Today, the Centers for Medicare & Medicaid Services (CMS) released the Calendar Year (CY) 2025 Advance Notice of Methodological Changes for Medicare Advantage (MA) Capitation Rates and Part C and Part D Payment Policies (the Advance Notice). CMS will accept comments on the CY 2025 Advance Notice through 6:00 PM Eastern Time on Friday, March 1, 2024, before publishing the final Rate Announcement on or before April 1, 2024.
Each year, CMS is required to update MA payment rates and regularly conducts technical updates to make improvements needed to keep MA payments up-to-date and accurate. CMS makes technical updates and improvements through the Advance Notice and Rate Announcement process for this purpose. If finalized, the proposed policies in the CY 2025 Advance Notice are projected to result in a net increase in MA payments to plans on average year-over-year in CY 2025. We also proposed updates, where applicable, to reflect the Part D redesign as required by the Inflation Reduction Act. The following FAQs support the public’s understanding of the release.
Net Payment Impact
The chart below indicates the expected impact of the proposed policy changes on MA plan payments relative to last year.
Year-to-Year Percentage Change in Payment
Impact
2025 Advance Notice
Effective Growth Rate
Change in Star Ratings 2
MA Coding Pattern Adjustment
Risk Model Revision and FFS Normalization 3
MA risk score trend 4
Expected Average Change in Revenue
+3.70%
1 Rebasing/re-pricing impact is dependent on finalization of the average geographic adjustment index and will be available with the publication of the CY 2025 Rate Announcement.
2 Change in Star Ratings reflects the estimated effect of changes in the Quality Bonus Payments for the upcoming payment year.
3 The impact of the update to the Fee-for-Service (FFS) normalization factors for MA risk adjustment is not shown in the fact sheet separately because there is considerable interaction between the impact of the MA risk adjustment model updates and the normalization factor update. Therefore, the combined impact is shown in the fact sheet.
4 The MA risk score trend is the average increase in MA risk scores, not accounting for normalization and coding pattern adjustments to MA risk scores, which are shown in separate rows. The risk score trend is calculated by using MA risk scores from 2018 – 2020, calculated using the risk adjustment models, with the appropriate blend for CY 2025. The trend is an industry average and individual plans’ experience will vary.
Growth Rates
The Effective Growth Rate reflects the current estimate of the growth in benchmarks used to determine payment for MA plans. This growth rate is largely driven by the growth in Medicare Fee-For-Service (FFS) per capita costs, as estimated by the Office of the Actuary. Included in the 2025 growth rate estimate is a technical adjustment to the per capita cost calculations related to indirect and direct medical education costs associated with services furnished to MA enrollees. In CY 2025, CMS is proposing to continue the three-year phase-in of this technical adjustment finalized in the CY 2024 Rate Announcement. We propose to apply 67 percent of the adjustment in CY 2025.
Part C Risk Adjustment Model
CMS finalized an updated Part C Risk Adjustment Model in the CY 2024 Rate Announcement and began a three-year phase-in of the use of that model, referred to as the 2024 CMS-HCC model, starting with CY 2024. For CY 2025, CMS is proposing to continue phasing in the updated risk adjustment model.
The updated model included important technical updates to improve the predictive accuracy of the model, including restructured condition categories using the International Classification of Diseases (ICD)-10 classification system (instead of the ICD-9 classification system), updated underlying FFS data years (from 2014 diagnoses and 2015 expenditures to 2018 diagnoses and 2019 expenditures), an updated “denominator year” in determining the average per capita predicted expenditures to create relative factors in the model, as well as applying our longstanding principles to make revisions focused on conditions that are subject to more coding variation.
For CY 2025, CMS is proposing to proceed with the phase-in as described in the CY 2024 Rate Announcement by blending 67 percent of the risk score calculated using the updated 2024 MA risk adjustment model with 33 percent of the risk score calculated using the 2020 MA risk adjustment model. Additionally, CMS is considering a more sophisticated calculation methodology for the FFS normalization factor that more accurately addresses the impacts of the COVID-19 pandemic without deleting data years. The FFS normalization factor is an adjustment to risk scores calculated using the models for CY 2025 payment to account for the expected growth in the average FFS risk score over time.
MA Risk Score Trend
The MA risk score trend is the average increase across plans in MA risk scores, not accounting for normalization and coding pattern adjustments to MA risk scores. The trend reflects increases in MA risk scores, which can be due to several factors, including changes in demographics and coding patterns. The trend is an industry average and individual plans’ experience will vary. We calculate the MA risk score trend for each model used in the payment year separately and blend based on the phase-in of the updated model. The risk score trend is 3.30 percent under the 2024 CMS-HCC model and 5.00 percent under the 2020 CMS-HCC model. CMS blended the MA risk score trends using the same blend proposed to be used to determine CY 2025 risk scores (i.e., 67 percent of the MA risk score trend under the 2024 CMS-HCC model and 33 percent under the 2020 CMS-HCC model). This blended MA risk score trend for CY 2025 is 3.86 percent.
Puerto Rico
The proportion of people with Medicare who receive benefits through MA (as opposed to FFS) is far greater in Puerto Rico than in any other state or territory. The policies proposed and under consideration and open for comment for CY 2025 would continue to provide stability for the MA program in the Commonwealth and to Puerto Ricans enrolled in MA plans. These policies include basing the MA county rates in Puerto Rico on the relatively higher costs of individuals in FFS who have both Medicare Parts A and B and applying an adjustment regarding the propensity of individuals with zero claims. We also seek comment on alternate adjustment approaches that may be appropriate in Puerto Rico.
