Remote monitoring of heart failure exacerbations using a smartwatch.

A deep learning model using Apple Watch data to predict peak oxygen uptake (pVO2) correlated strongly with CPET-measured pVO2 in a held-out validation set.

• The model was trained on data from 154 patients (46 women, 108 men) and validated on a held-out set of 63 patients (24 women, 39 men).
• Wearable-derived daily pVO2 correlated strongly with CPET-measured pVO2 with a Pearson's correlation of 0.85.
• Patients were observed over a median of 94.5 days in free-living conditions as part of the TRUE-HF study.

Each 10% drop in wearable-derived daily pVO2 was associated with a significantly increased hazard for unplanned healthcare events in the primary cohort.

• A 10% drop in wearable-derived daily pVO2 was associated with a 3.62-fold increased hazard ratio (HR) for unplanned healthcare events.
• The 95% confidence interval was 1.37–9.55 (P < 0.01).
• Unplanned healthcare events occurred at a median of 7.4 days after the first 10% drop in wearable-derived pVO2.
• This finding came from the TRUE-HF study cohort.

External validation in the All of Us Research Program cohort confirmed that drops in wearable-derived daily pVO2 were associated with unplanned healthcare utilization.

• A cross-platform model accounting for reduced-sensor capacities in the external cohort was used for validation.
• Drops in wearable-derived daily pVO2 were associated with unplanned healthcare utilization with an HR of 1.32 (95% CI 1.03–1.69; P = 0.03).
• Unplanned healthcare events occurred at a median of 21 days after the first 10% drop in wearable-derived pVO2 in the external cohort.

Wearable-derived daily pVO2 provided earlier and improved risk discrimination compared with existing wearable fitness estimates and established clinical markers.

• The comparison was made against existing wearable fitness estimates and established clinical markers of heart failure.
• The study population consisted of free-living patients with heart failure observed over a median of 94.5 days.
• The findings suggest a scalable and generalizable approach for longitudinal HF research and monitoring.

Heart failure involves cycles of remission and exacerbation that are poorly characterized by static disease measures, motivating the use of consumer wearables for daily monitoring.

• Consumer wearables have an 'understudied potential for daily monitoring of HF symptoms' according to the authors.
• The study used an observational cohort design with free-living patients to capture real-world dynamic disease states.
• Cardiopulmonary exercise testing (CPET)-measured pVO2 was used as the reference standard for disease severity.