Smartwatch AI Flags Signs of High Blood Pressure in New Study

An artificial intelligence system using ordinary smartwatch sensors can flag signs of high blood pressure without first being calibrated with an arm cuff, according to a study published Thursday in Nature. The research points to a possible way to encourage people at risk to seek a medical check, rather than diagnose hypertension from a watch alone.

Researchers from Google and collaborators including Beth Israel Deaconess Medical Center and Harvard Medical School tested the approach in a prospective study of 196 adults at five U.S. sites. Participants had no known cardiovascular disease and were not taking blood pressure medication. The watch results were compared with 24-hour ambulatory blood pressure monitoring, the study's reference standard.

Over seven days of monitoring, the system identified 65.8% of participants who had hypertension and correctly classified 90% of those who did not. That means it missed about a third of participants with hypertension. The results do not show that a watch can reliably rule out high blood pressure on its own.

Instead of producing an exact blood pressure reading at a particular moment, the system combines several days of signals from an optical pulse sensor and a motion sensor to estimate a person's likelihood of having hypertension. The researchers compared that screening approach with existing office and home cuff-based workflows.

Google described the work as early feasibility research and said the methods are exploratory, not evidence of an existing or upcoming commercial device capability. The paper does not establish that the system works equally well for people taking blood pressure medicine or those with known cardiovascular disease, who were not included in the validation group.

The researchers say a watch-based alert could prompt someone to visit a clinic for formal assessment. Standard cuff measurements and clinical follow-up remain necessary to confirm a diagnosis.