The intensity of free-living stepping activities: influence of key methodological choices on the magnitude and precision of its associations with health outcomes

Abstract

Steps are an intuitive and widely used measure of ambulatory movement, yet formal step-based health guidelines remain absent. While daily step amount has been extensively studied, stepping intensity has been overlooked because of methodological challenges in free-living settings. This thesis addresses research gaps through four studies. First, two cadence-based methods (one-level and two-level cadence) were compared to a machine learning (ML) method to validate physical activity intensity classification. Cadence methods based on minute-level time window performed relatively acceptably at the group level but poorly at the individual level, suggesting infeasibility in free-living intensity estimation. Second, these methods were used to compare associations with mortality, indicating that the ML approach may demonstrate superior accuracy in reflecting potential mortality gain. Third, the thesis compared various stepping intensity metrics and found that there was no material difference among the peak-cadence metrics and average cadence metrics in terms of the association with mortality risks. Fourth, the joint association of stepping intensity and daily step amount was examined, demonstrating that higher stepping intensity may prevent premature mortality in adults, particularly among those with daily step amount of less than 5,000. In summary, this thesis provides methodological insights into the appropriate metrics of stepping intensity in a free-living setting and supports the importance of higher stepping intensity for public health, especially for physically inactive individuals.