Medicare Study Shows Fair Payments for Diverse Populations Are Possible

Medicare’s payment system could be adjusted in new ways to ensure that health plans are paid more fairly when caring for those from different racial and ethnic backgrounds. In a new study published in JAMA Health Forum, it was found that changes to current risk adjustment formulas could reduce disparities without disrupting how the system functions overall.

Risk adjustment is a key part of how Medicare pays private health plans.

Some patients cost more to care for than others, and plans should not be punished for taking on sicker patients. Without accurate risk adjustment, plans can get fewer funds than they need to cover these higher costs, leaving them short financially to care for these patients.

To prevent that, Medicare uses algorithms that predict expected healthcare costs based on a person’s age, sex and diagnosed conditions, according to the study.

This process is especially important in Medicare Advantage (MA), which now covers more than half of Medicare beneficiaries and accounts for the majority of Medicare spending. Even small changes to how risk scores are calculated can have massive effects—shifting billions of dollars across the system and influencing who gets access to needed care.

Currently, Medicare uses a traditional method called least squares regression to generate risk scores. While functional, past research has shown that this approach doesn’t always capture the true costs of care for all groups.

Researchers have experimented with advanced statistical tools and machine learning models to improve accuracy. However, there are few studies that have directly tested whether risk adjustment can be altered to promote fairness across racial and ethnic groups while maintaining the financial health of the system.

While Medicare guarantees coverage after age 65, disparities in care and outcomes continue.

For example, authors explained that Black and Hispanic beneficiaries often have their risk scores adjusted in ways that lead to lower payments than what their actual care needs require.

These gaps don’t reflect lower need for care. Instead, they come from structural barriers such as reduced access to providers, underdiagnosis and differences in utilization.

Because Medicare Advantage payments are based on these spending patterns, the formulas can keep unfair gaps in place. That means risk adjustment isn’t just a math tool—it’s a way to either make health care fairer or keep it unequal.

Researchers of the new study analyzed claims data from more than 4.3 million Medicare beneficiaries between 2017 and 2020. They tested whether algorithmic tweaks could improve equity while maintaining accuracy.

Researchers focused on two approaches to make payments more fair. Constrained regression changed the formula so certain racial and ethnic groups would get payments that met set “fair spending” targets. Postprocessing kept the main formula the same but adjusted payments afterward, adding or subtracting money to meet fairness goals.

Both approaches were designed to be transparent and easy for policymakers to interpret.

Results revealed that each method provided more money to minority groups without affecting how well the model worked. Accuracy dropped very little—less than 0.1%—and average errors went up by no more than $21.

Under the current system, some groups, such as American Indian or Alaska Native beneficiaries, were underpaid by about $448 on average. The new methods added $300 to $600 for Black, Hispanic, Asian and American Indian or Alaska Native beneficiaries, closing much of the gap.
These changes still fit within CMS’s rules, showing that fairness and efficiency can work together. The constrained regression method could also help people in high-risk counties with poverty or housing challenges by increasing payments where health needs are higher.

Why This Matters for Plan Leaders

For health plan executives, the study highlights two key takeaways:

• Risk adjustment is an equity issue. If payments don’t match the needs of different groups, plans may not have enough money to care for vulnerable members, making gaps in access and outcomes worse.
• Fairer models can work. The study shows small changes to current formulas can make payments more balanced without breaking the system. Policymakers and plans can make payments fair and accurate at the same time.

As the Medicare population becomes more racially and ethnically diverse, pressure will grow on CMS to ensure that payment formulas evolve accordingly. Current models reflect on historical spending patterns, but future ones must be built with equity in mind.

The study highlighted that tools already exist to make progress. What remains is for CMS and health plan leaders to decide whether to act. In Medicare Advantage—where hundreds of billions of dollars and the health of millions of seniors are at stake—the opportunity to build a fairer system is too important to overlook.

The researchers recommend that policymakers should get input from many groups when setting fair spending goals and combine payment changes with other strategies. They also warned that algorithms by themselves cannot fix long-standing health inequalities.