Skeleton avatar technology as a way to measure physical activity in healthy older adults

Abstract

Abstract Background Nowadays, self-reported assessments (SA) and accelerometer-based assessments (AC) are commonly used methods to measure daily life physical activity (PA) in older adults. SA is simple, cost-effective, and can be used in large epidemiological studies, but its reliability and validity have been questioned. Accelerometer measurement has proven valid to provide accurate and reliable measurement of everyday life physical activities regarding frequency, duration, and intensity in older populations, but is expensive and requires a long-time measurement. Here is, furthermore, a lack of well-defined and reliable accelerometer cut-off points to measure PA among older adults. Therefore, there is a need to develop a simple and reliable method to complement/replace self-assessment methods of daily life physical activity and facilitate the future development of cut-off points to measure daily life physical activities among older adults. In this study, we explore how skeleton avatar technology (SAT) can be used to measure PA among older adults. Objectives 1. To explore the association between accelerometer data and self-reported assessment data of daily life physical activities in older adults, and 2. To explore how the SAT of a standardized functional (balance) test can be used to measure daily life physical activity among older adults. Method The correlation analysis was used to explore the association between response variables, and deep neural networks were used to predict the response variables (AC and SA outcomes). Results The results indicate that there is a moderate (r = 0.31) significant (p = 0.029) correlation between AC of PA and SA of PA. The functional balance test assessed with SAT was able to predict AC with 3.89% Mean Absolute Error (MAE), and SA with 11.07% MAE. Conclusion Overall, these results indicate that one functional balance test measured with SAT can be used to predict PA outcomes measured with accelerometer devices. SAT can predict PA outcomes better than SA outcomes within the same population. More research is needed to explore the ability of SAT predicting PA among older adults with various functional abilities, and how SAT can be developed using 2D recordings, such as mobile phone recordings, to predict PA efficiently.