2016 Poster Sessions : The Mobile Device as a Sensor for Physical Activity and Health from Personal to Planetary Scale

Student Name : Tim Althoff
Advisor : Jure Leskovec
Research Areas: Information Systems
Abstract:
Physical inactivity is a major public health problem contributing to approximately 5.3 million deaths per year worldwide (Lee et al., 2012). Regular physical activity helps to prevent heart disease, stroke, some forms of cancer, diabetes, and weight gain. National health surveillance efforts, such as NHANES, have been used to assess daily physical activity in the US population but at high financial costs and with limited scale and resolution. The growing popularity of consumer mobile devices, such as smartphones and wearable sensors, generates large volumes of data that could be key to better understanding physical activity patterns and health across millions of free-living individuals.

Our aim is to quantify the current state of daily physical activity and identify relationships between physical activity, sleep, weight status, and built environment factors. We present a preliminary analysis of data from Azumio’s Argus smartphone application, which includes physical activity, sleep data, and other health indicators from about 2 million people around the world. We show how to distill minute-by-minute, accelerometer-determined step counts into a novel and concise set of physical activity measures capturing volume, duration, and intensity. Within the subset of individuals in the U.S. (N=889,599; median age=35; 28.8% obese; 4365 mean daily steps), we find that longer bouts of activity (>=90 steps/min) were associated with lower BMI, lower resting he art rate, and longer and more efficient sleep. In addition, short bouts of activity (<=5 min) were also associated with lower BMI. Higher intensity physical activity was associated with lower BMI and resting heart rate. We also provide initial results connecting health indicators to environmental factors such as walkability and food supply composition on city, state, and country scales.

These initial results demonstrate that consumer devices provide data at unprecedented scale and resolution, allowing us to address questions around physical activity and health from a personal to planetary scale.

Bio:
Tim Althoff is a PhD candidate in the Computer Science Department at Stanford University where he works with Jure Leskovec and is part of SNAP, the InfoLab, and the Mobilize Center.
Tim enjoys analyzing large-scale traces of human activity to learn more about human behavior. Most recently, he has been working on computational linguistic methods to analyze counseling and conversation strategies and on data mining and social network analysis techniques to better understand exercise and health behaviors and their relation to health outcomes and the environment.