The Risk Adjustment Data Validation (RADV) final rule,¹ which was announced by the U.S. Department of Health and Human Services on January 30, 2023, radically changes MAO expectations for risk-adjusted funding. This Milliman whitepaper provides an overview of this new regulation and discusses some of its most impactful implications.

The RADV final rule focuses on the mechanism the federal government uses to determine how Medicare Advantage Organizations (MAOs) are paid. Approximately half of all Medicare-eligible beneficiaries, over 28 million people, are enrolled in Medicare Advantage (MA) plans.² The Medicare Advantage Organizations that offer these plans receive revenue from the Centers for Medicare and Medicaid Services (CMS) for each beneficiary enrolled in their plans. These revenue amounts are set based on the expected cost to care for the beneficiaries who enroll, rather than the actual cost of their care. To measure expected cost, CMS developed a risk adjustment system that considers each beneficiary’s medical diagnoses to determine which medical conditions the member has. In broad terms, Statute requires the risk adjustment system to make the same average payment in MA as would be made in traditional Medicare for a beneficiary with the same demographic status and health risks, technically referred to as “actuarial equivalence”.³ To achieve this, CMS assigns a payment relativity factor to each condition (or what CMS refers to as Hierarchical Condition Categories (HCC)) based on the historical cost to care for a beneficiary with that condition under the traditional Medicare fee-for-service (FFS) program.

To ensure that MAOs are only paid for each beneficiary’s actual conditions, CMS established standards for how medical diagnoses should be documented. Regulations require proper documentation for each diagnosis to be included for payment to MAOs via risk adjustment. The RADV final rule codifies how CMS will validate that the diagnoses submitted by MAOs comply with CMS documentation requirements, how CMS will determine the amount of any improper payments, and how CMS will recoup payments that it determines to be improper. Billions of dollars are at stake under the final rule; in the announcement of the final rule, CMS estimated it would collect $4.7 billion in additional recoveries from MAOs between 2023 and 2032.⁴ The specifics around how CMS calibrates the risk score model, conducts RADV audits, and applies (or does not apply) adjustments like the FFS Adjuster directly drive whether or not the equivalence of payments between traditional Medicare and MA is maintained.

Contents of the RADV final rule

The rule finalizes regulation focused primarily on three concepts:

• Should the RADV audit account for any coding errors in the FFS Medicare data used to calibrate the risk adjustment model?
• Can audit results extend beyond the audited diagnoses, and if so, how?
• Which RADV audit years are addressed by these changes?

No FFS adjuster

In its 2012 RADV Methodology⁵ CMS included what is referred to as a “Fee-For-Service (FFS) Adjuster” to adjust for the impact of unsupported diagnoses in the traditional Medicare data used to calibrate the risk adjustment model. In 2018, CMS proposed to remove the FFS Adjuster, and finalized that proposal in this final rule.

Removal of the FFS Adjuster has been the focus of intense debate among stakeholders for both technical reasons and because of the Social Security Act’s references to “actuarial equivalence”⁶ between traditional Medicare and MA in determining risk adjusted payments. A full discussion of this issue is beyond the scope of this white paper; however, we include an illustration in the sidebar that may help the non-actuarial reader envision the concept underlying the original intent of the FFS adjuster and why many stakeholders advocated for its inclusion in the final rule.

In simple terms, actuarial equivalence requires that on average, the same payment is made for two beneficiaries with similar demographics and health status. When creating a model to achieve this actuarial equivalence, it is critical to calculate the incremental cost of treatment for each condition using the same definition of the medical condition as will be used when the model is applied.⁷ The FFS Adjuster was introduced because the FFS data used to calculate the incremental cost of each medical condition is based on diagnoses included in claims submitted for payment, not the contents of medical records, whereas CMS regulations require MAO diagnoses to be documented in medical records. The FFS Adjuster was designed to account for that difference.