Inflation Reduction Act (IRA) Updates for 2025
The IRA made several amendments and additions to the standard Part D drug benefit for CY 2023 and subsequent years. Part D benefit-related IRA updates will be in place for CY 2025 and are described in the Advance Notice and related Draft CY 2025 Part D Redesign Program Instructions include the elimination of the coverage gap phase to affect a three-phase benefit (deductible, initial coverage, and catastrophic) and cap out-of-pocket costs at $2,000 for CY 2025. Other previously implemented IRA benefits will continue, including no cost sharing for enrollees in the catastrophic phase, a $35 monthly cap on enrollee cost sharing for each covered insulin product, and no cost sharing for adult vaccines recommended by the Advisory Committee on Immunization Practices that are covered under Part D. For more details, please see the Fact Sheet for the Draft CY 2025 Part D Redesign Program Instructions available at https://www.cms.gov/files/document/draft-cy2025-part-d-redesign-program-instruction-fact-sheet.pdf .
Part D Risk Adjustment
CMS is proposing updates to the Part D risk adjustment model to reflect the redesign of the Part D benefit as required by the IRA, including the increase in plan liability given the $2,000 cap on annual out-of-pocket spending for CY 2025, and the new Manufacturer Discount Program. These updates to the Part D risk adjustment model are essential for plan sponsors to develop accurate bids for CY 2025. In addition, the Advance Notice proposes to calibrate the model using newer data years as well as proposes updates to the normalization methodology to reflect differences between Medicare Advantage prescription drug (MA-PD) plan and stand-alone prescription drug plan (PDP) risk score trends.
Part C and D Star Ratings
In the Advance Notice, CMS provides information and updates in accordance with the Star Ratings regulations at §§ 422.164, 422.166, 423.184, and 423.186. In addition, CMS solicits input on future measures and concepts as we continue to enhance the Star Ratings over time.
Star Ratings updates in the CY 2025 Advance Notice include providing the list of eligible disasters for adjustment, non-substantive measure specification updates, and the list of measures included in the Part C and D Improvement measures and Categorical Adjustment Index for the 2025 Star Ratings. We are also soliciting initial feedback on substantive measure specification updates and comments on new measure concepts and demonstrating progress towards the Universal Foundation.
Process
Comments on the proposals set forth in the Advance Notice and the Draft CY 2025 Part D Redesign Program Instructions must be submitted by 6:00 PM Eastern Time on Friday, March 1, 2024. The 2025 Rate Announcement and the Final CY 2025 Part D Redesign Program Instructions will be published no later than Monday, April 1, 2024.
To submit comments or questions on the Advance Notice electronically, go to www.regulations.gov , enter the docket number “CMS-2024-0006” in the “search” field, and follow the instructions for ‘‘submitting a comment.’’
Please submit comments or questions on the Draft CY 2025 Part D Redesign Program Instructions to PartDRedesignPI@cms.hhs.gov . The Draft CY 2025 Part D Redesign Program Instructions can be found at https://www.cms.gov/files/document/draft-cy-2025-part-d-redesign-program-instruction.pdf .
Frequently Asked Questions (FAQs) on the 2025 Medicare Advantage and Part D Advance Notice
Each year, CMS is required to update Medicare Advantage (MA) payment rates and periodically conducts technical updates to risk adjustment in order to make improvements needed to keep MA payments up-to-date and accurate. CMS makes technical updates and improvements through the Advance Notice and Rate Announcement process for this purpose.
The updates in the Advance Notice include:
There is a statutory requirement to provide a minimum 30-day comment period and release the Rate Announcement, which finalizes payment updates and policies, not later than the first Monday in April. To comply with these requirements, comments are due by 6:00 PM Eastern Time on Friday, March 1, 2024, and then, after consideration of comments received, CMS will issue the CY 2025 Rate Announcement no later than April 1, 2024.
MEDICARE ADVANTAGE
1. How would the proposed changes in the 2025 Advance Notice impact payments to MA plans?
MA payments from the government to MA plans are expected to increase by 3.70 percent on average from 2024 to 2025, as proposed. This is over a $16 billion increase in expected MA payments for next year. This expected increase includes consideration of the various elements that impact MA payment, such as growth rates of underlying costs, 2024 Star Ratings for 2025 quality bonus payments, continued phase-in of risk adjustment model updates that were implemented in CY 2024, and increases to risk scores because of MA risk score trend, which can be driven by a number of factors including MA demographics and coding patterns. This increase represents the average expected payment update across plans, and thus, there will be variation among plans in terms of their plan-specific payment impacts, including plans that would see a larger or smaller impact year over year. As in past years, the projected change in payment can change between the Advance Notice and Rate Announcement, published no later than April 1, 2024.
2. How will the proposed changes impact individuals’ MA premiums and benefits in 2025?
If finalized, CMS anticipates stable premiums and benefits for individuals for CY 2025, as was the case for offerings in CY 2024, which was the first year of the updated risk adjustment model implementation. For CY 2024, average premiums and benefits for MA remained stable. The MA average monthly plan premium remained stable with an increase of less than one dollar on average, while plan choice and average supplemental benefit offerings across plans increased.
3. What are the updates for MA risk adjustment for CY 2025?
CMS proposes to continue implementing the three-year phase-in of the updated Part C Risk Adjustment model, referred to as the 2024 CMS-HCC model, finalized in the CY 2024 Rate Announcement. CY 2025 will be the second year of the phase-in, and CMS proposes to blend 67 percent of the risk score calculated using the updated 2024 MA risk adjustment model with 33 percent of the risk score calculated using the 2020 MA risk adjustment model.
