By Ciannah Correa and Neil Patil
For years, high coding intensity in the Medicare Advantage (MA) program, sometimes caused by “upcoding,” has served as what many policymakers and analysts view as a central flaw in the MA program. The Medicare Payment Advisory Commission (MedPAC) estimates that, in 2026, higher coding intensity by MA plans directly contributes to $22 billion out of a total $76 billion in higher payments to MA, as compared to what the government would have paid if those enrollees were instead in Medicare Fee-For-Service (FFS). While estimates of higher payments vary by year and methodology, the direction of the finding is not largely contested, despite the fact that the MA program was originally created to lower costs for the federal government while achieving the same or better-quality care as Medicare FFS by leveraging private insurers’ expertise.
Has the phase-in of v28 solved upcoding?
Changes to the MA risk adjustment model can reduce coding intensity and achieve federal savings. Since 2024, CMS has been phasing in version 28 (v28) of its hierarchical condition category risk-adjustment model. V28 aims to decrease MA plans’ rates of coding intensity and upcoding by reclassifying some of the most commonly coded diagnoses – those that led to higher payments without corresponding increases in care provided – to make them harder to code inappropriately, among other technical changes and updates. It has been fully phased in for 2026.
MedPAC annually revises its estimate of how MA payments compare to what federal spending would have been under Medicare FFS to incorporate more recent data. This year, the topline is striking: It found that higher MA coding intensity (and upcoding) will result in 4% higher payments to MA plans for 2026. That’s much lower than MedPAC’s previous estimates. Higher coding now accounts for less than a third of the total estimated 14% higher payments to MA plans than what spending would have been in Medicare FFS.
MedPAC estimates that uncorrected coding intensity, or coding intensity by MA plans above and beyond the 5.9% statutory coding pattern adjustment that resulted in overpayments to MA, reached its highest level at about 10% in 2022 and 2023. It’s tempting to compare that peak to the recent 2026 estimate of 4% – leading some to see those numbers and ask: Has v28 “solved” upcoding?
In this blog, we’ll break down why a conclusion that v28 has solved upcoding completely is incorrect and discuss where policymakers might consider going from here.
What v28 has effectively done so far:
- In 2025, MedPAC estimated that, before CMS’s statutory requirement to reduce MA payments by 5.9%, MA spending was 16.4% higher than what federal spending would have been if the same beneficiaries were in Medicare FFS. MedPAC annually estimates MA coding intensity for the latest available data year, and then projects trends forward to evaluate higher MA spending for more recent years. For example, to prepare estimates for its 2025 report, MedPAC analyzed MA payment data from 2023, data from when CMS used the previous v24 risk adjustment model.
- In its March 2026 report, MedPAC lowers its estimates for coding intensity in 2025 from 16.4% to 12.5% after incorporating newer data from 2024 — the first year of the v28 phase-in. This more recent data also allowed MedPAC to revise its estimates about how quickly risk scores have been growing over time – a dominant factor in the estimated decrease in coding intensity
- After deducting CMS’s 5.9% coding pattern adjustment and accounting for a slower risk score trend, MedPAC now finds that 4% of overpayments are likely because of higher coding intensity.
So yes, MedPAC’s early research shows that v28 has reduced the magnitude of coding intensity. MedPAC also found that the decrease has disproportionately affected the plans that had the highest-intensity coding practices relative to Medicare FFS.
Signs that there are still aspects of upcoding that v28 has not fully fixed:
- MedPAC estimates that coding intensity is still 10.3% higher in MA than Medicare FFS.
- Average MA risk scores are still increasing relatively year over year, as demonstrated by CMS’s annual calculation of a positive risk score trend in the Advance Notice.
- Drastically high coding intensity is still achievable under v28, with one MA organization reported to have over 20% higher coding intensity relative to Medicare FFS during the v28 phase-in as of 2024. MedPAC has reported a large range of variation in MA plans’ coding intensity, from a few MA organizations coding at a lower rate than Medicare FFS, to some reaching 10-20% higher coding intensity than Medicare FFS.
- Even though MedPAC projects a downward trend in the percent of MA overpayments caused by coding intensity from 2024 to 2026 (the time period where v28 was being phased in to replace v24), such a decrease is not guaranteed, nor likely, to continue forever. Further, 2025 and 2026 are projected estimates of coding intensity based on estimated impacts,
Upcoding has been slowed down, but it’s likely not yet gone away.
