Curious Iowa: How are AI tools being used in health care?

Dr. Shiny Mathewkutty, a cardiologist with Mercy Cardiology Clinic in Cedar Rapids, has noticed a shift in how her patients are absorbing the information she shares with them.

“I’ll sit down with the patient, I’ll review all the results with them … then they’ll go home, they’ll input all that information to ChatGPT and then call me back with the ChatGPT analysis,” Mathewkutty said, noting that ChatGPT is not used by health care professionals because it is not secure.

“So, care is changing, because the questions that we’re getting are very different from the questions that we got a year ago, and patients are understanding their disease process better,” Mathewkutty said. “They’re trying to be more involved.”

While doctors are not using chatbots like ChatGPT, they are using AI tools. Les Beck, a Gazette reader from Cedar Rapids, wondered how local hospitals are using AI, what they’re learning from using it, and how health care professionals anticipate AI tools will evolve. So, Beck wrote to Curious Iowa, a Gazette series that answers readers’ questions, to find the answer.

We spoke with doctors and executives from Mercy and University of Iowa Health Care to learn how they use AI tools.

When did area hospitals start using AI tools?

  Buy Photo

Dr. James M. Blum, anesthesiologist and chief health information officer with UIHC, told The Gazette that AI is not new in health care.

“We’ve used predictive analytics literally for decades inside health care … we’ve used mathematical techniques to determine risk for patients since the 1970s,” Blum said, acknowledging that there are different types of AI.

For example, generative AI uses large language models to generate or summarize text. agentic AI can replace actionable tasks. Image Analysis AI can examine images, like identifying the contours of a tumor or measurements of a wound.

The hospitals we spoke to had three main use areas for AI tools: note taking, chart summary generation and image analysis. These tools are largely used to cut down on time spent on administrative tasks.

“We’re not letting AI take actions on patients’ clinical care. It’s always guided and approved by a physician beforehand,” Jeff Cash, senior vice president and Mercy’s chief information officer, said. “It’s just providing support information that’s more helpful for them as they sift through large volumes of data or (making) behind the scenes tasks that could be assisted by something like AI in the right context.”

Both UIHC and Mercy began using generative AI tools in 2024. Both use AI tools in outpatient services and are trialing AI use in inpatient services.

“After a year of using this technology and caring for literally hundreds of thousands of patients … we have not had one instance of finding an area where a patient’s care was compromised through the use of this technology,” Blum said. “But I think that’s because of the education that our providers receive and the diligence that they’ve had in over-reading everything that’s generated from these records. And I think that’s very rare that you could say that about a new technology being introduced into health care.”

Which AI tools are Iowa hospitals using?

  Buy Photo

Tools like DAX Copilot, used by Mercy, and Nabla, used by UIHC, can write the first draft of a doctor’s notes during patient encounters, and all Mercy and UIHC doctors have access to this technology.

Both Mercy and UIHC have met with the companies of the ambient transcription tools to provide feedback in order to make improvements.

“Initially, the tone was robotic and automated and we don’t like that because our conversations are not like that,” Mathewkutty said, “so over time, the tone has improved. It captures more of the emotion of the conversation and it’s much better about figuring out exactly what was said in what context it was said, because it’s not just words. It’s also about the weight of the conversation itself.”

Mathewkutty said that humans write better notes, but with AI tools providing a first draft, doctors can complete notes quicker.

“And it’s particularly important because when my notes are done and signed, the patients can review it when they go home, and if, God forbid, … they happen to go to the emergency room the next day or that night for any reason, the doctors have access to my notes, which are all done, and this way they don’t have to wait for a few days to find out what happened,” Mathewkutty said.

UIHC was one of the first hospitals to pilot the tool Evidently, which helps doctors efficiently review a patient’s chart and aggregates the details. Mercy is in the trial phase of implementing a similar tool.

  Buy Photo

“In medical records, the data is scattered all throughout many different places … I think the average chart is now said to have something like 15,000 words,” Dr. Jason Misurac, a pediatric nephrologist with UIHC, said. “If you’re seeing 10, 15 patients in a day, that’s a volume of reading that you can’t keep up with … if they’re reviewing it manually.”

Misurac said that using tools like Nabla and Evidently has “made my life as a clinician a lot better,” increasing the amount of undivided attention he can provide patients and their families and decreasing the amount of time and mental energy spent writing notes.

Both hospitals use the electronic health record product EPIC, which has elements of AI built into it. Blum said AI helps draft responses to questions from insurers or patients.

“And the AI really helps us generate a better draft note, typically more polite, typically more humble than what you would see potentially with notes that are just drafted by a busy clinician or someone that’s trying to work in the back office to get someone’s procedure approved with an insurance company,” Blum said.

