OPINION
ROSEMARY M. COLLYER United States District Court
Health
insurance is provided to most seniors and many disabled
Americans through Medicare, paid for by taxes and
administered by the Centers for Medicare and Medicaid
Services (CMS). As amended, the Medicare statute (formally
part of the Social Security Act), includes a “Medicare
Advantage” program whereby Medicare-eligible
individuals can elect to receive their health insurance
coverage through a private insurance company. The insurance
company must provide at least the same coverage as
traditional Medicare, although it often expands coverage, and
is to make its profit from Medicare through efficiencies and
other cost-saving methods. The statute requires
“actuarial equivalence” between CMS payments for
healthcare coverage under Medicare Advantage plans and CMS
payments under traditional Medicare. In this case, a large
group of insurance companies that provide Medicare Advantage
coverage challenged a Final Rule, adopted in 2014, by which
the documentation used to set the rates to pay the insurance
companies is inconsistent with the documentation used to
determine if the insurers have been overpaid. The insurers
allege that the Final Rule will inevitably fail to satisfy
the statutory mandate of actuarial equivalence.
There
is a history to this dispute over actuarial equivalence. The
government previously had proposed an audit program for
Medicare Advantage insurers and some insurers challenged its
methodology for determining overpayments. Since government
records for traditional Medicare payments are used to set
rates but are not audited, the insurers contended that
imposing a 100% accuracy requirement on their records, on
pain of being required to return any “overpayment,
” would violate the statutory requirement for
actuarially equivalent payments between traditional Medicare
and Medicare Advantage. Heeding the advice of actuaries, the
government ultimately adjusted its audit plan to recognize
the different data sets. For the 2014 Final Rule at issue
here, however, CMS has refused to make such an adjustment
although the different data sets are again in use.
After
full briefing and oral argument, this Court concludes that
the 2014 Final Rule violates the statutory mandate of
“actuarial equivalence” and constitutes a
departure from prior policy that the government fails
adequately to explain. The Court will grant summary judgment
to the Medicare Advantage insurers and vacate the Rule.
I.
BACKGROUND
This
lawsuit is brought by Medicare Advantage (MA) organizations
in the UnitedHealth Group family of companies, the
nation's leading provider of Medicare Advantage health
benefits plans (collectively, UnitedHealth).[1] Known as
Medicare Part C, the Medicare Advantage program allows
Medicare-eligible individuals to receive healthcare benefits
through private insurance companies that have contracted with
CMS, a constituent agency of the Department of Health and
Human Services (HHS). Alex M. Azar II, HHS Secretary, is sued
in his official capacity. CMS administers traditional
Medicare and pays its benefits. However, some 20 million
Americans, approximately one-third of Medicare-eligible
individuals, have opted for Medicare Advantage coverage.
Medicare
Parts A, B and C are relevant here. Medicare Part A is
mandatory for senior Americans who take Social Security
benefits; Part A provides coverage for hospital expenses.
Medicare Part B is voluntary and provides partial coverage
for doctor expenses. Medicare Part C offers the Medicare
Advantage program through which private insurance companies
replace CMS and provide full Medicare coverage to
beneficiaries.
Initially,
Medicare paid all “reasonable costs” (“fee
for service”) to a hospital caring for a Medicare
beneficiary. See Methodist Hosp. of Sacramento v.
Shalala, 38 F.3d 1225, 1227 (D.C. Cir. 1994). Over time,
that standard has changed and Medicare now pays a hospital
based on the “Diagnosis-Related Group” (DRG)
shown by the patient's diagnoses at the time of
discharge. Medicare Part B also started by paying doctors a
reasonable “fee for service, ” but now pays them
according to fee schedules that limit the amount they may
charge and be paid for each defined service. See United
Seniors Ass'n, Inc. v. Shalala, 182 F.3d 965, 968
(D.C. Cir. 1999). Under Part B, doctors must submit diagnosis
codes to identify the reason a patient received treatment,
but “payments depend only on the services (or durable
goods) provided [office visit, examination, shot, etc.] and
not in any way on the diagnoses submitted.” Defs.'
Mem. in Support of Their Cross-Mot. for Summ. J. and
Opp'n to Pls.' Mot. for Summ. J. (CMS Mot.) [Dkt.
57-1] at 7.[2] In contrast, Medicare Advantage
insurers are not paid based on medical services but
“are paid a pre-determined monthly sum for each person
they cover, based in part upon the characteristics of the
particular beneficiary being covered.” Id.
