Uncovering Clusters of Social Determinants and Chronic Diseases: A Deep-Learning Approach

Jiani Yan, University of Oxford

This study employs a deep-learning algorithm to identify clusters of social determinants and chronic diseases using UK Biobank data collected at the time of first recruitment. The dataset includes 262 individual social factors—ranging from socioeconomic status, education and employment, Cognitive function summary, family history to lifestyle habits—and hospital inpatient records spanning the period prior to recruitment. Our primary objective is to investigate the impact of these social determinants on chronic disease outcomes. By applying an assumption-free, data-driven deep-learning approach, we aim to uncover complex, non-linear relationships between social factors and disease patterns that are often missed by traditional methods. Theoretically, this research also offers valuable insights into how specific social conditions contribute to disease progression, particularly across gender and age groups. Our preliminary results highlight clear differences in disease prevalence based on age and gender. The different disease counts by age, gender and education levels indicate that men and women may exhibit distinct disease patterns under similar social conditions. Additionally, the network mapping of disease connections reveals potential pathways for disease transitions, offering a novel way to understand chronic disease associations and progression over time.

Keywords: Health and Morbidity, Computational social science methods, Digital and computational demography

See extended abstract.