Accurately Predicting Patient-Level Risk With LSI & Individual Risk Assessments

Healthcare providers and payers rely on risk stratification to maximize the impact of population health, value-based arrangements, and health equity initiatives. Ultimately, healthcare leaders want to predict individual-level adverse events. However, current approaches to acquiring thorough, accurate, and reliable risk stratification face serious hurdles.

Standalone social needs assessment tools that are not standardized or widely used may only capture needs for certain populations, thereby creating a bias that could skew health risk scores and misclassify individuals. Most widely used area level social risk indices focus on limited factors, like social risk or economic status, versus the broader spectrum across clinical and socioeconomic risks.

A better approach to risk stratification is to utilize a comprehensive Census-tract-level risk score that individual risk assessments can complement as a data point.

Individual health risk assessments increase data input but can introduce bias

Individual social and health risk assessments are becoming very popular and with good reason. The goal is to assess every person at every health encounter and use that information to identify social needs and possible risks that can be addressed and remediated. However, the dedicated use of standardized individual risk assessments or Z-codes isn't occurring consistently or with the required accuracy.

If healthcare teams use data from health assessments collected inconsistently and from select patients, they inevitably introduce bias. The bias is compounded when the data are stratified and then when it becomes the basis for focusing on future patients or creating a reusable algorithm.

Singular health risk scores provide more data sources but are standalone indices

Several social risk indices are used regularly, but they give too narrow a view for informing or predicting an individual's risk level. For example, the Area Deprivation Index (ADI) focuses solely on poverty, and the Social Vulnerability Index (SVI) focuses on 4 socioeconomic themes.

These singular or narrow views limit the ability to comprehensively assess risk across clinical, financial, and social factors. That's where the Local Social Inequity (LSI) score generated by the RTI Rarity^® tool provides a more accurate, comprehensive, and reliable foundation for risk stratification.

When complemented with an organization's clinical data, the 2 tools create a robust foundation for predicting individual-level risk holistically.

Why a census- & area-level index based on many validated sources work better

The RTI Rarity tool creates and delivers the Local Social Inequity (LSI) score by drawing on 40+ public and private datasets from federal, state, non-profit, and academic resources. It also utilizes 200+ area-level variables across 10 domains at the census, ZIP code, and county levels.

In benchmarking and validation analyses, the RTI Rarity LSI score proved to be substantially better at explaining variance in life expectancy than other commonly used SDoH composite measures. Not only can LSI scores identify which communities are at high risk of poor health outcomes, but it also identifies the most critical social, behavioral, and contextual predictors for those outcomes within neighborhoods across the US.

The RTI Rarity LSI score provides a better foundation for predicting individual-level risk

The following 5 benefits of using the Local Social Inequity score are the main reasons why a singular risk index or a social risk assessment falls short of accurate risk stratification.

Provides granular, census-level insights

The LSI score provides a composite social risk score generated at the census tract level, which is very granular. That's critical to gleaning insights at the local level down to predicting at the individual level.

If you only used a county-based risk score, it would be inadequate to address individual-level needs. There's too much variation in health, health outcomes, and social needs at the county level.

LSI includes clinical and socioeconomic factors

Again, most indices focus on 1 or a few specific domains of social risks and needs, which can lead to an incomplete outlook.

The RTI Rarity tool generates an LSI score based on a plethora of economic, social, and health data. With the LSI score, teams can stratify risk by relying on a larger dataset that adjusts for physical and health factors as well. The LSI score is based on a more comprehensive set of validated datasets, encapsulating the prevalence of chronic diseases and the CDC's Healthy Days measures, among others.

LSI can help avoid bias

A predictive algorithm at such a very granular level captures essential differences that vary from neighborhood to neighborhood. This more inclusive approach provides a more complete and richer view of every individual within the context of their neighborhood.

Offers more accurate predictions

Whether the LSI score is used internally for strategic planning or externally for program development and implementation, an index is only as helpful and useful if it's accurate and unbiased. Through benchmarking analysis, the Local Social Inequity score proved more precise than other publicly available indices, particularly when explaining variance in life expectancy.

LSI can help prevent patient misclassification

Healthcare leaders must be careful when choosing a risk stratification model because they can be heavily biased against specific populations. They can even misclassify patients and underrepresent their risk level. For example, if an algorithm that predicts which individual will need a kidney transplant is biased against Black patients, the individuals at most significant risk may be left out or misclassified as low risk.

Use cases for predictive individual risk scores

Deploying a tool like RTI Rarity and its Local Social Inequity score, with or without additional individual patient-level data points, opens up a world of valuable use cases.

Create predictive risk scores to improve care and quality measures

The RTI Health Advance team uses the RTI Rarity tool as a foundation to build reusable risk predictive models that can be tuned to focus on at-risk patients more proactively to achieve prioritized outcomes like reducing avoidable ED utilization and 30-day acute care readmissions.

We can design a calculator that provides an immediate probability for a particular outcome based on individual risk. Because the datasets are comprehensive, they can fill in missing data points for an individual based on their Census tract.

Stratify existing patients

Many healthcare organizations want to build a predictive tool that can re-stratify patients into socially and clinically adjusted buckets of risk. They don't want to rely solely on clinical data or claims data but use their data to complement comprehensive area-level risk indices.

Supplementing the LSI score with patient-level information from EMR, claims, or self-reported data provides a complete view of an individual's risk because it's socially and clinically adjusted.

When an organization is equipped with an accurate predictive model for the patients they serve, it can create a replicable model that can be applied to everyone at risk.

Discover and engage new potential patients

What about the individuals who could become a patient or part of a value-based or population health management cohort? These individuals may have avoided care or had poor experiences in the past and aren't currently engaged with your system.

Lack of visibility into these individuals weakens risk stratification models and hides potential opportunities to provide better care to more patients. With area-level risk data, leaders can capture community health data at a very granular level, create unique strategies to engage individuals in communities, and build relationships with people who need care.

Take your risk stratification further with more predictive power

The RTI Rarity tool is an expandable and adaptable model-maker. Built on a powerful foundation, it can leverage or inform existing data points to provide a predictive, comprehensive individual risk score.

Today's healthcare organizations are moving towards achieving more accurate, reliable, and standardized individual assessment, but that methodology isn't standard practice and introduces bias. When coupled with the Local Social Insecurity score, leaders can access more sophisticated predictive analytics and risk-stratify individuals more accurately.

RTI created the Rarity tool for high-acuity risk prediction

RTI Health Advance supports organizations at the forefront of healthcare. Our team of experts covers the spectrum of issues faced by payers, providers, and vendors today. From health equity, data analytics, quality improvement, population health, and digital health technologies, we can help you achieve care, quality, and cost goals. Contact us to explore how our team can guide your health organization.