Banner Health v. Price

Full title: BANNER HEALTH, F/B/O BANNER GOOD SAMARITAN MEDICAL CENTER, F/B/O NORTH…

Court: United States Court of Appeals FOR THE DISTRICT OF COLUMBIA CIRCUIT

Date published: Aug 18, 2017

867 F.3d 1323 (D.C. Cir. 2017)

Opinion

No. 16-5129

08-18-2017

BANNER HEALTH, f/b/o Banner Good Samaritan Medical Center, f/b/o North Colorado Medical Center, f/b/o McKee Medical Center, f/b/o Banner Thunderbird Medical Center, f/b/o Banner Mesa Medical Center, f/b/o Banner Desert Medical Center, f/b/o Banner Estrella Medical Center, f/b/o Banner Heart Hospital, f/b/o Banner Boswell Medical Center, f/b/o Banner Baywood Medical Center, et al., Appellants v. Thomas E. PRICE, Secretary, U.S. Department of Health and Human Services, Appellee

Sven C. Collins, Denver, CO, argued the cause for appellants. With him on the briefs was Stephen P. Nash, Denver, CO. Robert L. Roth, James F. Segroves, Washington, DC, and John R. Hellow, Los Angeles, CA, were on the brief for amici curiae Hospitals in support of appellants. Benjamin M. Shultz, Attorney, U.S. Department of Justice, argued the cause for appellee. With him on the brief was Michael S. Raab, Attorney.

Per Curiam

Sven C. Collins, Denver, CO, argued the cause for appellants. With him on the briefs was Stephen P. Nash, Denver, CO.

Robert L. Roth, James F. Segroves, Washington, DC, and John R. Hellow, Los Angeles, CA, were on the brief for amici curiae Hospitals in support of appellants.

Benjamin M. Shultz, Attorney, U.S. Department of Justice, argued the cause for appellee. With him on the brief was Michael S. Raab, Attorney.

Before: Rogers, Griffith and Srinivasan, Circuit Judges.

Per Curiam:

This appeal challenges the implementation by the Secretary of Health and Human Services ("HHS") of the Medicare outlier-payment program in the late 1990s and early 2000s. The program provides "supplemental" payments to hospitals to protect them from "bearing a disproportionate share of the [ ] atypical costs" associated with caring for "patients whose hospitalization would be extraordinarily costly or lengthy." Cty. of L.A. v. Shalala , 192 F.3d 1005, 1009 (D.C. Cir. 1999). A group of twenty-nine non-profit hospitals ("the Hospitals") principally contend that HHS violated the Administrative Procedure Act ("APA"), 5 U.S.C. §§ 551 et seq. , by failing to identify and appropriately respond to flaws in its methodology that enabled certain "turbo-charging" hospitals to manipulate the system and receive excessive payments at the expense of non-turbo-charging hospitals, including appellants.

The court addressed similar challenges in District Hospital Partners, L.P. v. Burwell , 786 F.3d 46 (D.C. Cir. 2015), and, to the extent the Hospitals repeat challenges decided in District Hospital Partners , that decision controls here. See LaShawn A. v. Barry , 87 F.3d 1389, 1395 (D.C. Cir. 1996). As to the Hospitals' other challenges, we affirm the district court's denials of their motions to supplement the record and to amend their complaint, and its decision that HHS acted reasonably in a manner consistent with the Medicare Act in fiscal years ("FYs") 1997 through 2003, and 2007. HHS, however, has inadequately explained aspects of the calculations for FYs 2004 through 2006, and we therefore reverse the grant of summary judgment in that regard and remand the case to the district court to remand to HHS for further proceedings.

I.

A.

