There’s a growing awareness of the impact that social determinants of health (SDOH) have on our health: from our access to healthy food, to our social support networks, to our exposure to environmental toxins, to our ability to take time off work to see a doctor. All told, it’s estimated that up to 80% of health outcomes are linked to SDOH.

But when patients come in for care, all clinicians see is a list of chronic conditions, lab results, and prior procedures, with limited visibility on the external factors influencing our health. If all clinicians have is clinical data, that’s all they’re going to react to. Health systems are almost completely isolated from the external conditions in the communities they serve, and are often powerless to address the ways these conditions impact the health of the people they care for. 

That’s not to say the industry hasn’t tried. To date, integrating SDOH into clinical workflows has come in the form of more questions the care team has to ask patients directly (which are often uncomfortable and more work); more data elements populated throughout the patient medical record with no clear indication as to why they matter; or in the form of risk scores.  

All of these attempts have fallen short, although risk scores are growing in appeal. However, the problem is that a risk score is just that: a score. It offers no context on why the patient is at risk, or what actions care teams can take to reduce their risk. Without this context, the data is just more noise for care teams to sift through, leaving them just as disempowered to take action on SDOH as they were before. 

AI Mines Data to Deliver the Context for Action

Data in itself is not useless for addressing SDOH. In fact, a study we published last year in the American Journal of Managed Care showed that a machine learning model trained on SDOH data alone could accurately predict inpatient and emergency department utilization — without even considering clinical data. But we have to acknowledge that prediction is only the first step, and only useful if followed by action. 

This is the key distinction between predictive analytics and prescriptive clinical artificial intelligence (AI). Predictive analytics stops at predicting risk. Prescriptive clinical AI goes further by identifying the specific SDOH factors that put patients or populations at risk, and recommending actions to address these risk factors. 

The difference comes down to AI’s superior data processing power. Federal agencies such as the Census, the USDA, and the EPA provide detailed data on a wide range of socioeconomic and environmental risk factors at a geographic level, which can then be connected to the individual based on their address. This public community-level data can then be supplemented with commercially-available data for a more granular view of risk. 

By learning to recognize patterns in the relationships between these risk factors and the outcomes of millions of patients, AI can determine the relative influence of specific social determinants on health risk, identifying the most impactful social determinants for each patient. From there, AI can recommend clinically-validated interventions that address these risk factors.

What Does Taking Action on SDOH Look Like in Practice?

Applying AI to social determinants may sound promising, but does it work in practice? 

At Jvion, we’ve been incorporating SDOH into our AI insights for almost a decade. In this time, we’ve created solutions to address SDOH from the level of both community level investments and individual interventions. 

Last year, we debuted our SDOH map, a visualization tool built on Microsoft Azure Maps. The map overlays communities with a view of socioeconomic and environmental health risk factors, including determinants such as income, air quality, and food access. Healthcare organizations can drill down from the zip code level to the Census block level to see which neighborhoods are at greatest risk, and identify opportunities for investment in community benefit programs that address barriers to care. 

To give an example, health systems could use the SDOH map to determine where they can partner with food banks and food delivery services to address food insecurity or food deserts. In an analysis we conducted in Oklahoma, we found that by expanding access to food banks in counties where food security had a high impact on social vulnerability, Medicaid could save nearly $10 million per year statewide by reducing the average cost of healthcare for members.

When COVID hit, we built on the SDOH map to create our award-winning COVID Community Vulnerability Map, which leveraged our data on SDOH to show which communities were most vulnerable to hospitalizations and deaths in an outbreak of COVID-19, as well as the social determinants driving their risk. The Map has since been updated to show the communities that should be prioritized for vaccinations.

At the individual level, SDOH data can be combined with data from the individual’s medical record for a more powerful view of their risk of adverse outcomes. The Jvion CORE™ considers both and provides care managers with evidence-based interventions that are prioritized and personalized through the power of AI to address both clinical and social vulnerability risk. 

These insights can supercharge outreach efforts by promoting more informed and productive conversations with patients about their risk. As a result of these more targeted and efficient conversations, care managers at one national health plan were able to reduce the number of calls they had to make by 40-50%. What’s more, providers that implemented the CORE to take on SDOH have reduced avoidable utilization by 20-30%, reducing care costs proportionally. 

But perhaps more than improving patient outcomes and lowering costs, clinical AI presents the opportunity to treat patients more holistically and empathetically. By taking action on social determinants, we can treat patients with renewed compassion and with an understanding of their external circumstances that is so often lost in healthcare. 

For more details on how clinical AI can enable action on SDOH, with real-world examples and case studies, download our whitepaper.