Revolution or risk? The NHS’s AI balancing act
A GP tells us how and why he has embraced AI
AI success in the NHS hinges on change management.
After years of staff shortages, lack of tech investment and a pandemic, the NHS needs a fresh approach to keep providing each British citizen access to healthcare.
The government’s Plan for Change pictures AI as a “catalyst” to transform the NHS, but there’s a tougher reality behind the rhetoric. Winning over clinicians wary of disruption, and earning the public’s trust with something as sensitive as health data, will take more than ambition; it demands careful navigation through a minefield of scepticism, ethics and lived experience.
But, not all doctors are slow to adopt. While hospital employees have to work only with what limited NHS budgets can afford, those in primary care, i.e. general practice, are private contractors and can move quickly to adopt new tech tools.
Dr Rahul Goyal, chief medical officer at Elsevier and a practicing GP, first took AI on for transcription and note-taking in 2016, long before modern generative systems were available. His use has evolved over time and now he says AI has “kind of put my workflow on steroids.”
“[AI] gives me actionable insights within my workflow. I get insights to say, 'This patient has got a high risk of cardiovascular disease, but there isn't any blood pressure recorded in the last 12 months’ [and I can perform a test there]. It saves an appointment, and the patient’s care is more solid.
“It’s surfaced stuff that I was blind to or I had to dig in to look for; which is again aiding me in doing my work more efficiently and for better patient outcomes.”
Another use of AI both the NHS and government are keen to expand is lowering prescribing costs by suggesting switches to generic brands of medicine. Dr Goyal says this “nudging” behaviour is great for busy clinicians:
“It's a positive reinforcement, and it's a very good way of behaviour nudging for clinicians who are time poor.
“Instinctively, I wouldn't go and look for the cheapest alternative on BNF [British National Formulary], because that's not my workflow.”
In fact, the most successful modern clinical tools are those that nudge doctors towards “the right behaviour.”
"[Generative AI] is something very new, it only came three years ago... We shouldn't be umming and ahhing about what's right and what's wrong. We should act swiftly and create guardrails sooner rather than later, before bad behaviour becomes embedded in our workflows - especially for the younger ones who are still learning.”
Who is embracing AI, and who still has concerns?
It’s easy to imagine a future where something goes horribly wrong with an AI that has access to people’s medical records, but those concerns don’t translate to the practice room.
Take GP staff, for example. From receptionists to clinicians, they are time-poor; anything that can claw some back is welcome.
“Every minute saved is a minute added that we can probably have a coffee break that day. So, it's something as trivial as having a coffee break that motivates you to reduce the friction in your workflow.”
It’s little surprise that admin staff have been particularly positive about AI and require little encouragement to embrace the opportunities; whereas doctors, who are more wary, require change management. That’s a problem.
“We don't currently in the NHS have the infrastructure or the change management that it takes to have clinicians be taken along... For the size of the departments that NHS England has had, they've not been very good in the show and tell part,” says Dr Goyal.
“I think we need more of the comms for the change management, because clinicians are clinicians: they're all different shapes, sizes and colours, but the one thing that works and join them together is a value proposition. If you can give them a value proposition, it's a very hard case not to adopt.”
He believes that fear is holding some back, but emphasises that AI is not the industry-destroying phenomenon they might be imagining. He says the idea of software replacing doctors (by collating data from IoT sensors like smartwatches and glucose monitors) is overhyped.
“Yes, they will have certain use cases that we will use in the future, but it will not be the silver bullet that it's sort of perceived to be... With the best will in the world, a skew in data can introduce bias, and the best person currently to call that bias out is a human.”
And on the end user side, patients “love” AI. “They feel that I'm able to give them more face time than being submerged behind a screen all the time, typing away.”
As well as transcription and prescription support, Dr Goyal is using AI as “a neutral second opinion” that can reassure patients his recommendations are the right ones.
“I'm using technology as an aide-mémoire, more than anything else. It's something which I feel reassured by, and I can reassure my patient with, because I have the hard facts in front of me. So, I’m using it as a tool; as a servant, not the master...
“I feel patients get more confident in the healthcare service when they feel that a technology has been implemented that makes their care safer.”
“Democracy doesn't exist”
So admin staff and patients are eager to embrace AI, while doctors have reservations. But will they have a choice? The government has been outspoken in its opinion on AI’s value for the health service, and NHS England is developing its own AI models.
“It’s very rare that [decision-making] is by consensus [in the NHS], or even by, ‘This is what we’re planning to do, what do you think about it?' It’s usually an email on a Friday afternoon that on Monday this is happening.”
Unusually, Dr Goyal doesn’t think this is a bad thing. In an organisation the size of the NHS, there is no way to satisfy everyone: leadership has to be top-down and decisions must be made for the majority. Or, to put it another way: “If we look for the perfect solution before we do adoption, we would never get there.”
He adds, “I see more opportunity than risk from a top-down approach... The risk is obviously alienating your clinicians if you move too fast, too hard, but it cannot rely on a happy consensus at every step, either.”
The case for AI in the NHS is increasingly clear, but implementation is another story. Regulatory, financial, cultural and technical constraints (Dr Goyal’s surgery still has some PCs running on Pentium processors!) all stand in the way. It’s not just about convincing clinicians of AI’s value, but giving them the means to realise it.