The 2024 CMS-HCC model included important technical updates to improve the predictive accuracy of the model, including restructured condition categories using the International Classification of Diseases (ICD)-10 classification system (instead of the ICD-9 classification system), updated underlying FFS data years (from 2014 diagnoses and 2015 expenditures to 2018 diagnoses and 2019 expenditures), an updated “denominator year” in determining the average per capita predicted expenditures to create relative factors in the model, as well as applying our longstanding principles to make revisions focused on conditions that are subject to more coding variation.
For more information on the 2024 CMS-HCC model, please see the 2024 Advance Notice (specifically, Section G pages 43-57) and Fact Sheet and FAQs and 2024 Rate Announcement (specifically, Section G pages 65-115) and Fact Sheet with FAQs.
The CY 2024 MA risk adjustment model changes described above will support accurate MA payments in CY 2025. This CMS-HCC risk adjustment model improves payment accuracy by incorporating more recent utilization, coding, and expenditure patterns in the relative weights of HCCs in the model and reclassifying HCCs to reflect clinical cost patterns associated with ICD-10 codes.
4. How does the continued phase-in of the updated risk adjustment model impact dually eligible individuals?
CMS is committed to ensuring that all MA individuals, including dually eligible individuals and Special Needs Plans (SNP) enrollees, can access the care they need and that plans are paid accurately for the care they provide. Payment accuracy ensures MA plan payments better reflect the expected costs of care, with higher payments going to plans serving people with greater health care needs, including individuals dually eligible for Medicare and Medicaid. Enrollment of dually eligible individuals in MA has grown rapidly in recent years. In CY 2019, MA market penetration among full-benefit dually eligible individuals surpassed non-dually eligible Medicare individuals for the first time ever. In 2024, over half of all dually eligible individuals are in MA.
As discussed in the CY 2024 Rate Announcement, some commenters expressed concerns that the 2024 CMS-HCC model may negatively affect dually eligible enrollees and vulnerable populations (e.g., minority individuals, those under the federal poverty level, or those with complex, chronic health conditions), and the plans that serve them (e.g., special needs plans that serve dually eligible individuals (D-SNPs) or individuals with certain chronic diseases (C-SNPs)). Some commenters predicted that the supplemental benefits offered to SNP enrollees would be reduced and that MA organizations may offer fewer specialized plans.
When contemplating the continued phase-in of the updated model for CY 2025, CMS carefully considered and analyzed impacts on dually eligible enrollees and D-SNPs. CMS has concluded that continuing to implement the 2024 CMS-HCC model is necessary and appropriate and increases predictive accuracy of the risk adjustment model for these individuals. As CMS explained in the CY 2024 Rate Announcement, the updates to the model improved the model’s predictive accuracy and helped to ensure that higher payments are available to plans that serve enrollees with more costly health care needs. For reference, we have included in Attachment VII of the CY 2025 Advance Notice the 2024 CMS-HCC model predictive ratios that were published in the 2024 Rate Announcement, which shows improved predictive ratios across segments, including for enrollees entitled to Medicare because of age or disability who are dually eligible for Medicaid and Medicare.
Additionally, the updates in the 2024 CMS-HCC model did not change protective features in the CMS-HCC risk adjustment model, first implemented in CY 2017, that ensures plans that care for dually eligible individuals are paid more to reflect the expected cost of care for peoples’ health conditions. CMS has observed that, on average, predicted risk for dually eligible populations and those in SNPs are 45 to 55 percent higher than non-dually eligible individuals.
Further, the updates to the model also do not alter changes first implemented in CY 2020 that ensure that plans receive an additional increase in payment based on the number of chronic conditions a beneficiary has. Because the updated model more accurately predicts costs and continues to ensure that plans enrolling higher-need individuals receive higher payments, we believe that the updated model will pay appropriately for these individuals and support the delivery of care necessary for these high-need individuals.
Plan bidding signaled strong interest in the D-SNP market for CY 2024, with the number of D-SNPs increasing by approximately 8 percent. Additionally, plans projected in their bids that MA enrollment overall is expected to grow over 7 percent, with D-SNPs enrollment expected to grow by approximately 13 percent. Additionally, on average, rebate dollars that fund supplemental benefits for D-SNPs increased for 2024 and by more than the average increase in rebate dollars across all MA plans, indicating adequate payment levels that allow these plans to bid competitively below benchmark.
In addition, as we did at an industry level, CMS calculated the MA risk score trend for dually eligible individuals (i.e., full-benefit and partial-benefit dually eligible people residing in the community) and found that this trend is 4.33 percentage points higher than it is for non-dually eligible individuals accounting for the increased blend. As explained below, the MA risk score trend represents the average growth in MA risk scores for the payment year. It is imperative to consider the MA risk score trend in concert with the impact of the updated risk adjustment model (as well as other changes to payment factors) to accurately predict payment impacts in the following year. On average, this growth in MA risk scores will more than offset the impact of the new risk adjustment model and normalization for dually eligible individuals.
The risk adjustment model does not set payments to plans and does not reimburse plans for specific conditions. Plan bids project the average revenue needed to cover all Part A and B benefits, and the risk score adjusts this amount based on the health status of a plan’s enrollee population. The 2024 CMS-HCC model will accurately predict the differential between individuals who are expected to have low and high risk. If a specific HCC (or diagnosis code mapped to a specific HCC) is no longer included in the payment model, coefficients of other HCCs and demographic factors will be increased such that the model continues to predict the overall total expenditures.
Ultimately, the steps CMS is taking to update the MA risk adjustment model are intended to pay more for the sickest patients and spread relative weights for HCCs, and therefore payments, to plans accurately. Accordingly, CMS is confident that continued phase-in of the 2024 CMS-HCC risk adjustment model will continue to strengthen payment accuracy for individuals dually eligible for Medicare and Medicaid.
