The NHS spends £450 million each year on Health Checks to prevent cardiovascular disease, yet half of adults with high blood pressure or high cholesterol remain undiagnosed. This project aims to fix that mismatch by building machine-learning models that identify not just who is at risk, but who will actually act on the advice they receive. Current risk scores flag people with high cholesterol or hypertension, but many never change their behaviour or attend follow-up appointments. The researchers will train algorithms on NHS data from north-west London to predict both undiagnosed risk factors and a person’s likelihood of engaging with prevention programmes. They will then test whether this “impactability” approach would have improved outcomes in past patients before trialling it in real NHS Health Checks. If the models work, the NHS could redirect its £450 million budget toward people most likely to benefit, while also designing stronger support for those who need it most. The result would be fewer heart attacks and strokes without spending more money—just spending it smarter.
View original technical description
Cardiovascular disease (CVD) prevention remains a public health priority, yet traditional risk stratification has not consistently improved outcomes. "Impactability" refines this approach by identifying individuals’ propensity to benefit from interventions. In the UK, significant proportions of adults have undiagnosed risk factors—50% for hypertension (60% undiagnosed) and 50% for high cholesterol (only half diagnosed). Effective resource allocation targeting high-risk, high-impactability patients could improve health outcomes, while equally addressing disparities through intensified intervention for low impactability. The NHS Health Check, a £450 million-per-year CVD prevention programme, highlights the limitations of risk stratification alone. The effectiveness of detection varies widely, and without an impactability-driven approach, prevention costs can be excessive. Machine learning (ML)-based CVD risk prediction enhances stratification by addressing moderate predictive accuracy and scalability limitations. By pairing ML models with impactability modelling, healthcare systems can optimise patient selection and intervention impact. This project aims to: (1) develop an ML classification model to detect undiagnosed CVD risk factors; (2) create an impactability model incorporating NHS Health Check engagement and behavior modification likelihood; and (3) retrospectively validate these models before early prospective (feasibility) application. Using Whole Systems Integrated Care (WSIC) data, ML techniques (XGBoost, random forest, logistic regression) will be applied to enable a convergence of risk prediction and impactability models. The final phase will measure potential impact of this innovative approach in the context of technology-enabled NHS Health Checks. Findings will inform subsequent grant applications and broader implementation of impactability modelling for population health interventions.
Plain English summaries and category classifications on this site are generated by AI and may not perfectly reflect the original research.
Is something wrong? Let us know