Convergent validity of ActiGraph and Actical accelerometers for estimating physical activity in adults

• Published in: PloS one Vol. 13; no. 6; p. e0198587
• Main Authors: Duncan, Scott, Stewart, Tom, Bo Schneller, Mikkel, Godbole, Suneeta, Cain, Kelli, Kerr, Jacqueline
• Format: Journal Article
• Language: English
• Published: United States Public Library of Science 12.06.2018; Public Library of Science (PLoS)
• Subjects: Accelerometers; Accelerometry - instrumentation; Accelerometry - methods; Actigraphy - instrumentation; Actigraphy - methods; Adult; Adults; Analysis; Biology and Life Sciences; Calibration; Convergence; Data Interpretation, Statistical; Engineering and Technology; Estimation; Exercise; Exercise - physiology; Female; Health aspects; Humans; Male; Males; Medical research; Medicine and Health Sciences; Middle Aged; Physical activity; Physiological aspects; Public health; Research methodology; Sampling methods; Sensors; Systematic review; Thresholds; Young Adult; Young adults; United States; New Zealand; Denmark; California; United States > US
• ISSN: 1932-6203; 1932-6203
• DOI: 10.1371/journal.pone.0198587

The aim of the present study was to examine the convergent validity of two commonly-used accelerometers for estimating time spent in various physical activity intensities in adults. The sample comprised 37 adults (26 males) with a mean (SD) age of 37.6 (12.2) years from San Diego, USA. Participants wore ActiGraph GT3X+ and Actical accelerometers for three consecutive days. Percent agreement was used to compare time spent within four physical activity intensity categories under three counts per minute (CPM) threshold protocols: (1) using thresholds developed specifically for each accelerometer, (2) applying ActiGraph thresholds to regression-rectified Actical CPM data, and (3) developing new 'optimal' Actical thresholds. Using Protocol 1, the Actical estimated significantly less time spent in light (-16.3%), moderate (-2.8%), and vigorous (-0.4%) activity than the ActiGraph, but greater time spent sedentary (+20.5%). Differences were slightly more pronounced when the low frequency extension filter on the ActiGraph was enabled. The two adjustment methods (Protocols 2 and 3) improved agreement in this sample. Our findings show that ActiGraph and Actical accelerometers provide significantly different estimates of time spent in various physical activity intensities. Regression and threshold adjustment were able to reduce these differences, although some level of non-agreement persisted. Researchers should be aware of the inherent limitations of count-based physical activity assessment when reporting and interpreting study findings.