5. What is the impact of the updated MA risk adjustment model on enrollees with complex needs and preventive care?
The 2024 CMS-HCC model is not expected to reduce access to preventive or other necessary care for enrollees with complex needs. As is also discussed in the 2024 Rate Announcement, conditions in the model are used as predictors of relative costs, not as direct reimbursement for the treatment of each condition. Plan bids project the average revenue needed to cover all Part A and B benefits (subject to limited exclusions for Medicare benefits that are not covered by MA plans, such as the cost of kidney acquisition for transplant), and the risk score is used to assess the relative cost of a plan’s enrollee population. Further, it is the total risk score that predicts the relative cost of a beneficiary, and each condition predicts part of the costs; therefore, each relative factor cannot be assessed in isolation. If a specific HCC (or diagnosis code mapped to a specific HCC) is no longer included in the payment model, coefficients of other HCCs and demographic factors will be increased such that the model continues to predict the overall total expenditures. The updated risk adjustment model improves the predictive accuracy of the model by incorporating recent costs and utilization patterns and is developed using ICD-10 codes and ensures that plans that enroll individuals with more needs receive higher payments.
As stated in the 2024 Rate Announcement, the function of the CMS-HCC model is to minimize incentives for MA plans to compete for the healthiest individuals. To accomplish this objective, the risk adjustment model accurately predicts relative risk across subgroups of individuals and pays plans more for populations that have more complex health needs. For 2024, CMS updated the model to reflect the ICD-10 classification system, which has been used by the health care industry since 2015, and more recent underlying cost and utilization patterns to improve risk prediction. Last year, stakeholders submitted comments that asked several questions about the updated 2024 CMS-HCC risk adjustment model and the impacts on enrollees with complex needs, which we share again in the next few questions.
6. Why are some diagnosis codes that were in the 2020 CMS-HCC model not in the 2024 CMS-HCC model?
As previously stated, in order to update the model, CMS had to complete a reclassification to align the underlying HCCs with the ICD-10 classification system and risk adjustment principles. The MA risk adjustment models are based on diagnosis code groupings (called HCCs, Hierarchical Condition Categories) that are clinically related and have similar ability to predict Medicare costs. While all roughly 74,000 diagnosis codes are mapped to an HCC, only a subset of HCCs are included in the model for payment following well-established principles to determine which HCCs best predict Medicare costs. CMS developed the HCCs using empirical evidence on frequencies and predictive power; clinical input on relatedness, specificity, and severity of diagnoses; and professional judgment on incentives and diagnostic patterns relative to the classification system. Since the health care system transitioned to ICD-10 in 2015, the codes coming in from plans to CMS were ICD-10 codes, but before CMS could reclassify the HCCs in accordance with the ICD-10 classification system, CMS needed to wait for ICD-10 diagnosis and coding practices, treatment methods, and costs to stabilize.
As is done periodically, CMS rebuilt the HCCs from the ground up to take into account the clinical classification system changes, disease patterns, treatment methods and their costs, and coding practices, as well as compositional changes within the Medicare population. [1] Doing so improves the model's predictive ability. CMS followed well-established principles to determine which HCCs best predict Medicare FFS costs and distinguish between the level of severity and complexity of disease across large subpopulations. [2] After creating the new HCCs based on ICD-10, many of the HCCs resulting from this reclassification differ from the previous HCCs, with a number of diagnosis codes being mapped to different HCCs, as well as some HCCs being removed from the payment model and new HCCs created (see the 2024 Advance Notice Section G. Table II-4. and the 2024 Rate Announcement for more detail). The 2024 CMS-HCC model will accurately predict the differential between individuals who are expected to have low and high risk. If a specific HCC (or diagnosis code mapped to a specific HCC) is no longer included in the payment model, coefficients of other HCCs and demographic factors will be increased such that the model continues to predict the overall total expenditures.
In the finalized 2024 model, there were 115 HCCs in the payment model and 151 HCCs that were not in the payment model. It is typical that most HCCs are not in the payment model. The 2020 CMS-HCC model included 9,797 ICD-10 codes for payment or 13 percent of ICD-10 codes. The finalized 2024 CMS-HCC model included 7,770 ICD-10 codes, or 10.5 percent of ICD-10 codes. Over 95 percent of the 2,000 ICD-10 codes that no longer map to HCCs included in the payment model were moved to non-payment HCCs as part of updating the model to align with the ICD-10 classification system; the remainder are due to step (3), the clinical reclassification that we provide more detail on below. As part of this ICD-10 remapping process, ICD-10 has a much larger code set than ICD-9, which provides the opportunity to include more specificity, and for certain clinical categories (e.g., depression), the clinical concept used to classify these conditions is very different, so the HCCs for these conditions have changed more than for some other conditions.
The transition from ICD-9 to ICD-10 accounted for roughly 97 percent of the codes that were not included in the 2024 CMS-HCC payment model. The remaining codes that were not included were due to the application of ‘Principle 10’ (described as step (3) above), which involves making changes related to variation in coding, which led to the removal of certain HCCs and impacted the mapping of codes to HCCs. For example, we removed HCC 47 Protein-Calorie Malnutrition, HCC 230 Angina Pectoris, and HCC 265 Atherosclerosis of Arteries of the Extremities, with Intermittent Claudication since there is wide variation in coding and clinical implications of these HCCs suggest they not a consistently reliable predictor of prospective medical expenditures. Variation in coding between MA and FFS may indicate that clinical indicators for the diagnosis are broad or need significant interpretation, that the condition is being diagnosed and documented in situations where it has no clinical significance or where it does not require or affect patient care, treatment or management as required by the ICD-10 Coding Guidelines. Such diagnoses do not consistently and reliably predict costs as well because of the variation in their coding. Please see the ‘Table III-3 Summary Statistics for the 2020 CMS-HCC and 2024 CMS-HCC Classifications in the 2024 Rate Announcement’ on page 86 in the 2024 Rate Announcement for additional information on the underlying diagnosis code counts for the 2020 CMS-HCC model and the 2024 CMS-HCC model. Additionally, see pages 89-93 in the 2024 Rate Announcement for more information on HCC changes and Principle 10.
