A Reproducible Pipeline for Processing Commercial Wearable Step-Count Data in Aging Cohorts: Application and...

Area of Science:

• Gerontology
• Epidemiology
• Biomedical Engineering

Background:

• Wearable devices offer objective physical activity monitoring.
• Raw step-count data requires systematic processing for valid inference.
• Commercial wearable data lacks standardized processing pipelines.

Purpose of the Study:

• To describe a reproducible standard operating procedure (SOP) for processing wearable step-count data.
• To apply this SOP to Garmin device data in the STRRIDE-PD Reunion study.
• To compare wearable-derived physical activity with self-reported exercise in relation to health outcomes.

Main Methods:

• Developed a data pipeline to transform epoch-level step-count data into participant-level analytic variables.
• Standardized timestamps, reconstructed daily epoch grids, and inferred wear time.
• Applied a valid-day threshold (≥10 hours inferred wear time) for participant summaries.

Main Results:

• Wearable-derived step counts were associated with 9 of 16 cardiometabolic and fitness outcomes.
• Self-reported exercise was not significantly associated with any outcome.
• Regression calibration showed self-report systematically underestimated physical activity-health associations.

Conclusions:

• Measurement approach significantly impacts scientific conclusions in physical activity research.
• Reproducible wearable data pipelines are essential for aging epidemiology.
• Objective wearable data provides more robust associations with health outcomes than self-report.