(internal citation omitted).
A
Medicare Advantage insurer must provide, at a minimum, the
same level of benefits provided by traditional Medicare
itself, except for hospice care. See 42 U.S.C.
§ 1395w-22(a). Under a Medicare Advantage policy, the
insurance companies pay doctors, other healthcare providers,
and hospitals for their services and are reimbursed by CMS on
a per-member-per-month rate that is determined beforehand.
See Id. § 1395w-23(a).
By law,
CMS must pay Medicare Advantage insurers in a manner that
ensures “actuarial equivalence” between payments
for healthcare under Medicare and Medicare Advantage plans:
[T]he Secretary shall adjust the payment amount [of fixed
monthly payments to Medicare Advantage insurers] for such
risk factors as age, disability status, gender, institutional
status, and such other factors as the Secretary determines to
be appropriate, including adjustment for health status . . .,
so as to ensure actuarial equivalence.
Id. § 1395w-23(a)(1)(C)(i). Risk factors
represent the risk that a given beneficiary, or beneficiary
population, will need healthcare from doctors or hospitals in
the next year as it may be diagnosed. “A risk
adjustment model is required to translate the diagnosis data
into expected costs of coverage.” CMS Mot. at 14. For
this purpose, CMS relies on its model, the CMS Hierarchical
Condition Category (CMS-HHC) risk-adjustment model, to
“perform that conversion”:
CMS-HCC is a complex regression model built to estimate the
costs associated with certain characteristics of Medicare
beneficiaries. The inputs to the model are data from
individuals who receive their benefits through the
traditional, fee-for-service Medicare system. Its outputs are
a set of multipliers-that is, “coefficients”-that
“represent the marginal (additional) cost” of
each medical “condition or demographic factor (e.g.,
age/sex group, Medicaid status, disability status).”
The coefficients are added together to form a “risk
score, ” and then computed against a base payment rate
(which varies depending on geography and the bid submitted by
the insurer, among other things).
Id. (internal citations omitted).
By this
process, CMS calculates the average monthly expenditure for
an average beneficiary under traditional Medicare in the past
year. The “base rate establishes . . . what it would
cost to treat a beneficiary of average risk in a given
area.” See Transcript of Aug. 8, 2018 Motions
Hearing (Hearing Tr.) [Dkt. 73] at 5. CMS adds a geographical
differential, based on data from the past year, to calculate
an average per-capita monthly payment for each county in the
nation.
This is
no straightforward task. Each traditional Medicare
beneficiary has a “demographic risk coefficient”
which reflects that person's age, gender, institutional
status, and disability status, among others. See Id.
at 4. Additional coefficients represent the health status of
the beneficiaries in traditional Medicare, taken from their
diagnosis codes as reported to CMS by their doctors. Using
such CMS data, “the model estimates the marginal cost
of each disease and cluster of demographic characteristics. .
. . By mapping known expenditures . . ., the model calculates
the expected cost of each medical condition and demographic
factor.” CMS Mot. at 17. Using the data from the
demographic characteristics, reported diagnoses, and Medicare
expenses of the beneficiaries in traditional Medicare, the
model can estimate the marginal cost of each condition,
disease and cluster of demographic characteristics.
The
“average beneficiary” is given a risk score of
1.0, which is then adjusted upwards or downwards according to
the risk score determined by an individual's demographic
and health status information. For example, if a beneficiary
has a condition that CMS has determined based on its Medicare
data increases average costs by 20%, that person will have an
adjusted risk score of 1.2 and the Medicare Advantage payment
rate applicable to that person will be set at 120% of the
average benchmark rate. See, e.g., Advance Notice of
Methodological Changes for CY 2004 Part C Rates (Mar. 28,
2003) (2004 Advance Notice) at ¶ 3895-97 (describing how
CMS uses the model to “associate diseases categories
with incremental costs”).[3]Thus, the costs in a prior year of
the “risk coefficients” in the traditional
Medicare system are used to determine the costs of similar
risk coefficients for Medicare Advantage beneficiaries. The
underlying logic is that developing risk coefficients with
data from traditional Medicare, and then adjusting a Medicare
Advantage beneficiary's risk score (and the payment to
the Medicare Advantage insurer accordingly), will render the
cost to CMS under traditional Medicare and the cost to the
insurer under Medicare Advantage actuarially equivalent.
