Optimizing wearable device and testing parameters to monitor running-stride long-range correlations for fatigue management in field settings

Purpose: There are important methodological considerations for translating wearable-based gait-monitoring data to field settings. This study investigated different devices’ sampling rates, signal lengths, and testing frequencies for athlete monitoring using dynamical systems variables. Methods: Secondary analysis of previous wearables data (N = 10 runners) from a 5-week intensive training intervention investigated impacts of sampling rate (100–2000 Hz) and signal length (100–300 strides) on detection of gait changes caused by intensive training. Primary analysis of data from 13 separate runners during 1 week of field-based testing determined day-to-day stability of outcomes using single-session data and mean data from 2 sessions. Stride-interval long-range correlation coefficient α from detrended fluctuation analysis was the gait outcome variable. Results: Stride-interval α reduced at 100- and 200- versus 300- to 2000-Hz sampling rates (mean difference: −.02 to −.08; P ≤ .045) and at 100- compared to 200- to 300-stride signal lengths (mean difference: −.05 to −.07; P < .010). Effects of intensive training were detected at 100, 200, and 400 to 2000 Hz (P ≤ .043) but not 300 Hz (P = .069). Within-athlete α variability was lower using 2-session mean versus single-session data (smallest detectable change: .13 and .22, respectively). Conclusions: Detecting altered gait following intensive training was possible using 200 to 300 strides and a 100-Hz sampling rate, although 100 and 200 Hz underestimated α compared to higher rates. Using 2-session mean data lowers smallest detectable change values by nearly half compared to single-session data. Coaches, runners, and researchers can use these findings to integrate wearable-device gait monitoring into practice using dynamic systems variables.