Over 64 million Americans live in rural areas. Rural Americans are at higher risk of dying early than their urban counterparts from the five leading causes of death – many of which are preventable. Addressing urban-rural health disparities has become a national focal point, especially as states apply for funding to improve rural health through the $50 billion Rural Health Transformation Program. 

However, Medicare’s main tool for adjusting payments based on patient risk underpredicts mortality and overpredicts spending for rural beneficiaries, a new study by LDI Senior Fellows Kristin Linn, Amol Navathe, and colleagues found. But the tool works well for urban beneficiaries.

This weakness in the CMS Hierarchical Condition Category (HCC) model has serious policy implications. Specifically, the HCC model poorly captures individual health risk for rural beneficiaries, placing them at a disadvantage in payment formulas which Linn said “translates directly into reduced payments to rural providers and Medicare Advantage plans serving these communities.”

CMS-HCC Model Uses Spending to Predict Risk

Risk adjustment is a statistical method that facilitates fair performance comparisons between hospitals, health plans, or doctors by accounting for how sick their patients are. It estimates the cost to treat a patient and that estimate is used to allocate funds and modify payments according to individual health needs. Used in Medicare Advantage, Medicare fee-for-service, and other programs, the CMS-HCC Risk Adjustment Model considers demographic factors and diagnosis code groupings to predict one-year health spending.

But “spending may be a problematic outcome for a risk prediction model,” said Linn, “because, compared to more privileged populations with similar levels of clinical risk, some marginalized populations are more likely to experience barriers to accessing and utilizing care resources.” 

The same access issues also affect how likely certain groups are to have comorbidities diagnosed and coded. Those diagnostic codes, however, form the basis of the CMS-HCC model.

Given that care access and use are related to patient income and rurality, Linn, Navathe, and the team sought to investigate the relationship between the CMS-HCC model, observed spending, and observed mortality in rural and urban populations.

Study Finds That Spending is a Poor Proxy for Clinical Risk

The researchers used a nationally representative random sample of 4,170,277 traditional Medicare beneficiaries enrolled from 2018 to 2019, including both dually and non-dually eligible individuals for Medicare and Medicaid. They compared HCC-predicted spending risk with actual mortality rates—a more objective measure—and found that the HCC model systematically underestimated clinical risk for rural Medicare beneficiaries. Problems with using spending to predict health risk have been previously documented in other disadvantaged populations, as racial biases have been identified in clinical and population health algorithms.

Failure to Account for Social Determinants of Health Creates More Problems

While the CMS-HCC model works well for urban beneficiaries, the researchers note that the model’s inaccuracies for rural beneficiaries are likely rooted in a failure to fully account for social factors around access.

“Barriers such as reduced transportation access, limited diagnostic testing, and lower quality of clinical documentation disproportionately affect rural beneficiaries,” Linn said. “These factors result in under-coding of comorbidities and, consequently, under-adjustment of risk scores.”

On top of existing structural inequities, inaccurate spending and risk predictions result in lower payments to rural providers and plans, and can ultimately reduce access to care even further for rural beneficiaries.

Recommendations to Reduce Urban-Rural Bias

Linn, Navathe, and colleagues suggest that the CMS-HCC model could be improved by including high quality measures of social determinants of health (SDOH) and geographic factors. Specifically, the researchers recommend the inclusion of SDOH metrics that focus on individual access to health care. 

Enhanced data collection to capture underdiagnoses among rural populations can further improve the accuracy of predictive models and ensure that resources are being fairly allocated for rural beneficiaries.

“By adopting data-driven policy solutions,” said Linn, “CMS and other stakeholders can mitigate disparities and improve health equity for one of Medicare’s most vulnerable populations.”

---

The article “Unfairness Toward Rural Beneficiaries in Medicare’s Hierarchical Conditions Categories Score” appeared in Health Affairs Scholar. Authors include Ravi B Parikh, Kristin Linn, Junning Liang, Sae-Hwan Park, Torrey Shirk, Deborah S. Cousins, Caleb Hearn, Matthew Maciejewski, and Amol Navathe.

---

Author

---