UNITEDHEALTHCARE INSURANCE COMPANY, ET AL., APPELLEES v. XAVIER BECERRA, IN HIS OFFICIAL CAPACITY AS SECRETARY OF HEALTH AND HUMAN SERVICES, ET AL., APPELLANTS
No. 18-5326
United States Court of Appeals FOR THE DISTRICT OF COLUMBIA CIRCUIT
Argued November 3, 2020 Decided August 13, 2021
Appeal from the United States District Court for the District of Columbia (No. 1:16-cv-00157)
Weili J. Shaw, Attorney, U.S. Department of Justice, argued the cause for appellants. With him on the briefs were Ethan P. Davis, Acting Assistant Attorney General, and Mark B. Stern, Attorney. Michael S. Raab, Attorney, entered an appearance.
Daniel Meron argued the cause for appellees. With him on the brief was Matthew M. Shors.
David W. Ogden, Brian M. Boynton, and Kevin M. Lamb were on the brief for amicus curiae America‘s Health Insurance Plans in support of appellees.
Before: ROGERS, PILLARD and WALKER, Circuit Judges.
Opinion for the Court filed by Circuit Judge
PILLARD, Circuit Judge: UnitedHealthcare Insurance Company and other Medicare Advantage insurers under the umbrella of UnitedHealth Group Incorporated (collectively, UnitedHealth) challenge a rule the Centers for Medicare and Medicaid Services (CMS) promulgated under the Medicare statute,
As explained in more detail below, people who are eligible for Medicare may elect to receive their health insurance through a private insurer under Medicare Advantage rather than directly through the government under traditional Medicare, and approximately forty percent of beneficiaries have chosen Medicare Advantage. CMS pays private Medicare Advantage insurers, in a prospective lump sum each month, the amount it expects a month‘s care would otherwise cost CMS in direct payments to healthcare providers treating the same beneficiaries under traditional Medicare. For each Medicare Advantage beneficiary, CMS pays the insurer a per-capita amount that varies according to demographic characteristics and diagnoses that CMS has determined, based on its past experience
Payments to the Medicare Advantage program depend on participating insurers accurately reporting to CMS their beneficiaries’ salient demographic information and medically documented diagnosis codes. To better control erroneous payments, including those garnered from reported—but unsupported—diagnoses, Congress in 2010 amended the Medicare program‘s data-integrity provisions. The amendment specified a sixty-day deadline for reporting and returning identified overpayments and confirmed that such payments not promptly returned may trigger liability under the False Claims Act. See
UnitedHealth claims that it is unambiguous in the text of the Medicare statute that the Overpayment Rule is subject to a principle of “actuarial equivalence,” and that the Rule fails to comply. See
UnitedHealth contends that the actuarial-equivalence principle reaches beyond its statutory home to impose an implied—and functionally prohibitive—legal precondition on the requirement to return known overpayments. As UnitedHealth would have it, Congress clearly intended enforcement of the statutory overpayment-refund obligation, which the Overpayment Rule essentially parrots, to depend on a prior determination of actuarial equivalence. That principle, UnitedHealth says, prevents CMS from recovering overpayments under the Rule unless CMS first shows that the rate of payment errors to healthcare providers in traditional, fee-for-service Medicare is lower than the rate of payment errors to the Medicare Advantage insurer, or that CMS comprehensively audited the data from traditional Medicare before using it in the complex regression model—the CMS Hierarchical Condition Category (CMS-HCC) risk-adjustment model—that predicts the cost to insure Medicare Advantage beneficiaries.
There is no legal or factual basis for UnitedHealth‘s claim. Actuarial equivalence is a directive to CMS. It describes the goal of the risk-adjustment model Congress
Even if actuarial equivalence applied as UnitedHealth suggests, it would be UnitedHealth‘s burden to show the systematically skewed inaccuracies on which its theory depends, which it has not done. Also fatal to UnitedHealth‘s claim is that it never challenged the values CMS assigned to the risk factors it identified or the level of the capitation payments resulting from CMS‘s risk-adjustment model. It cannot belatedly do so in the guise of a challenge to the Overpayment Rule.
UnitedHealth‘s next claim relies on the Medicare statute‘s requirement that CMS annually compute and publish certain traditional Medicare data “using the same methodology as is expected to be applied in making payments” to Medicare Advantage insurers.
UnitedHealth‘s final claim is that the Overpayment Rule is arbitrary and capricious in violation of the Administrative Procedure Act (APA). That claim hinges on what UnitedHealth sees as an unexplained inconsistency between the Overpayment Rule and another error-correction mechanism to which Medicare Advantage insurers are subject: Risk Adjustment Data Validation (RADV) audits. With those audits, CMS proposed a systemic adjustment involving the traditional Medicare data used to model risk factors to account for any errors in that data set before requiring any contract-level repayments from insurers. UnitedHealth sees inconsistency in obligating repayments under the Overpayment Rule without any such adjustment. But the system-level adjustment that CMS said it would apply in the context of contract-level RADV audits came in direct response to concerns about actuarial equivalence. Because we hold that the actuarial-equivalence requirement does not pertain to the statutory overpayment-refund obligation or the Overpayment Rule challenged here, and the two error-correction mechanisms are plainly distinguishable in other ways, CMS‘s one-time intention to apply the adjustment in one context but not the other was reasonable.
