Artificial intelligence will revolutionise healthcare but it can’t replace doctors and nurses

Analysis: Machine learning offers incredible promise as a decision-making tool for clinicians but it cannot replace them, writes Shaun Lintern

The NHS is spending £250m on AI (Getty/iStock)

Millions of pounds of taxpayers money is being spent to prepare for and help the NHS to take advantage of the emerging revolution of artificial intelligence in healthcare.

Momentum has been building for several years and is now reaching a tipping point as a series of studies and trials show AI, or machine learning, can match or even improve on traditional techniques.

Most significantly for patients, machine-learning algorithms offer pattern-recognition solutions, allowing machines to spot and diagnose breast cancers, brain tumours and other diseases that stand out when compared to thousands or even millions of health scans.

A recent study into AI use in breast cancer diagnosis found the machine was at least as good as humans at spotting cancer and was better at avoiding false positives.

Now a new study from a team in New York has found AI can help surgeons by providing an accurate diagnosis of brain tumours in just minutes while the patient is still on the operating table.

These impressive results coupled with the fact AI technology works seven days a week, 24 hours a day without problematic unions or contracts to negotiate, means it is a tempting solution for the government and NHS England, which know they lack the number of nurses and doctors needed to cope with rising demand.

Health secretary Matt Hancock is an evangelist for greater technology in the NHS – support and reform for which is long overdue.

His in-house tech arm, NHSX, is spending £250m on artificial intelligence in healthcare as part of a new AI lab while NHS England chief Sir Simon Stevens has said he wants the NHS to be a world leader in AI to help axe millions of outpatient appointments.

But the health service’s record on technology and adopting IT at scale is not a convincing one. Labour’s national programme for IT, launched in 2002, was a disaster, costing taxpayers £10bn and failing to deliver on its promise.

The use of robots in surgery is another area of technological innovation which in practice has been questioned. The USA’s Food and Drug Administration issued a warning earlier this year over the use of robots for cancer surgery arguing the evidence was not clear on any benefits and in some cases showed patients fared worse.

Artificial intelligence is here and is already being used in healthcare, but it needs to be tested like any new medicine or drug.

If the emerging trends are true, the real strength of AI will not be as a replacement for doctors and nurses and other diagnostic staff but as an aid to their decision making.

In the New York study on brain surgery the research found the errors made by the AI and pathologists were different: meaning, while humans got it wrong, so did the machines – but in different patients.

The authors suggested if the two were combined the accuracy could increase to 100 per cent.

NHS staff are being worked into the ground by relentless patient demand, years of underfunding and a chronic lack of staff. Technology and artificial intelligence will not be able to replace the clinician, it won’t be able to factor in the personal needs of individual patients, change their catheter or make a differential diagnosis, and as studies show, it too will make potentially fatal mistakes.

The ambition must be to align the human and the machine, with one checking the other, and decisions being supported or challenged in either direction.

To achieve this the NHS needs to invest in its existing workforce, many of whom will still be working in the decades to come, to make sure they have the knowledge and critical thinking skills to be informed by, but not reliant on, what the computer is telling them.

They must be able to think for themselves and be empowered to challenge the machine, because healthcare is rarely a binary yes-no endeavour and reducing it to such a base level could come with unthinkable costs.