August 28, 2019
1310 G Street, N.W.

The Honorable Seema Verma
Administrator
Centers for Medicare and Medicaid Services
7500 Security Blvd.
Baltimore, MD 21244

Washington, D.C. 20005
202.626.4800
www.BCBS.com

Submitted via the Federal Regulations Web Portal, http://www.regulations.gov
RE: Medicare and Medicaid Programs; Policy and Technical Changes to the Medicare
Advantage, Medicare Prescription Drug Benefit, Program of All-Inclusive Care for the
Elderly (PACE), Medicaid Fee-for-Service, and Medicaid Managed Care Programs for
Years 2020 and 2021 (CMS-4185-P)
Dear Administrator Verma:
The Blue Cross Blue Shield Association (BCBSA) appreciates the opportunity to provide
additional comments on the Risk Adjustment Data Validation (RADV) portions of the Proposed
Rule: “Medicare and Medicaid Programs; Policy and Technical Changes to the Medicare
Advantage (MA) , Medicare Prescription Drug Benefit, Program of All-Inclusive Care for the
Elderly (PACE), Medicaid Fee-for-Service, and Medicaid Managed Care Programs for Years
2020 and 2021” as issued in the Federal Register on Nov. 1, 2018 (83 Fed. Reg. 54982). We
appreciate the additional time offered by the agency to submit comments on the proposals
related to MA RADV audits.
BCBSA is a national federation of 36 independent, community-based and locally operated Blue
Cross and Blue Shield (BCBS) companies (Plans) that collectively provide healthcare coverage
for one in three Americans. For 90 years, Blue Cross and Blue Shield companies have offered
quality healthcare coverage in all markets across America – serving those who purchase
coverage on their own as well as those who obtain coverage through an employer, Medicare
and Medicaid. Our Plans participate in MA and Medicare Part D, and many also contract with
the Centers for Medicare and Medicaid Services (CMS) as Medicare Administrative Contractors.
As BCBSA noted in our previous comments, submitted on Dec. 31, 2018, we do not support the
significant changes to the MA RADV audit methodology proposed by CMS in this Proposed
Rule. These changes are estimated by CMS to result in a $4.5 billion potential revenue
reduction to MA over the next 10 years, including $650 million in retrospective payments to the
Trust Fund. As we detailed in our previous comments, BCBSA believes that the Proposed Rule
relies on incomplete or improper analyses. Though BCBSA appreciates CMS’ release of
additional data, information and analyses to support its RADV proposal, we remain convinced
that these additional analyses do not adequately support the proposal to change the
methodology. Additionally, CMS’ proposal to retroactively apply this flawed methodology reflects
a breach of the trusted and successful public-private partnership between CMS and MA plans
and would threaten the Administration’s commitment to ensuring Medicare beneficiaries have
access to the enhanced benefits, care coordination and financial protections provided by MA.

 Our Plans are committed to accurate risk adjustment to ensure appropriate payments, and we
look forward to working with CMS to develop an alternative path forward on RADV that will
further this long-standing successful partnership.
Furthermore, we are aware that CMS has begun to undertake new RADV audits on a larger
number of contracts in recent months, all while the final repayment methodology is yet to be
determined. Some Plans that have been contacted for these audits have not yet received their
final settlements from previous RADV audits. While we respect the right of the agency to audit
programs that are under their jurisdiction, we believe that the “rules of the audit” or complete
protocols and potential repayments should be made known before any new audits commence.
This level of uncertainty is not conducive to a stable MA program.
If implemented, these proposed changes would result in unintended and unwanted
consequences for MA, beneficiaries and the Medicare program at large. More specifically, this
change will result in inflated audit recoveries, which would distort bidding behavior in a number
of ways that are detrimental to beneficiaries. For example, higher bids result in less ability to
reduce beneficiary cost-sharing and expand supplemental benefits to address social
determinants. Inflated audit recoveries also discourage plan participation, deterring new
entrants and constraining choice for beneficiaries. In addition, systematic underpayments due to
a flawed RADV methodology would result in slowed investments in value-based care
arrangements due to systemic underpayments and an increased burden on providers serving
MA beneficiaries.
BCBSA recommendations are as follows:
1) CMS should support MA stability by finalizing a RADV methodology that ends
uncertainty and avoid policies that undermine the program and negatively impact
beneficiaries.
2) CMS must apply a Fee-for-Service (FFS) Adjuster to RADV audits to ensure
actuarial equivalence and accurate risk adjustment validation or determine
another methodology that complies with statute. The CMS’ Study and Addendum
used as the basis of the proposal does not warrant removal of the FFS Adjuster and
should be withdrawn and re-evaluated.
3) BCBSA strongly opposes the proposals to retroactively apply changes to the
RADV audit methodology. RADV methodological changes should only be applied to
plan bids after a methodology is finalized. Furthermore, MA RADV audits for plan years
2011 to 2013 should be closed out with no further actions or liabilities required from the
audited contracts.
4) BCBSA urges CMS to make several significant modifications to the RADV audit
methodology. These changes include changing the criteria for the sample selection and
adjustments to account for variability in order to produce statistically valid and actuarially
sound results.
5) BCBSA supports future regulatory changes to the appeals process and
Hierarchical Condition Category (HCC) validation procedures that would improve
the overall accuracy and fairness of the RADV audit process.