In sum, nothing in the Medicare statute‘s text, structure, or logic applies actuarial equivalence to its separate overpayment-refund obligation, and thus the Overpayment Rule does not violate actuarial equivalence. For much the same reasons, we reject UnitedHealth‘s claim that the Rule violates the statute‘s “same methodology” requirement, and we also
BACKGROUND
Overpayment to Medicare Advantage insurers is a serious drain on the Medicare program‘s finances. In 2016 alone, audits of the data submitted by Medicare Advantage insurers to CMS showed that CMS paid out an estimated $16.2 billion for unsupported diagnoses, equal to “nearly ten cents of every dollar paid to Medicare Advantage organizations.” United States ex rel. Silingo v. WellPoint, Inc., 904 F.3d 667, 673 (9th Cir. 2018) (citing James Cosgrove, U.S. Gov‘t Accountability Off., GAO-17-761T, Medicare Advantage Program Integrity: CMS‘s Efforts to Ensure Proper Payments 1 (2017), https://www.gao.gov/assets/690/685934.pdf). UnitedHealth is the Nation‘s largest provider of Medicare Advantage plans. Meredith Freed et al., A Dozen Facts About Medicare Advantage in 2020, Kaiser Family Found. (Jan. 13, 2021), https://www.kff.org/medicare/issue-brief/a-dozen-facts-about-medicare-advantage-in-2020/.
A. Statutory and regulatory background
1.
Since 1965, most older adults and many people with disabilities in the United States have received their health insurance through Medicare, administered by CMS. In Medicare Parts A and B, or “traditional” Medicare, CMS itself acts as the insurer, paying healthcare providers directly for beneficiaries’ medical services. Medicare Part A covers inpatient hospital treatment and other institutional care and is generally provided without charge to Medicare-eligible individuals. But for outpatient services, like visits to doctors’ offices, the Medicare statute provides Medicare-eligible individuals a choice of whether and how to receive such coverage: They can receive that, too, by having the government pay providers for services, under Medicare Part B; or they can opt for private insurance paid for at least in part by the government, under Medicare Part C, also known as Medicare Advantage (and formerly known as Medicare+Choice).
Unlike Medicare Part A, coverage under Medicare Part B and Medicare Advantage generally requires payments from beneficiaries to the government or, if applicable, private insurance companies. Medicare Advantage insurers must provide coverage of at least the same services as Medicare-eligible individuals would receive through traditional Medicare,
Medicare Parts A and B and Medicare Advantage pay healthcare providers in different ways. Under Medicare Part A, CMS pays a hospital or institutional care provider based on a beneficiary‘s diagnoses at the time of discharge, which translate to a “Diagnosis-Related Group.” Under Medicare Part B, CMS pays outpatient providers on a fee-for-service basis under fee schedules that set the payment for each service provided, such as an office visit, examination, or immunization. A beneficiary‘s
Private Medicare Advantage insurers likewise pay healthcare providers based on the services provided to beneficiaries but, as noted above, under Part C those insurers themselves receive in advance a monthly lump sum from CMS for every beneficiary that they enroll, without regard to the services that the beneficiaries will actually receive. The prospective, lump-sum payment approach has the potential to curb costly and unnecessary overtreatment that the fee-for-service approach tends to encourage, and it favors preventative care and other health-protective measures, enabling cost efficiencies that can elude a fee-for-service system. See Advance Notice of Methodological Changes for CY 2004 Medicare+Choice Payment Rates, at 5 (Mar. 28, 2003), J.A. 115. The core idea is that a Medicare Advantage insurer that covers all of a beneficiary‘s health care at least as well as traditional Medicare but does so at lower cost may pocket the difference as earned revenue, or pass along that revenue to beneficiaries in the form of extra benefits meant to entice and retain subscribers.
2.
It is the Medicare statute that requires CMS to pay Medicare Advantage insurers in advance, on a monthly basis, for each of the Medicare-eligible beneficiaries that they insure.
shall adjust the payment amount . . . 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. The Secretary may add to, modify, or substitute for such adjustment factors if such changes will improve the determination of actuarial equivalence.
Specifically, identifying salient risk factors enables CMS to determine prospectively, based on Medicare Advantage beneficiaries’ actuarially relevant, known demographic and health characteristics, the per-capita payment rate that will fairly compensate that Medicare Advantage insurer. More broadly, the demographic- and health-adjusted, capitated payment scheme is designed to blunt the incentives to enroll only the healthiest,
To adjust the monthly payments, CMS uses a model—called the CMS Hierarchical Condition Category, or CMS-HCC, risk-adjustment model—that it periodically studies and improves based on clinical information and cost data. The model isolates demographic characteristics CMS has determined to be predictive of differing costs of care, including the risk factors expressly mentioned in the statute: age, sex, disability status, and whether the beneficiary lives in a long-term institutional setting. See
CMS‘s risk-adjustment model applies a regression analysis to the mass of data from traditional Medicare for a previous year to convert each demographic and health characteristic into an expected cost of coverage. See id. at 2, J.A. 519. CMS inputs traditional Medicare beneficiaries’ data, including the diagnosis codes that healthcare providers are required to report (even though, as noted above, CMS itself bases Medicare Part B payments on services, not diagnoses), along with the total cost for covering those beneficiaries. The model isolates the anticipated cost of care associated with each demographic and health characteristic by first determining the average marginal cost of that characteristic in dollars and then dividing that dollar amount by traditional Medicare‘s average cost per beneficiary. That process produces a “relative factor” for each demographic and health characteristic. The model “use[s] data from a large pool of beneficiaries (full sample sizes over 1 million for the CMS-HCC models) to estimate predicted costs on average for each of the component factors (e.g., age-sex, low income status, individual disease groups).” Id. at 5, J.A. 522. Using regression analysis on such a vast data sample mutes the effect of individual errors in traditional Medicare data, so long as errors are not so widespread or systemically skewed as to raise or lower the values of particular relative factors. See id.; see also Amy Gallo, A Refresher on Regression Analysis, Harv. Bus. Rev. (Nov. 4, 2015), https://hbr.org/2015/11/a-refresher-on-regression-analysis.
