[Federal Register Volume 88, Number 21 (Wednesday, February 1, 2023)]
[Rules and Regulations]
[Pages 6643-6665]
From the Federal Register Online via the Government Publishing Office [www.gpo.gov]
[FR Doc No: 2023-01942]


=======================================================================
-----------------------------------------------------------------------

DEPARTMENT OF HEALTH AND HUMAN SERVICES

Centers for Medicare & Medicaid Services

42 CFR Part 422

[CMS-4185-F2]
RIN 0938-AT59


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

AGENCY: Centers for Medicare & Medicaid Services (CMS), HHS.

ACTION: Final rule.

-----------------------------------------------------------------------

SUMMARY: This final rule announces certain policies to improve program 
integrity and payment accuracy in the Medicare Advantage (MA) program. 
The purpose of this final rule is to outline our audit methodology and 
related policies for the contract-level MA Risk Adjustment Data 
Validation (RADV) program. Specifically, this final rule codifies in 
regulation that, as part of the RADV audit methodology, CMS will 
extrapolate RADV audit findings beginning with payment year (PY) 2018 
and will not extrapolate RADV audit findings for PYs 2011 through 2017. 
We are also finalizing a policy whereby CMS will not apply an 
adjustment factor (known as a Fee-For-Service (FFS) Adjuster) in RADV 
audits. We are also codifying in regulation the requirement that MA 
organizations (MAOs) remit improper payments identified during RADV 
audits in a manner specified by CMS.

DATES: This final rule is effective on April 3, 2023.

FOR FURTHER INFORMATION CONTACT: Joseph Strazzire, 410-786-2775 or 
David Gardner, 410-786-7791.

SUPPLEMENTARY INFORMATION:

I. Executive Summary

    Contract-level Risk Adjustment Data Validation (RADV) audits are 
our main corrective action for overpayments made to Medicare Advantage 
organizations (MAOs) when there is a lack of documentation in the 
medical record to support the diagnoses reported for risk adjustment. 
The purpose of this final rule is to outline our audit methodology and 
related policies for the contract-level RADV program. Specifically, 
this final rule codifies in regulation our approach to the use of 
extrapolation, our decision to not apply an FFS Adjuster in RADV 
audits, and the payment years in which these policies will apply.
    We are finalizing that, as part of the RADV audit methodology, CMS 
will extrapolate RADV audit findings. We are not adopting any specific 
sampling or extrapolated audit methodology, but will rely on any 
statistically valid method for sampling and extrapolation that is 
determined to be well-suited to a particular audit. Rather than 
applying extrapolation beginning for payment year (PY) 2011 audits as 
we proposed,

[[Page 6644]]

we are finalizing a policy whereby we will not extrapolate RADV audit 
findings for PYs 2011 through 2017 and will begin extrapolation with 
the PY 2018 RADV audit. As a result, CMS will only collect the non-
extrapolated overpayments identified in the CMS RADV audits and 
Department of Health and Human Services Office of Inspector General 
(HHS-OIG) audits between PY 2011 and PY 2017, and will begin collection 
of extrapolated overpayment findings for any CMS and OIG audits 
conducted in PY 2018 and any subsequent payment year. We believe that 
this is an appropriate policy because it recognizes our fiduciary duty 
to protect taxpayer dollars from overpayments, and preserves our 
ability to collect on potentially significant amounts of overpayments 
made to plans beginning in PY 2018 using an extrapolation methodology. 
This final rule will also allow CMS to focus on conducting future RADV 
audits as soon as practicable after an MAO payment year concludes, 
which was the topic of significant public comment to the proposed rule. 
Lastly, we have determined that it is in the best interest of all 
parties to ensure that the contract-level RADV appeals process, which 
is also outlined in regulation, is able to successfully process all 
RADV appeals. By not using an extrapolation methodology prior to PY 
2018, we expect to better control the total number of active appeals 
that are submitted in the first few years following finalization of 
this rule, which will alleviate burden on MAOs and CMS.
    We are also finalizing a policy whereby CMS will not apply an FFS 
Adjuster in RADV audits because we have determined that an FFS Adjuster 
is not appropriate. As described at great length in this final rule, we 
have decided not to apply an FFS Adjuster in RADV audits because: (1) 
we believe, consistent with the D.C. Circuit's decision in 
UnitedHealthcare (UnitedHealthcare Insurance Co. v. Becerra, 16 F.4th 
867 (D.C. Cir. August 13, 2021, reissued November 1, 2021), cert. 
denied, 142 S. Ct. 2851 (U.S. June 21, 2022) (No. 21-1140)), 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 of 
MAOs to return improper payments (for example, payments for unsupported 
diagnosis codes); and (2) it would not be reasonable to read the Social 
Security Act (the Act) as requiring a reduction in payments to MAOs by 
a statutorily-set minimum adjustment in the coding pattern adjustment, 
while at the same time prohibiting CMS from enforcing longstanding 
documentation requirements by requiring an offset to the recovery 
amounts calculated for CMS audits.
    We are also codifying in regulation the requirement that MAOs remit 
improper payments identified during RADV audits in a manner specified 
by CMS. After the effective date of this final regulation, on a rolling 
basis (over a period of months, which will be communicated to MAOs by 
CMS), we will begin issuing the enrollee-level audit findings from the 
CMS RADV audits that have been completed, as well as recovering the 
enrollee-level improper payments identified in HHS-OIG audits.
    Nothing in this rule changes the longstanding principle that a 
diagnosis code that is not documented in a patient's medical record is 
not a valid basis for CMS risk adjustment payments to an MAO. 
UnitedHealthcare Ins. Co. v. Becerra, 16 F.4th 867, 869 (D.C. Cir. 
2021) (``Neither Congress nor CMS has ever treated an unsupported 
diagnosis for a beneficiary as valid grounds for payment to a Medicare 
Advantage insurer.''). Nor does this rule change the longstanding 
obligation of an insurer to refund payments to CMS if it learns through 
any means that a diagnosis lacks support in the beneficiary's medical 
record. Id.

II. Background

A. General Overview of Risk Adjustment Payments in the MA Program

    The Balanced Budget Act of 1997 (BBA), Public Law (Pub. L.) 105-33, 
established a new Part C of the Medicare program, known then as the 
Medicare+Choice (M+C) program, which became effective in January 1999. 
As part of the M+C program, the BBA authorized CMS to contract with 
public or private organizations to offer a variety of health plan 
options for Medicare beneficiaries. These health plans provide all 
Medicare Part A and Part B (also known as ``Original Medicare,'' or 
``Medicare FFS'') benefits, and most offer additional benefits beyond 
those covered under the Medicare FFS program. The M+C program in Part C 
of Medicare was renamed the Medicare Advantage (MA) program under the 
Medicare Prescription Drug, Improvement, and Modernization Act of 2003 
(MMA) (Pub. L. 108-173), enacted in December 2003. The MMA updated and 
improved the choice of plans for beneficiaries under Part C and changed 
the way benefits are established and payments are made. As of August 
2022, over 29 million individuals receive their Medicare benefits 
through MA, which represents nearly half of the total Medicare 
beneficiary population.\1\
---------------------------------------------------------------------------

    \1\ CMS, CMS Fast Facts, August 2022 Edition, pg.1, https://data.cms.gov/sites/default/files/2022-08/4f0176a6-d634-47c1-8447-b074f014079a/CMSFastFactsAug2022.pdf.
---------------------------------------------------------------------------

    Section 1853(a)(1)(C) of the Act requires that CMS risk-adjust 
payments made to MAOs. Risk adjustment strengthens the MA program by 
ensuring that accurate payments are made to MAOs based on the health 
status and demographic characteristics of their enrolled beneficiaries, 
and that MAOs are paid appropriately for their plan enrollees (that is, 
less for healthier enrollees who are expected to incur lower health 
care costs, and more for less healthy enrollees who are expected to 
incur higher health care costs). Making accurate payments to MAOs also 
ensures we are safeguarding Federal taxpayer dollars.
    The current risk adjustment model employed to adjust MAO payments 
is known as the CMS Hierarchical Condition Category (CMS-HCC) model. 
This model functions by categorizing International Classification of 
Disease, Clinical Modification (ICD-CM) \2\ diagnosis codes into 
disease groups called Hierarchical Condition Categories, or HCCs. Each 
HCC includes diagnosis codes that are related clinically and have 
similar cost implications. There are approximately 9,875 diagnoses 
mapped to 86 HCCs in the CMS-HCC Risk Adjustment Model for 2022.\3\ MA 
enrollee HCCs are assigned based on data submitted to CMS by MAOs. The 
HCCs contribute to an enrollee's risk score, which is used to adjust a 
base payment rate. Essentially, the higher the risk score for an 
enrollee, the higher the expected health care cost for the enrollee and 
the greater payment that is received by the MAO.
---------------------------------------------------------------------------

    \2\ The ICD-CM is a modification of the ICD, authorized by the 
World Health Organization, used as a source for diagnosis codes in 
the United States. The ICD-CM has been adopted by the Secretary as 
the standard medical data code set. See 45 CFR 162.1002.
    \3\ Source: 2022 Midyear Final ICD-10 Mappings at https://www.cms.gov/files/zip/2022-midyear-final-icd-10-mappings.zip.
---------------------------------------------------------------------------

    The CMS-HCC model was first used for payment in 2004 and has been 
recalibrated numerous times since then. When CMS recalibrates the CMS-
HCC risk adjustment model, it uses data from Medicare FFS claims, using 
diagnoses in one year to predict the following year's expenditures. 
Claims data from beneficiaries enrolled in the Medicare FFS program are 
used to calibrate the CMS-HCC model, which produces a set of 
coefficients (also known as risk

[[Page 6645]]

factors) that represent the marginal (additional) cost of each medical 
condition and demographic factor reported for a given beneficiary. (For 
additional information, see the Medicare Managed Care Manual, Ch. 7, 
section 70.1.\4\) Each beneficiary's risk coefficients are added 
together to form a risk score for that beneficiary that is used to 
adjust the insurer's base payment rate for that beneficiary.
---------------------------------------------------------------------------

    \4\ https://www.cms.gov/Regulations-and-Guidance/Guidance/Manuals/Downloads/mc86c07.pdf.
---------------------------------------------------------------------------

    The diagnosis data that MAOs submit to CMS do not undergo a 
validation review by CMS before being relied on by CMS to calculate 
each enrollee's risk score and make payments. Because there is an 
incentive for MAOs to potentially over-code diagnoses to increase their 
payments, that is, to code diagnoses not properly substantiated by 
medical record documentation, CMS conducts post-payment audits of MAO-
submitted diagnosis data from a selection of MAOs for specific payment 
years to ensure that the diagnoses they submitted are supported by 
their enrollees' medical records. These audits are called contract-
level Risk Adjustment Data Validation (RADV) program audits. While RADV 
audits are intended to identify improper risk adjustment payments, they 
are not specifically designed to detect fraud,\5\ nor are they intended 
to identify all improper diagnosis submissions made by MAOs for risk 
adjustment payment.\6\
---------------------------------------------------------------------------

    \5\ For example, the Department of Justice is responsible for 
pursing potential violations of the False Claims Act, which includes 
certain elements of knowledge.
    \6\ CMS contract-level RADV audits focus on specific MAO 
contracts to determine and recoup improper payments. The HHS-OIG 
also undertakes audits of MAOs, similar to RADV audits, as part of 
its oversight functions. CMS can collect the improper payments 
identified during those HHS-OIG audits, including the extrapolated 
amounts calculated by the OIG. CMS also oversees the Part C Improper 
Payment Measurement, previously referred to as ``national RADV,'' to 
determine a program-wide improper payment rate as required by the 
Payment Integrity Information Act of 2019 (Pub. L. 116-117). In 
addition to risk adjustment oversight conducted by CMS, HHS also 
oversees HHS-RADV, which was created by the Affordable Care Act to 
strengthen the integrity of the Affordable Care Act Marketplace by 
validating the accuracy of data submitted by issuers that is used to 
calculate the amount of funds transferred to insurers based on the 
actuarial risks of the individuals they enroll. Neither the Part C 
Improper Payment Measurement nor the HHS-RADV programs are subject 
to the provisions of this final rule.
---------------------------------------------------------------------------

B. Purpose and Description of Contract-Level RADV Audits

    The improper payment measurements conducted each year by CMS that 
are included in the HHS Agency Financial Report, as well as audits 
conducted by the HHS-OIG, have demonstrated that the MA program is at 
high risk of improper payments. In fiscal year (FY) 2021 (based on 
calendar year 2019 payments), we calculated that CMS made over $15 
billion in Part C overpayments, a figure representing nearly 7 percent 
of total Part C payments.\7\ The HHS-OIG has also released several 
reports over the past few years that demonstrate a high risk of 
improper payments in the MA program,\8\ and for several years has 
identified the MA program as one of the top management and performance 
challenges facing HHS due to the high amount of improper payments.\9\ 
The Medicare program, including MA, has also been identified by the 
Government Accountability Office (GAO) as a high-risk program due to 
the risk of substantial improper payments.\10\
---------------------------------------------------------------------------

    \7\ HHS, FY 2021 HHS Agency Financial Report, pg. 211, https://www.hhs.gov/sites/default/files/fy-2021-hhs-agency-financial-report.pdf. CMS made over $23 billion in total Part C improper 
payments. The improper payment measurement for the MA program in FY 
2021included both overpayments ($15 billion) and underpayments ($8 
billion).
    \8\ For example, see reports: Medicare Advantage Compliance 
Audit of Specific Diagnosis Codes That Anthem Community Insurance 
Company, Inc. For example, see reports: Medicare Advantage 
Compliance Audit of Specific Diagnosis Codes That Anthem Community 
Insurance Company, Inc. (Contract H3655) Submitted to CMS, May 21, 
2021, https://oig.hhs.gov/oas/reports/region7/71901187.asp; Medicare 
Advantage Compliance Audit of Specific Diagnosis Codes That Blue 
Cross Blue Shield of Michigan (Contract H9572) Submitted to CMS, 
February 24, 2021, https://oig.hhs.gov/oas/reports/region2/21801028.asp; Medicare Advantage Compliance Audit of Specific 
Diagnosis Codes That Highmark Senior Health Company (Contact H3916) 
Submitted to CMS, September 29, 2022, https://oig.hhs.gov/oas/reports/region3/31900001.asp; Medicare Advantage Compliance Audit of 
Specific Diagnosis Codes That Cariten Health Plan, Inc., (Contract 
H4461) Submitted to CMS, July 18, 2022, https://oig.hhs.gov/oas/reports/region2/22001009.asp; Medicare Advantage Compliance Audit of 
Diagnosis Codes That SCAN Health Plan (Contract H5425) Submitted to 
CMS, February 3, 2022, https://oig.hhs.gov/oas/reports/region7/71701169.asp; Medicare Advantage Compliance Audit of Diagnosis Codes 
That Humana, Inc., (Contract H1036) Submitted to CMS, April 19, 
2021, https://oig.hhs.gov/oas/reports/region7/71601165.asp.
    \9\ For example, see OIG, 2021 Top Management and Performance 
Challenges Facing HHS, pg. 13, https://oig.hhs.gov/reports-and-publications/top-challenges/2021/2021-tmc.pdf.
    \10\ GAO, Medicare Program & Improper Payments, https://www.gao.gov/highrisk/medicare-program-improper-payments.
---------------------------------------------------------------------------

    RADV audits are our main corrective action for overpayments made to 
MAOs when there is a lack of documentation in the medical record to 
support the diagnoses reported for risk adjustment. We select MAOs for 
RADV audits using a risk-based approach that focuses on HCCs that are 
more likely to be in error as identified by prior RADV audits, Part C 
Improper Payment Measurements, and OIG findings, and other 
vulnerability analyses. RADV audits occur after the final risk 
adjustment data submission deadline for the MA contract year and after 
CMS recalculates the risk factors for affected individuals to determine 
if payment adjustments are necessary, as described at 42 CFR 
422.310(g).\11\ RADV audits are intended to confirm the presence of 
risk adjustment conditions (that is, diagnoses that map to HCCs) as 
reported by MAOs in medical record documentation. RADV audits confirm 
the presence of the diagnoses related to the enrollee's HCC profile 
through the review of certain categories of medical records submitted 
by the MAOs for the purpose of a RADV audit; specifically, inpatient 
hospital, hospital outpatient facility, and physician/practitioner 
(excluding suppliers of durable medical equipment, prosthetics, 
orthotics, and supplies) medical records. Risk adjustment discrepancies 
are identified when an enrollee's HCCs used for payment, which are 
based on MAO self-reported data, differ from the HCCs assigned based on 
the medical record review performed by CMS through the RADV audit 
process. Risk adjustment discrepancies can be aggregated to determine 
an overall level of payment error. In turn, payment error for a sample 
of contract enrollees can be used to calculate a total payment error 
estimate, for the larger universe of enrollees within an MAO contract 
from which a sample is drawn, within specified confidence intervals 
using statistical extrapolation.
---------------------------------------------------------------------------

    \11\ See the May 23, 2014 final rule titled ``Medicare Program; 
Contract Year 2015 Policy and Technical Changes to the Medicare 
Advantage and the Medicare Prescription Drug Benefit Programs Final 
Rule'' (79 FR 29843, at 29926) for a more detailed discussion of the 
timing and execution of the RADV audit and appeals process.
---------------------------------------------------------------------------

C. History of the Contract-Level RADV Program

    RADV audits have existed in various forms and approaches for over 
20 years. RADV audits began for payment year (PY) 1999, when the amount 
of payment made to MAOs on a risk-adjusted basis was small (10 
percent). During the audit period from PY 1999 until PY 2003, our RADV 
activity had an educational focus and was primarily intended to provide 
information that could be used by MAOs to improve the accuracy of the 
risk adjustment data submitted to CMS for payment. Payment adjustments 
(recoveries) were limited to enrollee-level adjustments for those 
enrollees sampled in the audits and were not extrapolated to the 
overall error of the

