Differences among physical activity actigraphy algorithms in three chronic illness populations

Abstract

Objectives: In three chronic illness populations and in a combined sample, we assessed differences in two algorithms to determine wear time (WT%) and four algorithms to determine: Kilocalories, light physical activity (PA), moderate-to-vigorous PA (MVPA), and metabolic equivalents (METs).

Methods: Data were collected from 29 people living with HIV (PLHIV), 27 participants recovering from a cardiac event, and 15 participants with hypertension (HTN). Participants wore the ActiGraph^TM wGT3X-BT for > 3 days on their hip. Analysis of variance (ANOVA) was used to assess differences among the algorithms.

Results: No differences were found between the two algorithms to assess WT% or among the four algorithms to assess kilocalories in each of the chronic illness populations or in the combined sample. Significant differences were found among the four algorithms for light PA (p < .001) and METs (p < .001) in each chronic illness population and in the combined sample. MVPA was significantly different among the four algorithms in the PLHIV (p = .007) and in the combined sample (p < .001), but not in the cardiac (p = .064) or HTN samples (p = .200).

Discussion: Our findings indicate that the choice of algorithm does make a difference in PA determination. Differences in algorithms should be considered when comparing PA across different chronic illness populations.

Keywords: Actigraphy algorithms; chronic illness populations; cut points; metabolic equivalents; physical activity.