INTRODUCTION:
Wearables are increasingly used in population health research to objectively measure physical activity and sleep and to study their links to non-communicable diseases. Most large studies rely on a single 7-day wearable assessment, but it remains unclear how well this reflects individuals' usual movement behaviours. Here, we aimed 1) to assess the long-term reproducibility (i.e., agreement of repeat measurements within the same individuals over time) of accelerometer-derived physical activity and sleep, and 2) to demonstrate how within-person variability can attenuate observed associations between movement behaviours and health outcomes.

METHODS:
We analysed repeat accelerometer data from 3138 UK Biobank participants with up to four measurements collected over 3-4 years. Nine physical activity and sleep phenotypes were derived, and reproducibility was assessed using intraclass correlation coefficients (ICCs). To illustrate implications for disease risk, we estimated hazard ratios (HRs) for daily step count and incident coronary heart disease (CHD) using Cox regression (87038 participants; 3879 CHD events), before and after correction for within-person variability.

RESULTS:
Among the 3138 participants, 51% were women, and the mean (SD) age was 63.1 (9.4) years. Reproducibility was good for overall activity (ICC 0.75) and moderate for other behaviours, including sedentary time and sleep characteristics (ICCs 0.58-0.69). In our example, the inverse association between daily step count and CHD showed a 20% lower risk of CHD per usual 4000 steps (HR 0.80, 95% CI 0.76-0.85) after correcting for within-person variability, compared to 13% (HR 0.87, 95% CI 0.84-0.90) before correction.

CONCLUSIONS:
Wearable measurements show moderate-to-good stability over time. Ignoring within-person variability can lead to underestimation of associations between movement behaviours and health outcomes. These findings have important implications for population health research and prevention, suggesting that the health benefits associated with physical activity and sleep may be greater than previously estimated when usual behaviours are appropriately characterised.