Under the Medicare program, the federal government reimburses health care providers for medical services provided to the elderly and disabled. See Social Security Amendments of 1965 ("Medicare Act"), Pub. L. No. 89–97, tit. XVIII, 79 Stat. 286, 291 (1965). Initially, Medicare reimbursed hospitals for the "reasonable cost" of care provided. See 42 U.S.C. § 1395f(b)(1). This system, however, "bred ‘little incentive for hospitals to keep costs down’ because ‘the more they spent, the more they were reimbursed.’ " Cty. of L.A. , 192 F.3d at 1008 (quoting Tucson Med. Ctr. v. Sullivan , 947 F.2d 971, 974 (D.C. Cir. 1991) ) (brackets omitted). "To stem the program's escalating costs and perceived inefficiency," Congress revised Medicare's reimbursement system in 1983 to compensate hospitals prospectively at rates set before the start of each fiscal year. Id. Because the new system presented its own risk of under-compensating hospitals for the care of high-cost patients, Congress "authorized the Secretary [of HHS] to make supplemental ‘outlier payments.’ " Id. at 1009. Day outlier payments, which have since been phased out, were originally provided when a patient's length of stay exceeded a certain threshold. See 42 U.S.C. § 1395ww(d)(5)(A)(i), (v). Cost outlier payments are provided when a hospital's "charges, adjusted to cost" for a given patient exceed a certain "fixed dollar amount determined by [HHS]," after discounting any payments the hospital would normally receive. Id. § 1395ww(d)(5)(A)(ii). This appeal addresses cost outlier payments.

"[C]alculating [cost] outlier payments is an elaborate process," Dist. Hosp. Partners , 786 F.3d at 49, and some explication is necessary. First, by requiring that charges be "adjusted to cost" before determining whether a cost outlier payment is due, 42 U.S.C. § 1395ww(d)(5)(A)(ii), the Medicare Act "ensures that [HHS] does not simply reimburse a hospital for the charges reflected on a patient's invoice." Dist. Hosp. Partners , 786 F.3d at 50. HHS applies a "cost-to-charge ratio" "represent[ing] a hospital's ‘average markup’ " to a hospital's charges. Id. (quoting Appalachian Reg'l Healthcare, Inc. v. Shalala , 131 F.3d 1050, 1052 (D.C. Cir. 1997) ). "For example, if a hospital's cost-to-charge ratio is 75% (total costs are approximately 75% of total charges), [HHS] multiplies the hospital's charges by 75% to calculate the hospital's cost." Id.

Second, the "fixed dollar amount," 42 U.S.C. § 1395ww(d)(5)(A)(ii), commonly known as the "fixed[-]loss threshold," " ‘acts like an insurance deductible because the hospital is responsible for that portion of the treatment's excessive cost.’ " Dist. Hosp. Partners , 786 F.3d at 50 (quoting Boca Raton Cmty. Hosp. v. Tenet Health Care Corp. , 582 F.3d 1227, 1229 (11th Cir. 2009) ). The sum of the fixed-loss threshold and the standard payments a hospital would receive for a given treatment is known as the "outlier threshold." Id. "Any cost-adjusted charges imposed above the outlier threshold are eligible for reimbursement under the outlier payment provision," id. (citing 42 U.S.C. § 1395ww(d)(5)(A)(ii) ), although not at full cost, see 42 U.S.C. § 1395ww(d)(5)(A)(iii). For all years relevant to this appeal, "outlier payments have been 80% of the difference between a hospital's adjusted charges and the outlier threshold." Dist. Hosp. Partners , 786 F.3d at 50 ; see 42 C.F.R. § 412.84(j) (1997) ; 42 C.F.R. § 412.84(k) (2003).