To enable CMS to apply those relative factors to pay Medicare Advantage insurers at the correct risk-adjusted rate, the insurers must report to CMS the salient demographic and health characteristics of each of their Medicare-eligible beneficiaries.
CMS illustrates the operation of relative factors with an example:
[U]nder the 2014 model, a 72-year-old woman living independently (relative factor 0.348), with diabetes without complications (relative factor 0.118), and multiple sclerosis (relative factor 0.556) would have a total risk score of 1.022, which means that she is expected to cost Medicare slightly more than the average traditional Medicare beneficiary (who would by definition have a risk score of 1.0).
Gov‘t Br. 7 (citing Announcement of CY 2014 Medicare Advantage Capitation Rates and Medicare Advantage and Part D Payment Policies and Final Call Letter, at 67-68 (Apr. 1, 2013), J.A. 276-77). In other words, as a woman near the younger end of the Medicare-eligible population and living outside any long-term institutional setting, this sample beneficiary starts with a risk score well below the overall Medicare average. The fact that she suffers from diabetes raises her risk score, but not by much, presumably because she has not experienced complications and ordinary diabetes care is not as costly as many other conditions common among older Americans. The larger bump, putting her over the average predicted cost of care even for the cost-intensive Medicare population, is that she suffers from multiple sclerosis. A Medicare Advantage insurer providing coverage to this woman therefore “would be paid 102.2 percent of the relevant base rate.” Id. at 8.
This example illustrates the importance of risk-adjusted payment. Assume a similar woman, but without her diagnoses. With a risk score of just 0.348, her care would then be predicted to be far less expensive than that of the average Medicare beneficiary, whose risk score is, by definition, 1.0. If Medicare Advantage insurers were paid an unadjusted base rate for every beneficiary, they could receive an enormous, and unjustified, net surplus insofar as they enrolled beneficiaries with such low anticipated costs. Conversely, an unadjusted, per-capita base payment would likely fall far short of fairly compensating a Medicare Advantage insurer for the costs of care for the woman in the example with both of the posited diagnoses, and the shortfall would only grow with any added complications or diagnoses she developed.
There is some evidence that Medicare Advantage insurers in fact have tended to attract healthier-than-average beneficiaries—perhaps because of the additional premiums they may charge, and the well-established correlation between wealth and health. See Is Medicare Advantage More Efficient than Traditional Medicare?, Nat‘l Bureau of Econ. Rsch. (Mar. 2016), https://www.nber.org/bah/2016no1/medicare-advantage-more-efficient-traditional-medicare;
CMS determines the base payment rate—which, again, is the amount a Medicare Advantage insurer would receive for any beneficiary with a risk score of exactly 1.0, and which is the denominator for calculation of every capitated payment to Medicare Advantage—by reference to traditional Medicare‘s per-capita expenditures in a particular place and bids submitted by Medicare Advantage insurers. Each county in the United States has its own base rate, and every year Medicare Advantage insurers bid for contracts after CMS announces each county‘s benchmark for the coming year. See
3.
CMS‘s regulations have long obligated Medicare Advantage insurers to certify the accuracy of the data that they report to CMS. Since 2000, those regulations have made it “a condition for receiving a monthly payment” that a Medicare Advantage insurer
agrees that its chief executive officer (CEO), chief financial officer (CFO), or an individual delegated the authority to sign on behalf of one of these officers, and who reports directly to such officer, must request payment under the contract [with CMS] on a document that certifies (based on best knowledge, information, and belief) the accuracy, completeness, and truthfulness of relevant data that CMS requests.
But, as Congress has recognized, even accurate diagnosis codes that Medicare Advantage insurers report can lead to disproportionately high payments to insurers. That is because Medicare Advantage insurers have a financial incentive to code intensely—i.e., to make sure that they report to CMS their beneficiaries’ every diagnosis—given that their monthly, per-capita payments are higher to the extent that their beneficiaries have more or graver diagnoses. Meanwhile, healthcare providers to traditional Medicare beneficiaries lack that same incentive because their payments from CMS depend on services rendered, not diagnoses. See U.S. Gov‘t Accountability
UnitedHealth‘s challenge to the Overpayment Rule adverts to yet another data-integrity measure providing for Risk Adjustment Data Validation, or RADV, audits. To supplement the regulatory obligations on Medicare Advantage insurers to certify the accuracy of the diagnosis codes and other data they report to CMS, and because CMS cannot confirm in real time the data insurers submit for their millions of beneficiaries, CMS seeks to confirm that its payments to insurers are correct by retrospectively spot-checking the data submissions going back several years. See
CMS has conducted such audits for well over a decade, and their results show that a significant number of reported diagnoses are in fact unsupported. See, e.g., U.S. Dep‘t of Health & Human Servs., Off. of Inspector Gen., Risk Adjustment Data Validation of Payments Made to PacifiCare of Texas for Calendar Year 2007, A-06-09-00012, at 4 (May 2012), J.A. 471 (stating that the risk scores for forty-three out of 100 sampled beneficiaries of the audited insurer “were invalid because the diagnoses were not supported“); U.S. Dep‘t of Health & Human Servs., Off. of Inspector Gen., Risk Adjustment Data Validation of Payments Made to PacifiCare of California for Calendar Year 2007, A-09-09-00045, at i (Nov. 2012), J.A. 476 (stating that the risk scores for forty-five out of 100 sampled beneficiaries “were invalid because the diagnoses were not supported by the documentation that [the Medicare Advantage] insurer provided“).
