Comparison of physical activity measures using mobile phone-based CalFit and Actigraph

Area of Science:

• Biomedical Engineering
• Epidemiology
• Public Health

Background:

• Objective measurement of physical activity is crucial for epidemiological studies.
• Existing tools are often expensive, lack objectivity, or are not reproducible.
• Smartphone sensing technology presents a potential solution for accessible physical activity monitoring.

Purpose of the Study:

• To validate physical activity estimates from the smartphone-based CalFit software.
• To assess the usability of CalFit for large population-based studies.
• To compare CalFit's physical activity measures against a research-grade accelerometer.

Main Methods:

• 36 participants in Barcelona wore a smartphone with CalFit and an Actigraph GT3X accelerometer for 5 days.
• Compared usability and physical activity metrics (vertical axis counts, duration, energy expenditure) using statistical analyses.
• Utilized Kruskal-Wallis, Spearman correlation, linear regression, CCC, and Bland-Altman plots.

Main Results:

• High linear association and correlation for vertical axis counts (adjusted R(2)=0.85, correlation coefficient=0.932).
• High agreement (CCC from 0.83 to 0.91) for duration and energy expenditure measures.
• 69% of participants successfully recorded sufficient physical activity data with CalFit.

Conclusions:

• CalFit provides valid estimates for quantifying and classifying physical activity.
• Lower usability compared to Actigraph GT3X due to auto-start issues.
• CalFit is a potentially cost-effective and deployable tool for large-scale population health studies given widespread smartphone ownership.