How artificial intelligence can allow providers to get a better handle on social determinants of health data

Image: geralt/Pixabay

Two new and seemingly unrelated approaches to delivering healthcare are starting to take shape in the industry: the use of artificial intelligence, and the integration of social determinants of health in crafting care plans. Both trends are developing independently, but they're likely due to intersect; factoring in SDOH is possible due to data, and if AI shines in any one particular area, it's making sense of complex data sets.

If the social determinants are comprised of the socioeconomic factors that can influence a person's health -- income, education, access to transportation, etc. -- then AI has the potential to allow providers to make the best possible use of that information.

That becomes increasingly important as value-based care emerges. With reimbursement increasingly tied to health outcomes, providers have a real incentive to ensure they're delivering the best care possible.

Terry Ward, senior vice president of solutions at healthcare technology outfit Apixio, sees SDOH broken down into three distinct stages: identification, integration and application.

Identification comes first, as care teams need to pinpoint the specific SDOH factors that are relevant to a given case. Ward estimates social determinants can account for 60 to 70 percent of total health outcomes, which makes it imperative to identify social support structures, community resources, and other variables that can have a profound effect on health outcomes.

Next comes integration. A big chunk of the industry is still trying to figure out how to best utilize social determinants, and many organizations are developing libraries of responses to those factors. There have also been grants issued by the Centers for Medicare and Medicaid Services and other agencies to help shepherd these factors into quality formulas. The goal is to integrate those into the health system so providers can understand the risk involved and the impact of all of these different SDOH factors.

Then comes application, and where AI is concerned, that's where the rubber truly meets the road.

"We understand who, we understand what, by looking at the data," said Ward. "So much of that data collection can be resource intensive. AI (is) looking at scale to look at unstructured data, and then groom out the whole findings or factors related to social determinants. The crux is, how do we apply them? We're altering the case management the outreach services that are engaged with that member."

ACTIONABLE DATA

"The industry is changing," said Ward. "Social determinants haven't yet been part of a meaningful dialogue. Providers have known this is an impact, but have felt unprepared to respond. They haven't had the tools, and it's always been one interaction at a time. This is where technology comes into play. That information has always been meaningful and personal, but it's not scalable."

It's scalability that's crucial, and made possible, by AI. By ingesting materials ranging from electronic health records to scanned images, AI can use machine learning to train the model to hone in on the key social determinants relevant to a given patient. The personal touch will always be needed, but AI's ability to tease patterns out of large data sets allows SDOH integration to become more widespread, more relevant, and more actionable.

Different services will be more appropriate for different people, whether those services include case managers, home health visits or ridesharing access. Integrating SDOH data makes the current structures more insightful and meaningful, allowing providers to deliver the right resources to the right people at the right time.

Over the past several years, said Ward, there have been libraries of social support services that have emerged, including 211, a regionally focused organization that pinpoints local community resources. Through these libraries it's possible to identify, for example, senior centers that offer shuttles to people trying to make their doctors' appointments. It's also now possible to determine how well socialized people are in their communities, whether that be involvement with churches, food banks or other volunteer organizations.

"Where AI comes into play is now you're looking at the direct available resources and the social determinants and looking at use over time," said Ward. "Because that's where AI is superior. You can train it based on data and outcomes, and in this case it's connecting, what is the right service, the right component of that library in the community that's connecting to the needs of that member?"

EMERGING APPLICATIONS

In Ward's estimation, the healthcare industry isn't quite there yet when it comes to integrating SDOH data. CMS has tried to start integrating AI into the larger ecosystem, but from an analytics standpoint, Ward thinks healthcare is about 35 percent of the way there. So there's a ways to go before it's fully a part of the CMS model.

The development of tools and applications is a little further along, and is closer to being able to use structured and unstructured data for the identification of SDOH.

"As the elements start to impact risk, there will have to be standards for how the variables get applied to reimbursement," said Ward. "That's where we're going to see CMS, (the American Hospital Association) and others flex their muscle in these areas. You're going to see a lot of creative players out there try to be the best-in-breed AI.

"It's the Wild, Wild West in terms of what those social determinants are, so it loses its analytical value," he said. "What does it mean to have an absence of family and social supports? We'll have to create some standards."

At some point, Ward expects there will be some kind of legislative involvement, likely from the federal government. There will only be partial adoption of this approach to technology until there are universal standards for the use of SDOH information. Once that bridge is crossed, legislation is likely to follow since governmental bodies will use that information for reimbursable activities.

Those standards are slowly being developed even as models of care have begun to evolve.

"You're seeing that evolution onto the individual," said Ward. "This is part of that. It'll be a natural course, where we go from the population to the individual."
 

Twitter: @JELagasse

Email the writer: jeff.lagasse@himssmedia.com