The $84 Billion Question: Understanding Medicare Advantage Risk Adjustment and What It Means for Providers
Medicare Advantage has become the dominant healthcare delivery model for Medicare-eligible beneficiaries, now covering over 54% of all Medicare enrollees. With that growth has come unprecedented scrutiny of how these plans are paid—and specifically, how diagnosis coding affects those payments.
The numbers are staggering. The Medicare Payment Advisory Commission's 2025 Report to Congress estimates that Medicare will spend 20 percent ($84 billion) more for MA plan beneficiaries than if those enrollees were receiving care under Medicare Part A and Part B (traditional Medicare). That's not a rounding error. That's approximately the annual budget of a mid-sized federal agency.
The reason? Two factors that MedPAC identifies as "favorable selection" and "coding intensity." But for providers working with Medicare Advantage patients, the more pressing question isn't about policy debates in Washington—it's about understanding how risk adjustment works, what's expected in documentation, and why your diagnosis coding practices are under more scrutiny than ever before.
How Medicare Advantage Payment Actually Works
Unlike traditional Medicare, which pays healthcare providers directly based on services rendered, Medicare Advantage operates through Medicare Part C. Under this model, monthly payments are made to private insurance companies (Medicare Advantage Organizations, or MAOs) that provide coverage for Medicare beneficiaries who enroll in their plans.
These payments aren't flat rates. They're adjusted based on each beneficiary's demographic and health risk characteristics, including age, disability status, dual eligibility for Medicaid, institutional status, and—critically—prior health conditions coded using Centers for Medicare & Medicaid Services Hierarchical Condition Categories (CMS-HCCs).
The logic is sound: sicker patients cost more to treat, so plans covering sicker populations should receive higher payments to ensure they don't avoid enrolling high-risk beneficiaries. Risk adjustment is essential for maintaining access to care for Medicare beneficiaries with complex health needs.
The problem emerges in implementation. Most diagnoses used in risk adjustment calculations are reported on physician and outpatient claims. Prior to the emphasis on risk-adjustment coding, there was minimal scrutiny of diagnoses on professional or hospital outpatient claims, since reimbursement was based on procedure codes (CPT codes). Diagnoses were used primarily to demonstrate medical necessity and support evaluation and management coding.
Risk adjustment changed that equation fundamentally. Now, every diagnosis code submitted on a claim for a Medicare Advantage patient potentially affects how much the plan gets paid. And that creates powerful financial incentives.
The Coding Intensity Problem
Here's where the numbers get uncomfortable. MedPAC estimates that in 2022, MA risk scores were about 18% higher than scores for similar traditional Medicare beneficiaries due to higher coding intensity. MedPAC projects that in 2024, MA risk scores will be about 20% higher than scores for similar traditional Medicare.
"Coding intensity" refers to the tendency for more diagnosis codes to be recorded for MA enrollees compared to traditional Medicare beneficiaries with similar health status. As a result of insurers' "upcoding" patient diagnoses, the Centers for Medicare and Medicaid Services (CMS) overpaid Medicare Advantage by $50 billion (or 13 percent) in 2024, according to MedPAC.
To address this, Congress mandated that CMS apply a coding intensity adjustment to MA payments. CMS has counteracted this financial incentive by implementing a uniform "coding intensity adjustment," which deflates MA risk scores by 5.9% (as of 2023). But that minimum 5.9% adjustment hasn't kept pace with actual coding intensity growth. In 2020, risk scores for MA enrollees were already 13% higher than they should have been, generating $16 billion in overpayments. By 2022, scores and additional payments had jumped to 18% and $37 billion, respectively. In 2024, MA coding intensity is expected to be 20% higher than traditional Medicare, resulting in an extra $54 billion.
The gap keeps widening. March 2025 estimates project that coding intensity will account for $40 billion in MA overpayments, up from $23 billion in 2023.
Where the Coding Problems Are Concentrated
Not all diagnosis codes are equally problematic. Office of Inspector General (HHS-OIG) audits have revealed specific patterns of unsupported diagnoses. OIG audits found that up to 70% of submitted diagnosis codes lacked supporting medical documentation.