 Thank you for the opportunity to provide comments and recommendations. We would be
pleased to discuss our comments with you at your convenience. Questions regarding these
comments may be directed to Jane.Galvin@bcbsa.com.
Sincerely,

Kris Haltmeyer
Vice President, Legislative & Regulatory Policy
Office of Policy & Representation

 Medicare Advantage Risk Adjustment Data Validation Provisions
In the Medicare Advantage (MA) Technical Proposed Rule, released Nov. 1, 2018, CMS
proposed changes to the MA Risk Adjustment Data Validation (RADV) audit program. In this
Proposed Rule, CMS proposes two fundamental changes to the MA RADV audit process:
1. Removal of FFS Adjuster: Based on the results of its FFS Adjuster study published on
Oct. 26, 2018 (Study), CMS proposes to eliminate the FFS Adjuster from the RADV
audit methodology. CMS also proposes that the removal of the FFS Adjuster and
contractwide sampling and extrapolation methodology would apply to RADV audits
already conducted for plan years 2011, 2012 and 2013 as well as for any future audits.
(CMS reproduced this Study and added an addendum in June 2019).
2. Application of Contract-Level Payment Error Rate (Extrapolation): CMS invites
renewed feedback on its methodology for calculating contract-level payment errors in
RADV audits. This methodology was proposed in December 2010 and finalized in 2012.
In this Proposed Rule, CMS seeks public input on the original methodology as well as
other potential methodologies for sampling and extrapolation, including a subcohort
approach.
BCBSA appreciates the opportunity to provide detailed comments on the RADV provisions in
the Proposed Rule, which are outlined in the five key recommendation below.
Issue #1: Preventing Disruption in MA
The success of the Medicare Advantage market relies on accurate capitated payments to
Medicare Advantage Organizations (MAOs), including stable risk adjustment and data validation
policies. CMS conducts program integrity efforts in MA to prevent fraud and abuse and ensure
accurate payments. CMS conducts RADV audits as part of its program integrity audits to ensure
that diagnosis code data submitted by MAOs for risk adjustment purposes are supported by the
appropriate medical documentation. Additionally, the stability of MA relies on preventing
uncertainty for plans and other stakeholders by ensuring policies and procedures are clearly
established before bids must be submitted and other activities, such as audits, are conducted.
Recommendation #1:
CMS should support MA stability by finalizing a RADV methodology that ends
uncertainty and avoid policies that undermine the program and negatively impact
beneficiaries.
Rationale #1:
BCBSA supports program integrity efforts in MA to prevent fraud and abuse and ensure
accurate payments. Unfortunately, there has been persistent uncertainty due to lack of clarity

 and transparent rule-making on RADV audits. To end this uncertainty and support the growth of
MA, CMS should finalize a RADV audit methodology that is sound and ensures accurate
payment.
Furthermore, as stewards of the MA program, CMS should carefully consider the proposed
changes to the RADV methodology in the full context of the MA program, which is the choice of
over one in three Medicare beneficiaries. Notably, CMS already accounts for actual and
potential coding discrepancies under the MA program in several ways that reduce plan
payments:








Coding Intensity Adjustment: CMS applies a minimum annual across-the-board coding
intensity adjustment to MA risk scores designed to account for observed differences in
coding patterns completeness between the MA and FFS programs. For payment year
2019, plans are subject to a 5.91 percent reduction in risk scores due to the Coding
Intensity Adjustment.
FFS Normalization Factor: Payments to MA plans are further reduced by the FFS
normalization factor, which CMS applies annually so that the average expected risk
score in the applicable payment year will remain at 1.0. For example, from 2019 to 2020,
CMS estimates the FFS normalization factor update will result in an average -3.08
percent change in revenue for MA plans.
Transition from RAPS to EDS Data Files: CMS is currently transitioning between the use
of data files from the Risk Adjustment Processing System (RAPS) and the Encounter
Data System (EDS) to the use of only EDS data files as the sole basis for MA
beneficiary risk scores. This transition has resulted in lower than expected risk scores
and decreases on average in risk-adjusted revenue for plans. MA plan data reflects a
median 4 percentage point difference between EDS and RAPS-based risk scores in PY
2016, resulting in further decreases in risk scores and MA plan payments.1
Medical Loss Ratio: MAOs are prohibited from maintaining a Medical Loss Ratio (MLR)
less than 85 percent, and MAOs that report MLR percentages lower than that threshold
are required to repay the difference to CMS.

The cumulative impact of continually shifting risk adjustment and data validation processes and
requirements, as well as other policies, threaten the stability of the market and could stifle plans’
continued ability to modernize the program and deliver better value to beneficiaries.

Issue #2: Fee-for-Service Adjuster
A central question with the RADV audit methodology is how to address the fact that the FFS
data, upon which MA payment and risk adjustment is based, is not audited. MA risk scores are
constructed using a risk adjustment model with coefficients based on unaudited FFS data.
Simply auditing the diagnosis codes submitted by MA plans that CMS uses to calculate HCCs
and risk adjustment and then calculating an error rate against a perfect score (i.e. unaudited
1

D. Bell et al., Milliman Report: Impact of Transition from RAPS to EDS on Medicare Advantage Risk Scores (Jan.
13, 2017), available at: http://careers.milliman.com/uploadedFiles/insight/2017/2370HDP_Medicare-EDS-Survey.pdf.