The sidebar’s example would require an adjustment for the inclusion of crab apples in the original crate to achieve actuarial equivalence between per apple and per pound pricing. In a similar way, using a FFS Adjuster is one way to achieve actuarial equivalence in risk adjustment within the MA risk adjustment portion of the payment model. Clearly Medicare and the legal issues involved are more complex than our orchard example, but the concept underlying actuarial equivalence is the same.

Audit results will be extrapolated in 2018+

As finalized, CMS will continue to conduct audits by selecting a sample of beneficiaries from MAO contracts and auditing the medical records of those beneficiaries to determine whether any payments were made for unsupported diagnoses. If CMS determines overpayments were made for the sampled beneficiaries, then it will extrapolate the findings to all RADV-eligible beneficiaries in the MAO’s contract. That is, CMS will utilize statistical methods to estimate the total overpayment for the audited contract using the identified overpayment for the audited sample of beneficiaries.

While the proposed rule would have applied extrapolation to audits dating back to plan year 2011, the final rule moderated this in pursuit of a more timely completion of RADV audit appeals. Audits for plan years 2011 through 2017 will not include extrapolation, and extrapolation will generally be applied to audits beginning with plan year 2018. CMS reserved the right not to extrapolate, based on considerations specific to each audit, but noted that this would be limited and that extrapolation will be standard practice going forward.

Any valid statistical method

When CMS applies extrapolation, it will use any statistically valid sampling and extrapolation approach it deems appropriate. CMS stated its intent to improve transparency by documenting and releasing such approaches through the Health Plan Management System (HPMS) or another appropriate means.

Audit focus

CMS announced a shift from random selection of contracts for RADV audits to focusing on contracts and diagnoses at the highest risk for improper payments. That is, the audits may be targeted. CMS cited following U.S. Government Accountability Office (GAO) recommendations to target areas where unsupported diagnoses are most likely to exist.

Moreover, CMS may additionally elect to focus on specific types of beneficiaries (such as beneficiaries with specific HCCs) determined to be at higher risk of overpayment within a contract, rather than review the entire contract. In this case, extrapolation would be limited to the identified population of beneficiaries rather than the entire RADV eligible population.

CMS indicated both types of targeting are intended to increase the efficiency of recovering improper payments, to minimize the burden on MAOs with low improper payment rates, and to maximize recoveries.

OIG audit authority

Throughout the final rule, CMS included both CMS and the HHS Office of the Inspector General (OIG) audits as valid RADV audits, indicating that audits from both organizations would be enforced and extrapolation would apply beginning with plan year 2018 for both sets of audits.⁸ CMS stated that it can collect improper payments identified by the OIG but did not cite a specific authority.

Recovery of improper payments

CMS finalized the intent to collect lump sum payments via the Medicare Advantage/Prescription Drug System (MARx) for RADV audit extrapolated amounts. Such collections will not be tied to individual beneficiaries or submitted diagnoses.

Initial implications

MAOs holding liabilities for prior RADV audits

RADV audits have been ongoing for more than a decade. With the initial rules for the 2011 – 2013 RADV audits including an intention for CMS to extrapolate findings to all RADV eligible beneficiary revenue in a contract, some MAOs began holding liabilities on their financial statements to cover such potential payments.

The finalized rule waives extrapolation for payment years 2011 through 2017, generally enabling MAOs to release liabilities for extrapolated amounts. This may generate one-time favorable financial results for companies holding such liabilities.

OIG audits now have teeth

Readers may have noticed over the years that OIG RADV audits begin with a recommendation from the OIG to CMS to take action to recover funds from MAOs in accordance with the findings of the OIG audit. Prior to the release of this final rule, CMS has generally declined to act on the OIG’s recommendations. The new rule finalizes CMS’s intent for OIG audits to carry the same weight as CMS audits and for such amounts to be recovered from MAOs.

