Hello, my name is Emil.

Earlier this year, the 10-Year Health Plan for England outlined three major shifts in our health and care system: from hospital to community, from sickness to prevention and from analogue to digital.

The last one, sometimes described as moving from ‘bricks to clicks’, focuses on using technology to support more proactive, personalised care. It envisions a future where clinicians can act early, reaching out at the first signs of deterioration to prevent a crisis or hospital admission.

In Wakefield, we’ve already begun to make this vision a reality.

Our approach builds on the Wakefield Linked Data Model which securely brings together rich information from across the NHS, primary care and social care, but also from non-NHS partners such as education, housing and the voluntary, community and social enterprise (VCSE) sector.

Using this foundation, our team has developed a predictive tool to support the rollout of our neighbourhood health work. 

Initially we were asked to identify a cohort that represented 3 to 4 percent of the population and saw a high proportion of non-elective admissions. For us in Wakefield, the programme will start by focusing on individuals with chronic obstructive pulmonary disease (COPD), dementia and those receiving end-of-life care.

This was a significant caseload, so we have used a ‘risk stratification’ approach, based on Random Forest algorithms, to narrow down the cohorts to a more manageable size. Our approach draws on several features from our linked datasets to make predictions for patients’ risk of admission within following months and classifying these into high, medium and low risk categories.

Using these predictive models, we have seen encouragingly high accuracy and precision to correctly predict patients’ risk, allowing the system to utilise its resources efficiently. This enables our neighbourhood health teams to receive the latest patient lists by the level of risk and focus their efforts where they can make the biggest difference.

Looking ahead

In the future, similar methods can be employed to develop similar tools for other cohorts to help provide the right support at the right time.

Beyond risk stratification, there are lots of exciting opportunities for us to use data science, such as impact analysis. This would mean identifying not just who is most at risk, but what groups are most likely to benefit from a specific intervention or service.

For us working in analytics, much of our time goes towards developing and maintaining tools, reports and data models. Usually, the largest part of this involves acquiring and understanding the subject matter relating to the data, which we fortunately have plenty of expert colleagues to rely on, so that we can analyse the information appropriately.

Also because of the variety of tasks we do, I do partly intentionally and by accident get to continually learn about new techniques which keeps things interesting.

It is rewarding to be part of something that is at the forefront of our health and care system – a future that’s already being shaped here.

Thank you for reading.