Samsung Smartwatch Detected Moderate-to-Severe Sleep Apnea With 94% Sensitivity

The study appeared in the Journal of Clinical Sleep Medicine. It is a prospective observational study in people: researchers followed participants forward in time and compared a device with a reference test, without assigning any treatment. They enrolled 152 adults aged 22 or older. All had either a prior diagnosis of moderate-to-severe obstructive sleep apnea or a high chance of having it, based on a screening score called STOP-Bang of 3 or more. In total, 147 completed both overnight sleep studies.

Obstructive sleep apnea is a condition in which breathing repeatedly stops or narrows during sleep. According to the study authors, it is linked to long-term problems such as heart and brain disease, yet many cases go undiagnosed. The standard test is an overnight sleep study in a lab, called polysomnography. It is costly and hard to get.

Each participant wore a Samsung Galaxy Watch during an initial lab night, then for at least three nights at home with the watch alone, then during a second lab night. The lab nights totalled 1,850 hours of sleep. The watch's estimate was compared with the lab result. Moderate-to-severe apnea was defined as 15 or more breathing disruptions per hour of sleep.

The watch's overall ability to separate those with and without moderate-to-severe apnea scored 0.94 on a scale where 1.0 is perfect (95% confidence interval 0.889– 0.980). At its default threshold, sensitivity was 94.1%. That means it caught most true cases. Specificity was 66.7%. That means roughly a third of people without moderate-to-severe apnea were still flagged. These are the false alarms.

The researchers also tested a threshold tuned to this group. Sensitivity fell to 82.4% and specificity rose to 94.9%. So the watch caught fewer cases but raised fewer false alarms. Because that setting was optimized on this same cohort, it may not perform the same way elsewhere. That is our reading, not a claim in the study.

The team also looked at hypoxic burden. This measure combines how deep, how long and how often oxygen levels drop during sleep. The authors say it may reflect heart and metabolic risk better than the usual breathing-disruption count. Among people classed as high-risk by this measure, the watch reached 100% sensitivity and 100% specificity at the default threshold.

The study does not show the watch can diagnose sleep apnea, and the authors do not claim it can. Participants were people already diagnosed or likely to have the condition, not a general group of wearers. In a broader population, the balance of correct and false alerts could differ.

The American Academy of Sleep Medicine's press coverage said the watch "correctly distinguished" cases 94% of the time. The study reports a sensitivity of 94.1% and a separate overall score of 0.94. Those are different measures, and neither is a plain "94% accurate."

Principal investigator Dr. Robson Capasso of Stanford University is quoted in that coverage: "These devices are not diagnostic, but by bringing validated screening into something people already wear every night, we may be able to identify risk earlier and shorten the path to evaluation and to low-risk, well-established therapies such as positive airway pressure and weight management." The authors say wearables like this may help guide people toward formal diagnostic testing, which fits our reading that a watch alert is a prompt to seek that testing, not a diagnosis.

The coverage also discloses that Samsung provided financial support and materials. It adds that Capasso serves on the scientific advisory board of Restera Medical.

What to watch next

Studies that test the watch in people without a prior diagnosis, and that report how many false alarms lead to unneeded testing.