Extent and Impact of Accelerometer Time Drift on Linkage With Global Positioning System Data for Physical Activity Research

Abstract

Background: Accelerometer time may drift, whereas global positioning system time does not (due to constant satellite communication), potentially impacting linkage of these data; the impact of this issue on outcome measures is unknown. Our study assessed if time drift is consistent between devices and/or data collection waves and the impacts on common outcomes in built environment physical activity research. Methods: First, 11 ActiGraph wGT3X-BT accelerometers were manually or automatically shaken every 6 hr for 10 days, and this experiment was repeated twice (two waves). Accelerometer time was compared with local atomic clock time to identify time drifts. Next, we assessed the time-drift impact on estimates of park moderate to vigorous physical activity in an outdoor experiment, comparing no adjustment (status quo) to device-specific and average adjustment to account for time drift. Finally, we used data from an ongoing health study in Detroit, Michigan, to evaluate the impact of time drift adjustment on free-living linked accelerometer and global positioning system data. Results: There was an average of 6.1 ± 5.1 s of lag in accelerometer time after 10 days, but time drifts varied among accelerometers. Mean absolute differences in park moderate to vigorous physical activity were 0.4–0.5 min/day over the entire outdoor experiment or 0.8–0.9 min/day on Day 7. In the free-living study, overall differences in park moderate to vigorous physical activity were 0.0–0.1 min/day, but the maximum difference was 1.9 min/day. Conclusions: Time drift had little impact on estimates, although research requiring high temporal resolution should assess the time drift of individual accelerometers before data collection to account for drift, particularly when linking to global positioning system data.