A Comparison of Wrist and Hip Accelerometer Output at Different Walking Speeds

Background: Physical activity has been objectively measured using hip-worn accelerometers for decades. However, wrist-worn accelerometers are currently used in large-scale studies. Differences in wrist and hip dynamics during locomotion affect monitor output, which may impact how prediction models are built.

Purpose: To compare ActiGraph™ wrist and hip accelerations (g’s) at different locomotion speeds.

Methods: Participants (N = 7) wore ActiGraph™ GT3X+ accelerometers on the dominant wrist and hip (sampling rate 80Hz). They performed three 5-minute trials at self-paced (SP), slow (SL), and fast (F) over-ground walking speeds. Mean and standard deviation of the vector magnitude (VM) were calculated from two 20-s data windows per condition. Linear mixed-effects models were used to compare the relationship between speed and vector magnitude (VM) for the hip and wrist monitor locations.

Results: There was a significant difference between the slopes (speed vs VM) for the hip, (m = 0.052 [95% CI: 0.033, 0.103] compared to the wrist, m = 0.195 [95% CI: 0.160, 0.230], p<0.001).

Conclusions: The results show that the pattern of change in ActiGraph™ wrist and hip accelerations (g’s) differ as locomotion speed increases. There is a curvilinear increase in wrist VM as locomotion speed increases, whereas there is a linear increase in hip VM as locomotion speed increases. The pattern of change is different and more variable between subjects for the wrist VM, which may impact measurement error and model development. Additionally, wrist VM is more responsive to changes in speed than hip VM, suggesting that a wrist worn accelerometer may be more sensitive to locomotion intensity.