In March 2021, Julia Wilkins and I gave a talk online to an audience of 200, launching the machine learning community (login required) for data scientists and analysts. The community is based in AnalystX, part of the FutureNHS Collaboration Platform. NHS England has asked the Health Economics Unit (HEU) to lead this and we hope it will become a community where people can learn, exchange ideas and give feedback on each other’s work.
Why machine learning?
Some call machine learning the science of programming computers to learn from data; others call it an art. In the very simplest terms, machine learning teaches computers how to solve problems on their own by feeding them many solved examples (datasets). The computer uses these examples to guess the answers to similar questions. The more examples fed to the computer, the better it gets at coming up with the correct solution.
Machine learning has multiple applications within the health service, from predicting which patients are at risk of developing certain conditions to automating tasks and freeing up clinicians’ time.
One of our aims at HEU is to unlock the full potential of machine learning within the NHS, and the new workspace offers a great opportunity for this. At its launch we shared training materials for analysts wanting to start their machine learning path and showed how the platform will offer a space for expert data scientists to discuss their work and receive input. It will also act as a forum to share resources and ask for support.
We also shared examples of real-world applications, including one from Rob O’Neill, Head of Analytics at University Hospitals of Morecambe Bay NHS Trust. At Morecambe Bay, GP practices and primary care networks are using machine learning to predict patients at risk of high blood pressure.
Predictive models are crucial for the health service. If you can accurately predict who will get hypertension, you can prevent the problem instead of reacting to it. Once someone has hypertension, they are at much greater risk of developing other chronic conditions such as heart failure, COPD and diabetes.
Treating hypertension and its consequences is therefore far more costly than preventing it, not only in terms of the money spent on care, but from the perspective of quality of life for the individual. Predictive models have this potential to bring high levels of efficiency to the service.
Rob’s team involved clinicians from the beginning, which is important – it helps in building not only a better model but also one that resonates with end users and is therefore more likely to be used.
Gary Hutson, Head of Advanced Analytics, NHS Arden and GEM CSU, also joined us to talk through some of his recent machine learning projects, including a readmission avoidance tool that gives patients a readmission risk probability and supports resource allocation. This helps clinicians by augmenting their decision-making processes and freeing up their time for patient care.
Who will benefit from the new space?
The new space is designed for expert machine learning engineers and data scientists to discuss methods and approaches and receive feedback on their work. The NHS also has many analysts and business intelligence (BI) teams. We want to help these teams realise that they are not that far away from developing machine learning skills. We also want to give them the tools to show managers and commissioners why machine learning is important and what it can do in the real world.
Our aim is for this online space to attract hundreds of regular visitors. My biggest personal objective is to develop machine learning skills in the NHS workforce and help them understand their importance. If we can get close to that goal, then I’ll be very happy and it will be a job well done!
Author: Bruno Petrungaro
Helping the NHS use machine learning effectively makes Bruno Petrungaro happy.
Bruno brings his vast knowledge, experience and network to ensure we deliver outstanding products that achieve our clients’ aims.