One of the few positives to emerge from the pandemic has been how it has helped illuminate the often-stark health inequities faced by people of color and low-income Americans.
The COVID-19 pandemic exposed long-standing health inequities
In the early months of the outbreak, the heightened infection risk and limited care options confronting many essential workers and people from racial and ethnic minority groups generated considerable attention. But the discussion soon expanded to include entrenched health inequities that long pre-dated COVID-19. These ranged from dramatically higher rates of chronic disease, to maternal and infant mortality, to shorter overall life expectancy.
Clear and measurable benefits to improving health equity
In response to this wake-up call, healthcare stakeholders have been developing strategies to address systemic health disparities. The benefits of doing so are clear: Beyond an overriding moral and societal obligation to improve health equity, reducing disparities can decrease the total cost of care through improved prevention, earlier intervention, fewer emergency room visits, and fewer hospitalizations.
Value-based models need more sophisticated health equity data
Yet as the healthcare payment system continues to evolve from paying for volume to paying for value, existing risk adjustment mechanisms lag. The methods used to inform value-based payment are largely incapable of recognizing the many social determinants that produce health inequities.
Currently, the Centers for Medicare & Medicaid Services (CMS), through their CMS Innovation Center, is working to identify and assess available data sources. Their long-term goal is to require “model participants to collect self-reported demographic and social-needs data from beneficiaries through development of Fast Healthcare Interoperability Resources (FHIR)-based questionnaires, application program interfaces, and mechanisms for bulk data submission."
According to physician researchers in the New England Journal of Medicine, “The lack of granular and reliable data on race, ethnic group, and health-related social needs has hindered health equity efforts."
Developing tools that can more accurately identify social determinants and their accompanying risks is critical for creating more nuanced reimbursement models based on acuity and risk. At the same time, supporting researchers and data scientists in their pursuit of a better understanding of social determinants is an essential first step in moving toward a more equitable healthcare system.
A new approach to population health assessments
With these objectives in mind, we developed the RTI Rarity™ artificial intelligence tool. As a next-generation health equity measurement and analysis tool, RTI Rarity uses supervised AI and advanced machine learning capabilities to generate detailed insights, identifying populations at the neighborhood level where persons are at high risk for unmet social needs. This information supports targeted interventions and more accurate risk adjustment required for value-based contracts that entail more risk.
Limited existing SDoH tools
In a perfect world, collecting detailed information about social determinants would occur at the provider level. But gathering individualized social risk data at the point of care remains the exception, not the rule. This is due to the fragmented nature of the healthcare system, a lack of adequate primary care access, and often-overwhelmed physician offices.
Social risk can nonetheless be assessed through analysis of social and community data. The challenge is that, while numerous algorithms can measure social determinants, most are limited to a small set of social risk factors. This is inadequate to meet the data science need. Major databases offer ADI, SDI, and SVI data:
The Area Deprivation Index (ADI) includes 17 measures in four categories:
- Housing quality
The Social Deprivation Index (SDI) is based on seven variables collected in the American Community Survey:
- Poverty rate
- Adults without a high school diploma
- Single-parent households
- Living in rented housing unit
- Living in overcrowded housing unit
- Households without a car
The Social Vulnerability Index (SVI) incorporates 15 Census variables, including poverty, lack of access to transportation, and crowded housing within four domains:
- Socioeconomic status
- Household composition
Nationwide Local Social Inequity (LSI) scores
Unlike the limited scope of these analytic tools, RTI Rarity tool assesses more than 200 social, behavioral, environmental, and economic factors to produce more granular assessments of social determinant-driven risks.
The foundation of the RTI Rarity tool is the Centers for Disease Control's (CDC) Social Determinants of Health framework, which was expanded by our multi-disciplinary team from five to 10 domains. The team identified and curated more than 150 additional metrics that are found to contribute to inequities in health and healthcare.
Then, data are extracted from an array of publicly-available federal, state, and non-profit resources, including the American Community Survey, the U.S. Department of Agriculture's Food Environment Atlas, the CDC's Wide-ranging Online Data of Epidemiologic Research (WONDER), the U.S. Department of Housing and Urban Development, the Child Opportunity Index, the Opportunity Atlas, and others.
Examples of health equity data points incorporated include percentage of population receiving food assistance, median owner-occupied property values, percentage enrolled in Medicaid, and percentage with a college degree.
Leveraging the random forest ML algorithm
We have extensively used artificial intelligence and a highly flexible machine learning (ML) algorithm known as random forests (RF) to develop RTI Rarity. RFs combine the output of multiple decision trees to achieve a single result. When coupled with other AI tools, RFs offer significant advantages over traditional, regression-based analytical approaches. Chief among these is the ability to significantly expand the number of variables assessed.
The RTI Rarity score supports a range of health equity and PHM use cases
The RTI Rarity tool generates LSI scores that are valuable in many health equity and population health use cases:
- Identifying neighborhoods at highest risk of poor outcomes for better intervention and resource targeting
- Measuring the impact of healthcare innovations, payment models, and interventions on social drivers of health in high-risk communities
- Accounting for factors outside of providers' control for more fair and equitable performance and quality measurement, evaluation, risk adjustment, and reimbursement
Along with identifying broad patterns and risks, RTI Rarity reveals specific population-level outcomes like drug overdose mortality, infant mortality, and cancer mortality.
Ohio pilot of RTI Rarity LSI uncovers 10-20% more variability in life expectancy
During a pilot in Ohio, RTI Rarity tool used more than 140 publicly available variables specific to the state. The platform's LSI score explained 73% of neighborhood variation in life expectancy. This was a substantial improvement over the alternative indices -- the Area Deprivation Index (ADI), Social Deprivation Index (SDI), and Social Vulnerability Index (SVI) – that collectively explained for only 50-63% of life expectancy variation.
National benchmarking results were even more dramatic: When assessing life expectancy at birth nationwide, the RTI Rarity LSI score explained 67% of the variance versus all the other indices -- 26% for SVI, 29% for SDI, and 43% for ADI.
Addressing health disparities requires robust and intelligent data and tools
Confronting centuries-old social determinants to redress health disparities is an enormously complex task. Healthcare is leading the change, but it requires robust data and technology to build a solid understanding of where, why, and to what extent social risks exist. Using RTI Rarity, a breakthrough tool, our consulting team provides clients with detailed insight to support a wide range of needs and applications.
Contact RTI Advance Health to discuss your data approach to health equity and how we can accelerate action from findings.