Harmonization of Open-Source and Proprietary Accelerometry-Based Physical Activity Measures

• Activity Number: 190 - Novel Approaches for Assessment of Health Outcomes and Multi-Cohort Data Integration Using Wearable Devices in Large-Scale Biomedical Studies
• Type: Topic Contributed
• Date/Time: Monday, August 8, 2022 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Marta Karas* and John Muschelli and Andrew Leroux and Jacek K. Urbanek and Amal A. Wanigatunga and Jiawei Bai and Ciprian M. Crainiceanua and Jennifer A. Schrack
• Companies: Johns Hopkins University and Johns Hopkins University and University of Colorado and Johns Hopkins University and Johns Hopkins University and Johns Hopkins University and Johns Hopkins University and Johns Hopkins University
• Keywords: Wearables; Accelerometry; Physical activity measures; Open-source software

Abstract

The use of open-source accelerometry-based PA measures allows for reproducible and generalizable data collection, processing and analysis. However, their comparability with widely used proprietary measures such as ActiGraph activity counts (AC) is unknown.

We used subsecond-level accelerometry data collected from 655 participants in the Baltimore Longitudinal Study on Aging who wore ActiGraph GT9X Link device at wrist continuously for a week. Data were summarized at the minute level as ActiGraph AC and open-source MIMS, ENMO, MAD, and AI. Each pair of measurements were harmonized using nonparametric regression.

The marginal participant-specific correlation between AC and MIMS, ENMO, MAD, and AI were 0.988, 0.867, 0.913 and 0.970, respectively; significant yet small size effects of age, BMI, and sex on the correlations were observed. After harmonization, the mean absolute percentage error for predicting total activity counts from MIMS, ENMO, MAD, and AI was 2.5, 14.3, 11.3 and 6.3, respectively.

This measurement harmonization can be used to compare results from different studies and improve the translation and generalizability of accelerometry studies.