Longevity UniversityAsk the evidence

How accurate is the STOP-Bang questionnaire at detecting sleep apnea?

Answered November 2, 202612 sources

EvidenceBModerateGood evidence in people with one real weakness. New studies could change the answer.

STOP-Bang catches most people with sleep apnea in sleep clinics and before surgery, but it also flags many people who don't have it. A low score is more reliable than a high one.

Who it's for
People with suspected sleep apnea, seen at sleep clinics or before surgery.
What to expect
It catches about 9 in 10 moderate or severe cases at a score of 3.
Watch out for
Many false alarms, and weaker accuracy in some groups and regions.

Figure 3. Forest plot of pooled diagnostic odds ratio of STOP-Bang questionnaire for various OSA severities in surgical patients.

Figure 3 from Validation of the STOP-Bang questionnaire as a preoperative screening tool for…, BMC Anesthesiol 2022. Licence: CC BY 4.0

Figure 1. Forest Plot for Combined Sensitivity and Specificity for All Obstructive Sleep Apnea for All Included Studies in the Sleep Clinic FN indicates false-negative; FP, false-positive;…

Figure 1 from Use and Performance of the STOP-Bang Questionnaire for Obstructive Sleep Apnea…, JAMA Netw Open 2021. Licence: CC BY

What the studies show

  • A meta-analysis A study that pools the results of several earlier studies into one overall estimate. It is usually the strongest kind of evidence. pooling 47 studies and 26,547 adults referred to sleep clinics found a score of 3 or more caught over 90% of moderate to severe and severe cases. A low score ruled out severe apnea about 91% of the time1.
  • In that analysis, overall accuracy was high in every region except East Asia, where it was close to chance1.
  • A meta-analysis of 17 studies and 9,206 people found sensitivity of 90%, 94% and 96% for any, moderate to severe, and severe apnea in sleep clinics. Higher scores raised the chance of severe apnea, from 25% at a score of 3 to 75% at 7 or 82.
  • Before surgery (10 studies, 3,247 people), the questionnaire caught 85% to 90% of cases, but only 27% to 47% of people without apnea were correctly cleared. Accuracy for moderate to severe and severe apnea was modest4.
  • In the general population (five studies, 8,585 people), sensitivity stayed high and specificity was low to moderate, so false alarms are common3. A smaller study of 242 people in Singapore found lower sensitivity, about 66% to 69%8.
  • In 268 people hospitalized with a lung blood clot, discrimination was poor and 22% of low-risk people still had apnea10.
  • Adding a blood bicarbonate level raised specificity in surgical patients from about 30% to about 82%6.

These are group figures, so what they mean for one person depends on how common apnea is in their setting.

Key studies

StudyWhoWhat it found
Review of 47 studies, 2021126,547 sleep clinic adultsSensitivity over 90%; poor accuracy in East Asia
Review of 17 studies, 201529,206 adultsSensitivity 90–96%; higher scores meant higher risk
Review of 10 studies, 202243,247 surgical patientsSensitivity 85–90%; specificity 27–47%
Review of five studies, 202138,585 general populationHigh sensitivity, low to moderate specificity
Cohort study Researchers follow a group of people over time without giving them a treatment. It can show links, not cause and effect., 202410268 clot patientsPoor discrimination; 22% false negatives

Safety

The questionnaire itself carries no direct risk. A false alarm can lead to an unneeded sleep test, and a false negative can delay diagnosis3.

What we don't know

  • Why accuracy was so low in East Asian studies, and whether other cutoffs would work better there1.
  • How well it works in pregnancy and other special groups. A pregnancy trial of 108 people has been completed11, and a trial in surgical patients is under way12.

Try it yourself: Sleep apnea check

Ask a follow-up

More on tests

Written by AI from the studies listed, without review by a clinician, and graded with the GRADE method. It can be wrong, so check the studies it cites. Education, not medical advice.