Medicare Advantage insurers’ obligation to return mistaken payments pursuant to RADV audits differs from their obligation under the Overpayment Rule: With the former, insurers are required to refund payments based on unsupported diagnoses that CMS discovers through its audit,
In 2008, CMS announced an expansion of its RADV audit program for Medicare Advantage: Rather than requiring repayments only for the unsupported diagnosis codes identified in the limited sample itself, CMS would take the payment error in an audited sample, extrapolate that error rate across CMS‘s entire contract with that Medicare Advantage insurer, and require the insurer to make a repayment based on the extrapolated, or contract-level, degree of error. See Medicare Program; Policy and Technical Changes to the Medicare Advantage and the Medicare Prescription Drug Benefit Programs, 74 Fed. Reg. at 54,674, J.A. 96; see also Announcement of Calendar Year (CY) 2009 Medicare Advantage Capitation Rates and Medicare Advantage and Part D Payment Policies, at 22 (Apr. 7, 2008). (Because not all errors are created equal—that is, some are more costly than others—the extrapolated error rate would account for the magnitude of the errors by factoring in the difference between original and corrected payment amounts in an audited sample.) In late 2010, CMS sought comments on its proposal for contract-level RADV audits, and in early 2011 various commenters, including UnitedHealth and the American Academy of Actuaries, objected.
One criticism the commenters leveled at expanded RADV audits was that, “[u]nder sound actuarial principles, it is impossible to know whether [Medicare Advantage insurers] have been paid accurately by conducting a review of the medical records supporting [Medicare Advantage] coding, without also considering the medical records supporting [traditional Medicare] coding.” Aetna Inc.‘s Comments on Proposed Payment Error Calculation Methodology for Part C Organizations Selected for Contract-Level RADV Audits, at 4 (Jan. 21, 2011), J.A. 298. In other words, “CMS must audit and validate both [a Medicare Advantage insurer‘s data and the traditional Medicare data that goes into the risk-adjustment model] before extrapolating any potential RADV audit results” and requiring the insurer to return amounts thereby identified as excessive. Humana Inc., Comment on RADV Sampling and Error Calculation Methodology, at 3 (Jan. 21, 2011), J.A. 334. “If it does not, CMS will dramatically underpay [Medicare Advantage insurers] for the benefits they provided to Medicare beneficiaries,” in violation of the Medicare statute‘s actuarial-equivalence requirement. Id.; see also id. at 5, J.A. 336.
In a move that UnitedHealth describes as important context for this case, CMS responded to the comments by announcing in 2012 that it would apply a Fee-for-Service, or FFS, Adjuster before requiring repayments based on contract-level RADV audits. With the FFS Adjuster, Medicare Advantage insurers would be liable for repayments only to the extent that their extrapolated, contract-level payment errors, i.e., the dollar amounts that they received in error, exceed any offsetting payment error in traditional Medicare. CMS said that it would determine the actual amount of the FFS Adjuster “based on a RADV-like review of records submitted to support [traditional Medicare] claims data.” Notice of Final Payment Error Calculation Methodology for Part C Medicare Advantage RADV Contract-Level Audits, at 5 (Feb. 24, 2012), J.A. 398.
4.
Against the backdrop of concern about costly errors in the data reported by Medicare Advantage insurers, but before CMS even solicited comments on the proposed FFS Adjuster to contract-level RADV audits it ultimately deemed unnecessary, Congress enacted the provision that undergirds the Overpayment Rule.
In 2014, CMS promulgated the Overpayment Rule to implement the statutory requirement to report and return overpayments. The Rule similarly defines “overpayment” as “any funds that [a Medicare Advantage insurer] has received or retained under [the Medicare Advantage program] to which the [Medicare Advantage insurer], after applicable reconciliation, is not entitled.” Overpayment Rule, 79 Fed. Reg. at 29,958 (codified at
One commenter on the proposed Overpayment Rule, a Medicare Advantage insurer not a party to this case, had objected that it ran afoul of the Medicare statute‘s actuarial-equivalence requirement because it did not also require an adjuster akin to the FFS Adjuster that CMS had proposed two years earlier in the context of contract-level RADV audits. See id.; see also J.A. 50-51 (comment from Humana on proposed rule). In the final Rule, which does not provide for such an adjuster, CMS stated that it “disagree[d] with the commenter” because the “RADV methodology does not change [CMS‘s] existing contractual requirement that [Medicare Advantage insurers] must certify (based on best knowledge, information, and belief) the accuracy, completeness, and truthfulness of the risk adjustment data they submit to CMS.” Overpayment Rule, 79 Fed. Reg. at 29,921, J.A. 64. Nor, said CMS, did the statutory overpayment-refund obligation, as implemented by the Rule, “change the long-standing risk adjustment data requirement that a diagnosis submitted to CMS by [a Medicare Advantage insurer] for payment purposes must be supported by medical record documentation.” Id. at 29,921-22, J.A. 64-65.
B. Factual and procedural history
UnitedHealth filed this challenge to the Overpayment Rule in January 2016. Following the district court‘s denial of CMS‘s motion to dismiss in March 2017, the parties cross-moved for summary judgment. On September 7, 2018, the court granted UnitedHealth‘s motion in full and vacated the Overpayment Rule. See UnitedHealthcare Ins. Co. v. Azar, 330 F. Supp. 3d 173, 192 (D.D.C. 2018).
