Keele Research Repository

Introduction: The increasing prevalence of type 2 diabetes (T2DM) presents a significant burden on affected individuals and health-care systems internationally. There is, however, no agreed validated measure to infer diabetes severity from electronic health records (EHRs). We aim to quantify T2DM severity and validate it using clinical adverse outcomes.

Methods and Analysis: Primary care data from the Clinical Practice Research Datalink (CPRD), linked hospitalisation and mortality records between April-2007 and March-2017 for T2DM patients in England will be used to develop a clinical algorithm to grade T2DM severity. The EHR-based algorithm will incorporate main risk factors (severity domains) for adverse outcomes to stratify T2DM cohorts by baseline and longitudinal severity scores. Provisionally, T2DM severity domains, identified through a systematic review and expert opinion are: diabetes duration, HbA1c, microvascular complications, comorbidities, and co-prescribed treatments. Severity scores will be developed by two approaches: i) calculating a count score of severity domains; ii) through hierarchical stratification of complications. Regression models estimates will be used to calculate domains weights. Survival analysis for the association between weighted severity scores and future outcomes: cardiovascular events; hospitalisation (diabetes-related, cardiovascular); and mortality (diabetes-related, cardiovascular, all-cause mortality) will be performed as a statistical validation. The proposed EHR-based approach will quantify the T2DM severity for primary care performance management and inform the methodology for measuring severity of other primary care-managed chronic conditions. We anticipate that the developed algorithm will be a practical tool for practitioners, aid clinical management decision-making, inform stratified medicine, support future clinical trials and contribute to more effective service-planning and policy-making.