 FFS data) does not account for the known fact that FFS data are not perfect. Thus, since MA
coefficients are calibrated using unaudited FFS data, an adjustment is needed to account for the
inconsistencies in the FFS data.
In its February 2012 release, CMS stated that after using the outlined methodology to calculate
a preliminary payment recovery amount, CMS would then apply a FFS Adjuster before finalizing
a recovery amount – effectively creating the proxy needed to accurately compare the error
rates. CMS noted, “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).”2 This
conclusion was supported by industry stakeholders and the American Academy of Actuaries,3
who collectively agreed that the RADV audit methodology’s determination of payment errors on
the basis of validated medical records conflicts with the MA program’s risk adjustment model,
which is based on unaudited FFS claims. CMS stated it would conduct a, “RADV-like review of
records submitted to support FFS claims data” to develop the FFS Adjuster. In the years since,
stakeholders have requested public information on this review and findings related to the
development of the FFS Adjuster as described in 2012.
However, in the Proposed Rule, CMS reverses its stance on the FFS Adjuster. CMS cites its
Study as the primary rationale for no longer believing a FFS Adjuster is necessary.4 CMS
asserts that this analysis is the “RADV-like review” outlined in their 2012 memo.
Recommendation #1: CMS must apply a FFS Adjuster to RADV audits to ensure actuarial
equivalence and accurate risk adjustment validation or determine another methodology
that complies with statute.
Specifically:
a) CMS’ Study and Addendum do not warrant removal of the FFS Adjuster.
b) The FFS Adjuster is essential for an accurate RADV audit methodology.
c) CMS’s proposal to remove the FFS Adjuster represents an impermissible reversal
of agency position.
d) Only Subclause (iii) is relevant to the inclusion of a FFS Adjuster in RADV.
Rationale #2:
a. CMS’ Study and Addendum Do Not Warrant Removal of the FFS Adjuster

2

https://www.cms.gov/Research-Statistics-Data-and-Systems/Monitoring-Programs/recovery-audit-program-parts-cand-d/Other-Content-Types/RADV-Docs/RADV-Methodology.pdf
3 See American Academy of Actuaries, Comment on RADV Sampling and Error Calculation Methodology (Jan. 21,
2011), available at: https://www.actuary.org/pdf/health/RADV_comment_letter_012111_final.pdf.
4 On October 26, 2018, CMS released an Executive Summary and a Technical Appendix titled “Fee for Service
Adjuster and Payment Recovery for Contract Level Risk Adjustment Data Validation Audits.” Both of these
documents are herein collectively referred to as the “Study.”

 In its Study and subsequent Addendum, CMS concludes that diagnosis errors in the FFS
calibration data (which it refers to as the “audit miscalibration”) do not lead to systematic
payment error in the MA program, and, thus, a FFS Adjuster is unnecessary. However, this
conclusion is based on flawed methodologies and fails to provide sufficient conclusions to
permit a determination as to whether its results are reliable and reproducible. Additionally, the
Study utilizes improper assumptions and data inputs and employs an analysis that is both
statistically and actuarially unsound. BCBSA’s analysis, as well as that of respected
independent actuarial firms, have found that, in addition to not performing the “RADV-like” audit
of FFS claims, the Study and Addendum contains statistical and actuarial errors that undermine
its conclusions.
Below we have briefly summarized just some of flaws in the CMS FFS Study, including
technical problems and faulty assumptions and conclusions. We would be happy to discuss
these concerns in detail with you and your staff.


Failure to normalize the audited coefficients on an audited dataset: The Study
appears to use audited FFS data to calibrate new audited coefficients, but then
normalizes these coefficients using unaudited FFS claims to calculate the final audited
coefficients. This flaw leads to incorrect results. An appropriate analysis would be both
calibrated and normalized with audited FFS claims data to ensure accurate comparison.



Inconsistent comparison data: The MA RADV audit methodology involves the random
sampling of beneficiaries who satisfy specified criteria. By contrast, the Study is based
on claims, rather than complete member profiles, which were pieced together from three
distinct data sets. These claims represent only outpatient services and do not represent
a complete set of services. CMS does not identify all of the criteria it used to create its
sample or whether these claims were selected at random. Furthermore, the claims the
Study used predate CMS’ recalibration of the HCC model in 2013 and fail to capture or
discuss the impact of the phase in from ICD-9 to ICD-10 diagnosis coding or the phasein of encounter data as a diagnosis source.



Conversion of claim error rate to beneficiary error rate: CMS inappropriately makes
the assumption that HCC claim error rates are statistically independent and, thus, can be
raised to the power of the average number of claims per beneficiary to calculate the
beneficiary HCC error rate. Due to normal provider coding practices, claim error rates
will not be independent. The beneficiary error rate is greatly understated due to this
assumption. CMS should have directly calculated a beneficiary error rate using all claims
for a beneficiary for determining the FFS Adjuster. In addition, CMS inappropriately used
the mean of claim volumes per enrollee to estimate beneficiary-level error rates that
might result from their sampled claim-level error rates. However, BCBSA concludes it is
not correct to assume that all audited members would have an average volume of claims
due to the fact that within any patient population, beneficiaries will have a wide range of
claims within a calendar year. As claim volume increase, the likelihood of a validation
error shrinks significantly. Conversely, for beneficiaries with low claim volumes, the
beneficiary-level error rate approaches the claim-level error rate. For example, if a

 beneficiary only has one claim triggering a HCC, the beneficiary-level error rate should
be assumed to be equal to the claim-level error rate. As a result, the derived beneficiarylevel discrepancy rates are significantly lower than what would be found if beneficiaries
were audited via an appropriate “RADV-like” approach.


