United States Court of Appeals FOR THE DISTRICT OF COLUMBIA CIRCUIT
Argued September 6, 2022 Decided September 1, 2023
No. 20-5350
POMONA VALLEY HOSPITAL MEDICAL CENTER, APPELLEE/CROSS-APPELLANT
v.
XAVIER BECERRA, APPELLANT/CROSS-APPELLEE
Appeal from the United States District Court for the District of Columbia (No. 1:18-cv-02763-ABJ)
Sven C. Collins argued the cause for appellee/cross- appellant. On the briefs were Robert L. Roth and Kelly A. Carroll.
Stephanie R. Marcus, Attorney, U.S. Department of Justice, argued the cause for appellant/cross-appellee. On the briefs were Mark B. Stern and Brian M. Boynton.
Before: MILLETT, KATSAS, and WALKER, Circuit Judges.
Opinion for the Court filed by Circuit Judge KATSAS. 2
KATSAS, Circuit Judge: Hospitals receive greater payment if their Medicare patients are disproportionately low-income individuals entitled to federal supplemental security income benefits. Pomona Valley Hospital Medical Center contends that the Department of Health and Human Services undercounted the number of its Medicare patients who were entitled to SSI benefits and thus undercompensated the hospital for treating them. Prohibited from directly accessing the relevant SSI data, Pomona sought to prove the undercount through data from state benefit programs that piggyback on SSI. In an administrative proceeding, Pomona introduced expert testimony explaining how the state data derives from and overlaps with the federal SSI data. HHS offered no evidence in response. The Provider Reimbursement Review Board held that Pomona failed to prove the undercount, but the district court set aside its decision and remanded the case to the Board for further proceedings. We affirm the district court.
I
A
The Department of Health and Human Services administers Medicare, which provides health insurance to the elderly and disabled. 42 U.S.C. § 1395c. For treating Medicare beneficiaries, hospitals receive payments fixed by a statutory formula. Cape Cod Hosp. v. Sebelius, 630 F.3d 203, 205 (D.C. Cir. 2011). One input is the “disproportionate share hospital” adjustment, which increases payments to hospitals that serve “a significantly disproportionate number of low- income patients.” 42 U.S.C. § 1395ww(d)(5)(F)(i)(I). This adjustment depends in part on something called the “Medicare fraction,” which represents the percentage of a hospital’s Medicare patients who are entitled to SSI benefits. More 3 precisely, the numerator of this fraction is the number of patient days attributable to Medicare patients who are “entitled to supplement[al] security income benefits,” and the denominator is the total number of patient days attributable to Medicare patients. Id. § 1395ww(d)(5)(F)(vi)(I); see Becerra v. Empire Health Found., 142 S. Ct. 2354, 2359–60 (2022). The upshot is that hospitals may receive larger payments if more of their patients are entitled to SSI benefits.
The Social Security Administration administers the SSI program. It gives cash payments to needy individuals who are elderly, blind, or disabled. 42 U.S.C. § 1382(a). Eligibility is determined monthly and depends on an individual’s income. Id. § 1382(c)(1). Because income and thus eligibility may vary over time, SSA tracks monthly (1) whether individuals enrolled in the SSI program qualified for and received the payment and (2) the reason why or why not. SSA has developed several dozen codes for this purpose, which consist of a letter and a two-digit number. For instance, the code “N01” indicates that an enrollee failed to receive a payment for a particular month (“N”) because he or she had excess income during that time (“01”).
States may contract with SSA to provide further assistance to needy residents. 42 U.S.C. § 1382e. SSA makes the state supplementary payments (SSP) for the state, which then must reimburse SSA. Id. § 1382e(d). SSP benefits must go to all state residents receiving SSI benefits, but the state may choose to extend them to certain other residents. Id. §1382e(b).
