Just a year after one of the largest Part D benefit and risk adjustment model realignments in history,¹ significant model changes may again affect Medicare Advantage organizations (MAOs) and stand-alone prescription drug plans (PDPs) in 2026.

Introduction

This paper is a continued examination of major changes to the Part D prescription drug (RxHCC) model to align with Centers for Medicare and Medicaid Services (CMS) regulations codified in the Inflation Reduction Act of 2022 (IRA). Our initial analysis focused on the model’s recalibration following the payment year (PY) 2025 benefit structure redesign.² Here, we highlight impacts from additional, significant proposed changes in the 2026 Advance Notice.^3,4

Summary of the 2026 RxHCC model changes

Following recent revisions to align the 2025 RxHCC model with the Part D plan liability shift under the IRA, CMS is now proposing the following major changes in PY 2026:⁵

• Recognition of negotiated drug prices (maximum fair prices or MFPs).
• A retooling of the normalization factor calculation methodology.

Further, CMS provided information for two variations of the model—one reflecting the impact of MFPs (the Proposed model) and another including all proposed 2026 changes except the impact of MFPs (the Alternative model).

We discuss these items in greater detail next.

MFPS and risk adjustment

Overview

The IRA included a provision allowing CMS to negotiate Medicare drug prices,⁶ and, in 2026, 10 Part D drugs will be subject to these prices. We list the products with common associated conditions and MFP discounts in Figure 1.⁷

MFPs in the 2026 RxHCC model

The ten drugs with MFPs in 2026 represent about 20 percent of 2023 gross Part D costs.⁸ Given this volume, the reduced list prices will have a significant impact on the average cost to treat patients with the associated conditions. Figure 2 illustrates this dynamic for RxHCC188 (heart failure), which can be treated by Entresto.

As illustrated, the reduction in gross costs for heart failure patients treated by Entresto, in turn, reduces the average gross cost for all heart failure patients. Since RxHCC coefficients are calibrated to reflect plan liability, a reduction in average cost generally translates to a lower RxHCC coefficient.

The 2026 RxHCC coefficients for conditions associated with MFP drugs under the Proposed model are often significantly lower than the Alternative model, which reflects the reduction in average gross cost to treat the condition—whether for the related RxHCC itself or for other comorbid RxHCCs. Conversely, many Proposed model coefficients for conditions not associated with MFP drugs increase relative to the Alternative model coefficients, such that the average risk score across the entire Medicare-eligible population remains at the necessary 1.00 level.

Normalization factor updates

Initially introduced in 2025, CMS developed separate Part D normalization factors for Medicare Advantage prescription drug plans (MAPDs) and PDPs to reflect lower PDP risk score trends relative to MAPD.

While this approach remains, CMS fundamentally revised its 2026 methodology. Previously, CMS developed a simple linear regression and applied the implied annual risk score trend from this model (the slope) to a level representing a 1.00 across MAPD and PDP combined. Beginning in 2026, CMS is proposing:

• Switching to a multiple linear regression model to better account for years affected by COVID.
• Using the coefficients in its regression model as is, effectively targeting separate 1.00 levels for MAPD and PDP cohorts.

Figure 3 displays the resulting factors for both 2026 RxHCC models in the Advance Notice.

The effects are similar between the Proposed and Alternative models. However, the change in methodology produces normalization factors that drive the MAPD/PDP differential beyond the shift first introduced in 2025. While the 2025 normalization factors meant the same MAPD beneficiary would receive a 0.118 higher risk score if they instead enrolled in a PDP,⁹ the differential is approximately 0.300 under both proposed 2026 models.

Market impacts

As described above, the impact from the change in normalization is significant in total, but it’s also relatively apparent based solely on a comparison of the factors directly published by CMS.

The incorporation of MFPs, though, is more subtle and represents somewhat of a paradigm shift in the CMS approach. Previously, CMS never incorporated manufacturer rebates into the RxHCC model. However, since the MFPs themselves are based on net prices, the new adjustments inherently incorporate the impact of manufacturer rebates for MFP drugs.

