Improving disease subtyping and physiological characterization of adult-onset diabetes in electronic health records

ABSTRACT Diabetes is broadly classified as type 1 diabetes (T1D), with primary defect in insulin production due to autoimmune destruction of beta cells, and type 2 diabetes (T2D), with primary defect in insulin sensitivity in organs that regulate energy metabolism. There is an emerging conceptual change that diabetes is not composed of discrete sets of syndromes but a spectrum of heterogenous phenotypes with complex pathophysiology across different ethnic and racial populations. More than 90% of adult patients with diabetes are classified as T2D with variable clinical risk factors, prognosis, and treatment response. There is an unmet need to assess individual risk for developing diabetes, its complications, and drug response based on underlying pathophysiology, which change over time. This proposal will embark on studies of large-scale electronic health records (EHRs) with comprehensive phenome assessed during the course of diabetes, and multi-omics integration, to address the following Specific Aims. AIM 1: Classification of diabetes subtypes using clinical features in EHRs of diverse ancestries. AIM 2: Classification of diabetes subtypes in response to treatments for diabetes and cardiometabolic diseases in EHRs of diverse ancestries. AIM 3: Multi-omics characterization of diabetes subtypes in African Americans. Successful implementation of the proposal will build a framework to identify subtypes of diabetes and their underlying physiological drivers to impact clinical practice towards more precise diagnosis, prognosis, and effective intervention.