A treadmill-based evaluation of the validity of the ActivPAL4 for detecting steps in walking and running in healthy adults

Objective: To accurately characterize people’s physical activity validated measurement devices must be used. The activPAL4 physical activity monitor allows measurement of stepping. A systematic approach was taken to validate the monitor for walking and running, across the walk to run transition.

Methods: A cross-sectional sample of healthy young adults completed an incremental treadmill-based protocol (speeds 3.0–11.8 km/hr in 0.4 km/hr increments; 0.83–3.28 m/s). A video-based reference standard was used to evaluate the accuracy of the activPAL4 (thigh-mounted activity monitor) in detecting steps with mean percentage error, mean absolute percentage error, and agreement calculated. A linear regression modeled mean percentage error using sex, body mass index, height, cadence, and step length.

Results: Thirty-six participants (25.6 ± 3.7 years, 18 M/18 F) transitioned from walking to running between 1.61 and 2.28 m/s. Mean percentage error between activPAL4 and video-based step count within speeds ranged from 1.6% to 4.3% for walking and 1.1% to 20.5% for running. Overall, walking step detection was good (mean absolute percentage error 3.1%), but steps were undercounted by the activPAL4 during running (mean absolute percentage error 14.5%). Regression modeling indicated that in walking for males activPAL4 undercounted by 2.85%, for females 1.79%. In running, for males activPAL4 undercounted steps by 13.3%, for females 15.6%. Sex and body mass index explained only a small amount of variance(η²p < 0.34). Cadence was the most important predictor of percentage error in running (η²p = .322) contributing to an adjusted R² of .391.

Conclusion: The activPAL4 measures walking steps well across the range of walking speeds from 0.83 to 1.94 m/s. Running steps were undercounted by the activPAL4 in relation to video-based observation between 10% and 20% between 1.83 and 3.28 m/s.