Stratification of diabetes in the context of comorbidities, using representation learning and topological data analysis

Each year, diabetes and the complications associated with it contribute to 1.5 million deaths globally. Diabetes is a highly heterogeneous disease with unpredictable clinical manifestations, variable disease progression, and a range of complications. Recognising distinct phenotypes during diagnosis holds the potential to facilitate personalised treatment strategies and enhance patient outcomes.

This study, led by Dr Gosia Wamil, consultant cardiologist at Mayo Clinic Healthcare in London, analysed data from 9,967 patients with both diabetes and multiple comorbid conditions, obtained from the Clinical Practice Research Datalink. The researchers employed a deep learning model called BEHRT, which is able to identify patterns and relationships among the patients' comorbidities.

Dr Wamil and the team identified four distinct phenotypes of diabetic patients, each with different comorbidity patterns and risk profiles. Phenotype 1 and 2 were relatively younger diabetic patients with chronic inflammatory conditions or coronary artery disease, while phenotypes 3 and 4 were older and had a higher frequency of pre-existing cardio-renal diseases.

Within ten years of follow-up, a significant percentage of patients experienced MACE, mortality, and renal failure.

The study enhanced the performance of the established risk prediction model QRISK3 by incorporating features derived from the topological data analysis. The addition of specific TDA-derived phenotypes and improved QRISK3's ability to predict cardiovascular outcomes.

Dr Wamil’s research highlights the importance of considering comorbid conditions, particularly cardiac conditions, when stratifying the risk of a diverse group of patients with newly diagnosed diabetes. The machine learning approach successfully improved the prediction of clinical outcomes in these patients. These findings have the potential to improve patient care and risk management for individuals with diabetes and multiple comorbidities.

Dr Wamil is an internationally recognised consultant cardiologist and Assistant Professor at Mayo Clinic Healthcare in London. She specialises in heart failure, cardiomyopathies, hypertension and non-invasive assessment of coronary artery disease and also leads the delivery of multi-modality cardiac imaging services at Mayo Clinic Healthcare.

Our cardiology team in London tailors care to your patient’s needs while working closely with Mayo Clinic's top-ranked team of consultants in the United States. Mayo Clinic Healthcare experts employ a range of sophisticated cardiac imaging equipment to ensure a full understanding of your patient’s condition. On-site diagnostics include cardiac MRI, CT coronary angiogram, echocardiogram, electrocardiogram (ECG) and transoesophageal echocardiogram (TOE).

To learn more about Mayo Clinic Healthcare’s state of the art cardiology services and how to make a referral, call +44 (0) 207 871 2575 or email

To learn more about this research and access the full article please click here.