Registration Is Now Open!
ActiGraph Digital Data Summit 2021November 4 - 5 | Learn more
Challenges and opportunities for measuring physical activity and sedentary behaviour in children and adolescents: What do we know and how can we improve compliance?
- Presented on 2015
Introduction: This study examined agreement between activPAL3 (AP) and ActiGraph (AG) wrist and hip cut-points for assessing sedentary behaviour (SB) among 5-12 year-old children using direct observation (DO) as the criterion.
Methods: 30 children (9.2±2.1y, 53.3% boys) wore AG on both wrists and the right hip and a thigh-mounted AP while completing 15 5min semi-structured sedentary, light, and moderate-to-vigorous-intensity activities. Posture was classified by AP and coded from video for DO (1s epochs) during activities and transitions (151.9min±36.6). SB was defined using AG wrist cut-points developed for the vertical axis (VA) and vector magnitude by Crouter et al. (CR) and Kim et al. (KI), and for the hip (AG_hip, 25 counts(c)/15s). Analyses examined equivalence of time estimates (equivalence testing), individual level bias (Bland-Altman plots) and misclassification (receiver operating characteristic (ROC) curves).
Results: Although none of the methods were equivalent with DO (68.5min: equivalence range=61.7-75.4), AP (64.5min: 90%CI=58.6-70.5), CR_VA ≤.35c/5s; 64.1min: 90%CI=58.7-69.4), KI_VA (≤.1756c/60s) dominant (65.0min: 90%CI=59.4-70.6) and non-dominant (63.8min: 90%CI=58.1-69.6) demonstrated a small underestimation, while AG_hip overestimated SB (80.3min: 90%CI=73.0-87.6). Mean differences were smaller and 95% limits of agreement narrower for AP (4.0min: -9.7-17.7), CR_VA (4.5min: -14.0-23.0), KI_VA dominant (3.6min: -11.2-18.3) and non-dominant (4.7min: -11.6-14.2), compared to AG_hip (-11.8min: -30.0-6.5). Although area under the ROC curve for AP (0.995) and KI_VA dominant (0.884) and non-dominant (0.865) were higher than AG_Hip (0.853), CR_VA (0.787) was less accurate.
Conclusion: On group- and individual-level tests, the AG KI_VA wrist cut-point demonstrated similar agreement to AP and better agreement than AG_hip for estimating SB.