Targeted audits and reduced audit delays

In finalizing the intent and authority for CMS to target sub-populations and potential documentation issues, CMS has created the opportunity to quickly close several audits and reduce the large delay between the close of each contract year and the execution and completion of RADV audits. As a case in point, the OIG has recently issued a large series of narrowly targeted high-risk condition audits. This focused approach may enable CMS to quickly finalize these audits by focusing on the specific claims identified and bypassing MAO concerns with the extrapolation methodology. The federal government would then more quickly recover some amount of dollars for unsupported diagnoses for the targeted years. It may also allow CMS to close some audits, reduce the audit backlog to get more current with RADV audits, and recover some dollars for diagnoses that are not supported on medical records. In the process, this approach may leave other non-targeted subsets of the MA program unaudited.

FFS adjuster

The FFS Adjuster is likely the most impactful part of the final rule. CMS’s primary justification

“… for our decision not to apply an FFS Adjuster is because we believe that the actuarial equivalence provision of the statute applies only to how CMS risk adjusts the payments it makes to MAOs, and not to the obligation to return improper payments for diagnosis codes submitted by MAOs to CMS lacking medical record support.”

Many MAOs view this as a significant change in CMS policy. As noted above, CMS initially recognized this issue in MA payment and the RADV audits and documented one approach to correcting for the issue in CMS’s 2012 RADV Methodology⁹ document:

“The FFS adjuster accounts for the fact that the documentation standard used in RADV audits to determine a contract’s payment error (medical records) is different from the documentation standard used to develop the Part C risk-adjustment model (FFS claims). The actual amount of the adjuster will be calculated by CMS based on a RADV-like review of records submitted to support FFS claims data.“

To restate this, CMS finalized the rule without a FFS Adjuster under the argument that RADV audits are only a payment integrity tool and are not subject to statutory requirements for the Secretary of the U.S. Department of Health and Human Services (HHS) to create a risk-adjustment component of the payment methodology in MA that is actuarially equivalent to FFS.

In the proposed rule, CMS presented a technical analysis¹⁰ regarding the magnitude of the FFS adjuster. Because CMS finalized the rule with the determination that a FFS Adjuster is not appropriate for RADV, CMS did not address comments regarding the validity of the technical analysis. Rather, CMS stated:

“Despite our discussion of the FFS Adjuster study in the proposed rule and efforts to achieve transparency, we are not relying upon the study to reach our conclusion that an FFS Adjuster is not appropriate in the RADV context.”

CMS put forth an argument suggesting that FFS providers do not capture as many diagnoses as MAOs because of the difference in incentives and effort to capture diagnoses and that this difference offsets the unsupported diagnoses in FFS data. However, the MA payment system already includes an MA Coding Pattern adjustment, which Congress required CMS to apply to directly lower MAO risk scores because of the more complete level of coding in the MA program versus FFS. The mechanism to account for MAOs capturing additional diagnoses beyond what is captured in FFS is already accounted for separately.

Neither the final rule nor any other CMS document we are aware of quantified the magnitude of the impact of unsupported diagnoses in FFS data when RADV-like auditing standards are applied to diagnosis data. Various stakeholders have attempted to estimate this impact, but an actual review of all medical records for a sample of FFS beneficiaries is required to accurately estimate the impact, and only CMS has the authority to conduct such a review of FFS beneficiary medical records. The question of how large this impact is remains unanswered. The magnitudes of these individual items are required to understand if actuarial equivalence is maintained in the MA payment system.

Moreover, because there is no broad process to validate diagnoses on physician FFS medical claims, the magnitude of unsupported diagnoses in the FFS data has the potential to exceed the magnitude of unsupported diagnoses in MA diagnosis data, where MAOs take explicit action to support at least a portion of the diagnoses with medical records.

Certain analyses have been put forth as quantifying the magnitude of the impact of the difference in the documentation standards and while a full discussion is beyond the scope of this white paper, the use of claim level FFS audits that do not consider the whole beneficiary, like the Comprehensive Error Rate Testing (CERT) audits, are not able to quantify the impact of a RADV-like review of medical records on FFS data. This gap in information, including uneven documentation standards, an unknown magnitude of the effect of the documentation standard difference, selective enforcement of the documentation standard, and the potential ultimate decisions associated with audit appeals and legal action from MAOs combine to create significant uncertainty in the MA payment system.