But some codes have error rates that are even more alarming. According to OIG investigations through 2023, certain diagnosis codes were unsupported over 90% of the time:
Acute stroke: 96% error rate
Acute heart attack: 95% error rate
Breast cancer: 96% error rate
Colon cancer: 94% error rate
These aren't obscure codes. These are diagnoses for which documentation must carefully differentiate between current, acute conditions and historic conditions. The high error rates likely reflect a fundamental misunderstanding (or deliberate misapplication) of ICD-10-CM coding guidelines.
OIG stated in its report that the medical records for more than 61% of the 250 sampled enrollee-years did not support the diagnosis codes, which fall under seven high-risk groups. The pattern is consistent across multiple audits and multiple MAOs.
Health Risk Assessments: The Controversial Data Source
One of the most contentious issues in MA risk adjustment involves Health Risk Assessments (HRAs)—often conducted through in-home visits by companies contracted by MA plans.
Nearly two-thirds of MA coding intensity is attributable to diagnoses identified through HRAs, and 37% of HRA diagnoses are not documented on any other encounter data. Think about that: more than a third of diagnoses identified through HRAs never appear in any actual clinical encounter.
Using 2016 MA encounter data, prior OIG work identified two sources of enrollee diagnoses—health risk assessments and chart reviews—as vulnerable to misuse by MA companies. An OIG evaluation updated that work and determined that vulnerabilities persist regarding the appropriateness of resulting risk-adjusted payments.
The problem isn't with legitimate HRAs conducted as part of comprehensive patient care. The problem is with HRAs used primarily as diagnosis-finding missions designed to maximize risk scores without corresponding changes in treatment plans or clinical management.
What the ICD-10-CM Guidelines Actually Say
The confusion (or exploitation) around risk adjustment coding often stems from misunderstanding or deliberately misapplying the ICD-10-CM Official Guidelines for Coding and Reporting.
Section IV, Diagnostic Coding and Reporting Guidelines for Outpatient Services, provides clear direction:
Chronic diseases treated on an ongoing basis may be coded and reported as many times as the patient receives treatment and care for the condition(s)
Code all documented conditions that coexist at the time of the encounter/visit and that require or affect patient care, treatment, or management
Do not code conditions that were previously treated and no longer exist.
History codes may be used as secondary codes if the historical condition or family history has an impact on current care or influences treatment.
That third bullet is where most of the problems emerge. A patient with a history of breast cancer treated five years ago and currently in remission should be coded with a personal history code (Z85.3), not an active cancer code. But history codes don't generate HCC payments. Active cancer codes do.
Similarly, a patient who had an acute stroke two years ago and is now receiving ongoing stroke rehabilitation should be coded for late effects of cerebrovascular disease and whatever deficits remain—not for acute stroke. But acute stroke generates a much higher HCC payment than late effects.
The guidelines are clear. The financial incentives push in the opposite direction.
The Audit Tsunami
For years, CMS's Risk Adjustment Data Validation (RADV) audits moved slowly, reviewing a limited number of contracts and medical records annually. That's changing dramatically.
CMS will scrutinize all Medicare Advantage contracts each year and increase its auditing volume. CMS will dramatically increase the number of employees conducting audits from 40 to 2,000 and will utilize "enhanced technology" for the new enforcement regimen, partially to clear a backlog of audits for payment years 2018 through 2024.
The scale of this expansion is unprecedented. The agency will increase audit volume from 60 plans per year to all plans, and will increase record reviews from 35 per plan, per year to as many as 200, depending on the size of the plan.
And the financial stakes have increased exponentially. A 2023 final rule allows CMS to extrapolate RADV findings across an entire MA contract, rather than just recovering overpayments for the specific sampled records. That means a finding of improper coding in a sample of 200 charts can result in tens or hundreds of millions of dollars in extrapolated overpayments.
The Enforcement Landscape
Beyond RADV audits, MA risk adjustment has become a major focus of federal enforcement activity. Of 44 managed care audits conducted by HHS OIG since 2017, 42 have focused on issues with accurate diagnosis coding. Additionally, in March 2025, the No Unreasonable Payments, Coding, or Diagnoses for the Elderly (No UPCODE) Act, which was originally introduced in 2023, was reintroduced in Congress.
