WEDNESDAY, Nov. 4, 2026 (HealthDay News) -- A machine learning tool applied to continuous wearable blood pressure (BP) data can predict hypertensive crises up to 72 hours in advance, according to a study presented at the American Heart Association Scientific Sessions 2026, held from Nov. 6 to 9 in Chicago.
Waleed Razzaq Chaudhry, M.D., from the University of Maryland Capital Region Medical Center in Largo, and colleagues enrolled 4,190 adults with stage 2 hypertension in a prospective cohort study to examine whether machine learning applied to continuous wearable BP data could predict hypertensive crises 72 hours in advance. Participants (aged 40 to 85 years) were in continuous BP monitoring via a validated wrist-worn photoplethysmography device for 12 months. External validation was performed using a geographically distinct 20 percent hold-out cohort.
The researchers found that 418 crises occurred in 312 individuals (7.4 percent) over 12 months. In the validation cohort, the model achieved an area under the curve of 0.87. Sensitivity was 82.1 percent, specificity was 79.4 percent, and positive and negative predictive values were 31.6 and 97.4 percent, respectively, at the prespecified threshold (sensitivity ≥80 percent). Nocturnal nondipping and morning surge >35 mmHg were the strongest predictors (odds ratios, 4.82 and 3.96, respectively). Crises were reduced by 61.3 percent and emergency department utilization decreased by 44.7 percent when physicians were alerted and medications adjusted within 24 hours.
"These findings suggest that continuous blood pressure monitoring can help detect these blood pressure changes earlier, which may give patients and health care professionals more time to act before it turns into an emergency," Chaudry said in a statement.
