Cardiovascular disease (CVD) is an umbrella term used to categorise a range of diseases affecting the heart and blood vessels. The most common CVDs include atrial fibrillation (irregular heartbeat), heart attacks (blocked blood flow to the heart blocked), and heart failure (heart unable to pump blood). CVD accounts for around one-quarter of all deaths in the UK.
Many factors can increase the likelihood of developing CVD, including high blood pressure (hypertension), diabetes, high cholesterol, smoking and obesity. These ‘risk factors’ are often controllable and early interventions can greatly reduce the risk of developing CVD. Additionally, a range of demographic, social and environmental factors such as sex, ethnicity and pollution are known to increase the risk of CVD, usually through impacting one or more CVD risk factors. In the UK, research exploring social and environmental factors and their impact on CVD risk and CVD development is lacking. Understanding associations between social and environmental factors and CVD could help with the development of early interventions to reduce the risk of CVD.
This study will use primary care and hospital medical records to explore which and how demographic, social and environmental factors are linked to developing CVD across the UK. We will then explore whether taking these factors into account when predicting a patients likelihood of developing CVD in the next ten-years improves the accuracy of prediction. We will also explore whether artificial intelligence (AI) based models perform better than traditional statistical models in predicting a patients risk of CVD.
Cardiovascular disease (CVD) is a significant cause of mortality and morbidity, accounting for a quarter of all deaths in the UK. Traditional CVD risk models tend to utilise clinical measures and limited demographic information. However, systematic reviews of CVD determinants have recognised a need to take a more holistic view of what constitutes ‘CVD risk’ in prediction models. There is growing evidence on the impact of social and environmental factors and the risk of developing CVD.
There is limited research on how social and environmental determinants impact risk of CVD, in a UK context and how the addition of these factors benefits CVD risk prediction tools. Incorporating social and environmental determinants into CVD risk prediction models could be helpful to set nonmedical interventions and to lower the social inequities in health.
Utilising electronic health records linked to ONS data and deprivation measures. This study will:
1. Assess associations between demographic, social and environmental factors and CVD development.
We will assimilate findings from both Cox proportional hazard models and survival random forests analysis to identify social and neighbourhood determinants for CVD.
2. Explore whether CVD risk prediction models accounting for social and environmental factors perform better than currently used models (using traditional statistical and ML models).
Using traditional statistical models (i.e. Cox proportional hazards model), we will introduce the social and neighbourhood variables and compare performance to that of a comprehensive ML prediction model that we will develop (utilising algorithms such as neural networks, random forest, support vector machines etc.).
3. Investigate whether CVD clusters around specific social and environmental factors and whether clustering varies by type of CVD.
Using ML clustering techniques, we will explore whether the risk of developing CVD tends to cluster around specific social and environmental factors. We will then stratify by the five most common CVDs.
The primary outcome will be incidence of any CVD (such as coronary heart disease, stroke, peripheral arterial disease etc.) amongst our study population.
The secondary outcomes will be subsequent (post index CVD diagnosis) major cardiovascular outcomes (such as heart attack, stroke, and heart failure, etc), non-cardiovascular outcomes (such as pneumonia, renal failure, liver disease, cancer, and dementia, etc), all-cause and cause-specific mortality.
Jianhua Wu - Chief Investigator - Queen Mary University of London
Jack Brown - Corresponding Applicant - Queen Mary University of London
Harriet Larvin - Collaborator - Queen Mary University of London
Paris Baptiste - Collaborator - Queen Mary University of London
HES Admitted Patient Care;ONS Death Registration Data;Patient Level Index of Multiple Deprivation;Practice Level Rural-Urban Classification