Sleep trackers: a bedtime clock vs a score you lie awake to beat

The fight

There is a useful, boring job a wearable can do: mark when you got into bed and when you got up, and flag the nights you drank. That is the same job the CDC already assigns to a paper sleep diary. There is a different job the marketing sells: a proprietary “sleep score,” a pie chart of light / deep / REM, and the quiet implication that last night has been graded. Those are not the same product. One is a clock. The other is a costume of a lab.

I am not in a position to diagnose you. I can tell you what the opened papers and the labelling actually measured, and what they refused to claim. If you already own a ring or a watch, the honest use is the clock. If you are lying awake to beat last night, the device is no longer doing the clock’s job.

What the watch is actually measuring

The Food and Drug Administration’s general-wellness policy — issued 6 January 2026, superseding the 2019 note — puts “track sleep trends” in the wellness bucket. A product stays there when it is intended only for general wellness and is low risk. Software that encourages a healthy lifestyle and is unrelated to diagnosis, cure, mitigation, prevention or treatment of a disease is not a device under the FD&C Act §520(o)(1)(B). The moment the same gadget is intended to measure or report physiologic values for screening, diagnosis, monitoring, alerting or management of a disease, it is out. Naming a disease, a clinical threshold, treatment guidance, or “clinical grade” takes it out. Inclusion under the policy “does not establish that it has been shown to be safe and/or effective.” Wellness is a labelling category. It is not a validation stamp.

The American Academy of Sleep Medicine said the clinical half of that out loud in 2018. Consumer sleep technologies “purport to measure and even improve sleep.” Given the lack of validation against gold-standard polysomnography and the lack of FDA clearance, “CSTs cannot be utilized for the diagnosis and/or treatment of sleep disorders at this time.” They “may be utilized to enhance the patient-clinician interaction when presented in the context of an appropriate clinical evaluation.” President Ilene Rosen: consumer sleep devices currently are unable to diagnose sleep disorders. People with ongoing sleep problems or daytime sleepiness “should discuss this … with a licensed medical provider, regardless of what their wearable … tells them.” Data “should not replace validated diagnostic testing.” That statement predates the later Apple Watch apnea-risk notification clearance. It still describes every wellness score I opened.

Against overnight polysomnography, the devices are decent at one binary question — asleep versus awake — and they generally over-call sleep. de Zambotti and colleagues, reviewing the PSG comparison studies in 2019, put it as high sensitivity (above 90%) for detecting sleep and lower specificity for detecting wake: they overestimate total sleep time and underestimate wake after sleep onset, in line with actigraphy. Second-generation multisensor devices showed relatively higher agreement for PSG “light” (N1+N2) and REM, in the studies they tabulated around 60–75%, than for wake and N3. Firmware can change without notice. A validation of model X may post-date that model. “Insufficient evidence to consider consumer wearables as a potential stand-alone diagnostic tool for sleep disorders.” “No currently available consumer wearable devices are suitable for diagnosing OSA.”

The rows in that table are old hardware, and the ages are the wrong ages. ŌURA ring versus PSG, de Zambotti 2017, 42 healthy adolescents, PSG sleep efficiency over 90%: sensitivity 0.96, specificity 0.48; epoch agreement N1+N2 0.65, N3 0.51, REM 0.61; underestimated PSG N3 by 19.6 ± 41.2 minutes. Fitbit Charge 2, de Zambotti 2018, 44 healthy adults aged 19–61: sensitivity 0.96, specificity 0.61; agreement N1+N2 0.81, N3 0.49, REM 0.74; overestimated total sleep time 9 ± 24 minutes and N1+N2 34 ± 34 minutes; underestimated N3 24 ± 28 minutes. Greater nocturnal wake tends to widen the bias. That is exactly when men in this age band start collecting more wake. Chinoy 2021 ran seven consumer devices against PSG in 34 healthy young adults — 22 women, mean age 28.1 ± 3.9 years — across three lab nights including a disrupted-sleep night. Epoch-by-epoch sensitivity was all ≥0.93; specificity was low-to-medium, 0.18–0.54. Sleep-stage comparisons were mixed. Devices tended to perform worse on nights with poorer or disrupted sleep. Most of the set (Readiband, Fitbit Alta HR, EarlySense Live, ResMed S+, SleepScore Max) performed as well as or better than actigraphy on sleep/wake; the two Garmins performed worse. Chinoy’s sample is not men 45–65 with insomnia or apnea. Neither paper validates a proprietary 0–100 “sleep score.” Citing those numbers as “your 2026 ring is X% accurate” would be invention, and I am not going to invent it.

