Medicare Advantage (MA): Proposed Benchmark Update and Other Adjustments for CY2027 in Brief

Contents

• Introduction
• Determining Payments to Plans
• Benchmark Calculations
• Risk Adjustment
• Summary of Selected Benchmark Changes and Other Adjustments in the Advance Notice
• Proposed Benchmark Updates and Changes
• Proposed Updates and Changes to Risk Adjustment
• How Would These Changes Affect a Specific Congressional District?

Introduction

Medicare Advantage (Part C, or MA) is an alternative way for Medicare beneficiaries to receive covered benefits. Under MA, private health plans are paid a per person monthly amount to provide all Medicare-covered benefits (except hospice) to beneficiaries who enroll in their plans. Unlike under original Medicare,¹ where providers are paid for each item or service provided to a beneficiary, the same capitated monthly payment is made to an MA plan regardless of how many or how few services an enrollee actually uses. The plan is at risk if costs for all of its enrollees exceed program payments and beneficiary cost sharing; conversely, in general, the plan can retain savings if aggregate enrollee costs are less than program payments and cost sharing.² In 2024, 54% of Medicare beneficiaries chose to enroll in an MA plan.³

Capitated payments to plans are determined, in part, based on a benchmark, or maximum payment. Benchmarks are updated annually by a measure of Medicare spending growth and by other adjustments. The Secretary of the Department of Health and Human Services (hereinafter, the Secretary) published the Advance Notice of Methodological Changes for Calendar Year 2027 (Advance Notice) MA capitation rates on January 26, 2026,⁴ which provided preliminary estimates of the measures of spending growth used to update MA benchmarks, as well as other adjustments and proposals for updating benchmark rates. The Secretary estimated the change in revenue resulting from the proposed policies in the Advance Notice would increase plan payments by 0.09% before accounting for risk score coding trends. After accounting for estimated growth in plan risk scores, the Secretary projected average plan payments would grow 2.54% relative to payments in 2026.⁵ The final CY2027 benchmarks are expected to be published no later than April 6, 2026.

This report briefly describes how MA payments are determined through a comparison of a plan's estimated cost (bid) and the maximum amount Medicare will pay a plan (benchmark). The report then discusses the benchmark calculation, most recently amended by the Patient Protection and Affordable Care Act (ACA; P.L. 111-148, as amended) and related administrative action. It also describes the role of risk adjustment in determining payments. The report then summarizes selected provisions in the CY2027 Advance Notice that would adjust the benchmark, modify the risk adjustment model, or make other changes, some of which are specified statutorily and some of which are at the Secretary's discretion.

Determining Payments to Plans

As stated above, MA plans are paid a per person monthly amount. The Secretary determines a plan's payment by comparing its bid to a benchmark. A bid is the plan's estimated cost of providing Medicare-covered services (excluding hospice but including the cost of medical services, administration, and profit). In general, the Secretary has authority to review and negotiate plan bids to ensure they reflect revenue requirements.

A benchmark is the maximum amount the Center for Medicare and Medicaid services (CMS) will pay an MA plan to provide Medicare services in the plan's service area.⁶ Plan payments depend on how the bid compares to the benchmark. If a plan's bid is less than the benchmark, the plan's base payment equals its bid plus a rebate. The rebate represents a portion of the difference between the bid and the benchmark. The rebate must be returned to enrollees in the form of additional benefits, reduced cost sharing, reduced Medicare Part B or Part D premiums, or some combination of these options. Since CY2012, the size of the rebate has depended on plan quality; rebates range from 50% to 70% of the difference between the bid and the benchmark.⁷ If a plan's bid is equal to or above the benchmark, its payment equals the benchmark amount. In that scenario, each enrollee in that plan would pay an additional premium that is equal to the amount by which the bid exceeds the benchmark.⁸ After the base payment amount is determined through the comparison of the bid and benchmark, payments to plans are risk adjusted to account for the demographic characteristics and health history of those who enroll in the plan.