B
To determine the Medicare fractions of individual hospitals, HHS must rely on SSI data received from SSA. HHS makes these determinations through the Centers for Medicare 4 and Medicaid Services, which administers Medicare for HHS. In 2008, a district court held that CMS arbitrarily failed to use the best available SSI data in determining Medicare fractions. Baystate Med. Ctr. v. Leavitt, 545 F. Supp. 2d 20, 44 (D.D.C. 2008). In response, CMS promulgated a rule setting forth a new methodology for doing so. 75 Fed. Reg. 50,042, 50,275– 86 (Aug. 16, 2010) (2010 Rule). CMS applies this rule to determinations for years before 2010 as well as after.
Under the 2010 Rule, CMS uses two data sources to determine Medicare fractions. First, it maintains a Medicare Provider Analysis and Review (MedPAR) file, which contains information about hospital use by all Medicare beneficiaries. From this data, CMS determines a hospital’s total patient days attributable to Medicare beneficiaries—i.e., the denominator of its Medicare fraction. 75 Fed. Reg. at 50,277–78. Second, CMS obtains an expanded SSI-eligibility data file from SSA. Id. This file enables CMS to identify, on a month-by-month basis, SSI enrollees to whom SSA has assigned one of three codes: C01, M01, and M02. In the 2010 Rule, CMS analyzed the various SSA codes and concluded that these three—and no others—“accurately capture[] all SSI-entitled individuals during the month(s) that they are entitled to receive SSI benefits.” Id. at 50,281. CMS cross-checks whether Medicare beneficiaries listed in its MedPAR file have been assigned one of these three codes at the time of their hospitalization. The numerator of a hospital’s SSI fraction is the number of patient days attributable to Medicare patients who have been so assigned one of these codes.
The Medicare Prescription Drug, Improvement, and Modernization Act requires CMS to give each hospital the “data necessary” for the hospital to “compute the number of patient days” used in its Medicare fraction. Pub. L. No. 108– 173, § 951, 117 Stat. 2066, 2427 (2003). To that end, CMS 5 gives each hospital the MedPAR data for that hospital, together with “the results of the data match of SSI eligibility information.” 70 Fed. Reg. 47,278, 47,439 (Aug. 12, 2005). In other words, CMS tells the hospital which of its patient days recorded in the MedPAR file have been matched to patients entitled to SSI benefits when they were hospitalized. But CMS does not give hospitals the SSI eligibility file that it receives from SSA. According to CMS, federal privacy laws prohibit it from disclosing this information, as does CMS’s data-sharing agreement with SSA. Id. at 47,440. In the 2010 Rule, CMS once again declined to give hospitals “access to patient-level detail data, including SSI eligibility information.” 75 Fed. Reg. at 50,279.
C
To receive compensation for treating Medicare patients, a hospital must submit annual cost reports to a Medicare Administrative Contractor, which determines the hospital’s total annual reimbursement on behalf of CMS. 42 U.S.C. § 1395kk-1; 42 C.F.R. § 405.1801(b)(1). In making this determination, the Contractor uses the Medicare fraction determined by CMS. Id. §§ 405.1803, 412.106(b)(2).
Hospitals may appeal a Contractor’s reimbursement determination to the Provider Reimbursement Review Board, an administrative tribunal within HHS. 42 U.S.C. § 1395oo(a)(1)(A). Before the PRRB, the Contractor stands in for CMS. The Board must determine “whether the [hospital] carried its burden of production of evidence and burden of proof by establishing, by a preponderance of the evidence, that the [hospital] is entitled to relief on the merits of the matter at issue.” 42 C.F.R. § 405.1871(a)(3). 6 A final PRRB decision is reviewable in district court under the standards for judicial review set forth in the Administrative Procedure Act. 42 U.S.C. § 1395oo(f)(1).
