Smartwatch AI detects hypertension as well as office checks in Nature study

A study published in Nature on 5 November 2026 reports that an AI model reading signals from a widely available consumer smartwatch, with no cuff calibration, detected hypertension about as accurately as the office blood pressure check that usually starts a diagnosis¹. The model, called AI-PPG-ACC-HTN, works on two waveforms the watch already records: photoplethysmography, the optical pulse signal, and accelerometry¹.

What the study found

The researchers validated the model prospectively at several centres in 196 participants of diverse backgrounds. None had known cardiovascular disease and none were taking antihypertensive drugs¹. The reference standard was 24-hour ambulatory blood pressure monitoring, the accepted gold standard¹.

Across seven days of everyday wear, the smartwatch model picked up hypertension with a sensitivity of 65.8% (95% CI, 54.0%-76.3%), a specificity of 90.0% (95% CI, 83.2%-94.7%) and a positive predictive value of 80.6% (95% CI, 68.6%-89.6%)¹. Initial office blood pressure screening in the same group reached a sensitivity of 55.3% (95% CI, 43.4%-66.7%), the same 90.0% specificity, and a positive predictive value of 77.8% (95% CI, 64.4%-88.0%)¹.

Put simply, the watch found a larger share of people with hypertension than the office reading did, without raising the rate of false alarms. The confidence intervals overlap, however, so the study supports the authors' description of accuracy comparable to cuffed devices rather than a clear advantage¹.

The authors frame the problem in terms of awareness: more than half of adults with hypertension do not know they have it, and early treatment matters for preventing major cardiovascular events¹. They position the watch as a prompt that sends people at risk for clinical assessment, with diagnosis still confirmed by established cuff-based methods¹.

The study has clear limits. It enrolled 196 people, excluded anyone with known heart disease or already on treatment, and one in three people with hypertension were still missed by the watch¹. The abstract does not name the device maker or the countries involved.

What it means for the market

For health systems, the attraction is reach without new hardware. Hypertension case-finding currently depends on people turning up for a cuff reading, and this approach uses sensors already worn on millions of wrists¹. Primary care leaders and commissioners running cardiovascular prevention programmes will see a possible route to population screening that costs little per person.

The design also sets the terms for adoption. Because the authors keep cuff-based confirmation in the pathway, a positive watch result becomes extra demand for GP or pharmacy blood pressure checks and ambulatory monitoring¹. Services planning for this would need capacity for those confirmatory tests.

For wearable makers and digital health suppliers, a Nature-published prospective validation against ambulatory monitoring is the kind of evidence regulators and health technology assessors ask for. Larger studies in broader populations, including people already on treatment, will be needed before buyers treat a 65.8% sensitivity as good enough for routine screening¹.