The Centers for Medicare and Medicaid Services (CMS) issued a final rule on the use of extrapolation to determine overpayments in risk adjustment data validation (RADV) audits of Medicare Advantage organizations (MAOs) and for audits conducted by the Office of Inspector General (OIG), effective for payment year 2018. In doing so, CMS rejected application of a fee-for-service adjuster (FFS Adjuster) to allow for permissible threshold errors emanating from third-party provider medical record documentation.
The proposed rule was published on November 1, 2018, and the final rule was published on February 1, 2023.
It is expected that upcoming RADV audits will apply these refined audit standards at the same time that CMS will implement its 2024 capitation rates for MAOs. CMS published the calendar year 2023 Advance Notice on February 1, 2023, proposing coding changes from ICD-9 to ICD-10, especially for chronic conditions. CMS will announce the Medicare Advantage capitation rates and final payment policies for contract year 2024 by April 3, 2023. Because coding accuracy is at the heart of RADV audit exposure and the success of MAO and provider risk sharing arrangements, these two developments warrant significant attention.
MAO RADV audits assess whether the medical information and diagnoses submitted for risk adjustment purposes are fully supported by the patient’s medical record. Reliance on provider medical record documentation practices has an inherent risk of some level of incompleteness, inconsistency, and error that makes extrapolation of audit results legally and equitably inappropriate. The FFS adjuster operates to offset any impact of erroneous diagnosis codes in Medicare Part A and B data used to calculate an allowable level of payment error for risk adjustment modeling.
Debate surrounding audit extrapolation and the FFS Adjuster related to RADV audits has been ongoing for more than a decade. Over the years, CMS has changed its position on the necessity of the FFS Adjuster. Now having settled on not applying the FFS Adjuster, CMS will extrapolate audit findings to which MAOs are subject beginning with payment year 2018, estimating potential recoveries of $4.7 billion by 2032.
We previously discussed in a Health Law Scan blog post the full history related to extrapolation and the FFS Adjuster in the RADV audit program. In short, the risk adjustment model is based on diagnoses and program costs for individuals in traditional Medicare (i.e., FFS claims). CMS uses this claims data to calibrate the risk adjustment model and calculate risk factors, but the data is subject to minimal auditing. This means there are undocumented errors in the FFS data.
CMS first announced its intention to use extrapolated payment error rates in the RADV audit program in 2008. Commenters argued that by applying an extrapolated payment error rate to an MAO’s entire contract, CMS would effectively have audited all of the MAO’s data. Commenters further argued that, without an adjuster, there would be an actuarial difference in documentation standards between the MAO payments (based on unaudited FFS data) and the RADV standard (based on audited medical records). In other words, it would not be actuarially equivalent to, on the one hand, make payments based on unaudited data and, on the other hand, recoup overpayments based on audited data.
In 2012, CMS proposed to use an FFS Adjuster for extrapolated RADV audits to calculate a permissible level of payment error to offset the preliminary recovery amount. In 2018, CMS concluded that the errors in FFS claims data do not lead to systematic payment errors to MAOs. The subsequent proposed rule allowed for extrapolated RADV audits without the use of an FFS Adjuster. Commenters objected to the study’s methodology and the final rule was delayed multiple times.
The absence of a definitive rule from CMS also caused complications for courts in interpreting the Overpayment Rule, which requires MAOs to refund identified overpayments within 60 days, including unsupported diagnosis codes. Most recently, the US Court of Appeals for the District of Columbia Circuit upheld the Overpayment Rule in UnitedHealthcare Ins. Co. v. Becerra, concluding that “CMS has preliminarily decided not to use an FFS Adjuster for contract-level RADV audits after all.” The DC Circuit indicated that CMS did not have any obligation to consider the FFS Adjuster in the overpayment-refund context because RADV audits were an error-correction mechanism that was materially distinct from the Overpayment Rule.
