Comparison of stepping-based metrics from ActiGraph accelerometers worn concurrently on the non-dominant wrist and waist among young adults

Abstract

ABSTRACT Step counts can be estimated from wrist-worn accelerometers through the Verisense Step Count Algorithm. No study has assessed agreement between stepping metrics from ActiGraph accelerometers during free-living. Thirty-four participants (age: 22.9 ± 3.4 years) provided 24 h accelerometer data (non-dominant wrist) and waist. Agreement of two Verisense Algorithms (Verisense 1 & 2) for estimating daily steps, moderate-to-vigorous physical activity (MVPA), peak 1-min and 30-min accumulated steps, against the waist and ActiLife step-count Algorithm was assessed. Mean bias ± 95% limits of agreement (LoA) for daily steps was +1255 ± 3780 steps/day (mean absolute percent error (MAPE): 21%) (Verisense 1) and +1357 ± 3434 steps/day (MAPE: 20%) (Verisense 2). For peak 1-min accumulated steps, mean bias and 95% LoA was −17 ± 23 steps/min (MAPE: 17%) (Verisense 1) and −6 ± 5 steps/min (MAPE: 9%) with Verisense 2. For peak 30-min accumulated steps, mean bias and 95% LoA was −12 ± 45 steps/min (MAPE: 25%) (Verisense 1) and −2 ± 38 steps/min (MAPE: 13%) (Verisense 2). For MVPA steps/day, mean bias and 95% LoA was −1450 ± 3194 steps/day (MAPE: 420%) (Verisense 1) and −844 ± 2571 steps/day (MAPE: 211%) (Verisense 2). For MVPA min/day, mean bias and 95% LoA was −13 ± 27 min/day (MAPE: 368%) (Verisense 1) and −8 ± 24 min/day (MAPE: 209%) (Verisense 2). The Verisense 2 algorithm enhanced agreement for stepping intensity metrics but further refinement is needed to enhance agreement for MVPA against the waist.