What the STOP-BANG Score Actually Measures
This sleep apnea quiz is engineered to be wrong in one specific direction, and that design choice is the single most useful thing about it. The STOP-BANG questionnaire behind it was never built to tell you whether you have obstructive sleep apnea. It was built to make sure that almost nobody who does have it walks out of a clinic unflagged β and it pays for that with a spectacular rate of false alarms. Read the score knowing which direction it leans and it becomes genuinely useful. Read it the way most online quizzes present it and you will either panic over nothing or, worse, feel reassured by a number that was never meant to reassure you.

A Screening Test Built to Be Wrong (On Purpose)
Every diagnostic test trades sensitivity against specificity, and STOP-BANG sits at an extreme end of that trade. Pooled across general population and clinical samples, a score of 3 or more catches around 88% of moderate-to-severe cases, and the negative predictive value of a score under 3 lands near 93%. Those are strong numbers. Now the other side: when any three items are positive, specificity for moderate-to-severe apnea is roughly 31%. Roughly two out of three flagged people do not have the thing they were flagged for.
That is not a broken instrument. It is a triage instrument doing exactly its job. The cost of missing a real case β years of untreated hypertension, a road accident, atrial fibrillation β is enormous. The cost of a false positive is one sleep study. So the designers pushed the dial hard toward catching everyone, and the American Academy of Sleep Medicine backs that up with a strong recommendation against using questionnaires to diagnose apnea at all in the absence of a sleep study. The questionnaire opens a door. It does not walk through it.
Which is why the asymmetry panel in your results matters more than the number above it. A low score genuinely tells you something. A high score tells you to go and find out.
The Eight Items, and Which Half Carries the Score
The acronym splits cleanly in half, and the halves do different work. STOP is what you notice. BANG is what you happen to be.
| Item | What you're asked | What it's really a proxy for |
|---|---|---|
| Snoring | Loud enough to hear through a closed door | Turbulent airflow through a narrowed pharynx |
| Tired | Daytime fatigue or sleepiness | The downstream cost of fragmented sleep architecture |
| Observed | Witnessed pauses, gasping or choking | The closest thing to a direct observation of an apnea event |
| Pressure | High blood pressure, treated or not | Cardiovascular strain from repeated overnight oxygen dips |
| BMI | Above 35 | Fat around the pharynx plus reduced lung volume tugging on the airway |
| Age | Older than 50 | Age-related loss of upper-airway muscle tone |
| Neck | 40 cm (16 in) or more | The best single external measurement of soft-tissue crowding |
| Gender | Male | Airway length, fat distribution β and a diagnostic bias, see below |
Here is the part that tends to surprise people. The published refinement for ambiguous scores doesn't look at your symptoms at all β it takes your STOP score as a gate and then asks about your BMI, your neck and your sex. The anatomy items are carrying information the symptom items can't, largely because self-reported symptoms are unreliable and a tape measure isn't. That is also why untreated apnea shows up so reliably alongside metabolic problems; if you scored a point for blood pressure, the diabetes risk quiz is a reasonable second stop, because insulin resistance and nocturnal oxygen dips reinforce each other in both directions.
Three People, Three Scores of Four
Nothing exposes the weakness of a bare total like watching three people land on the same one. Each of these scores exactly 4 out of 8. Each means something completely different.
| Dan, 54 | Marta, 47 | Ken, 62 | |
|---|---|---|---|
| Snoring | Yes β loud | Yes β loud | No |
| Tired | No | Yes, daily | No |
| Observed pauses | Sleeps alone β unknown | Yes, partner has seen it | No |
| Blood pressure | Normal | Treated hypertension | On medication |
| BMI | 29 | 31 | 37 β |
| Age over 50 | β | No | β |
| Neck 40 cm+ | β | No | No |
| Male | β | No | β |
| Total | 4 | 4 | 4 |
| STOP subtotal | 1 | 4 | 1 |
Dan's four points are almost entirely anatomy, and his single unknown is the heaviest item on the form. He can't reach the refinement gate because his STOP score is 1 β but if a partner ever confirms he gasps at night, his STOP hits 2, and he then matches the refined criterion on two counts at once. His score is one observation away from moving a whole band.