In embracing audit extrapolation conceptually to address overpayments identified in RADV audits, CMS relied on its long-standing acceptance of flexible extrapolation approaches and judicial decisions that have permitted extrapolation in different circumstances. Notably, CMS declined to identify a particular methodology and reserved the discretion not to extrapolate errors if warranted by the circumstances. CMS also decided not to extrapolate RADV or OIG audits from 2011 to 2017 as it had previously proposed. Rather, CMS limited the extrapolation to RADV and OIG audits from 2018 to the present, noting that 2018 rather than 2011 is the most equitable starting line for the final rule. The use of an extrapolation remedy is controversial, especially where CMS also rejected application of the FFS Adjuster to account for a permissible threshold of errors related to provider medical record documentation.
CMS articulated two reasons for not applying an FFS Adjuster: (1) actuarial equivalence does not apply to MAOs’ obligation to return improper payments (citing to UnitedHealthcare); and (2) the Social Security Act cannot be consistently read to require both an FFS Adjuster and the coding pattern adjustment.
First, CMS acknowledged that the DC Circuit declined to address the RADV audit context in its decision in UnitedHealthcare. Nevertheless, CMS described its position as consistent with the DC Circuit’s conclusion that the actuarial equivalence requirement is not an ‘‘entitle[ment] . . . to a precise payment amount’’ for an MAO. According to the final rule, UnitedHealthcare held that the actuarial equivalence requirement applies only to how CMS risk adjusts the payments it makes to MAOs and not to the obligation of MAOs to return improper payments. CMS stated that this reasoning applies just as strongly in the RADV context and supports the agency’s decision not to apply an FFS Adjuster. In other words, CMS did not deny an actuarial inequivalence between the documentation standards; rather, the agency concluded that the RADV audit program is not subject to actuarial equivalence requirements.
However, CMS also characterized the FFS Adjuster as requiring the agency to “either adjust payment rates (by raising them) or adjust documentation standards (by loosening them) to resolve the alleged incompatibility between the payment rates and documentation standards.” CMS did not address why the actuarial equivalence standard would not apply to the implementation of an FFS Adjuster that affects payments CMS makes to MAOs. According to CMS’s interpretation of UnitedHealthcare, payments to MAOs are subject to actuarial equivalence and the FFS Adjuster may raise such payments. Therefore, by its own characterization, CMS cast the FFS Adjuster as affecting payments subject to actuarial equivalence.
Second, CMS pointed out that the Social Security Act establishes a “coding pattern adjustment” requiring that adjustments to MAO payments reflect differences in coding patterns between Medicare Advantage and FFS. According to the legislative history, the adjustment was intended to discourage upcoding. In managed care, reimbursement is tied to a beneficiary’s health status; therefore, there is an incentive to “upcode” or submit diagnosis codes inflating or exaggerating a beneficiary’s illness in order to receive greater reimbursement. Congress intended to guard against upcoding by comparing risk scores for similar populations in Medicare Advantage and FFS and implementing a downward adjustment percentage to minimize the effect of upcoding. However, the coding pattern adjustment does not audit FFS claims data to assess when a provider failed to submit a diagnosis code (i.e., downcoding).
According to CMS’s final rule, the Social Security Act cannot be consistently read to require a reduction in payment under the coding pattern adjustment and also require an offset to recovery audit amounts under the FFS Adjuster. However, it appears that the two are not mutually exclusive as the FFS Adjuster was, at least in part, meant to account for instances when diagnosis codes were not provided in FFS claims data and the coding pattern adjustment considers coding consistencies when diagnosis codes were provided in both FFS and Medicare Advantage. As such, an FFS Adjuster wouldn’t necessarily conflict with the coding pattern adjustment. Put simply, the coding pattern adjustment accounts for upcoding and the FFS Adjuster was meant to account for downcoding.
CMS emphasized that it is not relying on its 2018 FFS Adjuster study in the proposed rule. According to the final rule, the “theoretical study” had natural limitations based on the use of certain assumptions, estimations, and projections. Nevertheless, CMS disagreed with commenters who claim that the 2018 study or other counter-studies provide evidence that FFS errors systematically reduce payments to MAOs. CMS argued that the magnitude of over-coding (diagnosis codes unsupported by medical records) in the FFS data is much smaller than some commenters have suggested and cites to its disowned 2018 study, which found an average error rate of 3%.
CMS also argued that FFS data contains under-coding (supported codes that were not captured for risk adjustment) which offsets the effects of over-coding. Despite these arguments, CMS essentially forgoes the attempt to demonstrate the impact of potential error in FFS diagnoses data on Medicare Advantage, stating “even if systematic error exists, it would be inequitable to correct such errors in the payments made only to audited plans through the application of an FFS Adjuster.”