7. Why does the 2024 CMS-HCC model include fewer diagnosis codes for mental health conditions than the 2020 CMS-HCC model?
Because the ICD-10 classification system differed significantly from ICD-9 for mental health diagnosis codes, the mental health HCCs were reconstructed, and certain diagnosis codes that are not consistent predictors of costs, like mild depression in remission, were mapped to non-payment HCCs. The majority (the remaining 350) of these codes were kept in the payment model. Updating this HCC mapping does not mean that MA plans and contracted providers are no longer required to treat mild depression or other codes that map to non-payment HCCs; MA plans are required to provide all Medicare Part A and B-covered services (subject to limited exclusions). The update merely means that based on cost data, those diagnoses are not stable predictors of prospective cost and, thus, should not increase payments to MA plans. If we did not make this update, diagnosis codes that are poor or inconsistent predicters of cost would reduce the relative weight of other codes that are good predictors of cost, possibly resulting in underpayment for complex patients with depression. Thus, under the 2024 CMS-HCC model, higher MA payments will continue for the more severe manifestations of depression consistent with risk adjustment principles.
8. Why does the 2024 CMS-HCC model include fewer diagnosis codes for diabetes than the 2020 CMS-HCC model?
As part of updating the risk adjustment model, certain diabetes codes no longer map to payment HCCs because they are not consistently reliable predictors of costs. However, over 300 diabetes codes remain in the risk adjustment model. These diabetes codes remain in the risk adjustment model because they are consistently reliable predictors of cost where more payment may be needed for diabetes care. Additionally, in the updated model, we are making the same adjustment for diabetes across all Diabetes HCCs (HCC 36, 37, and 38). In the 2020 CMS-HCC model, the top two diabetes HCCs, 17 and 18, were already constrained. When developing the 2024 CMS-HCC model, empirical data showed that in the 2020 CMS-HCC model, HCC 18 had a substantially higher prevalence in MA than FFS, and HCC 19 had a lower prevalence; the shift of individuals to a higher HCC in the hierarchy led to further evaluation of the diabetes HCCs. Under the clinical criteria used to code HCC 18, individuals with a range of clinical burden and medical expenditures may be captured in the diabetes with chronic complications HCC resulting in a diminished capacity for the model to differentiate between disease severity. For example, there may be instances where a beneficiary is coded with the more severe manifestation of diabetes based on a laboratory finding (e.g., automated urinalysis test that finds protein in the urine), even if the finding is not clinically significant and has no implications for medical treatment. Thus, the constraint was expanded to all three diabetes payment HCCs in the 2024 CMS-HCC model.
Constraining the Diabetes HCCs allows expenditures to be attributed to well-defined and significant complications of diabetes captured in other payment HCCs (e.g., chronic kidney disease and diabetic eye disease). Additionally, the updated model predicts a higher risk for patients with diabetes who have complications associated with diabetes, like chronic kidney disease, heart disease, and diabetic retinopathy. In addition, there are other payment factors, such as the condition count variable, that increases payment when individuals have more comorbidities. Thus, as with all HCC-based models, the 2024 CMS-HCC model provides a more targeted and accurate adjustment for risk for a diabetic patient because it results in higher MA payments according to the patient’s full health profile rather than using only a diabetes diagnosis as a proxy for increased health care costs. This approach would help ensure that higher payments are directed to diabetic patients with the greatest health care costs.
9. What is FFS normalization?
At a high level, the purpose of FFS normalization is to account for trends in the FFS risk scores between the last time the model was recalibrated with new FFS data and the MA payment year.
For more context, once a risk adjustment model is developed when it is used to calculate risk scores over a number of years, the average risk score changes each year due to an underlying trend that reflects a variety of changes, such as the population enrolled, the health status of the population, and also how diagnoses are coded. Therefore, when a risk adjustment model predicts expenditures in years other than the year for which the model was established (denominator year), the average risk score may no longer be the same as in the denominator year (1.0). Accordingly, an adjustment must be applied to account for the risk score trend between the denominator year and payment year. CMS applies an FFS normalization factor to MA risk scores in the payment year to account for this trend in the average FFS risk score between the denominator year risk score and the payment year. The FFS normalization factor is a projection of this average FFS risk score trend, and we apply the factor by dividing each individual MA risk score in the payment year by the normalization factor.
For the normalization factor to work as intended, CMS must predict an average FFS risk score that is a reasonably accurate projection of the future payment year’s average FFS risk score, given the information available at the time the normalization factor is calculated. By setting the average FFS risk score to 1.0, we ensure that the risk scores used to pay MA organizations align with the risk scores used to set the MA benchmarks. CMS calculates benchmarks based on the average county-level FFS risk score that has been adjusted so that the overall FFS risk score across all counties is 1.0.