[[Page 6646]]

contract. As a result, for the few MA plans we audited, payment 
recovery amounts were small.
    Risk adjustment payments using the CMS-HCC risk adjustment model 
began for the first time in PY 2004. Because of various risk adjustment 
payment methodology changes required in the BBA and the Benefits 
Improvement and Protection Act of 2000 (BIPA) (Pub. L. 106-554), we 
provided a payment ``phase-in'' under the new risk adjustment 
methodologies from 2000 to 2007, when MAOs' payments were 100 percent 
risk-adjusted under the current methodology.\12\ Under the new 
methodology that began in PY 2004, MAOs were required to submit 
diagnoses from multiple sites of care, which increased the 
administrative data burden on MAOs. Because of this burden and the 
associated phase-in of the new methodology, CMS considered PYs 2004 
through 2006 as pilot years for the purpose of the RADV program and did 
not seek to recover improper payments for those payment years based on 
the audit results.
---------------------------------------------------------------------------

    \12\ CMS, Advance Notice of Methodological Changes for Calendar 
Year (CY) 2004 Medicare+Choice (M+C) Payment Rates, 4-5 (March 28, 
2003), https://www.cms.gov/Medicare/Health-Plans/MedicareAdvtgSpecRateStats/Downloads/Advance2004.pdf.
---------------------------------------------------------------------------

    Improper payment recovery resumed for PY 2007, when we conducted 
two sets of RADV audits: (1) Pilot 2007, which involved 5 MA contracts; 
and (2) Targeted 2007, which involved 32 MA contracts. CMS began with 
the Pilot 2007 audit to test the methodology and make any needed 
changes before conducting the Targeted 2007 audit. CMS selected MA 
contracts after measuring the weighted average change in disease scores 
(risk scores) over the preceding 3-year period and grouping MAO 
contracts as high, medium, or low relative to other MA contracts that 
were eligible for a RADV audit. Through these two sets of audits, we 
recouped $13.7 million. Payment adjustments were again limited to 
enrollee-level adjustments for those enrollees sampled in the audits 
and not extrapolated to the overall contract error. After CMS' findings 
were reported to each MAO, any MAO that disagreed with CMS' 
determinations could challenge them through an administrative dispute 
and appeals process that was established by regulation (75 FR 19678). 
This dispute and appeals process, as subsequently amended (75 FR 32858 
and 79 FR 29844), remains in effect and allows for the appeal of the 
medical record review determination and/or the payment error 
calculation through a three-level administrative review process, as 
outlined in 42 CFR 422.311. To date, CMS has not recovered based on 
RADV audit findings for audit years after PY 2007, as described more 
fully in this section of this rule.
1. Development of an Audit Methodology (PYs 2007 Through 2010)
    After the RADV audits were conducted for PY 2007, CMS paused RADV 
audits for PYs 2008, 2009, and 2010. CMS used those years to continue 
refining the methodology for the RADV audits, including the 
consideration of statistical methods to calculate extrapolated improper 
payments based on the individual errors identified. The use of 
extrapolation would enable us to make contract-level payment 
adjustments rather than simply adjusting payments for specific 
enrollees from an audit sample, as we had done previously.
    On December 20, 2010, we published an informal proposal on the CMS 
website that outlined our intended RADV methodology for: (1) selecting 
a statistically valid sample of enrollees from each audited MA 
contract; and (2) calculating a contract-level payment adjustment by 
extrapolating the results of that sample. We invited public comment on 
this proposed methodology.
2. Informal Proposal Comments and the FFS Adjuster
    In response to the December 2010 informal proposal, some MAOs 
suggested that CMS cannot lawfully enforce the requirement of medical 
record documentation for diagnosis codes while making payments at the 
published rates. These MAOs argued that there is a difference in 
auditing standards between Medicare FFS and MA diagnosis data because, 
in contrast to the MAO-submitted diagnoses data, Medicare FFS data is 
``unaudited'' by CMS. This difference purportedly exists because most 
FFS payments are made on the basis of the item or service provided and 
not the beneficiary's diagnosis or diagnoses. For example, an office 
visit is paid based on whether the evaluation and management service 
billed met Medicare coverage and payment rules, not based on what 
diagnoses are listed on the claim or in the medical record. As a 
result, they argued, the Medicare FFS data used to calculate MAO 
payments will understate the cost of treating various conditions and, 
because erroneous diagnoses in the FFS claims data are used to 
calibrate the MA payment model, CMS must 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. This proposed adjustment to the MAO payment 
rates and/or documentation standard is referred to as an ``FFS 
Adjuster.''
    To understand the MAOs' argument about why an FFS Adjuster is 
needed, some background is important. These MAOs ground their arguments 
in section 1853(a)(1)(C)(i) of the Act, which requires the Secretary to 
adjust payments to MAOs for demographic and health-related risk factors 
so as to ensure ``actuarial equivalence.'' As described previously, the 
Act requires that we calculate risk-adjusted payments to MAOs to ensure 
that MAOs are paid appropriately based on the enrollees' health status 
and demographic characteristics. The current risk adjustment model does 
this by calculating plan enrollees' risk scores and, in turn, using 
them to adjust the MAOs' base payment rates, which are the rates for 
the average beneficiary.
    This system of risk adjustment rests on two important principles. 
First, MAOs' payments are calculated using the CMS-HCC risk adjustment 
model, which is published each time it is updated (see section 1853(b) 
of the Act).\13\ Second, an MAO may only report a diagnosis when that 
diagnosis is properly supported by the beneficiary's medical records. 
As we noted in our April 15, 2022 Health Plan Management System (HPMS) 
memorandum, Reminder of Existing Obligation to Submit Accurate Risk 
Adjustment Data, MAOs must submit data that conforms to all relevant 
national standards, including the International Classification of 
Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) Guidelines 
for Coding and Reporting requirement that diagnoses be documented in 
patients' medical records. (See 42 CFR 422.310(d)(1); 45 CFR 
162.1002(c)(2) and (c)(3).) The diagnosis codes and other risk 
adjustment information that MAOs submit directly affect the calculation 
of CMS payments to the MAO. A diagnosis code that is not documented in 
a patient's medical record is not a valid basis for CMS risk adjustment 
payments to an MAO. UnitedHealthcare Ins. Co. v. Becerra, 16 F.4th 867, 
869, 877 (D.C. Cir. 2021). Medical records properly support a reported 
diagnosis when they comply with all CMS data and documentation 
requirements, which are described in current agency policy

[[Page 6647]]

documents, including the Medicare Managed Care Manual.\14\ In their 
annual contracts with CMS, MAOs agree to operate in accordance with 
applicable Federal statutes, regulations, and policies, including 
policies described in the Medicare Managed Care Manual. MAOs are also 
required to submit a sample of medical records for the validation of 
this risk adjustment data, as required by CMS (see 42 CFR 422.310(e)).
---------------------------------------------------------------------------

    \13\ https://www.cms.gov/Medicare/Health-Plans/MedicareAdvtgSpecRateStats/Announcements-and-Documents.
    \14\ https://www.cms.gov/Regulations-and-Guidance/Guidance/Manuals/internet-Only-Manuals-IOMs-Items/CMS019326.
---------------------------------------------------------------------------

3. The 2012 Methodology
    The feedback received from industry in response to the informal 
proposal in 2010 was considered by CMS, and on February 24, 2012, we 
issued on our website \15\ what we described as a final methodology for 
RADV contract-level payment error calculation, to begin with PY 2011 
RADV audits (referred to herein as the ``2012 methodology''). That 
methodology described sampling techniques and a statistical calculation 
to extrapolate from the sample selected, as well as the use of an FFS 
Adjuster.\16\ (Although the use of an FFS Adjuster beginning with PY 
2011 RADV audits was included in the 2012 methodology, CMS has not 
issued final RADV audit results for PY 2011 audits or any subsequent 
year, and therefore, an FFS Adjuster has not been applied to any RADV 
audits issued by CMS to date.)
---------------------------------------------------------------------------

    \15\ CMS, Notice of Final Payment Error Calculation Methodology 
for Part C Medicare Advantage Risk Adjustment Data Validation 
Contract-Level Audits, 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.
    \16\ Id. at 4-5.
---------------------------------------------------------------------------

    Sampling Technique: Under the 2012 methodology, up to 201 enrollees 
from each audited MA contract would be selected according to certain 
criteria. These criteria included, but were not limited to, the 
enrollee's: (1) continuous enrollment in the MA contract for the entire 
data collection year and January of the payment year; (2) lack of end-
stage renal disease (ESRD) or hospice status for the entire data 
collection year and January of the payment year; (3) enrollment in 
Medicare Part B coverage for the entire data collection year; and (4) 
assignment of at least one CMS-HCC based on diagnoses submitted by the 
MAO for risk-adjustment payment. The RADV-eligible enrollees would then 
be ranked by risk score and divided into three equal strata (low risk 
score, average risk score, and high risk score), with an equal number 
of enrollees randomly selected from each stratum (for example, 67 
enrollees per stratum in the case of an audit of 201 enrollees).
    Payment Error Calculation: After medical records were reviewed, 
payment errors would be calculated for each selected enrollee based on 
the number of months the person was enrolled in the selected MA 
contract (and also was not in ESRD or hospice status) during the 
payment year. A payment error amount for each stratum would be 
calculated, which could include both RADV-identified overpayments and 
underpayments, and an overall payment error estimate for the audited 
contract would be derived, along with a 99 percent confidence interval 
around the payment error estimate.
    FFS Adjuster: As part of the 2012 methodology, we also stated that 
we would apply an FFS Adjuster before finalizing audit recovery. The 
2012 methodology stated that the actual value of the FFS Adjuster would 
be calculated by CMS based on a RADV-like review of records submitted 
to support FFS claims data.
    CMS subsequently conducted an extensive study regarding the impact 
of such errors in Medicare FFS claims data for the purpose of 
determining the appropriate value of an FFS Adjuster. This study found 
that, in fact, errors in Medicare FFS claims data did not have any 
systematic effect on the risk scores calculated by the CMS-HCC risk 
adjustment model and, therefore, did not have any systematic effect on 
the payments made to MAOs. On October 26, 2018, we published an 
Executive Summary and Technical Appendix of our FFS Adjuster study 
findings on the CMS website, which are available at https://www.cms.gov/Research-Statistics-Data-and-Systems/Monitoring-Programs/Medicare-Risk-Adjustment-Data-Validation-Program/Resources.html. 
Additional information on this study can also be found in the November 
2018 proposed rule.
4. The 2018 RADV Proposed Rule
    In the 2018 proposed rule, to enhance transparency and provide 
ample notice to MAOs, we proposed to codify in regulation our 
methodological approach to RADV audits that would apply to all of the 
payment year audits that have not yet been finalized. These 
methodologies would apply to PY 2011 and subsequent years and include 
our proposals to use extrapolation and not apply an FFS Adjuster to our 
RADV audit findings.
5. Subsequent Federal Register Notices (2018, 2019, 2021, and 2022)
    Since publication of the 2018 proposed rule, we have published 
several related notices to further enhance transparency and encourage 
robust public comment:
     On December 27, 2018, we announced in the Federal Register 
(83 FR 66661) an extension of the comment period for the proposed RADV 
provisions until April 30, 2019, as well as a plan to release data 
underlying the October 26, 2018, FFS Adjuster Study.\17\
---------------------------------------------------------------------------

    \17\ https://www.federalregister.gov/documents/2018/12/27/2018-28070/medicare-and-medicaid-programs-risk-adjustment-data-validation.
---------------------------------------------------------------------------

     On March 6, 2019, we issued a notice in the Federal 
Register (84 FR 8069) announcing the release of additional data 
underlying the FFS Adjuster Study, both on the CMS website and to those 
organizations who established data use agreements (DUAs) with the CMS 
Office of Enterprise Data Analytics (OEDA).\18\
---------------------------------------------------------------------------

    \18\ https://www.federalregister.gov/documents/2019/03/06/2019-04052/medicare-program-release-of-data-underlying-risk-adjustment-data-validation-provisions.
---------------------------------------------------------------------------

     On April 25, 2019, we posted updates to existing 
documentation related to the study data, as well as additional data on 
the CMS website.\19\
---------------------------------------------------------------------------

    \19\ https://www.cms.gov/Research-Statistics-Data-and-Systems/Monitoring-Programs/Medicare-Risk-Adjustment-Data-Validation-Program/Other-Content-Types/RADV-Docs/NPRM-4185-Provisional-Data-Release-CPI-FFSA-Coefficients.xlsx.
---------------------------------------------------------------------------

     On April 30, 2019, we issued a notice in the Federal 
Register (84 FR 18215) granting an additional extension of the comment 
period for the proposed RADV provisions until August 28, 2019. We also 
announced that we would be releasing additional data underlying the FFS 
Adjuster study, including data containing Protected Health Information 
(PHI), to all parties who entered an applicable DUA with CMS and paid 
the required fee.\20\
---------------------------------------------------------------------------

    \20\ https://www.cms.gov/research-statistics-data-and-systems/files-for-order/limiteddatasets/.
---------------------------------------------------------------------------

     On June 28, 2019, we issued a notice in the Federal 
Register (84 FR 30983) \21\ that we replicated the FFS Adjuster Study 
and published a summary of that replication as an addendum to the study 
on the CMS website.\22\ The purpose of this replication was to allow us 
to test our initial results and release a more complete set of 
underlying data. (Certain intermediate data elements, not saved as part 
of the implementation of the initial study, were preserved and 
published in the addendum.) The

[[Page 6648]]

results of the replication were broadly consistent with the initial 
implementation of the study. In addition, the addendum contained 
further discussion of the study's assumptions and methodology. We also 
released the programming language used to implement the replication of 
the study, and a description of the technical requirements for use of 
that programming language.
---------------------------------------------------------------------------

    \21\ https://www.federalregister.gov/documents/2019/06/28/2019-13891/medicare-and-medicaid-programs-risk-adjustment-data-validation.
    \22\ https://www.cms.gov/Research-Statistics-Data-and-Systems/Monitoring-Programs/Medicare-Risk-Adjustment-Data-Validation-Program/Resources.html.
---------------------------------------------------------------------------

     In the October 21, 2021 Federal Register (86 FR 58245), we 
issued a notice that provided a 1-year extension of the timeline for 
publication of the final rule.\23\
---------------------------------------------------------------------------

    \23\ https://www.federalregister.gov/documents/2021/10/21/2021-22908/medicare-and-medicaid-programs-policy-and-technical-changes-to-the-medicare-advantage-medicare.
---------------------------------------------------------------------------

    As part of this extension, we explained our determination that we 
were unable to meet the 3-year timeline for publication.\24\ Based on 
extensive public comments received on the 2018 proposed rule and 
subsequent FFS Adjuster study and related data, along with delays 
resulting from the agency's focus on the COVID-19 public health 
emergency, we determined that additional time was needed to address the 
complex policy and operational issues that were raised. As such, we 
extended the timeline to publish the final rule from November 1, 2021 
to November 1, 2022.
---------------------------------------------------------------------------

    \24\ Section 1871(a)(3)(A) of the Act requires the Secretary to 
``establish and publish a regular timeline for the publication of 
final regulations based on the previous publication of a proposed 
regulation or an interim final regulation.'' Section 1871(a)(3)(B) 
of the Act provides that ``[s]uch timeline . . . shall not be longer 
than 3 years except under exceptional circumstances.'' The Secretary 
therefore may not ``establish'' a ``regular timeline'' for the 
finalization of a proposal or interim final rule that exceeds three 
years, absent exceptional circumstances. Section 1871(a)(3)(B) of 
the Act authorizes the Secretary to ``vary such timeline''--that is, 
to alter the ``regular timeline'' initially ``establish[ed]'' for 
finalization--by publishing a timely notice in the Federal Register 
with ``a brief explanation of the justification for such 
variation.'' As we have said, ``[t]he Secretary may extend the 
initial targeted publication date of the final regulation, if the 
Secretary provides public notice including a brief explanation of 
the justification for the variation no later than the regulation's 
previously established proposed publication date.'' 69 FR 78443.
    Under the plain text of the Act, no ``exceptional 
circumstances'' are required for the Secretary to extend the initial 
targeted publication date of the final regulation, but only ``a 
brief explanation of the justification'' for doing so. The Secretary 
has often extended such timelines without any reference to 
``exceptional circumstances.'' (See 86 FR 50263; 85 FR 55385; 85 FR 
52940; 85 FR 7; 79 FR 62356; 74 FR 8867; 72 FR 16794; 72 FR 13710.) 
But the Secretary has also said that the Act ``permits an extension 
of a published timeline under exceptional circumstances,'' 69 FR 
78442, and has invoked ``exceptional circumstances'' in extending 
such timelines, including in the notices published in this 
rulemaking. For the reasons explained in this note, the Act has 
never required exceptional circumstances for such extensions--though 
exceptional circumstances have often been present, as they were 
here, when such timelines have been extended.
---------------------------------------------------------------------------

     In the November 1, 2022 Federal Register (87 FR 65723), we 
issued a notice that provided a 3-month extension of the timeline for 
publication of the final rule.\25\ We explained that we were unable to 
meet the November 1, 2022, timeline for publication of the previously 
referenced RADV-audit related provisions. We explained that we 
continued to have ongoing delays resulting from the agency's focus on 
the COVID-19 public health emergency, and we determined that additional 
time continued to be needed to address the complex policy and 
operational issues that were raised. As such, we extended the timeline 
to publish the final rule from November 1, 2022, to February 1, 2023.
---------------------------------------------------------------------------

    \25\ https://www.federalregister.gov/documents/2022/11/01/2022-23563/medicare-and-medicaid-programs-policy-and-technical-changes-to-the-medicare-advantage-medicare.
---------------------------------------------------------------------------

    We received approximately 154 timely pieces of correspondence in 
response to the 2018 proposed rule and the subsequent notices and data 
releases. Summaries of the public comments that respond to the RADV 
provisions, and our responses to those public comments, are set forth 
in the discussion that follows. Additional public comments outside of 
the scope of the RADV proposed provisions were not considered and are 
not addressed in this final rule.

