With the emergence of promising new technologies and a flood of data, healthcare innovators face a new set of opportunities and challenges. These will define the next decade of operational efficiency in healthcare delivery. Our event ‘’Taking the pAIn out of AI’’ in collaboration with IET and WAI, looked at how the brilliant minds of Finn Catling, Eleonora Harwich, Jim Ritchie and Catherine Davies have navigated the challenges across the NHS. This was followed by an engaging Q&A with the panel. Set out below is a short summary. Some slides are also available on our website here.
George Batchelor, Director at Edge, outlined the challenge:
"Even the best ideas struggle to get data, or access to hospitals IT systems. If they manage that, there are still significant challenges with implementation – and that is before the technology has been fully developed or tested".
Building ML model to anticipate patients’ decisions
There is much potential to use AI in healthcare, particularly, in the intensive care unit – it can literally help save lives. But first, we need to be able to ask ourselves the right questions such as: "What problems our patients and staff face?" and "How can we help solve them?". Finn (Critical Care Doctor and ML researcher) built a machine learning model able to predict treatment decisions in the ICU, to define how to improve patients’ outcomes based on their current situation. The idea behind it is, when we are convinced the model can suggest sensible decisions that make sense from a doctor’s perspective, we can think of scaling and using it for supporting decisions across the unit and beyond.
Protecting patient confidentiality
Accessing the right data is hard. Eleonora (Head of Research at Reform) has been looking at this and gave an insightful presentation. However, once you have the data in hand, there is a bigger problem: companies are likely to face new challenges such as preserving patient privacy. And they could face potential backlash, while the NHS is still recovering from the controversy associated with care.data. It’s important to remember that patients should "trust private firms using their data, by design" Eleonora says. Interestingly, there is the option of using synthetic data – a clone of real data that can be used at the early stages of research and development. Since it is flexible and rich enough, it could help ML practitioners conducting fascinating experiments, where data are truly anonymised. Once the innovation is understood, it’s much closer to being ready for implementation!
Overcoming people challenges
While there is a lot of excitement around harnessing health innovations, hospital managers tend to face "resistance to change", as Jim (Programme Director at the Digital Control Centre) outlined. Healthcare innovators need to understand NHS processes better. However, to be successful, commitment and participation needs to be made at each stage of the NHS ladder. Currently, Trusts are bombarded with new regulations and initiatives based on a top-down approach, yet are not given the right tools to solve these problems. AI is the key to solve that pain, if we have the right support in place from the start.
The next wave of AI start-ups
At the same time, there is an explosion of start-ups, the NHS being on a financial strain has an impact on funding. Many of them fail due to investors’ lack of interest and give up on their innovations which the NHS could have benefitted from in the longer term. So, how do we solve that ambiguity? Catherine (Managing Partner at Monticle) proposes an optimistic model, based on Atul Gawande’s article on US incremental care: What if… start-ups of all sizes were given more support to understand the NHS? And were offered incentives to local payers to adopt the technology? What if NHSX, innovation organisation within the NHSI/E come together to offer regulatory consultancy, oversee, hold NHS accountable and encourage take-up? Would that not alleviate some of the strain put on our health system by allowing innovations to scale?
While there is still much frustration around the NHS, the momentum around AI keeps growing. There is a common agreement that without behavioural change, understanding the data governance model and working closely with the NHS, Taking the pAIn out of AI would be easier.
If you have attended, please do let us have your feedback here, so we can continue to ensure our events meets or exceeds your expectations. You can also see some of the footage on our twitter (#hospitalAI) and Instagram.