The 2 Big Myths About Using AI in Healthcare

 

Danielle Walsh, MD, FAAP, FACS, a professor of surgery and vice chair of Surgery for Quality and Process Improvement at the University of Kentucky College of Medicine in Lexington, believes there are two big myths about using AI in healthcare: that physicians don’t want AI and that patients don’t want AI.

“I admit, physicians really got burned in the transition from paper charts to electronic medical records (EMRs),” Walsh says. “It was a traumatic transition. And many physicians are still very frustrated with EMRs because they feel they spend more time documenting than caring for patients. Physicians worry that AI is going to similarly not deliver on its promises and create more burden than good. But I think the opposite: AI has the potential to undo much of the frustration of documenting in the EMR through tools such as ambient listening that not only transcribes but interprets and summarizes what is happening in a patient encounter.”

On the other hand, multiple surveys of the public indicate that patients fear AI in healthcare because of confidential information being leaked or used inappropriately. “Therefore, the onus is going to be on healthcare systems and oversight bodies to ensure well-built systems protect patient confidentiality, while allowing AI to use de-identified data to better understand patients and diseases,” Walsh says. “The public needs to become more comfortable and confident with AI tools through demonstration of safety.” 

A major challenge to implementing AI is the interoperability with existing systems. “You don’t want to have to log into seven different things to get through your one clinical encounter,” Shah says. “We need to integrate, including data integration, through AI tools and staff training. This a major paradigm shift, with costs that are not insignificant.”

Strong governance, collaboration between the key stakeholders (clinical, operational, privacy, security, finance, research) and investing in a strong data infrastructure are crucial to a successful AI program. “You also need to foster a culture of innovation and adaptability because things are changing rapidly in the AI space and will continue to,” Shah says.

As healthcare systems consider enterprise adoption of AI, its use should follow the needs, and not a race to be first, according to Greenhill. “Interestingly, we are only scraping the crust of AI’s full potential in healthcare, with mostly task-focused weak or narrow AI,” he says. “Even ChatGPT, for all its buzz, is still classified by many in the industry as weak AI.” And the industry is slightly restricted in using more robust AI tools, due to poor interoperability and inadequate data governance strategies.

Meanwhile, there are notable deficits in digital and emerging technology leadership competencies to create the AI-centric healthcare organization. “The reason that competencies for healthcare executives are important for AI in particular relate to what I refer to as the ‘unknown unknowns’ of the technology, plus knowing what questions to ask about the algorithms,” Greenhill says. “Knowing the right questions to ask is vital for technologies like AI, due to its far-reaching implications for patient safety, equity and quality care.”

Far-reaching implications indeed. “I am convinced that looking back 10 to 20 years from now, medicine as practiced today will look like the Middle Ages, thanks in large part to the looming advent of AI,” says Fuchs of Mount Sinai.”

“With AI, healthcare leaders need to rethink messaging to encourage the workforce and reduce fears associated with potential job losses due to the technology. I think most organizations want to maintain their best and brightest talent while having them work at the top of their training and licensure. AI tools will be an enabler for this, as well as create new opportunities,” says Greenhill.