Five takeaways on AI, trust, and money from CHIME’s Kansas City Innovation Summit

When WakeMed, a health system based in Raleigh, North Carolina, opened a cancer program, it was full within a week, said Pete Marks, the organization’s vice president and chief information officer. Two rival health systems had fought its expansion into cancer care before state regulators.

The resistance came down to money, Marks said. Every health system must protect its finances, he said, but in a fast-growing region, “there’s enough patients to go around.”

Marks told the story at the CHIME Innovation Summit, held Sept. 16 to 18 at KU Medical Center in Kansas City, Kansas. The theme of the event was co-creation, the argument that healthcare innovation depends on real collaboration across organizations and disciplines. Health IT leaders, clinicians, and a health plan executive spent a day and a half comparing notes, and they did not always reach the same conclusions.

Five takeaways from the Kansas City Innovation Summit

Money is the biggest threat to collaboration between health systems

Keith Fraidenburg, CHIME’s executive vice president and chief operating and innovation officer, asked a Friday panel to name in one word what most threatens healthcare collaboration. The answer? Money. 

Marks also pointed to referral management systems built to keep patients inside one organization, even when a competitor nearby offers a better-suited procedure or a far shorter wait. That pressure, he said, keeps patients from being sent where they would be best served.

Chris Harper, chief information officer and senior associate vice chancellor of AI at The University of Kansas Health System & Medical Center, hosted the summit and offered a counterexample. His health system and Children’s Mercy, the region’s pediatric health system, concluded years ago that Kansas City did not need two competing pediatric programs, and a cancer building in development will be a partnership between them. “We just can’t afford it anymore,” Harper said of competing.

Patients are turning to AI for health answers faster than clinicians are ready to trust it

Harper opened his talk with a line he credited to a patient safety leader. Clinicians adopt AI at the speed of trust, and patients adopt it at the speed of desperation.

He described his father, a military retiree with a heart condition, nodding as his cardiologist referred him to a neurologist and then asking on the way out what neurology was. Patients confused by their own care now ask a chatbot before they ask a provider, Harper said, which makes AI the front door to health information.

Harper’s trust-and-desperation line followed the group through both days. Priscilla Frase, MD, chief medical information officer and hospitalist at Ozarks Healthcare, said it cuts both ways. It argues for taking some risks a cautious organization would otherwise avoid, because no one wants desperate patients, but it also gives a reason to stay conservative in areas where trust is at stake.

When an AI agent goes down, health systems lose the workforce it replaced

During a panel on operational efficiency, a participant asked how health systems plan for the hours when an AI agent is unavailable. Two panelists called it a serious risk that the industry has not yet solved.

Brian Sterud, chief information officer at Faith Health, drew a distinction. When a traditional system goes down, the staff who use it are still in the building and can fall back on paper. When an agent goes down, the system and the workforce disappear together.

Seth Katz, chief operating officer of UnitedHealthcare’s Community Plan of Kansas, described the call center version. An organization shrinks its staff as the agent takes on more calls. Then the app fails, members start calling, and three people are left to answer. “Everything works great until it doesn’t,” he said, adding that his organization is still working out how to plan for that day.

Rural hospital partnerships depend on in-person trust before technology

Harper credited his health system’s standing with rural hospitals to a Kansas physician who, by his account, put 100,000 miles on his car in two years visiting hospital leaders across the state. To earn a rural provider’s trust, Harper said, “you have to sit with them and talk to them.”

Emily Warr, system administrator at Medical University of South Carolina Health, described a telehospitalist program that places an advanced practice provider on site at small hospitals while a remote hospitalist rounds on patients daily. When Medical University of South Carolina Health had no open bed for a patient in liver failure, the remote hospitalist cared for him overnight at a critical access hospital. He transferred the next day, received a transplant, and survived.

Frase, speaking as a rural leader, ended the summit still skeptical that partnership will bring lasting change for organizations like hers, though optimistic about what a different approach could produce.

Clinicians edit AI-generated notes less as they come to trust the tool

Melissa Jost, director of clinical informatics and clinician health and wellbeing at UC Davis Health, said her organization studied the quality of clinic notes drafted by an AI scribe. The published study found 94.7 percent of 356 physician-reviewed notes free of significant errors. Jost’s concern is what happens as clinicians come to trust the tool and fewer of them edit its drafts.

Denton Shanks, DO, associate chief medical information officer for health AI and technology innovation at The University of Kansas Health System, asked how many errors physician-written notes contained before AI. He recalled charting from 11 p.m. to 1 a.m. after a clinic day and reconstructing visits from memory. “I’m sure that I left out lots of things,” he said. Jost said her team never measured that baseline because reviewing those notes would have been too time-consuming and expensive.