Abstract
OBJECTIVE: Effective physical activity interventions for knee osteoarthritis (OA) require an understanding of the relationship between physical activity and pain. Using daily concurrent activity and pain measurements, we explored day-to-day changes and the bidirectional temporal association at individual and population level. DESIGN: This is a secondary analysis of step count and pain collected for 90 days using smartwatches in 26 people with knee OA. People reported pain twice daily on a Numerical Rating Scale (NRS 0-10). We used regression (individual level) and generalized linear mixed models (population) to explore same-day associations, as well as whether step count on one day predicted pain the next day, and vice versa. RESULTS: We analysed 1473 daily pain and step count measurements, recorded over a median 58 days. There were considerable day-to-day changes in individuals' median step count (range 423-7142) and pain (range 0-9). At individual level, associations varied in the strength and direction. At population level, a higher step count was associated with higher pain on the same day (0.04 NRS/1000 step increase, 95%CI 0.01-0.06) and following day (0.05/1000 step increase, 0.03-0.07). CONCLUSIONS: There was a modest association at population level between step count assessed on one day and pain assessed on the same and following day. However, there was variation in the strength and direction of associations when examined at the individual level. This exploratory analysis shows how smartwatches allow daily data collection that enables detailed exploration of complex time-varying relationships in OA.