Cardiologists and cardiothoracic surgeons at Cincinnati Children’s are exploring how to harness artificial intelligence (AI) to improve the hospital journey and outcomes for heart patients. Clinicians and researchers are studying possible ways AI could extract and analyze data from heart surgery “Flight Plans.”

“The potential of using AI to improve care is tantalizing, but the true application of these methods really needs to be thoughtful,” says Michael Gaies, MD, MPH, prior medical director, Acute Care Cardiology Unit and director of the Heart Institute Data Center. “We need to study them rigorously to make sure that any conclusions we generate from AI are valid and are safe.” Dr. Gaies is now the executive co-director of the Heart Institute and division director of Pediatric Cardiology.

Flight Plans are a type of “threat and error” system developed by NASA. A handful of healthcare systems are adapting the method to make surgeries safer for all patients. The nature of the potential Flight Plan AI platform offers opportunities for leveraging AI to make distinct connections about patient conditions and potential care plans and outcomes. 

Cincinnati Children’s has several efforts currently in the research and development phase, although patients and families won’t be exposed to the AI until proper vetting and approval takes place. 

“We’re on the cutting edge of this technology,” says Ryan Moore, MD, MSc, director, Digital Health Innovation. “We’re working to connect the dots where there are proven and potential solutions and look at how we move forward.”

Taking Flight Plan to the Cloud

Flight Plans at Cincinnati Children’s create a visual representation of a patient’s surgical journey. The plans are built by using a subset of clinical data about a patient. The entire patient care team—including cardiologists, interventionists, nurses, surgeons and others—contributes data to the Flight Plan during the Case Management Conference. 

A Flight Plan provides a way to review a patient’s surgical path through the hospital and identify areas that can be improved. It gauges outcomes and team performance and reveals areas for improvement. These are discussed at a weekly conference where every patient who undergoes a surgical intervention is discussed. 

“Flight Plans are a way we learn about our decision making, our execution and our team performance,” Gaies says. “This informs our decisions for the next patient, how we pick the best treatment strategy and how we tailor our team to execute that strategy.” 

It’s important to note that all available data, however, are currently limited by how they can be accessed. The methods are “wildly inefficient,” Gaies says, and limited by what a human can find in individual records.

To expand Flight Plan use, Moore; Brianne Reedy BSN, RN, Flight Plan program manager; and David Morales, MD, executive co-director of the Heart Institute, received three rounds of grant funding. Two rounds came from the Ohio Third Frontier Technology Validation and Start-Up Fund, to design and develop a cloud-based Flight Plan AI platform with a Python backend to leverage AI model integration. The funding allows the team to work closely with AI tech industry leaders SFL Scientific/Deloitte and Microsoft, to enable easy Flight Plan access through the cloud and embedded AI solutions to reach as many clinicians as possible.

Harnessing the Power of AI

Moving forward, Moore envisions using the Flight Plan AI platform to predict the course of a patient’s hospital stay and adjust as needed when changes arise by harnessing all available data. In addition, he believes that “Generative AI,” a category of AI recently popularized by ChatGPT, will revolutionize the way we interact with our data. 

“The ability to ask challenging questions to the Flight Plan AI database and receive an educated and immediate response could drastically improve patient care and open the door to more AI research,” Moore says. “However, there is an equal amount of concern and enthusiasm, so a lot of research will have to go into limiting bias in generated content and ensuring that the AI we are leveraging isn’t harmful or misused.” 

Moore is involved in many hospital level AI initiatives, including one to identify bias and inequity in AI Modeling, to ensure appropriate use is employed in the healthcare setting. 

In the future, AI might rapidly sort thousands of pages of medical text to extract information and answer questions. 

“AI can reveal patterns that can’t be observed even by the most experienced clinician,” Gaies says. 

Marrying AI capabilities with a cloud-based Flight Plan AI platform will also allow for what Moore calls federated learning—the ability for multiple centers to share and train AI models to further inform decision-making and hopefully improve individual outcomes, while still allowing each center to keep their patient-sensitive data local and secure.

Research Underway to Prove the Concept

Improving personalized care drives the push toward AI, which can potentially analyze information and provide insight on patients with unique diagnoses and underlying genetic makeup.

“There are subtleties to each individual patient that could impact the best treatment for them at a certain time of their journey,” Gaies says. 

Moore and Gaies envision a patient- and family friendly AI-enabled Flight Plan—a teaching tool that illustrates and tracks a patient’s journey through treatment—much like the screen on an airplane that shows where a plane is flying and where it’s headed next. 

Cincinnati Children’s has several research studies planned to explore how best to leverage AI. Research will answer questions such as:

• How does AI collect and coordinate data?

• How do physicians use it?

• How does an AI machine reach conclusions?

• How do we know if AI is responding to the right data?

Gaies is studying Flight Plan performance scores and how they relate to different patient populations. And he’s building a comprehensive data structure in the Heart Institute to incorporate as much clinical and genomics data as possible. As AI tools are developed, Moore sees opportunities to retrain an AI machine in real time, by using all the data available at any given moment. “We really hope to learn the best way to make decisions about patients and try to disseminate that widely across the field and around the world to all places where kids are cared for,” Gaies says.