iconimage - stock.adobe.com

Women In Data panel: NHS needs to get data basics right before rushing into AI

During a panel discussion at the recent Women In Data event, speakers from across the public healthcare sector outlined the groundwork that has to be laid for AI to really take the NHS by storm

The NHS is broken. After years of chronic underfunding, the waiting list stands at more than 7.5 million, and over 40% of those people have been waiting for longer than the target of 18 weeks. The public healthcare sector is also struggling with recruitment. As of December 2023, there were over 110,000 vacancies in secondary care in England, including 10% of all nursing posts.

Technology is often viewed as the solution to many NHS issues. There has been lots of discussion around how data and technology such as artificial intelligence (AI) could facilitate patient self-service and personalisation, offering efficiencies and freeing up time for doctors and nurses. 

Speaking on a panel at the recent Women in Data event in London, Ming Tang, chief data and analytics officer at NHS England, said the health service is under extreme pressure, but at the same time, it needs to find the money to invest in technology.

“We do need to invest in the data and the infrastructure,” she said. “That’s why a lot of our investment is now focused on – probably quite rightly – longer-term things that will help us get there rather than just fixing the band-aid on the things that are broken, because there’s too many things broken.”

The NHS also needs to have a healthy conversation with the public around the use of their data. While personalised medicine has great potential for improving treatments, it requires collecting better and additional data on the individual.

“[Without that] we are not going to make sure that we are reaching out to the inequalities, we don’t understand the population, we don’t understand the determinants of health, we can’t solve those problems,” said Tang.

“All of that is around public trust. The data is a real opportunity, but we do also need to explain to the public why their data is so important.”

Gaining the public’s trust

Building this trust becomes more important as the NHS pushes ahead with its plans for a federated data platform (FDP).

The FDP will be used by NHS trusts and the 42 integrated care boards across the NHS in England, with the aim of connecting existing patient data in a safe and secure way, and reducing the time spent checking and finding information about patients.

Tang explained the FDP is about creating an underlying cloud-based data infrastructure at the centre of the NHS, serving as a national instance, but also with individual versions of the software for the integrated care boards.

“It’s about having that connective tissue that allows us to share appropriately when it’s right,” she said. “If you’ve ever been on the ward, you’ll see nurses with pockets of bits of paper because they’re scribbling things down. It’s making sure all that information is captured in the record or for a transfer of care, giving the information from one piece of paper to another.

“We talk about it as workflow management, decision support and making sure that we make the lives of our frontline clinicians and the administrative staff much easier, and therefore that gives them more time for patients,” said Tang.

While there has been much discussion of how AI could transform the NHS, she said the organisation is just at the foothills of adopting the technology. There are lots of exploratory teams having a play with the tech, including looking at improvements in productivity. For example, simplifying the process of doing a discharge form.

“It’s implementation of AI with a human,” she said. “That’s really important as we learn to test and iterate rather than jump straight in and say AI’s going to replace the work. It’s not.”

Readying the data

Before the NHS can start major projects, its data needs a lot of foundational work so it is suitable for AI.

“People forget that you have to do the careful thinking,” said Tang. “Then really around the ethics and the preparation of the processes so we get best value out of what we do with AI.”

Also on the panel, Sophie Williams, lead data scientist at Barts Life Sciences, agreed that building the right data foundation for any NHS AI system is crucial.

“The UK probably has the best health data in the world in terms of diversity and coverage,” she said, meaning there’s huge potential for individualised care. But there needs to be a process ensuring the data that goes into AI models is fair, accurate, covers everybody and is the right data set.

“Statistics can be used, in the wrong hands, to tell a different story than one that’s actually true,” said Williams. “We’ve got to be super careful about that. What works with one population doesn’t work with another.”

Efficiencies and cost-savings

Once AI is in place, there’s huge potential for efficiencies and cost-savings. Another panellist, Helen O’Neill, CEO at Hertility, cited a London hospital that has a million-pound budget alone for sending letters.

Digitised self-service is another important aspect of where tech can further improve the NHS, enabling people to fill in their own personal data, symptoms and conditions.

Playing to the strengths of different groups means offering people who are data and digitally enabled the ability to front-load their appointment and fill out information before arriving.

“In the on-average nine-minute appointment, there is not enough time to ask all the relevant questions and for that patient to answer all of them to get the diagnosis,” said O’Neill.

“Then, what you get is this underlying blame culture where it’s [complaining] the doctor didn’t prescribe you, but the doctor has seen 18 people in the last few hours. It’s a fundamental flaw that we expect things from a human being, to give answers based on so many complicated symptoms.”

Read more about AI in the healthcare sector

As well as ensuring appointments aren’t wasted asking basic questions rather than diagnosing symptoms, there’s also the opportunity to gather more truthful data. According to O’Neill, Hertility has been able to gather such a truthful data set as it’s done via a digital interface, using encouraging language, validating the reasons for gathering the data and explaining how the data can be used.

“We tend to tell different answers when looking a stranger in the eyes, especially when it comes to intimate problems,” she added.

Digitisation and self-service are about making the NHS more convenient for modern living, too. Tang pointed out that, for many people doing several jobs, trying to get half a day off just to go to see a doctor is not part of our world anymore.

One huge obstacle that Tang and her team need to overcome regarding AI development is talent attraction and retention. “The difficult part is, we’re skint,” she said. “We don’t pay the right salaries for those skills.

“We have to be a bit more creative in how we bring people in to work with us,” said Tang. “So, rotations, getting involved in Girls In Data, getting involved in apprenticeships, but also growing our own. It’s trying to look at what are the skills and competencies that we need to develop.”

Read more on Artificial intelligence, automation and robotics

Data Center
Data Management