Smartwatch health alerts are better at warning you than reassuring you

Key takeaways

• In the Apple and Fitbit heart studies, which together enrolled nearly 875,000 people, only 1 in 200 to 1 in 100 got an irregular-rhythm alert, and about a third of those who wore a follow-up heart monitor had atrial fibrillation¹².
• Cleared watch alerts are tuned to avoid false alarms, so they miss many cases: Apple’s high blood pressure feature flagged 41% of people with hypertension in its validation study⁵.
• A quiet watch is not a clean bill of health: a 2026 analysis in JAMA estimated that about 1 in 3 adults 60 and older with no alert would still have high blood pressure⁶.
• Fall detection meets a real need, since 80% of people over 90 who fell in a British study couldn’t get up alone at least once, but the best real-world test we found used older waist-worn sensors, which caught just 57% of real falls¹⁴¹⁵.
• After any alert, see a clinician, don’t change a medicine on your own, and call 911 for chest pain or stroke signs, because a watch ECG can’t detect a heart attack¹⁰¹³.

In September 2025, the Food and Drug Administration cleared an Apple Watch feature that looks for signs of high blood pressure⁴. It uses no cuff, reading data from the watch’s light sensor in the background and sending a notice when the pattern suggests hypertension⁵.

In the company’s validation study, filed with the FDA, 1,863 adults had enough data to analyze. Of the people who had high blood pressure, 41% got the alert. The other 59% got nothing⁵.

That isn’t a scandal but a design choice. The feature was built so that people with normal blood pressure rarely get a false alarm, and 92% of them got no alert⁵. The trade-off is that a silent watch tells you very little.

This pattern shows up across the newest medical alerts on smartwatches. When a cleared alert fires, it’s often right. When it stays quiet, you’ve learned less than you might think.

For people who like to keep watch over their own health, that matters. A watch can be an early-warning helper, but it can’t be your checkup. Below, we look at what the big studies found, what each alert can do, and how to respond when your wrist buzzes.

A few watch features are medical devices, but the rest are wellness estimates

A smartwatch is really two products in one case. A handful of features have been reviewed by the FDA for one specific medical job. Everything else, from step counts to stress scores, is sold as a wellness estimate. Our guide “FDA Registered” Isn’t “FDA Approved” explains what each of those labels means.

The FDA keeps a public list of cleared sensor features. Here are the main ones you’ll find on watches today⁴.

Feature	What it does	Examples on the FDA list
ECG app	Records a one-lead heart tracing and sorts it as AFib or normal rhythm	Apple (2018), Fitbit and Samsung (2020), Garmin (2023), WHOOP (2025)4
Irregular rhythm notifications	Checks your pulse in the background and alerts you to a pattern that suggests AFib	Apple (versions cleared in 2021 and 2023), Fitbit (2022), Samsung (2023)4
Sleep apnea notification	Looks for breathing disturbances during sleep and flags possible moderate to severe sleep apnea	Samsung (February 2024), Apple (September 2024)48
Loss of pulse detection	If your pulse seems to stop and you don’t respond, it tries to prompt an emergency call	Fitbit (February 2025)49
Hypertension notification	Flags patterns that suggest undiagnosed high blood pressure, with no blood pressure numbers shown	Apple (September 2025)457

The FDA says this list is not complete⁴. Even so, it’s telling that it includes no fall detection feature and no blood-oxygen feature from the major watch brands⁴. Those features may be useful, but don’t assume they’ve had the same review for a medical claim.

Clearance also comes with fine print. Most of these features are for adults 22 and older, while Apple’s sleep apnea feature starts at 18⁵⁸⁹¹⁰.

The hypertension feature isn’t meant for people already diagnosed with high blood pressure or for use during pregnancy⁵. The ECG app isn’t meant to judge rhythm problems other than AFib¹⁰.

The two giant heart studies: alerts were rare, and usually right

Atrial fibrillation, or AFib, is an irregular and often fast rhythm in the heart’s upper chambers. The Centers for Disease Control and Prevention says it raises the risk of ischemic stroke, the kind caused by a blockage, about fivefold and causes about 1 in 7 strokes¹². Some people with AFib have no symptoms at all¹². That’s the promise of a wrist alert: catching a silent problem early.