10. What is CMS considering regarding the FFS normalization calculation methodology for CY 2025?
For CY 2025, CMS is again examining the inclusion of data years impacted by the COVID-19 pandemic and, given our findings, as discussed in the 2025 Advance Notice, we also assessed a methodology that we believe more reasonably accounts for the FFS average risk score trend since the beginning of the COVID-19 pandemic. Because excluding data years under the existing linear slope methodology does not produce reasonable projections and is no longer supportable when considering the actual 2023 average FFS risk score, CMS developed and is proposing a more sophisticated multiple linear regression methodology for calculating normalization factors for CMS-HCC models for CY 2025. This new methodology would allow CMS to incorporate the most recent average FFS risk scores in the calculation without excluding any years of FFS risk scores while making more reasonable projections of what the actual average FFS risk score will be in the payment year.
In CY 2025, for both MA and PACE End-Stage Renal Disease (ESRD) and Non-ESRD risk adjustment models, CMS is proposing to use a multiple linear regression methodology that incorporates historical FFS risk scores from the most current five years of average FFS risk scores and includes a flag that identifies whether a risk score is based on dates of service before 2020 or dates of service starting in 2020 after the onset of the COVID-19 pandemic to calculate the normalization factors. This methodology allows us to represent the historical FFS risk score trend more accurately and calculate a factor without excluding the years impacted by the COVID-19 pandemic. We believe this approach will respond to previous feedback about concerns with excluding years as well as lead to a reasonably accurate prediction of the FFS risk score in 2025.
Like CY 2024, the impact of the update to the normalization factor is not shown in the fact sheet separately because there is considerable interaction between the impact of the MA risk model updates and the normalization factor update. Therefore, the combined impact is shown in the fact sheet.
11. How do the risk adjustment model updates and FFS normalization interact to result in a negative 2.45 percent?
When CMS updates a CMS-HCC risk adjustment model, it may make a variety of updates, including updates to the underlying FFS diagnosis and expenditure data to use more recent years, which allows the model coefficients to reflect more recent costs and coding trends, revisions to the Conditions Categories in the model to improve predictive accuracy of the model, and update the denominator year, which brings the year in which the 1.0 FFS risk score is set to a more recent year.
Both the model updates – the updated data years and any revisions to Condition Categories – and the updated denominator year can result in changes in risk scores. Average MA risk scores may decrease, since the 1.0 is set in FFS, not MA, and plan-level risk scores can change given the clinical profile of each plan’s enrollees.
The denominator update will have a direct impact on the normalization factor, which serves the purpose of keeping the average FFS risk score at 1.0 in each payment year. After a risk model is implemented, there will be a trend in the risk scores calculated with that model over a period of years due to changes in population and coding practice compared to the denominator year. Therefore, for each payment year, CMS calculates a normalization factor for each model to project the average FFS risk score from the denominator year to the payment year. We apply this normalization factor to all risk scores from that model to account for underlying trends in FFS coding and population from the denominator year to the payment year and keep the average FFS risk score at the same average (1.0).
Because updating the denominator year brings the 1.0 year closer to the payment year, there are fewer years of change to account for through normalization, and the normalization factor will typically be smaller than the normalization factor used (in the prior year) with a risk model with an older denominator year. As a result, a model update is intertwined with the normalization factor, and we have included the combined impact in the fact sheet.
To break out the model impacts from the normalization impacts, we provide the change in raw (not normalized) risk scores under the risk adjustment models, along with the impact of the normalization factor. Raw risk scores do not measure actual payment impacts, and a good proportion of the difference in risk scores between models is due to differences in trend. The bottom-line impact of normalization is positive, offsetting the downward impact of the 2024 CMS-HCC model, because the normalization factor for the 2024 CMS-HHC model is lower than the normalization factor for the 2020 CMS-HCC model – and thus, it will reduce risk scores by less. Note that for the 2024 CMS-HCC model, first implemented in 2024, the normalization factor is higher for 2025 than it was for 2024.
If the raw risk adjustment model impact and the impact of the updated normalization factor for 2025 are broken out from one another, the impact of the raw risk adjustment model revisions phase-in is -4.44 percent, and FFS normalization is +1.99 percent. For plans to analyze their specific payment impact, they must both calculate their raw risk score impact of the model revisions with the phase-in and then net it with the FFS normalization impacts with the phase-in.
12. What is the technical adjustment being phased in related to medical education costs?
CMS finalized a technical adjustment to the growth rates related to medical education costs in the CY 2024 Rate Announcement and began phasing in the adjustment starting with CY 2024. Included in the CY 2025 Advance Notice is a proposal to continue phasing in the technical adjustment to the per capita cost calculations that result in a reduction in the growth rates. Historically, prior to the CY 2024 MA rates, the tabulation of per capita costs for FFS individuals had included indirect medical education and graduate medical education costs paid by CMS to inpatient facilities associated with services furnished to MA enrollees because CMS had not been able to separately identify these payments from those made for services furnished to FFS individuals. The baseline modeling supporting the growth rate calculations was updated to enable us to identify the historical and projected costs of medical education paid by CMS for MA admissions. As noted in the CY 2024 Rate Announcement, we initiated a phased-in approach beginning with the CY 2024 rates to remove these medical education costs related to services received by MA enrollees from the historical and projected expenditures supporting the FFS costs that are included in the growth rate calculations. As described in the Fact Sheet, we are proposing to follow the approach we described in the CY 2024 Rate Announcement. We propose to apply 67 percent of the adjustment for MA-related medical education costs in CY 2025.
