Processing UK Biobank High Resolution Accelerometry Data for Unsupervised Identification of Activity Profiles and...

Area of Science:

• Biomedical Engineering
• Physical Activity Research
• Data Science in Health

Background:

• Wearable devices generate objective, long-term accelerometer data on activity behavior.
• The Activity Types from Long-term Accelerometric Sensor data (ATLAS) index simplifies interpretation of this data.
• UK Biobank offers high-quality, extensive datasets for health research.

Purpose of the Study:

• To assess the feasibility of the ATLAS index with UK Biobank data for identifying activity behavior groups.
• To investigate if clinically relevant parameters differ between identified activity groups.
• To explore objective accelerometer data for health-related insights.

Main Methods:

• Processed raw accelerometer data from 6,400 UK Biobank subjects.
• Computed ATLAS index parameters: regularity, intensity, and duration of moderate-intensity physical activity.
• Applied hierarchical clustering to group subjects based on activity patterns.
• Evaluated differences in HDL cholesterol, BMI, and C-Reactive Protein (CRP) between clusters.

Main Results:

• Identified five distinct physical activity clusters using hierarchical clustering.
• Found statistically significant differences in HDL cholesterol, BMI, and CRP levels among these clusters.
• Demonstrated that ATLAS index parameters effectively differentiate physical activity behaviors.

Conclusions:

• The ATLAS index enables identification of distinct physical activity groups from objective accelerometer data.
• These identified groups exhibit variations in physiologically relevant health parameters.
• Further research, including causal inference, is needed to establish causal links between activity groups and health outcomes.