Many of the proposed changes for CY2027 in the Advance Notice relate to the benchmark—the maximum possible payment. Any change in an MA benchmark could indirectly affect plan payments, because the benchmark is used in conjunction with the bid to determine MA plan payments. For example, if an MA benchmark decreased from one year to the next and the plan bid the benchmark in each year, the plan payment would decrease. However, if in the same scenario a plan bid below the benchmark in each year, the plan payment (the bid plus the rebate) likely would be reduced but could remain the same or increase, depending on the size of the benchmark reduction and the size of the change in the plan bid in each year (e.g., the plan's bid is higher in the second year than in the first). If an MA benchmark decreased from one year to the next but the plan bid above the benchmark each year, the Medicare payment to the plan would decrease because the payment equals the benchmark amount when the plan bids above the benchmark. However, the total payment to the plan (the benchmark plus an additional premium from each enrollee) could increase, decrease, or remain the same, depending on how the plan's bid changes each year. If a benchmark increased from one year to the next and the plan bid below the benchmark, in most cases the plan payment also would increase and would decrease only if a plan bid substantially less in the second year. So, although proposed benchmark changes affect the maximum possible payment from CMS, benchmark changes alone do not determine changes in payments.

Some of the proposed changes for CY2027 refer to changes in risk adjustment. After the plan payment is determined through comparison of the bid and the benchmark, the payment is risk adjusted to account for the health history and demographic characteristics of the beneficiaries who actually enroll in a plan. Any changes to the risk-adjustment methodology, therefore, affect plan payments (because the risk-adjustment factor is multiplied by the non-risk-adjusted payment).

Benchmark Calculations

Separate benchmarks are calculated for each county. Although the methodology for calculating the benchmarks is applied consistently across counties, benchmark levels vary across counties. The level of the benchmark in any particular county can be affected by the practice of medicine in original fee-for-service (FFS) Medicare and how that affects spending in original Medicare in the county relative to other areas of the country. This section of the report discusses the calculation of benchmarks,⁹ as well as administrative action affecting benchmarks.

MA county benchmarks are set using a percentage of estimated FFS spending in each county. To project per capita FFS spending in each county for the upcoming calendar year, the Secretary first calculates historic spending data from original Medicare claims files and then estimates a trend to determine the projected growth (or percentage increase) in national FFS Medicare per capita spending (also known as growth in fee-for-service United States per capita costs, or FFS USPCC). The growth in FFS USPCC for 2027 is estimated to equal 5.10%. This figure is calculated as the percentage increase between the prior projected national FFS USPCC ($1,230.52 in CY2026) and the current projected FFS USPCC ($1,293.23 in CY2027): 5.10% = ($1,293.23 - $1,230.52) / $1,230.52 × 100.

To calculate per capita spending for each county, the national estimated level of FFS per capita cost ($1,293.23 for CY2027) is multiplied by a county-level geographic index (the average geographic adjustment, or AGA) to determine the relative difference in the estimated FFS per capita spending in each county. The AGA is calculated using a five-year rolling average of claims data for beneficiaries in original Medicare in each county and includes weighting for enrollment and average risk scores.

In addition, several adjustments are made to the county per capita FFS estimates, which are either specified in statute or made at the Secretary's discretion, to more accurately reflect estimated spending for the year in question. These adjustments are discussed in more detail in the "Summary of Selected Benchmark Changes and Other Adjustments in the Advance Notice" section of this report.

Two adjustments are then applied to the per capita FFS estimates of spending for each county for the benchmark calculation. First, FFS estimates for each county are multiplied by a percentage specified in statute—95%, 100%, 107.5%, or 115%—with higher percentages applied to counties with the lowest FFS spending,¹⁰ while the counties with highest FFS spending receive the lowest percentage. In other words, the 25% of counties with the lowest FFS spending will receive the highest percentage (115%) of per capita FFS as their MA benchmark. The 25% of counties with the highest FFS spending will receive the lowest percentage (95%) of per capita FFS.