III. Provisions of the RADV Final Rule

A. Extrapolation of RADV Audit Findings

1. Use of Extrapolation in the Medicare Program
    Extrapolation, or the act of estimating a value (such an 
overpayment amount for a Medicare provider) based on a statistically 
valid sample of units (such as Medicare claims), has historically been 
a standard part of auditing practice at CMS. There is significant 
guidance, including case law and best practices from HHS and other 
Federal agencies, stating that extrapolation may be utilized as a valid 
part of calculating improper payments. In particular, courts have held 
that sampling and extrapolation are a valid method of calculating 
improper Medicare payments, so long as statistically valid methods are 
used. See United States v. Lahey Clinic Hosp., Inc., 399 F.3d 1, 18 
n.19 (1st Cir. 2005) (noting that ``sampling of similar claims and 
extrapolation from the sample is a recognized method of proof'' for the 
United States in an affirmative case seeking recovery under a common-
law theory). See also Ratanasen v. California Dep't of Health Servs., 
11 F.3d 1467, 1469-71 (9th Cir. 1993) (collecting cases in which 
sampling and extrapolation have been approved in the Medicaid context, 
and ``join[ing] other circuits in approving the use of sampling and 
extrapolation as part of audits in connection with Medicare and other 
similar programs''); Chaves Cnty. Home Health Serv. v. Sullivan, 931 
F.2d 914, 917-23 (D.C. Cir. 1991). The authority to use sampling and 
extrapolation in Medicare audits is grounded in our statutory and 
regulatory authority to audit providers and recoup improper payments. 
See Chaves, 931 F.2d at 919 (interpreting the Medicare statute to allow 
for a ``sample adjudication procedure'' followed by extrapolation from 
that sample, which ``is reasonable given the logistical imperatives 
recognized by courts in other comparable circumstances'').
    Sampling and extrapolation have been used to calculate improper 
payments in Medicare FFS (Part A and Part B) for decades. CMS formally 
approved of this technique in 1986 (HCFA Ruling 86-1), but Medicare 
Administrative Contractors (MACs), which are responsible for 
determining medical necessity and paying Medicare FFS claims, have been 
using it ``at least since 1972.'' Chaves, 931 F.2d at 921; see id. at 
913 (explaining that ``sample adjudication has been used in previous 
instances involving post-payment review of `coverage determinations' 
under Part A,'' and that HCFA Ruling 86-1 ``simply reiterated [the 
agency's] belief that it had the latitude to employ sample audits on 
post-payment review to efficiently recoup overpayments for non-covered 
services''). In 1991, the United States Court of Appeals for the 
District of Columbia Circuit, in Chaves, upheld the use of this audit 
methodology against arguments that the Medicare statute required 
individualized review of claims submitted by providers (Id. at 922).
    The MMA imposed limits on the use of sampling and extrapolation in 
Medicare payment decisions in the context of Part A and Part B, when a 
settlement to resolve improper payments is not reached. Since 2003, 
Medicare Part A and Part B extrapolation under section 1893(f)(3) of 
the Act has been limited to instances in which the Secretary determines 
either that ``there is a sustained or high level of payment error'' or 
that ``documented educational intervention has failed to correct the 
payment error.'' No similar limitation applies to the MA program.

[[Page 6649]]

    As previously discussed, sampling and extrapolation is a generally 
accepted audit technique in the Medicare context, and the Act does not 
apply any limits to the use of extrapolation in the MA program. 
Therefore, we believe that CMS has the authority to implement this 
audit methodology in RADV audits for any case in which a RADV audit 
identifies improper risk-adjusted payments. We also believe that this 
is a reasonable approach to our RADV audits, given the sustained and 
high level of risk adjustment payment error, as previously described.
2. Summary of Proposed Rule
    In the 2018 proposed rule, CMS proposed to extrapolate contract-
level RADV audit findings using statistically valid random sampling 
techniques. CMS proposed to extrapolate findings in PY 2011 and all 
subsequent payment years, but specifically sought comment on how to 
treat the audits for PYs 2011, 2012, and 2013. In the proposed rule, we 
explained that we had conducted RADV audits for PYs 2011-2013 according 
to the sampling and extrapolation methodology described in the 2012 
methodology but that these audits were not yet finalized because we had 
not yet issued the audit findings to the MAOs.\26\ For PYs 2011 through 
2013, we estimated that audited MA contracts received $650 million in 
improper payments.
---------------------------------------------------------------------------

    \26\ See 83 FR 55038.
---------------------------------------------------------------------------

    In the 2018 proposed rule, we stated that, given the amount of 
improper payments identified under the MA program, interest in 
determining an accurate recovery amount for each audited MA plan, and 
importance of protecting the overall integrity of the program, we 
believed that it was in the public interest for CMS to apply the RADV 
payment error methodology(ies) adopted through this rulemaking to PY 
2011 and all subsequent years. We stated that CMS would be acting in 
compliance with the improper payment obligations under the Act (most 
recently updated as part of the Payment Integrity Information Act of 
2019 (PIIA)), as well as our fiduciary responsibility to recover funds 
due to the Medicare Trust Funds. We also noted that our February 2012 
publication put MAOs on notice that CMS expected to calculate a 
contract-level payment error for PY 2011 and subsequent payments years 
by extrapolating from its review of a statistically valid sample of 
enrollees, and that MAOs have never been entitled to receive or retain 
payments associated with HCCs that cannot be validated by medical 
records.
    We also proposed that MAOs would be required to remit extrapolated 
recovery amounts from RADV audit findings through CMS' payment system, 
the Medicare Advantage and Prescription Drug system (MARx), as offsets 
to MA plans' monthly capitation payments. In the event that the 
recovery amount exceeds the payment in one month, we proposed that the 
recovery would be spread across adjustments for multiple months until 
the full amount is recovered. We also proposed that CMS might likewise 
require MAOs to remit such recovery amounts based upon audit findings 
by the HHS-OIG.
    We explained in the 2018 proposed rule that CMS is not required to 
set forth the methodology for calculating an extrapolated payment error 
through regulatory provisions. However, we explained that, in the 
interest of transparency, we were choosing to inform MAOs about our 
plans to use various sampling and extrapolation methodologies in RADV 
audits, as CMS deems appropriate, through rulemaking.
    In addition to codifying in regulation our existing authority to 
use extrapolation techniques in the RADV context, we also used the 2018 
proposed rule as a means to gather public feedback on sampling 
methodologies that could be employed for purposes of extrapolation. We 
explained that, in addition to the contract-level approach described in 
the 2012 RADV Methodology, we have identified other potential 
methodologies for sampling and extrapolation that are based on a 
particular sub-cohort or sub-cohorts in a given payment year. For 
example, a sub-cohort could be the enrollees for whom a particular HCC 
or one of a related set of HCCs (such as the three diabetes HCCs) was 
reported.

                         Table 1--Diabetes HCCs
------------------------------------------------------------------------
                    HCC category description                       HCC
------------------------------------------------------------------------
Diabetes with acute complications..............................       17
Diabetes with chronic complications............................       18
Diabetes without complication..................................       19
------------------------------------------------------------------------

    After choosing an MA contract and a sub-cohort or sub-cohorts to 
audit, we would select a statistically significant sample of enrollees 
in the sub-cohort or sub-cohorts. After reviewing these enrollees' 
medical records that are submitted by the MAO, we would use statistical 
extrapolation to calculate and recoup the improper payments made to the 
audited MA contract for all enrollees in the sub-cohort or sub-cohorts 
in that payment year.
    We noted in the 2018 proposed rule that using a sub-cohort 
methodology, such as one focused on enrollees with high-risk HCCs, 
could allow us to use a much smaller sample size to calculate a 
statistically valid extrapolated improper payment amount. This is 
possible because, when selecting a sample from a smaller population 
(that is, a sub-cohort of enrollees), one can still achieve an 
acceptable level of statistical confidence with that smaller sample 
size. This sub-cohort-based audit methodology would also allow us to 
spread our audit resources across a wider range of MA contracts and 
focus on cohorts of enrollees that raise programmatic concerns, while 
also reducing operational burden on both CMS and the MAOs due to the 
reduced sample size needed to calculate improper payments.
    In the 2018 proposed rule, we invited comment on both the contract-
level audit methodology published in February 2012 and our proposal for 
an extrapolated audit methodology based on sub-cohorts of enrollees. We 
also sought comment on whether there are particular situations in which 
one methodology may be preferable to the other. We emphasized that 
neither proposed methodology was meant to displace our longstanding 
authority to audit the medical records of particular enrollees who we 
believe may be associated with improper payments or to use any 
statistically valid audit methodology. We also stated that, if we 
finalize one or more sampling and extrapolation methodologies through 
this rulemaking, we would announce any future changes to that 
methodology (or those methodologies) through the Health Plan Management 
System (HPMS).
    In addition, we stated that we may begin to conduct RADV audits for 
PYs 2014 and 2015 before finalizing the policies in the proposed rule, 
pursuant to our longstanding authority to review the medical records of 
any MA enrollee and recoup improper payments identified. We also sought 
comment on whether the use of sampling and extrapolation for certain 
payment years would require the exercise of our statutory authority to 
engage in retroactive rulemaking, as set out in section 1871(e)(1)(A) 
of the Act, which authorizes retroactive application of rules where 
``failure to apply the change would be contrary to the public 
interest.''
    We also discussed proposed changes to our RADV dispute and appeals 
regulations in 42 CFR 422.311 to conform with the finalized RADV 
provisions. Specifically, consistent with

[[Page 6650]]

our other proposed policies, we proposed to amend Sec.  422.311 by 
adding language to clarify that recovery of improper payments from MAOs 
will be conducted according to the Secretary's payment error 
extrapolation and recovery methodologies, and that CMS will apply 
extrapolation to RADV audits beginning with PY 2011. We also requested 
comment on whether to explicitly expand the MAOs' RADV appeal rights, 
such as by permitting appeal of the RADV payment error calculation 
methodology used in a RADV audit, similar to practices in Medicare FFS. 
A summary of the comments received and our responses follow.
3. Summary of Public Comments
    Comment: Several commenters supported CMS' proposal to use 
extrapolation in RADV audits, as well as our proposal to begin 
extrapolation for PY 2011 audits. Commenters indicated that this is the 
most effective way to address improper payments in MA.
    Response: We thank commenters for their support. While we plan to 
finalize our proposal to apply extrapolation to RADV audits, we are 
making a change to the years in which to apply extrapolation to achieve 
what we believe is an appropriate final policy that still takes into 
consideration our obligation to address potentially significant 
improper payments in the MA program. Extrapolation will now begin with 
the PY 2018 RADV audits rather than PY 2011, as proposed. This change, 
as further described in this section of this rule, is being made due to 
our fiduciary duty to protect taxpayer dollars from overpayments, 
certain operational considerations, and public comments on the 
timeliness of RADV audits.
    Comment: Several commenters opposed the use of extrapolation in 
RADV audits. Some commenters questioned whether we had the statutory 
authority to use sampling and extrapolation in RADV audits. These 
commenters suggested that, because section 1893(f)(3) of the Act grants 
CMS the authority to use sampling and extrapolation in certain 
circumstances when conducting audits in Medicare Part A and Part B, CMS 
cannot use those techniques in Part C audits without an equivalent 
grant of statutory authority.
    Several commenters challenged the statistical and methodological 
validity of both the contract-level sampling and extrapolation 
techniques described in the 2012 methodology, as well as an approach 
based on sub-cohorts of enrollees. A commenter stated that it is more 
difficult for plans to determine results from extrapolation in MA than 
in Medicare FFS because RADV audits can include the review of multiple 
medical records to validate one diagnosis from various providers with 
``disparate methods of documentation.''
    Some comments focused on the application of extrapolation beginning 
in PY 2011. Several commenters asserted that increased liabilities of 
MAOs from retroactive application of an extrapolated payment error 
recovery would deter future participation by MAOs in the MA program and 
reduce benefits to beneficiaries. Several commenters expressed concern 
that extrapolation for past payment years will destabilize physician 
care. Specifically, the concern is that providers participating in 
risk-sharing contracts with MAOs that have not yet completed a final 
settlement may be at risk for losses. The same commenters believe that 
recovering improper payments when the audit methodology has been 
revised several times is inequitable to the MAOs.
    Response: We appreciate these comments and considered them when 
finalizing the timing and content of these extrapolation policies. As 
discussed previously, CMS has the authority to use sampling and 
extrapolation in its RADV audits. Federal courts have held that 
sampling and extrapolation are a valid method of calculating improper 
Medicare payments, so long as statistically valid methods are used. The 
MMA added section 1893(f)(3) of the Act, which specifically applies to 
Medicare Part A and Part B and limits the use of extrapolation to 
determine overpayment amounts for recoupment under certain 
circumstances. This provision did not confer new authority to use 
extrapolation, but limited our preexisting audit authority in Medicare 
Part A and Part B. No similar limitation has been applied to audits in 
Medicare Part C. However, CMS will continue to focus its RADV efforts 
on MAOs identified as being at higher risk of improper payments.
    In the implementation of this authority to use sampling and 
extrapolation in RADV, CMS will employ statistical methods to determine 
statistically valid sample sizes, accurately identify payment error, 
and extrapolate to the universe of enrollees from which the sample is 
selected. These statistically valid methods may include applying one or 
more RADV audit methodologies for any given RADV audit. In addition, 
while CMS views extrapolation as a statistically valid methodology for 
RADV audits, the agency may, at times, use its discretion to not 
utilize extrapolation in a particular instance. For example, there may 
be unforeseen circumstances in which the statistical validity of the 
sample is disturbed (such as the need to exclude a large number of 
cases from the sample due to the loss of medical records in a natural 
disaster) and extrapolation is no longer possible, despite the initial 
intent to do so. There may be other limited instances in which CMS 
seeks to collect overpayments associated only with enrollees in a given 
sample, or wishes to perform only a probe sample of RADV reviews 
without the use of a statistically valid sample and yet will seek to 
recover any identified, non-extrapolated overpayments. The OIG may also 
independently decide not to extrapolate for reasons outside the control 
of CMS, and CMS will still recover those overpayments in accordance 
with the provisions in this final rule. To account for this, we are 
finalizing Sec.  422.311(a)(2) to read ``CMS may [emphasis added] apply 
extrapolation to audits for payment year 2018 and subsequent payment 
years,'' rather than ``CMS will apply extrapolation . . .'' as 
proposed. This language is not intended to signal that it would be a 
frequent occurrence to not extrapolate in PY 2018 and future audits; 
rather, extrapolation is expected to be the standard practice for RADV 
audits beginning in PY 2018.
    As previously stated, we believe that it is in the best interest of 
the Federal Government and our efforts to protect taxpayer dollars to 
extrapolate in our RADV audits, given the substantial amount of 
improper payments in MA and the fact that RADV is CMS' main corrective 
action used to address the submission of inaccurate diagnosis data. 
However, we also have decided not to extrapolate for PY 2011 through 
2017 audits, as originally proposed, due to certain operational 
considerations and public comments on the timeliness of RADV audits. 
The reasoning for this decision is discussed in greater detail later in 
this final rule.
    In addition, we do not agree with the comment that RADV audits 
include the review of multiple medical records with ``disparate methods 
of documentation.'' We reemphasize that the policies we are finalizing 
in this rule do not impose new documentation requirements on providers. 
The core component of a RADV audit is ensuring that all diagnoses 
reported to CMS are properly supported by medical record documentation. 
CMS' existing regulatory documentation standards, 42 CFR 422.310(d)(1); 
45 CFR 162.1002(c)(2) and (c)(3), including the RADV-specific authority 
to validate risk

[[Page 6651]]

adjustment data through the review of a sample of medical records at 
Sec.  422.310(e), remain unchanged under this final rule and are 
described in current agency policy documents, including the Medicare 
Managed Care Manual (with which MAOs agree, in their MA contracts, to 
comply). MAOs are also already required to ensure that contracted 
providers meet MA documentation requirements.
    We respectfully disagree with commenters' assertions that 
liabilities will increase. We are not imposing additional liabilities, 
penalties or retroactive application of new requirements or policy. We 
only seek to recover improper payments received by MAOs for HCCs that 
are not substantiated by enrollees' medical records. We continue to 
rely on existing program methods to establish auditing practices that 
encourage proper payment recovery consistent with established audit 
practices. We recognize that MAOs enter into agreements with providers, 
including those with a risk-sharing component, and we encourage all 
parties to those agreements to take steps to mitigate the submission of 
diagnosis codes that are not properly supported in the medical record.
    We emphasize that nothing in this rule changes the longstanding 
principle that a diagnosis code that is not documented in a patient's 
medical record is not a valid basis for CMS risk adjustment payments to 
an MAO. Nor does this rule change the longstanding obligation of an 
insurer to refund payments to CMS if it learns that a diagnosis lacks 
support in the beneficiary's medical record.
    Comment: Many comments were received on the proposed extrapolation 
methodologies, mainly focused on our proposed sub-cohort approach. Some 
commenters requested clarity on the sub-cohort methodology, while 
others expressed support for this methodology with various suggestions 
to improve it. Commenters questioned whether the proposed sub-cohort 
methodology will replace the existing contract-level methodology, which 
utilizes a general, non-targeted sampling methodology, and how CMS will 
determine which HCC groups will be used in the identification of sub-
cohorts. A commenter requested that CMS confirm whether RADV will 
consist of a single audit methodology or whether MAOs will be subject 
to multiple audit methodologies.
    Some commenters believe that applying a sub-cohort extrapolation 
methodology of enrollees would produce inaccurate results in RADV 
audits because of differences between plans with regard to size and 
risk characteristics. For example, several commenters argued that plans 
with a higher than average risk score are at increased risk for RADV 
audit because high-risk enrollees are more likely to have more HCCs. 
Other commenters believe that a small sample size, which CMS sees as a 
benefit of a sub-cohort methodology, will result in inaccuracies. 
Others commented that an extrapolation methodology based on sub-cohorts 
of enrollees would violate the statutory mandate of ``actuarial 
equivalence'' between payments made under MA and Medicare FFS because 
it would generate recoveries based on random outcomes without regard to 
specific characteristics of MA plans' diagnostic mix, enrollment size, 
and risk scores. A commenter requested that, if CMS adopts a sub-cohort 
extrapolation methodology, it uses a pilot period first before 
implementing the program on a large scale and extrapolating results.
    Other comments spoke to extrapolation methods more generally, 
including the appropriate confidence interval, potential for plans of 
certain sizes to be unduly chosen for RADV audits, and perceived 
inability to assess potential liability for RADV audits already 
performed if CMS abandons the extrapolation methodology set forth in 
the 2012 methodology.
    Other comments on our proposed extrapolation methodologies were 
focused on the impact of underpayments. A commenter objected to the 
RADV audit sampling methodology, arguing that it results in a purported 
payment recovery bias against MAOs. The commenter believes the results 
of the RADV audit sample are ``asymmetric,'' thus incorrectly 
representing the improper payment rate. More specifically, the 
commenter asserted that ``[t]hough there is no upper limit for how high 
the payment recovery amount can be, there is no balancing negative 
recovery amount.'' In other words, the commenter objected that MAOs 
cannot receive a payment from CMS based on a RADV audit if, overall, 
the risk scores should have been higher because, for instance, there 
were more supported diagnoses that had not been submitted (that is more 
under-coding) than unsupported diagnoses that had been submitted (that 
is over-coding). Other commenters shared these concerns, as well as 
voiced concern that RADV audit samples do not account for the reported 
bias that exists for enrollees who have no diagnosis codes submitted 
during the year but have existing documentation to support a diagnosis 
that could have been submitted. The same commenters perceive the audit 
methodology as being random and indiscriminate, believing that the 
results will incorrectly estimate the risk profile of enrollees.
    A commenter requested information related to the sampling 
methodology used to select enrollees for the PY 2014 RADV audit. 
Specifically, the commenter requested details on the development of the 
regression model used to predict payment error and on the sampling 
criteria from which the RADV audit currently extrapolates. This 
commenter also contended that the PY 2014 methodology appears to 
maximize the probability of selecting individuals with coding errors.
    Response: As previously explained, extrapolation is an established 
auditing practice and remains a valid method for addressing audit 
recoveries. In this final rule, we are clarifying the scope of our 
authority to strengthen the integrity of the MA program by identifying 
improper payments. Our initiatives are designed to ensure fair and 
accurate recovery efforts by focusing on the areas at highest risk of 
improper payments. We will use statistically valid methodologies to 
extrapolate improper payment findings to the universe of enrollees from 
which a sample is selected. These statistically valid methodologies may 
include applying one or more RADV audit methodologies for any given 
RADV audit. As previously discussed, we may also determine that 
extrapolation will not be applied in certain limited instances. We 
emphasize that, in this final rule, we are not adopting either the 
contract-level sampling and extrapolation technique described in the 
2012 methodology or a specific extrapolated audit methodology based on 
sub-cohorts of enrollees. Instead, for future RADV audits, CMS will 
rely on any statistically valid method for sampling and extrapolation 
that it determines to be well-suited to a particular audit. We 
described the sub-cohort methodology in the 2018 proposed rule to 
provide the industry with transparency on potential audit 
methodologies. In addition, while not required, CMS will continue to 
disclose our extrapolation methodology to MAOs through HPMS memos or 
other appropriate means, providing MAOs with the information sufficient 
to understand the means by which CMS extrapolated the improper payment 
determination.
    Any sampling and extrapolation methodologies adopted by CMS for 
RADV audits will be focused on MAO contracts and enrollees' HCCs that, 
through statistical modeling and/or data

