Associations between sensor-based physical activity behaviour features and health-related parameters

Area of Science:

• Biomedical Engineering
• Public Health
• Data Science

Background:

• Wearable actimetry devices are increasingly used for objective physical activity (PA) assessment in research.
• The clinical significance of distinct PA groups identified through advanced sensor data analysis is not well-established.
• This study investigates the relationship between objectively measured PA patterns and health risk parameters.

Purpose of the Study:

• To determine if distinct physical activity (PA) groups, identified via sensor data analysis, differ in key health risk indicators.
• To evaluate the medical relevance of novel PA group classifications.

Main Methods:

• Utilized data from the NHANES 2005-06 study, including physical activity sensor data and health outcomes.
• Applied data pre-processing, including outlier elimination and feature computation, incorporating a novel PA regularity measure.
• Employed the x-Means clustering algorithm to identify PA groups and analyzed differences in C-reactive protein (CRP), body mass index (BMI), and HDL cholesterol.

Main Results:

• Analyzed data from 7,334 participants, identifying four distinct physical activity (PA) groups.
• Found statistically significant differences in CRP and BMI across the identified PA groups (p<0.001).
• No significant differences in HDL cholesterol levels were observed between the PA groups (p=0.67).

Conclusions:

• Physical activity (PA) groups derived from objective accelerometer data exhibit significant variations in health risk parameters.
• A novel PA regularity measure shows potential for future PA assessments, particularly for low-intensity activities.
• Further research into pattern recognition and analytical algorithms for multi-sensing PA devices is warranted.