Two studies tested that promise at enormous scale.

The apple heart study. Stanford researchers enrolled 419,297 people over eight months. Only 0.52% got an irregular pulse notification, and age made a big difference¹. About 3.1% of people 65 and older got one, compared with 0.16% of those ages 22 to 39¹.

People who got an alert were offered an ECG patch, a stick-on heart monitor. Only 450 sent back a usable patch. AFib showed up in 34% of them, but in just 18% of those under 40¹. When an alert went off while the patch was on, the patch confirmed AFib 84% of the time¹.

The Fitbit heart study. Researchers led by Steven Lubitz enrolled 455,699 people in 2020². About 1% got an alert, rising to 4% of people 65 and older. Among 1,057 who wore a patch afterward, 32% had AFib². When 225 people got another alert while wearing the patch, 221 of them were in AFib at that moment, or 98%².

Lubitz said most undiagnosed AFib episodes happened during sleep, so the device works best if you wear it to bed³.

Both studies had company sponsors. Apple paid for its study and owns the data, though Stanford statisticians ran the analysis on their own¹. Fitbit sponsored its study³.

Neither study compared people who got alerts with similar people who didn’t. So neither can tell us whether the alerts prevent strokes¹³. The Apple team also noted that its algorithm wasn’t built to catch short AFib episodes¹.

Built to avoid false alarms, so silence proves little

Sensitivity is the share of people with a condition who get flagged. Specificity is the share of people without it who correctly get no alert.

Newer watch alerts lean hard toward specificity, so they rarely cry wolf. The cost is that they miss a lot. This is what the companies reported to the FDA.

Alert	Share of true cases flagged	Share of healthy people correctly left alone	What the FDA filing says
High blood pressure (Apple, 2025)	41%	92%	No alert doesn’t mean you don’t have hypertension5
Moderate to severe sleep apnea (Apple, 2024)	66%	98.5%	No alert doesn’t mean you don’t have sleep apnea8
Loss of pulse (Fitbit, 2025)	69%	99.97% of days	It isn’t meant to catch every event9

The blood pressure feature did worse in some groups. It caught fewer cases in adults under 60 and in people with a body mass index of 30 or less⁵.

In February 2026, researchers writing in the journal JAMA put these numbers to work. They applied the FDA figures to national health survey data that represent about 127 million U.S. adults who haven’t been told they have high blood pressure⁶. Among people who would get an alert, about 69% would truly have hypertension⁶.

The worrying part was the silence. With no alert, the chance of still having high blood pressure was 10% for adults under 30 and 34% for those 60 and older⁶. The authors warned that false reassurance could keep some people from getting checked⁶.

Houston Methodist, a hospital system in Houston, adds a practical point: the watch shows no blood pressure numbers and can’t do a spot check on demand⁷. Eleonora Avenatti, a cardiac prevention specialist there, says an alert should lead to a reading on a medical-grade device and a talk with a physician⁷. So the feature can’t replace a cuff. It can only nudge you to use one.

Fall detection meets a real need, but the proof is still thin

A fall is dangerous, but being stuck on the floor afterward can be worse.

For a year, British researchers tracked 110 Cambridge residents over age 90¹⁴. Six in ten had at least one fall. Among the fallers, 80% had a fall they couldn’t get up from alone, and 30% waited on the floor for an hour or more at least once.

Call alarms were common in this group, yet in most of those long waits nobody pressed one. The long waits went along with serious injuries, hospital admissions and moves into care homes¹⁴.

That’s the case for automatic fall detection: you don’t have to press anything.

The catch is that lab results can flatter these systems. In a 2012 study, an international team tested 13 published fall-detection methods on 29 real falls recorded by body-worn sensors¹⁵. Most of the falls came from patients with a rare neurological disease and a high risk of falling.

On average the methods caught 57% of the falls, far below the near-perfect results often reported on falls staged by healthy volunteers¹⁵. Over a day of monitoring, they also raised 3 to 85 false alarms, depending on the method¹⁵.