[[Page 6652]]

analytics, are identified as being at highest risk for improper 
payments. This is an appropriate approach to any Federal MA audit that 
seeks to recoup taxpayer dollars that have been inappropriately paid to 
MAOs for diagnoses that are not supported in the medical record. This 
approach was also recommended by the GAO in a 2016 report titled 
``Fundamental Improvements Needed in CMS's Effort to Recover 
Substantial Amounts of Improper Payments.'' \27\ The GAO recommended 
that CMS ``modify [its] selection of contracts for contract-level RADV 
audits to focus on those contracts most likely to have high rates of 
improper payments by taking actions such as the following: selecting 
more contracts with the highest coding intensity scores; excluding 
contracts with low coding intensity scores; selecting contracts with 
high rates of unsupported diagnoses in prior contract-level RADV 
audits; if a contract with a high rate of unsupported diagnoses is no 
longer in operation, selecting a contract under the same MAO that 
includes the service area of the prior contract; and selecting some 
contracts with high enrollment that also have either high rates of 
unsupported diagnoses in prior contract-level RADV audits or high 
coding intensity scores.'' \28\
---------------------------------------------------------------------------

    \27\ GAO, at 26, https://www.gao.gov/assets/gao-16-76.pdf (April 
2016).
    \28\ Id.
---------------------------------------------------------------------------

    We also note that the purpose of RADV audits is to validate that 
diagnoses submitted by MAOs for risk-adjusted payment are properly 
supported by medical record documentation. See 42 CRF 422.310(e). RADV 
audits are the main corrective action used to address the submission of 
inaccurate diagnosis data. Occasionally, upon review of these medical 
records, CMS will uncover ``additional'' diagnoses supported by the 
medical records that were not submitted for payment by MAOs during the 
data collection period for enrollees selected in the sample. Under 
current contract-level RADV policy, when CMS uncovers these additional 
diagnoses that map to CMS-HCCs during medical record review of audited 
CMS-HCC(s), these newly-discovered diagnosis codes are used to 
recalculate risk scores in certain circumstances, which may result in 
an updated (reduced) improper payment calculation.
    MAOs are required by CMS regulations (Sec. Sec.  422.503 and 
422.504) and MAO contracts to establish compliance programs and 
processes to ensure accurate diagnosis coding and the submission of 
accurate diagnosis data. These processes should enable MAOs to identify 
not only instances where diagnoses submitted for risk-adjustment 
payment are not supported by the medical record, but also diagnoses 
that may not have been submitted to CMS. MAOs can submit additional 
diagnoses for risk-adjusted payment up until the final risk adjustment 
data submission deadlines described at Sec.  422.310(g)(2)(ii). As with 
overpayment recoveries under the Affordable Care Act and CMS's 
Overpayment Rule, the purpose of RADV audits is not to reopen 
submission deadlines and for CMS to make additional payments.\29\ RADV 
audits identify overpayments after the final risk adjustment data 
submission deadline.
---------------------------------------------------------------------------

    \29\ Section 6402 of the Affordable Care Act (Pub. L. 11-148) 
established section 1128J(d) of the Act. Under the Part C and D 
Overpayment Rule (79 FR 29844), which implemented section 6402 of 
the Affordable Care Act, MAOs are required to correct overpayments 
by self-reporting and returning payments associated with MAO 
diagnosis codes not supported by medical record documentation. 
Although MAOs are required to correct identified overpayments after 
the final risk adjustment data submission deadline in order for CMS 
to conduct reruns and recover the overpayments, MAOs are not 
permitted to submit additional diagnoses for payment after the 
submission deadline.
---------------------------------------------------------------------------

    Comment: Some comments were focused on the scope and number of 
plans selected for RADV audit. A commenter objected to an increase in 
the number of plans selected for the RADV audits. Another commenter 
requested an explanation of how sample sizes will be determined for 
Program of All-Inclusive Care for the Elderly (PACE) organizations, 
most of which have fewer than 500 enrollees.
    Response: As previously described, any extrapolation methodology 
adopted by CMS for RADV audits will be focused on MAO contracts that, 
through statistical modeling and/or data analytics, are identified as 
being at highest risk for improper payments. Examples of MAO contracts 
that may be deemed higher risk for the purposes of RADV audit selection 
are discussed later in this section. This is also the best approach to 
ensure that MAOs that do not show indications of being at high risk of 
improper payments are not exposed to audit burden to the exclusion of 
higher-risk plans. In addition, as noted previously, such an approach 
was recommended by the GAO in its April 2016 report.\30\ CMS does not 
currently subject PACE organizations to RADV audits and CMS' selection 
methodology for each year will describe any adjustments made for PACE 
or other low enrollment contracts.
---------------------------------------------------------------------------

    \30\ GAO, at 26, https://www.gao.gov/assets/gao-16-76.pdf (April 
2016).
---------------------------------------------------------------------------

    Comment: Several commenters noted that implementing these proposed 
policies would lead to more audit burden for providers because of an 
increase in documentation standards for treating providers. For 
example, commenters believe that this is a ``more stringent audit 
expectation'' that will increase administrative burden at a time in 
which there is already a physician shortage, thereby impacting 
patients. Another commenter contended that our extrapolation 
methodology should reflect that certain HCCs are more difficult to 
substantiate in medical record documentation than others.
    Response: RADV audits will not impose new documentation 
requirements on health care providers and, therefore, we believe there 
will be no additional audit impact on providers that contract with MAOs 
to provide services to MA plan enrollees. As previously stated, nothing 
in this rule changes the longstanding principle that a diagnosis code 
that is not documented in a patient's medical record is not a valid 
basis for CMS risk adjustment payments to an MAO. In addition, there is 
a longstanding requirement under Sec.  422.310(e), in place since the 
beginning of the MA program, that ``[MAOs] and their providers and 
practitioners will be required to submit a sample of medical records 
for the validation of risk adjustment data, as required by CMS,'' which 
is unaffected by this final rule. This requirement is consistent with 
longstanding requirements applicable to Medicare Part A and Part B 
providers that they furnish sufficient information to support payment. 
42 U.S.C. 1395(g) (Effective July 7, 2004) (``[No]. . . . payments 
shall be made to any provider unless it has furnished such information 
as the Secretary may request in order to determine the amounts due such 
provider . . .''); Clinic Res. Mgmt. v. Burwell, 2015 WL 3932657, at *2 
(S.D. Tex. June 26, 2015) (``The provider is responsible for 
maintaining and submitting adequate information to substantiate medical 
necessity and entitlement to payment.'').\31\
---------------------------------------------------------------------------

    \31\ Under section 1853(a)(3) of the Act, the Secretary must 
require MAOs to submit data regarding inpatient hospital services 
and other services, as well as other information as the Secretary 
deems necessary to calculate MA risk adjustment payments. This 
authority has been implemented at Sec.  422.310, which requires MAOs 
to submit ``data necessary to characterize the context and purposes 
of each item and service provided to a Medicare enrollee by a 
provider, supplier, physician, or other practitioner.'' Sec.  
422.310(b). MAOs must submit data that conforms to CMS' requirements 
for data equivalent to Medicare FFS data, when appropriate, and to 
all relevant national standards. The International Classification of 
Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) 
Guidelines for Coding and Reporting is the existing national 
standard. (See Sec.  422.310(d)(1); 45 CFR 162.1002(c)(2) and 
(c)(3)). This is consistent with obligations imposed on hospitals 
and providers in Medicare Parts A and B, who are required to furnish 
proper documentation and comply with the ICD Guidelines. See, for 
example, 42 U.S.C. 1395g and 1395n.

---------------------------------------------------------------------------

[[Page 6653]]

    Comment: A commenter contested CMS' proposal to recover contract-
level payment adjustments through a lump-sum reduction in the plans' 
monthly payments through MARx. The commenter noted that, for example, 
CMS currently makes retroactive, beneficiary-specific adjustments 
related to miscellaneous corrections to beneficiaries' status (such as 
eligibility, State and county of residence, date of death, etc.) 
outside of the RADV process. The commenter requested that CMS seek only 
beneficiary-level recoveries through RADV audits so as not to overlap 
with these non-RADV recoveries.
    Response: While we appreciate the commenter's consideration of the 
other areas in which CMS may make adjustments to MA payments, we do not 
believe that current and proposed RADV efforts overlap with non-RADV 
adjustments. RADV audits only validate diagnoses associated with a 
beneficiary's medical record documentation, not a beneficiary's 
demographic characteristics. If an HCC cannot be validated with medical 
records, MAOs are not entitled to the risk-adjustment payment 
associated with that HCC.
    Comment: Several commenters opposed the application of our 
extrapolation methodology to past payment years claiming that, pursuant 
to section 1853(b)(2) of the Act, this would be considered a 
retroactive application of policy and CMS must disclose our RADV audit 
methodology changes prior to any payment year RADV audit. Some 
commenters also asserted that the application of this rule to past 
payment years would alter the actuarial soundness of payments 
previously received by MA contracts, as existing contracts relied on 
the RADV audit methodology we announced in the 2012 RADV Methodology. 
Other commenters also characterized this approach as contrary to the 
Supreme Court's holding in Azar v. Allina Health Servs., 139 S. Ct. 
1804, 204 L. Ed. 2d 139 (2019), which emphasized that a substantive 
legal standard must go through a notice-and-comment process.
    Response: First, as a fundamental concept, this policy does not 
impose any new requirements on MAOs that could be construed as 
retroactive. The 2012 RADV Methodology did not create a different 
``documentation standard'' for MA plans than the standard that applies 
to traditional Medicare providers, nor did we state that an FFS 
Adjuster should set a permissible rate for the submission of erroneous 
codes. There is only one documentation standard for diagnosis coding, 
as discussed previously: proper medical record documentation is 
required for any reported diagnosis code to be valid. That is the 
consistent policy throughout the Medicare program (see previous 
discussion).
    The RADV auditing methodology has not fundamentally changed the 
longstanding requirement that a diagnosis submitted to CMS by an MAO 
for payment must be properly supported by medical record documentation. 
See UnitedHealthcare Ins. Co. v. Becerra, 16 F.4th 867, 869, 877 (D.C. 
Cir. August 13, 2021, reissued November 1, 2021), cert. denied, 142 S. 
Ct. 2851 (U.S. June 21, 2022) (No. 21-1140). Rather, it only enforces 
the well-established regulatory requirement that MA diagnoses be 
validated under that longstanding documentation standard. (For 
additional information, see Sec.  422.310(e); 83 FR 55037 (and 
authorities cited therein).)
    We also noted in the 2018 proposed rule that we may begin to 
conduct RADV audits for additional payment years (specifically, 2014 
and 2015) before this proposal is finalized, pursuant to our 
longstanding authority to review the medical records of any MA enrollee 
and recoup any improper payments identified.
    Even if this methodology was determined to be a retroactive 
application of policy, a position with which we do not agree, it is 
still necessary to comply with statutory requirements and is in the 
public interest for CMS to apply extrapolation to past payment years, 
and, therefore, is authorized under the Act. CMS has the authority, in 
accordance with section 1871(e)(1)(A) of the Act, to apply retroactive 
changes in regulations, manual instructions, interpretative rules, 
statements of policy, or guidelines of general applicability to items 
and services furnished before the effective date of the change, if the 
Secretary determines that ``such retroactive application is necessary 
to comply with statutory requirements or failure to apply the change 
would be contrary to the public interest.'' We believe that recovering 
extrapolated improper amounts is necessary to comply with statutory 
requirements and advances the public interest by protecting the overall 
integrity of the MA program. We have a statutory mandate under the PIIA 
to reduce improper payments and a fiduciary responsibility to recover 
funds due and owed to the Medicare Trust Funds.
    As previously discussed, HHS and the GAO have identified a 
significant volume of improper payments in the MA program,\32\ and RADV 
audits are the main way CMS ensures payment accuracy to MAOs. As 
further discussed in the Regulatory Impact Analysis section of this 
final rule, CMS estimates extrapolated improper payment recoveries of 
approximately $479 million per audit year beginning with the PY 2018 
audit. We also believe that there will be an additional sentinel effect 
of RADV audits on the improper payment rate as MAOs improve their 
processes to report only those diagnoses that meet CMS requirements for 
risk adjustment payment.
---------------------------------------------------------------------------

    \32\ For example, the FY 2021 HHS Agency Financial Report, pg. 
211, https://www.hhs.gov/sites/default/files/fy-2021-hhs-agency-financial-report.pdf, states that Part C Improper Payment 
Measurement (IPM) estimated approximately $15 billion in 
overpayments for calendar year 2019 risk-adjusted payments to MAOs.
---------------------------------------------------------------------------

    In addition, as discussed previously, RADV audits will not impose 
new documentation requirements on health care providers. The core 
component of a RADV audit is ensuring that all diagnoses are properly 
supported by medical records. We only seek to recover improper payments 
received by MAOs for HCCs that are not substantiated by enrollees' 
medical records. MAOs have never been entitled to receive or retain 
payments associated with HCCs that cannot be validated by medical 
records. Therefore, applying the rule under the public interest 
exception in section 1871(e)(1)(A) of the Act would not upset any 
settled or reasonable reliance interests. This all serves the public 
interest by reducing the improper allocation of taxpayer dollars that 
can otherwise be used for other purposes within the Federal Government, 
including solvency of the Medicare Trust Funds. Thus, applying the rule 
retroactively is necessary to comply with statutory requirements and in 
the public interest within the meaning of section 1871(e)(1)(A) of the 
Act.
    Comment: Several comments provided input on the potential 
promulgation of rules permitting administrative appeals of RADV audit 
methodology. A commenter opined that such procedures were unnecessary 
because stakeholders had an opportunity to participate in the 
development of our methodology through the notice-and-comment

