Using Accelerometer Data to Identify Physical Activity Profiles in the Osteoarthritis Initiative

Abstract

Introduction: Osteoarthritis is a chronic joint disease with no current cure, affecting around 500 million people worldwide. Evidence suggests that physical activity has beneficial effects, but the ideal “prescription” of exercise remains unknown. The Osteoarthritis Initiative is a longitudinal, prospective, observational study with over 10 years of follow-up with accelerometer data from 2712 subjects. In this project, we used a data-driven approach to identify physical activity patterns based on daily accelerometer information. Using these patterns, we established profiles and determined if associations exist between the created profiles and known correlates and symptomatic outcomes of osteoarthritis such as pain.
 Methods: Physical activity curves for subjects with accelerometer data from the Osteoarthritis Initiative were aligned using curve registration methods. Once curves were aligned, subjects were grouped into profiles using k-medoid clustering. The outcome of pain was measured by the total Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) score. A linear regression model was used to determine if an association exists between the defined profiles and the total WOMAC score.
 Results: K-medoid clustering of physical activity curves resulted in two profiles. However, there was no association between the defined profiles and the total WOMAC score at the alpha level of 0.05.
 Discussion: The developed profiles are not associated with the total WOMAC score. While the created profiles are not associated with symptomatic outcomes, it is of interest to explore if they provide prediction abilities for structural outcomes such as joint space width and radiographic classification of osteoarthritis such as the Kellgren-Lawrence score.