A newly published real-world validation study found that an AI-enhanced smartwatch algorithm can estimate the apnea-hypopnea index, the key metric used to diagnose obstructive sleep apnea, with 92.3% sensitivity and 92.6% specificity for moderate-to-severe cases, matching much of the accuracy of AI-scored clinical polysomnography. The research adds fresh evidence that consumer wearables are closing in on medical-grade diagnostic tools for one of the most underdiagnosed conditions in medicine.
Why Sleep Apnea Screening Has Been So Hard
Obstructive sleep apnea affects an estimated hundreds of millions of people worldwide, causing repeated breathing interruptions during sleep that are linked to hypertension, heart disease, stroke and daytime fatigue. Yet diagnosis has traditionally required an overnight stay in a sleep lab hooked up to a polysomnography machine, an expensive, inconvenient process with long waitlists in many health systems. That barrier has left a large share of people with sleep apnea undiagnosed for years, sometimes decades.
How the Study Was Designed
Researchers enrolled 90 adults in Korea who underwent simultaneous Level 1 polysomnography, the clinical gold standard, and smartwatch recording on the same night. Of those, 53 datasets with at least three hours of valid watch data were analyzed. The smartwatch’s AI-derived apnea-hypopnea index, which the researchers call eAHI, was compared against both AI-scored and expert-scored polysomnography readings. For detecting moderate-to-severe cases, defined as an index of 15 or more events per hour, the smartwatch algorithm achieved 92.3% sensitivity, 92.6% specificity and 92.5% overall accuracy, with a strong statistical correlation to both AI- and expert-scored clinical results.
Where the Technology Still Falls Short
The study’s authors were careful to flag a significant limitation: Bland-Altman analysis, a statistical method for comparing two measurement techniques, revealed that the smartwatch systematically underestimated the true apnea-hypopnea index, particularly in patients with mild sleep apnea. That means while the device performed strongly for catching more severe cases, it may miss or downplay milder cases that still warrant clinical attention, a gap researchers said needs further algorithm refinement before the technology could be relied upon as a standalone diagnostic tool.
What Clinicians Are Saying
Sleep medicine specialists have generally described the results as encouraging evidence that wearables could serve as an accessible first-line screening tool, helping identify patients who should be prioritized for a formal sleep study rather than replacing polysomnography altogether. Given how few people currently get screened for sleep apnea at all, proponents argue that even an imperfect but scalable screening tool represents a meaningful public health improvement. Skeptics counter that a tool with known blind spots for mild cases could create a false sense of security among wearers whose apnea nonetheless requires treatment.
Part of a Wider Wearable Health Trend
The sleep apnea findings fit into a broader 2026 pattern of wearables pushing into clinical-grade diagnostics, alongside Stanford’s SleepFM model, which predicts risk for more than 130 diseases from a single night’s sleep data, and neurofeedback devices like the Bía smart sleep mask that actively intervene to improve sleep quality. Collectively, these developments suggest smartwatches and other consumer sleep devices are shifting from passive fitness trackers toward genuine, if still imperfect, medical screening tools.
What’s Next
Researchers say larger, more diverse validation studies across different populations and smartwatch models are needed before health systems formally incorporate wearable-based apnea screening into clinical pathways. If further refinement addresses the mild-case underestimation problem, sleep medicine experts say wearable screening could eventually help funnel far more undiagnosed patients toward proper sleep apnea treatment than the current lab-based system manages today.
Comparing Watch-Based Screening to the Gold Standard
Sleep medicine specialists emphasize that while a 92.5% overall accuracy rate sounds impressive, it should be understood in context: polysomnography remains far more comprehensive, tracking dozens of physiological signals simultaneously rather than the more limited sensor set available on a consumer smartwatch. Even so, given that only a small fraction of people with sleep apnea ever undergo a formal sleep study, researchers argue that a screening tool people already wear every day, even an imperfect one, could dramatically expand how many undiagnosed cases get identified and referred for proper evaluation and treatment in the first place. Sleep medicine researchers say the next generation of these algorithms will need to be trained on larger and more demographically diverse populations than the 90-adult Korean cohort used in this validation study, since apnea presentation and severity can vary meaningfully across different body types, ages and ethnic backgrounds.