Those sensors sat on the waist or lower back, not the wrist, and they ran research methods from 2012 or earlier, so the numbers don’t describe any watch sold today. We found no independent real-world test of watch fall detection to take their place. The lesson carries over, though. A feature that works in a demonstration may miss a slow slide out of a chair.

Treat watch fall detection as a helpful backup, not a guarantee. If you live alone, pair it with a daily check-in call or a neighbor who knows your routine.

When the alert itself becomes the problem

Alerts can also cause harm of a quieter kind.

In 2024, researchers at the University of North Carolina compared 172 people with AFib, average age 73. About half used a wearable¹¹. Over nine months, wearable users were more preoccupied with their heart symptoms and more worried about their AFib treatment¹¹.

One in five users felt anxious after irregular rhythm alerts and always contacted their doctor¹¹. Users also had more AFib-related tests and procedures, and they called and messaged their clinics more¹¹. The study can’t say whether those visits helped or were needed, and the authors called for randomized trials to weigh the net effect¹¹.

The Apple study saw the same pull toward care. Of notified people who answered a survey, 57% contacted an outside health provider¹.

Some of that care is good, and some of it may be worry turned into appointments. If alerts leave you checking your wrist all day, that’s a reason to talk with your clinician about which ones you need.

What to do when your watch sends an alert

Decide your plan before the alert, not during it, and review this starting plan with your clinician.

If your watch says	Do this soon	Call 911 instead if
Irregular rhythm or possible AFib	Note the time and how you feel. If your watch has an ECG app, take a reading and save it. Book a visit and bring the record. In both big studies, follow-up meant a stick-on ECG patch12.	You have chest pain or pressure, or any sign of stroke1013; also fainting or severe shortness of breath
Possible high blood pressure	Check with a validated, medical-grade cuff and share the readings with a clinician7	You have chest pain or stroke signs13
Possible sleep apnea	Bring the alert to your clinician. The notice is not a diagnosis or a replacement for a sleep study8.	You have severe trouble breathing
Loss of pulse or a hard fall	If you’re fine, cancel the call. If someone else is unresponsive, call 911 yourself.	Anyone is unresponsive, badly hurt or can’t get up

Three rules apply to every alert. Don’t stop or change a medicine because of a watch reading¹⁰. Don’t let a normal-looking watch ECG talk you out of care, because the FDA label for Apple’s ECG app says plainly, “It cannot detect heart attacks”¹⁰. And learn the stroke signs: sudden face drooping, arm weakness, trouble speaking, trouble seeing or loss of balance mean call 911 and note the time¹³.

What the evidence supports, feature by feature

How the research stacks up, using four levels: strong evidence, promising but early, not enough evidence, and evidence says no.

Claim	Rating	Why
Irregular rhythm alerts usually reflect real AFib when they fire	Strong evidence	Two studies of more than 400,000 people each12
Getting AFib alerts prevents strokes or deaths	Not enough evidence	The big studies had no comparison group; researchers have called for randomized trials111
A hypertension notification finds undiagnosed high blood pressure	Promising but early	Caught 41% in the maker’s study; real-world use not yet tested56
A sleep apnea notification finds moderate to severe apnea	Promising but early	Caught 66% in the maker’s study, with very few false alarms8
Watch fall detection reliably catches real-life falls	Not enough evidence	No independent real-world watch test found; older waist-worn detectors caught 57% of real falls15
No alert means you’re healthy	Evidence says no, or unsafe	Cleared features miss a third to more than half of true cases568
A watch ECG can spot a heart attack	Evidence says no, or unsafe	The FDA label says it can’t10

Your body’s signals still come first

Many people buy a smartwatch because they believe in prevention and want to know their own bodies. That instinct is sound. The watch is best used to support it, not replace it.

The CDC lists the habits and conditions tied to AFib: high blood pressure, which is linked to about 1 in 5 cases, along with obesity, diabetes, moderate to heavy drinking and smoking¹². Most of those respond to daily choices about food, movement, alcohol and tobacco. A watch can’t make those choices. It can remind you why they matter.

