AI isn't magic
When it comes to artificial intelligence healthcare solutions, it's important to manage expectations for consumers and users.
"The first thing that's important to realise is that AI isn't magic," said David Champeaux, chief growth officer, Cherish Health, during a panel at the HIMSS & Health 2.0 Europe Digital Conference.
Though AI can improve people's lives, said Champeaux, developers and stakeholders shouldn't overstate its capabilities.
The panel, 'AI Solutions for Consumers', was moderated by Orcha cofounder Tim Andrews and featured Medical Realities cofounder and Chief Medical Officer Prof Shafi Ahmed, ResApp Health CEO Tony Keating, Skinvision business development director Gavin Matthews, and IESO Digital Health Chief AI Officer Valentin Tablan.
Champeaux noted that AI tools can, and should, be designed in ways to augment or facilitate an existing routine. To that end, he stressed that tools should be designed to fit the life of a user, not the other way around.
In doing so, he said, "You'll be freeing a lot of time for the interactions – human-to-human – that AI can't replace."
He also emphasised the importance of avoiding over-promising, saying that users are more likely to be on board with tools that have clear advantages, even small ones, in the shorter-term.
"If you can offer a little bit right now, that builds confidence," he said.
Ahmed explained that much of his communication around AI tools involved engaging with "hope, hype and reality," and matching the former two with the latter.
Framing AI in healthcare as a replacement for clinicians is the wrong move, he said, particularly with regard to building rapport or fostering empathy.
Rather, it's "one technology amongst many others … all coalescing to create better healthcare."
Keating, Matthews and Tablan walked attendees through a demonstration of their products, which leverage AI technologies in order to further patient treatment.
Tablan, too, noted that AI should be viewed as one of many tools. The IESO tool, which helps cognitive behavioural therapists provide care to National Health Service patients in the UK, uses terminology that clinicians are well-versed in to evaluate the success of sessions.
"What we built should be similar to [satellite navigation]," he said. "It tells you where to go, but it doesn't drive your car."
Ahmed also raised the importance of using a diverse data set to train AIs, noting that, for instance, skin condition diagnostic apps might be more accurate for lighter skin.
Matthews, who had demonstrated the SkinVision app for the panel, said that existing studies are lacking in that regard.
It's "very much something that has to be addressed," he said.
For his part, Keating said that although it's important to have an understanding of one's patient base, the studies ResAppDX relied on for its respiratory-disease-detecting algorithm had not flagged major differences for patients of different races or ethnicities.
Champeaux also flagged the issue of potential bottlenecks, noting that the insights generated by solutions might trigger downstream demands for follow-up investigations.
An important question for creators to consider, he said: "How have you engaged in the health systems where you're launching … so that you're managing the end-to-end pathway?"