Randomly removing HCCs introduces errors: CMS’ analysis simulates an audit of
FFS data by randomly determining when to remove particular HCCs from particular
members in the FFS Medicare data set, thus treating as irrelevant the differences in
expected claims’ costs associated with beneficiaries who have substantiated HCCs and
those who do not. This assumption is flawed since it means that the analysis is as likely
to remove the condition from a person with high health costs as a person with low health
costs, when, in reality, it is more likely that the person with lower health costs would
have an unsubstantiated diagnosis.



The Study uses an inappropriately small sample size: The data set used for
discrepancy analysis was a very small sample size, resulting in low to no credibility for
many HCCs. The analysis attempted to adjust for this by using averages for those HCCs
without credibility. We believe that an accurate assessment cannot be made by simply
using averages.



Flawed assumption conflating payment error and error in recoveries: CMS
contends that any issues with a certain HCC due to unaudited FFS input data are
smoothed out across the entire HCC model. This does not translate to eliminating the
need for a FFS Adjuster when comparing MA error rates with FFS error rates for the
purpose of recoveries – these are independent issues.

b. The FFS Adjuster is Essential for an Accurate RADV Audit Methodology
In the February 2012 release, CMS announced plans to develop a FFS Adjuster that would
apply to its RADV audits. While committing to develop the FFS Adjuster,5 CMS noted that “[t]he
actual amount of the adjuster will be calculated by CMS based on a RADV-like review of
records submitted to support FFS claims data.”6 Therefore, the FFS Adjuster was essential to
calculate a permissible level of payment error and limit RADV recoveries to payment errors in
excess of that level. The absence of a FFS Adjuster when making RADV payment error
recoveries would lead to systematic MA plan underpayments, in violation of the Social Security
Act’s (SSA’s) mandate for “actuarial equivalence.” 42 U.S.C. § 1395w-23(a)(1)(C)(i). Since the

5

CMS has repeatedly acknowledged data inconsistency concerns in reference to the FFS Adjuster. See, e.g., CMS,
Notice of Final Payment Error Calculation Methodology for Part C Medicare Advantage Risk Adjustment Data
Validation Contract (Feb. 24, 2012) (“Payment recovery amounts will be subject to a fee-for-service adjuster. The fee
for service adjuster accounts for the fact that the documentation standard used in RADV audits to determine a
contract’s payment error is different from the documentation standard used to develop the Part C risk-adjustment
model.”); CMS5881, “Model Calibration Factor,” at p.6. (“In RADV audits, we expect coding perfection from MA plans
. . . [but] [i]n FFS Medicare, some portion of diagnoses on FFS claims are not documented in medical record[s] . . . .
Under RADV, plans are being held to a different (higher) standard for diagnoses.”)
6 https://www.cms.gov/Research-Statistics-Data-and-Systems/Monitoring-Programs/recovery-audit-program-parts-cand-d/Other-Content-Types/RADV-Docs/RADV-Methodology.pdf

 2012 release, MA stakeholders have repeatedly requested the public release of the FFS
Adjuster as well as the data and analysis CMS used to calculate the FFS Adjuster.
However, in the Proposed Rule and accompanying Study, CMS contends that diagnosis errors
in FFS claims data do not create systematic payment errors, and that, even if they did, it would
be inappropriate to correct for this problem through payment adjustments applied to audited
contracts. As mentioned above, CMS contends it may not matter if a MA HCC coefficient is
allocated an inflated value because such an error may balance out with the remaining HCCs
and demographic coefficients relative to the benchmark average costs. However, in the context
of a RADV audit recovery, CMS is holding MA plans liable for any and all coding errors (i.e., 0
percent error rate standard) while using a significant error rate to create the values associated
with the diagnosis codes targeted in the audit. The application of a FFS Adjustor to RADV
recoveries is necessary to cure the impact of FFS errors on the values attributed to codes (and
related coefficients) found in error that are then extrapolated to the entire MA contract.
Consequently, without a FFS Adjuster, audited MA plans would be unfairly penalized by
incurring inflated recoveries that ultimately stem from miscalibrated coefficients.
The unique characteristics of the CMS-HCC risk adjustment model calibration and the RADV
audit methodology combine to create a fundamental flaw in the calculation of RADV audit
recoveries. CMS creates disease coefficients based on unaudited FFS data—which, even if
they potentially balance out in terms of calculating original payments to MA plans—necessarily
lead to inflated calculations of RADV recoveries, thereby penalizing MA plans. During RADV
audits, CMS unfairly holds MA plans to a perfect (i.e., 0 percent) error rate standard, despite the
fact that a significant error rate in the FFS data associated with a specific disease will impact the
MA plan’s contract-level extrapolated recoveries. In addition, payments to MA plans are already
reduced by the coding intensity adjustment as previously discussed, which accounts for
differences in coding completeness between FFS and MA.
BCBSA again concludes that the application of the FFS Adjuster is necessary to ensure proper
RADV audit recoveries and prevent the sort of systematic errors that would otherwise result.
CMS is required by law to ensure that MA plans are compensated at a level that is actuarially
equivalent to FFS. Evaluating the results from a RADV audit using HCC coefficients based on
an unaudited dataset is not actuarially equivalent as required by law. While this issue is
overlooked in the Proposed Rule and accompanying Study, BCBSA recommends that future
rule-making on RADV audits must address this critical shortcoming. To remedy this problem,
CMS must preserve the FFS Adjuster or entirely reformulate the CMS-HCC risk adjustment
model and RADV audit methodology.
c. CMS’s Proposal to Remove the FFS Adjuster Represents an Impermissible Reversal of
Agency Position
BCBSA urges CMS to reconsider its proposal to eliminate the FFS Adjuster. In the meantime,
CMS has not engaged in the steps necessary and sufficient to properly effectuate its reversal in
agency position. In the absence of any adequate justification, the regulatory implementation of
CMS’ proposed new policy position would be arbitrary and capricious.

