Empowering Techquity: The Role of Generative AI in Bridging the Health Equity Divide

Despite growing attention to the issue of health equity, there are still stark disparities in healthcare experiences and outcomes for historically underserved populations with socioeconomic barriers to care. 

Members of Black, Indigenous, and People of Color (BIPOC) communities are more likely to be uninsured, more likely to struggle with care access, and more likely to experience everything from chronic diseases to cancer compared to non-Hispanic white people.

Solving for these challenges to improve overall health equity will require a multifaceted, system-wide approach that combines policy, reimbursement, education, and technology.  

The advent of generative artificial intelligence (AI) – algorithms that can produce new narrative, audio, or visual content in response to prompts – represents a promising opportunity to take a major step forward in addressing equity gaps across the care continuum.

Generative AI can analyze vast amounts of healthcare data to identify patterns and predict outcomes, leading to more comprehensive insight into where disparities exist and how to close the gaps. However, it is crucial to deploy these tools carefully and deliberately, since hidden biases in the source data could propagate into the results, thereby exacerbating issues instead of resolving them.

Recently, Divurgent and CHIME convened a roundtable of healthcare CIOs from organizations serving a wide range of diverse communities across the United States to discuss the challenges and opportunities of leveraging generative AI to reduce health disparities. 

 

CHIME President and CEO Russell Branzell moderated the discussion group, joined by Divurgent’s COO Shane Danaher, Executive VP of Delivery Joe Grinstead, and VP of Delivery Dana Locke.

CIO participants included:

• Julia Zhou, CIO of Community Behavioral Health in Philadelphia
• CIO of a large tribally operated health system in Oklahoma
• Donna Roach, CIO of University of Utah Health
• Kevin Starnes, CIO of The Center Orthopedic & Neurosurgical Care in Oregon
• Linda Stevenson, CIO of Fisher-Titus Health in Ohio
• Beth Hunkeler, CIO of Dayton Physicians Network
• Chani Cordero, CIO of Brooke Army Medical Center in Texas
• Inderpal Kohli, VP & CIO of Englewood Health in New Jersey

Together, these experts shared their ideas about how to define health equity in the modern healthcare environment and how generative AI might be able to assist providers in achieving their equity goals.

 

Defining health equity against a background of disparities

The Centers for Disease Control and Prevention defines health equity as “the state in which everyone has a fair and just opportunity to attain their highest level of health.”

But that broad statement may not capture the nuances of what it takes to realize true health equity in the real-world health system. 

“We tend to think of health equity as equal access to care, but it has to go deeper than that,” said Inderpal Kohli, VP & CIO of Englewood Health. “Even when care is accessible and the opportunity is there, outcomes are not the same across different groups.  The access could be there, but the data we have is now revealing that there are inherent disparities in the way populations experience care and how those experiences translate into outcomes. For me, the definition of health equity must include a measurement of outcomes, not just accessibility.”

Donna Roach, CIO of University of Utah Health, agreed that the working definition of health equity must get more granular to avoid reinforcing the “two-tiered health system” that many communities currently face.

“Right now, society is more divided than ever into the haves and have nots. That is clearly reflected in our healthcare,” she explained. “If we want to overcome that, there has to be a baseline of services that we are providing to everyone, no matter who they are or what barriers they face: vaccines, dental health, preventive screenings, mental health. It should be part of every health system’s charter to ensure we’re doing that consistently and equitably across all our patient populations.”

“Prevention plays a critical role in advancing health equity. Promoting access to preventive services will be key to closing gaps in access, outcomes, and experiences for populations across the socioeconomic spectrum,” added Beth Hunkeler, CIO of Dayton Physicians Network based in Ohio. “And it has to happen in a big way.” 

“It’s not only up to health system leaders and clinicians to address these issues,” she pointed out.  “It’s up to all of us — the big societal ‘us’ — to provide education to students in school, to new parents, to seniors, about what wellness looks like and how to achieve it.  If we can do that up front, we’re going to save a lot of time, effort, and money on the back end.” 

“Equity starts with education within the community,” beth said. “Then we can start to bring in interventions to catch diseases earlier and apply technologies to maximize our resources when those diseases require medical care.”

These technologies will include algorithmic intelligence in its various forms, including generative AI, natural language processing, and other machine learning techniques. Stakeholders across the healthcare industry are already employing some forms of these technologies to address equity concerns, but applying the right tools at the right point in the patient journey has been challenge so far, said Linda Stevenson, CIO of Fisher-Titus Health in Ohio.

