ViVE 2024: Is Your Healthcare Organization Ready for AI Adoption?

How to Set Up Your Health System for AI Success

Combining innovation and AI raises the ceiling for healthcare leaders’ ability to be creative in problem-solving, according to Chang.

“Traditional problems are mired in traditional thinking. Generative AI really gives you some innovative solutions you hadn’t thought of because you didn’t have time to read about what the world is doing to solve that problem,” he said. “AI, combined with human intelligence, is particularly powerful.”

Marty Paslick, senior vice president and CIO of HCA Healthcare, said organizations should undergo an internal audit to determine if they’re ready, from a governance perspective, to partner in the AI space. Once the organization is ready, whoever is leading the AI charge should develop a portfolio of AI opportunities that has a fair amount of discrete items that can demonstrate short-term value. Implementing these tools can be foundational for delivering midterm and long-term value.

READ MORE: Healthcare embraces AI as industry leaders seek support for clinicians.

“It’s important to set expectations,” said Paslick. “For HCA, we’re not used to investing in technology that we’re not 98 or 99 percent sure is going to deliver value.”

He added that part of his role is to education the organization on the need to be comfortable with failure: “We’ll fail in a way that paves the way for greater achievements in the future.”

Jessica Beegle, health technology expert and former CIO of LifePoint Health, agreed that failure is an important part of innovation. She said that while no one wants to be known as a failure, it’s important to fail, and fail fast. That can be a difficult concept for healthcare leaders to feel comfortable with, especially because in healthcare, failure often means poor patient outcomes.

“We’re not going to hit all our shots on goal, and that’s okay because we’ll do so in a well-established, containerized manner so it’s not running wild,” she said.

Artisight President Dr. Stephanie Lahr; Jessica Beegle, health technology expert and former Lifepoint CIO; HCA Healthcare CIO Marty Paslick; and Dr. Anthony Chang, Chief Intelligence and Innovation Officer at CHOC Children’s Hospital discuss healthcare organizations’ AI readiness at ViVE 2024 in Los Angeles.

Chang suggested gaining buy-in from the C-suite or hospital leadership by showing them that AI can take on the organization’s two toughest and solvable problems. Leadership will need to be supportive in terms of giving the team a chance to succeed. They will also need to have lower expectations accompanied by a longer-term vision, he said.

Beegle pointed out that if an organization’s data isn’t organized and accessible, then it isn’t worth having that data. She emphasized the importance of doing an assessment to understand the health system’s core infrastructure and preparing it so that the organization can be ready and agile as new language models come out.

“Building a strong foundation now will really enable you to succeed in that area,” she said.

When it comes to buying or building AI solutions, Chang recommended doing both.

“There’s a special synergy when you have an in-house team, however small, and off-the-shelf solutions,” he said, explaining that those off-the-shelf solutions often don’t have aggressive maintenance schedules, and an in-house team is needed to ensure the solution is maintained properly.

According to Chang, health systems can’t afford to not have other resources when considering AI adoption because many organizations don’t have a data scientist in-house. He said the first step is for the C-suite to decide how they want to move forward as they think about investing in the future.

Finding out where to invest resources is key. Chang suggested that creating one or two data- or AI-related positions could make a big difference. Later, once AI tools are implemented effectively to support administrative tasks, leadership can choose not to replace people who decide to relocate or retire.

“There are ways to trim costs without creating panic,” he said.

DISCOVER: An expert explains how to pick the right AI solution in healthcare.

Quality AI Relies on Quality Healthcare Data

As organizations move toward AI adoption, they need to ensure their data is clean and of high quality. During a session titled “Garbage In, Garbage Out,” healthcare leaders discussed the impact messy and incomplete data can have on AI initiatives.

Richard Clarke, senior vice president and chief analytics officer at Highmark Health, explained that the best way to improve the quality of data is to start using it for something that will benefit the organization, because then people will care more about it. One of the limiting factors in the adoption of generative AI has been the data to enable those use cases.

“If you can fully represent the problem you’re trying to solve in the data, then these tools can be amazing,” said Clarke. However, he added that it’s becoming apparent that without complete and high-quality data, these tools aren’t as effective.

From a federal government perspective, Elisabeth Myers, deputy director of policy at the Office of the National Coordinator for Health IT, explained that it’s important for healthcare organizations to understand how they can get data that’s useful from the beginning of their AI journey.

“We talk a lot about AI, and one of the things that goes along with that is understanding what AI should be good at and how to set it up to do that. There’s not a magic wand that’s going to make that data good. We need to think about the data that is necessary for AI to function and to get to the use case that you’re looking for,” said Myers. “You don’t want AI making inferences on inaccurate and incomplete data, especially when we start talking about health equity and disparities.”

Ultimately, Myers said, it’s not just about quality of data but also collected granular data that gives health systems more insight into their patient populations.

However, Dr. Ann Cappellari, chief medical information officer at SSM Health, pointed out that clinicians are often the ones left to capture that granular data.

“The person who cares least about granular data is expected to collect it with the most integrity,” she said, adding that their feelings are understandable because health systems haven’t prioritized quality data collection in the past.

Dr. Fatima Paruk, chief health officer at Salesforce, used to be a practicing physician, and she explained that clinicians are not trained to respect data. Rather, they are just trying to get through their workflows to see the next patient, which means the organization’s goals are misaligned with clinicians’ reality.

“We have a moral responsibility to understand that clinicians are burned out. They shouldn’t be using their time and energy toward data collection,” said Paruk. “I don’t see why the clinician also has to be the data entry personnel.”

She added that much of the data needed to understand the causes of health inequities, such as social determinants of health, are not captured in the electronic health record.

EXPLORE: Healthcare organizations can prepare for the future of AI.

At ONC, Myers said, the organization doesn’t care how the data is captured and remains user agnostic. She suggested that if standards are in place across the board, even the patient could be responsible for that granular data entry because they would know best.

ONC is working to improve health equity by creating a minimum set of data required for health systems and policy constructs. This includes requiring the collection of basic patient care and access data.

Clarke agreed that standards are needed and that innovations that use data in compliance with standards can be applied to multiple use cases rather than creating data sets that are only relevant to one use case.

One issue for the healthcare industry as a whole is that EHR companies allow customization, which might go against data quality measures.

“The EHR direction is customization. For a construct pretending to be a data gathering application, they allow customization again and again and again,” said Cappellari. “Codifying data sets to serve as a foundation is fine, no one will argue against that, but where are you going to fix it in your health system? That’s both financially and clinically driven. If you start in quality, everything else follows.”

Click the banner below to learn how a modern data platform supports decision making.

Another way to improve data quality is to collaborate with clinicians. Clarke said aligning with the care team on which data elements it wants to capture can improve quality immensely. However, it can be challenging if those data points aren’t explicitly tied to reimbursement frameworks.

Myers explained that better aligning research and clinical data can also better serve the organization in its data and AI goals.

As organizations begin to adopt AI tools, it’s crucial that they ask where the data for that algorithm came from. Paruk explained that having transparency into what went into determining that patient is at the high-risk end of the population creates more respect for the data.

AI transparency is also important to the ONC, says Myers. In addition to defining source attributes and making them transparent, she said, healthcare organizations need to have risk mitigation strategies in place to mitigate bias and other errors. Identifying which governance models work best and creating harmony across different governance models can help bridge that gap to ensure AI isn’t contributing to the “garbage in, garbage out” concept.

Keep this page bookmarked for our ongoing coverage of ViVE 2024. Follow us on X at @HealthTechMag and join the conversation at #ViVE2024.