Posted on January 26, 2017 by staff

How should the NHS adopt artificial intelligence?


Medical diagnosis has been identified as one of the areas where artificial intelligence and machine learning could have most impact – but how should the NHS proceed?

The likes of University College London Hospitals NHS Foundation Trust, Royal Free NHS Trust and Moorfields Eye Hospital have teamed up with Google DeepMind to work on better patient outcomes for those suffering from certain cancers, or for those who have had sight loss.

According to Orlando Agrippa, deputy CIO at Barts Health NHS Trust, the NHS could benefit tenfold if it leveraged AI.

“It could be used to increase accuracy over things like prescriptions, interventions and early diagnosis,” he told BusinessCloud.

The Commons Science and Technology Committee suggested last year that supercomputers assisting doctors with medical diagnoses could be one of the key impact areas of AI, but that government leadership in the fields of robotics and AI had been lacking.

Agrippa, who is also CEO of Draper and Dash, believes that the lack of leadership has resulted in anyone interested in the field having to ask ‘how do we do it?’

But the answer to that question isn’t necessarily to do with technical specifications – there are other important considerations to make. For example, data governance is critical – the AI system has to be able to capture and use the data effectively.

This can be a particular issue for NHS trusts where the electronic medical records are still being migrated, as they have to decide whether to use any data they currently have to plough ahead with an AI tool, or wait until the data is all migrated and then make the leap towards AI.

Agrippa says that amongst the challenges for the NHS is a “mental block” because there isn’t a business plan or strategy for AI, and a lack of infrastructure to deal with it.

“It would be difficult for the NHS to invest that heavily on the infrastructure, so it would need to come from a combination of the NHS and larger firms like Google or Amazon, or one of the big tech companies would have to make an investment on their own to create a movement on this,” he said.

Privacy is an issue that the organisations working with Google DeepMind had to contend with, and is likely to continue to feature in debates when the NHS signs contracts with other AI vendors.

“It’s got the potential to be life-saving and transformational, so we’re going to have to give up several of our basic assumptions of what privacy means – because this needs a lot of data… and without that hurdle being overcome we will struggle to see some meaningful uptake for outcomes and patients with AI,” claimed Agrippa.

Matteo Berlucchi, CEO and co-founder of personal health assistant Your.MD – which has over one million users – has taken a different approach.

“We’re looking at creating tools to put into the hands of patients to reduce the load on the NHS,” he said.

The application uses an AI engine which gives the probability of any individual with a given set of symptoms having a certain condition.

“Normally this is something the doctor would do with a gut feeling, but we can do it 24/7, a million times per minute, and we’re using the same approach but are less likely to make human mistakes,” he added.

The company has had early discussions with the NHS and other healthcare organisations around the world about using the Your.MD engine.

Berlucchi believes that it could streamline inefficient interactions between people and doctors: he claims that 90 per cent of visits to a GP shouldn’t require an interaction with the NHS at all.

Jeff Spight, president of collaborative health systems at healthcare organisation Universal American, worked with the NHS a few years ago.

He believes that some clinical commissioning groups (CCGs) are further along than others in being able to exploit AI.

“In terms of aggregating data, managing it and having some analytic overlay on top of it, my guess is that a third are ready to go for this from a technical stand point,” he said. But the bigger issue, he believes, is a cultural one.

“The biggest hurdle for the NHS and other providers is culturally for doctors to embrace this. How can you get them to understand that it can make their staff better – they can’t see it as an attack on them,” he stated.

“AI should raise [mid-level NHS staff] to level-up [with their seniors], it’s not to replace but to verify or validate; it gives them a better sense of when to escalate to the physician.”

The fact that all levels of medical staff could benefit from AI gives Spight enough reason to ensure that they are all engaged in any procurement activity from the beginning.

“You have to get them to understand why it’s there, how it can help and you really need champions who can get involved from the start.

“We certainly make the call about capital or finance but they are the ones who make the all-important clinical and patient safety calls, as well as how it would fit in operationally.”