Aggregate level

Combined, the impacts of MFPs and normalization affect risk scores to varying degrees. Focusing on the Proposed model, Figure 4 displays the distribution of risk score changes by market and broad plan type.

The boxes represent the middle 50 percent risk score change percentiles for each cohort, while the solid lines inside the boxes identify the median changes and the bottoms and tops show the 25th and 75th percentile changes, respectively. The lines extending outward from each box reflect the upper and lower extremes of the results, while the solid dots indicate outlier data points. Taken together, the taller the box and the farther the lines extend from the edges, the higher the variability of the risk score changes within each cohort.

From this figure, we note the following:

• Despite the outlier data points, the overall variability across all plans and markets is reasonably low, as suggested by both the narrow height of the boxes and the lines above and below—much less than the 2025 model change outlined in our initial analysis last year.
• Although generally low, the variability between the 25th and 75th percentiles for individual PDPs is noticeably higher compared with other markets and plan types. This increased variability is likely due to a more diverse member mix among individual PDPs available in the Medicare market.
• As anticipated from the normalization factors, we expect MAPDs will be more unfavorably affected.
• The individual and employer markets appear to be impacted similarly.

Population level

We next provide 25th, 50th, and 75th percentile risk score changes for specific member population cohorts within the individual market, again for the Proposed model.

We observed the following from this figure:

• As we saw earlier, MAPDs are universally adversely affected, while PDP impacts are neutral or favorable.
• Further, the narrowness of the spread across percentiles that was apparent in Figure 4 at the broader category level carries through to the population cohort level.
• Across both MAPDs and PDPs, the patterns are generally the same, with non-low-income members affected the least favorably and other cohorts affected more favorably, starting with institutional members, then low-income members, and finally new-to-Medicare members.

Condition level

Now that we’re grounded in the macro-level risk score changes, we move next to the condition level. Since MFP drugs are more strongly correlated with certain conditions, we expect select RxHCCs in the proposed model to change specifically from the MFP plan liability impact. These changes would usually have been masked by other 2026 model updates were it not for CMS publishing both the Proposed (with MFP impact) and Alternative (without MFP impact) models.

Using weighted average¹⁰ model coefficients, we observed 23 (out of 84) RxHCCs with a coefficient change (positive or negative) of roughly 1.0 or more between the 2026 Proposed and 2025 models. By then comparing the coefficient changes between the Proposed and Alternative models, we can impute a loose “contribution,” or portion of the change explained, from MFP drugs. This contribution represents a substantial portion of the Proposed model impact for certain conditions. Figure 6 highlights these concepts for seven example RxHCCs.

The Average Model Coefficient Change columns in the figure show the change in coefficient for each RxHCC listed and the implied change explained by MFP. The Observed Risk Score Change section shows the change in total risk score for members that have the listed RxHCC, regardless of what other RxHCCs they might also have.

While the Average Model Coefficient Change section in the table above is interesting and highlights potential drivers of changes to the risk scores applied directly by the RxHCC model, what matters in practice is how the new coefficients affect actual, observed risk scores for members with each RxHCC (the rightmost column).

These observed risk scores don’t completely align, as you may initially expect, with the average model coefficient change between the 2026 Proposed and 2025 models (the second column). This apparent misalignment is driven by the various interactions that happen in practice for a member with a given condition (as their risk score includes more than just the pure coefficient for that condition, such as member demographics, other identified conditions, model interactions, etc.).