13. What drives the Medicare Advantage payment Growth Rates?
As required by statute, the growth rates used in the calculation of the MA rates reflect the growth in per capita costs for non-ESRD individuals enrolled in either FFS or Medicare health plans [3] . The growth ratesare based on the expected change in United States Per Capita Costs in Fee-For-Service (FFS USPCC) and in Medicare overall (both FFS and MA) and, as such, are largely driven by trends in per capita costs for individuals in Medicare FFS. The Effective Growth Rate in the Fact Sheet is a national average of expected change in the per capita costs year over year. The main driver of the Effective Growth Rate is the FFS USPCC, with the Total USPCC used to calculate the Pre-ACA benchmark cap amount for each county. We note that the Effective Growth Rate shown in the chart at the beginning of this Fact Sheet includes the continued phase-in of the technical adjustment to the growth rates related to medical education costs described above.
14. Why is the MA risk score trend so important to understanding year-over-year impacts?
The intent of the bottom-line table in the Fact Sheet is to give the public a measure of the all-in impacts of the policies we are proposing regarding MA payment for the following year. While not a policy proposal, the MA risk score trend is a key factor in the level of overall MA payments. The MA risk score trend accounts for the average annual increase in MA risk scores and is driven by MA demographics and diagnosis coding patterns. It represents the estimate, based on historical data, for how risk scores will increase for the next year, which results in higher payments to plans. It is calculated using the risk adjustment model(s) we are proposing for the payment year and is an important piece of information we provide to the public. Like all aspects of the bottom-line table, the risk score trend is an industry-wide average, and thus, individual MA plans may have a different experience. Historically, the risk score trend has steadily increased over time, even in years when CMS implemented updated risk adjustment models. The MA risk score trend is an average estimate of growth, and we have found it to be a reasonable measure of risk score growth.
15. How does CMS calculate the MA risk score trend?
CMS calculates the MA risk score trend by using MA risk scores over three prior years and then calculating the average annual change in risk scores across those three years. For CY 2025, the MA risk score trend was calculated using MA risk scores from 2018 through 2020, before MA risk score data was impacted by the pandemic. Because risk scores are calculated from prior year diagnoses, the 2020 risk scores (based on 2019 diagnoses) were minimally affected by the disruption resulting from the COVID-19 pandemic. Our analysis finds that 2020 risk scores are appropriate to include in the risk score trend.
All three years of risk scores are based on the risk adjustment model(s) to be used in the upcoming payment year. For CY 2025, the MA risk score trend was calculated using MA risk scores from 2018 through 2020. The risk score trend is 3.30 percent under the 2024 CMS-HCC model and 5.00 percent under the 2020 CMS-HCC model. CMS blended the MA risk score trends using the same blend proposed to be used to determine CY 2025 risk scores (i.e., 67 percent of the MA risk score trend under the 2024 CMS-HCC model and 33 percent under the 2020 CMS-HCC model). This blended MA risk score trend for CY 2025 is 3.86 percent.
16. Why are changes to payments for 340B drugs being discussed in the Advance Notice?
Following the Supreme Court’s decision in American Hospital Association v. Becerra on June 15, 2022, and the district court’s remand to the agency, CMS issued the Hospital Outpatient Prospective Payment System Remedy for the 340B-Acquired Drug Payment Policy for Calendar Years 2018-2022 Final Rule, CMS-1793-F, on November 2, 2023. CMS outlined that the agency will make a one-time lump sum payment calculated as the difference between what covered entities were paid for 340B drugs (ASP minus 22.5 percent) and what they would have been paid had the 340B payment policy not been applied (generally ASP plus 6 percent) from 2018 through September 27, 2022. CMS will maintain budget neutrality by reducing non-drug outpatient item and service prospective payments beginning in CY 2026. The remedy payments are reflected in spending for years other than 2025. Therefore, the CY 2025 growth rates are not expected to be impacted.
MEDICARE PART D
17. What proposals does CMS provide for the Part D Benefit in the CY 2025 Advance Notice?
In the CY 2025 Advance Notice, CMS provides information on the Part D benefit parameters that are to be updated or eliminated because of amendments to the Social Security Act made by the Inflation Reduction Act of 2022 (IRA). CMS also provides information on proposed updates to the Medicare Part D (RxHCC) risk adjustment model used to calculate direct subsidy payments for Part D benefits offered by stand-alone Part D prescription drug plans (PDPs) and Medicare Advantage prescription drug plans (MA-PDs). The proposed model is updated to reflect the 2025 benefit changes under the IRA, including the increase in plan liability given the $2,000 cap on annual beneficiary out-of-pocket spending and the new Manufacturer Discount Program. These Part D risk adjustment model updates are essential for plan sponsors in developing accurate bids for 2025.
Part D benefit-related updates under the IRA that will be in place for CY 2025 and that are described in the Advance Notice and related Draft CY 2025 Part D Redesign Program Instructions include the elimination of the coverage gap phase to effect the introduction of a new three-phase benefit (deductible, initial coverage, and catastrophic) and the lowering of the annual out-of-pocket threshold to $2,000. As was instituted in CY 2024, there will be no cost sharing for enrollees in the catastrophic phase, enrollee cost sharing for covered insulin products will be limited, and there will be no enrollee cost sharing for adult vaccines recommended by the Advisory Committee on Immunization Practices and covered under Part D. For more information on these items, see the CY 2025 Advance Notice, Attachment III, and the related Draft CY 2025 Part D Redesign Program Instructions.
18. Has CMS updated the Part D risk adjustment model to reflect the redesign of the Part D benefit?
Yes, the proposals in the CY 2025 Advance Notice include an update to the Part D risk adjustment model that reflects the 2025 Part D benefit design as required by the Inflation Reduction Act. This update reflects changes such as the lowering of the annual out-of-pocket threshold, limits on beneficiary cost sharing for covered insulin products and recommended adult vaccines covered under Part D, the elimination of the coverage gap phase, changes to beneficiary cost sharing and Medicare reinsurance obligations in the catastrophic phase, the sunsetting of the Coverage Gap Discount Program and establishing the new Manufacturer Discount Program.