“We have the ability to conduct risk stratification better than we ever have before. With the help of analytics, including AI-driven tools, we have a more precise idea of where to focus that education and how to design interventions that produce tangible results for our communities,” she said. “We just need to do a better job of using these strategies in a predictive way rather than a reactive way after people start getting sick.”

 

Overcoming AI growing pains in pursuit of health equity

Artificial intelligence has the potential to accelerate progress toward a more equitable health system, but only if it is developed and deployed correctly.

“AI is anything but simple, especially because of the pace at which it is evolving,” said Joe Grinstead, Executive VP of Delivery at Divurgent. “Our guiding idea should be that AI will allow us to spend more of our dollars and our time on providing actual healthcare rather than completing administrative tasks.  If we can cut our administrative costs in half, that gives us 50 percent more budget to spend on equitable preventive care, which will in turn save us time and money on avoidable acute care down the line. That’s where the real difference can be made.”

Unfortunately, early attempts at using AI in healthcare and other industries have been plagued with unintentional biases due to limited datasets that are not diverse and representative enough of real-world populations.

“In more acute care settings, for example, the data is going to be skewed toward those higher-severity cases,” explained Julia Zhou, CIO of Community Behavioral Health in Philadelphia. 

“It might give you some insight into how intensive interventions affect severe cases, but it’s not going to do much to tell you about what happens before that: what progression looks like in people of different backgrounds; what the impact of preventive care could be in the general community. To design effective AI, we need that data – and we need it from traditionally underserved communities as well as those that have access to high-level care for behavioral health and psychiatric needs.”

Wariness around bias has largely resulted in a guarded approach among CIOs and others responsible for bringing these technologies into the care environment.

“I’m pretty cautious about ‘disruptive’ technology in healthcare,” said the CIO of a large health system in a sovereign Native American tribe in Oklahoma and surrounding areas.  “It needs to be managed and controlled so we can avoid bias and ensure we’re making safe and accurate decisions.” 

“Right now, AI is best suited to being a backend technology performing tasks like database cleanup, security management, scheduling optimization, and maybe some basic patient engagement tasks,” he continued. “I don’t believe generative AI is ready to help us make medical decisions just yet. There isn’t enough data, and the data sets we do have aren’t good enough to provide an informed opinion that clinicians or patients can trust.”

Kevin Starnes, CIO of The Center Orthopedic & Neurosurgical Care in Oregon, also believes healthcare organizations need to take a measured approach to their adoption of generative AI.

“Right now, models like ChatGPT are moving very quickly, because big tech companies want to capitalize on the interest and the potential,” he said. “In healthcare, we have to have more stringent guardrails because we can get into a lot of trouble if something goes wrong with the way we deploy these tools.”

CIOs need to balance their drive for innovation, which can be a positive force, with a recognition that new technologies need to be thoroughly vetted before they can produce results.

“CIOs are always going to have to deal with ‘bright shiny object syndrome’ every time a new AI tool comes out,” said Hunkeler. “Every leader is tempted to believe that the latest and greatest algorithm is going to solve all their problems, but that may not be true. Within a carefully constructed ecosystem, AI can be a game-changer. But we need to slow down and make certain we have a deep understanding of the problems we’re trying to solve before bringing in solutions.”

Evaluating AI products carefully will be essential for making the right choices. CIOs should ask vendors detailed questions about the data used to train AI models, as well as how the algorithm makes decisions or recommendations, before investing in a partnership.

 

Looking to the future of an equitable, AI-driven healthcare system

The healthcare industry is just entering the first phase of adopting generative AI, and exploration is ongoing. Despite concerns about what might happen if AI goes wrong, the CIO participants were optimistic that doing right could bring significant benefits to health systems and communities facing socioeconomic barriers to care.

“I definitely see opportunities in reducing physician burdens so they have more time to focus on talking to their patients about socioeconomic barriers and how to address them,” said Chani Cordero, CIO of Brooke Army Medical Center in Texas.

Other roundtable participants highlighted opportunities in mining radiology data, translating natural conversations in the clinic into structured documentation, and providing personalized education to patients based on their health status and identified needs.

The CIOs were confident that AI is here to stay, and it will only get more advanced as developers, care providers, payers, and patients work together to create a new ecosystem of more equitable, data-driven care.

“Our success with AI is going to start with conversations like these about how organizations can develop best practices and interact with one another – and with our communities – so we can move forward together,” concluded Stevenson. 

“When we talk with each other and better understand our shared needs and challenges, we’ll be able to make progress on health equity issues and create a better system for more accessible care, better outcomes, and better experiences no matter who our patients are or what their backgrounds might be.”