However, examining the observed risk score changes more closely, many of the RxHCCs in our example have either smaller positive values than other RxHCCs not shown or negative values precisely because they are related to drugs subject to MFPs. These conditions include:

• Those directly treated by MFP drugs, such as diabetes with complications (Januvia, Fiasp/Novolog, Farxiga, and Jardiance), leukemia (Imbruvica), and heart failure (Entresto).
• Those comorbid with another condition treated by MFP drugs such that the modeled plan liabilities for all related conditions are affected (hypertension).
• Those for which MFP drugs are less commonly prescribed (for instance, inflammatory bowel disease treated by Stelara).¹¹

Again, not all the patterns highlighted in Figure 6 can be attributed to MFP drugs (even if our simplistic “contribution” is high) and can result from other market events, such as changes in treatment and coding patterns between the old and new calibration periods (2022 to 2023 in this case), new drug launches, price changes, and patent losses.¹²

Conclusion

The significant increase in direct subsidy payments in 2025 heightened the focus on the Part D program and the role the RxHCC model plays in adjusting plan revenue. The CMS proposals in 2026, as well as expected changes in the coming years to reflect MFPs in 2027 and beyond, add an extra layer of complexity to an already complicated program.

While the nationwide variance inherent in the risk scores produced by the 2026 proposed model within each market, plan type, and population type is relatively low at the contract level, there is significant variance in changes among conditions (RxHCCs). Plan sponsors must consider the risk adjustment implications from both the shifting normalization factors and the Medicare Drug Price Negotiation Program when preparing 2026 Part D bids, forecasting revenue, and setting future budgets. The RxHCC model will maintain its critical role in the finances of Part D plans for years to come—with its impacts becoming only more pronounced over time.

Appendix A: Methodology

We developed the estimates in this paper by first calculating beneficiary level Part D risk scores under the 2025, proposed 2026, and alternative 2026 RxHCC models, using nationwide 2022 Medicare Advantage encounter and fee-for-service (FFS) claims data as well as 2023 MAPD and PDP eligibility data in the CMS research identifiable files (RIFs). We then summarized the data and created a comparison of the risk score changes among various cohorts.

We performed the following steps in the risk score calculation:

• Filtered claims consistent with the CMS Encounter Data System (EDS) logic.
• Assigned member status using CMS eligibility files.
• Assigned drugs to an MFP category (i.e., MFP vs. non-MFP) using National Drug Codes (NDCs) provided by CMS.
• Ran each member through the respective CMS RxHCC model software.
• Mapped the proper demographic and condition risk score to each member based on the RxHCC outputs and our member status mapping.
• Excluded the following plan types: Program of All-Inclusive Care for the Elderly (PACE), Cost, Private Fee-for-Service (PFFS), Limited Income Newly Eligible Transition (LI NET) and Direct Contract PDPs.

After the risk score calculations, we assigned plan and product to each contract/plan benefit package (PBP) using published CMS information. For our numerical assessments, we compared calculated risk scores by the selected cohorts.

Appendix B: Caveats and limitations

Milliman developed certain models to estimate the values included in this analysis. The intent of the models is to estimate marketwide risk scores under the proposed 2026 CMS RxHCC models. It may not be appropriate for any other purpose. We reviewed the models, including the inputs, calculations, and outputs. We believe they are consistent, reasonable, appropriate to the intended purpose, and compliant with generally accepted practice and relevant actuarial standards.

The models reflect data as inputs. We relied on the following information:

• 2022 diagnoses and 2023 eligibility CMS RIF data.
• The CMS RxHCC model software for various benefit years.
• Publicly available data and information on the Part D risk adjustment program from CMS, including the identification of eligible MFP drugs.
• CMS 2023 public use files (PUFs).
• Medicare Advantage and Part D Advance Notice and Rate Announcement documents.
• Other publicly available information.
• Our interpretation of federal guidance.

We accepted this information without audit but reviewed it for general reasonableness. Our results and conclusions may not be appropriate wherever information is not accurate.

