From AI-driven diagnostic software aiding rural healthcare centers to machine-learning models enhancing radiology and pathology, technology is automating tasks once solely performed by doctors
Medical practice is based on protocols that help doctors diagnose conditions. A protocol is an algorithm and it appears that machines might be gaining an advantage in diagnosing conditions efficiently and quickly. It is increasingly being recognised that a digitally connected healthcare ecosystem can improve treatment outcomes, especially in large parts of rural India where the lack of specialists often comes in the way of timely diagnosis and medication. Software, developed by doctors and technicians at AIIMS Delhi, is a significant initiative in this respect.
The software can extend specialised diabetes consultation to primary healthcare centres (PHCs) in rural areas. Studies have suggested that India’s diabetes burden, already the second highest in the world, is underreported because half the patients, even in urban areas, are not even aware of their condition. The AIIMS software is one step towards addressing these deficits. It requires the local level healthcare professional to feed in patient data on risk factors like blood pressure, cholesterol, and blood sugar. It then uses AI to process the information to suggest the treatment or further consultation, as the case may be. All this, however, does not attenuate the significance of the attending physician. The doctor will be required to adapt the software’s advice. The digital intervention’s potency, therefore, is dependent on the quality of analogue care.
The country’s healthcare system requires accessible PHCs run by quality professionals, empowered by digital resources. With this statement, I will talk about the shape of jobs in the field of medicine in this era of AI and GenAI. The advent of AI is posing serious concerns about jobs in future. Some people will become more productive and some roles will be replaced. I have a different take on this.
I hypothesise that AI is automating tasks, not jobs. Let us take the example of a Radiologist—he reads X-rays, sees patients, maintains machines, and mentors younger doctors. Computers analyse X-rays and other medical diagnostics more efficiently and accurately than human experts. The task of reading x-rays can be done by AI, but not the other parts of a radiologist’s job. So I suggest that while identifying areas that will be impacted by AI, we need to break down a job into tasks.
We have to accept that most jobs will likely have at least some of their tasks affected by LLMs.Pathologists are receiving scanned slides of biopsies accompanied by critical data. Then the pathologist only has to focus on the more complex issues to generate a diagnosis. The positive side of this is that speed and accuracy will increase. The downside is that if pathologists are no longer required to evaluate the basic histology elements themselves, the skill to do so will gradually be lost and ultimately they will become dependent on machines, the output of which they will not be able to even corroborate. In this context, there is a need to redefine the job profiles, and we will have to give opportunity to health workers to develop new skill sets relevant to shifting economies and labour markets. Emerging market realities call for major changes to the medical college curriculum. The advent of AI threatens to alter not only how subjects are taught but also what is taught. A recent survey reveals that 60 per cent of medical students are already using AI. However, they need to appreciate that all information given by AI is not accurate; AI and GenAI also suffer from the GIGO rule (Garbage In Garbage Out). Doctors also need to be taught where and how they should use AI and GenAI tools and apps.
There is an urgent need to revamp the medical educational architecture of our country, integrating it with modern technologies and national priorities.Software like the one developed by AIIMS can become more potent if synced with the electronic repository of the National Health Programme for Diseases. Data portability is in line with the Digital Health Mission’s objective of facilitating seamless interaction between medical experts.
A second caveat: Given the asymmetrical relationship between patients and health service providers, the system should be insulated against data confidentiality breaches. Data is crucial for AI to evolve. Machine learning is the field that trains computers through analysis of large quantities of data. There is plenty of data available in our country. This data is being acquired by seducing the public to part with it voluntarily, people are tantalised with online carrots and free to use apps and their responses are monitored, tracked and recorded. Most people happily and voluntarily give up their private data, often without realising it. The digital capitalists use several pretexts and constantly reassure the public that data collection is for their good.
For instance, in the name of developing custom diets, lots of personal medical information is being captured. Genetic data from a country’s immense biodiversity is being collected as part of studies and medical collaborations with foreign organisations.
Another interesting development, COVID-19 forced physicians and their patients to use telemedicine, and the response has been universally appreciated. It might trigger your thoughts when I say that telemedicine can capture new kinds of patient data such as facial expressions, emotional states, etc. The pandemic also accelerated the commercial availability of health and fitness wearables which facilitated doctors—supervised health monitoring, remote diagnostics, and even certain treatments. With these new kinds of devices, there will be new kinds of data captured as well. A word of caution here—we need a techno-legal approach to unlocking data. We need to work out modalities of Confidential Data Clean Room—Sangam of data and model where secrecy is maintained.The cause of concern is that in this game one player is largely ignorant. The suppliers of digital services understand the game and play it skillfully, while most doctors are not even aware that the interests of patient may be at odds with those of the digital and pharma companies. AI has entered the medical field so rapidly and unknowingly that it seems that doctors have accepted interactions with it without due diligence or in depth consideration. They are giving away their knowledge, research and data to AI apps voluntarily. AI may be capable of outstanding performance in terms of speed and accuracy, but all of its operations are built on data, ie, knowledge derived from experts in the field ie the doctors. We should realise that while AI is essential for meaningful progress in diagnosis and understanding of disease mechanisms, keeping AI within boundaries is essential for the survival of ethics in medical profession.
(The writer is UP’s Agriculture Production Commissioner. Views expressed are personal)
