Accelerometry Harmonization Across Research-grade And Consumer Wearable Devices: A Comparison Of MIMS, ENMO And MAD

PURPOSE: Accelerometry metrics, including Euclidean Norm Minus One (ENMO), Mean Amplitude Deviation (MAD), and Monitor-Independent Movement Summary (MIMS) units, are meant to harmonize across devices with different sampling rates and dynamic ranges. Current evidence exploring the potential of these metrics to be device-agnostic has focused on ENMO and MAD and has relied on shaker table studies. The objective of this study was to identify which metric (ENMO, MAD, MIMS) best harmonizes across devices with varying sampling rates and dynamic ranges in children. METHODS: Children (n = 239; 9.3 ± 2.1 years, 47% female, 30% Black) wore Actigraph GT9X (+/- 8 g, 50 Hz) research-grade accelerometers and were randomized to wear two of three consumer wearables including Apple Watch Series 7 (+/- 16 g, 50 Hz), Garmin Vivoactive 4S (+/- 8 g, 25 Hz), and Fitbit Sense (+/- 4 g, 50 Hz) on their non-dominant wrist, while participating in 60 minutes of simulated free-living activities (i.e., walking, running, soccer). The standard deviation (SD) across z-scores for ENMO, MAD and MIMS of each device was calculated at the second level as an estimate of variability across devices. Lin’s Concordance Correlation Coefficient (CCC) was calculated for each combination of devices by metric to determine agreement. RESULTS: The SD of z-score across devices was the lowest, indicating better harmonization, for MAD (0.12 ± .23), followed by ENMO (0.24 ± .50), and then MIMS (0.26 ± .39). Lin’s CCC was strongest across all combinations of devices for MAD, with coefficients ranging from 0.89 for Apple and Fitbit to 0.96 for Apple and Actigraph. Lin’s CCC for MIMS ranged from 0.70 for Apple and Fitbit to 0.83 for Apple and Actigraph, and was lowest for ENMO, ranging from 0.62 for Garmin and Actigraph to 0.76 for Apple and Actigraph. CONCLUSIONS: MAD appears to perform best at harmonizing across research-grade and consumer devices with different sampling rates and dynamic ranges in children. Supported by: This publication was made possible in part by Grant Number T32-GM081740 and T32-GM145226 from NIH-NIGMS. Research reported in this publication was also supported in part by the National Institute of Diabetes and Digestive and Kidney Diseases Award Number R01DK129215. Co-author White was supported in part by National Institute of Diabetes and Digestive and Kidney Diseases Award Number F31DK136205. Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the NIGMS or NIH.