Finally, in calculating the fixed-loss threshold, HHS must ensure that the total amount of outlier payments is not "less than 5 percent nor more than 6 percent" of total payments "projected or estimated to be made" under the inpatient prospective payment system that year. 42 U.S.C. § 1395ww(d)(5)(A)(iv). HHS "complies with this provision by selecting outlier thresholds that, ‘when tested against historical data, will likely produce aggregate outlier payments totaling between five and six percent of projected [non-outlier prospective] payments.’ " Dist. Hosp. Partners , 786 F.3d at 51 (quoting Cty. of L.A. , 192 F.3d at 1013 ). For all years relevant to this appeal, HHS has used 5.1% as its target percentage. See Banner Health v. Burwell , 126 F.Supp.3d 28, 43, 50 (D.D.C. 2015) (" Banner Health 2015 "). To account for the costs of the outlier-payment program, HHS also must reduce the standardized prospective payment rates for non-outlier payments by the same target percentage used to establish the fixed-loss threshold. 42 U.S.C. § 1395ww(d)(3)(B). In County of Los Angeles , 192 F.3d at 1017–20, the court held that HHS reasonably interpreted the Medicare Act not to require retroactive adjustments to the outlier threshold if total actual payments fell above or below the target percentage given the prospective nature of the system.

B.

Two sets of implementing regulations govern a hospital's qualification for outlier payments: (1) payment regulations determining when individual patient cases qualify for outlier payments, see 42 C.F.R. §§ 412.80 – 86 ; and (2) annual threshold regulations determining the fixed-loss threshold and other criteria used to define "outlier cases" for the upcoming fiscal year, see 42 C.F.R. § 412.80(c). The latter regulation sets the threshold based on the payment regulations and other factors.

During the early years of the outlier-payment program, HHS made a number of program-design decisions that are pertinent to this appeal. In the late 1980s, HHS revised the payment regulations at 42 C.F.R. § 412.84 to adopt hospital-specific cost-to-charge ratios in lieu of a national cost-to-charge ratio. See FY 1989 Final Rule, 53 Fed. Reg. 38,476, 38,503, 38,507 –09, 38,529 (Sept. 30, 1988). The purpose of this change was to "greatly enhance the accuracy with which outlier cases are identified and outlier payments are computed, since there is wide variation among hospitals in these cost-to-charge ratios." Id . at 38,503. HHS also provided that a hospital would default to an average statewide cost-to-charge ratio if its hospital-specific ratio fell outside reasonable parameters—three standard deviations above and below the statewide average—assuming that "ratios falling outside this range are unreasonable and are probably due to faulty data reporting or entry." Id. at 38,507 –08. Because "Medicare costs are generally overstated on the filed cost report and are subsequently reduced as a result of audit," HHS specified that cost-to-charge ratios were to be based on the "latest settled cost report" (that is, the latest audited cost report) and the associated charge data. Id. at 38,507. HHS acknowledged that this meant the data could be "as much as three years old," but nonetheless concluded that it was "the most accurate available data." Id .

In 1993, HHS decided to change how it adjusted its data for inflation in predicting future outlier payments. Until then, HHS had been inflating the prior year's charge data and then applying hospital cost-to-charge ratios to predict cost-adjusted charges for the upcoming fiscal year. FY 2004 Final Rule, 58 Fed. Reg. 46,270, 46,347 (Sept. 1, 1993). This is referred to as using a "charge[-]inflation factor." Id. Recognizing that charges were consistently increasing at a faster rate than costs and thus cost-to-charge ratios were declining, HHS switched to a "cost[ ]inflation factor," which it would apply after adjusting hospital charges by the cost-to-charge ratios. Id. HHS expected that this change would address a tendency in the model to "overestimat[e] outlier payments in setting the thresholds," that had been causing actual outlier payments to come in below the target percentage of total inpatient prospective payments. Id.

Over the next several years, actual outlier payments began to grow in relation to total inpatient prospective payments, see FY 1997 Final Rule, 61 Fed. Reg. 46,166, 46,229 (Aug. 30, 1996) ; FY 1998 Final Rule, 62 Fed. Reg. 45,966, 46,041 (Aug. 29, 1997), and exceeded the target percentage for the first time in FY 1997, see FY 1999 Final Rule, 63 Fed. Reg. 40,954, 41,009 (July 31, 1998). Observing that it was now consistently underestimating the thresholds necessary to hit the 5.1% target and attributing these miscalculations to the fact that charges were continuing to increase faster than costs, HHS switched back to a charge-inflation methodology beginning in FY 2003. See FY 2003 Final Rule, 67 Fed. Reg. 49,982, 50,123 –24 (Aug. 1, 2002). But, unbeknownst to HHS, the outlier-payment system had, in fact, "beg[u]n to break down in the late 1990s." Dist. Hosp. Partners , 786 F.3d at 51. As recounted in District Hospital Partners ,

