The heavy burden of chronic diseases, such as cancer, diabetes, and cardiovascular disease, has become a major global public health challenge, severely impacting patients’ quality of life and creating immense socioeconomic pressure. Regular physical activity is a key intervention for improving physical and mental health and reducing disease risk. However, current research often relies on macro-level metrics like total volume and average intensity, which limit our deeper understanding of its health benefits. These broad measures can obscure critical details, failing to distinguish between short, vigorous bursts of activity and prolonged, moderate exertion, which may have different physiological impacts. We posit that more granular daily activity patterns, such as intensity variations, bout structures, and circadian rhythms, have independent health effects. Their complexity has made them difficult to analyze with traditional methods, but advanced machine learning now enables the systematic parsing of these fine-grained behavioral patterns from large-scale accelerometer data, opening a new frontier in behavioral epidemiology.
Based on this, our study has the following objectives:
1.To quantify granular physical activity features using machine learning and assess their prospective association with the risk of major chronic diseases.
2.To investigate the impact of different physical activity patterns on disease progression and long-term prognosis in individuals with diagnosed chronic conditions.
3.To integrate multi-omics data to identify biological pathways mediating the association between physical activity and health, as well as gene-activity interactions.
4.To evaluate whether incorporating these novel physical activity features into existing models can significantly improve the accuracy of chronic disease risk prediction.
This study aims to test the core hypothesis that the complex daily physical activity patterns are the key drivers of chronic disease risk.
