Health equity is 'the greatest challenge with AI'

Orlando, Florida - Artificial intelligence emerged as the dominant topic of the HIMSS Global Health Conference & Exhibition, with leaders discussing tools to detect disease earlier and improve outcomes.

Just as many in healthcare are excited about the potential of AI, healthcare leaders also fear that advances in AI may not be available to everyone. Many warn that if AI isn’t developed with health equity as a guiding principle, then disparities could worsen among disadvantaged communities.

Jennifer Stoll, chief external affairs officer for OCHIN, says she’s greatly concerned about the possibility of AI leading to greater inequities in healthcare. OCHIN, a nonprofit organization, works with Federally Qualified Health Centers and rural hospitals to improve their technology capabilities.

“This is going to be the greatest challenge with AI,” Stoll told Chief Healthcare Executive® in an interview at the HIMSS Conference. “If not done thoughtfully, it will create a whole unique set of haves and have nots.”

OCHIN is partnering with Microsoft and more than a dozen leading hospital systems to form the Trustworthy & Responsible AI Network, also dubbed TRAIN. OCHIN wants to ensure that providers with fewer resources in underserved communities are eventually getting access to those AI tools.

"AI could be a wonderful source for those that are serving rural and underserved communities, driving a tremendous amount of efficiencies, improve outcomes, bring in knowledge," Stoll said. "But it also can be a big destroyer of worlds, if it's not carefully and thoughtfully deployed."

(Healthcare leaders talked with us about AI and health equity in this video from the HIMSS Conference. The story continues below.)

AI helps, if ‘done correctly’

Robert Garrett, CEO of Hackensack Meridian Health, talked about AI’s potential to eventually improve the health of billions of people worldwide during a keynote address at the HIMSS Conference last week. Garrett said AI in healthcare could improve outcomes, as well as access and equity, but he also included an important qualifier.

“There's no question in my mind that AI can help to drive health equity if it's done correctly,” Garrett said.

“I think about health equity in terms of even identifying the social determinants of health, who's at risk for one or more of the social determinants of health, and how can we really address that by linking people who have those exposures to great care and to great resources,” he added. “AI can really bridge that gap and hopefully close some of the disparities that exist in healthcare outcomes today.”

Anna Schoenbaum, Penn Medicine’s vice president of applications, predictive health and digital health, stressed the importance of considering health equity in the development of AI tools.

“It is our responsibility to ensure that AI includes health equity,” she said in an interview at the HIMSS conference. “I think it is a huge piece of AI tools. We need to make sure that it is inclusive.”

At Penn, as researchers look at using AI in predictive models, they are also looking closely to validate the data, Scheonbaum notes.

Healthcare leaders stress that AI tools rely on the accuracy of data, so if they are using data that is inaccurate or reflects bias against racial groups, inequities can continue and even be magnified. The World Health Organization has urged caution in the use of AI tools in medical decisions.

Researchers found that chatbots offered answers reflecting racial bias, in a study, published in Digital Medicine last October. The Coalition to End Racism in Clinical Algorithms has found bias in screening in areas such as kidney function, delaying Black patients from getting better treatment. The coalition’s work has prompted hospitals and health systems to make changes.

‘A mirror of humanity’

Heather Lane, senior architect at athenahealth, said in an interview at the HIMSS conference that the healthcare industry is going to need to take “deliberate action” to ensure that health equity isn’t worsened by advances in AI.

“There have definitely been cases out there where people have shown that when carefully done, AI can be used to correct human biases, which is lovely,” Lane said. “But when done naively, it just absorbs all of our biases. I guess the way that I often put it is that, in many ways, current AI is a mirror. It's a mirror of humanity. And it reflects our good selves, but it also reflects our worst selves.”

Health systems can't naively build AI solutions that deepen inequities, Lane said. She said the easy - and incorrect - approach is simply to use existing data, which may include system inequities.

"If you're not thoughtful about it, if you don't approach it carefully, and with measurement and safety and equity in mind, then you build a system that just increases the gap," Lane said.

"But if you approach it in the right way, with the right metrics, and optimized carefully and thoughtfully, then you can build a system that can actually decrease those inequities and is more fair than the input data that you started with," she added. "And that's the goal that I think we should all be aiming for."

Brendan Watkins, chief analytics officer at Stanford Medicine Children’s Health, said there is the danger that biases could be embedded in algorithms. He said he is encouraged by the amount of discussions around AI and health equity.

“People I talk to at least do have this at the top of their mind,” Watkins said. “So I think that's very good.”

Schoenbaum noted that AI can leverage data on social determinants of health that have been collected by health systems, and share that information across other health systems to provide better care and improve patient outcomes.

“With that, I think we can address and promote health equity in our health solutions,” Schoenbaum said.

Hopes with TRAIN

With the emergence of TRAIN, Stoll said OCHIN can bring more data from disadvantaged populations into AI, including data on Black and Hispanic patients, and patients with chronic homelessness.

Stoll is hopeful about the potential of ensuring organizations with limited resources can gain from AI. She notes smaller hospitals don’t have the people, resources or governance to develop some of their own AI tools, but need to have that shared experience.

Some smaller hospitals and providers could be a long way from using AI tools to support clinical decisions, but could get welcome help from tools to help their business operations, Stoll said. Many under-resourced providers would get the most benefit from AI in reducing burdens on clinicians and staff, by using tools to easily file claims or summarize patient conversations.

“We see our community being best positioned to take advantage of the operational efficiencies, the administrative simplifications and the reduction of burden on our providers,” Stoll said.

“Where OCHIN is going to be a little slower is in the clinical, very slowly looking at it from the research side,” Stoll said. “But the operational efficiencies, we are moving very, very quickly in that space. But it can create haves and have nots, and that will only accelerate health inequity, especially based on where we are.”

While OCHIN aims to be a voice for health equity in the TRAIN initiative, Stoll also hopes to see other groups serving disadvantaged communities to participate.

“I think OCHIN really wants to make sure that we're bringing other voices in to help with this equity conversation,” Stoll said. “I mean, we have a large, large population we support, and we're growing every day. But we need more communities coming in.”

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

The federal government also can play an important role in providing funding to help smaller hospitals and providers develop AI capabilities. Smaller providers didn’t get nearly enough assistance in the move to electronic health records, Stoll noted.

“We need to continue to invest in health information technology and the infrastructure needed for our rural and underserved communities,” Stoll said. “That is going to be the biggest problem in leaving people behind. If we don't help everyone get to a level playing field, we're going to drive disparities.”