Comparative assessment of ActiGraph data processing techniques for measuring sedentary behavior in adults with COPD

Abstract

Objective. The ActiGraph is commonly used for measuring sedentary behavior (SB), but the best data processing technique is not established for sedentary adults with chronic illness. The purpose of this study was to process ActiGraph vertical axis and vector magnitude data with multiple combinations of filters, non-wear algorithm lengths, and cut-points and to compare ActiGraph estimates to activPAL-measured sedentary time in sedentary adults with chronic obstructive pulmonary disease (COPD). Approach. This study was a secondary analysis of adults ≥50 years (N = 59; mean age: 69.4 years; N = 31 males) with COPD. Participants wore ActiGraph GT9X and activPAL3 for 7 d. ActiGraph vertical axis and vector magnitude data were processed using combinations of filters (normal, low frequency extension (LFE)), non-wear algorithm lengths (60, 90, 120 min), and cut-points for SB previously validated in older adults (two for vertical axis and three for vector magnitude data). The Bland–Altman method was used to assess concordance between sedentary time measured with 30 ActiGraph techniques and activPAL-measured sedentary time. Main results. Agreement between the two devices was moderate to strong for all techniques; concordance correlations ranged from 0.614 to 0.838. Limits of agreement were wide. The best overall technique was vector magnitude data with LFE filter, 120 min non-wear algorithm, and <40 counts/15 s SB cut-point (concordance correlation 0.838; mean difference −11.7 min d−1). Significance. This analysis supports the use of ActiGraph vector magnitude data and LFE filter in adults with COPD, but also demonstrates that other techniques may be acceptable with appropriate cut-points. These results can guide ActiGraph data processing decisions.