In
conducting these analyses, CMS relies entirely on the
diagnosis codes submitted by healthcare providers under
traditional Medicare. “[T]he risk adjustment model is
built on unaudited [traditional Medicare] data . . . which
must contain errors.” CMS Mot. at 37. Indeed, doctors
treating traditional Medicare patients are paid based on
their services and not the diagnosis codes they might submit
to report why the patient saw the doctor. As
UnitedHealth's counsel explained at argument, physicians
bill traditional Medicare by procedure, not diagnosis codes,
so that “physicians are essentially indifferent to the
diagnosis . . . . There's no financial incentive to be
particularly careful.” Hearing Tr. at 13. “[W]hat
matters is the procedure they did.” Id. at 14;
see also CMS Mot. at 7 (agreeing that traditional
Medicare payments to doctors “depend only on the
services . . . and not in any way on the diagnoses
submitted”). Given this incentive scheme, it can be no
surprise that diagnosis reports for Medicare Part B are
considered much less reliable than hospital diagnosis reports
for Part A. See CMS Mot. at 7 (noting “the
quality of the Part B diagnosis data is generally understood
to be inferior to the Part A diagnosis data”).
Medicare
Advantage insurance companies bid annually after CMS issues
notice of each county's benchmark rate for the
forthcoming year. See 42 U.S.C. §
1395w-23(b)(1)(B). The insurers are paid on a per-capita
basis for each covered individual, including applicable risk
scores. As a result, a Medicare Advantage insurer undertakes
to provide insurance coverage at least identical to Medicare
at annual fixed rates even though the health care needs of
the covered populations, mostly the elderly, vary greatly.
Humans
being human, diagnoses in healthcare records may be miscoded,
inappropriately added, or otherwise faulty by accident or mal
intent. UnitedHealth suggests that the error rate can be as
high as 20%. See Compl. [Dkt. 1] ¶ 38; see
also Hearing Tr. at 28. In the past, neither CMS nor the
insurers made efforts to review proactively the diagnosis
codes assigned by healthcare providers. Indeed, as stated
above, CMS treats diagnosis codes as categorically valid for
its own purposes under traditional Medicare, including for
setting rates for Medicare Advantage. Nonetheless, CMS has
long required Medicare Advantage insurers to certify
“based on best knowledge, information and belief”
that the information they provide to CMS, including all
diagnosis codes, is “accurate, complete, and
truthful.” 42 C.F.R. § 422.504(1)(2). CMS contends
that this pre-existing regulation, and other existing agency
practices, have long required that diagnosis codes submitted
by Medical Advantage insurers be supported by underlying
medical records (i.e., patient medical charts).
UnitedHealth responds that neither this pre-existing
regulation, nor any other law or regulation, has previously
obligated the insurance companies who provide Medicare
Advantage insurance to validate independently the underlying
medical records that support diagnosis codes submitted by
health care providers.
For
more than a decade, CMS has conducted audits of a subsection
of insurers in the Medicare Advantage program, through which
it has compared the diagnosis codes in bills paid by the
insurance companies to the underlying patient medical charts
and records, which it requires the insurers to obtain for
this purpose. It has then required repayment to CMS of any
costs that were based on unsupported diagnosis codes. In
2008, CMS announced that it would begin applying these
“Risk Adjustment Data Validation (RADV)” audits
to extrapolate the error rate in the audited sample across an
entire insurance contract. [4]The insurer would be responsible for
returning any overpayment to CMS, based on the extrapolated
error rate.
When
CMS sought comments on its new methodology for conducting
RADV audits, Medicare Advantage insurers immediately
protested that the rates paid for each diagnosis code are
based on traditional Medicare records that are not audited or
verified in any way; requiring repayment of all amounts
seemingly “overpaid” to a Medicare Advantage
insurer based on audited records would ignore errors in CMS
records and violate the statutory requirement of actuarial
equivalence.[5]
This
argument ventures deep into the weeds of actuarial science
but is not actually disputed by the parties. Nor could CMS
really debate it: as a result of the comments it received,
CMS adopted a “Fee-for-Service Adjuster” or
“FFS Adjuster” to the results of RADV audits of
Medicare Advantage insurance contracts. The FFS Adjuster
reflects CMS's own estimate of the error rate in risk
factors and diagnosis codes submitted by healthcare providers
and paid by CMS for its traditional Medicare participants;
applied to the results of a RADV audit of a Medicare
Advantage insurer, it is designed to achieve actuarial
equivalence between the two. Thus, Medicare Advantage
providers must return to CMS any audited
“overpayments” to the extent that the
insurer's errors exceed the estimated error rate in CMS
payments under traditional Medicare. See Notice of
Final Payment Error Calculation Methodology for Part C
Medicare Advantage Risk Adjustment Data Validation
Contract-Level Audits (Feb. 24, 2012) (RADV Final
Methodology) at ¶ 5311-15.