 As noted above, the Proposed Rule and Study do not address the underlying data inconsistency
issue that led to CMS’ adoption of the FFS Adjuster. CMS’ failure to address this issue is
surprising given that it is statutorily required to comply with the SSA’s “actuarial equivalent” and
“same methodology” requirements. Indeed, CMS fails to fully address its policy positions and
related issues, not simply with respect to the FFS Adjuster, but also in regards to several of the
RADV issues mentioned in the Proposed Rule. The background information and agency has
provided is incomplete and unclear, and stakeholders are frustrated in attempts to provide
productive comments.
CMS’ efforts to explain its departure from prior policy regarding the FFS Adjuster are primarily
dependent upon its Study, which for the reasons noted above, is clearly deficient. The
subsequently released data and addendum were also not sufficient. Abandoning its prior
commitment to conduct a “RADV-like audit” to determine the amount of the FFS Adjuster, CMS,
instead, undertook an analysis BCBSA believes is actuarially and statistically invalid. CMS is
required, at a minimum, to thoroughly support and explain its reversal in agency position and
release the information needed for the public to reproduce the agency’s analysis.7 This
reproduction, however, is impossible without access to the complete Study, including the
underlying data and methodology.

d. Only Subclause (iii) is Relevant to the Inclusion of a FFS Adjuster
On June 28, 2019, CMS sought public comment on the extent to which 42 U.S.C. § 1395w23(a)(1)(C) is relevant to whether CMS should include a FFS Adjuster in the RADV
methodology. Subclause (i) of this provision is relevant because it establishes the actuarial
equivalence requirement, which compels CMS to include a FFS Adjuster in its proposed RADV
methodology. The rest of this statutory provision, however, is not relevant to CMS’ decision to
include a FFS Adjuster.
Subclause (ii) requires CMS to ensure risk adjustment reflects differences in coding patterns
between Medicare Advantage plans and providers. CMS has implemented this provision by
adopting the coding intensity adjustment, which CMS has explained is exclusively intended to
address differences in the completeness of MA plan and FFS provider coding, not any
differences in coding accuracy (needed for actuarial equivalence).
Subclause (iii) is also irrelevant. It provides that CMS may not use the “default risk score for new
enrollees” in Special Needs Plans (SNPs) for individuals with chronic conditions, but rather “a
risk score that reflects the known underlying risk profile and chronic health status.” Congress
adopted this provision to address the concern that if new enrollees in SNPs were given a default
risk score that failed to take into account the very severe or disabling chronic conditions
required to qualify for a SNP in the first place, their risk score would be too low. Furthermore,
Subclause (iii) does not modify the actuarial equivalence requirement in Subclause (i) and, thus,
7

Pub L. No. 106-554 § 515.; Office of Management and Budget, Guidelines for Ensuring and Maximizing Quality,
Objectivity, Utility and Integrity of Information Disseminated by Federal Agencies; Republication, 67 Fed. Reg. 8451,
8459-60 (Feb. 22, 2002)

 has no bearing on whether CMS should include a FFS Adjuster in the proposed RADV
methodology.

Issue #3: Retroactive Application of Proposed RADV Methodology
CMS proposes to apply the new RADV audit methodology, except for the subcohort proposal,
retroactively for all years after 2010, including plan years 2011, 2012 and 2013, for which RADV
audits have already been conducted. CMS argues that applying this methodology reflects
compliance with the Improper Payments Elimination and Recovery Improvement Act of 2012
(IPERIA) as well as its fiduciary responsibility to recover funds due to the Medicare Trust
Funds.
Recommendation #3:
BCBSA strongly opposes the proposals to retroactively apply changes to the RADV audit
methodology. RADV methodological changes should only be applied to plan bids after a
methodology is finalized. Furthermore, MA RADV audits for plan years 2011 to 2013
should be closed out with no further actions or liabilities required from the audited
contracts.
Rationale #3:
BCBSA’s strong opposition to applying the proposed RADV methodology to years where bids
have already been submitted is compounded for plan years 2011 to 2013, for which plans have
not only submitted bids, but plans have already been audited for those years. Application of the
new and more expansive RADV extrapolation procedures to prior payment years would
undermine the soundness of prior MA plan payments, negatively affect the applicable MA plans
and the beneficiaries they serve, and undermine CMS’ position as a good business partner with
plans. The publication requirement in Section 1853 of the Act serves an important purpose: It
ensures that MAOs have the information that they need to submit accurate bids for the following
year.8 If applied retroactively, the Proposed Rule would contravene this important objective and
exceed the authority that Congress granted to CMS. Additionally, Section 1871(e) prohibits
CMS from applying rules retroactively unless “retroactive application is necessary to comply
with statutory requirements” or “failure to apply the change retroactively would be contrary to the
public interest.9
Threat to Actuarial Soundness of Payments to MA Plans
During the annual MA bidding process, MAOs are required to attest to their current and
projected costs to furnish coverage for eligible beneficiaries. MA plan bids must include costs
associated with MA program compliance, which includes “best practices” to prepare for and
8
9