MA markets will likely continue prioritizing the coding arms race in response to v28
The market is still shifting and adapting to a very new, very recent market pressure (v28). It has done this before: By the early 2010s, most MA organizations had invested heavily in, and many had mastered, the risk adjustment and diagnosis coding side of the business, and coding intensity differences began to take off. Then, in 2018, the 5.9% statutory minimum coding intensity adjustment first took full effect – and 2018-2019 were the last years where MedPAC recorded just 4% higher payments to MA plans compared to Medicare FFS caused by MA plans’ higher coding intensity practices. MA plans adapted to this adjustment by using technology and developing effective practices – they achieved ever-growing overpayments relative to Medicare FFS attributed to coding from 2020 through 2023, and continuously higher rebates per enrollee, from $1426 annually in 2020 to $2,388 in 2026.
Now, the MA market is in the very first year of the complete v28 phase-in. Novel technology investments and adaptive methods are likely to be explored and mastered in response to v28. While v28 has, so far, effectively diminished the easier gains MA plans had previously adapted to achieve, v28 does not fundamentally change the payment structure MA plans are competing on.
MA plan representatives have explained that under the current payment structure, success in MA is still heavily influenced by how effectively a plan can translate enrollee characteristics into the highest-yield risk-adjusted payments, which is sometimes more an indicator of success than how effectively plans provide care. This is not a moral critique of MA organizations; it is a predictable outcome of the incentives under the MA program. Even CMS implicitly acknowledges this by estimating for increases in coding intensity trends year over year, even under v28. The current system effectively forces plans to compete on coding sophistication.
And because payment structures are staying the same (for now), insurers have continued and will continue to leverage artificial intelligence (AI) and strategic investments (including third party chart review organizations) to maximize coding diagnoses. Since the introduction of v28, companies have started offering training programs, specialized AI technology, diagnostic-coding manuals, and other risk-score-optimization strategies for the transition to v28 risk model. The market will adapt in this risk-score coding arms race, so long as it remains feasible to do so.
Therefore, today, upcoding has not been solved, despite mathematical estimates and hypothetical projections that it could be.
Even with v28, the costs are high, and CMS has taken more steps forward
“Four percent in MA higher payments compared to what would have been paid in Medicare FFS” translated to just $7-10 billion in overpayments in 2018-2019. By comparison, 4% in MA higher payments (in 2026) now amounts to $22 billion annually in higher spending on MA plans. This increase has led to an unprecedented average of over $2,300 in rebate payments per enrollee per year. As the government continues to spend more on Medicare at large, and as more adults age into Medicare and live longer, the amount of money at stake may continue to rise, even if the percentage of overpayment attributed to coding appears to be the same or smaller than previous years.
CMS has taken note. The Calendar Year (CY) 2027 Advance Notice has proposed a long-standing recommendation to exclude diagnoses from unlinked chart reviews from risk adjusted payments, a major step to disincentivize plans from the risk-score coding intensity arms race by prohibiting a key technique plans frequently misuse to get ahead in that race.
CMS expects that, on its own, excluding diagnoses from unlinked chart reviews from risk adjusted payments will decrease year over year payments to MA plans by -1.53%, beyond the effectiveness of v28. Such a decrease represents $7.12 billion in net savings to the federal government in CY 2027.
In tandem, the v28 phase-in and the proposed changes in the CY 2027 Advance Notice could very nearly shutter the financial effects of MA coding intensity. However, in response to comments, it is possible that CMS modifies or chooses not to finalize proposals in the Advance Notice. Even if the policies are finalized, the mere implementation of these changes, however, should not be mistaken for a final resolution in the broader context of MA overpayment.
How can policymakers continue to move the needle to right size MA payment?
The most important thing policymakers can do is to watch diligently to see how the MA market responds before formulating an opinion on the overall impact of v28. Policymakers may consider:
- Monitoring how MAOs adapt to v28, rather than relying solely on static projections. Whether coding intensity continues to decline as expected, and for whom, will be key to observe.
- Evaluating how smaller MA plans respond to the risk adjustment changes. Competition among smaller MA organizations is an important consideration for CMS, as stated in the RFI released in November 2025. CMS and policymakers should therefore monitor and evaluate how smaller MA organizations respond to the risk adjustment changes and potentially consider policies to ensure policies continue to promote competition.
As MA markets adapt to v28 and (potentially) to the exclusion of diagnoses from unlinked chart reviews, policymakers may consider future legislation and/or regulatory policies to continue improving MA payment accuracy according to updated market evidence and research.
For descriptions and quick analyses of other policy proposals being considered to reform risk adjustment, see MPI’s Compendium of Medicare Advantage and Part D proposals.