What is AHI? The apnea-hypopnea index, explained

Two kinds of breathing event go into the index.

An apnea is a near-complete stop in airflow. The American Academy of Sleep Medicine (AASM) scores an apnea when airflow drops by 90 percent or more from baseline for at least 10 seconds.

A hypopnea is a partial reduction. The AASM's recommended rule scores a hypopnea when airflow drops by at least 30 percent for at least 10 seconds and that drop is accompanied by either a 3 percent fall in blood oxygen or an arousal from sleep. An older, still-used alternative rule requires a 4 percent desaturation and does not count arousals. See hypopnea explained for why that difference matters.

Add the apneas and hypopneas together, divide by hours, and you have AHI. An AHI of 4.2 means roughly four events every hour across the night.

For adults, the widely used bands are:

These bands come from a diagnostic sleep study, not from a CPAP machine. They describe how often you stop breathing without treatment. Once you are on therapy, the number you see each morning is your residual AHI: what is left after the machine has done its work. The target for treated apnea is usually a residual AHI under 5, and many people run well below that. What is a good AHI on CPAP? goes into this in more detail.

In polysomnography you are wired for brain activity (EEG), eye movement, muscle tone, airflow at the nose and mouth, chest and abdominal effort belts, and pulse oximetry. A technologist scores each event by hand, or checks automated scoring, against the AASM rules. Because the lab can see when you are asleep, the denominator is true sleep time. Because it can see oxygen and arousals, it can apply the full hypopnea rule.

A home sleep apnea test records fewer channels, usually airflow, effort, oximetry and position. It cannot see sleep stages, so it reports events per hour of recording, which tends to run a little lower than a lab AHI because some of that recording time is spent awake.

Your machine has one sensor that matters here: a flow sensor in the blower. It sees the air moving through the tube. From that single signal it has to infer whether you are breathing normally, partially, or not at all, and whether a pause was obstructive or central.

That means three important differences from a lab number.

The denominator is usage, not sleep. The machine counts events per hour the mask was on and the blower was running. If you lie awake for an hour reading with the mask on, that hour is in the denominator with few events in it, which lowers the AHI.

There is no oxygen and no EEG. The machine cannot see a desaturation or an arousal, so it cannot apply the AASM hypopnea rule as written. Manufacturers use flow-only definitions instead. ResMed, for example, scores a hypopnea on a sustained reduction in airflow alone. A hypopnea on your machine and a hypopnea in a lab are therefore related but not identical events.

Wake breathing can look like events. Sighs, swallows, talking and shifting position can produce flow patterns the algorithm flags. Most of these are brief and get filtered, but some will be counted.

Studies comparing machine-estimated AHI with simultaneous polysomnography generally find good agreement when leak is low, with the machine tending to under- or over-count in specific situations such as high leak or a high proportion of central events. The practical reading is simple: a machine AHI is a reliable trend indicator and a reasonable estimate, not a diagnosis.

If a machine AHI is an estimate, why look at it? Because it is measured every single night, in your own bed, for as long as you use therapy. A sleep study gives you one night in a strange room. Your machine gives you thousands of nights. Trends over weeks and months, and changes after a settings or mask change, are exactly what the machine number is good at.

It is also the number your clinician sees when they pull your data, so understanding it puts you on the same page.

A few things reliably move the number on CPAP.

• Leak. When a lot of air escapes around the mask, the machine struggles to measure flow accurately and may miss events or mark them as unclassified. See CPAP leak rate explained.
• Pressure. Too little pressure leaves obstructive events behind. Too much can, in some people, provoke central events. Your pressure range is a clinical decision. See CPAP pressure settings explained.
• Sleep position and alcohol. Back sleeping and alcohol both relax the airway. Many people see a clear difference in their nightly AHI.
• Illness and congestion. A blocked nose changes breathing patterns and can raise events for a week.
• Time in bed awake. More awake time with the mask on dilutes the index.

On your dashboard, last-night AHI appears as a headline tile with a colour band: Great, Mild, Moderate or High. Below it, the event composition ring breaks the night into obstructive, central, hypopnea and unclassified so you can see what kind of events made up the number, not just how many.

The 30-night rolling AHI trend sits beside it, and the Heads up panel flags an unusual night in plain words, for example "AHI 8.5 — 2.9× your usual ~3.0". That comparison to your own baseline is often more useful than the absolute number.

In Trends you can stretch the view to 90 nights, a year or all time, compare one period against another, and overlay leak, pressure and usage to see what moved together. Pro users can open any night in the breath-by-breath viewer and see the flagged events on the airflow trace itself.

All of this is read from the SD card of a ResMed AirSense, AirCurve or S9 machine through the web SD card import.

• My residual AHI is usually around X. Is that where you would expect it to be for me?
• Most of my events are hypopneas / central / unclassified. Does that change anything?
• My AHI jumped after I changed masks. Could leak be behind it?

Bring the trend, not a single night. One bad night is weather; four bad weeks is climate.