Marta is the interesting one, and the reason this section exists. All four of her points come from the STOP half. She snores loudly, she's exhausted every day, her partner has watched her stop breathing, and she's on blood pressure medication. By the letter of the published refinement she stays at intermediate risk, because that rule only checks BMI, neck and male sex β and she matches none of them. A woman with a perfect symptom score gets filed next to a man with none. If your result panel told you your score was built from symptoms rather than measurements, that is the situation you are in, and it should be treated like a 5.
Ken is Marta inverted. He reports no symptoms whatsoever. His four points are a BMI of 37, an age over 50, blood pressure medication and being male β pure risk, zero evidence. That is a legitimate reason to test him, since a large share of apnea patients genuinely don't notice their own symptoms, but it is a different conversation from Marta's. The number is identical. The meaning isn't remotely.
The composition is the information.Take "I scored 4, and 4 of those came from symptoms" to a doctor. Never just the total.
What to Do When You Land on 3 or 4
The original scoring left an awkward hole. Zero to two is low risk, five to eight is high risk, and three to four is β what, exactly? Chung and colleagues published a second step for precisely this band in their 2016 review in Chest, and it is the least-known and most useful part of the whole instrument. Instead of counting any three items, you require at least two positives from the STOP half plus one specific body item. The effect on specificity is dramatic:
| Combination | Specificity for moderate-to-severe OSA |
|---|---|
| Any three positive items | ~31% |
| STOP β₯ 2 plus male sex | ~77% |
| STOP β₯ 2 plus neck over 40 cm | ~79% |
| STOP β₯ 2 plus BMI over 35 | ~85% |
From 31% to 85% is not a tweak. It is the difference between a screening tool that mostly cries wolf and one a clinician can act on. The quiz above runs this check for you and says so explicitly when it applies, because almost no consumer sleep apnea quiz implements it β most stop at the raw total and leave half their users in a band that was never meant to be a final answer.
Why the Questionnaire Hands Men a Free Point
It looks crude, and the objection people raise is fair: why should sex alone be worth the same as witnessed apneas? The honest answer is that the item is doing real predictive work in the populations the tool was validated on. The problem isn't the point. It's everything that point sits next to.
Women with obstructive sleep apnea tend to present differently. Instead of loud snoring and dramatic witnessed pauses, the complaint is more often insomnia, morning headache, low mood and fatigue that no amount of sleep touches. Those symptoms map poorly onto a form built around snoring and choking, so women score lower on the STOP half as well as the G item. The consequence was measured decades ago and hasn't moved much: Young and colleagues found in 1997 that 93% of women and 82% of men with moderate-to-severe sleep apnea had never been clinically diagnosed, in a population with no obvious barrier to healthcare.
If that pattern sounds familiar, it should. It is the same structural problem that runs through the women's heart attack symptom quiz β an instrument calibrated on male presentation, applied to everyone, quietly under-scoring half the population. The practical fix is unglamorous: if you're a woman with a 3 or a 4 built out of symptoms, ignore the band label and ask for the test.
The One Question You Can't Answer About Yourself
There is a genuine hole at the centre of every self-administered apnea screener, and nobody likes to mention it. The observed-apnea item carries more diagnostic weight than any other symptom on the form, and it is the one item that requires another conscious human in the room. If you sleep alone, you are asked to leave your heaviest question blank.
Paper forms handle this badly. A blank gets scored as a "no", which silently drops your total and can move you down a band on the basis of nothing at all. The quiz above scores the blank as zero too β it has to, or the instrument stops being the instrument β but it flags it back to you rather than pretending the zero was information.
The workaround costs nothing. Leave a phone recording audio beside the bed for two or three nights and scrub through the file. You are listening for the shape of it: snoring that builds, then silence, then a sharp resumption. Consumer wearables that track overnight blood oxygen give a second, cruder signal. Neither is diagnostic and neither should be treated as one. But in practice, hearing your own breathing stop is the thing that finally gets people to book the appointment they have been putting off for two years.
Five Nights That Feel Like Apnea and Aren't
Exhaustion is a lousy diagnostic clue because almost everything produces it. That is why this quiz scores a second, independent reading from ten of your answers β five patterns that arrive at the same wrecked morning by different routes.