As a final consideration, CMS noted that non-extrapolated recoveries average $8.2 million per payment year and would result in the agency losing money because the cost of administering the RADV audit program is $51 million per payment year. Although there were alternatives to extrapolation, CMS rejected them because they would result in negative net recoveries for the agency.
While extrapolation is new in the RADV context, its use in the traditional Medicare fee-for-service program may shed some light on what MAOs can expect in RADV audits from 2018 going forward. Under Medicare Part A and Part B, Medicare Administrative Contractors (MACs) typically authorize payment of claims immediately upon receipt of the claims, so long as the claims do not contain glaring irregularities. Post-payment audits conducted by Medicare contractors, such as recovery audit contractors (RACs) and Unified Program Integrity Contractors (UPICs), under the Medicare Integrity Program are intended to catch any improper payments or overpayments.
Under the Social Security Act, a Medicare contractor may not use extrapolation to determine overpayment amounts to be recovered by recoupment, offset, or otherwise unless it determines that (1) there is a sustained or high level of payment error; or (2) documented educational intervention has failed to correct the payment error. The Medicare Program Integrity Manual indicates that, in order to use statistical sampling for the purposes of extrapolation, a sustained or high level of payment error is determined to exist in a variety of circumstances, including where there is a “greater than or equal to 50 percent [error rate] from a previous pre- or post-payment review.”
Statistical sampling used in extrapolated audits has been the subject of much criticism over the years. A 2020 OIG report found that Medicare contractors were not consistent in how they reviewed extrapolated overpayments in the provider appeals process. OIG found that “CMS did not always provide sufficient guidance and oversight to ensure that these reviews were performed in a consistent manner.” One of the most significant inconsistencies (i.e., the use of a type of simulation testing) resulted in at least $42 million in extrapolated overpayments that were overturned in fiscal years 2017 and 2018.
Stakeholders have asserted that extrapolations often include errors that skew the results, exaggerate alleged overpayments, and would render the extrapolation outright invalid in an academic or accounting setting. Challenging extrapolations can be difficult, as administrative exhaustion, flexible regulations and manuals, and case law protect even the most flawed extrapolations.
In the final RADV rule, CMS declined to identify a specific extrapolation or adopt a uniform methodology, instead opting for ad hoc sampling and extrapolation particular to each audit. In fact, CMS reserves the right not to extrapolate RADV results depending on the circumstances of the particular audit. Commenters to CMS’s 2018 proposed rule asserted that extrapolation authority under the Social Security Act only extends to Medicare contractors auditing providers under Parts A and B and that CMS failed to cite to its legal authority to extrapolate for RADV audits. Such challenges may spark further litigation. Additionally, proper methods of statistical sampling, which may allow the results from a sample size of a few thousand patients to be applied to total populations in the millions, will be the subject of much debate.
CMS aims to select MAOs for RADV audits using a risk-based approach based on the results of prior RADV audits—meaning that better RADV audit performance should result in fewer audits for particular MAOs. To achieve the goal of fewer audits, MAOs may consider a number of options.
MAO and provider contractual relationships, especially value-based and risk sharing arrangements, will be challenged to more effectively manage RADV risk, including any audit adjustments that impact payments. This could include more intensive provider training and contracted penalties for non-compliance related to coding accuracy and medical record documentation by providers.
The magnitude of big data associated with MAO claims processing, and the punitive results for extrapolated RADV audit results, compel much greater innovation in risk management, including embracing data analytics to identify errant coding or other data trends. While CMS may employ more traditional audit methods retrospectively in RADV audits, MAOs need to be in a position to proactively identify and correct errant trends in real time before submissions.
Providers and MAOs should be alert to implementation of the Advance Notice of 2024 Capitation Rates moving Medicare Advantage diagnosis coding to ICD-10, which may change or delete Hierarchical Condition Categories (HCC) related to chronic conditions and impact RADV audits if not accounted for. These circumstances should be addressed as an operations issue and not simply as part of an audit protocol.
If you have any questions or would like more information on the issues discussed in this LawFlash, please contact any of the following: