The inconvenient truth about AI in health
Health is probably not ready to use Artifical Intelligence because the technology cannot re-engineer the incentives, systems and structures that drive behaviours in the health system, says the head of digital health at Healthcare Holdings.
Lloyd McCann gave a keynote presentation on AI in Health at the HiNZ 2019 Conference last month where he told attendees there are a number of questions yet to be answered in this space.
He said the real inconvenient truth about AI and health is that, “we're probably not ready to actually use this type of technology, because we're still designing and deploying these solutions into a current system: utilising our current paradigms in our current mindsets around this.
“Despite the potential and the awesome opportunities AI offers, AI cannot re-engineer the incentives or the systems and the structures that exists today that drive our behaviours within our health systems."
McCann said there is already a lot of AI being used in New Zealand’s health system today.
ACC has deployed an algorithm that assesses all of the claims that go into the corporation; eMental Health platform Clearhead uses an AI chatbot to screen users for mental health issues; and Healthcare Holdings is collaborating with three organisations in the AI space, using the technology to augment clinical decision making.
“There is no sign that the use of AI, or machine learning algorithms is going to slow down, the opposite is in fact going to happen, it's likely to speed up,” he told the audience.
A recent report released by AI forum on AI in healthcare demonstrated that there is a lot of money flowing in the sector within this domain in health.
“And yet there is a relative paucity of evidence to support its use in healthcare”, said McCann.
However, while there is a lack of robust evidence in the space, he also raised issues with the way evidence is collected as the traditional model of randomised clinical trials does not take into account that technology can allow things to be assessed in almost real time.
Other issues were around the lack of consensus around what AI means in the healthcare context and who is going to be responsible for the decisions these algorithms make?
Also, “how do we manage bias in algorithm development?” asked McCann.
“There is a narrative that almost tries to blame the algorithm for that bias, the blame doesn't sit with the algorithm, the blame sits with us and the datasets we've utilised to develop those algorithms because our data sets, aren't necessarily representative of the populations we're trying to serve.
He also spoke to issues around equity and when to deploy AI.
“We have not necessarily truly defined the problems we're trying to solve by utilising this technology,” McCann said.
“And because we've not defined the problems we're trying to solve. We can't prioritise which problems are the most important.”
Watch Lloyd McCann’s presentation on the eHTV
Note – you must be a HiNZ member to access this