Detecting signs of hypertension using wearable sensors

Hypertension often goes unnoticed, especially among younger adults, because it may cause no obvious symptoms. A proposed smartwatch system aims to passively identify persistent signs of high blood pressure without the cuff calibration required by many existing wearable tools. Rather than estimate exact blood pressure at a particular moment, it combines physiological signals gathered over several days to reduce the influence of short-lived changes. It uses photoplethysmography (PPG), which records pulse-related changes in blood volume, and accelerometry, which captures movement and subtle vibrations caused by the heart’s pumping. A self-supervised neural network learned patterns from 492,347 hours of sensor recordings from 2,109 participants. The system aggregates those patterns across days, combines them with a person’s height, and produces a hypertension probability score.

A prospective study at five U.S. sites tested the system in 196 adults without known cardiovascular disease who were not taking blood pressure medication, using 24-hour ambulatory monitoring as the reference standard. For initial screening, the smartwatch had 65.8% sensitivity, 90.0% specificity, and an 80.6% positive predictive value, compared with 55.3% sensitivity for office screening. Among 48 people with elevated clinic readings, it detected sustained hypertension with 78.4% sensitivity, 90.9% specificity, and 96.7% positive predictive value, outperforming repeat office measurements on these measures and showing performance competitive with multi-day home monitoring. The findings support a possible role as a screening aid, not as a replacement for established diagnosis.