Your own sensations count, too. Fatigue, a fluttering or pounding heartbeat, lightheadedness and shortness of breath are AFib symptoms worth reporting, with or without an alert¹².

The same skepticism readers bring to drug-company research belongs here. Hold the marketing to the evidence, and the evidence so far shows watches catching problems, not improving health.

Questions readers ask

The bottom line

A smartwatch alert deserves attention. When a cleared AFib alert fires, it’s often right, and a clinician can confirm it with a proper heart recording. When the watch stays quiet, though, it hasn’t cleared you of anything. The newest alerts are built to avoid false alarms, so they miss many real cases.

Use the watch as a backup to habits and checkups, not a stand-in for them. Make an alert plan ahead of time, keep your cuff and your clinician in the loop, and let chest pain or stroke signs send you to 911 no matter what your wrist says.

Keep exploring: digital therapeutics, interpreting wearable data, fitness tracker accuracy and wellness devices FDA.

Related reading

• “FDA registered” isn’t “FDA approved”: a buyer’s map to wellness devices
• Your wearable gets resting pulse right, while calories and sleep are another story
• The symptoms that mean call 911 now
• What a medical test can and cannot tell you
• When extra medical tests do more harm than good

Sources

1. Perez MV, Mahaffey KW, Hedlin H, et al. “Large-Scale Assessment of a Smartwatch to Identify Atrial Fibrillation.” New England Journal of Medicine, 2019 (full text via eScholarship). escholarship.org
2. Lubitz SA, Faranesh AZ, Selvaggi C, et al. “Detection of Atrial Fibrillation in a Large Population Using Wearable Devices: The Fitbit Heart Study.” Circulation, 2022 (abstract via UMass Chan repository). pubmed.ncbi.nlm.nih.gov
3. HCPLive. “Fitbit Heart Study Provides Further Backing for Wearables in Detecting AFib.” November 15, 2021. hcplive.com
4. U.S. Food and Drug Administration. “Medical Devices that Incorporate Sensor-based Digital Health Technology.” Updated September 2026. fda.gov
5. U.S. Food and Drug Administration. 510(k) Summary K250507, Hypertension Notification Feature (Apple Inc.). September 2025. accessdata.fda.gov
6. Cohen JB, Addo DK, Jacobs JA, et al. “Impact of a Smartwatch Hypertension Notification Feature for Population Screening.” JAMA, February 2026. pubmed.ncbi.nlm.nih.gov
7. Houston Methodist. “Smartwatch Hypertension Notifications: How Seriously Should You Take It?” On Health blog, May 12, 2026. houstonmethodist.org
8. U.S. Food and Drug Administration. 510(k) Summary K240929, Sleep Apnea Notification Feature (Apple Inc.). September 2024. accessdata.fda.gov
9. U.S. Food and Drug Administration. 510(k) Summary K242967, Loss of Pulse Detection (Fitbit). February 2025. accessdata.fda.gov
10. U.S. Food and Drug Administration. De Novo Decision Summary DEN180044, ECG App (Apple Inc.). 2018. accessdata.fda.gov
11. Rosman L, Lampert R, Zhuo S, et al. “Wearable Devices, Health Care Use, and Psychological Well-Being in Patients With Atrial Fibrillation.” Journal of the American Heart Association, 2024 (abstract via DOAJ). pubmed.ncbi.nlm.nih.gov
12. Centers for Disease Control and Prevention. “About Atrial Fibrillation.” 2024. cdc.gov
13. Centers for Disease Control and Prevention. “Stroke Signs and Symptoms.” 2026. cdc.gov
14. Fleming J, Brayne C. “Inability to Get Up After Falling, Subsequent Time on Floor, and Summoning Help: Prospective Cohort Study in People Over 90.” BMJ, 2008. pubmed.ncbi.nlm.nih.gov
15. Bagalà F, Becker C, Cappello A, et al. “Evaluation of Accelerometer-Based Fall Detection Algorithms on Real-World Falls.” PLOS ONE, 2012. journals.plos.org