The district court held that the Overpayment Rule violated the Medicare statute‘s requirement of “actuarial equivalence.” Id. at 187. It concluded that the Rule would “inevitabl[y]” lead to the loss of actuarial equivalence, id. at 185, because “payments for care under traditional Medicare and Medicare Advantage are both set annually based on costs from unaudited traditional Medicare records, but the 2014 Overpayment Rule systematically devalues payments to Medicare Advantage insurers by measuring ‘overpayments’ based on audited patient records,” id. at 184. The court emphasized that CMS had actually “recognized and mitigated” “the same actuarial problem” when, in 2012, it provisionally committed to using an FFS Adjuster for contract-level RADV audits to account for the fact that extrapolating an error rate across a Medicare Advantage insurer‘s entire contract effectively corrected for any unsupported codes in the insurer‘s data. Id. Relying on much the same reasoning, the court held that the Rule also violated the Medicare statute‘s “same methodology” requirement. Id. at 187. The court then deemed the Rule arbitrary and capricious in violation of the APA as an unexplained departure from CMS‘s prior policy, namely, its stated intent to use an FFS Adjuster in the context of contract-level RADV audits. Id. at 187-90. The court noted only in passing that CMS had not yet determined an appropriate amount of any FFS Adjuster for contract-level RADV audits. See id. at 188.
The district court also rejected the Overpayment Rule‘s imposition of a negligence standard of liability for failure to identify and report an overpayment. The Rule as promulgated provided that a Medicare Advantage insurer “has identified an overpayment when the [insurer] has determined,
In November 2018, CMS moved for partial reconsideration, which the court denied in January 2020. CMS based that motion on the results of the October 2018 study of the error rate in traditional Medicare, conducted as groundwork for the anticipated FFS Adjuster for contract-level RADV audits. As noted above, the results of that study were made public several weeks after the district court‘s summary judgment ruling in this case. The study revealed that “errors in [traditional Medicare] claims data do not have any systematic effect on the risk scores calculated by the CMS-HCC risk adjustment model,” undermining the case for an adjuster. CMS Study at 5, J.A. 731; see also UnitedHealthcare Ins. Co. v. Azar, No. 16-cv-157, 2020 WL 417867, at *1, *3 (D.D.C. Jan. 27, 2020), J.A. 801, 805. In denying the motion, the district court stated that it “need not linger on the details of the[] arguments” regarding the validity of the study and CMS‘s preliminary conclusion not to apply any FFS Adjuster to contract-level RADV audits. UnitedHealthcare, 2020 WL 417867, at *5, J.A. 811. The court deemed it “sufficient to say that [UnitedHealth‘s] arguments [opposing the study] are fully explained and the government does not adequately respond.” Id.
CMS timely appealed on November 6, 2018, and we removed the case from abeyance in February 2020 following the district court‘s denial of reconsideration.
Finally, it bears noting that the issue of actuarial equivalence has come up in other litigation between the parties. The federal government and qui tam plaintiffs have pursued several False Claims Act cases against Medicare Advantage insurers in the last several years, charging failures to report and return overpayments that the insurers knew were based on unsupported diagnoses. At least some such cases are still pending. See, e.g., United States ex rel. Poehling v. UnitedHealth Grp., Inc., No. 16-cv-8697 (C.D. Cal.); United States ex rel. Osinek v. Kaiser Permanente, No. 13-cv-3891 (N.D. Cal.). Medicare Advantage insurers, including UnitedHealth, have raised actuarial equivalence as a defense to False Claims Act liability. See Appellees Br. 55. At least one court has
DISCUSSION
We review a district court‘s grant of summary judgment de novo. See, e.g., Clarian Health W., LLC v. Hargan, 878 F.3d 346, 352 (D.C. Cir. 2017). Under the APA, we must “hold unlawful and set aside agency action” that is “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law” or “in excess of statutory jurisdiction, authority, or limitations, or short of statutory right.”
A. The Overpayment Rule does not violate the Medicare statute‘s requirement of “actuarial equivalence”
UnitedHealth‘s central challenge to the Overpayment Rule is that it violates the Medicare statute‘s command to CMS to adjust payment amounts to a Medicare Advantage insurer based on risk factors “so as to ensure actuarial equivalence” between that insurer‘s beneficiary population and the traditional Medicare beneficiaries whose healthcare cost data CMS uses to calculate capitated, monthly payments to the insurer.
UnitedHealth claims, and the district court agreed, that before CMS may lawfully apply the Overpayment Rule, it must implement one of two measures to remedy the claimed imbalance. First, CMS could devise and apply an adjuster akin to the FFS Adjuster it had intended to use (but since has preliminarily decided is unwarranted) in the context of contract-level RADV audits of Medicare Advantage insurers’ risk-adjustment data. In that scenario, Medicare Advantage insurers would be liable for overpayments only to the extent that their payment error rate exceeded that of traditional Medicare. Alternatively, CMS could comprehensively audit traditional Medicare data before using it in
There are two main problems with UnitedHealth‘s argument. First, nothing in the Medicare statute‘s text, structure, or logic makes the actuarial-equivalence requirement in
1.
We have not previously decided any case involving “actuarial equivalence” as referenced in
The parties disagree about whether the Overpayment Rule even implicates the actuarial-equivalence requirement. UnitedHealth assumes the Overpayment Rule creates a sweeping obligation that effectively requires Medicare Advantage insurers to self-audit all their data. It thus asserts that, because of actuarial equivalence, before CMS may police overpayments in the manner of the Overpayment Rule, CMS must either audit traditional Medicare data before it goes into the risk-adjustment model or, alternatively, adopt a systemic corrective similar to the FFS Adjuster CMS contemplated in the context of proposed contract-level RADV audits. In the context of the RADV audit expansion, the insurers’ objection was that applying a sampled payment error rate across an entire contract would effectively audit all of an insurer‘s data while leaving unaudited the traditional Medicare data used to set monthly payments in the first place, thus requiring the application of an adjuster that would also effectively audit all of the data on the traditional Medicare side. Here, UnitedHealth asserts much the same: that the Overpayment Rule essentially requires insurers to audit all of the
UnitedHealth‘s premise is unsupported. Nothing in the Overpayment Rule obligates insurers to audit their reported data. As the district court held, see UnitedHealthcare, 330 F. Supp. 3d at 190-91, and CMS does not here dispute, see Gov‘t Br. 22, 30, the Rule only requires insurers to refund amounts they know were overpayments, i.e., payments they are aware lack support in a beneficiary‘s medical records. That limited scope does not impose a self-auditing mandate.