[[Page 6654]]

rulemaking process, and that permitting challenges to our methodology 
in the administrative appeals context would generate ``numerous 
unnecessary practical problems'' for us. Another commenter supported 
the expansion of RADV audit appeals to allow MAOs to demonstrate that 
alternative methodologies would be more accurate, and to show that 
cohorts sampled for RADV audits might not be representative of the 
contract population.
    Response: We appreciate the commenters' input and concerns. We do 
not believe it would be appropriate to expand our appeals regulations 
to allow MAOs to appeal the RADV audit methodology, as revisions to the 
appeal regulations were not part of our proposed rule and stakeholders 
did not have the opportunity to provide comments on specific proposed 
policies. As such, MAOs will continue to be able to use the RADV 
appeals process currently set forth in Sec.  422.311. Any future 
changes to our appeals process would occur through separate notice and 
comment rulemaking.
    Comment: Several comments outside the scope of the proposed rule 
were received, including those related to the RADV program and other 
CMS programs. Out-of-scope comments pertaining to the RADV program 
included recommendations for changes to RADV documentation requirements 
and procedures; requests that CMS prohibit MAOs from auditing providers 
for patient records within the RADV cohort during the course of RADV 
audit; a request to expand the hardship exception to account for delays 
in acquiring medical records resulting from providers who are 
``traveling, sick, or deceased;'' a request to implement a schedule 
whereby RADV audits would be performed within 2 years of the applicable 
dates of service; challenges in collecting medical records created 
several years before the RADV audit; and requests for clarification of 
how CMS treats ``non-unique'' diagnosis codes during RADV audits when, 
even if one code is in error, there may be one or more diagnoses that 
substantiate the same HCC.
    Other out-of-scope comments pertained to the RADV dispute and 
appeals processes. These comments included requests for CMS to provide 
MAOs with more time to appeal a RADV audit finding; expand MAOs' appeal 
rights by removing the current limitation cited in Sec.  
422.311(c)(2)(iv) that allows MAOs, for each audited HCC, to appeal 
only one medical record that has undergone a RADV review; use an 
independent third party to reconsider disputed HCCs and/or payment 
error calculations; allow additional flexibility in disputing medical 
record interpretation during the appeals process and for MAOs to 
supplement medical records with documents that could not be obtained at 
the time of the audit; and allow MAOs to file complaints of 
underpayments by CMS.
    Other comments were received unrelated to RADV, such as requests to 
make burden-reducing changes to the Medicare Part C Recovery Audit 
program requirements and requests for payment parity between MA and 
Medicare FFS.
    Response: While we appreciate this feedback, these comments do not 
directly relate to the proposed changes to the RADV audit program, 
which is focused on our policies related to the use of extrapolation 
and the non-application of an FFS Adjuster, and are therefore outside 
of the scope of this final rule. Updated resources on RADV rules and 
methodologies are available on the CMS website at https://www.cms.gov/Research-Statistics-Data-andSystems/Monitoring-Programs/Medicare-Risk-Adjustment-Data-Validation-Program/Resources. We also encourage 
stakeholders to engage with CMS throughout the course of an audit cycle 
and to provide feedback on programmatic improvements that can be 
considered outside of this rulemaking process.
4. Summary of Final Policies
    After consideration of public comments, we are finalizing the use 
of extrapolation under the contract-level RADV program. However, we are 
modifying our proposed policy to extrapolate beginning in PY 2011. We 
are instead finalizing our ability to extrapolate beginning in PY 2018 
due to considerations of appropriateness in light of public comments 
and certain operational concerns, as well as our obligations to protect 
the sustainability of the Medicare program. We are announcing, through 
this final rule, our interpretation of our statutory and regulatory 
authority as authorizing the use of sampling and extrapolation in RADV 
audits. We are not adopting any particular statistical sampling 
methodology in this final rule. As previously noted, CMS will use 
statistically valid methods for sampling and extrapolation that we 
determine to be well-suited to a particular RADV audit.
    After reviewing comments and considering the matter further, we 
also believe that the use of sampling and extrapolation to calculate 
audit recoveries would not be retroactive within the requirements of 
section 1871(e)(1)(A) of the Act. The use of sampling and extrapolation 
for prior payment years is not retroactive because the substantive 
requirement of proper medical record documentation of all diagnoses 
submitted for payment remains unchanged, whether we calculate audit 
recoveries on an enrollee-by-enrollee basis or use a statistically 
valid sample of enrollees to extrapolate. Enrollee-level audit 
recoveries and extrapolated audit recoveries are simply two different 
ways of enforcing the same medical record documentation requirement 
under Sec.  422.310(e).
    While we believe that the use of sampling and extrapolation for 
prior payment years is not a retroactive application of policy, even if 
it was somehow interpreted as retroactive, we still believe that 
recovering extrapolated improper amounts is necessary to comply with 
statutory requirements and advances the public interest by protecting 
the overall integrity of the MA program. We have a statutory mandate 
under the PIIA to reduce improper payments and a fiduciary 
responsibility to recover funds due and owed to the Medicare Trust 
Funds. The RADV program was developed as one of the primary methods to 
address CMS' responsibility to recover improper payments in the MA 
program.
    In addition, although we stated in the proposed rule that we 
intended to apply any finalized RADV payment error methodology or 
methodologies to PY 2011 and all subsequent years, we have decided to 
begin to exercise our authority to collect extrapolated recoveries with 
the PY 2018 RADV audit. Based on our review of a number of factors, CMS 
determined it is in the overall best interests of the RADV program and 
ultimately the Part C program itself to limit all RADV improper payment 
recoveries for PYs 2011 through 2017 to enrollee-level adjustments for 
those enrollees sampled in the payment validation audits. Our reasoning 
for this decision follows.
    First, after careful consideration of the comments received, we 
believe that the most appropriate decision is to begin extrapolation 
with the PY 2018 audits. As a result, CMS will not collect extrapolated 
overpayments identified as a result of either CMS RADV or HHS-OIG 
audits for payment years prior to PY 2018, but will collect enrollee-
level overpayments identified in those audits. As previously described, 
we believe that beginning extrapolation for PY 2018 RADV audits 
represents an appropriate policy because it recognizes our

[[Page 6655]]

fiduciary duty to protect taxpayer dollars from overpayments and 
preserves our ability to collect on significant (extrapolated) amounts 
of overpayments made to plans beginning in PY 2018. This final rule 
will also allow CMS to focus on conducting future RADV audits as soon 
as practicable after an MAO payment year concludes, which was the topic 
of significant public comment to the proposed rule.
    Lastly, we have determined that it is in the best interest of all 
parties to ensure that the contract-level RADV appeals process, which 
is also outlined in regulation, is able to successfully process all 
RADV appeals. By not using an extrapolation methodology prior to PY 
2018, we expect to better control the total number of active appeals 
that are submitted in the first few years following finalization of 
this rule, which will alleviate burden on MAOs and CMS. This includes 
appeals that result from CMS RADV audits, as well as CMS recoveries 
made based upon improper payments identified in HHS-OIG audits of MAOs. 
When this rule is finalized, we will begin issuing the enrollee-level 
audit findings from the CMS RADV audits that have been completed (that 
is, CMS RADV audits for PYs 2011 through 2013, followed eventually by 
PY 2014 and PY 2015 audits), as well as recovering enrollee-level 
improper payments identified in HHS-OIG audits. The release of these 
results in quick succession could result in an unprecedented influx of 
MAO appeals into the RADV appeals process. HHS' past experience with 
appeals backlogs, particularly for Medicare FFS claims, has 
demonstrated that proactive steps to avoid large volumes within an 
abbreviated period of time is key to ensuring the timely processing of 
all appeals. Depending upon the number of RADV audit appeals filed by 
plans, there may be a possible appeals backlog that could lead to 
significant burden on MAOs and CMS. It can also divert government 
resources away from other important activities that could also reduce 
MAO burden, such as finding ways that RADV audits can be performed in 
quicker succession to the conclusion of any payment year reconciliation 
period, resulting in future RADV audits being more contemporaneous, 
which was the topic of significant public comments to the proposed 
rule.
    At the same time, this finalized policy also recognizes our 
fiduciary duty to protect taxpayer dollars from overpayments and 
preserves our ability to collect on significant (extrapolated) amounts 
of overpayments made to plans beginning in PY 2018. We understand that 
this decision means that certain amounts of improper payments will be 
left uncollected in those earlier payment years (PYs 2011 through 2017) 
because we will only be collecting the non-extrapolated improper 
payments identified for PYs 2011 through 2017 and not the extrapolated 
overpayments that we will be collecting for PY 2018 and subsequent 
payment years. However, for the reasons previously described, we 
believe that the overall long-term success of the RADV program and 
ultimately the Part C program requires us to consider several issues 
and balance the collection of extrapolated improper payments with the 
practical realities of the current RADV program.
    We are finalizing our RADV regulations as proposed, with the 
exception of a change to the payment year in which extrapolation will 
begin. Specifically, we are--
     Revising Sec.  422.300 to include ``collection of improper 
payments;''
     Amending Sec.  422.310(e) to announce that extrapolation 
may be applied in RADV audits for PY 2018 forward and by adding a 
requirement for MAOs to remit improper payments based on RADV audits in 
accordance with a manner specified by CMS; \33\ and
---------------------------------------------------------------------------

    \33\ See discussion regarding the use of ``may'' in Sec.  
422.310(e). This language is not intended to signal that it would be 
a frequent occurrence to not extrapolate in PY 2018 and future 
audits; rather, extrapolation is expected to be the standard 
practice for RADV audits beginning in PY 2018. This will allow CMS 
with flexibility to not extrapolate in certain limited instances the 
Agency determines to be appropriate.
---------------------------------------------------------------------------

     Amending Sec.  422.311 by clarifying that recovery of 
improper payments from MAOs will be conducted according to the 
Secretary's payment error extrapolation and recovery methodologies and 
that CMS may apply extrapolation to RADV audits for PY 2018 and 
subsequent payment years.
    While we appreciate the comments received as to potential 
expansions of MAO appeals rights, we are not finalizing any other 
changes to the RADV appeals process as part of this final rule because 
no specific appeals-related policies were proposed.

B. Fee-For-Service Adjuster

1. Description of an FFS Adjuster
    As previously described, risk adjustment ensures that MAOs are paid 
appropriately for their plan enrollees, and section 1853(a)(1)(C) of 
the Act requires that we calculate risk-adjusted payments to MAOs based 
on specific criteria, such as age, disability status, gender, 
institutional status, and health status. As discussed earlier, MAOs' 
payments are calculated using the CMS-HCC risk adjustment model, which 
is published each time it is updated (see section 1853(b) of the 
Act).\34\ Additionally, an MAO may only report a diagnosis, and claim 
the associated payment, when that diagnosis is properly supported by 
the beneficiary's medical records. Medical records properly support a 
reported diagnosis when they comply with all CMS data and documentation 
requirements, which are described in current agency policy documents, 
including Chapter 7 of the Medicare Managed Care Manual.\35\ Plans are 
also required to submit a sample of medical records for the validation 
of this risk adjustment data (see 42 CFR 422.310(e)).
---------------------------------------------------------------------------

    \34\ https://www.cms.gov/Medicare/Health-Plans/MedicareAdvtgSpecRateStats/Announcements-and-Documents.
    \35\ https://www.cms.gov/Regulations-and-Guidance/Guidance/Manuals/Downloads/mc86c07.pdf.
---------------------------------------------------------------------------

    Some MAOs have suggested that CMS cannot lawfully enforce the 
requirement of medical record documentation for diagnosis codes while 
making payments at the published rates. These MAOs argue that there is 
a difference in auditing standards between Medicare FFS and MA 
diagnosis data. In contrast to the MAO-submitted diagnosis data, these 
MAOs claim that Medicare FFS data is ``unaudited'' by CMS and 
presumably contains erroneous diagnosis codes not properly supported by 
beneficiaries' medical records. As a result, they argue, the Medicare 
FFS data used to calculate MAO payments will understate the cost of 
treating various conditions. To address the presence of erroneous 
diagnoses in the FFS claims data used to calibrate the MA payment 
model, MAOs argue that CMS must raise payment rates to MAOs or relax 
the documentation standard that CMS applies to reported medical 
diagnoses to ensure accurate payments. MAOs refer to this concept of a 
proposed adjustment to the payment rates and/or documentation standard 
for MAOs as an ``FFS Adjuster.'' These MAOs ground their arguments in 
section 1853(a)(1)(C)(i) of the Act, which requires the Secretary to 
adjust payments to MAOs for demographic and health related risk factors 
so as to ensure ``actuarial equivalence.'' According to these MAOs, an 
FFS Adjuster would either adjust payment rates (by raising them) or 
adjust documentation standards (by loosening them) to resolve the 
alleged incompatibility between the current rates and current 
documentation standards. In the 2012 methodology, using the term 
somewhat differently,

[[Page 6656]]

CMS said that it would ``apply a Fee-for-Service Adjuster (FFS 
Adjuster) amount as an offset to the preliminary recovery amount'' 
calculated for RADV audits under that methodology.
2. Summary of 2018 Proposed Rule
    In the 2018 proposed rule, we proposed not to include the FFS 
Adjuster described in the 2012 methodology in any final RADV payment 
error methodology. We stated that a study that we conducted found that 
errors in Medicare FFS claims data do not lead to systematic payment 
error in the MA program and that, even if there was evidence of 
systematic payment error, it would be inequitable to only correct 
payment errors made to audited contracts. We sought comment on our 
proposal not to use an FFS Adjuster. We also sought comment in our June 
28, 2019 Federal Register notice and request for additional comment (84 
FR 30983) regarding how the statutory minimum levels of the coding 
pattern adjustment set at section 1853(a)(1)(C)(ii) of the Act bear on 
the issue of whether or not to apply an FFS Adjuster.
3. Summary of Public Comments
    We received numerous comments regarding our proposal to not include 
an FFS Adjuster in RADV.
    Comment: Several commenters expressed support for CMS' proposal not 
to apply an FFS Adjuster, including the Medicare Payment Advisory 
Commission (MedPAC).\36\ These commenters discussed the study results 
demonstrating that errors in FFS Medicare claims data do not 
systematically bias MA risk scores, and said that if such bias existed, 
applying an FFS Adjuster to RADV would not be the appropriate remedy to 
address that bias because only a small number of MA plans undergo RADV 
audits each year. These commenters further asserted that any potential 
bias from undocumented FFS diagnoses is negligible and that the 
application of an FFS Adjuster would require significant effort for 
negligible benefit.
---------------------------------------------------------------------------

    \36\ https://www.medpac.gov/wp-content/uploads/import_data/scrape_files/docs/default-source/comment-letters/08122019_medpac_ma_radv_comment_v3_sec.pdf.
---------------------------------------------------------------------------

    Response: We thank these commenters for their support of not 
applying an FFS Adjuster to the RADV methodology. We agree with these 
comments for the reasons described throughout this final rule.
    Comment: Some commenters contended that an FFS Adjuster is required 
to ensure ``actuarial equivalence'' between payments to MA plans and 
payments under the Medicare FFS program. Some commenters also contended 
that the ``same methodology'' provision of section 1853(b)(4)(D) of the 
Act requires the application of an FFS Adjuster in RADV. Other 
commenters argued that CMS needs to apply an FFS Adjuster to comply 
with the district court's holding in UnitedHealthCare Insurance Co. v. 
Azar, 330 F. Supp. 3d 173 (D.D.C. 2018), rev'd sub nom. 
UnitedHealthcare Insurance Co. v. Becerra, 16 F.4th 867 (D.C. Cir. 
August 13, 2021, reissued November 1, 2021), cert. denied, 142 S. Ct. 
2851 (U.S. June 21, 2022) (No. 21-1140). A commenter requested that CMS 
suspend ongoing RADV audits and not begin any new RADV audits until an 
FFS Adjuster is developed for use in RADV audits and in MAOs' 
calculations of improper payments.
    Response: As a general matter, we believe that it is in the best 
interest of the Federal Government and taxpayers for CMS to continue 
RADV audits for the purpose of addressing the high dollar amounts of 
improper payments, as well as to employ a RADV methodology that does 
not include the application of an FFS Adjuster. Further, the 
``actuarial equivalence'' requirement under section 1853(a)(1)(C) of 
the Act and ``same methodology'' provision under section 1853(b)(4)(D) 
of the Act do not require the use of an FFS Adjuster. First, as 
described by the D.C. Circuit, these provisions do not apply to the 
obligation to return improper payments for MAO diagnosis codes that are 
unsupported by medical records. Although the D.C. Circuit did not 
address the RADV audit context in its decision in UnitedHealthcare, 
this position is consistent with the D.C. Circuit's reasoning in that 
case. (See UnitedHealthcare, 16 F.4th at 869, 891-92.) Second, it would 
be unreasonable to interpret the Act as requiring a minimum reduction 
in payments in one provision (the coding pattern provision), while at 
the same time prohibiting CMS in an adjacent provision (the actuarial 
equivalence provision) from enforcing those longstanding documentation 
requirements (by requiring an offset to the recovery amount calculated 
for CMS audits). (Section 1853(a)(1)(C)(ii) of the Act requires a 
minimum coding pattern adjustment to reduce the risk scores of all MA 
beneficiaries, and therefore, MA payment rates. Such a minimum coding 
pattern adjustment accounts for differences in coding patterns between 
MA and Medicare FFS, given that MAOs have a greater incentive than FFS 
providers to report diagnoses.) These points are further explained 
later in this section.
    The first basis 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. This position is consistent with the D.C. Circuit's decision 
in UnitedHealthcare. There, a group of MAOs challenged the Secretary's 
Part C Overpayment Rule (the ``Overpayment Rule'') (79 FR 29844), which 
implemented section 6402 of the Affordable Care Act and required MAOs 
to self-report and return payments associated with MAO diagnosis codes 
not supported by medical record documentation. The district court 
invalidated the Overpayment Rule. UnitedHealthcare, 330 F. Supp. 3d at 
192.
    However, the D.C. Circuit reversed the district court, holding that 
the actuarial equivalence provision applies only to how CMS risk 
adjusts the payments it makes to MAOs, and not to the obligation of 
MAOs to return improper payments for diagnosis codes, submitted by MAOs 
to CMS, lacking medical record support. (See UnitedHealthcare, 16 F.4th 
at 883-887.) The D.C. Circuit also held that even if the actuarial 
equivalence provision applied, plaintiffs' claims would still fail 
because they did not meet their burden in showing, either through 
empirical evidence or persuasive logic, that application of the 
Overpayment Rule would lead to systematic underpayment of MAOs. (Id. at 
887 through 891.)
    While the D.C. Circuit decision pertained only to the Overpayment 
Rule and declined to address RADV audits, its reasoning applies just as 
strongly in the RADV context and supports our conclusion that an FFS 
Adjuster is not appropriate in a RADV audit. ``The role of the 
actuarial-equivalence provision is to require CMS to model a 
demographically and medically analogous beneficiary population in 
traditional Medicare to determine the prospective lump-sum payments to 
[MAOs].'' (Id. at 870.) The RADV program, like the Overpayment Rule, 
applies after the fact to require MAOs to refund any payment to which 
they are not entitled, based on diagnoses that lack support in the 
medical record. The purpose of RADV audits is to recover payments that 
were made improperly based on diagnoses not supported by medical record 
documentation. If a payment is made to an MAO based on a diagnosis code 
not supported by

