University of Toronto and UHN Lead Groundbreaking Research on Wearable Technology for Heart Failure
Researchers from the University of Toronto's Temerty Faculty of Medicine and University Health Network (UHN) have unveiled a transformative study demonstrating how everyday smartwatches can forecast hospitalization risks for heart failure patients. Published on the cover of Nature Medicine on March 20, 2026, this work from the Ted Rogers Centre for Heart Research and UHN's Peter Munk Cardiac Centre leverages consumer-grade devices like the Apple Watch to monitor cardiopulmonary fitness in real-world settings. By analyzing data on heart rate, activity levels, and oxygen saturation, an AI model predicts declines that signal impending health crises days or even weeks in advance.
Heart failure, a condition where the heart cannot pump blood effectively to meet the body's needs, affects approximately 750,000 Canadians and leads to over 100,000 new diagnoses annually. It ranks among the top five causes of hospitalization in the country, with one in five patients readmitted within a month of discharge and nearly half passing away within five years of diagnosis. Globally, 64 million people grapple with this chronic illness, underscoring the urgency for innovative monitoring tools that bridge gaps between clinic visits.
The Burden of Heart Failure in Canada and Beyond
In Canada, heart failure imposes a staggering economic toll, with projections estimating 1.69 million hospitalization episodes between 2019/2020 and 2039/2040, costing the healthcare system $19.5 billion. Traditional management relies on periodic assessments like NT-proBNP blood tests, six-minute walk tests (6MWT), or New York Heart Association (NYHA) class evaluations—static snapshots that often miss subtle deteriorations. Patients experience unpredictable cycles of stability and sudden exacerbations, leading to emergency interventions. This UofT-UHN study addresses that void by enabling continuous, passive surveillance through wearables.

Wearable technology integration into cardiology represents a shift toward proactive care, particularly vital in a vast country like Canada where rural patients face barriers to frequent specialist access.
Unpacking the TRUE-HF Study: Methodology and Innovation
The Ted Rogers Understanding Exacerbations of Heart Failure (TRUE-HF) study enrolled 217 ambulatory heart failure patients, providing them with Apple Watch Series 6 devices and iPhones. Over a median of 94.5 days, participants continued normal routines while wearables captured metrics including step count, exercise time, heart rate variability (HRV), and peripheral oxygen saturation (SpO2). Clinic visits bookended the period with gold-standard cardiopulmonary exercise testing (CPET) to measure peak oxygen uptake (pVO2)—a robust indicator of cardiorespiratory fitness.
A novel deep learning model, dubbed TRUE-HF, processed 30-day sliding windows of 90-minute aggregated data alongside demographics (age, sex, weight) and clinical variables (medications, ejection fraction). This autoregressive transformer architecture output daily absolute pVO2 estimates (in liters per minute), correlating strongly with CPET (Pearson's r = 0.85, mean absolute error = 0.25 L/min). The threshold for concern: a ≥10% pVO2 drop, informed by prior evidence linking such declines to heightened hospitalization risk.
- Training cohort: 154 patients
- Internal validation: 63 patients
- Key inputs: Steps, active energy, HRV, SpO2
Striking Results: Precision Prediction and Risk Stratification
In the held-out cohort, a 10% pVO2 decline heralded a 3.62-fold hazard ratio (HR) increase for unplanned healthcare utilization—hospital admissions, urgent visits, or IV diuretics—with events occurring a median 7.4 days later (AUROC 0.77). Sensitivity reached 88% at this threshold, outperforming static markers like baseline pVO2 (AUROC 0.66), NT-proBNP (0.61), and 6MWT distance (0.64). Even Apple's cardio fitness estimate lagged (AUROC 0.52 for decline detection).
External validation on 193 U.S. patients using Fitbit data (TRUE-HF-RS reduced-sensor model) confirmed generalizability: HR 1.32 per 10% drop, median lead time 21 days. This cross-platform robustness highlights scalability beyond proprietary ecosystems.Read the full Nature Medicine paper.
| Metric | AUROC (TRUE-HF pVO2) | AUROC (Apple Estimate) |
|---|---|---|
| ≥10% CPET Decline Detection | 0.82 | 0.52 |
| Unplanned Utilization Prediction | 0.77 | N/A |
Real-World Validation and Broader Applicability
The All of Us cohort validation underscores the model's versatility, adapting to Fitbit's sparser sensors (steps, HR) via knowledge distillation. Despite reduced fidelity, it retained predictive power, suggesting viability for diverse wearables. Lead researcher Dr. Heather Ross notes, "Heart failure often worsens silently between clinic visits. By tapping into information captured through everyday wearable tech, this study shows we can detect significant changes much earlier, and potentially intervene before a health crisis occurs."
Participant Paula Vanderpluym, living with hypertrophic cardiomyopathy, shared: "The whole idea that doctors could use this data to predict if you're going to get worse... was something I was more than happy to participate in." Her experience illustrates equitable potential for remote areas.UHN press release details patient impact.
Implications for Heart Failure Management and Healthcare Equity
This innovation could slash readmissions, a persistent challenge costing billions. By flagging deteriorations proactively, clinicians might adjust diuretics, optimize devices like LVADs, or schedule telehealth sooner. Chris McIntosh emphasizes, "We couldn't have done this anywhere else. This work reflects UHN's commitment to translating innovation into clinical tools through a highly interdisciplinary team."
In Canada, where heart failure drives top hospitalization costs per CIHI data, wearables democratize monitoring. Rural Ontarians, like those in remote UHN networks, gain from passive data collection—no clinic treks required.
Context Within Global Wearable Cardiology Research
This builds on pioneers like LINK-HF (predicting HF hospitalizations via wearables) and Apple Heart Study (AFib detection). Fitbit studies correlate activity intolerance with exacerbations, but UofT-UHN's pVO2 focus offers superior granularity over steps or HR alone. Ongoing trials like WeRoaM explore textiles for pediatric HF, signaling maturation.Explore related trials.

Stakeholder Perspectives: Clinicians, Patients, and Policymakers
Cardiologists hail the shift to dynamic metrics. Patients value unobtrusive tracking—Eddy Lam, a study participant, relies on his smartwatch post-HF. Policymakers eye cost savings; integrating into provincial health systems could align with Canada's digital health strategy.
Future Outlook: From Research to Routine Care
UHN eyes clinical trials for intervention arms, patenting TRUE-HF for apps. Collaborations with Apple/Fitbit may yield FDA approvals. Challenges: Data privacy, equity in access, algorithm bias mitigation. Yet, as Ross states, "The findings... are a potential game-changer."
For Canadian higher education, this exemplifies translational research prowess, fostering careers in AI-health intersections at institutions like UofT.



