Abstract P320: Validation of Smartphone-Recorded Physical Activity for Predicting Cardiorespiratory Fitness

Abstract

Introduction: While most smartphones can track physical activity, whether this readily available data is associated with clinically meaningful outcomes is unknown. We tested the hypothesis that smartphone-recorded physical activity is associated with cardiorespiratory fitness as measured by peak metabolic equivalents of task (METs) achieved on an exercise stress test. Methods: We recruited 51 Apple iPhone users from the Beth Israel Deaconess Cardiovascular Stress Testing Laboratory between September 2017 - June 2018. We securely downloaded iPhone activity episodes, or epochs, and calculated cumulative steps, distance, and duration; peak speed and duration; and average daily peaks for specific intervals preceding stress tests. Age, gender, height, weight, and METs were obtained from the medical record. In a 70% training set, we used linear regression to determine relationships between individual components of physical activity and METs and to derive a multivariable prediction algorithm for METs. We then tested the best performing algorithm (based on Akaike Information Criterion) in a 30% test set. Results: We collected over 1.1 million activity epochs with a median of 17,103 epochs per participant. Twenty participants were female, median age [interquartile range] was 67 [53, 71] years, and median time preceding stress test [range] was 535 [13, 1,147] days. Average daily peak speed at 180 days most strongly correlated with METs (R 2 =0.44, p<0.001), and correlation was similar at 3 days (R 2 =0.41, p<0.001). The best-fitting multivariable model included age, gender, height, weight, and average peak speed at 180 days. Conclusion: Smartphone-recorded physical activity is readily available and strongly associated with cardiorespiratory fitness, even in an external validation set.