There is a large administrative burden placed on hospitals when working with insurance companies. Cash said that AI is being used to look through doctors’ notes as bills are prepared to ensure that the bill is at the right level of specificity.

What do hospitals consider before acquiring software that uses AI?

Both Mercy and UIHC said that new software goes through a standard security review, but additional oversight is given to programs with AI.

“And we have a special AI addendum … that asks a lot of deeper questions like where’s the data stored, is it accessible by people other than us, those sort of things,” Cash said.

Blum is apart of UIHC’s nine-person AI acquisition team. Members have training on AI, he said, to ensure “that we aren’t buying snake oil.”

“That’s what our focus is, that we’re buying technology that seems appropriate to the use case, that’s being marketed for things that are FDA approved,” he said.

Mercy has a similar group of doctors that makes up a technical advisory team. These doctors are trained on AI and test AI programs before they are fully rolled out for physicians.

How do AI tools protect patient privacy?

  Buy Photo

Cash said the systems they use store patient data in “private tenants,” meaning data is not shared with outside organizations.

“When we send stuff in to be processed with AI and it comes back, the data then gets removed from the AI system,” Cash said, “so it doesn’t sit there as a big repository of that kind of data.”

Additionally, AI systems are trained on non-specific patient data to build workflows.

Blum said that while Nabla creates a transcript of a conversation between a provider, and patient it doesn’t contain a voice recording of the encounter. Additionally, Nabla deletes the transcript from its servers after 14 days.

“So in the very unlikely event that there was a breach, there would be a limited number of patients affected,” Blum said.

Misurac said retaining the transcript for two weeks is helpful to providers.

“So we treat what comes from Nabla as a draft, but it’s been really nice because I’m less concerned I may forget something,” Misurac said.

How have patients reacted?

Patients are informed when a provider uses an ambient transcription tool like DAX Copilot and Nabla and they have the opportunity to opt out.

Blum said the number of patients expressing reservation and asking that AI tools not be used has been “shockingly small.”

“I don’t know what it is here in Iowa, we just seem to have patients that respect our decisions and that we’re doing the right thing for them and they have trust in our health system,” Blum said, ”because when I talk with colleagues around the country, there’s more concern from patients than we see here.”

Mathewkutty has had a similar experience to Blum.

“And I think one of the reasons why (patients react positively) is when I walk in, … I’m talking to patients face to face, so they like that a lot better than me sitting there staring at a computer screen,” she said.

Blum said he makes himself available to talk with patients about their concerns over AI use and explain how the technology works. He said in the year that UIHC has been using these generative AI tools, he’s had just one conversation.

How do providers see AI use evolving in health care?

In the near future, Cash said that Mercy will be trialing chart summary tool Ambient for Nursing. Equipping nurses with AI tools aims to alleviate the burden of administrative tasks and improve efficiency while hospitals across the country face a nursing shortage.

Cash also noted that AI programs are constantly undergoing upgrades in the background as companies receive feedback from health care providers using the tools.

“I think you’ll see a renewed focus in diagnostics and the combination of AI with other features to help us better diagnose and create correlations between certain elements that we don’t have as much information on today, like genetic information,” Blum said, “and how does that information really deal with your prognosis as a patient in developing a disease state much later in life.”

Blum said the combination of increased information and genetic testing will “fundamentally change the way that we look at disease” and allow doctors to focus more on precision medicine and targeting people before they get sick.

At its annual meeting in August in Verona, Wisc., EPIC announced 160 new AI projects, including patient-facing chatbot Emmie. EPIC is the largest electronic health vendor in the U.S.

Cash said that being an EPIC customer gives Mercy access to a valuable collection of data.

“They’ve built their own private cloud of data and they have almost 300 million patients that are in there today with actively validated, de-identified patient records,” Cash said.

This is beneficial to providers because of the level of the detail included — down to the medications used and outcomes.

“We’re able to start to predict a lot better where we should be providing interventions and care for patients,” Cash said. “It’ll be interesting to see what they (are) able to produce out of that data over the next couple of years.”

Blum said it’s “definitely conceivable” that we’ll see devices with AI assisted diagnostics before the end of the decade.

“And then I do think we will get into the point in the 2030s of really starting to see more intervention by AI inside health care, potentially with things where AI will assist with potentially surgeries,” Blum said, giving the example of a robot identifying certain structures and labeling them for the surgeon and telling the surgeon “don’t cut here.“

Have a question for Curious Iowa?

Tell us what to investigate next.

Comments: bailey.cichon@thegazette.com

Sign up for the weekly Curious Iowa newsletter to read answers to reader questions — along with behind-the-scenes observations.