- Fotolia

Sample datasets key to successful AI at NHS

IT providers will need access to datasets from NHS Digital to create artificial intelligence algorithms, but they should be held accountable for the systems they develop, says report

A report from think-tank Reform has recommended that NHS trusts apply artificial intelligence (AI) to support transformation initiatives.

But the report, Thinking on its own: AI in the NHS, warned that the piecemeal approach taken to procuring IT has left trusts relying on individual providers to support AI in their applications.

Reform urged NHS England and the National Institute for Health and Care Excellence to set out a clear framework for the procurement of AI systems.

The report said this would ensure that complex-to-use and unintuitive products are not purchased because they could hamper service transformation and become burdensome for the healthcare professionals.

As Computer Weekly has reported previously, industry experts have warned that poor-quality datasets in the NHS could hamper the development of AI algorithms for healthcare.

Data-sharing arrangements, such as the agreement between Google’s DeepMind and the Royal Free Hospital NHS Foundation Trust for the Streams application, were found to have infringed the UK Data Protection Act, as patients need to give explicit consent for their data to be shared.

The Reform report recommended that NHS Digital create a list of training datasets, such as clinical imaging datasets, which it should make more easily available to companies that want to train their AI algorithms to deliver better care and improved outcomes. It also called for the development of a specific framework specifying the conditions required to securely access this data.

Reform also recommended that IT companies operating AI algorithms in the NHS should be held accountable for system failures, just as other medical device or drug companies are held accountable under the Medicine and Healthcare Products Regulatory Agency framework.

Benefits of AI in healthcare

The report highlighted several areas where AI could be applied in healthcare. One example is wearables, monitored by AI, that could help to keep patients well within their own communities, rather than in hospital.

These devices can monitor the number of steps a patient has taken, or vital signs such as heart rate.

“AI can interpret this information to give people greater access to knowledge about their physical condition,” wrote the report’s authors, Eleonora Harwich and Kate Laycock. “One in seven UK adults owns wearable fitness trackers, reflecting the UK’s appetite for wellbeing.”

Digesting medical research is another opportunity for AI. “Watson could process existing literature alongside patient data to aid diagnosis and then recommend treatment options to clinicians,” the report said. “This has the potential to standardise high-quality care as all health professionals would have improved access to relevant research and guidance.”

Similarly, AI could be applied to augment X-ray diagnosis. For instance, it has been widely reported that a high proportion of mammograms yield false positive results when interpreted by radiologists, leading to one in two healthy women being told they may have cancer. “AI is enabling interpretation of mammograms 30 times faster than humans and with greater accuracy,” said Harwich and Laycock.

In the area of augmenting treatment, the report noted that the NHS is investing in an AI smartphone app developed by Ieso Digital Health to deliver online cognitive behavioural therapy. “So far, nearly 17,000 people have been treated and industry evidence shows it is reducing treatment time by 50%,” the report’s authors said.

Read more about AI in healthcare

Apps have also been developed that use AI to process blood sugar readings from people with diabetes. “After learning about the individual, the programme sends guidance and information to help them manage their disease,” the report said.

Last year, research from Nuance Communications, based on Freedom of Information requests across 45 NHS trusts, showed that half of NHS trusts (43%) were investing in AI enabling patients to “self-help” when accessing services. The trusts are harnessing technology such as virtual assistants, speech recognition technology and chatbots to ease the pressure on healthcare workers, according to Nuance’s research.

AI also offers the NHS an opportunity to optimise back-office, non-clinical process. Nuance’s research found that the vast majority of NHS workers still rely in some way on pen and paper to build patient records, with 93% using traditional-word processing tools to type up electronic patient records.

The Reform report cites research from the Royal College of Nursing which found that 17-19% of nursing time is spent on “non-essential” paperwork. Harwich and Laycock said: “Interviewees for this paper explained that intelligent virtual assistants, such as Amelia, could also support medical staff by placing individuals on pathways, book appointments, automatically compose letters, and send patients reminders.”

David Champeaux, director of global cognitive health solutions at IPSoft, the company that makes Amelia, said: “AI may also be a long-term cure for the [NHS] resource crisis, as reportedly there is a shortage of at least 50,000 doctors and nurses. Cognitive agents can hold a human-like conversation with patients, providing the most up-to-date, personalised and timely guidance, based on the insights available to patients and the care team.

“When these agents join the care team, this relieves the healthcare workforce from repetitive, administrative tasks that can be automated, which allows doctors and nurses more time to care for those in need.”



Enjoy the benefits of CW+ membership, learn more and join.

Read more on Artificial intelligence, automation and robotics

Start the conversation

Send me notifications when other members comment.

By submitting you agree to receive email from TechTarget and its partners. If you reside outside of the United States, you consent to having your personal data transferred to and processed in the United States. Privacy

Please create a username to comment.