II
Pomona Valley Hospital Medical Center is an acute-care hospital located some 30 miles east of downtown Los Angeles. This case involves Pomona’s Medicare fractions for fiscal years 2006 to 2008. Suspecting that CMS’s determinations were too low, Pomona sought to redo them. To that end, it obtained the matched MedPAR data from CMS. But CMS refused to give Pomona any underlying data from the SSI eligibility file, consistent with its longstanding regulations. SSA also refused to give Pomona the relevant data.
Undeterred, Pomona sought to determine its Medicare fractions with data obtained from state agencies administering two benefit programs that piggyback on SSI. One of these is the program affording state supplemental payments to needy Californians. The other is Medi-Cal, through which California participates in Medicaid and thus provides health insurance to needy Californians. The state SSP benefit has a higher income ceiling than does the federal SSI benefit, so some Californians receive SSP but not SSI benefits. No Californian receives SSI but not SSP benefits. Californians who receive either SSI or SSP benefits are also eligible for Medi-Cal during months in which they receive these benefits. The California Department of Healthcare Services, which administers Medi-Cal, ascertains this population of beneficiaries through monthly data provided by SSA. Upon receiving the SSA data, Medi- Cal assigns codes 10, 20, or 60 to individuals who are aged, blind, or disabled respectively and who are eligible for either 7 SSI or SSP benefits. The data provided by SSA does not distinguish between SSI and SSP eligibility, so neither do the Medi-Cal codes. Medi-Cal makes its data and codes available to providers, which use them to bill Medi-Cal for services provided to beneficiaries. And Medi-Cal uses the data to determine how much reimbursement it can claim from Medicaid. Expert testimony established all of this, and none of it is disputed.
Pomona determined its Medicare fractions using Medi-Cal codes 10, 20, and 60 to measure the relevant numerators—i.e., the number of its patient days attributable to patients receiving SSI benefits at the time of their hospitalization. To eliminate patients who received SSP but not SSI benefits, Pomona turned to data obtained from the California Department of Social Services, which coordinates with SSA to administer the state SSP program. This Department gave Pomona data showing the total number of Californians receiving SSI and SSP benefits, as well as the total number of Californians receiving only SSP benefits. Using this data, Pomona reduced the numerators in its calculations by about 16.5 percent, which, over the years in question, was the statewide percentage of individuals receiving SSP but not SSI benefits.
Pomona compared its results (based on adjusted Medi-Cal data) with CMS’s results (based on SSI data). Over the three fiscal years at issue, the Medicare fractions determined by CMS were about 20 percent lower than those determined by Pomona. That difference equates to disputed Medicare reimbursements of over $3 million.
Pomona proposed a way to settle the dispute. It offered to have CMS or SSA review a sample of 50 records from its Medi-Cal dataset, including 20 randomly selected records and the ten longest unmatched admissions—i.e., the ten admissions 8 of Medicare patients with the lengthiest hospital stays for which Pomona and CMS disagreed about whether the patient was SSI-eligible at the time of hospitalization. Pomona offered to be bound by the results of this sample review. CMS declined, citing “workload concerns.” J.A. 44.
As required by law, the assigned Contractor used the Medicare fractions calculated by CMS in determining the reimbursement owed to the hospital.
Pomona appealed to the PRRB. It did not challenge the methodology set forth in the 2010 Rule, including the protocols for selecting and matching the relevant SSA and CMS data. Instead, it claimed that CMS must have miscalculated because of inaccurate transmission of data from SSA to CMS, coding errors, or other systemic problems with the matching process.
