Why Medicare’s Coding Rules May Widen Health Gaps

When a patient walks into a clinic, their diagnosis doesn’t just determine treatment—it determines how much money flows through the health system. Every diagnosis code added to a chart affects how Medicare pays hospitals, physician groups, and other healthcare providers. But what if the way we code diseases—rather than the diseases themselves—has been quietly reshaping the fairness of health plan payments across the country?

That’s the question raised by a new study from Harvard and Stanford researchers published in Health Services Research. The team introduces new ways to measure “coding intensity”—the pace and persistence with which health systems add diagnoses to patient records—and warns that our current system may be inflating spending for some plans and providers more than others.

Their work offers a simple yet powerful tool to help public health leaders, payers, and policymakers spot—and potentially fix—inequities built into the data that drives billions in Medicare payments.

Why Coding Matters for Health Plan Payment Equity

Medicare utilizes the CMS–HCC (Hierarchical Condition Category) model to determine payments for both Medicare Advantage (MA) plans and Accountable Care Organizations (ACOs). The goal is to adjust payments fairly based on the severity of patients’ illnesses, so that providers caring for people with more complex needs aren’t penalized.

But here’s the catch: the system assumes that diagnosis codes reflect medical reality. In practice, they reflect behavior—how intensely different organizations document patient conditions.

For example, one ACO might code every instance of heart arrhythmia, while another records it less consistently. Under Medicare’s risk adjustment formula, that difference translates into real dollars: more codes mean higher predicted costs and therefore higher payments. The result? A built-in incentive to “code harder,” even without any change in patients’ actual health.

A New Way to Track Coding Dynamics

The research team analyzed two years of claims data from nearly 1.8 million Medicare beneficiaries assigned to ACOs.

Instead of simply counting how many people had a given diagnosis (the usual approach), they broke coding behavior into two parts:

• Incidence — how often new diagnosis codes appear among patients who didn’t have them before.
• Persistence — how often diagnosis codes remain from one year to the next.

Together, these measures predict what the authors call “steady-state prevalence”—the level of coding that would be reached over time if current patterns continued unchanged.

This is a breakthrough insight: even if coding practices remain unchanged today, the proportion of people with a given diagnosis code will likely continue to grow for years. That built-in momentum explains why program spending can rise even without new illnesses or policy shifts.

Key Insight: The Coding Tide Keeps Rising

Take “specified heart arrhythmias,” one of the most common and costly categories. In 2018, 18.7 percent of Medicare beneficiaries in ACOs had this code. However, if the current coding pace continues, that number would eventually rise to 28 percent, even without any change in behavior (Health Services Research, 2025).

For conditions such as diabetes and vascular disease, the same pattern holds: coding prevalence naturally increases over time.

This creates a major challenge for health plan payment equity. It means that some ACOs or MA plans may appear to have sicker populations—and therefore receive higher payments—not because their patients are actually sicker, but because their coders are more diligent in coding.

Equity Implications: When Incentives Skew Fairness

If one health plan codes aggressively while another codes conservatively, the system tilts. Federal payments rise overall, but the playing field becomes uneven.

As lead author Oana Enache explains, “When coding practices differ across organizations, the same patient could generate very different payments depending on who documents their care.”

That misalignment doesn’t just affect budgets—it shapes access and incentives across the entire system. Overpayment to certain plans can draw resources away from traditional Medicare or underfund organizations serving historically undercoded populations, such as smaller rural clinics or providers caring for racially and economically diverse groups.

In short, coding behavior becomes a hidden equity issue.

What This Means in Practice

Public health professionals and policymakers can take action now to monitor and address inequities hidden in coding data. Here are practical takeaways:

• Use incidence and persistence metrics. Don’t just track how many people have a diagnosis—track how quickly codes appear and how long they persist. This helps distinguish between real disease patterns and documentation drift.
• Monitor “steady-state prevalence.” Comparing steady-state and current coding levels can flag conditions likely to keep growing even without behavioral changes—an early warning for budget or equity distortions.
• Audit for consistency. Encourage state Medicaid agencies, ACOs, and community-based programs to benchmark their coding persistence against population health data.
• Incorporate equity lenses. Ensure that coding-based payments don’t systematically reward organizations with more administrative resources or penalize those serving marginalized populations with less intensive documentation capacity.

When Coding Becomes a Mirror for Structural Racism

Medicare’s payment formulas are designed to be neutral—just math. But behind those numbers lie long-standing inequities in who gets diagnosed, documented, and ultimately reimbursed.

Communities of color often receive less comprehensive diagnostic workups, fewer follow-up visits, and less consistent documentation in electronic health records. These gaps aren’t about individual bias—they’re the cumulative result of structural racism in care delivery: fewer resources in safety-net hospitals, less access to specialists, and time-constrained primary care visits.

Therefore, when health plan payments heavily depend on diagnosis codes, the data itself reflects—and reinforces—those inequities. Health systems serving largely white, higher-income, or privately insured populations often appear to have “sicker” patients on paper, because they have more intensive coding practices, not necessarily higher disease burdens.

This creates a dangerous feedback loop:

• Undercoded communities appear healthier and receive less funding.
• Overcoded systems receive more resources, allowing them to expand services, technology, and administrative capacity.
• The cycle repeats, widening the resource gap.

Researchers, such as McGuire and colleagues, have shown that even small, one-time shifts in coding intensity can compound over time. Without guardrails, these trends can amplify racial and socioeconomic disparities—a subtle but powerful form of structural bias built into the nation’s largest public payer.

🧭 Policy takeaway: Achieving health plan payment equity means addressing the inequity inside the data itself. Tracking coding incidence and persistence through a racial equity lens is not optional—it’s essential to ensure that payment reform doesn’t hardwire injustice into the system.

The Policy Frontier: Guardrails and Reform

Medicare’s risk adjustment system has already faced scrutiny for upcoding—diagnosis inflation used to boost reimbursement. Policymakers have attempted to trim payment multipliers and introduce audits. However, the study reveals that even small, one-time increases in coding intensity can have lasting effects.

Future reforms might include:

• Integrating non-claims data, such as patient-reported health or electronic health record indicators, to balance coded information.
• Creating equity adjustments that account for disparities in documentation capacity.
• Transparent public reporting of incidence and persistence trends by plan type.

These approaches can help prevent the “steady-state creep” that quietly undermines both fiscal sustainability and fairness.

What’s Next—and What We Don’t Yet Know

The study presents a descriptive, rather than prescriptive, perspective. It doesn’t label any coding practice as “right” or “wrong.” Instead, it provides tools to understand how patterns evolve—and how policies might unintentionally magnify inequities over time.

Open questions remain:

• How can CMS use these measures to adjust payments without discouraging accurate documentation?
• Could steady-state analysis uncover where coding incentives diverge most sharply by region, race, or provider type?
• And how can community health systems build capacity to code accurately without inflating costs?

The Bottom Line

For decades, Medicare’s risk adjustment was seen as a neutral formula—just math. This study reminds us that coding is not neutral. It’s a behavior, shaped by incentives, capacity, and culture. And when those forces differ across the system, health plan payment equity suffers.

By rethinking how we measure and monitor coding intensity, public health leaders can better align payment with reality—and ensure that fairness, not paperwork, drives the flow of healthcare dollars.

💬 Reflect and Discuss

• How might your organization track coding incidence and persistence to ensure fair comparisons across plans?
• What barriers might prevent smaller or safety-net providers from coding at the same intensity as larger systems?
• Could steady-state analysis become a routine equity check in Medicare and Medicaid policy?