[o]utlier payments were supposed to be made only in situations where the cost of care is extraordinarily high in relation to the average cost of treating comparable conditions or illnesses. But hospitals could manipulate the outlier regulations if their charges were not sufficiently comparable in magnitude to their costs. [HHS] issued a notice of proposed rulemaking (NPRM) [in February 2003] to address these concerns.

In the NPRM, [HHS] described how a hospital could use the time lag between the current charges on a submitted bill and the cost-to-charge ratio taken from the most recent settled cost report. A hospital knows that its cost-to-charge ratio is based on data submitted in past cost reports. If it dramatically increased charges between past cost reports and the patient costs for which reimbursement is sought, its cost-to-charge ratio would be too high and would overestimate the hospital's costs. Some hospitals took advantage of this weakness in the system. [HHS] identified 123 hospitals whose percentage of outlier payments relative to total [non-outlier prospective] payments increased by at least 5 percentage points between [FYs] 1999 and 2001. The adjusted charges at those 123 hospitals increased at a rate at or above the 95th percentile rate of charge increase for all hospitals over the same period. And during that time, the 123 hospitals had a mean rate of increase in charges of 70 percent alongside a decrease of only 2 percent in their cost-to-charge ratios. The 123 hospitals are referred to as turbo-chargers.

[HHS] published the final rule three months after the NPRM. As relevant here, [HHS] adopted two new provisions to close the gaps in the outlier payment system. First, a hospital's cost-to-charge ratio was to be calculated using more recent cost reports. This change reduced the time lag for updating cost-to-charge ratios by a year or more and ensured that those ratios accurately reflected a hospital's costs. Second, a hospital's outlier payments were to be subject to reconciliation when its cost report coinciding with the discharge is settled. Outlier payments were still disbursed based on the best information available at that time.

Id . (citing 2003 Outlier NPRM, 68 Fed. Reg. 10,420 (Mar. 5, 2003) and 2003 Outlier Final Rule, 68 Fed. Reg. 34,494 (June 9, 2003) ) (internal quotation marks and alterations omitted).

HHS opted not to change the then-in-effect FY 2003 threshold to reflect the new methodology. 2003 Outlier Final Rule, 68 Fed. Reg. at 34,506. In rejecting this option in the NPRM, HHS cited the "extreme uncertainty regarding the effects of aggressive hospital charging practices on FY 2003 outlier payments to date." 2003 Outlier NPRM, 68 Fed. Reg. at 10,427. Noting, however, that outlier payment data "for the first quarter of FY 2003" would "be available soon," the NPRM allowed that HHS might adjust the fixed-loss threshold at some point in the future. Id . In the Final Rule, however, HHS estimated an immediate adjustment would increase the FY 2003 threshold "by approximately $600," 2003 Outlier Final Rule, 68 Fed. Reg. at 34,505, and determined that any benefits in accuracy were outweighed by the potential that "for disruption and the fact that there was only a "limited amount of time remaining in the fiscal year," id. at 34,506. Although unknown to the public at the time, the decision not to adjust the FY 2003 threshold represented a departure from HHS's initial thinking as documented in a draft interim final rule ("draft IFR") submitted to the Office of Management and Budget ("OMB") about month before the NPRM was issued. The draft IFR, which otherwise was largely consistent with the changes ultimately adopted but would have been implemented earlier in the fiscal year, would have also immediately lowered the FY 2003 outlier threshold from $33,560 to $20,760.

