Measurements of certain 'risk factors' are currently used to estimate how likely someone is to develop cardiovascular disease, and to prioritise people for intervention. However, the personalised use of these risk factors could be greatly improved by the application of machine learning methods. These techniques will allow for systematic identification of patients at high risk of future disease, improving the accuracy of intervention and treatment.
One of six British Heart Foundation funded projects.
Explaining the science
Cardiovascular disease is the UK's leading cause of death and disability, and overall burden of it on health services is predicted to increase as a result of an ageing population and increasing prevalence of risk factors such as obesity and diabetes. Although the UK has introduced a national screening programme (NHS Health Check) for adults without a history of cardiovascular disease aged 40-74 years, many of its features have not been founded on robust evidence and could be optimised (or re-designed) with the emergence of new data.
This project will use novel deep learning and statistical machine learning methods for modelling trajectories of historically recorded risk factors available in primary care records to systematically identify patients at high risk of future disease and to incorporate competing risks for modelling the risk of cause-specific cardiovascular disease. These novel methods will be applied to large and complex datasets such as the Clinical Practice Research Datalink, the European Prospective Investigation into Cancer and Nutrition (EPIC), and the UKBiobank.
Cardiovascular disease 'risk factors' include being of an older age, male, smoking, high blood pressure, and high cholesterol. This project aims to address key unresolved questions in the use of these risk factors for identifying individuals that are more likely to suffer from cardiovascular events. These include:
- What is the value of using repeated information on cardiovascular disease risk factors already available in primary care records?
- How can we identify individuals at greatest risk of experiencing certain type of cardiovascular diseases so that the most suitable intervention could be used?
These questions will be addressed by adapting cutting-edge statistical and machine learning methods, and applying these techniques to powerful relevant datasets, such as computerised general practice records and large population studies with detailed biological information. These techniques will systematically identify patients at high risk of future disease and allow for the prioritisation of people for intervention (e.g. medication, lifestyle advice, or referral).
Such advances will have benefits for both public health and healthcare resource use, and are complementary to population-level interventions related, for example, to diet and exercise.