AI will have to prove itself to gain greater traction in healthcare sector, say industry players
Until there is greater confidence in technology, it can still be used in areas that are less critical such as capacity planning
ARTIFICIAL intelligence (AI), especially in frontier technologies such as large language models, has gained traction in many different industries as potential game changers.
Yet, if history is any indication, any new technology in the medical field will have to prove itself through rigorous testing in order to see wider use.
Dr Fong Sau Shung, a specialist in general surgery at Raffles Medical Group, said that historically, technologies have to clear very high bars of scrutiny before they are widely adopted.
For instance, he said, a surgical robot was introduced in the 2000s to much fanfare. But it was only in the 2010s that it finally gained prominence in the field of rectal cancer treatments. This was after South Korean doctors began making a name for themselves with such surgery.
Dr Fong described adoption in Singapore as “trickling down” and believes that today, maybe slightly over half of surgeons would say that the robot does a better job than a surgeon conducting surgery with their bare hands.
“You need a lot of time for the studies to be done and verified, published, and then accepted by doctors before it becomes used by the masses,” he said.
Dr Fong said that certain technologies, such as the GI Genius system that is implemented with colonoscopies at Raffles Hospital, are more readily accepted as it has a narrower use case. The system has been proven to be 30 per cent more likely to detect a polyp.
GI Genius works by processing images taken from a colonoscopy and uses machine learning to highlight potential polyps and tumours on screen, just as a camera’s facial recognition software would detect faces.
However, he believes that current AI projects that aim to scan a patient’s history and come up with diagnoses will take a long time to reach the masses.
“We’re all human beings and there will always be a certain belief that the human is the best,” he said. “Unless it is proven irrevocably, acceptance will be affected.”
Encouraging engagement
Dr Dean Ho, director of the Institute for Digital Medicine (WISDM) at the Yong Loo Lin School of Medicine, described AI prediction as “critical”. But without implementation, it will be “purely academic”, he said.
For such technologies to be more widely adopted, he said that technologists will have to work closely with the medical field and other stakeholders, such as regulators and even the insurance industry, to implement solutions.
He noted that the first person that WISDM hired was the head of health innovation Yoann Sapanel, who had experience in the insurance industry.
“Let’s think practically. At the end of the day, the payer is one of the ultimate decision makers for any technology,” he said.
Dr Ho said that the problem with the current approach to AI research in healthcare is that researchers are thinking chronologically. Currently, researchers will build algorithms based on data sets that they have access to, generate results for research papers, and then hope that doctors adopt these algorithms.
Instead, he said WISDM researchers have been conducting parallel user engagement studies with not just patients, but doctors, nurses and other healthcare professionals as they build workflows that incorporate AI.
“Ultimate explainability is going to the doctor on day one, not on day 300, and saying ‘how can we engineer a workflow you can actually use’,” he said.
Dr Ho added that researchers should not see regulators as “us versus them” and should instead see the regulatory process as an opportunity for mutual learning.
Singtel’s technology services arm NCS’ healthcare lead Zee Yoong Kang said that even though the stakes for the use of AI in healthcare are much higher, technology can still be used in areas that are less critical, such as capacity planning.
Who’s accountable?
“If the AI misdiagnoses even a routine fever because it missed certain things, who’s accountable?
“We are far from, at this juncture, a universe where people are comfortable to entirely let AI take over. But certainly, things like rostering people, appointments, scheduling and so forth… there are obvious benefits,” he said.
Still, Bain & Company’s Southeast Asia head of health and life sciences, Alex Boulton, said that customer attitudes towards receiving healthcare from digital healthcare companies have improved after the pandemic.
He noted that in markets such as India and China, hospitals were trusted as the single point of care to manage patients’ healthcare. Since the pandemic, digital health companies have “rocketed” up the trust rankings; narrowing the large gap between such companies and hospitals.
“When digital health groups are receiving that level of trust, doctors have to be open to change if they’re going to survive in that new world,” he said.
In certain markets, he believes that AI can reduce costs and increase the ability of the healthcare system to meet people’s needs.
He added that healthcare costs will continue to rise as populations age, with people aged 55 and up spending about six times more on healthcare than those aged between 25 and 35.
He said that there could come a time when healthcare systems will need to adapt to the use of AI, with Ai-empowered doctors potentially requiring less training so that healthcare capacity can be expanded at a faster rate.
While that future may take some time to come to fruition, Boulton said that companies should begin to look at the business problems they can solve and the datasets that they have to see how AI can be integrated into their workflows.
“Test and learn because there’s a lot of upside from deploying this,” he said. “If you could imagine taking away 20 per cent of the nurse’s admin time and putting it towards patient-facing time, that’s huge. That has real revenue upside and potential.”
Healthcare groups such as IHH Healthcare have taken note. The company has allocated US$100 million towards innovation and digital transformation from 2022 to 2025.
Viable solutions
The company’s group chief information officer Linus Tham said that the company seeks to experiment with and provide viable solutions across both clinical and nonclinical settings.
For instance, IHH Singapore’s data team developed an in-house solution that analyses historical patient demand data to minimise understaffing issues to optimise nurse staffing levels.
As an additional upside, it has also created fairer and more transparent work schedules so that work is more equitably distributed among the workforce.
“We find that AI can potentially contribute by elevating nurses’ well-being and satisfaction through strategic resource planning and data-driven decisionmaking, in turn empowering our nurses to provide better patient care,” Tham said.
On May 8, IHH also announced that it has invested in Belun Technology for an undisclosed amount. The company produces an Ai-powered, medical grade at-home wearable sleep test device and software that has been cleared by the US Food and Drug Administration.