Second, for the benchmark calculation, per capita FFS estimates of spending for each county after being adjusted as above are adjusted additionally by plan quality ratings.¹¹ Starting in CY2012, plans with at least a 4-star rating on a 5-star quality-rating scale established by CMS have received an increase in their benchmark.¹² In CY2027, a plan receiving 4, 4.5, or 5 stars on a 5-star quality-rating system may receive a 5-percentage-point increase in its benchmark. This means a plan that otherwise might have had a benchmark of [100% × per capita FFS] could receive a benchmark set at [105% × per capita FFS] if the plan had a star quality rating of 4 or more stars. The benchmark quality increases are doubled for qualifying plans in a qualifying county.¹³

The ACA also requires that benchmarks (including any quality adjustment) be capped at the level they would have been in the absence of the ACA. In CY2026, in one-third of U.S. counties, the MA benchmark adjusted by a 5-percentage-point quality bonus is constrained by the pre-ACA benchmark cap. In some cases, this means the quality bonus for plans with 4 or more stars may be less than 5 percentage points (or possibly no increase at all). In other cases, the benchmark for plans with fewer than 4 stars (or 0-percentage-point quality adjustment) also may be constrained by the pre-ACA benchmark levels. The payment cap is a statutory provision,¹⁴ and the Secretary indicated in the Advance Notice that the provision would remain in effect for CY2027.

Risk Adjustment

After the base payment amount for an MA plan is calculated, payments are risk adjusted to compensate plans for the relatively higher medical costs associated with enrollees who are older or sicker and the relatively lower medical costs associated with enrollees who are younger or relatively healthy. The size of the payment adjustment is based on a mathematical model that predicts the relative effect of specified diagnosis groupings and demographic factors in a base year on subsequent health care spending in the following year.¹⁵ Several different risk adjustment models have been used under MA, with each successive model increasing in complexity and explanatory power. The current model is hierarchical (accounting for more significant manifestations of diseases), additive (summing effects across unrelated disease states), and interactive (accounting for situations where having two specific diseases results in health care expenditures that are greater than their sum). This model is referred to as the CMS Hierarchical Condition Category (CMS-HCC) model. The model has different segments used to estimate risk based on an enrollee's entitlement to Medicare (aged or disabled) and whether an enrollee is also eligible for Medicaid (as a full-benefit dual eligible or a partial-benefit dual eligible) or is not eligible for Medicaid at all. The model is created with claims data from beneficiaries in original Medicare.

To illustrate how the model is used to adjust payments to MA plans, consider an MA enrollee who is female, age 67, living in the community, and does not have Medicaid. Consider also that she had a hip fracture the prior year (HCC402), and she was also diagnosed with breast cancer (HCC23), and mild dementia (HCC127). For this person the CMS HCC model predicts the relative effect of her demographic characteristics (0.318), plus the relative effect of her breast cancer (0.184), her hip fracture (0.455), and mild dementia (0.345). In this example, because her cancer, hip fracture, and dementia are unrelated, the estimates from the model can be summed for a total of (1.302). The payment to the plan would be the (risk-neutral) MA payment (for example, $1,200 per person per month) multiplied by the sum of the relative risks produced by the risk model (1.302) resulting in a risk adjusted monthly payment of $1,562.40 (or $1,200 x 1.302 = $1,562.40).

While CMS regularly updates the CMS-HCC model with newer data, it occasionally undertakes a more comprehensive reexamination. This last occurred for CY2024 with model version 28.¹⁶ CMS updated the model to include more recent data, account for the transition from the International Classification of Diseases (ICD) 9 to ICD-10 diagnostic codes, changed condition categories used for payments (including additions, deletions, and revisions), and removed diagnosis codes that were disproportionally used by MA plans compared with original Medicare. CMS indicated that the updated model improves or maintains predictive accuracy for subgroups of Medicare beneficiaries relative to the prior model. In general, changes to the model were expected to reduce MA plan payments, though CMS estimated a positive average change in revenue for CY2024 overall.¹⁷ The update was phased in over three years and is completely phased in for CY2026.

