I was lucky enough to be in a medical school with a mentor-ship program. My mentor was a seasoned physician and an expert diagnostician, and he certainly knew it! In an OPD that was more reminiscent of the Kumbh Mela than the sterile, aesthetic hospitals on the silver screen, he would call a patient in and perform a comprehensive clinical examination while simultaneously taking a clinical history. At the end of 90 seconds he would begin writing a prescription and ask me what my probable diagnosis was. “Tuberculosis”, I would invariably quip given that in India you can safely blame the mycobacterial menace for absolutely any sign or symptom in the human body and you certainly wouldn’t be wrong. He would then shoot me a look of utter repugnance and theatrically go on to describe ‘raspy’ coughs, ‘pink frothy’ sputums and ‘woody’ dullness-es . Short of mentioning ‘smoky flavours’ and ‘tannins’, one would be forgiven for assuming that he was referring to a fine bottle of Bordeaux rather than a pathology. He was from a generation that is said to be on the decline (his own words, not mine). He would always end with something along the lines of, “the more patients you see the better the clinician you will be!”
It’s been a couple of years since those days of yonder and technology seems to have defied Moore’s Law in every regard. The latest clinical quandary is ‘Will artificial intelligence render doctors obsolete?’
This ‘leitmotif’ has been doing the rounds since time immemorial, first it was X-rays, then it was the internet, now its AI. The replacement of the healthcare practitioner has always been predicted but in vain. It has always been met with raspy sarcasm and scepticism by the clinical community. I’m curious though, if AI can do everything from accurately predicting my shopping behaviour, travel inclinations, choice of books, genres of music, cinematic inclinations, food predilection, pub proclivities, dating and matrimonial preferences to driving a car and efficiently running a factory, why can’t it give me accurate health advice?
I went out and asked doctors what they thought and in keeping with my non-conformist nature prepared a list of counter arguments. I must add that this is in no way a rant and just a way for clinicians/non-believers to see the other side of the debate.
AI needs to be trained and needs a lot of information: Using the estimate of 30 patients per day, 6 days a week, from the age of 25 to the age of 60, a doctor would have seen 321,300 patients which is a huge number indeed. However when the wave of EHR (Electronic health records) reaches a stage of maturity, and it will reach Stage 7 as per WHO sooner rather than later the world over, that will seem but an infinitesimally small digit. To add salt to wounds, AI can now learn almost as fast as a human and process the said information much more efficiently than we can. We’re heading toward the era of ‘Precision Medicine’, strap in!
AI won’t take over in my lifetime: One of the gurus of AI, Geoffrey Hinton, said, “People should stop training radiologists now. It is just completely obvious that within five years, deep learning is going to do better than radiologists.” This was said in 2016 and he predicted the demise of ‘conventional’ radiology given the sensitivity and specificity of image recognition that can be achieved with current technology. Radiology, Pathology and Dermatology are said to be within firing range.
AI can’t replace a doctor’s gut instinct: Clinical medicine in its current form is so rigid with regard to adherence to guidelines and protocols that any deviation from the beautifully designed flowcharts and treatment algorithms is met with chastisement and damnation. Medical litigation (and the occasional physical thrashing) has compounded this stringency and has ensured that medicine is about as gutsy as a goldfish in a bowl.
There is no data in developing countries: Coincidentally I hear this from a person who abhors ‘poor quality’ textbooks and inhales the so called Bible’s of medicine (Note: not the Bhagwad Gita’s) including Harrison, Sabiston, Bailey’s and follows American and European guidelines to the to treat his/her Indian patients with pharmaceuticals meant to cater to Caucasians. In other non-satirical news the Centre for Development of Advanced Computing in India is working on various projects in the field of Health Informatics, involving collating and usage of electronic health data.
AI can’t take a history and conduct a clinical examination: ‘2 minute consultations in India’, ‘The lost art of the clinical examination’ , Abraham Varghese’s much acclaimed Ted Talk, ‘A doctors’ touch’, pretty much encapsulate the supposed declining ‘Art of clinical medicine’. This is besides the fact that the number of investigations conducted year on year is growing astronomically as is the dependence on them as compared to contemporary medicine.
AI can’t replace human trust: The lack of faith in doctors, policy makers and health systems as a whole has been brewing, with the occasional outburst from time to time. In the age of Amazon, Netflix and Uber, clinical medicine being subjected to more and more consumerism; Practo and other ‘physician rating apps’ are a great example. Research states that 40-50% of patients seek a second opinion, over 50% of patients in total, and well over 80% of millennials, will cross check their doctors advice on Google. Trust indeed.
Privacy and confidentiality: We trust Google, Netflix, Amazon and Apple as well as our beloved smartphones with such dogged tenacity that we allow molestation and manipulation of our orifices in the name of security at an airport, if need be, but we’re oh so worried about phones, emails and Aadhar data being hacked. As a species we’ve lost the plot.
AI isn’t accurate: To paraphrase Instagram: #NoFilterNeccessary
Artificial Intelligence Reads Mammograms With 99% Accuracy (Futurism)
Digital Diagnosis: Intelligent Machines Do a Better Job Than Humans (Singularity Hub)
Google AI in rural India in resource constrained settings with no Ophthalmologists
I do not deny that it is an incredibly complex topic with abundant intricacies. There are large parts of medicine and physiology we still cannot comprehend, including the placebo effect, the powerful ’empathetic hand on the shoulder effect’, the ability of the mind to heal and cure to name a few. The artificial neural networks will be (at least initially) restricted by our cognitive inadequacies. I did not open this Pandora’s box to vex those resistant to this change but in the hope that some may begin accepting that if not replacement, ‘humanistic AI’, at the very least, is in the offing . Let’s face the facts, doctors being replaced is an inevitability, if not in a decade by Alexa or Siri, then by a Homeopath in a month or two!
Doctorness 