Cardiovascular health

We have developed a machine learning algorithm called ‘FIND-AF’ (Future Innovations in Novel Detection of Atrial Fibrillation), with collaborators at the University of Leeds. FIND-AF uses the data routinely held in patient health records to identify people who may develop the heart condition atrial fibrillation in the next six months. 

Atrial fibrillation causes an irregular heartbeat and blood clots and carries a fivefold increase in the risk of stroke if a patient is not given anticoagulant medication. Using the anonymised electronic health records of over 2.1 million people, Professor Jianhua Wu’s team trained the FIND-AF algorithm to find warning signs that suggest patients are at high risk of developing atrial fibrillation within the next six months, meaning they can be approached by their GP for testing and potential treatment. FIND-AF has been validated on large datasets in England (using linked data from primary and secondary care via CPRD), Canada, Hong Kong, Israel and Japan. 

We are using big data from electronic health records to analyse treatments, outcomes, hospitalisations and deaths for people with atrial fibrillation. By including demographic information from health records in the analyses, we can study inequalities in outcomes, and establish whether particular treatments are working less well for some groups of people. 

We are using process mining techniques to model how different diagnosis and treatment pathways are used in practice. And hypergraphs – a type of network analysis – to analyse the influence of AF on developing other conditions. 

The QRISK prediction model, developed by our colleague Julia Hippesley-Cox, estimates a patient’s 10-year risk of developing a cardiovascular disease such as heart attack or stroke. We are working to expand the model to include wider determinants of health (initially deprivation and local air pollution) and see whether this improves the model’s ability to predict cardiovascular disease. 

To study the effect of air pollution, we are working with data from SAIL Databank in Wales. This dataset holds information on people’s health, and their interactions with government and public services.

REAL-Health Cardiovascular is a major research and clinical effectiveness programme, initiated by funding from Barts Charity in 2018. Through this programme, Dr John Robson and his team produced guidelines, educational materials and clinical software tools, including the CEG Atrial Fibrillation Tool. These resources focus on optimising medications to control blood pressure, prevent heart attacks and strokes, and reduce gastrointestinal bleeding in patients on antithrombotic treatment. Since its launch in 2016, the CEG Atrial Fibrillation Tool (formerly known as 'APL-AF') has been used extensively in east London and more widely, including in several award-winning initiatives.