The No UPCODE Act would require CMS to consider two years (rather than one year) of patient diagnostic data for risk adjustment and would prohibit coding submissions from chart reviews and health risk assessments. While the legislation hasn't passed, its reintroduction signals continued congressional scrutiny of MA payment integrity.
Department of Justice enforcement actions based on alleged false diagnosis codes have resulted in substantial settlements. There have been significant DOJ enforcement actions resulting in settlements based on alleged false diagnosis codes, some as recently as March and April 2025.
The MAO Response: Actuarial Equivalence
Medicare Advantage Organizations have pushed back hard against OIG audit findings, and their objections aren't frivolous. Most MAOs refuse to accept OIG audit conclusions, arguing that the methodology is fundamentally flawed.
Humana stated that the OIG's findings are "systematically skewed towards identifying overpayments rather than underpayments, rendering OIG's results inherently unreliable." Humana acknowledges this endless loop: "OIG and the MA industry therefore appear to be at an impasse on this critical issue".
The core of the MAO argument is "actuarial equivalence"—the principle that MA payments should be comparable to traditional Medicare costs when accounting for variations in coding practices. MAOs argue that audits focusing solely on unsupported diagnosis codes ignore situations where conditions may be under-reported, which would balance any over-reporting.
United argued that the government had no evidence that the physician-submitted codes were unsupported other than the fact that they weren't identified by the chart reviews despite other explanations (e.g., missing medical record pages, chart review coder mistakes). United stated that "after more than a decade of litigation in a case in which [the government] seeks to punish United for billions of dollars, [the government] has chosen instead not to review a single chart—not even a sample" to prove that any diagnosis codes were unsupported.
There's validity to some of these concerns. RADV audits that only look for overpayments without adjusting for potential underpayments may indeed produce skewed results. CMS-HCCs are based on a capitated model covering care delivered over a year, not a single encounter. Fee-for-service claims capture moments in time.
But the scale of the coding discrepancies revealed in audit after audit makes it difficult to dismiss the findings entirely as methodological artifacts. When 70% of sampled diagnosis codes lack supporting documentation, the problem extends beyond audit methodology.
What This Means for Providers
If you're a physician or other healthcare provider seeing Medicare Advantage patients, you're caught in the middle of this fight. You're not submitting risk adjustment data directly to CMS—your diagnosis codes go to the MA plan, which then submits encounter data for risk adjustment. But your documentation is what gets audited.
Here's what you need to understand:
1. Your Documentation Will Be Audited
With CMS expanding RADV audits to cover all MA contracts and dramatically increasing the number of medical records reviewed, the statistical probability that your documentation will be pulled in an audit has increased substantially. You should assume your charts will be audited and document accordingly.
2. Code What You're Treating
This is the fundamental principle that protects you in an audit. If you're actively treating or managing a condition during the encounter, it's codeable. If you're documenting a historical condition that doesn't affect current care, use a history code or don't code it at all.
A patient with well-controlled diabetes that you're monitoring and managing? Code it at every visit. A patient who had gestational diabetes 15 years ago, which resolved after pregnancy, and has no current impact on care. Don't code it as active diabetes.
3. Be Specific in Your Documentation
"History of stroke" is ambiguous. Is that an acute stroke or late effects? "Patient had a stroke in 2022, currently with residual right-sided weakness affecting mobility" is specific and supports coding for late effects and the specific deficits.
"Breast cancer" without additional context doesn't tell a coder whether this is active cancer under treatment or historical cancer in remission. "Patient completed treatment for stage II breast cancer in 2023, currently disease-free on surveillance" supports a personal history code, not an active cancer code.
4. Understand That You're Not Coding for the MA Plan
This is critical. You're not employed by the MA plan (unless you actually are, in which case different considerations apply). Your obligation is to code accurately based on ICD-10-CM guidelines and the care you're providing. You're not responsible for maximizing the plan's risk score.