Before the Board, Pomona presented testimony supporting its use of Medi-Cal data to count patient days attributable to patients receiving SSI or SSP benefits. Pomona’s star witness was Stan Rosenstein, who oversaw Medi-Cal eligibility issues for the California Department of Healthcare Services for over a decade. Rosenstein outlined the operation of, and relationship among, the SSI, SSP, and Medi-Cal programs. He explained how Medi-Cal, insofar as it affords benefits to recipients of SSI or SSP benefits, identifies those beneficiaries through eligibility data provided to it monthly by SSA. He explained how Medi-Cal, based on that data, assigns codes 10, 20, and 60 to patients who are eligible for either benefit. He explained how healthcare providers, Medi-Cal, and Medicaid rely on these codes in deciding which government agencies will pay for which treatments for which patients—and how much each agency will pay. He explained how Pomona used 9 the Medi-Cal data to determine which of its patient days were attributable to patients entitled to SSI or SSP benefits at the time of their hospitalization. And he opined that the Medi-Cal data and coding systems, which have been in use for four decades, are reliable for these purposes.
Pomona also presented testimony supporting its use of statewide averages to eliminate patient days attributable to its SSP-only patients. According to Rosenstein, the SSI and SSP populations do not vary much over time. And if anything, Pomona’s use of statewide averages likely overcorrected for its SSP-only patients: The hospital serves a relatively poor part of the state, so it likely had a disproportionately high percentage of patients in the lowest band of income, thus qualifying them for SSI as well as SSP benefits.
Finally, Pomona presented testimony supporting its conclusion that the observed counting discrepancies indicated some systemic problem in how CMS was acquiring SSA data and matching it to CMS’s own MedPAR data. Despite decades of experience with Medi-Cal eligibility issues, Rosenstein had never before seen such discrepancies and could think of no other way to reconcile them. So, he concluded that there had to be “something wrong” with the CMS data. J.A. 193. Pomona’s other principal witness corroborated this view. Tzvi Hefter, who once headed the CMS division that determined DSH adjustments, testified that the “really huge” difference between the Medi-Cal and CMS data indicated “some kind of systemic problem” with the latter. J.A. 167. In sum, the experts could think of no other reason why the Medi-Cal data revealed thousands of Medicare patient days attributable to SSI-eligible patients that were somehow dropped in the data- matching process between SSA and CMS. 10 The Contractor presented no countervailing evidence or explanation. Instead, it criticized Pomona for relying on “data from a secondary source, instead of the primary databases maintained by the SSA and CMS.” J.A. 276.
The Board concluded that Pomona had failed to prove an undercount. It “recognize[d] Pomona’s difficulty” because neither SSA nor CMS would provide it with any SSI eligibility data and because neither agency would consider a sample of Pomona’s proxy data based on Medi-Cal eligibility. J.A. 44. Nonetheless, the Board saw possible differences among the Medi-Cal, SSA, and CMS data, which it thought Pomona had not adequately addressed. The Board thus concluded that the hospital “did not submit sufficient quantifiable data in the record to prove that the SSI percentages calculated by CMS … were flawed.” Id. at 46.
On review, the district court held that the Board’s decision was unsupported by substantial evidence, so it granted summary judgment to Pomona. The court noted that the Contractor “did not even bother to produce evidence at the hearing” to justify CMS’s Medicare fractions. Pomona Valley Hosp. Med. Ctr. v. Azar, No. 18-2763, 2020 WL 5816486, at *11 (D.D.C. Sept. 30, 2020). The court also faulted the Board for placing too much weight on “minor” objections to Pomona’s affirmative case. Id. The court held that CMS had to do more because it had sole access to SSI eligibility records that could conclusively settle the dispute. Applying our decision in Atlanta College of Medical and Dental Careers, Inc. v. Riley, 987 F.2d 821 (D.C. Cir. 1993), the court imposed on CMS what it characterized as a shift in the burden of producing evidence on remand. Specifically, the court ordered that CMS would have to provide either “countervailing 11 evidence or a reason, not based on the insufficiency of the movant’s showing,” for rejecting Pomona’s affirmative case. 2020 WL 5816486, at *11 (cleaned up).
Pomona asked the district court for various forms of additional relief, including the imposition of an adverse inference establishing that Pomona’s determination of its Medicare fractions was correct. The court declined to impose an adverse inference. Instead, it stopped at setting aside the PRRB’s decision and remanding the case to the Board for further proceedings consistent with its opinion. 2020 WL 5816486, at *12.
