Heart and circulation diseases cause a quarter (>160,000) of all deaths in the UK every year. In England, healthy people between 40 and 74 years old can have an NHS Health Check, where information is entered into a risk calculator that works out that person’s risk of having a stroke or heart attack in the next 10 years. If their risk is 10% or higher, they discuss ways to reduce this risk with their GP. The health check has reduced the number of people who have had a stroke, and spurred people on to make healthy lifestyle changes. Despite this, many people do not fully understand their risk score, or find the lifestyle advice they are given too impersonal.
Research suggests that people are more likely to change their behaviour if it is clear how taking specific actions will reduce their risk, and if they are supported to take these actions. Unfortunately, existing risk calculators are not designed to work out how much, e.g., losing weight or reducing blood pressure, would reduce someone’s risk of having a stroke or heart attack.
We will develop an improved risk calculator using a statistical technique called ‘causal inference’. This makes it possible to ask ‘what if?’ questions, like ‘what if I lost 2 stone?’, and will give accurate estimates of how much lifestyle changes (such as losing weight or stopping smoking) could reduce someone’s risk of a stroke or heart attack.
Primary prevention of cardiovascular and cerebrovascular diseases (CVD) in England involves provision of lifestyle advice and use of a CVD risk calculator (QRISK2) to guide statin prescribing. Psychological research suggests individuals are more likely to change behaviour if: risk information is personally relevant, it is clear how doing so will reduce risk, and they are supported appropriately. Models like QRISK can help identify patients at high risk, and whilst clinicians use these to illustrate the reduction in risk associated with intervention, there are flaws in this approach.
We will use causal inference to calculate more precise and interpretable absolute risks following lifestyle changes and medical interventions, in an open adult cohort, using primary care data to inform on risk factors and medications. Incident CVD (outcome) will be derived from Aurum primary care data and hospital episode statistics,. The study period is from 1/1/2005 to 31/12/2021. Individuals enter the cohort at the latest of: start of study period, attaining at one year of registration with an Aurum practice, and attaining age 18. Individuals exit the cohort at the earliest of: end of study period, diagnosis of CVD, death, deregistration with practice, and last data upload.
The outcome is time to first CVD event and for prediction we will use flexible parametric survival models, Cox regression models, and pooled logistic regression models. Marginal structural models and targeted maximum likelihood estimation will be used to estimate to causal effects of medications, while causal effects of lifestyle interventions will primarily be imported from other sources (such as randomised controlled trials). The causal estimates and absolute risks will be combined to produce a decision support tool that estimates risk of CVD under a range of lifestyle and medical interventions, which could be used to improve primary prevention of CVD.
Primary outcome: Incident cardiovascular and cerebrovascular disease (composite outcome of coronary heart disease, ischaemic stroke, or transient ischaemic attack)), referred to as CVD in this application (in common with previous tools such as QRISK).
Matthew Sperrin - Chief Investigator - University of Manchester
Alexander Pate - Corresponding Applicant - University of Manchester
Alexander Pate - Collaborator - University of Manchester
Brian McMillan - Collaborator - University of Manchester
David Stables - Collaborator - Endeavour Health Charitable Trust
Evangelos Kontopantelis - Collaborator - University of Manchester
Joyce (Yun-Ting) Huang - Collaborator - University of Manchester
Kritchavat Ploddi - Collaborator - University of Manchester
Maurice O'Connell - Collaborator - University of Manchester
Michael Cook - Collaborator - University of Manchester
Niels Peek - Collaborator - University of Manchester
Kritchavat Ploddi - Collaborator - University of Manchester
Maurice O'Connell - Collaborator - University of Manchester
HES Admitted Patient Care;ONS Death Registration Data;Patient Level Index of Multiple Deprivation