42 C.F.R. § 422.521
42 U.S.C. § 1395hh(e)(1)(A))

 successfully participate in RADV audits, as well as the liability that can result from RADV audits.
To this end, for most of the past seven years, MA plans and their actuaries have relied on the
contract-level RADV audit methodology published by CMS in February 2012 for purposes of
estimating RADV compliance costs incorporated into their annual bids.10
The Proposed Rule endeavors to dramatically reshape the RADV audit methodology in a
manner that would increase both the probability and size of audit penalties. If applied
retroactively, MA plans would be susceptible to a far greater amount of financial liability, which
MA plans could not have accounted for in their prior year bids, thereby potentially rendering
their prior payments actuarially unsound. In other words, the resources previously furnished to
MA plans may no longer be sufficient to cover the total costs of MA plan operations and
beneficiary coverage.
At no point in the Proposed Rule does CMS address the retroactive application’s impact on the
actuarial soundness of prior year bids. This is a significant omission given the central
importance of actuarial soundness to the Medicare program and its incorporation by law into the
MA program bidding process. Therefore, to properly apply its new methodology retroactively,
CMS needs to explain and account for the impact to prior year bids that have already been
approved.
Negative Impact on MA Beneficiaries
BCBSA further notes that under CMS’ proposed methodology, projected risk scores for MAO
bids would be lower, leading to higher bids, which would, in turn, result in fewer rebate dollars,
higher cost-sharing and premiums, as well as benefit reductions. Extrapolation, and retroactive
application of the proposed methodology would further exacerbate the fundamental problems
that would stem from finalization of the proposed changes to the RADV methodology. As a
consequence, MA plans would face dramatically inflated and unanticipated financial liabilities,
which will have negative collateral consequences for beneficiaries and the public interest.
BCBS Plans all maintain comprehensive compliance programs dedicated to meeting MA
requirements. Increasing the risk associated with RADV audits will necessarily divert resources
away from innovations that improve beneficiary care and towards compliance activities. To
sustain these increased efforts, Plans must receive clear and adequate notice of applicable
legal and operational requirements to ensure the delivery of high-quality care to beneficiaries.11

10

CMS, Notice of Final Payment Error Calculation Methodology for Part C Medicare Advantage Risk Adjustment
Data Validation Contract-Level Audits (Feb. 24, 2012), available at: https://www.cms.gov/Research-Statistics-Dataand-Systems/Monitoring-Programs/recovery-audit-program-parts-c-and-d/Other-Content-Types/RADV-Docs/RADVMethodology.pdf
11 For example, in its FY 2018 Financial Report, which is based on data through June of 2017, HHS calculated an
average gross payment error rate of 8.10 percent and an average net payment error rate of only 1.37 percent. This
reflected a marked improvement from FY 2016 results, which calculated an average gross payment error rate of 9.99
percent and average net payment error rate of 4.19 percent. Compare HHS, Fiscal Year 2018, Agency Financial
Report (Nov. 2018), at pp. 207-208, available at: https://www.hhs.gov/sites/default/files/fy-2018-hhs-agency-financial-

 However, CMS has taken a sudden change in direction regarding the FFS Adjuster, the
elimination of which will force MA plans to comply with a potentially higher error and
documentation standard than that applied in the FFS. Even if applied prospectively, the negative
ramifications of this reversal, as exponentially increased through extrapolation, would be
substantial. If the proposed retroactive approach were implemented, MA plans would also be
held to a more stringent standard for a multi-year period that preceded the existence of that
standard. This result would be unfair for MA plans. Further, because only a small number of
randomly selected plan contracts were audited in the 2011 to 2013 time period, those plans
would unfairly bear the full negative effect of this reversal in policy.
Finally, the increase in MA plan liabilities that would flow from the Proposed Rule’s RADV
methodology may deter future MA program participation. Those plans that choose to participate
will be forced to reserve a greater portion of their revenue to account for expanded and
retroactively-applied RADV audit liabilities. To accommodate these increased reserve needs, a
MA plan may be forced to:






Decrease supplemental benefits to their beneficiaries
Increase beneficiary cost-sharing amounts and premiums
Reduce resources for the development of new value-based payment models
Decrease their provider networks
Offer fewer benefit packages for beneficiaries

Thus, the ultimate harm stemming from the new methodology would impact not simply MA
plans, but also their beneficiaries and the Medicare program as a whole.
BCBSA believes that the negative repercussions for retroactive application of the revamped
RADV methodology to both MA plans and beneficiaries are significant. Nonetheless, CMS does
not address these concerns in the preamble to the Proposed Rule. When referencing the public
interest, CMS only cites the agency’s objective to obtain greater recoupments from audited
plans, without recognizing that the overwhelming portion of MA plan payments are devoted to
furnishing MA coverage and benefits. By contrast, BCBSA believes that the public interest is
best served by a robust MA program, which offers beneficiaries a variety of plans and benefit
packages, and that retroactive application of the RADV audit methodology hinders rather than
promotes this interest.
Therefore, because any potential benefits garnered by audit recoupments would be
substantially outweighed by the negative effects to MA plans and beneficiaries, BCBSA also
contends that CMS lacks statutory authority to retroactively apply the proposed RADV
methodology. The public interest would be especially harmed by retroactive application to
report.pdf with HHS, Fiscal Year 2016, Agency Financial Report (Nov. 2016), at pp. 213-214, available at:
https://www.hhs.gov/sites/default/files/fy-2016-hhs-agency-financial-report.pdf.

 payment years 2011 through 2013, during which time MA plans were already subject to RADV
audits. By applying the new methodology to RADV audits to years when audits have already
been completed, CMS would be imposing unforeseen financial liabilities on MA plans, thereby,
exacerbating collateral effects to the MA program and beneficiaries.