If an MA plan sends you HRA results documenting diagnoses you're not treating, you're not obligated to code those diagnoses on your claims unless they're relevant to the care you're providing. If a chart review company hired by the MA plan asks you to "confirm" diagnoses you haven't assessed or aren't managing, you should be very cautious about signing off on coding for conditions you haven't independently evaluated.
5. Chronic Conditions Require Ongoing Treatment or Management
Just because a patient has a chronic condition doesn't mean it's automatically codeable at every visit. The guidelines say chronic diseases "treated on an ongoing basis" may be coded when the patient receives treatment and care for those conditions.
If you're seeing a patient for an acute respiratory infection and their diabetes, hypertension, and chronic kidney disease aren't addressed, evaluated, or affecting the care provided for that visit, the argument for coding those conditions on that specific encounter is weaker. This is a gray area, and practices vary, but the safest approach is coding what you're actually addressing.
The Clinical Documentation Integrity Role
For hospitals and health systems offering MA plans or seeing significant MA volume, Clinical Documentation Integrity (CDI) programs have become essential. But CDI in the MA context is different from CDI for inpatient billing.
Traditional hospital CDI focuses on ensuring documentation supports appropriate MS-DRG assignment and captures severity and complications that affect hospital payment under the Inpatient Prospective Payment System. MA CDI focuses on ensuring outpatient and physician documentation supports accurate risk adjustment coding.
This has led to the development of risk-adjustment coding as a subspecialty within medical coding. CDI professionals working in MA settings need expertise in:
CMS-HCC models and how diagnoses map to HCCs
ICD-10-CM coding guidelines as applied to risk adjustment
Documentation requirements that differentiate active conditions from historical conditions
The distinction between conditions that affect care versus conditions that are simply part of a patient's history
The challenge is maintaining the integrity of documentation and coding while working in an environment where every additional HCC potentially generates substantial additional revenue.
Policy Reforms on the Horizon
Several policy changes are being discussed to address MA coding intensity:
Two-Year Diagnosis Data Model
MedPAC has proposed using two years of diagnosis data for risk scores instead of one. This change would increase the number of diagnoses captured for traditional Medicare beneficiaries by allowing more time for claims data to reflect care provided to a beneficiary for such diagnoses. Expanding the lookback period would help narrow the coding intensity gap between MA and traditional Medicare.
Excluding HRA-Only Diagnoses
To prevent HRA-driven inflated payments, MedPAC has recommended excluding diagnoses identified solely through HRAs from risk adjustment calculations. This would address the problem of diagnoses documented on HRAs but never appearing in any actual clinical encounter.
Increasing the Coding Intensity Adjustment
Increasing the minimum risk score reduction from 5.9% to 20% could reduce Medicare spending by more than $1 trillion over 10 years, according to the Congressional Budget Office's December 2024 estimates. This is the most aggressive proposed reform and would effectively penalize all MA plans for the coding practices of the most aggressive plans.
Plan-Specific Coding Intensity Adjustments
The degree of upcoding varies a great deal across Medicare Advantage organizations. MedPAC estimated a 15-percentage-point variation in coding intensity among the eight largest MAOs and found 10 MAOs with coding intensity more than 20 percent higher than levels in traditional Medicare.
Rather than applying a uniform coding adjustment to all plans, some policy experts have proposed plan-specific adjustments that would penalize aggressive upcoding while rewarding plans that maintain coding practices closer to traditional Medicare levels. The MAO-specific coding intensity factor could be increased (for example, by 20 percent) for plans that both upcode aggressively and engage in problematic practices—as identified by MedPAC, CMS, the CMS Office of Inspector General, and the Government Accountability Office—such as using home visits or AI to maximize the identification and reporting of diagnoses.
The 2024 Risk Adjustment Model Changes
CMS has been implementing changes to the CMS-HCC risk adjustment model to address some of these concerns. In 2024, the risk score was a blend with 67% weight on the prior model and 33% weight on the new model. In 2025, CMS increased the weight on the new model from 33% to 67% and estimated that this would lead to a 2.45% reduction in MA risk scores, which implies that the new model could result in MA risk scores that are approximately 7.2% lower when fully implemented.
