An HSJ webinar explores how generative AI can enhance NHS efficiency, focusing on administrative tasks and practical steps for trusts to adopt generative AI technology, with insights from Imperial College Healthcare Trust

Human technology interaction

Generative AI offers huge potential for the NHS to work more efficiently but it is often difficult for organisations to take the first step. While much has been said about some of the clinical applications of generative AI – such as in interpreting scans – there are many more routine tasks where it could be used to improve efficiency and reduce the impact of staff shortages.

In association withNuance-Microsoft-lockup-stacked-300x239 (1)

An HSJ webinar, in association with Microsoft, sought to tease out where trusts could most benefit from adopting generative AI and how they should go about it.

One of the trusts which have been at the forefront of using generative AI is Imperial College Healthcare Trust. However, while Imperial has been developing and evaluating the use of generative AI in clinical areas such as radiology, imaging and pathology, it has also been looking at how it could support efficiencies in areas such as administration and other back office functions.

Erik Mayer, clinical reader, Imperial College London, and a consultant surgeon at both Imperial and the Royal Marsden, said one of the successes had been in using natural language processing to look at unstructured data and free text narratives such as patient feedback. Generative AI could analyse themes and sentiments from these in 15 minutes whereas a human would take six to eight days.

“What that means is people can then use that information to drive quality improvement initiatives that are patient-derived,” he said. “The staff get to see all the positive things that patients say about them when often they only hear about the negative things.”

Other applications included using unstructured analysis to look at recruitment processes to ensure that the trust was compliant with the EDI agenda and how discharge summaries could be tailored for patients in multiple languages. He added that around 80 per cent of information within Imperial’s electronic health record was free text.

For Malte Gerhold, director of innovation and improvement at the Health Foundation, demographic changes made looking at technology vital. Making use of the data the NHS has offers an opportunity to improve care – and generative AI is a powerful tool to understand and analyse that data, he added.

But he urged realism about what could happen and when. “I don’t think the AI doctor will see us anytime soon. I don’t think it’ll revolutionise healthcare in the next two, three, four or five years,” he said.

Mr Gerhold explained that much of the focus of generative AI at the moment was around the areas where there was a lot of data but that work was required in the areas that had less well-curated data, of poorer quality and consistency.

But the biggest challenge would be how to change ways of working and culture. Announcements by chancellor Jeremy Hunt around investment in IT had focused on the kit with no mention of investment in change capacity, he added.

“We know that even across introducing electronic patient records, it’s one thing to have them, it’s another one to change the ways of working on top of them, whether clinical or non-clinical, and that requires staff that requires analysts, et cetera,” he said.

One concern often raised about AI was that it could inhibit empathy and the use of emotional intelligence – and that may make some in the NHS more reluctant to adopt it

But what will make the difference for trusts who want to make use of generative AI? Anna Dijkstra, director of innovation for healthcare and life sciences at Microsoft, said the organisations she saw which seemed to be progressing in the use of AI had some common characteristics.

One was executive and clinical sponsorship which helped to focus the organisation on the biggest challenges. These organisations also needed to have quality data and bring together domain experts to figure out what the “core use” cases were and how to address them.

Knowing where to start can be hard; she recommended looking at,  first, the technical complexity to implement, and second, the value it creates for the organisation.

“Starting small can help your IT staff get more familiar with the technology and also help your clinical staff and others understand the technology and its potential a little bit better as you progress towards more complex use cases,” she said. Help was available from Microsoft and others for trusts which wanted to take those first steps, she said.

One concern often raised about AI was that it could inhibit empathy and the use of emotional intelligence – and that may make some in the NHS more reluctant to adopt it. Mr Mayer said that trust was really important and arose from engaging and co-designing with patients and the community.

Mr Gerhold added: “It’s quite easy to say ‘oh technology is all about productivity, efficiency, process’. But a lot of the aspects of those problems are also about the quality of care and the experience of patients. And in a lot of those areas, there’s great opportunity.”

Freeing up time from admin could mean clinicians were able to engage more with patients and could mean their experience of care improved, he said. “We may just help them to have a slightly better conversation and more human conversation with the patient, which will not show in any productivity statistics and that is okay.”

Ms Dijkstra pointed out how generative AI could adapt information to match the reading age of the patient or deliver it in a different language – giving them information they could understand and use.

But with so much work going on in generative AI, she saw a problem arising with different organisations developing their solutions independently. “I would love to see a centre of excellence where different organisations are sharing all these different use cases they’re creating,” she added.

So, what would the panellists’ advice be to a trust thinking about dipping its toe in the water of generative AI? Mr Mayer said there were several challenges to be worked through but the most obvious areas to focus on were in non-clinical operational administration and were less risky because they were not directly involved with patient care. This could help mature thinking and provide an evidence base before uses closer to patient care were tried.

Mr Gerhold urged listeners to think about what problems they wanted to solve, to look at what others had done and then have a conversation with staff.

And Ms Dijkstra suggested that anybody interested in exploring GenAI use cases, potentially through hackathons or skilling, to reach out to technology partners like Microsoft, or to Imperial to learn about their experience with effective governance and implementation.

This webinar is now available on demand.

To access the recording, visit here and click play.

If you had previously registered as a viewer for the event, you will be able to view the recording immediately.

If you have not previously registered, please complete the registration form at the foot of this page to access the recording.