[[Page 6657]]

medical record documentation, the entire payment for that code is in 
error and should be recovered in full because the payment standard has 
not been met. RADV audits only address issues relating to diagnoses 
that are not supported by valid medical record documentation.
    Comment: Several commenters expressed concern that our proposal to 
extrapolate without applying an FFS Adjuster to payment recoveries 
achieved through RADV audits will overlap with coding pattern 
adjustments or create a double-recovery by CMS.
    Response: Section 1853(a)(1)(C)(ii) of the Act requires the 
implementation of a minimum coding pattern adjustment to reduce risk 
scores of all MA beneficiaries, and therefore MA payment rates. This 
minimum coding pattern adjustment accounts for differences in coding 
patterns between MA and Medicare FFS, given that MAOs have a greater 
incentive than FFS providers to report diagnoses. To meet this 
requirement, each year, CMS has implemented an adjustment to offset the 
effects on MA risk scores of higher levels of coding patterns in MA 
relative to FFS. (See section 1853(a)(1)(C)(ii) of the Act.) Under 
section 1853(a)(1)(C)(ii)(III) of the Act, the minimum adjustment 
factor for 2019 and each subsequent year is 5.90 percent. CMS has, each 
year, implemented the minimum coding pattern adjustment reduction 
required by statute.
    As CMS has explained in its annual MA advance notices and rate 
announcements, the coding pattern adjustment, unlike RADV, is not 
intended to address unsupported or inaccurate codes reported by MAOs in 
particular instances but only the general practice, relative to 
Medicare FFS, of reporting codes with greater intensity, including 
codes that are nonetheless accurate.\37\ Contrary to some commenters' 
assertions, the coding pattern adjustment provision of the statute 
actually supports our decision not to apply an FFS Adjuster, and we 
rely on that conclusion here as a second basis for our decision not to 
apply an FFS Adjuster. We briefly review the history of that provision:
---------------------------------------------------------------------------

    \37\ Announcement of Calendar Year (CY) 2011 Medicare Advantage 
Capitation Rates and Medicare Advantage and Part D Payment Policies 
and Final Call Letter at 19 (April 5, 2010); see also Announcement 
of Calendar Year (CY) 2017 Medicare Advantage Capitation Rates and 
Medicare Advantage and Part D Payment Policies and Final Call Letter 
at 54 (April 4, 2016); Announcement of Calendar Year (CY) 2012 
Medicare Advantage Capitation Rates and Medicare Advantage and Part 
D Payment Policies and Final Call Letter at 37-38 (April 4, 2011).
---------------------------------------------------------------------------

     The coding pattern adjustment was enacted as part of the 
Deficit Reduction Act of 2005. (Pub. L. 109-171 (February 8, 2006), 
codified at 42 U.S.C. 1395w-23(a)(1)(C)(ii)(I) and (II).)
     The Health Care and Education Reconciliation Act of 2010 
(Pub. L. 111-152), section 1853(a)(1)(C)(ii) of the Act was amended to 
require that the adjustment be at least 4.71 percent in 2014, rising 
annually to at least 5.7 percent in 2019. (Pub. L. 111-152, tit. I, 
subtit. B, section 1102(e), 124 Stat. 1046.) (For payment years 2010 to 
2013, CMS applied a 3.41 percent adjustment.\38\)
---------------------------------------------------------------------------

    \38\ Announcement of Calendar Year (CY) 2010 Medicare Advantage 
Capitation Rates and Medicare Advantage and Part D Payment Policies 
at 19-29 (April 6, 2009).
---------------------------------------------------------------------------

     Section 1853(a)(1)(C)(ii)(III) of the Act was subsequently 
amended again in the American Taxpayer Relief Act of 2012 to require 
the Secretary to make a reduction of at least 4.91 percent in 2014, 
rising to at least 5.9 percent by 2019. (Pub. L. 112-240, tit. VI, 
subtit. C, section 639, 126 Stat. 2357.)
    CMS audits reinforce longstanding documentation requirements. We 
believe it would be unreasonable to interpret the Act as requiring a 
minimum reduction in payments in one provision (the coding pattern 
provision), while at the same time prohibiting CMS in an adjacent 
provision (the actuarial equivalence provision) from enforcing those 
longstanding documentation requirements (by requiring an offset to the 
recovery amount calculated for CMS audits). To the contrary, because 
the Act requires CMS to reduce payments to MAOs by at least a specific 
minimum percentage, the only reasonable interpretation of the Act is 
that CMS would pay MAOs at those reduced rates, under the existing 
payment model,\39\ and enforce the longstanding documentation 
requirements through CMS' audits.
---------------------------------------------------------------------------

    \39\ Any changes to the CMS-HCC payment model are published in 
the annual payment notice.
---------------------------------------------------------------------------

    Comment: Several comments disputed our suggestion that addressing 
any diagnosis error in FFS Medicare claims through a RADV FFS Adjuster 
would introduce inequities between plans that are audited and plans 
that are not audited. Specifically, commenters discussed that not 
applying an FFS Adjuster would be a disadvantage to the MA plans 
selected for RADV audits because the audited plans are held to a 
higher, inappropriate standard of medical documentation than unaudited 
plans.
    Response: As we stated in the proposed rule, the purpose of RADV 
audits is to recover improper payments resulting from diagnoses that 
are not supported in the medical record documentation, which is a 
longstanding documentation standard that applies to all plans equally 
and regardless of whether the plan is subject to a RADV audit. The 
objective of an audit is to promote fair and impartial recovery of 
improper payments due to insufficient documentation in accordance with 
regulations. As we stated in the proposed rule, 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. We also do not intend for this conclusion to suggest that we 
believe an FFS Adjuster is appropriate or necessary outside of the RADV 
context.
    Our position is consistent with the conclusion of the D.C. Circuit, 
which is that the actuarial-equivalence requirement is not an 
``entitle[ment] . . . to a precise payment amount'' for a Medicare 
Advantage insurer, but only ``an instruction to the Secretary regarding 
the design of the risk adjustment model as a whole . . . describ[ing] 
the type of `payment amount[s]' that the risk adjustment model should 
produce''; ``[i]t does not directly govern how CMS evaluates the 
validity of diagnoses or defines `overpayment.' '' (UnitedHealthcare, 
16 F.4th at 885-86).
    Comment: Several commenters asserted that moving forward without an 
FFS Adjuster would render the RADV auditing requirements flawed, 
unclear, stringent and unrealistic, and increase the burden placed on 
providers to ensure accuracy as a result. Specifically, commenters 
believe this ``more stringent audit expectation'' during a physician 
shortage would not serve the public interest and would be detrimental 
to the MA program. A commenter argued that increased auditing 
requirements for MA providers would be contrary to CMS' other efforts 
focused on reducing unnecessary provider burden. Other commenters also 
noted burden for patients, while others believe that this policy will 
have a disproportionate impact on smaller, not-for-profit special needs 
plans with fewer resources to pay audit recoveries.
    Response: This final rule does not impose a new documentation 
standard on MA providers, nor is there a distinction in the 
documentation standards between the MA and FFS Medicare programs. 
Section 1815(a) of the Act (Medicare Part A) states that ``no such 
payments shall be made to any

[[Page 6658]]

provider unless it has furnished such information as the Secretary may 
request in order to determine the amounts due such provider under this 
part for the period with respect to which the amounts are being paid or 
any prior period.'' Additionally, Section 1833(e) of the Act (Medicare 
Part B) states that ``[n]o payment shall be made to any provider of 
services or other person under this part unless there has been 
furnished such information as may be necessary in order to determine 
the amounts due such provider or other person under this part for the 
period with respect to which the amounts are being paid or for any 
prior period.'' Section 1172 of the Health Insurance Portability and 
Accountability Act (HIPAA) (Pub. L. 104-191) also requires both 
providers and health plans to use standard content, formats, and coding 
for health care transactions. In addition, the Secretary has adopted 
various organizations' formats and code sets, including the ICD-10 and 
the ICD Guidelines, which is the national standard for both FFS and MA. 
See 45 CFR 162.1002. CMS has always required proper medical record 
documentation in order for any reported diagnosis code or claim to be 
valid. (See, for example, Becerra, 16 F. 4th at 869 (``[n]either 
Congress nor CMS has ever treated an unsupported diagnosis for a 
beneficiary as valid grounds for payment to a Medicare Advantage 
insurer'').) That is the consistent policy throughout the Medicare 
program, including MA and FFS.\40\ (See 42 CFR 422.310 (``MA 
organizations must submit data that conform to CMS' requirements for 
data equivalent to Medicare fee-for-service data, when appropriate, and 
to all relevant national standards.'').) As such, we do not believe 
that RADV audits impose any new level of burden on providers or violate 
any initiatives to reduce that burden.
---------------------------------------------------------------------------

    \40\ FFS Medicare claims are subject to error correction and 
payment adjustment when they are based on diagnosis codes not 
supported by the medical record. See Medicare Program Integrity 
Manual sections 3.3.1.1, 3.3.2.1, 3.6.2.4, 6.5.2, 6.5.3., https://www.cms.gov/Regulations-and-Guidance/Guidance/Manuals/internet-Only-Manuals-IOMs-Items/CMS019033.
---------------------------------------------------------------------------

    This rule, rather than the 2012 methodology, will govern CMS' 
conduct of RADV audits. Nonetheless, we did not intend the 2012 
methodology to suggest that contract-level RADV audits create a 
different ``documentation standard'' for MAOs than the standard that 
applies to traditional Medicare providers, or that any FFS Adjuster 
should set a permissible rate for the submission of invalid diagnosis 
codes. After a lengthy consideration of these issues, and more than a 
decade of additional experience with the Medicare Advantage program, we 
have decided not to apply an FFS Adjuster in RADV audits because: (1) 
we believe, consistent with the D.C. Circuit's decision in 
UnitedHealthcare, 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 of MAOs to return improper payments 
(that is, payments for unsupported diagnosis codes); and (2) it would 
not be reasonable to read the Act as requiring a reduction in payments 
to MAOs by a statutorily-set minimum adjustment in the coding pattern 
adjustment, while at the same time prohibiting CMS from enforcing 
longstanding documentation requirements by requiring an offset to the 
recovery amounts calculated for CMS audits.
    Comment: A commenter opined that the cost to stakeholders of 
extrapolating payment error recoveries without an FFS Adjuster 
outweighed any benefits to the rule. The commenter noted that CMS' 
analysis of the regulatory impact in the proposed rule ignored changes 
in MA bids, including reduced or eliminated product availability, 
increased administrative costs to MAOs for auditing provider medical 
record documentation and coding, and the cost of responding to RADV 
audits. Other commenters argued that extrapolation, along with the 
elimination of the FFS Adjuster, would threaten the MA program more 
generally through consequences on the bidding process, reduced 
incentives for cost savings, reduced benefits to enrollees, and 
increased premiums. A commenter requested that CMS consider that 
selecting contracts that represent a disproportionate amount of an 
MAO's business for RADV audits may drive smaller organizations out of 
the MA program.
    Response: It is our objective to strengthen the MA program by 
ensuring that the payments received by MAOs are accurate and that the 
Federal Government recovers any funds, representing taxpayer dollars, 
to which an MAO was not entitled. Our RADV audit methodology, which 
will not include an FFS Adjuster, should not have any material impact 
on MAOs' bidding practices or offerings because any funds recovered 
under RADV would be for payments to which the MAO was never entitled. 
Consistent with a prior GAO recommendation to focus on MAO contracts 
most likely to have high rates of improper payments, we have also 
shifted our RADV approach from a largely untargeted, random sampling 
from a universe of most of an audited MAOs' enrollees to a more 
targeted, risk-based approach that incorporates risk factors, such as 
HCCs that were more likely to be in error. This current approach 
enables the Federal Government to focus its limited auditing resources 
on areas where improper payments are more likely to be found, and 
reduces audit burden on those MAOs that are not at high risk of 
improper payments. We believe, for example, that MAOs that implement 
meaningful steps to reduce the reporting of unsupported diagnoses will 
be less likely in the future to be chosen for a CMS RADV audit because 
the indicators of potential improper payment risk will be greatly 
reduced in the risk adjustment data.
    Comment: A commenter requested that CMS withdraw the proposed RADV 
provisions and develop a new audit procedure in concert with industry 
stakeholders. Several commenters noted that CMS has announced no plans 
to address FFS Medicare diagnosis errors in the original payments to 
plans. These commenters assert that CMS' failure to provide a general 
adjustment for payment bias does not justify our proposal not to apply 
an FFS Adjuster for audited plans.
    Response: We believe these comments are outside the scope of the 
proposed rule's provisions. The RADV program enforces the longstanding 
medical record documentation regulatory requirement as it relates to 
risk adjustment, not the analyses performed to determine the risk 
adjustment coefficients used to calculate risk scores, and thus risk-
adjusted payments. It would be inappropriate to address these 
determinations and calculations via this final rule's RADV payment 
error methodology.
    Comment: Several commenters requested that we provide additional 
disclosures of information related to our FFS Adjuster study to enhance 
transparency, some arguing that the Information Quality Act (Pub. L. 
106-554) requires disclosure of such materials. For example, a 
commenter requested copies of the medical records reviewed during the 
FFS Adjuster study and diagnostic coding protocols followed by 
reviewers, citing the Information Quality Act as the justification for 
this request. Another stated that additional data is needed in order to 
provide a meaningful response, such as the HCCs mapped from diagnoses 
on the claims from Medicare FFS data. A commenter argued that the RADV 
provisions violated the Administrative Procedures Act (APA) due to the 
disclosure of insufficient

[[Page 6659]]

methodology or data to support these policies. Another criticized the 
extension of the proposed rule comment period beyond 60 days as 
favoritism by CMS for MAOs as opposed to other stakeholders. Finally, a 
commenter asserted that the study was not compliant with actuarial 
professional standards because CMS did not identify a qualified actuary 
involved in the study and did not release information about how the 
study or proposed policy complied with the Actuarial Standards of 
Practice.
    Response: Our approach after the release of the proposed rule was 
to ensure as much transparency as possible so that stakeholders could 
provide meaningful comment to our proposal not to apply an FFS 
Adjuster. To this end, we maximized data availability to the public and 
provided extended time for stakeholders to examine and opine on the 
data used in the study. As stated previously, since the publication of 
the FFS Adjuster Study on October 26, 2018, and the 2018 proposed rule 
on November 1, 2018, we published data and several related notices to 
further enhance transparency and to encourage robust public comment, 
including enhanced discussions of the methodology and assumptions used 
to conduct the study, extensions to the comment period of the proposed 
rule, and the release of the results of a replicated study. The data 
and methodology we disclosed should sufficiently allow for stakeholders 
to evaluate and comment on the study.
    Comment: As part of the comments received, MAOs analyzed and 
assessed our FFS study and the data, assumptions, and methodology it 
relied on. Many of these comments provided lengthy analysis and 
critique, and some commenters performed counter-studies. Commenters 
criticized CMS' recalibration of the CMS-HCC model, the Inflated Post-
Audit Risk Score (IPARS) adjustment, and the decision to convert claim-
level discrepancy rates to beneficiary-level discrepancy rates.
    Response: We appreciate the lengths that commenters went to examine 
and provide comment on our study, and we agree that any study that 
relies on assumptions, estimates, and projections has inherent 
limitations. However, the finalization of our proposal not to apply an 
FFS Adjuster does not depend on the results of our study. Even if 
systematic payment error exists, it does not impact the requirement 
that submitted diagnoses must be adequately supported by medical 
records. An adjustment factor to account for hypothetical systematic 
payment differences would not be appropriately applied in the RADV 
context, even if such systematic differences existed. Additionally, our 
decision relies on our reading of the coding pattern adjustment 
statutory provision and its minimum levels.
    Further, although we are not relying on the empirical findings of 
our study as the basis for our decision not to apply an FFS Adjuster, 
we do not agree with those commenters who claim that our study or their 
counter-studies provide evidence that FFS errors systematically reduce 
payments to MAOs.
    First, the magnitude of over-coding (diagnosis codes unsupported by 
medical records) in the Medicare FFS data is much smaller than some 
commenters have suggested. While some have claimed that the rate is as 
high as over 30 percent, our study calculated beneficiary-level 
discrepancy rates for each HCC that were on average only about 3 
percent, with a median of 1.8 percent. The beneficiary-level error 
rate, and not the claim-level error rate, is the appropriate measure of 
inappropriate coding because an HCC is supported if just one claim in 
the relevant year for that beneficiary is supported.
    Second, the FFS data contains significant under-coding (unreported 
diagnosis codes that have medical record support), which would likely 
offset the effects of FFS over-coding, to the extent any such effects 
exist. Although accurate coding supported by the medical record is 
required in Medicare FFS, Medicare FFS providers have less of an 
incentive to report all valid, supported codes because this does not 
increase their payments as directly as it does for MAOs in Part C. This 
is supported by the extant literature.\41\ Significantly, the 
commenters' counter-studies purporting to show that Medicare FFS errors 
systematically reduce payments to MAOs do not adequately address the 
offsetting effects of Medicare FFS under-coding.
---------------------------------------------------------------------------

    \41\ Kronick and Welch found that positive coding intensity in 
the MA risk scores increased faster than comparable FFS risk scores. 
Richard Kronick & Pete Welch, Measuring Coding Intensity in the 
Medicare Advantage Program, Medicare & Medicaid Research Review, 
2014 Vol. 4, No. 2, at E1-E19. https://www.cms.gov/mmrr/downloads/mmrr2014_004_02_a06.pdf.
    Frogner et al. examined the impact of incomplete FFS coding in 
the context of the CMS-HCC model and found that it biases payments 
to MAOs upwards. Bianca K. Frogner, Gerard F. Anderson, Robb A. 
Cohen & Chad Abrams, Incorporating New Research Into Medicare Risk 
Adjustment, 49 Medical Care 295 (2011). https://journals.lww.com/lww-medicalcare/Fulltext/2011/03000/Incorporating_New_Research_Into_Medicare_Risk.11.aspx.
    Welch et al. found that regional variation of diagnostic coding 
in FFS was related to case-fatality. H.G. Welch, S.M. Sharp, D.J. 
Gottlieb, J.S. Skinner & J.E. Wennberg, Geographic Variation in 
Diagnosis Frequency and Risk of Death Among Medicare Beneficiaries, 
305 JAMA 1113 (2011). That is, FFS Medicare enrollees have variable 
diagnostic coding. https://jamanetwork.com/journals/jama/fullarticle/646152.
    Finally, MedPAC (1998) demonstrated that the persistence in 
diagnostic coding for FFS beneficiaries was low from year to year, 
even for conditions that were serious and permanent, documenting 
incomplete coding for FFS enrollees. Medicare Payment Advisory 
Commission, Report to the Congress: Medicare Payment Policy, Vol. 1 
at 32, Vol. 2 at 15-18 (1998).
---------------------------------------------------------------------------

