Healthcare and AI: Thinking about the problems and avoiding mistakes

Many discussions in healthcare today almost inevitably mention artificial intelligence, and sometimes, it doesn’t take long for that to happen.

Calum Yacoubian of IQVIA says hospitals and health system need to think about using AI to solve specific problems.

Calum Yacoubian, MD, director of healthcare AI strategy at IQVIA, takes a nuanced look at AI and says organizations need to have a thoughtful, considered approach.

Still, he tells Chief Healthcare Executive®, “I think the hype, the possibility and the potential of large language models and generative AI are completely fair.”

“The potential is huge and that potential is huge for lots of reasons, but namely, because it has essentially democratized the ability to do advanced analytics and AI to just about anyone who has an idea,” Yacoubian says.

But he also doesn’t discount the potential harms for improper use of AI.

“The potential for harm and the risks, particularly within the medical domain, are super important,” he says. “And it has to be at the forefront of anyone's mind who's thinking about trying to translate some of this promise and potential into actionable, patient-impacting outcomes.”

In a recent interview, Yacoubian talks about ways to utilize AI, a successful use case with Endeavor Health, and some mistakes for organizations to avoid.

• Read more: For now, AI’s biggest potential to help healthcare isn’t the sexy stuff

Thinking beyond the tech

Healthcare organizations that want to utilize AI effectively have to think about more than the technology, Yacoubian stresses.

“The AI is a piece of the puzzle but it is by no means the entire picture,” Yacoubian says.

Particularly with clinical practice, he says, “You need to have the right array of stakeholders, and people to make sure that the technology sings and the right process that makes sure that makes sure that those people are able to use the technology appropriately.”

Yacoubian says organizations need to think equally about people, process and technology when it comes to AI and focus on solving a specific problem. Otherwise, he says, “It's very hard to be successful with AI.”

He says that’s a mistake some healthcare organizations make with their approach to AI. While many people have good ideas, without thinking about the impact or utilizing AI on a specific problem, he says, “You never really get beyond the proof of concept stage.”

• Read more: Hospitals and AI: Legal questions, liability and consent

A successful venture

Endeavor Health, which operates nine hospitals and many clinics in the Chicago area, worked with IQVIA and developed a successful project to utilize AI, and it’s helping more patients get the help they need, Yacoubian says.

In this project, Endeavor Health used a form of AI called natural language processing to scour vast amounts of patient data and find patients who are more vulnerable and need additional services. Endeavor used AI to examine patient records to identify patients who may be struggling with food insecurity, transportation issues, social isolation, and a host of other factors affecting their quality of life.

“It’s identifying social determinants of health and prioritizing patients based on those,” he says. “It's actually using AI to decrease social biases or to ensure that patients with social risk factors are managed appropriately.”

Endeavor’s program illustrates AI’s potential to improve population health, he says. It’s also cutting down on time physicians and staff are examining their records to detect at-risk patients, enabling the system to intervene more quickly.

“I think that potential for AI essentially to be impacting clinician workflow, but actually to be getting vulnerable patients more likely to be seen by relevant clinicians, because you're reducing the burden of work and surfacing otherwise hidden information, for us is definitely something that we're really proud of and something that we're really excited to have been a part of,” Yacoubian says.

He also praised Endeavor’s deliberate approach to solving a defined problem.

“That was a big key to success. Endeavor Health invested in technology, people and process,” Yacoubian says.

• Read more: Health equity is ‘the greatest challenge with AI’

Centering on patients

While Yacoubian understands, and shares, enthusiasm for using AI tools, he says he sees cases where organizations are working backwards.

He says some are looking to utilize AI technology and essentially saying, “I have a solution. I need to find a problem, rather than the other way around.”

Organizations looking to do more with AI need to consider their governance framework, and gather leaders and staff from different areas to consider uses and appropriate ethical considerations, he says.

Hospitals and health systems need to have safeguards in place to mitigate the risk of AI technologies leading to patient harm, including inaccurate suggestions on treatments. AI models have the potential to reflect racial bias in healthcare, he says.

Organizations also need to think more about protecting the data of patients as they utilize AI.

“There are also some significant direct consequences of patient data being used inappropriately and patient-centered information being leaked,” he says.

Large organizations aiming to deploy AI technology across their electronic medical records system need to be sure that data is de-identified.

Health systems also need to think about “a defensible approach to AI,” he says. Organizations should make sure they fully consider ethics, governance and privacy as they utilize AI, and be able to explain their approach, he adds.

As a physician, Yacoubian talks about the importance of focusing on patients.

“Everything that we do really should be kind of centering back on the impact to the patient or the communities that patients live in,” he says.