The machines are coming, and no, it’s not a Terminator-like apocalypse. Artificial intelligence is poised to transform healthcare — in some areas, it’s already doing so — and that means doctors and clinicians are wrapping their heads around automation, machine learning, advanced algorithms and predictive analytics. The future is here.
The technology has the potential to improve patient care, and provider performance, by helping clinicians make decisions based on reams of data and patterns. Automated tools are simplifying tasks and allowing doctors to focus more on the clinical side of things.
But to get the most out of AI, doctors and nurses need to be properly trained. Their already-important skill sets need to be expanded if the technology is to be leveraged properly. That’s a challenge, and it requires a culture shift.
WHAT THE EXPERTS SAY
Dr. Clemens Suter-Crazzolara, vice president of product management for health and precision medicine at SAP, said any retraining efforts need to keep in mind the various stakeholders affected by AI. After all, the software should be there to support the person, not the other way around.
“That means you have to take this person on the entire journey with you,” he said, “and make sure they understand, ‘This is how I do things at the moment, this is how the algorithm is being built, this is the KPI that is being measured.’ (With) a lot of the AI algorithms, you have to be very open about it.”
Because of that, retraining efforts need to begin while the algorithm is being built, not after. The more closely management educates people and works with them throughout the transition, the more seamlessly it will be integrated into clinical practice.
PeriGen CEO Matthew Sappern sees retraining as a challenge that’s largely predicated on the usability of the AI interface, particularly when it comes to things like alarms. And the designers of the software need to keep in mind that things have to fit relatively seamlessly into an existing workflow.
“That user interface is so important,” said Sappern. “How intuitive can you make this information? … When we try to design tools to look at the patient, to us it’s the degree of abnormality over time — what’s the trend?
“The ability to generate with great accuracy a readout on a patient is critical, but even more critical is how you communicate that information,” he said. “It has to fit into the workflow, because you can’t really change the workflow. It’s like pushing a string uphill. You just can’t do it. How do I put the alert where the eyeballs are? That’s a real challenge.”[…]