Comparing Location Data From Smartphone and Dedicated GPS Devices: Implications For Epidemiologic Resea1rch

Abstract

This paper compares location data from a dedicated Global Positioning Systems (GPS) device with location data from smartphones. Data from the INTerventions, Equity, and Action in Cities Team (INTERACT), a study examining urban form changes on health in Victoria, Vancouver, Saskatoon, and Montreal, were used. A total of 337 participants contributed data from the Ethica Data smartphone app and the SenseDoc dedicated GPS. Participants recorded an average total of 14,781 Ethica locations (SD=19,353) and 197,167 SenseDoc locations (SD=111,868). Dynamic time warping and cross-correlation were used to examine spatial and temporal similarity of GPS points. Four activity space measures derived the smartphone and dedicated GPS device were compared. Analysis showed that cross correlations where above 0.8 at 125m resolution for survey and day levels and increased as cell size increases. At the day or survey level, there were only small differences between the activity space measures. Based on our findings, we recommend dedicated GPS devices for studies where the exposure and outcome are both measured at high frequency and when the analysis will not be aggregate. When the exposure and outcome are measured or will be aggregated to the day level, the dedicated GPS device and smartphones provide similar results.