III
We review the district court’s grant of summary judgment de novo. Forsyth Mem’l Hosp. v. Sebelius, 639 F.3d 534, 537 (D.C. Cir. 2011). Like that court, we must review the PRRB’s decision under familiar APA standards. 42 U.S.C. § 1395oo(f)(1). We thus must consider whether the Board’s decision was supported by “substantial evidence when the [administrative] record is viewed as a whole.” Id. § 1395oo(d); see also 5 U.S.C. § 706(2)(E).
Substantial evidence is “such relevant evidence as a reasonable mind might accept as adequate to support a conclusion.” NLRB v. Ingredion Inc., 930 F.3d 509, 514 (D.C. Cir. 2019) (cleaned up). In other words, “substantial evidence” is evidence that is “enough to justify, if the trial were to a jury, a refusal to direct a verdict when the conclusion sought to be drawn … is one of fact for the jury.” Ass’n of Data Processing Serv. Orgs., Inc. v. Bd. of Governors of the Fed. Reserve Sys., 745 F.2d 677, 684 (D.C. Cir. 1984) (Scalia, J.) (cleaned up); see also Sec’y of Labor v. Knight Hawk Coal, LLC, 991 F.3d 1297, 1308 (D.C. Cir. 2021) (“substantial evidence” turns on 12 whether a “reasonable factfinder” could have reached the agency’s conclusion) (cleaned up). Because the parties both take the position that CMS was required to use the “best available data” when calculating Pomona’s Medicare fraction, we assume without deciding that CMS was so obligated. See Dist. Hosp. Partners, L.P. v. Burwell, 786 F.3d 46, 56 (D.C. Cir. 2015).
Before the PRRB, Pomona bore the “burden of production” and the “burden of proof by establishing, by a preponderance of the evidence,” that it was “entitled to relief on the merits.” 42 C.F.R. § 405.1871(a)(3). Thus, to set aside the adverse Board decision on substantial-evidence review, Pomona must show that the evidence was so one-sided as to compel the Board to resolve the disputed factual issues in its favor—the administrative equivalent of a directed verdict for the plaintiff. That is a decidedly difficult standard, but we conclude that Pomona met it here.
IV
The only evidence before the Board was Pomona’s affirmative case that the CMS matching process missed thousands of Medicare patient days attributable to patients who qualified for SSI benefits. As summarized above, we think that case was substantial. The Board levelled three criticisms against it, which we address in turn.
First, the Board noted that the Medi-Cal data included patient days for patients receiving SSP but not SSI benefits. But as the Board itself acknowledged, Pomona “eliminated these SSP only days” based on statewide averages of the relative size of the SSI and SSP populations. J.A. 44. 13 Rosenstein testified that this reduction likely overcorrected for SSP-only beneficiaries, given the income demographics of the nearby area relative to the state as a whole. Rosenstein and Hefter—with collective expertise about Medi-Cal and CMS— both testified that the remaining gap could be explained only by some systematic error in the SSA/CMS matching process. And the Board gave no reason to question the adjustment for SSP-only patients based on statewide averages.
The Board also pointed to two specific populations who temporarily receive or retain Medi-Cal eligibility during a period when they are ineligible for SSI benefits. The first group involves long-term nursing-home residents. Individuals who live in nursing homes for more than three months may lose SSI benefits because payments made to the nursing-home count as income to the patient. Yet such patients temporarily retain their Medi-Cal eligibility while that program determines whether there is any independent ground for covering them. The second group involves first-time applicants for SSI and Medi-Cal benefits. A discrepancy could arise because Medi- Cal eligibility begins at the beginning of the month when the person applies, while SSI eligibility begins at the beginning of the following month. The Board marked down Pomona’s case because the hospital did not estimate the size of these two populations “from its own records.” J.A. 44–45.