The score is not a sleep study

Oura, on its own safety page updated 19 June 2026: “Oura Ring is not a medical device and is not intended to diagnose, treat, cure, monitor, or prevent medical conditions or illnesses.” The Sleep Score is 0–100 with in-app bands — 85+ “optimal,” 70–84 “good,” under 70 “off” — built from seven proprietary contributors: total sleep, efficiency, restfulness, REM, deep, latency, timing. That sentence is on a company blog. It is not a cleared diagnostic. WHOOP: except as otherwise expressly indicated, the services and the device are not a medical device; content is “for general fitness and informational purposes only”; heart rate, sleep analytics and other biometric indicators “are not guaranteed to be error-free or accurate”; intended solely for personal wellness tracking and not for making any medical decisions. Fitbit / Google Health: unless otherwise specified, not medical devices, not intended to diagnose, treat, cure or prevent any disease; accuracy “is not intended to be equivalent to medical devices or scientific measurement devices.” Google’s vitals help: well-being information, not for medical purposes. Apple’s Health intro: “iPhone, iPad, and Apple Watch aren’t medical devices.” The Sleep app “can estimate the time you spent in each sleep stage—REM, Core, and Deep.” Sleep score is based on estimates of how long you slept, the consistency of when you fell asleep relative to recent history, and the amount and duration of awake periods. Respiratory rate: “not intended for medical use.” Apple’s own September 2023 stages paper: validation median binary sleep sensitivity 97.9%, specificity 75.0%; four-stage kappa 0.63 (166 subjects, 299 nights, mean age 47). In a clinical PSG cohort of 236 patients, kappa 0.55. Explicit: “Apple Watch sleep stages aren’t intended for clinical use.” Conclusion frames stages as a device for wellness. Mean age 47 is closer to this readership than Chinoy’s 28. Kappa 0.55 against clinical PSG is not a lab.

The 2018 AASM line still holds for those scores. It is incomplete for one later clearance I will come to. Do not write the score as if AASM has blessed it. They have not.

There is a name for what happens when the score becomes the night. Baron, Abbott, Jao, Manalo and Mullen coined “orthosomnia” in 2017 for a preoccupation with perfecting wearable sleep data. Three clinic cases. Not a prevalence study. Not an RCT that “trackers cause insomnia.” Mr. R, 40: felt pressure so the tracker would show at least eight hours; goal “at least 8 h of deep sleep each night”; symptoms “only” on days the tracker showed under eight hours. Ms. B, 27: Fitbit sleep efficiency around 60%; after in-lab PSG with no sleep-disordered breathing (AHI/RDI 0.7) and increased N3, she asked, “Then why does my Fitbit say I am sleeping poorly?” Belief unchanged. Mr. S, 69: tracker “restless” sleep led to PSG, severe OSA (AHI 54), CPAP — and he still chased tracker quality. All three were extending time in bed to raise tracker duration, which “may have exacerbated their insomnia.” The authors note consumer devices are unable to accurately discriminate stages and have poor accuracy detecting wake after sleep onset. Trackers “tend to overestimate sleep,” which “may have served to reinforce poor sleep habits.” Limitation they state themselves: they do not know sleep before the tracker, so they cannot prove the device caused the problem. Their advice to clinicians is the one I will repeat: use the tracker as a pattern — time in bed — not as minutes of light and deep. Pair it with a written log. If the device itself is the stressor, take it off.