Summary of Selected Benchmark Changes and Other Adjustments in the Advance Notice

The Advance Notice contains estimated values for some of the factors that update the MA benchmarks, as well as the Secretary's proposed methodological changes to the benchmarks and risk adjustment. This section describes a selection of these factors and proposed changes. It divides the provisions into those that adjust the benchmark and those that pertain to risk adjustment.

Proposed Benchmark Updates and Changes

• Growth in the Fee-for-Service United States Per Capita Cost: The FFS USPCC is a measure of the growth in original Medicare spending used to calculate national per capita FFS spending, which is part of the benchmark calculation. For CY2027, the value is preliminarily estimated to increase by 5.10% relative to the CY2026 FFS USPCC.
• National Per Capita MA Growth Percentage (NPCMAGP): The NPCMAGP is a measure of the overall growth in Medicare spending. It applies to the calculation of benchmarks for plans under the Program of All-Inclusive Care for the Elderly (PACE), which are not subject to the ACA methodology.¹⁸ It also applies to pre-ACA benchmarks, which are the caps for MA benchmarks. For CY2027, the value is preliminarily estimated at a 4.04% increase to the previous year's (pre-ACA) benchmark.
• Phaseout of Indirect Medical Education (IME):¹⁹ A provision in the Medicare Improvements for Patients and Providers Act of 2008 (P.L. 110-275) required the Secretary to phase out the value of IME from the MA benchmarks.²⁰ Prior to CY2008, the value of IME payments to hospitals was included in the calculation of MA benchmarks. At the same time, an IME payment also was made from CMS to eligible teaching hospitals when an MA enrollee was admitted. Effectively, CMS was adjusting for IME twice—once directly to the MA plans through an adjustment to the MA benchmark, and once directly to the teaching hospital. The statutory phaseout removes the IME component from MA benchmarks. This adjustment will affect benchmarks differently depending on the value of IME that is to be phased out, but the reduction will not be greater than 10.8% of the per capita FFS rate in a county.
• Kidney Acquisition Cost Carve-Out: Since CY2021, all Medicare beneficiaries with End-Stage Renal Disease (ESRD) have been allowed to enroll in an MA plan.²¹ This change was specified in the 21^st Century Cures Act (Cures Act; P.L. 114-255). The Cures Act further specified that original Medicare is to pay for the cost of kidney acquisition for MA enrollees receiving a kidney transplant. It also directed the Secretary to remove the estimated cost of kidney acquisition from MA benchmarks. This adjustment will affect benchmarks differently based on variation in related costs and kidney transplants.
• New Data for FFS Estimates: County-level estimates of per capita FFS spending used in benchmark calculations are based on a five-year rolling average of original Medicare claims data used to calculate the average geographic adjustment (AGA). For CY2027, the Secretary is to rebase, or update, the claims data used to calculate the AGA by dropping the CY2019 data from the five-year rolling average calculation and adding one additional year CY2024 claims data. Thus, for CY2027, the AGA is to be based on claims data from CY2020 to CY2024. This change may increase benchmarks in some counties and decrease them in others.
• Adjustment to County FFS Estimates to Reflect Current Prices: County-level per capita FFS estimates are calculated using historical claims data, which account for the prices and quantities of items and services used. Starting in 2014, the Secretary has been taking into account current payment policies and applying these policies to the historical claims data upon which the FFS estimates are based to better reflect expected expenditures under current program rules. Since then, the practice of adjusting historical data has expanded to include current payment policy adjustments related to hospital inpatient and outpatient services, skilled nursing facilities, home health, physician services, disproportionate share hospital payments, shared savings payments and losses (or other methods of payment) under specified Center for Medicare & Medicaid Innovation models and demonstrations, and physician fee schedule bonuses for physicians working in Health Professional Shortage Areas. For 2027, the Secretary proposes to include an adjustment for the facility payment received by Rural Emergency Hospitals (REH). REHs are small (not more than 50 beds), outpatient-only providers that furnish emergency services and meet statutory and regulatory requirements. Medicare REH payments include a facility payment that is not otherwise captured in underlying claims data. The adjustment is expected to increase benchmarks in some counties and decrease them in others.

