Personalized physical activity recommendations for people with axial spondyloarthritis using wearable activity tracker data: an exploratory study

Objective Benefits of physical activity (PA) on sleep in people with axial SpondyloArthritis (axSpA) are largely unknown. Our aim is to explore the relationships between PA and sleep on both a group level and an individual level using Wearable Activity Trackers (WATs) and machine learning. Methods A sample of 64 axSpA participants received a WAT to monitor their PA and sleep. Participants with more than 30 days data of PA and sleep duration were included in the analyses. Spearman's correlation and the machine learning technique Subgroup Discovery were used to determine relationships between PA during the three prior days and light and deep sleep duration. Results Number of daily steps (n = 64) was (median (first quartile (Q1) - third quartile (Q3) )) 4026 (1915 - 6549), total sleep (daily light and deep sleep) duration of the participants was 7 h 29 min (6 h 41 min - 8 h 8 min). Nearly 30% (n = 18) of the participants were eligible for inclusion in analyses (> 30 days of data). No significant relationships between prior PA and sleep were obtained on a group level. On an individual level, for 8 of the 18 included participants, significant relationships (p < 0.05) could be identified between PA during the three prior days and daily sleep duration. These significant relationships differed from participant to participant with a varying qualification of PA (number of steps, intensity level PA) and relevant time window (previous one, two or three days). Conclusion Significant relationships between PA and daily sleep duration could be obtained on an individual level with details of the significant relationships varying between participants.