Representation and Bias in Self-reported Physical Activity Across Social Media Platforms

Nina Cesare , Boston University
Christan Grant, University of Oklahoma
Elaine Nsoesie, Boston University

This study seeks to assess bias and representation within self-reports of physical activity across multiple social media platforms. While many acknowledge that adoption and usage variation within and between social media platforms introduces bias into social media platforms, scientists have only begun to quantify this bias. Using data from three social media-based, digital sources of self-reported physical activity – Twitter, Strava and Instagram – this study will estimate the gender composition of users represented by each platform, assess whether the age and gender composition of these samples is significantly different than the counties from which the posts originate, and assess – in combination with county-level environmental and sociodemographic variables – how effectively each source may be used to estimate physical inactivity. Results have important implications for researchers seeking to use digital data to assess physical activity trends, as well as researchers interested in understanding how individuals within specific demographic groups use digital platforms for health and wellness.

See extended abstract

 Presented in Session 5. Health & Mortality 1