The overlap most worth knowing about is insomnia. Comorbid insomnia and sleep apnea has its own acronym now, COMISA, because roughly 30 to 50% of people with obstructive sleep apnea also report insomnia symptoms, and 30 to 40% of insomnia patients meet criteria for apnea. Treating one and ignoring the other is a well-documented route to someone abandoning their CPAP machine in a cupboard within a month. If your insomnia bar came back high alongside a high STOP-BANG score, that combination is the finding, not a contradiction.
The timing pattern is the other one worth separating out. If your sleep is perfectly fine when nobody sets an alarm and catastrophic on a Tuesday, you may be looking at circadian misalignment rather than a breathing problem β a different mechanism with completely different treatment. The chronotype quiz is the better tool for that, and the exhaustion-that-sleep-doesn't-fix pattern is worth checking against the burnout stage quiz too. And then there is the most common and least interesting explanation of all: not enough hours. No apnea questionnaire asks how long you were actually in bed, so no apnea questionnaire can ever catch it.
Every Band and Every Pattern, Explained
Low risk (0 to 2).The band the questionnaire answers confidently. A score under three carries a negative predictive value around 93% for moderate-to-severe apnea, and in the original series this band ran at roughly an 18% probability of moderate-to-severe disease and about 4% for severe. If you scored here and you're still exhausted, the cause is almost certainly one of the look-alike patterns rather than your airway.
Intermediate risk (3 to 4). The unresolved middle, and the band where composition matters most. Four points of symptoms is a different animal from four points of body measurements. This is where the refined criteria apply and where a symptom-heavy score deserves to be escalated by hand.
Intermediate, reclassified high. You cleared the STOP gate with at least two symptom items and matched one of BMI over 35, a neck over 40 cm, or male sex. Specificity for moderate-to-severe apnea jumps from about 31% to somewhere between 77% and 85% in this configuration. Practically, this belongs in the same queue as a raw 5.
High risk (5 to 8). Enough items are positive that a sleep study is warranted. In the original series, the top band carried roughly a 60% probability of moderate-to-severe apnea and about 38% for severe β high, and still not a diagnosis. Bring the item breakdown to the appointment, not the headline number.
Airway collapse pattern. Snoring, gasping, nocturia, unrefreshing sleep however long it lasts, and a dramatic worsening after alcohol. This is the shape STOP-BANG was designed to detect, and when this bar tops out alongside a high score the two readings are agreeing with each other.
Nose and throat pattern. Congestion, mouth-breathing, a dry morning mouth, snoring that vanishes on your side. This sits upstream of the collapse rather than replacing it, and it frequently coexists with genuine apnea. It also tends to be the cheapest thing to address first.
Wired-and-awake pattern. Long sleep latency, 3am wake-ups, dread at bedtime, a drink that knocks you out and then abandons you. That is insomnia, and the COMISA overlap means it very often rides alongside apnea rather than excluding it.
Not-enough-hours pattern. Under six hours on a work night, a two to four hour catch-up at weekends, and mornings where you feel fine except for wanting more time. Unglamorous, extremely common, and invisible to every apnea screener ever published.
Wrong-clock pattern. Sleep that works beautifully on your own schedule and collapses against an alarm β asleep at 2am and up at midday given the chance, or asleep at 9pm and awake at 4am. That is a timing disorder, not a breathing one, and light timing rather than a CPAP machine is the lever that moves it.
Getting Tested Is Less Trouble Than It Used to Be
The mental image most people carry β a night in a lab, wired to thirty electrodes, somebody watching on a monitor β is now the exception rather than the rule. For an uncomplicated adult with a high pre-test probability, a home sleep apnea test posted to your door is accepted practice. You wear a chest band, a nasal cannula and a finger probe in your own bed and post the box back.
In-lab polysomnography is still the right call in specific circumstances: significant heart or lung disease, suspected central rather than obstructive apnea, neuromuscular conditions, or another sleep disorder muddying the picture. It is also the usual next step when a home test comes back negative in someone whose symptoms are loud, since home testing can understate severity.
One thing genuinely doesn't wait. If you answered that you have fallen asleep at the wheel or come close to it, that answer outranks the entire questionnaire, whatever your score. Say it out loud at the appointment β those exact words move you up a waiting list in a way that "I scored 6 on an online quiz" never will.
