Using GPS-enabled mobile phones to characterize individuals’ activity patterns for epidemiology applications

Abstract

GIScience 2016 Short Paper Proceedings Using GPS-enabled mobile phones to characterize individuals’ activity patterns for epidemiology applications E.-H. Yoo and Y.-S. Eum University of Buffalo, SUNY, Buffalo, NY, USA Email: {eunhye, yeum}@buffalo.edu Abstract We assessed the potential of global positioning system (GPS)-equipped mobile phones for health-related studies. We demonstrated the use of GPS data as a means of collecting individuals’ activity patterns for personal exposure assessments and public health surveillance. The widespread use of mobile phones has enabled investigators to conduct exposure studies and to detect infectious disease at the individual level on a massive scale. However, still substantial uncertainties are present in converting raw GPS data into relevant information. To address these issues, we proposed three algorithms for pre-processing and classification of raw GPS data, and demonstrated their applications to real world data in a case study. 1. Introduction Exposure models typically impose unrealistic assumptions such that subjects within a neighborhood are equally exposed to air pollution and/or most individuals spend their time at their residences. Similarly, a lack of understanding of human movement, which is an important component of disease transmission, has been considered as an obstacle to develop effective national communicable disease control programs. In exposure modelling, some improvements have been achieved by adopting a microenvironment (ME) approach where individuals’ time spent at MEs, such as outdoors, residence, and workplace, was explicitly taken into account. However, collecting the information on individuals’ time-activity patterns has been cost-, time-, and labor-intensive with limited reliability and accuracy. Comparably, aggregated data have limited efforts to reconstruct the complex and dynamic nature of real- world contact networks, which plays a critical role of contact network in an outbreak of dangerous infectious disease. The emergence of lightweight, low-cost, and accurate GPS devices has provided a promising tool for objectively assessing the geographic positions of the environmental context in which health-related behaviors take place (Schipperijn et al. 2014). GPS technology enabled investigators to capture daily trajectories of individuals with higher temporal resolution at increasing locational precision (Gerharz et al. 2013, Dias & Tchepel 2014), although the use of GPS data in health research is not without challenges. As reviewed by Krenn et al. (2011), the positional accuracy of GPS data collected in real world is often unacceptable in health studies, especially, over longer study periods. The data quality of GPS traces depends on the amount of data lost from signal drop-outs, loss of device battery power, and poor adherence of participants to following the specific research protocol. Despite the advancement in GPS technology, signal acquisition is still affected by the presence of tall buildings and significant uncertainties associated with the processing and classifying raw data are present. In this study we focus on the GPS data collected from a mobile phone with and without data connection. Our primary goal is to identify major MEs associated with health-related activities using GPS data. First we developed and applied a “selective resampler” to raw GPS data for pre-processing. Using the processed data, we identify significant places and travel modes using the two automated classification schemes.