Editor’s note: This article includes insights from Healthcare Dive’s recent live event, “AI in Healthcare.” You can watch the full event here.
Although healthcare embraced artificial intelligence-backed tools when they hit the scene four years ago, actually implementing the technology among payers and providers has come with its own set of challenges.
The tools — which promise to alleviate clinician burnout, streamline claims processing for payers, help with administrative tasks and more — are expensive. They also raise ethical considerations. What happens if an AI tool makes an error? And what happens if that error is catastrophic for patients?
Healthcare organizations, including payers and providers, need to consider multiple fronts when looking at AI investments, including clinician trust, governance and costs, experts said during panels hosted by Healthcare Dive on Nov. 4.
Here are four big considerations healthcare organizations should consider when adopting AI tools.
1. How to get clinicians to trust AI
Clinician buy-in for AI tools is one of the largest bottlenecks toward the technology’s adoption in the industry. As hospitals and health systems look to expand their use of AI tools, clinicians have remained skeptical about the technology: About 41% of 1,000 physicians surveyed by the American Medical Association in 2023 said they were equally excited and concerned about healthcare AI tools.
AI tools are accepted the most among clinicians when they’re developed by clinicians themselves, according to Micki Tripathi, chief artificial intelligence implementation officer at the Mayo Clinic.
At Mayo Clinic, a tool designed to support nurse-to-nurse patient handoff summaries during shift changes achieved “tremendous adoption” among clinicians — largely because it was built by nurses themselves.
“They themselves thought of the idea, they conceptualized it, they worked on the development of the product,” Tripathi said. “They were able to socialize it with their colleagues to be able to roll that out”
Still, it’s not easy to get widespread buy-in, especially among clinicians who work in specialized fields, Tripathi said.
For a cardiologist at Mayo, which has 43 cardiac divisions, regular AI tools just might not have enough expertise to help with their job.
“That’s where we face some of the bigger adoption issues,” Tripathi said.
2. How cash-strapped hospitals can adopt agentic AI
Agentic AI has been the big buzzword in healthcare for the past several years. The technology, which can complete more autonomous tasks, has been heralded as a solution for the industry’s complex workflows.
But hospitals struggling with their financial viability, especially rural hospitals, often lack the means to adopt the technology, even if those facilities could benefit the most from agentic AI.
If cash-strapped hospitals are deciding where to invest their limited budgets into agentic AI, they should prioritize administrative technologies, like those that can handle prior authorizations with insurers, according to Dr. Alexander Fortenko, vice president of innovation at Ochsner Health. Investing in prior authorization would alleviate some pain points for both patients and providers.
Strained hospitals should also look toward tools that help with revenue cycle management, said Lydia Daniels, VP of analytics and AI at Epic.
Revenue cycle tools can translate directly to returns on investment, unlike other AI tools.
“Coders have capacity to do more, to take on additional specialties. You’re not necessarily lowering your headcount, but you’re doing more with what you have,” Daniels said. “And in rural medicine, that is the only strategy you can have, is to do more with what you have.”
3. Does AI actually save healthcare costs?
For all the talk of AI creating efficiencies in healthcare, it’s unclear how much money the technology actually saves. Worse, some payers have been warning that AI, particularly when it’s used by providers, can drive up healthcare costs because the technology is allowing providers to code for more conditions when evaluating patients.
The technology has become another pain point in contentions between payers and providers, with both sides in an arms race to embrace AI to their advantage in claims processing.
Still, even if AI is raising healthcare costs now, it might actually be beneficial for the whole system in the long run, according to Shri Santhanam, chief product officer at Machinify, which makes AI-backed payment tools.
“Does this crisis almost work to our benefit in that it forces a dialog, which would otherwise have been lurking in the background?” Santhanam said. “Because this adversarial dynamic you talked about is not new, right? That’s been as old as death and taxes.”
4. How to view AI governance without federal regulation
A dearth of federal regulation on AI is creating a patchwork of state laws that could dampen adoption, experts said.
There is “no single comprehensive regulatory framework to date” from the federal government, according to David Arnold, senior vice president and chief AI and data officer at Hackensack Meridian Health.
But, “things are coming fast and wild on the state side,” he said, as states attempt to tackle AI regulation themselves. That’s creating a logistical challenge for companies that operate across multiple states, or for others attempting to comply with AI laws that aren’t specific to the industry.
“Without any preemptive legislation at the federal level, we are kind of stuck with the wild west in the states,” Arnold said.
For larger health systems that span multiple states, this can make AI governance “unsustainable,” said Aashima Gupta, global director of healthcare at Google Cloud.
Industry has stepped in to regulate the technology, but companies may be tasked with creating their own governance structures to ensure AI does its job safely.
The best AI governance investments to start with involve education, said Lucy Orr-Ewing, director of policy and strategy at the Coalition for Health AI.
CHAI, which works across the industry to develop guidance for AI, focuses on educating the healthcare workforce on AI literacy. This training allows people to catch potential errors early and comprehensively, Orr-Ewing said. Although CHAI trains nurses and physicians, it also educates cybersecurity, IT and data teams to make up a “multidisciplinary front.”
