Associating physical function and capacity tests to free-living sensor data: a systematic review on technology and methods

Physical function and capacity tests are widely used for assessing health across various clinical conditions. However, traditional assessments may not accurately capture real-world health conditions reliably and frequently. Sensors, smartphones and wearable devices offer the potential to bridge this gap by collecting data in everyday life that may better reflect participants’ physical capabilities, and could be used to predict clinical outcomes and the performance of physical tests. However, there is a lack of comprehensive reviews and consensus in the field. This work reviews the literature on passively collected data from digital health technology in relation to physical function and capacity tests and informs future investigations in this domain. A systematic literature search was conducted following the PRISMA guidelines on 3 databases. Our analysis identifies cardiovascular and neurodegenerative diseases as the most frequently studied conditions, and wearables embedding inertial sensors as the most common device type. Most studies rely on one week-long data collection. Associations between physical test outcomes and metrics such as step count and activity intensity show correlations as high as 0.89 when machine learning is introduced. This review provides a comprehensive summary of current research on the use of digital health technology in free-living conditions and the clinical significance of data when associated with physical tests.