Development and validation of an activPAL accelerometry count-based model of physical activity intensity in adults

Abstract

The activPAL linear cadence-metabolic equivalents (METs) equation poorly estimates activity intensity. The magnitude of acceleration in three directional planes may be a superior predictor of activity intensity than stepping cadence, with accelerometry count thresholds developed in children/adolescent populations. We extracted the proprietary accelerometer-derived information to develop a counts-METs model and cross-validates it in laboratory and free-living conditions. Forty adults (25±6 years) wore an activPAL during a 7-stage progressive treadmill protocol (criterion: indirect calorimetry). Tri-axial accelerometry-derived activity counts (vector magnitude) and METs data from a subset of participants (n = 20) were modelled (R2=0.76) and the regression equation evaluated in the remaining participants (n = 20). Thirty-two of these participants wore the activPAL during free-living conditions (n = 192-d; criterion: PiezoRxD monitor). The absolute percent error of the counts-METs model in the laboratory cross-validation was 18±13%, with equivalence testing determinining equivalent MET values to indirect calorimetry during the slowest (1.5 mph) and fastest (4.0-4.5 mph) stages. In free-living conditions, the model accurately quantified light- and moderate-intensity physical activity but underestimated vigorous-intensity activity (6.5±11.3 vs. 5.5±20.8 mins/day; p < 0.001). We developed and present a data analysis method using the activPAL tri-axial accelerometry counts to improve estimations of physical activity intensity in controlled laboratory settings and uncontrolled free-living settings.