| Class | Samsoon and Young wording, as reproduced in airway texts | The other widely published rendering |
|---|---|---|
| I | Soft palate, uvula, fauces and pillars all visible | Identical in both |
| II | Soft palate, fauces and the majority of the uvula visible | Faucial pillars and soft palate visible; the tongue base masks the uvula |
| III | Soft palate and the base of the uvula visible | Soft palate visible |
| IV | Only the hard palate visible | Soft palate not visible |
Published inter-rater agreement for this grade
| Comparison | Kappa | Interpretation |
|---|---|---|
| Two specialists | 0.80 | Good |
| Dental students against a dentist | 0.54 | Moderate, 77 per cent agreement |
| Two residents | 0.41 | Moderate |
| The same gastroenterologist, same patient, twice | 0.420 | Moderate — intra-rater, not inter-rater |
| Patient, physician assistant and anaesthetist together | 0.195 (Fleiss) | Poor; patient and anaesthetist agreed 50 per cent of the time |
| Two gastroenterologists | 0.120 | Slight |
| Gastroenterologist against anaesthetist | 0.103 | Slight — described as no better than chance |
Three seconds to grade, and seven published kappas to interpret
Mallampati and colleagues described a clinical sign in 1985: with the patient sitting, mouth open and tongue protruded, how much of the oropharynx can be seen predicts how difficult direct laryngoscopy will be. Their series of 210 patients gave three classes. Samsoon and Young split the middle class and added a fourth in 1987, and that four-class modification is what almost everybody now means by a Mallampati grade — though a chart recording “class 3” without naming the scale is genuinely ambiguous between them, and the two performed very differently in meta-analysis.
The technique is not a formality. The grade is obtained sitting upright, with the mouth opened as wide as it will go, the tongue protruded as far as it will go, and without phonating. Phonation lifts the soft palate; partial tongue protrusion leaves the pharynx hidden. Both departures shift the grade, and both shift it in the reassuring direction, so a grade obtained carelessly is not merely noisy but biased. This page asks about the technique before it asks about the view, and reports a non-standard examination as not comparable with the literature rather than grading it anyway.
Inter-rater reliability is the measure that says how much the number means, and for this scale it is the most interesting thing about it. Published kappas: 0.80 between specialists, 0.41 between residents, 0.54 between dental students and a dentist, 0.120 between two gastroenterologists, and 0.103 between gastroenterologists and anaesthetists — described in that study as no better than chance. A three-way Fleiss kappa across patient, physician assistant and anaesthetist was 0.195, with patient and anaesthetist agreeing exactly half the time. Most telling: the same gastroenterologist examining the same patient twice achieved 0.420. If a rater agrees with themselves only moderately, the grade is partly a property of the rater.
The accuracy figures are correspondingly modest and are often quoted from the wrong population. Mallampati’s own 210-patient series reported a specificity of 0.995 and a positive predictive value of 0.933. Applied to 1,518 unselected adults, among whom difficult intubation occurred in 48 — 3.2 per cent — the specificity was 0.824 and the positive predictive value fell to 0.107. The test did not get worse; the prevalence fell, and a test applied to a low-prevalence population produces mostly false positives whatever its specificity. Across 42 studies and 34,513 patients the modified test reached a summary ROC area of 0.83 for difficult intubation, and that review’s own conclusion was that the Mallampati tests have limited accuracy alone and are not useful as standalone screening tests. The larger review of 133 studies and 844,206 participants found a sensitivity of about 0.5 — while still rating it the best single bedside predictor of difficult intubation available. A score’s output is a cohort frequency, not this patient’s probability: a stratum in which 9 per cent had an event describes that stratum, not which 9 per cent. This page reports published figures and recommends no action.
Frequently asked questions
What are the modified Mallampati classes?
Class I, soft palate, uvula, fauces and pillars all visible; class II, soft palate, fauces and most of the uvula visible with the pillars hidden; class III, soft palate and the base of the uvula only; class IV, soft palate not visible. Mallampati’s 1985 original had only three classes; Samsoon and Young added the fourth in 1987.
How should a Mallampati grade be obtained?
Patient sitting upright, mouth opened as wide as it will go, tongue protruded as far as it will go, and not phonating. Phonation and incomplete tongue protrusion both change the view and both make it look better than it is, so technique is part of the measurement rather than a detail.
How reliable is the Mallampati grade between assessors?
Published kappas run from 0.103 between gastroenterologists and anaesthetists — described as no better than chance — up to 0.80 between specialists, with 0.41 between residents and a three-way Fleiss kappa of 0.195 across patient, physician assistant and anaesthetist. The same rater examining the same patient twice managed 0.420, which is moderate agreement with oneself.
Does a Mallampati class III or IV mean the intubation will be difficult?
No. In an unselected cohort of 1,518 adults, where difficult intubation occurred in 3.2 per cent, the positive predictive value of a high grade was 0.107 — roughly one high grade in nine. Mallampati’s own series reported 0.933, but that was a different and much higher-prevalence population. A low grade is more informative, with a negative predictive value of 0.986 in the same unselected cohort.
Is the Mallampati grade enough on its own?
No, and the meta-analysis of 42 studies says so explicitly: the tests have limited accuracy alone and are not useful as standalone screening tests. It is one component of a structured assessment alongside mouth opening, thyromental distance, neck movement, mandibular protrusion, dentition and any history of previous difficulty.
Related calculators
References
- 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(4):429–434.
- Samsoon GLT, Young JRB. Difficult tracheal intubation: a retrospective study. Anaesthesia. 1987;42(5):487–490.
- Lee A, Fan LTY, Gin T, Karmakar MK, Ngan Kee WD. A systematic review (meta-analysis) of the accuracy of the Mallampati tests to predict the difficult airway. Anesth Analg. 2006;102(6):1867–1878.
- Roth D, Pace NL, Lee A, et al. Airway physical examination tests for detection of difficult airway management in apparently normal adult patients. Cochrane Database Syst Rev. 2018;(5):CD008874.
- Adamus M, Fritscherova S, Hrabalek L, Gabrhelik T, Zapletalova J, Janout V. Mallampati test as a predictor of laryngoscopic view. Biomed Pap Med Fac Univ Palacky Olomouc Czech Repub. 2010;154(4):339–343.
- Vannucci A, Rosenblatt WH, et al. Are physician assistant and patient airway assessments reliable compared to anesthesiologist assessments in detecting difficult airways in general surgical patients? Perioper Med. 2017;6:21.
Not medical advice. For healthcare professionals and education. Reference intervals vary by laboratory and assay — always use your own laboratory's. Never base a dose or a treatment decision on this page alone. Full disclaimer at calcengines.com/disclaimer/