Smartwatches predict heart failure
A new deep‑learning study validated consumer smartwatches for remote monitoring that can predict heart‑failure exacerbations and higher hospitalization risk by analyzing real‑time biometrics. This points toward more accessible, preventive cardiology outside clinic walls. (nature.com)
Nature Medicine received the manuscript on 21 February 2025, accepted it on 23 January 2026 and published it online on 20 March 2026. (nature.com) The trial—named the Ted Rogers Understanding Exacerbations of HF (TRUE‑HF) study—followed free‑living heart‑failure patients for a median of 94.5 days and trained a deep‑learning model to estimate daily peak oxygen uptake (pVO2) from smartwatch data against in‑clinic CPET measurements. (nature.com) Model development used 154 patients for training (46 women, 108 men) and a held‑out validation set of 63 patients (24 women, 39 men), for a combined cohort of 217 participants. (nature.com) Wearable‑derived daily pVO2 correlated with CPET‑measured pVO2 with Pearson’s r = 0.85 in the validation set. (nature.com) A 10% drop in wearable‑derived daily pVO2 was linked to a 3.62‑fold higher hazard for unplanned healthcare events (95% CI 1.37–9.55; P < 0.01), with events occurring a median 7.4 days after the first 10% decline; an external All of Us cohort using a reduced‑sensor model showed HR 1.32 (95% CI 1.03–1.69; P = 0.03) with a median 21 days to event. (nature.com) The authors report that wearable‑derived daily pVO2 offered earlier and improved risk discrimination compared with standard wearable fitness estimates and established clinical markers, framed against a backdrop of heart failure affecting an estimated 64.3 million people worldwide and costing about US$346 billion annually. (nature.com)
