Olvera Barrios, Abraham; (2025) Statistical modelling for prediction of diabetes complications. Doctoral thesis (Ph.D), UCL (University College London).

Preview

Text Olvera Barrios_10214437_thesis.pdf Download (3MB) | Preview

Abstract

Background: Diabetes is a major health issue affecting 10% of the population, causing significant morbidity and healthcare costs, with NHS expenditures exceeding £1.5 million per hour. Diabetic retinopathy (DR), a common microvascular complication of diabetes, doubles healthcare costs compared with people living with diabetes (PLD) but no DR. In 2021, there were 6.7 million global diabetes-related deaths, a third among working-age individuals. Notably, not all PLD develop complications, and times-to-progression vary. Accurate risk stratification at the point of care could improve resource allocation, recruitment for clinical trials, and outcomes, yet current approaches remain limited. Objectives: To utilise diabetic eye screening (DES) and electronic health records (EHR) data to: i) quantify visual impairment in people with DR, ii) identify sociodemographic factors associated with DES non-attendance and sight-threatening DR, and iii) develop predictive models for sight-threatening DR and all-cause mortality. Methods and results: Visual impairment prevalence and certification rates among people with DR were quantified using ophthalmic EHR data (aim i). Sociodemographic factors linked to DES non-attendance were identified, and prognostic factors for sight-threatening diabetic retinopathy (STDR) were examined within a large, ethnically diverse cohort from DES using multivariable logistic regression. Transition probabilities to STDR were calculated (aim ii). I linked EHR data to an identified cohort undergoing DES in the North East of London between 01/01/2012 to 31/12/2021 to develop and rigorously evaluate 5-year Cox regression predictive models for sight- threatening DR and mortality (aim iii). Conclusion: I have shown visual impairment in people with DR remains a significant public health problem, underscoring a need of continued resource allocation. By inte- grating DES and EHR data, this work provides valuable insights into critical points in the diabetes disease course. It aims to identify high-risk individuals early, enabling targeted interventions and improved strategies for clinical trial recruitment, ultimately enhancing diabetes care and outcomes.

Download activity - last month

Export as XMLJSONCSV

Download activity - last 12 months

Export as JSONXMLCSV

Downloads by country - last 12 months

Export as CSVXMLJSON

Archive Staff Only

View Item