Actual results will differ from those developed in the paper for a variety of reasons, and the following limitations should be considered when analyzing our results:

• The datasets used in the analysis represent historical data with its own mix of population types, plan selections, utilization, and acuity. This data may not reflect any one Medicare population in a given state/market or future claim patterns or cost levels in future periods (despite using the CMS risk adjustment model from future benefit years).
• Our calculated risk scores will not tie exactly to any published CMS risk scores.
• The datasets used do not have the same information available to CMS to assign risk scores for revenue payments, including, but not limited to:
• Imperfect mapping of member institutional status.
• Potentially incomplete retroactive data adjustments.
• We excluded certain populations from the analysis.
• We applied the following hierarchy when assigning a member’s population type: new enrollee, then institutional, followed lastly by income status.
• Composite results in a given cohort (e.g., market or plan type) reflect an average across many members, and any one member’s experience will likely deviate from the average of the cohort.
• We did not adjust for the potential impact of newly added ICD-10 diagnosis codes applicable in PY 2025 and PY 2026 that did not exist during the 2022 diagnosis period. This, generally, will cause our CMS model implementation to under-identify conditions.
• At any time, CMS could update or refine risk adjustment rules, guidance, and/or regulations such that the results presented in this analysis no longer apply.

---

¹ Centers for Medicare and Medicaid Services. (2024). Announcement of Calendar Year (CY) 2025 Medicare Advantage (MA) Capitation Rates and Part C and Part D Payment Policies. Retrieved March 3, 2025, from https://www.cms.gov/files/document/2025-announcement.pdf.

² Robb, M., Petroske, J., & Rodrigues, D. (April 26, 2024). A prescription for change: How the 2025 Medicare Part D risk adjustment (RxHCC) model overhaul will affect risk scores [Milliman white paper]. Retrieved March 3, 2025, from https://www.milliman.com/en/insight/prescription-for-change-2025-medicare-part-d-risk-adjustment-model.

³ Centers for Medicare and Medicaid Services. (2025). Advance Notice of Methodological Changes for Calendar Year (CY) 2026 for Medicare Advantage (MA) Capitation Rates and Part C and Part D Payment Policies. Retrieved March 3, 2025, from https://www.cms.gov/files/document/2026-advance-notice.pdf.

⁴ The model changes affect Program of All-Inclusive Care for the Elderly (PACE) and non-PACE organizations differently. Since we focus only on non-PACE organizations, all descriptions, analyses, and metrics will only reflect those populations.

⁵ In addition to these major changes, CMS plans to incorporate minor revisions for both the 2026 Part D defined standard benefit changes and the availability of more recent data for model calibration (2022 diagnoses and 2023 claims).

⁶ Cline, M., Karcher, J., Klaisner, J., & Robb, M. (August 18, 2022). Weathering the reform storm: The Inflation Reduction Act’s changes to Medicare and other healthcare markets. Retrieved March 3, 2025, from https://www.milliman.com/en/insight/Weathering-the-reform-storm.

⁷ Centers for Medicare and Medicaid Services. (August 15, 2024). Medicare drug price negotiation program: Negotiated prices for initial price applicability year 2026 [fact sheet]. Retrieved March 3, 2025, from https://www.cms.gov/newsroom/fact-sheets/medicare-drug-price-negotiation-program-negotiated-prices-initial-price-applicability-year-2026.

⁸ Ibid.

⁹ Robb, M., Petroske, J., & Rodrigues, D. (April 26, 2024). A prescription for change: How the 2025 Medicare Part D risk adjustment (RxHCC) model overhaul will affect risk scores [Milliman white paper]. Retrieved March 3, 2025, from https://www.milliman.com/en/insight/prescription-for-change-2025-medicare-part-d-risk-adjustment-model.

¹⁰ Based on 2023 nationwide enrollment by cohort and non-disabled RxHCC coefficients.

¹¹ Brewer, A. (September 24, 2024). Stelara for Crohn’s disease. Medical News Today. Retrieved March 3, 2025, from https://www.medicalnewstoday.com/articles/drugs-stelara-for-crohns-disease.

¹² For example, the reduction in the coefficient for multiple sclerosis may be at least partially attributable to a patent loss in 2023 for a leading product in that class (Aubagio). See National Multiple Sclerosis Society. (April 18, 2023). First generic versions of Aubagio® (teriflunomide) now available. Retrieved March 3, 2025, from https://www.nationalmssociety.org/news-and-magazine/news/first-generic-versions-of-aubagio.