Proposed Updates and Changes to Risk Adjustment

• Coding Intensity Adjustment: In general, MA plan payments are risk adjusted to account for variation in the cost of care. Risk adjustment is designed to compensate plans for the increased cost of treating older and sicker beneficiaries and thus to discourage plans from preferential enrollment of healthier individuals. In part because MA plan payments are affected by enrollee diagnoses, MA plans tend to identify more diagnoses for a given patient than providers in original Medicare, some of whom are paid not by diagnosis but by unit of work. The Deficit Reduction Act of 2005 (P.L. 109-171) required the Secretary to adjust for patterns of diagnosis coding differences between MA plans and providers under Medicare Parts A and B for plan payments in CY2008, CY2009, and CY2010. The ACA required the Secretary to conduct further analyses on the differences in coding patterns and to adjust for those differences after CY2010. It specified minimum coding intensity adjustments starting in CY2014 that apply until the Secretary implements risk adjustment using MA diagnostic, cost, and use data.²² For CY2027, the coding intensity adjustment is a reduction of 5.90%, which is the statutory minimum applied to MA enrollee risk scores. This adjustment is the same amount as the one applied in CY2026.
• Proposed Risk Adjustment Model Update: When creating the CMS-HCC risk adjustment model, the estimates of relative cost for the different condition categories sometimes are constrained to be equal across model segments or across condition categories; this might be the case, for example, if there are insufficient data to make the calculation otherwise. For CY2027, the Secretary proposes removing the constraint previously applied to two measures of kidney disease (HCC 328 Chronic Kidney Disease, Moderate [Stage 3B], and HCC 329 Chronic Kidney Disease, Moderate [Stage 3, except 3B]). The Secretary has determined that there are now sufficient data to reflect the distinctive relative costs of those condition categories.
  Additionally, the Secretary proposes updating the model with more recent diagnosis data (from CY2023 rather than CY2018) and more recent expenditure data (from CY2024 rather than CY2019).
  Finally, the Secretary also proposes excluding diagnosis codes from audio-only services from the risk adjustment model. Diagnosis codes are included in the CMS-HCC risk adjustment model when they are identified during eligible face-to-face encounters. An eligible face-to-face encounter includes real-time, interactive audio and video telecommunications; it does not include audio-only communication. Recent changes to the Current Procedural Terminology (CPT) code set allow identification of real-time, interactive audio-only communications. The Secretary proposes eliminating diagnosis codes collected through audio-only communications from the risk adjustment model to align the CMS-HCC model with a proposal to exclude diagnoses from audio-only services from use in risk-adjusting plan payments, as explained in more detail below. This proposal may increase or decrease the risk scores of individual plans. Due to the interaction between the risk model changes described here and risk model normalization (described below), CMS published a single estimate for the two components, which is an average expected reduction in revenue of 3.32%.
• Risk Model Normalization: As discussed above, CMS uses a model to determine how different demographic characteristics and diagnoses affect the relative cost of enrollees for the purpose of risk adjusting MA payments. When CMS calibrates the risk adjustment model, it does so for a specific set of FFS data and a specific total expenditure in a particular year, and it standardizes the model so that a beneficiary with average Medicare spending has a risk score of 1.0. (A beneficiary who is older and sicker than average, and thus has higher-than-average health spending, would have a risk score greater than 1.0; a beneficiary who is younger and healthier than average, and thus has lower-than-average health spending, would have a risk score of less than 1.0.)
  In years when the model is not recalibrated, it has to be normalized to account for population and coding pattern changes since the calibration year. For example, if the population and coding pattern changes had resulted in a 3% increase in risk codes since the calibration year, then the plans would be overpaid by 3% relative to a normalized population and spending level if CMS did not normalize the model. If the normalization factor was 1.03, then the risk score for each beneficiary would be divided by 1.03; a beneficiary with a risk score of 1.2 would have a normalized risk score of 1.165 (or 1.2 / 1.03 = 1.165). So the beneficiary would still be above average risk (1.0), but not quite as high as the raw score suggested (1.2). For CY2027, to calculate the normalization factor, the Secretary proposes to continue using the same methodology as has been employed in the prior two years. The proposed normalization factor for the CMS-HCC model is 1.058. This proposal is expected to decrease risk scores, which are multiplied by plan payments.
