Wrist Accelerometer Cut Points for Classifying Sedentary Behavior in Children
Christiana van Loo,A. Okely,M. Batterham,Trina Hinkley,U. Ekelund,S. Brage,J. Reilly,S. Trost,Rachel A. Jones,X. Janssen,D. Cliff
DOI: https://doi.org/10.1249/MSS.0000000000001158
2017-04-01
Medicine & Science in Sports & Exercise
Abstract:Introduction This study aimed to examine the validity and accuracy of wrist accelerometers for classifying sedentary behavior (SB) in children. Methods Fifty-seven children (5–8 and 9–12 yr) completed an ~170-min protocol, including 15 semistructured activities and transitions. Nine ActiGraph (GT3X+) and two GENEActiv wrist cut points were evaluated. Direct observation was the criterion measure. The accuracy of wrist cut points was compared with that achieved by the ActiGraph hip cut point (⩽25 counts per 15 s) and the thigh-mounted activPAL3TM. Analyses included equivalence testing, Bland–Altman procedures, and area under the receiver operating curve (ROC-AUC). Results The most accurate ActiGraph wrist cut points (Kim; vector magnitude, ⩽3958 counts per 60 s; vertical axis, ⩽1756 counts per 60 s) demonstrated good classification accuracy (ROC-AUC = 0.85–0.86) and accurately estimated SB time in 5–8 yr (equivalence P = 0.02; mean bias = 4.1%, limits of agreement = −20.1% to 28.4%) and 9–12 yr (equivalence P < 0.01; −2.5%, −27.9% to 22.9%). The mean bias of SB time estimates from Kim were smaller than ActiGraph hip (5–8 yr: 15.8%, −5.7% to 37.2%; 9–12 yr: 17.8%, −3.9% to 39.5%) and similar to or smaller than activPAL3TM (5–8 yr: 12.6%, −39.8% to 14.7%; 9–12 yr: −1.4%, −13.9% to 11.0%), although classification accuracy was similar to ActiGraph hip (ROC-AUC = 0.85) but lower than activPAL3TM (ROC-AUC = 0.92–0.97). Mean bias (5–8 yr: 6.5%, −16.1% to 29.1%; 9–12 yr: 10.5%, −13.6% to 34.6%) for the most accurate GENEActiv wrist cut point (Schaefer: ⩽0.19 g) was smaller than ActiGraph hip, and activPAL3TM in 5–8 yr, but larger than activPAL3TM in 9–12 yr. However, SB time estimates from Schaefer were not equivalent to direct observation (equivalence P > 0.05) and classification accuracy (ROC-AUC = 0.79–0.80) was lower than for ActiGraph hip and activPAL3TM. Conclusion The most accurate SB ActiGraph (Kim) and GENEActiv (Schaefer) wrist cut points can be applied in children with similar confidence as the ActiGraph hip cut point (⩽25 counts per 15 s), although activPAL3TM was generally more accurate.