    Third, the effects of Medicare FFS over-coding are also offset by 
the increased costs associated with that over-coding. As noted 
previously, Medicare FFS claims are subject to error correction and 
payment adjustment when they are based on diagnosis codes not supported 
by the medical record. (See Medicare Program Integrity Manual sections 
3.3.1.1, 3.3.2.1, 3.6.2.4, 6.5.2, 6.5.3.) Thus, if CMS were to delete 
the unsupported Medicare FFS codes used to calibrate the risk 
adjustment model, it would also have to remove certain expenditures 
associated with those codes that should have been denied for payment. 
The purpose of the IPARS adjustment was to account for this 
relationship and the offsetting effects of costs associated with FFS 
over-coding.\42\ The commenters' counter-studies did not adequately 
address these effects.
---------------------------------------------------------------------------

    \42\ We note that applying the IPARS adjustment rather than 
directly studying this effect empirically is an inherent limitation 
of our study. As a result, our study's empirical findings are 
limited to the conclusion that attenuation bias, an effect described 
in the June 28, 2019 Addendum, does not systematically reduce 
payments to MAOs.
---------------------------------------------------------------------------

    Fourth and finally, we note that the counter-studies purporting to 
prove that an FFS Adjuster in a specific amount is required employed 
widely differing methodologies and arrived at widely varying estimates 
for their FFS Adjuster. For example, one commenter claimed that an FFS 
Adjuster of 9 percent would be appropriate based on the analysis they 
conducted, while another claimed the appropriate amount would be 33 
percent based on their analysis. The fact that these studies can be 
conducted in various different ways and produce such a wide range of 
results raises the question whether an FFS Adjuster is even a 
reasonable or practical means of addressing any risk adjustment 
coefficients that were too low and any that were too high, and if that 
was because of any over- and/or under-coding by FFS providers. It also 
further shows the complexity of the issues in measuring the effects of 
both under-coding and over-coding in FFS, and the fact that any related 
study must rely on assumptions, estimates, and projections,

[[Page 6660]]

and will, therefore, have inherent limitations.
    Thus, we do not agree with commenters who claim that our study or 
their counter-studies provide evidence that Medicare FFS errors 
systematically reduce payments to MAOs. For a complete discussion of 
the study methodology and all of its conclusions, see the November 1, 
2018, proposed rule, the FFS Adjuster Study and Technical Appendix 
published on October 26, 2018, the study Addendum published June 28, 
2019, and the other study documents previously described in this rule.
4. Summary of Final Policies
    We are finalizing our proposal to not apply an FFS Adjuster to RADV 
audits because the ``actuarial equivalence'' and ``same methodology'' 
provisions do not apply to the obligation of an MAO to report and 
return improper payments for diagnoses lacking medical record support, 
including those improper payments identified during a RADV audit. We 
have also concluded that it would not be reasonable to interpret the 
Act as requiring a reduction in payments to MAOs by at least a 
statutorily-set minimum percentage pursuant to the coding pattern 
adjustment, while at the same time prohibiting CMS from enforcing 
longstanding documentation requirements by requiring an offset to the 
recovery amounts calculated for CMS audits.
    While the D.C. Circuit's decision in UnitedHealthcare pertained to 
the Part C Overpayment Rule, its reasoning supports our conclusion that 
an FFS Adjuster is neither required nor appropriate in the context of 
RADV. ``The role of the actuarial-equivalence provision is to require 
CMS to model a demographically and medically analogous beneficiary 
population in traditional Medicare to determine the prospective lump-
sum payments to [MAOs].'' (UnitedHealthcare, 16 F.4th at 870.) The RADV 
program, like the Overpayment Rule, applies after the fact to require 
MAOs to refund any payment to which they are not entitled, based on 
diagnoses that lack support in the medical record.
    In the proposed rule, we also discussed a study that we conducted 
that concluded that diagnosis error in FFS claims data does not lead to 
systematic payment error in the MA program. We also stated that, even 
if systematic error exists, it would be inequitable to correct such 
errors in the payments made to audited contracts only. Furthermore, in 
the interest of transparency, CMS publicly released additional data 
underlying the study cited in the proposed rule related to the FFS 
Adjuster, provided information on a replication of our original study, 
and extended the comment period to allow more time for stakeholders to 
review the data and provide comment.
    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. We recognize that any study that aims to 
demonstrate the impact of potential error in Medicare FFS diagnoses 
data on MA requires the use of certain assumptions, estimations, and 
projections, and that any theoretical study has natural limits that 
must account for those assumptions. However, that does not change our 
ultimate conclusion that, even if systematic payment error exists, an 
adjustment factor to account for this error would not be appropriately 
applied in the RADV context. We also do not intend for this conclusion 
to suggest that we believe an FFS Adjuster is appropriate or necessary 
outside of the RADV context.
    Our position is consistent with the conclusion of the D.C. Circuit, 
which is that the actuarial-equivalence requirement is not an 
``entitle[ment] . . . to a precise payment amount'' for a Medicare 
Advantage insurer, but only ``an instruction to the Secretary regarding 
the design of the risk adjustment model as a whole . . . describ[ing] 
the type of `payment amount[s]' that the risk adjustment model should 
produce''; ``[i]t does not directly govern how CMS evaluates the 
validity of diagnoses or defines `overpayment.''' (UnitedHealthcare, 16 
F.4th at 885-86.)

IV. Collection of Information Requirements

    As defined under 5 CFR 1320.3(b) and (c) of the Paperwork Reduction 
Act of 1995 (PRA's) (44 U.S.C. 3501 et seq.) implementing regulations, 
this final rule does not impose any new or revised ``collection of 
information'' requirements or related ``burden.'' More specifically, 
the utilization of extrapolation will not affect the existing process 
for MAOs submitting medical record documentation pursuant to RADV 
audits under Sec.  422.310(e). The existing requirements for MAOs 
submitting medial record documentation are active and approved by OMB 
under control number 0938-1000 (CMS-10191). As this final rule is not 
imposing any new or revised ``collection of information'' requirements 
or related ``burden'', this rule is not subject to the requirements of 
the PRA.

V. Regulatory Impact Analysis

A. Statement of Need

    This final rule clarifies certain program integrity policies in the 
MA program, specifically, the recovery of improper payments identified 
during RADV audits, and aligns with the Administration's focus on the 
fiscal sustainability of the MA program and the interests of Medicare 
beneficiaries, providers, and MAOs.
    The improper payment measurements conducted each year by CMS, which 
are included in the HHS Agency Financial Report, as well as audits 
conducted by the HHS-OIG, have demonstrated that the MA program is at 
high risk of improper payments. In FY 2021 (based on CY 2019 payments), 
we calculated that the agency made over $15 billion in erroneous 
overpayments.\43\ (The improper payment measurements CMS conducts for 
all programs include both overpayments and underpayments.) The HHS-OIG 
has also released several reports over the past few years that also 
demonstrate a high risk of improper risk adjustment payments in the MA 
program,\44\ and has identified the MA program as one of the top 
management and performance challenges facing HHS for several years due 
to the high rate of improper payments.\45\ The Medicare program, 
including MA, has also been identified by the GAO as a high-risk

[[Page 6661]]

program due to the risk of substantial improper payments.\46\
---------------------------------------------------------------------------

    \43\ HHS, FY 2021 HHS Agency Financial Report, https://www.hhs.gov/sites/default/files/fy-2021-hhs-agency-financial-report.pdf.
    \44\ For example, see reports: Medicare Advantage Compliance 
Audit of Specific Diagnosis Codes That Anthem Community Insurance 
Company, Inc. (Contract H3655) Submitted to CMS, May 21, 2021, 
https://oig.hhs.gov/oas/reports/region7/71901187.asp; Medicare 
Advantage Compliance Audit of Specific Diagnosis Codes That Blue 
Cross Blue Shield of Michigan (Contract H9572) Submitted to CMS, 
February 24, 2021, https://oig.hhs.gov/oas/reports/region2/21801028.asp; Medicare Advantage Compliance Audit of Specific 
Diagnosis Codes That Highmark Senior Health Company (Contact H3916) 
Submitted to CMS, September 29, 2022, https://oig.hhs.gov/oas/reports/region3/31900001.asp; Medicare Advantage Compliance Audit of 
Specific Diagnosis Codes That Cariten Health Plan, Inc., (Contract 
H4461) Submitted to CMS, July 18, 2022, https://oig.hhs.gov/oas/reports/region2/22001009.asp; Medicare Advantage Compliance Audit of 
Diagnosis Codes That SCAN Health Plan (Contract H5425) Submitted to 
CMS, February 3, 2022, https://oig.hhs.gov/oas/reports/region7/71701169.asp; Medicare Advantage Compliance Audit of Diagnosis Codes 
That Humana, Inc., (Contract H1036) Submitted to CMS, April 19, 
2021, https://oig.hhs.gov/oas/reports/region7/71601165.asp.
    \45\ See OIG, 2021 Top Management and Performance Challenges 
Facing HHS, pg. 13, https://oig.hhs.gov/reports-and-publications/top-challenges/2021/index.asp.
    \46\ https://www.gao.gov/highrisk/medicare-program-improper-payments.
---------------------------------------------------------------------------

    RADV audits are CMS' main corrective action for improper 
overpayments in the MA program made to MAOs when there is a lack of 
documentation in the medical record to support the diagnoses reported 
for risk adjustment. The RADV audits confirm the presence of the 
diagnoses related to the enrollee's HCC profile through the review of 
certain categories of medical records submitted by the MAOs for the 
purpose of a RADV audit. Risk adjustment discrepancies are identified 
when an enrollee's HCCs used for payment (which is, again, based on MAO 
self-reported data) differ from the HCCs assigned based on the medical 
record review performed by CMS through the RADV audit process. Risk 
adjustment discrepancies can be aggregated to determine an overall 
amount of payment error for sampled enrollees. In turn, this payment 
error for the sample of contract enrollees can be extrapolated to 
calculate a payment error estimate for the universe of enrollees from 
which the sample is selected, within specified confidence intervals.
    The policies in this final rule are essential to having an 
effective RADV program that protects taxpayer dollars and ensures 
oversight of the MA program.

B. Overall Impact

    We examined the impact of this final rule as required by Executive 
Order 12866 on Regulatory Planning and Review (September 30, 1993), 
Executive Order 13563 on Improving Regulation and Regulatory Review 
(January 18, 2011), the Regulatory Flexibility Act (RFA) (September 19, 
1980, Pub. L. 96-354), section 1102(b) of the Act, section 202 of the 
Unfunded Mandates Reform Act of 1995 (UMRA) (March 22, 1995; Pub. L. 
104-4), Executive Order 13132 on Federalism (August 4, 1999), and the 
Congressional Review Act (5 U.S.C. 804(2)).
    Executive Orders 12866 and 13563 direct agencies to assess all 
costs and benefits of available regulatory alternatives and, if 
regulation is necessary, to select regulatory approaches that maximize 
net benefits (including potential economic, environmental, public 
health and safety effects, distributive impacts, and equity. Section 
3(f) of Executive Order 12866 defines a ``significant regulatory 
action'' as an action that is likely to result in a rule: (1) having an 
annual effect on the economy of $100 million or more in any 1 year, or 
adversely and materially affecting a sector of the economy, 
productivity, competition, jobs, the environment, public health or 
safety, or State, local or Tribal governments or communities (also 
referred to as ``economically significant''); (2) creating a serious 
inconsistency or otherwise interfering with an action taken or planned 
by another agency; (3) materially altering the budgetary impacts of 
entitlement grants, user fees, or loan programs or the rights and 
obligations of recipients thereof; or (4) raising novel legal or policy 
issues arising out of legal mandates, the President's priorities, or 
the principles set forth in the Executive Order.
    A regulatory impact analysis (RIA) must be prepared for major rules 
with significant regulatory action/s and/or with economically 
significant effects ($100 million or more in any 1 year). Based on our 
estimates, OMB's Office of Information and Regulatory Affairs has 
determined this rulemaking is ``economically significant'' as measured 
by the $100 million threshold, and hence also a major rule under 
Subtitle E of the Small Business Regulatory Enforcement Fairness Act of 
1996 (also known as the Congressional Review Act). Accordingly, we have 
prepared a Regulatory Impact Analysis that to the best of our ability 
presents the costs and benefits of the rulemaking. Finally, in 
accordance with the provision of the Executive Order 12866, this final 
rule was reviewed by the Office of Management and Budget.
    Section 202 of UMRA also requires that agencies assess anticipated 
costs and benefits before issuing any rule whose mandates require 
spending in any one year of $100 million in 1995 dollars, updated 
annually for inflation. In 2022, that threshold is approximately $165 
million. This final rule would not impose a mandate that will result in 
the expenditure by State, local, and Tribal governments, in the 
aggregate, or by the private sector, of more than $165 million in any 
one year.
    The RFA requires agencies to analyze options for regulatory relief 
of small entities, if a rule has a significant impact on a substantial 
number of small entities. For purposes of the RFA, small entities 
include small businesses, nonprofit organizations, and small 
governmental jurisdictions. The great majority of hospitals and most 
other health care providers and suppliers are small entities, either by 
being nonprofit organizations or by meeting the SBA definition of a 
small business (having revenues of less than $8.0 million to $41.5 
million in any 1 year. This final rule affects MAOs with a minimum 
threshold for small business size of $41.5 million (see the Small 
Business Administration's website at http://www.sba.gov/content/small-business-size-standards). This final rule additionally affects 
hospitals (NAICS subsector 622) and a variety of provider categories, 
including physicians and specialists (NAICS subsector 621).
    To clarify the flow of payments between these entities and the 
Federal Government, note that MAOs submit bids (that is, proposed plan 
designs and projections of the revenue needed to provide those 
benefits, divided into three categories--basic benefits, supplemental 
benefits, and Part D drug benefits) in June for operation in the 
following contract year. These bids project payments to hospitals, 
providers, and staff as well as the cost of administration and profits. 
These bids in turn determine the payments from the Medicare Trust Fund 
to the MAOs that pay providers and other stakeholders for their 
provision of covered benefits to enrollees in MA plans. Consequently, 
our analysis will focus on MAOs.
    There are various types of Medicare health and drug plans, 
including MAOs, demonstrations, section 1876 cost plans, Part D 
prescription drug plans (PDPs), and PACE organizations. There are a 
variety of ways to assess whether MAOs meet the $41.5 million threshold 
for small businesses. The assessment can be done by examining net 
worth, net income, cash flow from operations, and/or projected claims 
as indicated in their bids. Using projected monetary requirements and 
projected enrollment for 2018 from submitted bids, 32 percent of the 
MAOs fell below the $41.5 million threshold for small businesses. 
Additionally, an analysis of 2016 data shows that 32 percent of all 
MAOs fall below the minimum threshold for small businesses.
    If a rule potentially has a significant impact on a substantial 
number of small entities, the rule must discuss steps taken, including 
alternatives, to minimize the burden on small entities. While some of 
the entities affected by this rule are not-for-profit organizations and 
small businesses, the impact is not significant. No changes are made to 
long-standing audit documentation standards as a result of this rule; 
therefore, there is no significant impact to small entities (or any 
entities). MAOs provide medical record documentation to CMS as a normal 
business practice pursuant to RADV audits. Consequently, the Secretary 
has certified that this final rule will not have a significant economic 
impact on a substantial

[[Page 6662]]

number of small entities, and we have met the requirements of the RFA.
    In addition, section 1102(b) of the Act requires us to prepare a 
regulatory impact analysis if a rule may have a significant impact on 
the operations of a substantial number of small rural hospitals. This 
analysis must conform to the provisions of section 604 of the RFA. For 
purposes of section 1102(b) of the Act, we define a small rural 
hospital as a hospital that is located outside of a metropolitan 
statistical area and has fewer than 100 beds. Therefore, the Secretary 
has certified that this final rule will not have a significant impact 
on the operations of a substantial number of small rural hospitals, and 
as a result we are not preparing an analysis for section 1102(b) of the 
Act.
    Executive Order 13132 establishes certain requirements that an 
agency must meet when it promulgates a final rule that imposes 
substantial direct requirement costs on State and local governments, 
preempts State law, or otherwise has federalism implications. Because 
this final rule does not impose any substantial costs on State or local 
governments, the requirements of Executive Order 13132 are not 
applicable.

C. Regulatory Review Cost

    If regulations impose administrative costs on reviewers, such as 
the time needed to read and interpret this final rule, then we should 
estimate the cost associated with regulatory review. There are 
approximately 750 MA contracts (of which, 65 MA contracts include 
PDPs). We assume each entity will have one designated staff member who 
will review the entire rule. Other assumptions are possible and will be 
reviewed after the calculations.
    Using the 2021 wage information from the Bureau of Labor Statistics 
(BLS) for medical and health service managers (code 11-9111), we 
estimate that the cost of reviewing this rule is $115.22 per hour, 
including fringe benefits and overhead costs (http://www.bls.gov/oes/current/oes_nat.htm). Assuming an average reading speed for technical 
material of 200 words per minute, we estimate that it will take 
approximately 2 hours for each person to review this final rule. For 
each entity that reviews the rule, the estimated cost is therefore, 
$230.44 (2 hours * $115.22). Therefore, we estimate that the total cost 
of reviewing this regulation is $172,830 ($230.44 * 750 reviewers).
    Note that this analysis assumes one reader per contract. Some 
alternatives include assuming one reader per parent entity. Using 
parent organizations instead of contracts would reduce the number of 
reviewers to approximately 500 (assuming approximately 250 parent 
organizations), and this would reduce the total cost of reviewing by a 
third. However, we believe it is likely that reviewing will be 
performed at the contract level. The argument for this is that a parent 
organization might have local reviewers; even if that parent 
organization has several contracts that might have a reader for each 
distinct geographic region, to identify effects of provisions specific 
to that region.

