Online Program Home
  My Program

All Times EDT

Abstract Details

Activity Number: 205 - Applications of Machine Learning
Type: Contributed
Date/Time: Tuesday, August 4, 2020 : 10:00 AM to 2:00 PM
Sponsor: Section on Statistical Learning and Data Science
Abstract #313088
Title: Opportunities and Challenges in the Use of Smartphone and Smartwatch-Based Step Count Measures in Studies of Physical Activity and Health
Author(s): Briana Cameron and Teresa Filshtein Sonmez* and Stella Aslibekyan and Robert Gentleman
Companies: 23andMe and 23andMe and 23andMe
Keywords: mobile; activity; bias; health

The increasing use of mobile devices presents an unprecedented opportunity to collect activity data at a large scale. This data provides a near-continuous measure of activity across large cohorts over long periods of time, providing better temporal coverage and allowing for a more comprehensive investigation into the role of physical activity in health. At 23andMe, we are able to collect phone-based step count data through Appleā€™s HealthKit integration. With months of step count data collected on a cohort of over 10,000 individuals, we derive a number of activity phenotypes including daily step count, peak 60-minute cadence, and percent of daytime active. We then investigate both within- and between-individual variability of each phenotype, examining factors such as sex, height, device type, and day of the week. We also explore how device usage patterns impact inferences derived from phone-based step count data and use this information to investigate and quantify the bias associated with phone-based measures of activity. Our results can be used to inform strategies for analyzing mobile device-based longitudinal data and to guide the development of predictive models of health.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2020 program