The evidence currently in the administrative record suggests that these two groups, whatever their exact size, do not bridge the sizable gap between Pomona’s and CMS’s respective numbers—a difference of several thousand inpatient days attributable to SSI-eligible patients. Consider the timing difference for first-time applicants. To appear as false positives in Pomona’s calculation, such individuals would have to apply for both SSI and Medi-Cal benefits for the first time in the same month, receive fast Medi-Cal approvals, and be hospitalized in 14 that same month as Medicare patients. As Rosenstein explained, SSI enrollment is “fairly stable” in California, patients do not generally apply for SSI and Medi-Cal “for the first time at the same time,” and the impact of these patients on Pomona’s calculation thus “would be minimal.” J.A. 54. The same is true for the timing issue involving nursing-home residents, which covers only long-term residents hospitalized as Medicare patients in the interval between when their SSI and Medi-Cal codes change. Using statewide statistics and some back-of-the-envelope math, Pomona estimated that fewer than 10 such patients would likely show up in its SSI-fraction calculations in any given year. And neither the Board nor the Contractor countered these estimates. Given the lack of contrary evidence in the record, such discrepancies appear immaterial and suggest no substantial flaw in Pomona’s methodology.
Finally, the Board faulted Pomona for failing to provide what it described as a “crosswalk” between the relevant Medi- Cal and SSI codes, to show that Medi-Cal codes 10, 20, and 60 “identified only those individuals with an SSI code of C01, M01, or M02.” J.A. 46. But Pomona never suggested that the relevant state and federal codes bear this relationship. Instead, it explained that the relevant Medi-Cal codes establish eligibility for SSI or SSP benefits, with many SSP-only beneficiaries presumably coded by SSA as “N01” to reflect ineligibility for SSI benefits due to excess income. Moreover, Pomona provided expert testimony that the Medi-Cal codes have reliably served to establish SSI or SSP eligibility for decades and that they support commercial and governmental decisions made every day to provide and pay for medical care for needy Californians. Pomona then made a more-than-fair adjustment to eliminate Medicare patients who were SSP-only beneficiaries, which still left thousands of patient days attributable to SSI-eligible patients unaccounted for by CMS. 15 Pomona provided uncontroverted evidence that two potential difficulties with its approach amounted to little more than rounding errors. And it proffered creditable testimony from two experts indicating that the only explanation for the discrepancy was some error in CMS’s collection or matching of data. By contrast, the Contractor remained silent, failing to give even a hint as to why Pomona’s adjusted data might be overinclusive. Given the strength of the hospital’s showing, and the absence of any countervailing evidence, the Board’s conclusion that Pomona had failed to prove an undercount was unreasonable.
Atlanta College reinforces our conclusion. It involved schools expelled from a student-loan program on the ground that their default rates were too high. In administrative proceedings, the schools’ expert identified “hundreds of suspected errors” in the agency’s calculation of the rates. 987 F.2d at 825. The agency rejected the schools’ submission because it “did not control for other factors” that might affect the relevant calculations. Id. Yet that analysis “placed the schools in a ‘catch-22’ situation: the schools did not have the necessary material in their own records” to address the concerns identified by the agency. Id. at 826.
The district court in Atlanta College set aside the administrative expulsion order, and this Court affirmed. We expressed concern that the schools’ statutory right to appeal ineligibility determinations “would be meaningless” if the Secretary of Education “could reject for ‘hypothetical’ reasons a school’s showing of a mistaken calculation backed up by all the information to which the schools had access.” 987 F.2d at 830. In ruling for the schools, we stressed three critical features of the case. First, the schools had “gone about as far as they 16 can go” in presenting “specific allegations supported by all the information available to them.” Id. at 831 (cleaned up). Second, the schools’ submission was “sufficient, if undisputed, to require the Secretary to reverse his eligibility determination.” Id. Third, the Secretary had failed “to produce countervailing evidence or a reason, not based on the insufficiency of the school’s showing, that explains why the school’s allegations have not been accepted.” Id. (cleaned up).