No part of the Medicare statute or the Overpayment Rule supports UnitedHealth‘s challenge. The statute‘s actuarial-equivalence requirement does not apply to the separate statutory obligation on insurers to refund overpayments they erroneously elicit from CMS; nor, by the same token, does actuarial equivalence apply to the Overpayment Rule that implements that statutory obligation and, in relevant part, essentially parrots it. Compare
More specifically, nothing in either provision renders actuarial equivalence a defense against the obligation to refund any individual, known overpayment. Notably, Congress through the Affordable Care Act strengthened Medicare Advantage insurers’ data-reporting obligations by requiring insurers to report and return overpayments within sixty days of their discovery, and it made specific provision for False Claims Act liability for those that do not. In so doing, Congress made no reference to the Medicare statute‘s longstanding actuarial-equivalence requirement, let alone any suggestion that it could be interposed as a defense. See
If anything, the text of
Stephens v. U.S. Airways Group, Inc., cited by the district court in support of its holding, see UnitedHealthcare, 330 F. Supp. 3d at 185-86, actually cuts the other way. There, we held that an ERISA actuarial-equivalence requirement did not obligate the airline to pay pensioners interest on requested lump-sum payments made well after annuity payments would have begun had the same benefit been disbursed periodically. Stephens, 644 F.3d at 440. When we held that interest was required under IRS regulations regarding unreasonable delay of such payments, id.; see also id. at 442, we were also clear that the lump-sum payments did not violate actuarial equivalence where the airline “accurately calculated [the] lump sums to be the ‘actuarial equivalent’ of the annuity option as of the annuity start date,” id. at 440. Because the actuarial equivalence of the annuity and lump-sum payments had been calculated based on a common initial payment date, and the statute was silent on whether interest was owed when an otherwise actuarially equivalent pension was paid later, we declined to grant the interest claim on that basis. Id.
Here, the Medicare statute is similarly silent, as it speaks not at all to whether the actuarial-equivalence requirement in
As CMS points out, 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
The actuarial-equivalence requirement and the overpayment-refund obligation apply to different actors, target distinct issues arising at different times, and work at different levels of generality. The actuarial-equivalence provision directs CMS to develop a system of relative factors to use in adjusting the amount of the monthly payments to each Medicare Advantage insurer. See id. It calls on CMS to use its expert judgment to identify cost-predictive risk factors in the Medicare population and to analyze the data accumulated in traditional Medicare to determine average costs associated with those factors.
The point of that exercise is to enable CMS to pay only as much for coverage of Medicare Advantage beneficiaries as it would if they were instead enrolled in traditional Medicare, notwithstanding differences between the actual populations—for example, that Medicare Advantage populations have tended to be healthier than traditional Medicare‘s population. See Reply Br. 20-21 (citing Pope et al., Risk Adjustment of Medicare Capitation Payments Using the CMS-HCC Model, at 119, J.A. 487). Thus, the actuarial-equivalence requirement is focused on accounting for the distinct profiles of each insurer‘s beneficiary population, listing “age, disability status, gender, institutional status, and . . . health status” as potentially relevant considerations in the risk-adjustment model.
Finally, applying actuarial equivalence to the Medicare statute‘s separate obligation to refund particular, known overpayments would seriously undermine that obligation, with the potential for absurd consequences. As UnitedHealth acknowledged at oral argument, under its view of actuarial equivalence as a defense against its obligation to reimburse CMS for known overpayments, a Medicare Advantage insurer could be entitled to retain payments that it knew were unsupported by medical records so long as CMS had not established that the insurer‘s overall payment error rate was higher than traditional Medicare‘s payment error rate. See Oral Arg. Tr. 50:12-18. Indeed, under that line of thinking, a Medicare Advantage insurer could knowingly submit unsupported diagnosis codes and retain payment for them unless and until CMS established—based on fully audited data of both traditional Medicare and the Medicare Advantage insurer at issue—that the particular overpayment resulted in a net gain to the insurer relative to traditional Medicare. There is no basis on which we can conclude
UnitedHealth asks us to rewrite the statutory overpayment-refund obligation, which was the basis for the Overpayment Rule, by narrowing the capacious “any funds” to which a Medicare Advantage insurer “is not entitled,”
2.
Even if the Medicare statute could theoretically support UnitedHealth‘s reading, we lack the necessary grounds here to invalidate the Overpayment Rule as a violation of actuarial equivalence. Recall that UnitedHealth‘s claim is that CMS cannot demand that UnitedHealth refund overpayments unless CMS shows it meets what UnitedHealth posits as a symmetrical auditing or error-correction obligation regarding traditional Medicare. But Congress has spelled out distinct obligations for traditional Medicare and Medicare Advantage, such as the Coding Intensity Adjuster that applies to the latter program but not the former, see
Congress‘s and CMS‘s use of measures tailored to the differing structures of and incentives in the two programs makes sense; indeed, it could be irrational not to use distinct tools as needed to respond to different problems. UnitedHealth does not challenge the Coding Intensity Adjuster imposed by Congress. And UnitedHealth has never taken the opportunity that arises annually to challenge the accuracy of the risk-adjustment model or pricing when CMS announces the relative factors and base payment rates that it will use for the upcoming year. See Oral Arg. Tr. 12:12-13:16; see also Ormsby, 444 F. Supp. 3d at 1068 n.442. We accordingly accept the unchallenged validity of the overall design of the model, the risk factors considered by CMS pursuant to its discretion under
UnitedHealth has failed to provide any logical or empirical basis to question the accuracy of traditional Medicare data. UnitedHealth asserts that the obligation to refund overpayments, at least as defined by the Overpayment Rule, leads to systematic underpayment of Medicare Advantage insurers relative to traditional Medicare. But it is by no means “inevitable” that Medicare Advantage insurers will be underpaid without the correctives that UnitedHealth
UnitedHealth complains of “a substantial number” of unsupported diagnosis codes in the minimally audited traditional Medicare data set. Appellees Br. 26. But UnitedHealth identifies no reason why the traditional Medicare data that goes into the risk-adjustment model would suffer systematically from unsupported codes like those the Overpayment Rule targets, i.e., codes lacking substantiation in medical records. If anything, the fact that providers for traditional Medicare beneficiaries are generally paid based on services, not diagnoses, would seem to tend toward underreporting, not overreporting, of diagnoses within traditional Medicare. The underlying premise of UnitedHealth‘s overall position is that traditional Medicare data includes a significant rate of unsupported diagnosis codes that ultimately depresses the payments to Medicare Advantage insurers. But the different ways the programs’ reimbursement schemes work in practice make that premise implausible.
