Comparing Step-Counting Algorithms for High-Resolution Wrist Accelerometry Data in Older Adults in the ARIC Study.

Area of Science:

• Gerontology
• Biomedical Engineering
• Public Health

Background:

• Wrist accelerometry is key for physical activity research.
• Open-source algorithms exist for step counting, but comparisons are needed for older adults.

Purpose of the Study:

• To compare five open-source step-counting algorithms using wrist accelerometry data.
• To provide recommendations for algorithm use in older adults.

Main Methods:

• 1,282 participants (mean age 83.4) wore ActiGraph GT9X devices for 7 days.
• Five algorithms (ADEPT, Oak, SDT, Verisense, Stepcount) were applied to 80-Hz triaxial accelerometry data.
• Step count distributions and associations with health outcomes were analyzed.

Main Results:

• Mean daily step counts varied significantly across algorithms (988-23,607 steps).
• All algorithms showed strong correlations (r=0.52-0.96) and similar associations with health outcomes after z-scoring.
• Step counts declined with age (119.0-142.8 steps/year).

Conclusions:

• Step-counting algorithms are highly correlated and show similar health associations post-standardization.
• The wide variation in raw step counts necessitates careful algorithm selection for older adults.
• Algorithm choice impacts interpretation for health monitoring and clinical applications.