AI can diagnose ‘just as well as human doctors’ – but cannot replace them

As I read the latest medical research earlier this week, I came across a headline that made me flinch.
 
Educated AI can diagnose disease just as well as human doctors
 
The story reported on the first meta-analysis and systematic review of artificial intelligence (AI) technology.
 
The research, published in The Lancet Digital Health, analysed 14 studies that compared AI performance with that of health professionals. It found that AI known as ‘deep learning’ correctly detected disease in 87% of cases, compared to 86% by human professionals, though the authors did note several limitations in their methodology and reporting.
 
Now, don’t get me wrong; I welcome advances in AI and can see its potentially vital role in the future of medicine.
 
I can also understand how such algorithms may outperform humans when it comes to detecting disease in certain situations.
 
And medical imaging seems, to me, to be an area where I can understand its role.
 
RACGP President Dr Harry Nespolon also applauds such technology.
 
He voiced his thoughts on the matter at the Future Forum event at the recent General Practice Training and Education Conference (GPTEC).
 
‘AI, to me, is going to be one of the great things to help general practice become a much broader speciality,’ he said.
 
‘It will empower us rather than disempower us.’
 
But he also asked a vital question about AI and its role in medicine: What is the value in a consultation?
 
‘It’s not ordering an ultrasound and getting an answer, it’s what you do with it,’ he said.
 
‘There are a million apps out there to tell you if a mole is a melanoma, but until AI can actually cut it out you have to come see a GP.
 
‘The better technology gets, the more a GP can take over an enormous range of tasks.’
 
While I welcome advances in AI and its role in medicine, my worry more specifically lies in how such news is reported and interpreted.
 
The headline of the story – suggesting that AI can diagnose disease ‘just as well as human doctors’ – concerns me.
 
Even though the headline only talks about the role of AI in diagnosing patients, not treating them, I fear it could be a slippery slope to the next implication: that doctors can be replaced in other areas.
 
And that is problematic.
 
Because patients don’t go to the doctor to simply get a diagnosis, nor to seek the right treatment. Seeing a GP offers so much more than that.
 
There is the opportunity for preventive care, checking in on a patient’s mental health, following up previous issues, and offering a supportive role for future healthcare.
 
Associate Professor Vicki Kotsirilos agrees, saying that information regarding AI needs to be conveyed in a manner that doesn’t diminish the role of doctors.
 
She told newsGP she was ‘astounded’ when she read the headline stating that AI can diagnose disease ‘just as well as’ doctors, which she feels implies that doctors can be replaced by such technology.
 
‘The patient–doctor relationship can never be replaced by AI,’ she said. ‘It’s a relationship based on trust, connection and understanding.
 
‘It’s a special relationship that’s part of the healing process that you can never replace.’
 
Dr Wendy Burton was equally outraged at the way the story was framed, calling it ‘clickbait’ and ‘inflammatory’.
 
She told newsGP she can see a place for AI in medicine, but also worries about the implication of these kinds of stories.
 
She said AI can be useful in certain parameters, such as making a diagnosis on medical imaging.
 
‘But what a machine cannot do, is it cannot take that diagnosis and then put that into context for the individual in front of them with their infinite variabilities and other comorbid factors and their personality,’ Dr Burton said.
 
‘A machine can’t do that. It’s just too variable.
 
‘I think it’ll be quite some time before we’re comfortable with allowing a machine to make a diagnosis.’
 
Even if AI can diagnose specific issues as well as a doctor, Associate Professor Kotsirilos believes nothing should stand in the way of a patient–doctor relationship.
 
‘It’s not just about a diagnosis,’ she said. ‘When we know the patient well, GPs recognise when stress or psychosocial issues are behind a disease, something AI may not be able to recognise in a diagnosis alone.
 
‘Studies have shown that when we regularly see a patient over a long period of time and we’ve got that trust and connection, they’re more likely to heal.
 
‘You can’t replace that.’

artificial intelligence healthcare technology