Issue #4: Proposed RADV Methodology
In its December 2010 memo, CMS proposed a RADV audit methodology that it purported would
be a statistically valid sample of beneficiaries from each audited MA contract and then an
extrapolation from the results of that sample audit would be used to calculate a contract-level
payment adjustment.12 As CMS stated, though this methodology was described as final in 2012,
it was never implemented – audits for payment years 2011, 2012 and 2013 have been
conducted according to this methodology, but contract-level recoveries have not yet been
collected. CMS is again asking for comment on this issue. In the Proposed Rule, CMS stated
that, in addition to the “final” methodology from 2012, it has identified other potential
methodologies for sampling and extrapolation. Specifically, CMS contemplates a “subcohort”
audit methodology where the calculation of improper payments would be made on the audited
MA contract for a particular subcohort or subcohorts of enrollees in a given payment year. CMS
asks for comment on both the contract-level audit methodology published in February 2012 and
the new proposal for an extrapolated audit methodology based on subcohorts of enrollees. CMS
also seeks comment on whether there are particular situations when one of these
methodologies might be preferable to the other. Finally, CMS indicates that it intends to apply
the finalized RADV payment error methodology or methodologies to payment year 2011 and all
subsequent years, but seeks comments on whether the agency should revise the contract-level
audits that have already been conducted, but not finalized, for payment years 2011, 2012 and
2013.
Recommendation #4:
BCBSA urges CMS to make several significant modifications to the RADV audit
methodology related to sample selection and substantiation in order to produce
statistically valid and actuarially sound results.
Specifically:
a) CMS should remove the criteria that a “RADV-eligible” beneficiary has at least one
HCC for the applicable payment year.
b) CMS should adjust the audit methodology to account for variability among HCCspecific substantiation rates, but not finalize its policy to adopt the subcohort
approach until further detail is provided.

“Medicare Advantage Risk Adjustment Data Validation Notice of Payment Error Calculation Methodology for Part C
Organizations Selected for RADV Audit – Request for Comment.”
12

 Rationale #4:
While CMS suggests that it is not required by law to describe its RADV methodology through a
formal regulatory process, there is no debate within stakeholders that RADV audits must employ
valid sampling and extrapolation methods. To that end, BCBSA is providing comments to
ensure that the RADV audit methodology, in addition to incorporating an appropriate FFS
Adjuster, employs sampling and extrapolation methodologies that ensure actuarial equivalence
and prevent unintended consequences. BCBSA believes that certain aspects of CMS’ contractlevel audit methodology, which were incompletely described in the agency’s five page February
2012 publication and revisited in the Proposed Rule with additional approaches, undermine its
overall statistical and actuarial soundness. Many of these shortcomings were detailed in a July
2018 report published by the Wakely Consulting Group, which revealed that CMS’ contract-level
audit methodology will lead to inequitable treatment of MA plans due to characteristics unrelated
to coding accuracy.13 The report describes simulations that reflected a high degree of
randomness among outcomes, even for contracts with identical error rates, and large
sensitivities to small variations in diagnostic mix.
Furthermore, CMS does not provide adequate detail of its proposal to conduct RADV audits of
certain patient subcohorts for BCBSA to be able to provide substantive comments at this time.
We urge CMS not to proceed with this proposal until adequate information and analysis is
released and stakeholders are able to provide actionable feedback.
a. Sample Selection – “RADV-Eligible” Members
BCBSA disagrees with the methodology’s use of a “RADV-eligible” criteria that includes a
requirement that a beneficiary have at least one diagnosis code that resulted in an assigned
HCC for the applicable payment year, which biases the sample payment error rate upwards.
This distortion results in an inaccurate error rate across the entire contract (RADV-eligible and
non-eligible). In contrast to the other five selection criteria, which are designed to include fully
participating beneficiaries in an applicable MA contract while excluding outlier beneficiaries (i.e.,
beneficiaries with End Stage Renal Disease (ESRD) and hospice status), the justification for
excluding beneficiaries who lack an assigned HCC is not readily apparent.
b. Variability among HCC-Specific Substantiation Rates
CMS’ proposed methodology should consider the fact that certain HCCs are more difficult to
substantiate in documentation in a given medical record. CMS noted that “there are attributes
intrinsic to a specific HCC that drive measurement errors higher or lower. For example, some
underlying diseases may be more difficult to evaluate and therefore the discrepancy rate or

13

Wakely Consulting Group, Medicare RADV: Review of CMS Sampling and Extrapolation Methodology (July 2018),
available at: https://www.ahip.org/wp-content/uploads/2018/07/Wakely-Medicare-RADV-Report-2018.07.pdf. This
report and its underlying technical evaluation of the CMS RADV methodology was commissioned by America’s
Health Insurance Plans (AHIP).