You do not need a gadget for the two habits the public-health pages actually name. CDC, 15 May 2024: go to bed and get up at the same time every day; avoid large meals and alcohol before bedtime; also caffeine later in the day, a cool dark room, screens off 30 minutes before bed. Adults 18–60: at least 7 hours; 61–64: 7–9 hours. For suspected sleep disorders, “Your provider can run tests, including sleep studies.” The suggested self-record is a sleep diary — bed, night wakes, alcohol/caffeine, meds — not a consumer score. NIAAA: “Even moderate doses of alcohol may alter the physiology of sleep, for instance by reducing the duration of rapid eye movement sleep.” Alcohol “may aggravate sleep-disordered breathing” and periodic limb movements. These alterations “may be subacute or chronic, recovering only after 30 or more days of abstinence.” Using alcohol as a sleep aid is a pathway into worse sleep. A higher score does not prove the habit worked. CDC and NIAAA are not trials of Oura versus no Oura.

Apnea: a notification is not a diagnosis

A wellness ring or watch cannot diagnose obstructive sleep apnea. de Zambotti 2019: CMS/AASM diagnostic equipment types for OSA are polysomnography or specified home tests with airflow, effort and oximetry reviewed by a sleep physician. Consumer wearables were not among them.

One opened product is in a different regulatory bucket, and even that bucket is not a diagnosis. Apple’s Sleep Apnea Notification Feature (SANF) is 510(k) K240929, decision 13 September 2024, Class II, 21 CFR 868.2378, product code QZW. Indications for use: a software-only mobile medical application that analyzes Apple Watch sensor data to identify patterns of breathing disturbances suggestive of moderate-to-severe sleep apnea and provides a notification. Over-the- counter, adults 18 and over who have not previously received a sleep apnea diagnosis. It “is not intended to diagnose, treat, or aid in the management of sleep apnea. The absence of a notification is not intended to indicate the absence of sleep apnea.” Input is the accelerometer over discrete 30-day windows — not a nightly AHI. The intended use of the class: not a standalone diagnosis, not a replacement for traditional methods (e.g. polysomnography), not to assist clinicians in diagnosing sleep disorders, not an apnea monitor. Clinical study: 1,499 enrolled, 1,278 in the notification analysis, versus Nox T3s home sleep apnea testing. Notification sensitivity 66.3% (95% CI 62.2–70.3) for AHI ≥15; specificity 98.5% (98.0–99.0) for AHI <15; “SANF did not falsely notify any subjects with normal AHI (AHI <5).” Sensitivity of 66.3% is high specificity bought with missed cases: about one in three moderate-to-severe in that study will not be notified. Apple Support: “Your Apple Watch is not a medical device and is not intended to replace talking with your doctor for diagnosis, treatment, or monitoring of sleep apnea.” Not for already- diagnosed apnea or age under 18. The SANF clinical study did enrol men in this age band: 50–64 was 32.8% of the sample, 65 and over 10.2% — closer than Chinoy’s 28-year-olds, and still not a diagnosis. I did not open a 510(k) for Oura, Whoop or Fitbit on this feature. A quiet ring is not a clean bill.

Do you need to buy anything

No. A paper diary does the job the evidence actually supports: bed, wake, alcohol nights, caffeine, meds. If you already own a watch or a ring, use it as that diary. Treat the number as a weather report, not a lab. If the device increases pre-sleep checking, time-in-bed stretching, or arguments with a partner about “deep sleep,” take it off. That is Baron’s own clinical advice, not a wellness slogan.

I am not going to sell you a proprietary sleep score as “your sleep study.” I am not going to imply a ring diagnoses or rules out apnea, insomnia, or “low testosterone sleep.” I am not going to quote a 2026 accuracy percentage I do not have. If you want a gadget, buy it as a schedule cue — the same humility as a cheap alarm clock — and with a hard line that the score is not care. There is no opened RCT that giving a 50-year-old man a ring improves sleep versus a paper diary plus the CDC habits.

When this is not the article

Daytime sleepiness. Loud snoring. Gasping or witnessed apneas. Resistant hypertension. A tracker “breathing” flag. Ongoing sleep problems regardless of what the wearable tells you. That list is a reason to talk to a licensed medical provider, and if they want a test, the test is a sleep study — HSAT or PSG — not a pie chart. Only Apple SANF among the opened products is cleared even to notify possible moderate-to-severe apnea, and it still is not a diagnosis, and no notification is not a clearance.

That is not me diagnosing you. It is the short list of things that deserve a professional look rather than a score. A GP or a sleep clinician will get further with it in ten minutes than I can here.