Nor has UnitedHealth established another premise of its position—that the unsupported codes it posits in traditional Medicare would both be materially analogous to those the Overpayment Rule targets, and would cause UnitedHealth to be underpaid. To start, it is not even clear which kind of payment error in traditional Medicare, relative to Medicare Advantage, UnitedHealth believes is overlooked to its detriment. UnitedHealth identifies the problem in traditional Medicare as “a substantial number” of unsupported codes, id., though, as discussed more below, it does not specify what, if any, payment implications it sees as necessarily attending them. To the extent that unsupported codes in traditional Medicare would be associated with erroneous payments that CMS need not have made to healthcare providers—i.e., overpayments analogous to any CMS makes to Medicare Advantage insurers and targets with the Overpayment Rule—that kind of error would, if anything, tend to raise, not lower, overall payments to Medicare Advantage insurers. That is, because CMS‘s expenditures on traditional Medicare contribute to setting the base rate later used to make payments to Medicare Advantage insurers, the more money CMS spends on traditional Medicare, the higher the baseline for its expenditures on Medicare Advantage.
UnitedHealth nonetheless defends its position and the district court‘s ruling as founded “on straightforward math: Including unsupported codes when allocating costs on the traditional Medicare side, then excluding those same codes when determining payment amounts on the [Medicare Advantage] side, will underpay plans.” Id.
UnitedHealth‘s twin example ignores that unsupported codes are likely to occur for different reasons and with differing effects in the two programs: Unlike an unsupported diabetes code associated with Twin B in Medicare Advantage, which leads to an unwarranted increase in payment to the insurer, the mere existence of an unsupported diabetes code for Twin A in traditional Medicare does not mean CMS spent more money on that beneficiary. That is, CMS‘s expenditure for Twin A (at least in fee-for-service Part B) is not likely to have been higher if she were miscoded as diabetic than it would be without that error. CMS‘s expenditure on the twin in traditional Medicare would increase only if CMS paid for treatment corresponding to that unsupported code. But if Twin A‘s unsupported diabetes code is only an administrative error that does not correspond to treatment actually provided and paid for, UnitedHealth‘s hypothetical uses the wrong starting point, and so the wrong figures, for Twin A‘s side of the comparison. Her costs in traditional Medicare from the outset (and even if her unsupported diabetes code is never caught) would be at the same, lowered level as Twin B‘s in Medicare Advantage once that diabetes code was removed—in both cases, the payment level appropriate for a non-diabetic.
Even assuming Twin A‘s unsupported diabetes code were associated with erroneous payment by CMS, one would need to know more about the nature and scale of such errors to determine whether they could have affected the results of the regression analysis used to calculate relative factors, and in what direction. For example, if UnitedHealth is assuming that Twin A‘s unsupported diabetes code triggered payment for treatment that had no medical purpose, UnitedHealth still has not made its case of inevitable underpayment. Specifically, if an unsupported code in traditional Medicare pairs with diabetes treatment for which CMS paid, UnitedHealth has not explained how, in coding it as just that—a cost of diabetes treatment, however unnecessary—CMS would inevitably depress the value of the relative factor for diabetes. As UnitedHealth sees it, every unsupported diabetes code in traditional Medicare lowers the value of the relative factor for diabetes, as CMS‘s expenditure on diabetes is divided among more and more beneficiaries. But UnitedHealth does not account for the possibility of an unsupported code associated with payment by CMS, which would enlarge both the total costs and the beneficiary pool in traditional Medicare and thus, if anything, tend to keep constant the value of the relative factor at issue.
The implications of any unsupported diabetes code in traditional Medicare are quite different from those of the same unsupported code in Medicare Advantage. The former will not lead to Medicare Advantage insurers’ inevitable underpayment because, as already noted, any erroneous code in traditional Medicare is aggregated with millions of others in the regressions called for under the risk-adjustment model. Errors that are isolated and random, not systemic, cannot alone be assumed to affect the value of a relative factor that bears on how much CMS will pay Medicare Advantage insurers for beneficiaries with any particular condition. An unsupported code submitted by a Medicare Advantage insurer, in contrast, triggers overpayment in every case. That is because individual codes in that program are used to determine payments, not as data points in a complex and rigorous statistical model.
In sum, UnitedHealth has given no reason to think that miscoding in traditional Medicare necessarily leads to any inflated or deflated relative factors and, if it did, which ones are affected in which direction. We cannot assume based on UnitedHealth‘s reasoning alone that Medicare Advantage insurers are inevitably underpaid under any of the circumstances possible in its example.