D. Detailed Economic Analysis

    This final rule creates regulations to govern the collection of 
extrapolated audit findings in MA. As we develop our approach to 
statistical sampling and extrapolation, we are taking account of the 
recommendations of the 2016 GAO report entitled, ``Fundamental 
Improvements Needed in CMS' Effort to Recover Substantial Amounts of 
Improper Payments.'' The GAO recommended that CMS select plans based on 
the risk for improper payments. Prior to the GAO report, CMS selected 
stratified random samples of enrollees during RADV audits, including 
our 2011 to 2013 audits for which we proposed to apply the policies in 
this rule. However, beginning with the 2014 audit year, CMS began 
incorporating the potential risk of improper payments to MAOs, based on 
past audit findings and other factors, into selecting enrollee samples 
for audits. Accordingly, CMS expects to be more effective in 
identifying improper payments in future audit years.
    To clarify in more detail how the final rule impacts the recovery 
audit process, we note the following:
     The Part C Improper Payment Measurement audits are 
conducted annually to measure payment error in the Medicare Part C 
program. After defining the eligible population, a representative 
sample of beneficiaries from risk adjustment eligible contracts are 
selected for medical record review. MAOs submit medical record 
documentation to substantiate the CMS-HCCs payments sampled by CMS for 
each year's Part C Improper Payment Measurement. Certified coders code 
the medical records, and the findings are used to recalculate risk 
scores for each sampled beneficiary. The difference between the payment 
risk scores and the recalculated risk scores is termed Risk Adjustment 
Error. Validation results from the sample are extrapolated to the 
broader Part C population to produce payment error estimates that meet 
the PIIA requirements for the payment year. No recoveries are made 
through these audits.
     Findings from the Part C Improper Payment Measurement and 
contract-level audits are used to help identify cohorts of 
beneficiaries for which CMS may be most at risk for making improper 
payments to MAOs. While CMS has flexibility to decide how to focus 
audits, CMS intends to focus audits on such MAOs in the future, and has 
been taking a more focused approach on areas of high risk of improper 
payments starting with the PY 2014 RADV audits.
     By better targeting contract-level RADV audits based on 
MAOs' risk of receiving improper payments, CMS expects to have a 
sentinel effect and reduce the historical Part C improper payment rate 
over time.
1. Expected Impact of These Provisions
    While we cannot fully estimate the quantitative impact of this 
provision, we can clearly identify certain components of impact. We 
start with some basic facts:
     With extrapolation applied to audit findings for payment 
years 2018 and later, we would realize a positive return on investment. 
The annual cost per year for the contract-level RADV audit program 
activities, with or without the changes finalized in this rule, is 
approximately $51 million.
     Extrapolating audit findings does not increase the cost 
burden on the plan. The cost to the plan of complying with a RADV audit 
is neither the subject of nor affected by this provision.
     We estimate that findings from audits of MAO contracts for 
PYs 2011, 2012, and 2013 will identify a total of $683.2 million in 
extrapolated improper payments. This $683.2 million represents a 
transfer from the Federal Government to insurers, because it reflects 
improper payments for human coding error which CMS paid to MAOs. 
Although we will not exercise our authority to seek extrapolated 
contract-level recoveries for these payment years, we refer to the 
$683.2 million in improper payments to estimate future expected 
recoveries from finalizing this rule.
     30 contracts per year were audited in PYs 2011 through 
2013.
     Approximately 80 percent of the audited contracts in 2011 
through 2013 had findings of improper payments.
    Using this data, we can conclude as follows:
     $683.2 million divided by 3 audit years is $227.7 million 
per audit year.
     $227.7 million per audit year divided by 24 contracts (30 
contracts multiplied by 0.80) with audit findings

[[Page 6663]]

per year is approximately $9.5 million in findings per contract per 
year.
     As we are adopting GAO recommendations by focusing on 
contracts at higher risk for improper payments, if the average level of 
audit findings per contract, at a minimum, holds constant, the $9.5 
million per contract with audit findings per year multiplied by 30 
contracts with audit findings per year would produce approximately $285 
million in improper payment recoveries per audit year.\47\
---------------------------------------------------------------------------

    \47\ The $285 million amount is a theoretical estimated amount 
for the audit of PY 2014; however, as we have previously explained, 
CMS will begin extrapolation with the PY 2018 RADV audits. The $285 
million amount is the baseline amount from which CMS begins 
adjusting estimated improper payment recoveries for inflation beyond 
PY 2014. Note, if CMS conducts more than one payment year audit 
annually, savings estimates will be higher in subsequent years.
---------------------------------------------------------------------------

    With extrapolation applied to audit findings beginning with 2018 
payment year audits, the expected level of recovery in calendar year 
2025 (the year in which we project to initiate improper payment 
recoveries for PY 2018 audits) would produce $428.4 million in net 
recovery (that is, $479.4 million minus the annual cost of the RADV 
program of $51 million). However, we note that while non-extrapolated 
recoveries would likely result in an average of $8.2 million in 
estimated improper payment recoveries associated with each audited 
payment year, the RADV audit program would not achieve positive net 
recoveries per year without the RADV rule (see Table 2).
     Improper payment recoveries in years 2025 and later 
increase based on projected rates of growth in MA spending. The 10-year 
impact of this final rule is estimated in Table 3. Estimating recovery 
amounts per year is difficult for the following reasons:
     The improper payment rate per year, as indicated in the 
reports of the CMS Chief Financial Officer, have been declining and are 
likely to continue to decline due to the impact that these RADV audits 
have on MAO efforts to reduce the reporting of unsupported HCCs.
     The aggregate amount paid to MAO contracts is increasing 
due to enrollment growth and other cost inflationary factors. The 
Office of the Actuary at CMS annually publishes a Trustees Report that 
contains projected annual MA enrollment in aggregate. All other things 
being equal, the increase in enrollment will cause nominal dollars in 
error to increase. The historical decline in the error rate may or may 
not offset the increase due to increasing enrollment, making a 
projection difficult.
     We previously indicated that acceptance of GAO 
recommendations would facilitate auditing contracts with cohorts of 
enrollees associated with higher degrees of risk for CMS making 
improper payments, and therefore assume there would be findings in all 
contract audits.
    For the reasons cited previously in this section, we are increasing 
the annual estimate of recoveries of improper payments to the Medicare 
Trust Fund at the same rate as the projected growth in MA spending 
stated in the FY 2023 President's Budget, beginning with $479.4 million 
for 2025 (when we anticipate beginning to receive extrapolated 
recoveries). In 2023 and 2024, we estimate receiving approximately 
$13.1 million and $28.0 million, respectively, in non-extrapolated 
recoveries from 2011 through 2013 and 2014 and 2015 payment year 
audits. Accordingly, the result would be negative net recovery amounts 
of $37.9 million ($13.1 million minus the $51 million annual cost of 
the RADV audit program) in 2023 and $23 million ($28 million minus $51 
million) in 2024.
    In total, the estimated recovery amount from 2023 through 2032 is 
$4.7 billion (see Table 3). This money is a reduction in spending of 
the Medicare Trust Fund resulting mostly from recoveries (or transfers) 
from MAOs to the Federal Government; there will be no money transferred 
to enrollees.
    The intent of this rule is to protect taxpayer dollars and ensure 
oversight of the MA program, in part by reducing the Part C improper 
payment rate.
2. Alternatives Considered
    This rule includes transfers from MAOs to the Federal Government. 
The aggregate impact of each of these over 10 years is approximately 
$4.7 billion (see Table 3). Various alternatives to this rulemaking 
were considered, including the use and timing of extrapolation, as well 
as the application of an FFS Adjuster. These alternatives are described 
in this section of this rule.
a. Alternatives Related to the Extrapolation of RADV Findings
    As an alternative to our decision to extrapolate our RADV audits 
beginning in PY 2018, we considered policies whereby we would not 
extrapolate and would only collect improper payments associated with 
sampled enrollees as a result of RADV audits. While such a policy would 
likely be favorably received by MAOs, it would result in a drastic 
reduction in potential recoveries and dilute the sentinel impact that 
the RADV program has on reducing the Part C improper payment rate. 
Specifically, annual net recoveries of improper payments (that is, 
estimated collections from past audits minus the estimated annual audit 
program costs) would be reduced from approximately $234 million \48\ to 
negative $42.8 million (see Table 2). Given the overall cost of $51 
million per year to administer the RADV program, this would result in a 
negative return on investment of approximately $6.2:1 (negative $51 
million divided by $8.2 million). This would be in direct conflict with 
our responsibilities under the PIIA to reduce improper payments and 
fiduciary responsibility to recover improper payment from the Medicare 
Trust Funds, and therefore, this alternative was not an acceptable 
alternative to CMS.
---------------------------------------------------------------------------

    \48\ $234 million in net recoveries is derived by subtracting 
$51 million (cost of administering the CMS RADV audit program) from 
the theoretical estimated amount of extrapolated recoveries ($285 
million) that would have been collected if extrapolation was applied 
for the PY 2014 audits.
---------------------------------------------------------------------------

    We also considered whether to apply extrapolation beginning in PY 
2011, as proposed, as well as other payment years after PY 2011. 
Beginning extrapolation in PY 2011 would result in the collection of 
approximately $2 billion in improper payments for PYs 2011 to 2017, in 
contrast to the $41.1 million in improper payments we estimate to 
collect for these years as a result of this final rule. While we 
believe that applying extrapolation to RADV findings beginning in PY 
2011 (or other payment year after PY 2011) would be a supportable 
decision and consistent with our mandate to protect taxpayer dollars, 
we determined that the overall long-term success of the RADV program 
(and ultimately the MA program) requires us to consider the projected 
level of effort and likelihood of collecting improper payments along 
with other practical realities.
    As previously described, we believe that beginning extrapolation 
for PY 2018 RADV audits represents an appropriate policy because it 
recognizes our fiduciary duty to protect taxpayer dollars from 
overpayments and preserves our ability to collect on significant 
(extrapolated) amounts of overpayments made to plans beginning in PY 
2018. This final rule will also allow CMS to focus on conducting future 
RADV audits as soon as practicable after an MAO payment year concludes, 
which was the topic of significant public comment to the proposed rule. 
Lastly, we have determined that it is in the best interest of all 
parties to ensure that the contract-level RADV appeals process, which 
is also outlined in regulation, is able to

[[Page 6664]]

successfully process all RADV appeals. By not using an extrapolation 
methodology prior to PY 2018, we expect to better control the total 
number of active appeals that are submitted in the first few years 
following finalization of this rule, which will alleviate burden on 
MAOs and CMS.
b. Alternatives Related to the Application of an FFS Adjuster to RADV 
Improper Payment Determinations
    As an alternative to our decision to not apply an FFS Adjuster to 
our RADV overpayment determinations, we considered whether to finalize 
a policy whereby we would apply an FFS Adjuster to RADV overpayment 
determinations. While we contemplated adoption of an FFS Adjuster as 
part of our 2012 Methodology, we believe that finalizing such an 
approach through regulatory or other means would be an unsupportable 
and unreasonable interpretation of the Act.
    As previously described, we have determined that the ``actuarial 
equivalence'' and ``same methodology'' provisions do not apply to the 
obligation of an MAO to report and return overpayments that they have 
identified, including overpayments due to lack of medical record 
support for diagnoses, or their obligation to return overpayments 
identified based on a RADV audit. In UnitedHealthcare, the D.C. Circuit 
held that actuarial equivalence and same methodology do not apply to 
the MAOs' obligation to report and return overpayments that they have 
identified, including overpayments arising from the MAOs' submission of 
and payments based on diagnoses unsupported by their beneficiaries' 
medical records. Although UnitedHealthcare addressed the enforceability 
of the Part C overpayment regulation, its reasoning applies just as 
strongly in the RADV context and supports our conclusion that the use 
of an FFS Adjuster is neither required nor appropriate for an RADV 
audit.
    We have also concluded that it would be unreasonable to interpret 
the Act as requiring a minimum reduction in payments in one provision 
(the coding pattern provision), while at the same time prohibiting CMS 
in an adjacent provision (the actuarial equivalence provision) from 
enforcing those longstanding documentation requirements (by requiring 
an offset to the recovery amount calculated for CMS audits). To the 
contrary, because the Act requires CMS to reduce payments to MAOs by at 
least a specific minimum percentage, the only reasonable interpretation 
of the Act is that CMS would pay MAOs at those reduced rates, under the 
existing payment model,\49\ and enforce the longstanding documentation 
requirements through CMS' audits.
---------------------------------------------------------------------------

    \49\ Any changes to the CMS-HCC payment model are published in 
the annual payment notice.

 Table 2--Expected Net Recoveries of CMS RADV Improper Payments per Year
                          Without Extrapolation
------------------------------------------------------------------------
                                Amount ($ in millions)--
    Label            Item           non-extrapolated        Source or
                                                           calculation
------------------------------------------------------------------------
(A)..........  Estimated Non-                     $41.1
                Extrapolated
                Collections
                for 2011-2015
                audits.
(B)..........  Number of                              5
                years, 2011-
                2015.
(C)..........  Estimated                           $8.2  (C) = (A)/(B).
                Average Non-
                Extrapolated
                Collections
                per year.
(D)..........  RADV audit                           $51  Estimated costs
                programs costs                            of RADV
                per year.                                 program in
                                                          which
                                                          statistically
                                                          valid samples
                                                          are pulled to
                                                          audit sub-
                                                          cohorts of
                                                          enrollees for
                                                          a minimum of
                                                          30 contracts
                                                          per year.
(E)..........  Estimated net                    ($42.8)  (E) = (C)-(D).
                recoveries of
                improper
                payments per
                year without
                extrapolation.
------------------------------------------------------------------------


                                  Table 3--Impact on Estimated Collections of Improper Payments per Year From RADV Rule
                                                                     [$ in millions]
--------------------------------------------------------------------------------------------------------------------------------------------------------
                                    2023       2024       2025       2026       2027       2028       2029       2030       2031       2032      Total
--------------------------------------------------------------------------------------------------------------------------------------------------------
Estimated Non-Extrapolated            13.1       28.0       11.6       10.9       12.7       13.5       14.4       15.4       16.4       17.5      153.5
 Collections Assumed Without
 RADV Final Rule Changes.......
Estimated Collections from            13.1       28.0      479.4      447.5      522.6      557.2      594.0      633.2      675.0      719.5    4,669.5
 Audits Completed in Prior
 Years With RADV Final Rule
 Changes.......................
Additional Estimated                   0.0        0.0      467.8      436.6      509.9      543.7      579.6      617.8      658.6      702.0    4,516.0
 Collections as a Result of
 RADV Final Rule...............
--------------------------------------------------------------------------------------------------------------------------------------------------------

E. Accounting Statement and Table

    As required by OMB Circular A-4 (available at https://obamawhitehouse.archives.gov/omb/circulars_a004_a-4/), Table 4 shows 
the costs and transfers associated with the provisions of this final 
rule for calendar years 2022 through 2031.

[[Page 6665]]



  Table 4--Accounting Statement--Classification of Estimated Transfers
------------------------------------------------------------------------
                                     Discount rate
            Category            ----------------------   Period covered
                                     7%         3%
------------------------------------------------------------------------
Transfers:
    Annualized Monetized              $410       $433  CYs 2023-2032.
     Transfers ($ in Millions).
------------------------------------------------------------------------
    From Whom to Whom..........        MAOs to Federal Government.
------------------------------------------------------------------------

    We estimate that from 2022 through 2031 this final rule will 
generate Federal annualized monetized transfers of $410 million and 
$433 million, at the 7 percent and 3 percent discount rates 
respectively, from MAOs back to the Medicare Trust Fund.
    This final rule is subject to the Congressional Review Act 
provisions of the Small Business Regulatory Enforcement Fairness Act of 
1996 (5 U.S.C. 801 et seq.) and has been transmitted to the Congress 
and the Comptroller General for review.
    Chiquita Brooks-LaSure, Administrator of the Centers for Medicare & 
Medicaid Services, approved this document on January 24, 2023.

List of Subjects in 42 CFR Part 422

    Health facilities, Health maintenance organizations (HMO), 
Medicare, Penalties, Privacy Reporting and record keeping requirements.

    For the reasons stated in the preamble, the Centers for Medicare & 
Medicaid Services proposes to amend 42 CFR part 422 as follows:

PART 422--MEDICARE ADVANTAGE PROGRAM

0
1. The authority citation for part 422 continues to read as follows:

    Authority: 42 U.S.C. 1302 and 1395hh.

Subpart G--PAYMENTS TO MEDICARE ADVANTAGE ORGANIZATIONS

0
2. Section 422.300 is revised to read as follows:


Sec.  422.300   Basis and scope.

    This subpart is based on sections 1106, 1128J(d), 1852, 1853, 1854, 
and 1858 of the Act. It sets forth the requirements for making payments 
to MA organizations offering local and regional MA policies, including 
calculation of MA capitation rates and benchmarks, conditions under 
which payment is based on plan bids, adjustments to capitation rates 
(including risk adjustment), collection of risk adjustment data, 
conditions for use and disclosure of risk adjustment data, collection 
of improper payments and other payment rules. Section 422.458 specifies 
the requirements for risk sharing payments to MA regional 
organizations.
0
3. Section 422.310 is amended by revising paragraph (e) to read as 
follows:


Sec.  422.310   Risk adjustment data.

* * * * *
    (e) Validation of risk adjustment data. MA organizations and their 
providers and practitioners are required to submit a sample of medical 
records for the validation of risk adjustment data, as required by CMS. 
There may be penalties for submission of false data. MA organizations 
must remit improper payments based on RADV audits, in a manner 
specified by CMS. For RADV audits, CMS may extrapolate RADV Contract-
Level audit findings for payment year 2018 and subsequent payment 
years.
* * * * *
0
4. Section 422.311 is amended by revising paragraph (a) to read as 
follows:


Sec.  422.311   RADV audit dispute and appeal processes.

    (a) Risk adjustment data validation (RADV) audits. In accordance 
with Sec. Sec.  422.2 and 422.310(e), the Secretary annually conducts 
RADV audits to ensure risk-adjusted payment integrity and accuracy.
    (1) Recovery of improper payments from MA organizations will be 
conducted in accordance with the Secretary's payment error 
extrapolation and recovery methodologies.
    (2) CMS may apply extrapolation to audits for payment year 2018 and 
subsequent payment years.
* * * * *

    Dated: January 26, 2023.
Xavier Becerra,
Secretary, Department of Health and Human Services.
[FR Doc. 2023-01942 Filed 1-30-23; 4:15 pm]
BILLING CODE 4120-01-P