Other · Sleep medicine — over-the-counter screening
Sleep Apnea Notification Feature (SANF)
Apple Inc.
Almost certainly the largest deployment of a regulated sleep-apnoea screening algorithm to date. Cleared by the FDA in September 2024 on the strength of a 1,448-participant validation against home sleep apnoea testing, Apple's notification feature is engineered for specificity — 98.5 per cent overall, at the cost of missing more than half of moderate cases — a deliberate trade for a consumer device deployed at population scale (Apple states availability in more than 150 countries). Clinicians will meet this device not in procurement reviews but in the consulting room, carried in by patients; the relevant question is what a notification, or its absence, actually means.
Performance Metrics
Clinical Evidence
The pivotal evidence is Apple's clinical validation, reported in the September 2024 whitepaper "Estimating Breathing Disturbances and Sleep Apnea Risk from Apple Watch" and reviewed by the FDA in the K240929 clearance: 1,448 participants spanning apnoea-hypopnoea index values from normal to severe were assessed against home sleep apnoea testing (HSAT) as the ground truth — a minimum of two nights of recordings with a type 3 HSAT device; 1,499 were enrolled and 1,448 completed. Polysomnography was used only in the separate algorithm-development dataset, not in the validation. The algorithm's operating point is asymmetric by design. Weighted overall sensitivity for moderate-to-severe obstructive sleep apnoea was 66.3 per cent (95 per cent CI 62.2–70.3) against a weighted overall specificity of 98.5 per cent (98.0–99.0); sensitivity was 89.1 per cent for severe disease but roughly 43 per cent for moderate disease, while specificity in participants with normal sleep was reported at 100 per cent (99.7–100). Read clinically: a notification is a strong signal that warrants a proper sleep evaluation, and the absence of one excludes nothing — over half of moderate cases and about one in ten severe cases will never trigger it. Two caveats keep the evidence in frame. First, the validation is a manufacturer study reviewed by a regulator, not a peer-reviewed publication; no independent peer-reviewed validation of the shipped feature has been located as of this review. The nearest independent work — a 2025 pilot in Sleep and Breathing (PMID 39891814) that trained its own machine-learning models on Apple Watch inertial signals in 61 adults undergoing polysomnography — supports the plausibility of accelerometry-based apnoea detection (per-subject correlation with AHI of 0.93) but tested a custom research algorithm, not Apple's. Second, there is no outcome evidence: whether population-scale notification changes diagnosis rates, treatment uptake, or downstream cardiovascular outcomes is unstudied in the public literature.
| Study | Design | n | Sensitivity | Specificity | AUC | Published |
|---|---|---|---|---|---|---|
| Apple Inc. (SANF clinical validation) | ProspectiveProspective | 1,448 | 66.3% | 98.5% | — | Apple whitepaper, September 2024, reviewed in FDA 510(k) K240929; validation against type 3 home sleep apnoea testing (minimum two nights of HSAT as ground truth) across the AHI spectrum, 1,499 enrolled and 1,448 completed; sensitivity 89.1% for severe and ~43% for moderate OSA; not peer-reviewed |
Inside the algorithm
Editorial featureHow a watch screens for sleep apnoea without touching your breath.
Five stages — from raw input to verdict — drawn from manufacturer documentation and the public regulatory record.
- INGEST
- NORMALISE
- DETECT
- LOCALISE
- VERDICT
Stage 01 · INGEST
The only sensor is an accelerometer.
During sleep tracking the watch records its triaxial accelerometer — no airflow cannula, no oximeter, no microphone. The signal captures coarse body movement and, between movements, the fine rise and fall of the wrist that accompanies breathing. Everything downstream is inference from that motion trace, which is what makes the feature deployable on hardware people already wear.
Input
Wrist accelerometry during sleep
Inference
On device, nightly
Inside the Auris+ Listing
Five more sections complete this device’s Auris+ Listing.
Decision Ledger
ProPro unlocks a private, cross-vendor log of every case you read on this device — what the AI called, what you concluded, and a one-line reason if you overrode it. Ready for the EU AI Act's deployer logging obligations when they land in 2028.
Clinical Evidence Deep Dive
ProPro unlocks the structured clinical-evidence summary — study count, target patient population, and a tabular accuracy-metrics view drawn from peer-reviewed sources.
Peer-Reviewed Publications
ProPro unlocks the curated peer-reviewed publication list with PubMed cross-links — the citation backbone of every editorial verdict.
Post-Market & Regulatory Conditions
ProPro unlocks the post-market surveillance summary, recall record, and the conditions of approval that bound real-world use.
AI Algorithm Version History
ProPro unlocks the chronological record of algorithm version changes — what changed when, drawn from manufacturer changelogs and regulatory filings.
Regulatory Approvals
Safety Record
No recalls, FDA safety communications, or adverse-event reports attributable to the Sleep Apnea Notification Feature were identified in public sources as of August 2026 — read as none found rather than an exhaustive audit, since the FDA recall and MAUDE databases could not be queried directly from this environment. The feature's principal clinical risk is intrinsic to its operating point: false reassurance. With sensitivity around 43 per cent for moderate disease, a wearer who never receives a notification may wrongly conclude they have been screened clear. The indication's exclusions carry the same weight — the feature is not for users with a prior diagnosis, does not monitor treatment, and is no substitute for clinical testing when symptoms are present.
Intended Use & Indications
The Sleep Apnea Notification Feature is software-only and runs across Apple Watch and iPhone for users aged 18 and over who have not been diagnosed with sleep apnoea. During sleep tracking the watch computes a nightly Breathing Disturbances metric from accelerometer signals and classifies each night as elevated or not elevated. Approximately every 30 days the feature evaluates the accumulated nights; if elevated disturbances recur often enough to be consistent with moderate-to-severe obstructive sleep apnoea, the user receives a notification summarising the pattern, with supporting data exportable for a clinical consultation. The feature is explicitly not intended to diagnose, treat, or manage sleep apnoea, is not a substitute for a sleep study, and does not monitor for acute events in real time. Users with an existing diagnosis are outside its indication.