What‘s more, the empirical evidence that we do have—CMS‘s October 2018 study concerning an FFS Adjuster in the context of contract-level RADV audits—suggests that Medicare Advantage insurers are not underpaid relative to traditional Medicare, contrary to UnitedHealth‘s and the district court‘s belief that underpayment is inevitable. Through that study, CMS “found that errors in [traditional Medicare] claims data do not have any systematic effect on the risk scores calculated by the CMS-HCC risk adjustment model, and therefore do not have any systematic effect on the payments made to [Medicare Advantage] organizations.” CMS Study at 5, J.A. 731. In fact, CMS determined that the impact of errors in traditional Medicare data “is less than one percent on average and in favor of the [Medicare Advantage] plans.” Id.
Together with its opposition to CMS‘s motion for partial reconsideration before the district court, UnitedHealth submitted a declaration from an actuarial expert “reflect[ing] [the expert‘s] professional interpretation” of CMS‘s study. Declaration of Julia Lambert at 2, UnitedHealthcare Ins. Co. v. Azar, 2020 WL 417867 (No. 16-cv-157), J.A. 771. UnitedHealth‘s expert criticized the study by asserting that the underlying data in fact showed that, “if you take [a Medicare Advantage insurer] with risk profiles identical to those in the [traditional Medicare] data, the [insurer] would be underpaid if the relative factors generated using both supported and unsupported data [from traditional Medicare] were applied only to supported codes in the [insurer‘s] data.” Id. at 19, J.A. 788. But neither CMS‘s study nor UnitedHealth‘s expert‘s declaration tells us what
The burden of proof is UnitedHealth‘s to show that the Overpayment Rule is unlawful. See, e.g., Abington Crest, 575 F.3d at 722 (citing City of Olmstead Falls, 292 F.3d at 271). In the absence of such proof—or even persuasive logic in UnitedHealth‘s favor—we could not here invalidate the Overpayment Rule as violating actuarial equivalence even if we held that such requirement bore on the overpayment-refund obligation.
B. The Overpayment Rule does not violate the Medicare statute‘s requirement of “same methodology”
UnitedHealth‘s second claim—that the Overpayment Rule violates the Medicare statute‘s “same methodology” requirement in
As UnitedHealth acknowledges, the annual computation and publication requirement in
The “same methodology” requirement plays a specific role in the computation and publication of data to aid the bidding process. It does not impose a substantive limit on the operation of the risk-adjustment model, which is governed by a separate provision. Nor does it have any bearing on whether a particular payment to a Medicare Advantage insurer constitutes an “overpayment.” Rather, the requirement to “us[e] the same methodology” clarifies
C. The Overpayment Rule is not an unexplained departure from prior policy
UnitedHealth‘s third and final claim on appeal is that CMS‘s response to a comment calling for the use of an adjuster under the Overpayment Rule was arbitrary and capricious in violation of the APA. That comment advocated “appl[ication of] the principles adopted by CMS in the RADV audit context” to argue that “the sole instance in which an ‘overpayment’ can be determined” is when CMS first has shown that the overall payment error for a given Medicare Advantage insurer is higher than that in traditional Medicare. Overpayment Rule, 79 Fed. Reg. at 29,921, J.A. 64.
In 2012, CMS proposed to use an FFS Adjuster in the context of contract-level RADV audits used to review Medicare Advantage insurers’ risk-adjustment data. It did so in response to objections by Medicare Advantage insurers and the American Academy of Actuaries that failure to use an adjuster would violate the Medicare statute‘s requirement of “actuarial equivalence.” Specifically, those commenters had argued that the actuarial-equivalence requirement prohibited CMS from using traditional Medicare data—which is subject to minimal auditing—to make monthly payments to Medicare Advantage insurers in the first instance, but then requiring an insurer to return some portion of those payments once CMS had effectively audited all the insurer‘s data by applying an extrapolated payment error rate to its entire contract with CMS. See, e.g., Aetna Inc.‘s Comments on Proposed Payment Error Calculation Methodology for Part C Organizations Selected for Contract-Level RADV Audits, at 4 & 18-22, J.A. 298 & 312-16; Humana Inc., Comment on RADV Sampling and Error Calculation Methodology, at 2-5 & 12, J.A. 333-36 & 343. Notably, the Academy did not object to the proposed Overpayment Rule based on actuarial equivalence, and CMS has preliminarily decided not to use an FFS Adjuster for contract-level RADV audits after all because “errors in [traditional Medicare] claims data do not have any systematic effect on the risk scores calculated by the CMS-HCC risk adjustment model.” CMS Study at 5, J.A. 731.
Because, as discussed above, the Overpayment Rule does not violate, or even implicate, actuarial equivalence, CMS had no obligation to consider an FFS Adjuster or similar correction in the overpayment-refund context. Contract-level RADV audits, which would effectively eliminate—and require repayment for—all unsupported codes in a Medicare Advantage insurer‘s data, are an error-correction mechanism that is materially distinct from the Overpayment Rule challenged here, which requires only that an insurer report and return to CMS known errors in its beneficiaries’ diagnoses that it submitted as grounds for upward adjustment of its monthly capitation payments. Thus, CMS was not required to provide further explanation of its decision. See Motor Vehicles Mfrs. Ass‘n v. State Farm Mut. Auto. Ins. Co., 463 U.S. 29, 43 (1983). CMS‘s response to the comment reiterated Medicare Advantage insurers’ longstanding obligations, under
* * *
For the foregoing reasons, we hold that the Overpayment Rule does not violate the Medicare statute‘s “actuarial equivalence” and “same methodology” requirements and is not arbitrary and capricious as an unexplained departure from prior policy. We accordingly reverse the judgment of the district court and remand this case with orders to enter judgment in favor of Appellants.
So ordered.