Discovering Prevalent Chronic Disease Profiles with Advanced Clustering Methods

Jiani Yan , University of Oxford

Chronic non-communicable diseases (NCDs) have become pervasive, representing a significant public health challenge in the United Kingdom, particularly due to rising multimorbidity among older populations. Existing research often lacks a holistic, population-level perspective on multimorbidity patterns and associated social determinants. Using the UK Biobank dataset, this study employs a bottom-up analytical approach combining Uniform Manifold Approximation and Projection (UMAP) and Hierarchical Density-Based Spatial Clustering (HDBSCAN) to explore chronic disease patterns across 316 chronic conditions. Nine distinct multimorbidity clusters were identified, characterised by various dominant conditions: Healthy, Asthma, Hypertension, Allergic Rhinitis, Hypertension+Respiratory, Depression, Other Prevalent Diseases, and Heavy Cardiovascular Disease (CVD). Subsequent analysis of social determinants revealed significant differences between healthy and disease-affected groups, notably influenced by age, gender, and drinking behaviours. Clusters such as Heavy CVD showed particularly high disease burdens among older, predominantly male populations, while mental health-related clusters like Depression+ were closely linked with adverse psychosocial factors. This integrated approach highlights the necessity of considering comprehensive social determinants alongside detailed multimorbidity profiles to inform targeted public health interventions and personalised clinical practices.

See paper

 Presented in Session 51. Health, Morbidity and Wellbeing