 measurement error would be higher.” This issue is reflected in the Study, which identified wide
variation in discrepancy rates for HCCs mapped from diagnoses in the sample set.
Issue #5: Expansion of MA Plan Appeal Rights
BCBSA appreciates the steps that CMS has taken to potentially enhance the administration of
the RADV appeals process, including expansion and clarification of MA plan appeal rights.14 In
the Proposed Rule, CMS announced that it is “considering whether to explicitly expand the MA
organizations’ RADV appeal rights,” in light of anticipated audits and recoveries under the new
RADV standards.
Recommendation #5:
BCBSA supports future regulatory changes to the appeals process and HCC validation
procedures that would improve the overall accuracy and fairness of the RADV audit
process.
Specifically, CMS should propose the following changes in future rule-making:
a) Increase the time available for plans to initiate RADV appeals
b) Allow the same set of validating evidence for appeals as is allowed in initial audits
c) Allow plans to use any available source to document the validity of a diagnosis
code
d) Expand the overly restrictive hardship exception request policy
Rationale #5: Reforms are warranted to ensure that the appeals process and HCC validation
procedures are accurate and fair, especially in light of methodological and procedural changes
proposed for 2020.
a. Increase the time available for plans to initiate RADV appeals.
Currently, MA plans must file a written request for an appeal within 60 days of the RADV audit
report. Under the extensive changes to the RADV model proposed in this rule, MA plans would
need additional time to review issues on which an appeal might be based given the increased
complexities and risks. This period does not give plans sufficient time to conduct a thorough
review of the audit report, determine if there are payment calculations and/or medical record
issues that warrant appeal, and develop an appeal request. In view of these practical
challenges, we believe that CMS should lengthen the appeals deadline to afford MA plans a
reasonable amount of time to prepare and file appeals.
Similarly, the deadline governing the reconsideration stage of the appeals should be expanded
to 90 days. Under the current rules, upon request for a CMS Administrator review, the hearing
officer’s ruling from the prior stage of appeal becomes final if the Administrator declines to
review that ruling or does not make a decision within 60 days. This process does not afford the
14

CMS, Final Rule, Medicare Program; CY 2015 Policy and Technical Changes to the Medicare Advantage and the
Medicare Prescription Drug Benefit Programs, 79 Fed. Reg. 29844, 29926 (May 23, 2014).

 MA plan a meaningful opportunity to appeal a hearing officer’s unfavorable decision if it is
finalized due to the Administrator’s inaction. Lengthening the deadline by 30 days would provide
both MA plans and CMS more time to properly administer and engage in the appeals process.

b. Allow the same set of validating evidence for appeals as is allowed in initial audits.
With the announcement of the RADV audit extrapolation methodology in February 2012, CMS
slightly broadened its medical records procedures to allow MA plans to submit more than one
medical record to validate an audited HCC. However, this procedure has remained unduly
restrictive in the audit context, and CMS continues to apply its “one best medical record” policy
with respect to audit appeals. CMS should revisit this policy in light of proposed changes to the
RADV methodology that substantially increase the potential repayment risks for MA plans.
Notably, there is an inconsistency between permitting MA plans to submit more than one
medical record during the RADV audits and forbidding the submission of multiple records on
appeal. Unless this inconsistency is remedied by broadening the basis of documentation on
appeal, the reviewer would obtain an incomplete picture of the clinical data that the plan initially
submitted to support the HCC. To ensure the integrity and accuracy of risk adjustment payment
data, RADV audit procedures—at both the audit and appeals phase—should permit MA plans to
build their case for HCC validation based on all pertinent underlying medical records.
c. Allow plans to use any available source to document the validity of a diagnosis code.
The purpose of RADV audits is to ensure the integrity and accuracy of risk adjustment payment
data and to recover for unsubstantiated MA plan payments. However, this purpose is
undermined by current rules that forbid appealing MA plans from including HCCs, medical
records or other documents beyond the audited HCC, the RADV-reviewed medical record, and
any accompanying attestation. This rule imposes an arbitrary and onerous documentation
standard that does not further CMS’ goal of properly assessing risk adjustment data.
When a MA plan lacks a medical record documentation that supports an audited HCC, CMS
should allow the plan to add any evidentiary support demonstrating that the patient has the
diagnosis in question. Such additional support could be found in Medicare Part D prescription
drug data, laboratory results, prior or current claims data or supplemental documentation from
the patient’s current treating physicians attesting to the existence of the disease. This is
especially important for certain chronic conditions that may require lifelong treatment, such as
Parkinson’s disease, Huntington’s disease, aortic atherosclerosis, Type 1 diabetes or
complications from an amputation. Treating providers often fail to explicitly identify these
conditions in the medical record because it may be presumed or readily apparent that the
patient has the condition in question. For example, if a beneficiary with Type 1 diabetes is in the
RADV audit sample, but the MA plan does not have medical record documentation from that
audit time period specifically establishing the condition, the plan should be able to furnish proof
in the form of other claims data and/or prescription drug records for insulin.
While we provide this recommendation within the context of appeals, BCBSA believes that MA
plans should be allowed (but not required) to provide alternative and supplemental forms of

 proof in both the audit and appeals stage of the RADV process to enhance CMS’ ability to
determine the accuracy of payment.
d. Expand the overly restrictive hardship exception request policy.
CMS should broaden the list of allowable hardship exceptions that may prevent an MAO from
meeting RADV audit requirements. The list of hardship exceptions should provide
accommodations for delays in obtaining medical records or attestations as a result of providers
who are traveling, sick, or deceased.