Continued uncertainty

The final rule does not resolve some legal and technical criticisms of the proposed rule. Notably, the rule is unlikely to satisfy critics, such as MAO industry organizations, of CMS’s handling of the FFS Adjuster and actuarial equivalence between FFS and MA payments.

RADV has been subject to significant uncertainty over the last 10 years, and although officially final, this rule appears unlikely to settle the issue based on early stakeholder reactions. Potential legal challenges and requests for clarifications from CMS are likely to result in ongoing uncertainty over the upcoming years.

Qualifications

Rob Pipich and Michael Rothschild certify we are Members of the Academy of Actuaries and meet the Academy's qualification standards for this type of analysis.

This white paper is intended as an update regarding the final rule, not as a comprehensive summary of the history and issues involved. Readers are assumed to have a working knowledge of the complex issues involved. Milliman does not intend to benefit or create a legal duty to any reader and readers should consult their own qualified experts.

The authors would like to thank Jason Karcher and Michael Polakowski for their contributions and peer review.

¹The full text of the final rule is available at https://www.govinfo.gov/content/pkg/FR-2023-02-01/pdf/2023-01942.pdf

²KFF (Aug 25, 2022). Medicare Advantage in 2022: Enrollment Update and Key Trends. Retrieved Feb 7, 2023, from https://www.kff.org/medicare/issue-brief/medicare-advantage-in-2022-enrollment-update-and-key-trends/

³Section 1853(a)(1)(C) of the Social Security Act

⁴Table 3 in the final rule estimates $4.7 billion in additional recoveries between 2023 and 2032.

⁵CMS (February 24, 2012). Notice of Final Payment Error Calculation Methodology for Part C Medicare Advantage Risk Adjustment Data Validation Contract-Level Audits. Retrieved February 3, 2023, from https://www.cms.gov/Research-Statistics-Data-and-Systems/Monitoring-Programs/recovery-audit-program-parts-c-and-d/Other-Content-Types/RADV-Docs/RADV-Methodology.pdf.

⁶The actuarial equivalence requirement of Section 1853(a)(1)(C) of the Social Security Act requires risk-adjusted payments for benefits to be aligned with the cost to provide those benefits to that class of individuals under FFS Medicare. Use of the phrase “actuarial” typically implies that a member of the American Academy of Actuaries should opine on the equivalence and follow Actuarial Standards of Practice as promulgated by the American Academy of Actuaries.

⁷Actuarial Standard of Practice (ASOP) #45 The Use of Health Status Based Risk Adjustment Methodologies identifies specific considerations to be used in risk adjustment modeling and includes a variety of considerations around the consistency of how a risk adjustment model is developed and applied. The full text of ASOP 45 is available at http://www.actuarialstandardsboard.org/asops/use-health-status-based-risk-adjustment-methodologies/.

⁸Many OIG RADV audits for 2017 and prior incorporated extrapolation, but only the identified overpayment for audited members will be collected for these audits

⁹CMS (February 24, 2012). Notice of Final Payment Error Calculation Methodology for Part C Medicare Advantage Risk Adjustment Data Validation Contract-Level Audits. Retrieved February 3, 2023, from https://www.cms.gov/Research-Statistics-Data-and-Systems/Monitoring-Programs/recovery-audit-program-parts-c-and-d/Other-Content-Types/RADV-Docs/RADV-Methodology.pdf.

¹⁰CMS (October 26, 2018) Fee for Service Adjuster and Payment Recovery for Contract Level Risk Adjustment Data Validation Audits. Retrieved February 8, 2023, from https://www.cms.gov/Research-Statistics-Data-and-Systems/Monitoring-Programs/Medicare-Risk-Adjustment-Data-Validation-Program/Other-Content-Types/RADV-Docs/FFS-Adjuster-Excecutive-Summary.pdf