Insights On Free-living Sedentary Behavior Estimates Using A Hip- Or Wrist-worn Accelerometer

Abstract

PURPOSE: The primary purpose was to validate existing methods to estimate sedentary behavior (SB) under free-living conditions using ActiGraph GT3X+ accelerometers (AG). The secondary purpose was to identify method-specific systematic errors that result in the misclassification of SB. METHODS: Forty-eight participants (age:20.4±1.3 years, 45.8% male) were video-recorded during four 1-hour sessions in different settings (home, community, school, environment) while wearing an AG on the right hip and non-dominant wrist. Videos were coded for postural orientation and activity type (e.g. walking). Observed time in sitting and lying postures were classified as SB (criterion measure). Twelve methods were applied to hip and wrist accelerometer data to estimate time spent in SB (see Figure 1). Repeated measures linear mixed models were used to estimate method bias (estimate - criterion SB) and a 95%CI around the bias. RESULTS: On average, participants spent 34.1 of the 57.2 minutes/session in SB. Four of the hip methods were unbiased (Soj1x, Soj3x, CPM100, CPM150), however SB was underestimated using CPM200vm (-5.5 minutes, 95%CI: -7.1, -3.8) and overestimated using ENMO47.4 (12.2 minutes, 95%CI: 9.9, 14.5). For the wrist, Sed Sphere was the only unbiased method. SB was overestimated using ENMO44.8 (3.7 minutes, 95%CI: 1.8, 5.5) and underestimated using Wrist RF, CP15s376vm, Wrist TR, and CPM1853vm, ranging from -9.5 to -5.7 minutes. The majority of misclassified SB occurred during standing or sitting behaviors (67.0-96.7%). CONCLUSION: Accurate estimates of SB from a hip-worn AG can be achieved using either simpler count-based approaches (CPM100, CPM150) or machine learning models (Soj1x, Soj3x). Only the Sedentary Sphere may be suitable to estimate SB from the non-dominant wrist. Future work to distinguish standing from SB may lead to improvements in estimating SB. Supported by NIH NIDDK 1R01DK110148Figure 1: Mean bias and 95%CI for hip- and wrist-method estimated time spent in sedentary behavior. *indicates method was applied to low-frequency extension processed accelerometer data.