Understanding your AHI number: what mild, moderate and severe mean

Mild obstructive sleep apnea is an apnea-hypopnea index of 5 to 14 events per hour, moderate is 15 to 29, and severe is 30 or more. Those bands are the ones the AASM diagnostic guideline (Kapur 2017) uses as AHI cutoffs of 5, 15, and 30 throughout its evidence tables, and the ones population studies such as Wisconsin and Sleep Heart Health use when they label mild, moderate, and severe disease. The integer is useful for talking to payers. It is a blunt instrument for talking about you. Home tests usually undercount. Symptoms can disagree with the number. Device eligibility windows, including Inspire’s, exist because insurers needed a threshold, not because 14 and 16 are different diseases.

What is the AHI actually counting?

Apneas plus hypopneas, divided by hours of sleep on a lab study, or by hours of recording on many home tests. An apnea, in AASM scoring, is a drop of at least 90 percent of airflow for at least 10 seconds. A hypopnea is a 30 percent drop for at least 10 seconds plus either a 3 percent desaturation or an arousal (recommended rule) or a 4 percent desaturation (the rule CMS uses). Change the hypopnea rule and the same night prints a different AHI. That is not lab error. That is two legal definitions.

Kapur 2017 is a diagnostic-testing guideline, not a scoring manual, but it is explicit that AHI depends on which hypopnea definition you pick, and that the discrepancy “introduces complexity in the evaluation of evidence.” Prevalence estimates in the paper’s background (about 14 percent of men and 5 percent of women in one population study) used AHI of 5 or more with 4 percent desaturation hypopneas plus symptoms. Use 3 percent plus arousals and those percentages move.

On a home test the denominator is often recording time, including wakefulness, so the index is frequently labeled REI. Kapur lists the reasons REI underestimates “true” AHI: no EEG, so no arousal-only hypopneas and no accurate sleep time; sensors that fall off; poor signal. A home REI of 8 can be a lab AHI of 18. A home REI of 0, if the night was technically bad or you slept two hours, is not a clean bill of health. Kapur’s strong recommendation: if a single HSAT is negative, inconclusive, or technically inadequate, do polysomnography.

Where do mild, moderate, and severe come from?

From convention that Kapur’s CPG treats as the field’s cutoffs. The guideline’s diagnostic-accuracy tables are organized at AHI of 5 or more, 15 or more, and 30 or more. Increased risk of “moderate to severe OSA,” for choosing a home test, is a clinical phrase in Kapur tied to sleepiness plus snoring, witnessed apneas, or hypertension, and the numeric home of “moderate to severe” in this literature is AHI of 15 or more. ICSD-3-TR, as summarized in the 2025 women-and-scoring review, states the same bands: mild from 5, moderate from 15, severe from 30. Young’s Wisconsin mortality paper used cut points at 5, 15, and 30 to name mild, moderate, and severe SDB. Punjabi’s Sleep Heart Health mortality paper wrote them out as 5.0-14.9, 15.0-29.9, and 30 or more.

So if your report says “moderate OSA, AHI 22,” it is using that shared ladder. It is not using a Westerville-specific scale. Two labs can still disagree if one scored 4 percent hypopneas and one scored 3 percent plus arousals. Always read the methods line.

Mild is not “almost normal” in every patient, and severe is not “you will not wake up tomorrow.” Mortality associations in Wisconsin and SHHS clustered at the severe end. Symptom burden does not always. People with AHI 12 can be wrecked. People with AHI 35 can insist they sleep well. Believe the symptoms and the tracing together.

How much does a home test undercount?

Kapur does not print a single correction factor such as “add 30 percent.” It reports that 26 validation studies suggested potential for clinically significant misclassification of HSAT versus PSG, while randomized management trials in selected uncomplicated patients found similar sleepiness, quality of life, and PAP adherence when a negative home test was followed by PSG. The undercount is real. The clinical pathway that includes a backup lab night is what made home testing acceptable in the uncomplicated group.

Who is not uncomplicated is listed in Kapur recommendation 4: significant cardiorespiratory disease, neuromuscular weakness, hypoventilation, chronic opioids, stroke, severe insomnia. Those patients should not have their AHI assigned by a home box. Neither should someone whose first home test was technically inadequate. Rosen and colleagues, cited in Kapur, found that 30 percent of people with technically inadequate HSATs and 16 percent with a low AHI on HSAT did not proceed to the protocol PSG. That drop-off is how undercounts become false reassurance.

If you already have a home-test number and it feels too low for the snoring the house reports, say so. Sleep testing in this practice includes in-lab studies for exactly that disagreement.

Why is AHI an imperfect severity meter?

It ignores how long events last, how low oxygen goes, which stage they occupy, whether they are supine-only, and how sleepy you are. Two people with AHI 20 can have different hypoxic burden. REM-only disease can look mild overnight and still wreck REM. Position-dependent disease can look mild if you lucked into a side-sleeping night. Punjabi has argued in print that AHI is not the best single metric of sleep-disordered breathing; hypoxic burden and other measures are active research. Clinical reports still lead with AHI because guidelines, Medicare, and device labels do.

Kapur’s questionnaire section is the other half of “imperfect”: tools that try to guess AHI from neck size and snoring misclassify too many people to diagnose without a recording. The recording is necessary. The integer it produces is not a personality.

Treatment response is also not a race to AHI 0. SURMOUNT-OSA dropped mean AHI by about 25 to 29 events per hour from baselines near 50 and still left residual disease in the mean. PAP titration aims for control of events and symptoms, not a trophy number. Residual AHI of 4 on a download with a leaky mask is not the same as residual AHI of 4 with a dry seal and seven hours of use.

Why do Inspire and Medicare care about the integer?

Because coverage rules need a threshold they can audit. FDA labeling for Inspire upper-airway stimulation, after the June 2023 expansion, describes an adult indication window of AHI 15 to 100 for eligible patients who have failed or not tolerated PAP (plus anatomy and DISE constraints). Medicare LCD L38310, which Ohio Medicare contractors use, covers hypoglossal nerve stimulation when AHI is 15 to 65, BMI is under 35, central plus mixed events are under 25 percent of AHI, DISE shows no complete concentric collapse, and CPAP failure or intolerance is documented. Those two windows are both “moderate to severe” in Kapur’s language, but 70 events per hour can be on-label for FDA and off-coverage for Medicare.

That is the example, not a full eligibility lecture (the Inspire therapy page is the place for criteria). It shows why the integer is used: a payer cannot write a policy that says “treat people who look tired.” It can write “AHI 15 to 65.” Your job in clinic is not to worship the cutoff. It is to know which cutoff which payer is using, and to know that an AHI of 14.8 versus 15.2 is a scoring-night accident, not a biological cliff.

Medicare PAP coverage itself uses AHI or REI of 15 or more, or 5 to 14 with symptoms or comorbidities, and it uses the 4 percent hypopnea definition. A lab that only scores 3 percent arousals can diagnose you and still not satisfy the DME documentation. Ask for both numbers when the difference would change coverage.

How should you read your own report?

Find the AHI or REI, the hypopnea definition, the total sleep time or recording time, the REM AHI, the supine AHI, the oxygen nadir, and the time spent below 90 percent if printed. If the summary says “mild” and you cannot stay awake in the afternoon, believe your afternoon. If it says “severe” and you feel fine, still treat the severe number as a medical finding, not as a mood. Bring the full PDF, not a patient-portal snippet that stripped the methods.

For a walkthrough of a report you already have, or to get a first study, call Westerville Sleep Center at (614) 898-9340. The general sleep apnea page is the disease overview. This page is what the integer on that overview is made of.