The biggest risk of artificial intelligence to our health decision-making framework is not that it displaces humans but that we fail to embrace its potential, particularly in making 'clunky' institutional processes more efficient and accessible.
AI in health technology assessment (HTA), for example, should not be additive to an already broken process.
An additive approach would just create new levels of uncertainty and reasons for delays and denials of reimbursement. The obvious risk is that it makes an already inaccessible and exclusionary process more inaccessible and exclusionary.
For some, adding complexity is a profound good, while simplicity and speed are always the enemy. Why? It is part cynical ruse to protect the broken status quo.
Yet it is also a reflex response. The one certainty is that discussions about HTA always lead to more HTA. AI can fundamentally reshape this longstanding reality. With the broad adoption of AI, some form of more HTA might be good.
The argument that some hyper-level of caution is required on the adoption of AI also belies the lessons of history.
Is AI likely to be any more disruptive than the adoption of computing or the internet?
The only sensible approach to its adoption, including in our funding decision-making frameworks, is to contemplate how it can be embraced and applied to make things better. This can only mean better for patients.
Its adoption in HTA and evaluation processes will obviously create winners and losers, as it should. If we are sensible in embracing its potential, there will be far more winners than losers.
As for the constant refrain that AI should be subordinated to humans, which humans?
Have humans really done such a great job? The humans who have created an intransigent and ossified institutional decision-making framework in which it costs over $250,000 just to make an HTA submission? The humans who thought an HTA Review, now meandering into its fifth year, should end with 'Delphi' surveys on the wording of select technical inputs? The humans who have so complicated HTA that only a pharmaceutical company has the financial resources to make a submission and navigate the process?
History shows that the tasks people perform in their professional lives have never been fixed in stone. They are constantly evolving, including with the emergence of new technologies.
After the release of the first electronic spreadsheet in 1979, some predicted accountants would face mass unemployment. Instead, according to Professor Eldar Maksymov at the University of Arizona, after adjusting for population growth, the number of accountants in the US has increased fourfold over the past 40 years.
“In every major occupational group that adopted computers heavily, employment grew faster than in groups that did not,” wrote Professor Maksymov. “Computers eliminated specific tasks within jobs — but the resulting cost reductions created so much new demand that the occupations expanded overall.”
More productivity means more demand.
According to Professor Maksymov, based on the accounting experience, "If the paradox holds again, demand for judgment-intensive knowledge work will expand as AI makes previously uneconomical work viable."
Why?
"AI is now making intelligence itself — the capacity to analyse, judge, and assure — dramatically cheaper," he said.
A recent New York Times article about the potential for AI to end coding as a career was based on interviews with 70 software developers at Google, Amazon, Microsoft and small start-ups. It found that while the technology may negatively impact some coders, many of those interviewed are enthusiastic adopters of AI because it means they can be more productive.
According to the report, “a coder is now more like an architect than a construction worker.” One coding executive interviewed said they can now work further into their 'to-do' list and make better progress on a broad range of new ideas. Others suggested that AI would make software development more accessible to companies and organisations.
“Several developers suggested that the number of software jobs might actually grow," the article said, because the issue has never been the lack of good ideas, but the scale to investigate them.
In other words, AI is making previously unaffordable work affordable. The scale opportunity has made these coders more valuable, not less, because they can do more. They can create more value for more organisations and people.
Apply that to HTA.
As access to HTA becomes cheaper, as it should with the adoption of AI, people will not use less of it. They will find new uses for it, and more will start using it, because what was previously unaffordable and inaccessible will become affordable and accessible.
The roles of HTA professionals will evolve, and we might need more of them.
It will terrify some to imagine an even larger HTA industrial complex. Yet AI has the potential to fundamentally change its character and make it far more accessible. It can also reshape the relationship between HTA 'experts' and the end-user. The end-user, the patient, can regain primacy.
As long as we do not muck it up by using a contrived caution to protect the status quo. The critical question we have to ask is clear. What can we do with AI-powered HTA that we have not been able to do before?
Boehringer Ingelheim's new platform, Patient Input Buddy, released earlier this week, is a critical first step in the potential of AI to open an institution that is about as accessible as an exclusive private golf club.
Using AI to support patient input to HTA evaluation processes is a very positive start. But what next? What about patients using AI to develop their own HTA submissions?
So many talk about patient-centred HTA decision-making. What about patient-dominated decision-making? Why limit the use of AI to the strictures of the broken status quo that treats patients as mostly spectators, invited to participate at a tiny number of points in the process convenient to the institution?
We have to embrace AI because of the immense opportunity. It is critical that we avoid a reflexive caution or an approach that bolts it onto an already broken system. Making it additive would be an unforgivable and generational lost opportunity for the people whose lives depend on better, faster choices, and that means patients.
