Evaluating The Agnostic Nature Movement Summary Indicators In Free-living

Various movement summary indicators have been proposed to provide a standard way to process raw accelerometer data including the Monitor-Independent Movement Summary (MIMS), Mean Amplitude Deviation (MAD), Euclidean Norm Minus One (ENMO), Activity Index (AI), and Actigraph Counts (AC). PURPOSE: The objective of this study is to evaluate the agnostic nature of these indicators using matched monitor data collected from the Free-Living Study for Health (FLASH). METHODS: A sample of 110 participants (age 24.7 ± 9, BMI 24.7 ± 4, 63.6% female) wore 4 wrist-worn devices: ActiGraph GT3x monitors on both wrists and counterbalanced GeneActiv or Axivity monitors on dominant and non-dominant wrists. Each was configured to record at 100 Hz. Participants wore devices for 24 hours and reported their physical activity (PA) via the ACT24 web survey the following day. Summary metrics were obtained using open-source data processing methods. Equivalence testing at 10% and MAPE were used to assess agreement between matched pairs of devices worn on the same wrists. RESULTS: The MIMS indicator yielded statistically equivalent values for all pairwise combinations using standard 10% equivalence bands. The MAD indicator also yielded equivalent values, but the intervals were wider. The ENMO, AI and AC estimates were mostly not equivalent for the pairwise comparisons. The average MAPE values for the 4 pairwise combinations of total daily summary scores were as follows: MAD (22.3%), MIMS (25.02%), AI (27.08%), ENMO (29.2%), AC (36.4%); However, based on ACT24 intensity classifications, MAPE values for MIMS and MAD were consistently low for periods of vigorous PA (8.3-9.4%) and moderate PA (13.8-14.4%) but considerably higher for Light PA (24-23.3%), Sedentary (26.4-23.6%) and Sleep (29.6-19.8%). The patterns were similar for the ENMO, AI, and AC but MAPE values were consistently higher and less consistent. The variability in agreement across intensity was confirmed by Bland-Altman plots where tighter relationships for MIMS and MAD were evident. CONCLUSION: The results supported the use of MIMS and MAD for agnostic monitoring of PA, but the weaker agreement for lower intensities should be noted. Performance was weaker for the other summary indicators. The differences may be attributed to pre-processing and standardization steps used.