Summary
A prospective multicentre study tested an AI method using smartwatch optical and motion signals to screen for hypertension without cuff calibration. Against 24-hour ambulatory blood-pressure monitoring, it recorded 65.8% sensitivity over seven days.
Researchers have evaluated an AI method that uses signals from a consumer smartwatch to screen for hypertension without cuff calibration. In a prospective, multicentre study of 196 diverse participants, the method detected hypertension with 65.8% sensitivity and 90.0% specificity over seven days of real-world monitoring, compared with 24-hour ambulatory blood-pressure monitoring as the reference. The study was published in Nature on 5 November 2026 as peer-reviewed accepted research, in an early version subject to further edits.
The authors present the approach as a way to prompt people at risk to seek clinical assessment—not as a replacement for established cuff-based confirmation.
How the smartwatch screening was evaluated
The AI method, named AI-PPG-ACC-HTN in the paper’s abstract, combines photoplethysmography (PPG) and accelerometer waveforms recorded by a consumer smartwatch. PPG is an optical measurement of changes in blood volume near the skin; an accelerometer records movement. The system used these signals without being calibrated against a cuff reading.
The researchers evaluated it prospectively at multiple centres in 196 participants who were free from known cardiovascular disease and were not taking antihypertensive medication. Participants were monitored in real-world conditions for seven days. The results were assessed against 24-hour ambulatory blood-pressure monitoring, in which a cuff records blood pressure repeatedly across a day and night.
The smartwatch method had 65.8% sensitivity (95% confidence interval, 54.0%–76.3%), 90.0% specificity (95% CI, 83.2%–94.7%) and a positive predictive value of 80.6% (95% CI, 68.6%–89.6%). Sensitivity is the share of people with hypertension whom a screening test identifies; specificity is the share without it whom the test correctly screens as negative. Positive predictive value is the share of positive results that corresponded to hypertension in this study.
What the results suggest
For comparison, initial office blood-pressure screening in the study had 55.3% sensitivity (95% CI, 43.4%–66.7%), 90.0% specificity (95% CI, 83.2%–94.7%) and a positive predictive value of 77.8% (95% CI, 64.4%–88.0%). The smartwatch method therefore had a higher sensitivity estimate in this comparison, while the reported specificity was the same. These figures describe the study results; they do not establish how performance would compare in every setting.
The potential value is opportunistic screening: a wearable already used outside a clinic could flag a possible risk and encourage follow-up. The study’s authors suggest that such screening could help identify people who might otherwise not know they have hypertension. A positive smartwatch result should lead to clinical assessment and confirmation using established cuff-based methods, rather than being treated as a diagnosis on its own.
The findings apply to the tested group and seven-day monitoring period. Participants with known cardiovascular disease or taking antihypertensive medication were excluded, so the reported performance should not be assumed for those groups.