The proposal in the CY 2025 Advance Notice is to update the Part D risk adjustment model to reflect the expected increase in plan liability associated with the IRA and other changes to the Part D standard benefit. The updated model being proposed uses the most recent available data for model calibration, which we believe best reflects what the patterns in drug spending will be in 2025.
CMS presents predictive ratios for the proposed model in the Advance Notice, and our analysis found that the new model generally predicts well for individuals with the highest predicted risk. While the model does not make assumptions about behavioral changes under the IRA benefit structure, future model calibrations will be able to use more recent data to address any changes that are unknown at this time.
19. Which years of data did CMS use to calibrate the Part D risk adjustment model?
For MA-PDs and standalone PDPs, the proposed Part D risk adjustment model is calibrated on 2021 diagnoses to predict 2022 expenditures. For PACE organizations, the proposed Part D risk adjustment model is calibrated on 2018 diagnoses to predict 2019 expenditures.
20. What change is CMS proposing for Part D normalization?
CMS is proposing to maintain the existing linear slope methodology for calculating Part D model normalization factors—which is to calculate a slope using five years of risk scores and then projecting the slope by the number of years between the denominator year to the payment year—but to do this calculation separately for MA-PD plans and PDPs. Applying separate normalization factors to risk scores used to pay MA-PD plans and PDPs will more accurately reflect Part D costs in each of these two sectors of the Part D market that are driven by a variety of market-based variables, including the overall benefits that they are able to manage, lack of an ability of PDPs to affect the submission of diagnoses in FFS, and available strategies used to manage costs.
21. Why is CMS not proposing the multiple linear regression methodology for Part D normalization?
The risk scores used in the trend to calculate the Part D normalization factors are the five most recent MA-PD risk scores and the five most recent years of PDP risk scores available. For this reason, the availability of risk scores used to calculate RxHCC model normalization factors are lagged one year, relative to CMS-HCC risk scores, meaning that the most recent final RxHCC risk score is for 2022 (using diagnoses from 2021 dates of service).
Because we do not have a 2023 risk score for the RxHCC normalization factor calculation, we cannot yet evaluate the accuracy of the linear regression approach, as we did for the CMS-HCC models, and we do not believe it is prudent at this time to alter the methodology as has been proposed for the CMS-HCC models. If we applied the multiple linear regression approach to the RxHCC model for CY 2025, the “post-COVID” portion of the trend would only reflect the change from 2021 to 2022. We believe that this is not a typical year of risk score growth as it includes the rebound of risk scores after the COVID-19 pandemic when utilization began to increase, and the rebound would disproportionately affect the multiple linear regression approach when this year of growth is used as the sole basis for the “post-COVID” portion of the trend. In future years, when more post-pandemic risk scores are available for RxHCC models, we will evaluate the multiple linear regression approach, but at this time, we believe that using that approach for the RxHCC models could distort the normalization factor.
22. Does the revised Part D risk adjustment model accurately predict cost for individuals with a low- income subsidy?
The proposed Part D risk adjustment model update reflects the changes to plan cost sharing for low-income individuals as part of the IRA Part D benefit design. Risk scores for low-income individuals are expected to increase because plan liability for these individuals are expected to increase more than it does for non-low-income individuals.
23. Does the revised Part D risk adjustment model accurately predict for individuals using high-cost drugs?
Yes, CMS presents predictive ratios for the proposed Part D risk adjustment model in the Advance Notice, and our analysis found that the new model generally predicts well for individuals with the highest predicted risk.
The Part D risk adjustment model is not intended to predict costs for specific drugs. Part D plan sponsors submit bids that account for the projected revenue needed to cover the expected per beneficiary costs of their enrollee population. Risk adjustment is used to adjust these plan bids and payment is based on health status and demographic characteristics such that plans are paid more for individuals predicted to have higher drug costs on average. Plan sponsors are expected to manage the risk of their enrolled populations, and the Part D risk adjustment model is not intended to influence decisions about drugs prescribed to individuals.
24. Will software or risk scores be posted to the CMS website so that plans can evaluate the impact of the new Part D risk adjustment model?
Model software and risk scores for the proposed Part D risk adjustment model will be posted when the Advance Notice is published. Specifically, CMS will post software for the Part D risk adjustment models presented in the Advance Notice and risk scores for three versions of the Part D model for MA-PDs and PDPs: 1) the current Part D model (first implemented for payment in 2023), 2) the proposed 2025 Part D model (2021/2022 calibration), and 3) an alternate calibration of the 2025 Part D model (2018/2019 calibration). CMS will also include risk scores for two versions of the PACE RxHCC model: 1) the current Part D model (first implemented for payment in 2020) and 2) the proposed 2025 Part D model (2018/2019) calibration.
[1] CMS has updated the underlying data in the model many times since the initial implementation of the CMS-HCC risk adjustment model (see the Advance Notices and Rate Announcements for 2007, 2009, 2013, 2014, 2017, and 2019). In addition, in the 2014 Advance Notice, CMS proposed a model (the “2014 model”) with similar updates to those in the 2024 (e.g., data year update, a clinical revision that resulted in newly built HCCs, and updates based on a review a diagnoses codes that showed indication of variable coding and removal of certain codes and associated HCCs). See section G of the 2014 Advance Notice for more information.
[3] “Medicare health plans” include MA plans, Cost plans, PACE plans, and MMP plans.