Cardiovascular disease in Type 2 Diabetes Mellitus: A precision medicine approach applying artificial intelligence for heart failure and mortality prediction

Kaur, Narinder (2025) Cardiovascular disease in Type 2 Diabetes Mellitus: A precision medicine approach applying artificial intelligence for heart failure and mortality prediction. PhD thesis, University of Glasgow.

Abstract

Cardiovascular diseases (CVDs) are the leading cause of morbidity and mortality worldwide, despite substantial advances in diagnosis and treatment. People who suffer from cardiovascular disease often have multiple risk factors and other chronic conditions. Additionally, medical events may be strongly influenced by socioeconomic status. Patient information can be obtained from electronic medical records (EMRs) that, unlike data from clinical trials and registries, provide a broad range of patient characteristics representative of the general population. EMRs covering a population of ~1.1 million people in Greater Glasgow & Clyde (GG&C) Health Board NHS over 50 years (the age at which the incidence and prevalence of disease affecting older people increase rapidly) were used. Information such as demographics, laboratory tests, primary-care prescriptions, hospitalisations and mortality was retrieved. Several steps were required to ensure that the extracted information was appropriate for analysis and transformed for investigations beyond traditional statistics. Accordingly, data on patients with type-2 diabetes mellitus (T2DM) were obtained to examine their health trajectories, including, incident heart failure and death. Novel risk prediction models were built to help understand the development of heart failure (HF) in patients with T2DM. The models were developed using random survival forest (RSF) methodology. This research highlights the limitations of traditional regression models and demonstrates the improvement of risk prediction with RSF methods, which outperformed traditional approaches in both discrimination and calibration. State-of-the-art machine learning interpretation was applied to discover key contributing factors to the development of heart failure and to all-cause mortality. External validation was applied by acquiring EMRs from Hong Kong, Special Administrative Region (SAR) China. The inclusion of two diverse populations found little evidence of ethnicity-related differences in risk factors. GG&C key risk factors for incident HF were loop diuretics, atrial fibrillation (AF), history of coronary artery disease (CAD), older age, lower levels of estimated glomerular filtration rate (eGFR), haemoglobin and serum albumin. Similarly, for Hong Kong, key risk factors were use of loop diuretics, insulin, lower serum albumin, haemoglobin, lymphocyte counts and eGFR. The model based on Hong Kong data showed slightly better performance compared to the Glasgow cohort for incident heart failure (C-index 0.88 and 0.87) and all-cause mortality (0.85 and 0.83). In both cohorts’ older women were more likely to be prescribed loop diuretics. Whether loop diuretics are just a marker of undiagnosed heart failure or whether they accelerate the progression of cardiovascular and renal disease is uncertain. Another key similarity was that patients had prevalent chronic kidney disease (CKD) events in the prescribed loop diuretics groups. Treatment with loop diuretics was strongly associated with all-cause mortality in GG&C and Hong Kong. (GG&C: adjusted hazard ratio: 2.93, (95% CI: 2.821 to 3.04); Hong Kong: adjusted hazard ratio: 1.75 (95% CI: 1.72 to 1.77). Only a minority of patients prescribed loop diuretics had a diagnosis of heart failure, end-stage renal disease or resistant hypertension. Finally, further investigation of social deprivation in GG&C underlined that 41% patients with T2DM were in the most deprived socioeconomic quintile and that they had a 36% higher rate for all-cause mortality compared to those who were least deprived (adjusted HR: 1.36, 95% CI 1.24–1.50, p < 0.005).

• Item Type: Thesis (PhD)
• Qualification Level: Doctoral
• Subjects: R Medicine > R Medicine (General)
• Colleges/Schools: College of Medical Veterinary and Life Sciences > School of Cardiovascular & Metabolic Health
• Supervisor's Name: Cleland, Professor John G.F., Deligianni, Dr. Fani and Pellicori, Dr. Pierpaolo
• Date of Award: 2025
• Depositing User: Theses Team
• Unique ID: glathesis:2025-85292
• Copyright: Copyright of this thesis is held by the author.
• Date Deposited: 03 Jul 2025 14:10
• Last Modified: 03 Jul 2025 14:31
• Thesis DOI: 10.5525/gla.thesis.85292
• URI: https://theses.gla.ac.uk/id/eprint/85292