Similar considerations are present here. Only CMS or SSA possess the SSI-eligibility data that would definitively establish the correct numerators for Pomona’s Medicare fractions. So, Pomona went about as far as it could, in attempting to reverse-engineer the SSI-eligibility data from publicly available data regarding SSI and SSP eligibility, plus publicly available data regarding the relative size of those two populations. Moreover, as shown above, Pomona’s affirmative case was sufficient, if undisputed, to require the PRRB to rule for it. And neither the Contractor nor the PRRB produced countervailing evidence or a reason for rejecting Pomona’s case other than its supposed factual insufficiency. To the contrary, the Contractor merely criticized Pomona for relying on “data from a secondary source,” when the primary data was available only to SSA or CMS. J.A. 276. And the PRRB, as explained above, merely speculated about unsubstantiated problems regarding use of the Medi-Cal eligibility data, despite Pomona’s showing that this data was generally reliable and that all known concerns with it could be reasonably accounted for (elimination of the SSP-only population) or were immaterial on this record (timing issues regarding nursing-home residents and first-time applicants). 17 V
In its cross-appeal, Pomona contends that the district court erred in refusing to impose on CMS an adverse inference based on CMS’s refusal to disclose relevant data in its SSI-eligibility file. Adverse inferences rest on the theory that if a party unjustifiably refuses to produce relevant evidence, the party likely views the evidence as unfavorable. See Int’l Union v. NLRB, 459 F.2d 1329, 1336 (D.C. Cir. 1972). We cannot draw such an inference here, where the disputed evidence involves sensitive details about the financial or other circumstances of individual patients and where CMS is prohibited by contract, if not by law, from disclosing the data. See 75 Fed. Reg. at 50,279; 70 Fed. Reg. at 47,440.
Pomona also invokes language from Atlanta College indicating a “shift” in the “burden of production.” 987 F.2d at 831 n.14. Pomona notes that if a burden of production in civil litigation shifts to the defendant, and if the defendant then fails to produce any evidence, the court must enter judgment for the plaintiff. See, e.g., Tex. Dep’t of Cmty. Affairs v. Burdine, 450 U.S. 248, 254 (1981). Pomona then reasons that because the Contractor failed to produce evidence to counter Pomona’s case before the PRRB, despite the opportunity and incentive to do so, the Contractor should not get a second bite at the apple on remand.
The short answer is that Atlanta College forecloses that argument. Like this case, it involved judicial review of agency action, rather than civil litigation conducted in a court. The schools had proven a case that was “sufficient, if undisputed,” to compel an administrative judgment in their favor. 987 F.2d at 831. And the Secretary of Education had failed to produce countervailing evidence to dispute it. See id. Yet we did not order the administrative adjudicator to accept the schools’ 18 position on remand. To the contrary, we affirmed a district court order that merely required the adjudicator to consider the schools’ position more carefully. See Atlanta Coll. of Med. & Dental Careers, Inc. v. Alexander, 792 F. Supp. 114, 123 (D.D.C. 1992). And in doing so, we made clear that the schools would be entitled to an administrative judgment if, but only if, the Secretary on remand failed “to produce countervailing evidence or a reason, not based on the insufficiency of the school’s showing, that explains why the school’s allegations have not been accepted.” 987 F.2d at 831.
The same course is appropriate here. Although Pomona’s case compels a ruling in its favor “if undisputed,” we do not foreclose the possibility that CMS may be able to dispute it successfully. And if CMS does introduce evidence to dispute it, the burden of proof will remain with Pomona. 42 C.F.R. § 405.1871(a)(3). All we hold today is that Pomona’s showing was robust enough to require some response from the agency.
VI
The district court correctly set aside the PRRB’s order and remanded to the Board for further proceedings without any adverse inference.
Affirmed.