Emerging technology allows patients to measure and record their heart rate (HR) remotely by photoplethysmography (PPG) using smart devices like smartphones. However, the validity and expected distribution of such measurements are unclear, making it difficult for physicians to help patients interpret real-world, remote and on-demand HR measurements. Our goal was to validate HR-PPG, measured using a smartphone app, against HR-electrocardiogram (ECG) measurements and describe out-of-clinic, real-world, HR-PPG values according to age, demographics, body mass index, physical activity level, and disease. To validate the measurements, we obtained simultaneous HR-PPG and HR-ECG in 50 consecutive patients at our cardiology clinic. We then used data from participants enrolled in the Health eHeart cohort between 1 April 2014 and 30 April 2018 to derive real-world norms of HR-PPG according to demographics and medical conditions. HR-PPG and HR-ECG were highly correlated (Intraclass correlation = 0.90). A total of 66,788 Health eHeart Study participants contributed 3,144,332 HR-PPG measurements. The mean real-world HR was 79.1 bpm ± 14.5. The 95th percentile of real-world HR was ≤110 in individuals aged 18–45, ≤100 in those aged 45–60 and ≤95 bpm in individuals older than 60 years old. In multivariable linear regression, the number of medical conditions, female gender, increasing body mass index, and being Hispanic was associated with an increased HR, whereas increasing age was associated with a reduced HR. Our study provides the largest real-world norms for remotely obtained, real-world HR according to various strata and they may help physicians interpret and engage with patients presenting such data.

Early detection of structural heart disease is critical to improving outcomes, but widespread screening remains limited by the cost and accessibility of imaging tools such as echocardiography^1,2. Recent advances in machine learning applied to heart rhythm recordings have shown promise in identifying disease^3,4, although previous work has been limited by development in narrow populations or targeting only select heart conditions⁵. Here we introduce a deep learning model, EchoNext, trained on more than 1 million heart rhythm and imaging records across a large and diverse health system to detect many forms of structural heart disease. The model demonstrated high diagnostic accuracy in internal and external validation, outperforming cardiologists in a controlled evaluation and showing consistent performance across different care settings and racial and/or ethnic groups. The models were prospectively evaluated in a clinical trial of patients without previous cardiac imaging, successfully identifying previously undiagnosed heart disease. These findings support the potential of artificial intelligence to expand access to heart disease screening at scale. To enable further development and transparency, we have publicly released model weights and a large, annotated dataset linking heart rhythm data to imaging-based diagnoses.