• Source of Diagnoses for Risk Score Calculation: Currently, MA organizations submit all diagnosis codes to CMS when they submit medical encounter data into the Encounter Data System (EDS). Some diagnosis codes submitted to the EDS are used to adjust plan payments, because they predict future health care spending; other codes are not used for plan payment because they are for short-term maladies or do not otherwise predict future expenditures. CMS filters the incoming EDS data to identify diagnoses eligible to be used for risk adjusting the plan payment. Starting in CY2020, CMS allowed MA organizations to submit diagnoses for risk adjustment of payments if the visit met all specified requirements, including being from a face-to-face encounter, which included telehealth visits with real-time, interactive audio and video. Audio-only encounters were to be marked with a modifier code indicating they were audio-only and thus were not eligible for use in risk adjustment of the payment. Similarly, chart review records (CRRs) for audio-only services also were required to include the modifier code; CRRs allow an MA organization to add additional diagnosis codes or delete diagnosis codes previously reported for an encounter. Starting in CY2027, consistent with the change discussed above which eliminates diagnoses from audio-only services from the CMS-HHS model, the Secretary proposes excluding those diagnoses obtained from audio-only encounters from use in risk-adjusting payments and from CRRs identified by the specified modifiers when no other diagnosis on the record is risk-adjustment eligible.
  Additionally, the Secretary proposes excluding diagnoses found on unlinked CRRs from the risk score used to calculate plan payment. CRRs may be submitted to the EDS either linked to an existing encounter data record—meaning the new diagnosis is associated with a prior item or service—or unlinked, meaning the diagnosis is not associated with any particular item or service. The Advance Notice indicates that research by CMS and the Department of Health and Human Services, Office of Inspector General, found that in 2022, few enrollees had unlinked chart reviews (0.05%), but those enrollees were spread over approximately 58% of all MA contracts.²³ The Advance Notice also refers to Medicare Payment Advisory Commission (MedPac) research indicating that between 2020 and 2023, about half of the higher propensity of MA organizations to identify diagnosis codes (relative to original Medicare) could result from diagnoses in chart reviews and health risk assessments.²⁴ However, the MedPac research did not distinguish between linked and unlinked CRRs. The proposals to change the source of diagnosis codes for risk score calculation are expected to result in an average reduction in MA payments of 1.53%.

How Would These Changes Affect a Specific Congressional District?

The final benchmarks for CY2027 are expected to be published no later than April 6, 2026. CMS does not provide estimated benchmarks with the Advance Notice. Until those benchmarks are released, it would be difficult to estimate district-level effects.

Several factors contribute to this uncertainty. For example, the measure of growth estimated in the Advance Notice may change in the final announcement and some of the proposed adjustments might not be included in the final announcement. Additionally, some of the adjustments proposed in the Advance Notice would change the benchmarks in different areas by different amounts. In other words, it would not be informative to multiply the 2026 per capita FFS spending data for each county by the growth in the FFS USPCC (for 2027 estimated to grow by 5.10%), because that national measure of growth would not incorporate the additional proposed changes to the average geographic adjustment factor, which will not be published until the final announcement. The average geographic adjustment takes into account that the growth of spending in FFS Medicare is faster in some counties and slower in others. Further, the effect of the changes proposed in the Advance Notice would depend, in part, on various factors related to plan behavior. For example, the effect of proposed changes could depend on a plan's star quality rating, which can change from year to year, or its diagnosis coding practices, which are not publicly available. For these reasons, the district-level impact of the proposed CY2027 changes cannot be estimated until CMS releases the final benchmark values.

Footnotes