Influence of pre-processing criteria on analysis of accelerometry-based physical activity

Area of Science:

• Biomedical Engineering
• Public Health
• Sports Science

Background:

• Accelerometers are standard tools for measuring physical activity (PA).
• Pre-processing accelerometry data is crucial before statistical analysis.
• Inconsistent pre-processing can significantly alter PA outcomes and study findings.

Purpose of the Study:

• To investigate how different accelerometer data pre-processing criteria affect physical activity study outcomes.
• To assess the impact of wear-time, intensity, and bout definitions on PA metrics.

Main Methods:

• Utilized ActiGraph accelerometry data from 538 Latino adults.
• Examined four key pre-processing domains: wear-time, minimum wear-time, intensity level, and modified bouts.
• Analyzed effects on sample size, moderate-to-vigorous physical activity (MVPA) outcomes, and regression models for age and gender.

Main Results:

• Most pre-processing criteria had minimal impact on PA outcomes.
• Stringent minimum wear-time requirements substantially influenced MVPA analysis and reduced statistical power.
• Intensity calculations using vector magnitude differed significantly from those using vertical axis counts.

Conclusions:

• Adjustments to minimum wear-time criteria can alter MVPA behavior analyses, impacting intervention assessments.
• Vector magnitude and vertical axis intensity data are not interchangeable.
• Recommends sensitivity analyses to ensure the reliability of accelerometry data findings.