UnitedHealth
asserts that the 2012 FFS Adjuster works to counteract the
fact that per-capita payments to Medicare Advantage insurers
are based on a less precise set of data- belonging to
CMS-than that which is reviewed during an audit. Their
argument, and CMS's eventual concurrence, are supported
by the American Academy of Actuaries, which strongly advised
CMS that it was not actuarially sound to compare unaudited
figures to calculate per-capita payments and then audited
figures to calculate overpayments. See Academy of
Actuaries Comment at ¶ 5236 (“This type of data
inconsistency not only creates uncertainty, it also may
create systematic underpayment, undermining the purpose of
the risk-adjustment system and potentially resulting in
payment inequities.”).
The
passage of the Patient Protection and Affordable Care Act
(ACA), Pub. L. No. 111-148, 124 Stat. 119 (2010), is also
directly relevant here. The ACA imposed an obligation on
Medicare Advantage insurers to report and return any
overpayments that an insurer discovers on its own.
See 42 U.S.C. § 1320a-7k(d)(1) (2012). This
section of the ACA defined “overpayment” as
“any funds that a person receives or retains under
[Medicare Advantage] to which the person, after applicable
reconciliation, is not entitled.” Id. §
1320a-7k(d)(4)(B). The law further required that any
“overpayment . . . be reported and returned [within] 60
days after the date on which the overpayment was
identified.” Id. § 1320a-7k(d)(2). If an
insurer in the Medicare Advantage program fails to return
such a discovered overpayment within 60 days of identifying
it, that failure renders the insurer's initial but faulty
claim for payment a violation of the False Claims Act (FCA).
Id. § 1320a-7k(d)(3) (“Any overpayment
retained by a person after the deadline for reporting and
returning the overpayment . . . is an obligation (as defined
in section 3729 (b)(3) of title 31) for purposes of section
3729 of such title.”); cf. False Claims Act,
31 U.S.C. § 3729(b)(3). Claims for overpayments under
the False Claims Act carry the potential for treble damages,
civil penalties, and debarment from Medicare. See 31
U.S.C. § 3729(a)(1)(G) (providing for civil penalties
and treble damages); 42 C.F.R. § 424.535(a) (describing
grounds for revocation of enrollment in the Medicare
program). Further, nongovernment qui tam plaintiffs
may bring FCA claims in federal court. See 31 U.S.C.
§ 3730(b).
The
Affordable Care Act established a basic statutory framework
but left several crucial terms undefined. It did not define
at what point an insurer might be said to have
“identified” an overpayment, thus triggering the
60-day clock; nor did it outline the scope of
“applicable reconciliation” or state how
“overpayments” and “actuarial
equivalence” in payments are related.
We come
to the 2014 Final Rule at issue here. CMS issued a notice of
proposed rulemaking in January 2014 and sought
comments.[6] CMS proposed to “clarify the
statutory definition of overpayment” with a new
regulation titled “Reporting and Returning
Overpayments, ” to be codified at 42 C.F.R. §
422.326. See 79 Fed. Reg. at 1996, 2055-56 (June 29,
2000) (AR80 at ¶ 139-40).
CMS
published its Final Rule on May 23, 2014, and in so doing
finalized 42 C.F.R. § 422.326 concerning
overpayments.[7] Under the 2014 Overpayment Rule,
any diagnostic code that is inadequately documented
in a patient's medical chart results in an
“overpayment.” Id. at 29, 921 (AR1313).
Further, an overpayment is “identified” whenever
a Medicare Advantage insurer determines, “or should
have determined through the exercise of reasonable diligence,
” that it had received an overpayment. Id. at
29, 923 (AR1315). CMS further defined reasonable diligence as
requiring “at a minimum . . . proactive compliance
activities conducted in good faith by qualified individuals
to monitor for the receipt of overpayments.”
Id. UnitedHealth alleges that these obligations
apply a simple negligence standard for purposes of False
Claims Act liability, which is contrary to the standards in
the False Claims Act itself. See 31 U.S.C. §
3729(b)(1) (defining “knowing” and
“knowingly” to include “actual knowledge,
...