Mapping Social Inequity: How AI & Data Science Can Identify Populations With High Social Risk

The COVID-19 pandemic spotlighted the entrenched—and often invisible—inequities in the United States health system. As awareness of these healthcare gaps emerged, policy experts, the medical community, and the American public sounded the alarm, calling for structural and societal change.

That increased focus on inequities accelerated the urgency to understand and address the social determinants of health (SDoH), or the nonmedical factors that influence someone's health outcomes. More and more research is exploring how the places where people live, work, and study can impact their health. These forces and systems, which include economic policies, social norms and policies, racism and climate change, can significantly shape the conditions of daily life, notes the Centers for Disease Control and Prevention.

Pandemic heightened the need for SDoH work

The pandemic underscored the importance of health policy interventions that address SDoH and their far-reaching impact on everything from health programs to tailored interventions and policy changes.

In response to the inequities exposed through COVID-19's disproportionate toll, the federal government set in place a series of measures, such as allocating American Rescue Plan Act funding to bolstering community health workers, school nurses, and organizations in historically marginalized communities.

State and local governments also used pandemic funding to address social needs, such as a North Carolina program that provided transportation to and from medical services and shelters, offered home delivery for medications, and supplied financial relief for housing, childcare, and living expenses, described in this Health Affairs article.

Understanding complex social risk factors matters for health equity

The urgency of the recent public health crisis offers compelling evidence of SDoH's instrumental role in shaping outcomes, such as the disproportionate burden of the disease in Black and other historically marginalized communities. But the role of SDoH extends far beyond a single health crisis.

Understanding a population's complex characteristics is essential for program planning, epidemiologic studies, and other public health practices, according to the CDC. Gathering data to better understand social risk factors can help inform health efforts in myriad ways, including:

• It can identify populations and geographies at risk for acute and chronic illness.
• It can characterize a community's preparedness and potential impact for a public health emergency.
• It can help us understand the factors that influence environmental exposures and human health.
• It can assess inequities over time, helping to measure the impact of policies intended to address them.
• It can help policymakers allocate public health resources.

Accounting for myriad SDoH poses challenges

Fully understanding the role of SDoH requires engaging with a number of social and economic forces, from education and income to race, ethnicity, and housing, explain the authors of a paper published in the Journal of Urban Health. While increasing amounts of data offer tremendous potential in informing healthcare decisions, challenges persist in both availability and uniformity across these drivers.

Indeed, with so many complex and nuanced factors, how can the healthcare system even begin to understand the ways in which patients' SDoH impact health outcomes?

Going beyond limited data sources for SDoH

Given the numerous and interconnected factors that influence SDoH, it's important to go beyond single data sources to capture the complex factors that influence someone's health experiences and outcomes. Even commonly referenced indices, such as the Social Vulnerability Index, use a limited set of variables.

Gathering an individual's specific information at a primary care provider level makes intuitive sense, but this approach has limitations. A fragmented healthcare system, primary care access challenges, medical mistrust, and clinician workload can hinder widespread data collection at the personal level.

Harnessing the potential of AI and data science for health equity

Artificial Intelligence (AI), machine learning (ML) and data science tools offer tremendous potential to improve and innovate in these arenas.

“Given the predictive capacity of ML, it offers new opportunities for health, behavioral, and social scientists," wrote the authors of a paper in SSM Population Health that reviewed the prospects of machine learning in SDoH research.

At RTI Health Advance, we are already harnessing the power of these exciting advances through an innovative tool called RTI Rarity^TM. This health equity measurement can capture more than 200 SDoH variables sourced from more than 40 databases. The technology generates a composite local social inequity score (LSI) using supervised AI and random forest modeling, or a supervised ML method that excels at assessing multiple variables.

How RTI Rarity measures, predicts, and adjusts SDoH

Our team used the CDC's Social Determinants of Health framework as the tool's foundation, and then identified and curated 200 additional metrics that contribute to health inequities.

The data is sourced from broad-ranging publicly-available data, such as the American Community Survey, the US Department of Agriculture's Food Environment Atlas, the CDC's Wide-ranging Online Data of Epidemiologic Research (WONDER), the US Department of Housing and Urban Development, the Child Opportunity Index, the Opportunity Atlas, and other resources.

As a result, RTI Rarity is able to create high-resolution SDoH composite scores constructed with a health equity focus.

Outperforming other health measures

Incorporating so many factors into a single composite score leads to a more accurate measure of someone's experiences—and ultimately, health outcomes. In fact, the LSI outperforms other stand-alone measures such as the Social Vulnerability Index (SVI) or the Area Deprivation Index.

The tool can look at a neighborhood, an area defined as a Census tract, with the greatest social risk, and the top factors driving health outcomes in that area. The ability to zoom into that granular data is bolstered by research that's increasingly revealing the value of using extremely local-level data to map health inequities.

Putting the scores to use for health equity

Once a score is generated, health plans and providers can better understand and identify the social risk of the populations they serve. They can also take into account the social factors that drive those risks, which makes them better equipped to identify suitable interventions.

Healthcare organizations can leverage the tool to identify populations with high social risk. This way, they can prioritize interventions and adjust resources to improve population health.

For example, the tool can help public health advocates identify neighborhoods that have a high level of health risk and tailor interventions to address those specific challenges. On the health system side, the tool can be used to measure the impact of healthcare innovations or value-based payment models. It can also help with risk adjustments, accounting for factors outside of provider control that may impact outcomes.

Looking forward to innovation in healthcare

LSI scores are helpful in mitigating bias in clinical decision-making models because they provide a more comprehensive AI-based algorithm that accounts for both social and clinical needs among neighborhoods. The decisions are promoted by data, not human bias. LSI scores can also be used in clinical trial studies to ensure sufficient representation and inclusion of people with social needs.

Investing in a future with more health equity

Creating a more equitable future will acknowledge and account for the injustices historically marginalized populations have experienced. To begin that journey, we need to better understand the economic, social, and geographical obstacles that continue to widen these gaps.

At RTI Health Advance, we can harness the power of data and RTI Rarity to help you understand—and address—the social risk factors impacting your patient population. Collecting and stratifying data on SDoH is instrumental in the progression toward a health equity-oriented future.