Insights: AI and Data
Time to deliver on the promise of AI
20 September 2024
There is little doubt that AI has the potential to transform various aspects of health and care. Technologies using AI are already in use in the NHS, for example to enhance clinical imaging and diagnosis. There is also growing acceptance among the public and healthcare staff that AI has a role to play in healthcare. According to a report published in July by The Health Foundation, on balance, the public and NHS staff support the use of AI, and most NHS staff look forward to using it in their work.
Real-world challenges remain – not least the issues of outdated infrastructure, tight resources, continuing questions over privacy and cybersecurity, and, of course, our old friend interoperability. But the urgent need to improve healthcare may have created a shift – making it easier to make the business case for AI-based technologies. Experts agree that AI, properly deployed, could transform both healthcare and population health – and some are optimistic that this will happen sooner rather than later.
Part 1
The promise of data and AI
“I’m a huge advocate of AI and machine learning,” says Monica Jones, chief data officer for the University of Leeds and associate director for HDR (Health Data Research) UK North. “It’s been on its way for a long time, but it’s here, and it’s here to stay. But AI technologies rely on high quality, well-managed data, and in order to be able to leverage AI’s transformative potential, we’ve got to fortify our data foundation: bad data – bad AI; good data – good AI.”
Simon Noel, head of nursing informatics at Oxford University Hospitals NHS Foundation Trust, says “the sky’s the limit” when it comes to the potential of AI.
“Healthcare by its very nature is data hungry and generates an enormous amount of information about an enormous amount of people across an enormous amount of subject areas – not just to do with diagnostics, but also operational performance, people, trends, population health,” says Noel, who sits on Digital Health’s CNIO advisory panel, which is putting together a position statement on the use of AI in the production of clinical notes. “It’s a rich and dynamic environment which is a natural area for AI to take hold within.”
There’s a huge opportunity to predict what’s going to happen with people and populations. Simon Noel
He is particularly excited about the initiatives coming through in the management of complex data, for example using MRI scans to predict whether patients will eventually develop cardiac disease. “There’s a huge amount of opportunity to predict what’s going to happen with people and populations, and if we are able to apply that information appropriately, it can help us enormously to target what we need to do in healthcare, but also to prevent people getting sick in the first place.”
Dan Midgley, UK and Ireland sales manager with Orion Health, says there are some misunderstandings around what AI does and means, but that the actual promise is huge. “In the next few months and years it’s really looking at how you can use AI to assist clinicians, to assist in decision-making, scheduling and prioritisation, rather than taking over roles,” he says. “I think some people have concerns about replacing the clinician-patient relationship, but we don’t see it that way. Rather, it’s about if we embrace the promise of AI, how can clinicians make better decisions, how can patients take ownership of their own care, and how can we increase capacity and get people to the right place at the right time?”
Looking further ahead, he adds, it’s about increasing productivity and reducing administrative tasks for the workforce. “It’s also making sure that care becomes more personalised and patient-centric, and ideally more proactive as well – starting to identify where an intervention is needed before a patient is really sick, rather than waiting for them to turn up in the emergency room.”
Part 2
Barriers to progress
Although progress is being made with AI, particularly in traditionally pioneering areas such as radiology, roll-out is far from universal.
Market analysis from Digital Health Intelligence (DHI) published earlier this year points to early successes in using AI in clinical imaging, saying it is one of the best examples of technology directly impacting the delivery of care. But it also identifies key challenges, and says action is needed, for example, to ensure data integrity and enhance transparency. This, the analysis says, is essential for building trust and ensuring responsible use of AI in healthcare.
At the moment we are having to jump through hoops and make compromises to work with AI.
Gerald Lip
Consultant radiologist Dr Gerald Lip points to challenges including the prevalence of older IT infrastructure in the NHS, which can make it difficult to integrate AI solutions.
He also cites concerns about data privacy and ownership, and cybersecurity. “At the moment we are having to jump through hoops and use alternative methods or make some compromises to work with AI,” says Lip, clinical director, North East Scotland Breast Screening Programme and vice chair, Royal College of Radiologists British Society of Breast Radiology. “But in the next 10 or 20 years, as more modern systems come in, the facility to anonymise, share the data, and have research environments as well as innovation environments [will come] to the fore.”
