A Population-derived Maximal Heart Rate Method For Moderate-to-vigorous Physical Activity Monitoring Using Consumer Wearables

PURPOSE: To present a novel data-centric approach to age-predicted maximal heart rate (APMHR) estimation using moderate-to-vigorous physical activity (MVPA) population statistics over direct in-lab exertion testing. Published evidence using self report and accelerometry suggests that absolute PA declines with age, but MVPA measured using heart rate reserve (HRR) based on existing APMHR formulae shows a systematic upwards skew with age. We derive a new APMHR formula based on the assumption that the distribution of daily MVPA by age should be similar in a large and diverse population of wristworn wearable consumers with similar fitness profiles across ages. METHODS: Up to 47 million hours of wrist-worn wearable data covering an eight-week period is retrospectively analyzed for a 52,000 subset of consenting Google Pixel Watch® users in the United States between the ages of 20 and 80 years (20s: 11.27%, 30s: 36.14%, 40s: 27.24%, 50s: 13.89%, 60s: 8.12%, 70s: 3.35%). Measured data includes heart rate (sampled at 0.2 Hz to 1 Hz), physical activity and demographics. MVPA is calculated using a Karvonen protocol based on personalized HRR, per-minute mean heart rate and a fitted APMHR formula, parameterized linearly with age such that skew between cumulative daily MVPA across age groups is minimized. RESULTS: A simple linear APMHR parameterization is found which rewards MVPA equitably in adults 20 to 80 years old and is similar to best-established formulae. Validation across eight weeks using a reserved 30% subset of the cohort shows reduced age-attributed overestimation compared to the Fox formula (220 - age) by 87% with 95% confidence interval (79%, 95%). The versatility of this data-driven approach is demonstrated by modeling relative adherence of a population to physical activity guidelines. CONCLUSION: Consumer wearables enable valuable insights into real-world physical activity behaviors through passive monitoring. The value of using HRR based approaches are well recognized. The breadth and magnitude of available data encourages a novel technique for investigating the relationship between age and maximal heart rate; one that does not necessitate the demanding requirements of lab-based graded exercise experiments, whose generalizability are hindered by cost, scale, invasiveness and repeatability.