The new V28 model was designed to minimize the impact of discretionary coding on risk scores. But model changes alone haven't solved the underlying problem: financial incentives that reward more diagnosis codes regardless of whether those codes reflect active clinical management.
What Billing Companies Need to Know
If you're working in medical billing—whether in-house for a provider practice or for a billing services company—risk adjustment coding for MA patients requires a different level of scrutiny than traditional fee-for-service Medicare coding.
Understand the Downstream Impact
Your diagnosis coding doesn't just support medical necessity for the services billed. For MA patients, those codes flow into encounter data that affects risk-adjusted payments to the plan. That means your coding is more likely to be audited and scrutinized.
Don't Code Beyond the Documentation
If the provider's documentation says "history of breast cancer," don't code active breast cancer, hoping to help the revenue cycle. Code what's documented. If the documentation is ambiguous, query the provider for clarification before submitting the claim.
Educate Your Providers
Many physicians don't fully understand how their documentation affects MA risk adjustment. Provide education on:
The difference between active conditions and historical conditions
When chronic conditions should be coded on a claim
How to document specificity that supports accurate coding
Why are they receiving queries about diagnosis clarification
Build Audit-Ready Processes
Your coding should be defensible in a RADV audit. That means:
Coding only conditions documented in the medical record
Ensuring documentation supports the level of specificity coded
Maintaining a clear audit trail between documentation and code assignment
Having internal quality audits that mirror external audit methodology
Stay Current on Model Changes
The CMS-HCC model changes periodically. What qualified as an HCC under the V24 model may not qualify under the V28 model. Stay current on model changes and how they affect risk adjustment coding.
The Bigger Picture
Risk adjustment in Medicare Advantage isn't going away. With MA enrollment now exceeding traditional Medicare enrollment, risk-adjusted payments are fundamental to how Medicare pays for care for the majority of Medicare beneficiaries.
But the system as currently structured creates powerful incentives for overcoding, and the evidence from hundreds of audits shows that those incentives are influencing behavior. "Taxpayers fund billions of dollars in overpayments to MA companies each year based on unsupported diagnoses for MA enrollees," the OIG wrote. "Unsupported diagnoses inflate risk-adjusted payments and drive improper payments in the MA program".
For providers, the path forward is clear: code accurately based on what you're treating and documenting, understand that your documentation will likely be audited, and don't allow financial incentives—whether from MA plans or from your own organization's revenue goals—to push you toward coding practices that can't be defended in an audit.
For the MA program as a whole, meaningful reform will require changes that reduce financial incentives for aggressive coding while maintaining appropriate payments for plans that enroll and care for high-risk beneficiaries. That balance has proven difficult to achieve, but with $84 billion in annual overpayments projected for 2025, the pressure for reform is only going to intensify.
Sources
Bipartisan Policy Center - Medicare Advantage Risk Adjustment analysis and policy options
Medicare Payment Advisory Commission (MedPAC) - March 2024 and March 2025 Reports to Congress on Medicare Advantage coding intensity and favorable selection
USC Schaeffer Center for Health Policy and Economics - February 2025 analysis on improving Medicare Advantage by accounting for upcoding variations across plans
Centers for Medicare & Medicaid Services - Medicare Advantage capitation rate announcements and risk adjustment model updates
Department of Health and Human Services Office of Inspector General - Multiple Medicare Advantage compliance audits and risk adjustment data validation reports (2017-2025)
Modern Healthcare - May 2025 reporting on CMS's expanded RADV audit program
Medicare Rights Center - Analysis of Medicare Advantage payment policies and coding intensity
American Health Insurance Plans (AHIP) - May 2025 response to CMS on coding intensity methodology
Wolters Kluwer - Risk adjustment audit preparation guidance based on OIG work plans
Norwood - November 2024 analysis of OIG audit methodologies and MAO responses
Cotiviti - Risk adjustment compliance audit findings and takeaways
Morgan Lewis - April 2025 analysis of risk adjustment enforcement trends
Healthcare Finance News - OIG audit findings on Medicare Advantage overpayments
BMC Health Services Research - Peer-reviewed research on coding intensity variation in Medicare Advantage (published 2024)