There are cultural issues at play too. “There’s fear of the unknown,” says Jones. “There can be a feeling that if you can’t see it, and can’t touch it, how can you trust it. I think it’s beholden on us as healthcare professionals and working in the health service to be able to explain and not try to dupe patients and the public, to show that it’s not just some AI bot looking at your records and then prescribing something willy nilly without a qualified healthcare professional providing oversight.”
Part 3
Clinical imaging is at the forefront of AI development, with technologies already making a difference to diagnosis and interventions, and more solutions coming on board all the time.
At NHS Grampian, Lip is working with Annalise.ai, a chest X-ray AI company, which is used to help detect possible lung cancers at an early stage, and has also used AI as a “safety net” in mammograms. He is clearly impressed. “If we screen 10,000 women, you expect to find about 100 cancers, but we found 11 or 12 more cancers using the AI as well,” he says.
As a radiologist, it’s very rewarding to use AI to get better results and outcomes, he adds. “I became a doctor because I wanted to help people, and AI is just another technology that we’re able to use now to help people.”
Paula Lender-Swain, director UKI public sector sales with Hewlett Packard Enterprise (HPE), says it’s now much easier to make the business case for AI-based technologies. “There’s a couple of things that are helping accelerate innovation. One is that new technology is cost-effective in terms of power consumption – I’ve seen cases where the power consumption for technology in a hospital is so high that the new equipment cost is exactly what they were paying in monthly bills to run the old equipment. We can help organisations access new, modern infrastructure, which brings the benefits of improved cyber security and AI-readiness, for the same (or lower) cost as their current infrastructure, because of the savings on energy.
Whatever tool you bring into a system, it’s the humans who will be using it who should be the focus.
Beatrix Fletcher
“The second thing is that we have a better understanding, I think, since Covid, that we feel the urgency more deeply to do things better. As a society across the board we are taking health and bringing it to the centre… People are more likely to spend the money if it brings value and saves lives. These are the outcomes that technology can promise and deliver, and people are moving towards that in a much more positive and urgent manner.”
Solutions aren’t only technology – they are also about people. Beatrix Fletcher, programme manager for the Fellowship in Clinical AI initiative based at Guy’s and St Thomas’ NHS Foundation Trust, is helping to create a cadre of AI-literate clinicians equipped to spread knowledge across the NHS. Open to clinicians across the UK, the fellowship has myriad benefits.
“Whatever tool you bring into a system, it’s the humans who will be using it,” says Fletcher. “So actually, they should be the focus originally. What we’re already seeing is that people exit the programme and go back to their trusts and they become an AI expert within the trust, which is quite rare.”
This means, she says, not only that they can raise awareness and share their expertise with colleagues, but that they know the points to raise and questions to ask when a vendor wants to sell an AI product. They also have the specialist knowledge to be able to identify where an AI solution might help solve a problem – and so be able to go out to the commercial world and explain what they want them to build.
What is important, she says, is to identify where AI is actually most useful, and where a human still has the edge. “Humans are good at certain things – we can look at a cat and know it’s a cat. We don’t need to spend hours working that out. AI should never be used for things like that – it’s a waste of time and energy.
“But what it’s really good at is taking a lot of very complicated data from lots of different data points that are extremely detailed, and seeing a common pattern, then synthesising it down to a meaningful output, like a single statement or a suggested action. From that point of view, I think its greatest use in healthcare is often improving the workflow of our clinicians – and that’s got to be welcome.”
- Data quality and validation: The accuracy of AI algorithms depends on the quality and diversity of training data. Ensuring data integrity, and that AI models are validated in real-world clinical settings, is crucial.
- Transparency: AI algorithms can be complex and opaque, making it challenging to understand their decision-making processes. Enhancing transparency is essential for building trust and ensuring responsible use of AI in healthcare.
- Ethical considerations: AI in healthcare raises ethical concerns, such as potential bias, discrimination, and patient autonomy. Careful ethical analysis and guidelines are needed to ensure AI is used responsibly and ethically.