HHS subsequently calculated annual fixed-loss thresholds in accordance with the new methodology, making only minor modifications to this approach along the way. See FY 2004 Final Rule, 68 Fed. Reg. 45,346, 45,476 –77 (Aug. 1, 2003); FY 2005 Final Rule, 69 Fed. Reg. 48,916, 49,276 –78 (Aug. 11, 2004); FY 2006 Final Rule, 70 Fed. Reg. 47,278, 47,493 –94 (Aug. 12, 2005); FY 2007 Final Rule, 71 Fed. Reg. 47,870, 48,148 –51 (Aug. 18, 2006). For FY 2004, HHS attempted to predict which hospitals would be subject to reconciliation and project their cost-to-charge ratios accordingly. See FY 2004 Final Rule, 68 Fed. Reg. at 45,477. HHS subsequently abandoned this approach, however, concluding that the majority of hospitals would not be subject to reconciliation and it was difficult to predict which would be in any given year. See FY 2005 Final Rule, 69 Fed. Reg. at 49,278. For FY 2007, HHS announced that it applied an adjustment factor to cost-to-charge ratios to account for the fact that hospitals' charges had consistently been growing faster than their costs, causing cost-to-charge ratios to consistently decline between when the threshold was estimated and when payments would actually be made. See FY 2007 Final Rule, 71 Fed. Reg. at 48,150. For further elaboration on the details of these rules, see Parts VI through IX.

C.

The Hospitals appealed their final outlier payment determinations between 1997 and 2007. See 42 U.S.C. § 1395oo (a). Challenging the validity of the governing regulations, they were granted expedited judicial review and filed a complaint in district court. See Banner Health , 797 F.Supp.2d 97, 103–04 (D.D.C. 2011) (" Banner Health 2011 "); 42 U.S.C. § 1395oo (f)(1). HHS moved to dismiss the complaint for lack of subject matter jurisdiction and for failure to state a claim for which relief can be granted. The district court ruled that the Hospitals "stated sufficient facts ... to support their standing" at that time to pursue claims under the Medicare Act, but dismissed claims brought under the Mandamus Act. Banner Health 2011 , 797 F.Supp.2d at 107. The district court otherwise declined to reach the merits of the Hospitals' remaining challenges in the absence of an administrative record and ordered the Hospitals to file a "notice of claims" identifying each "discrete agency action" being challenged. Id. at 118.

Over the next several years the district court pared down the Hospitals' claims, see Banner Health v. Sebelius , 905 F.Supp.2d 174, 182–87, 188 (D.D.C. 2012), and the parties engaged in discovery, see Banner Health v. Sebelius , 945 F.Supp.2d 1, 13–15, 17–39 (D.D.C. 2013). During this process, HHS advised the court that it had lost and was unable to recover multiple boxes of public comments submitted in connection with the FY 2004 rulemaking. Id. at 19. The district court concluded this loss was insufficient to defeat the "presumption of regularity" to be afforded to HHS's action, id. at 20, but granted the Hospitals' motion to supplement the record with certain materials, including the draft IFR that HHS prepared but abandoned in 2003, see id. at 27 ; see also id. at 33, 34, 36, 38. The Hospitals sought leave to amend and supplement their complaint to add claims under 5 U.S.C. § 553 regarding HHS's failure to disclose the draft IFR and its contents during the 2003 outlier rulemaking, but the district court denied this motion as futile. Banner Health v. Burwell , 55 F.Supp.3d 1, 7, 12 (D.D.C. 2014) (" Banner Health 2014 ").

In September 2014, the parties filed cross-motions for summary judgment. The Hospitals also filed a motion for judicial notice or, in the alternative, for extra record consideration of documents and other related relief. Banner Health 2015 , 126 F.Supp.3d at 36–37. The district court granted the latter motion insofar as the court would "take judicial notice of the publicly available materials subject to the motion, as relevant" id. at 37 ; see id. at 62, but otherwise denied the Hospitals' record-related requests, id. at 37 ; see id. at 60–64. On the merits, the district court remanded the FY 2004 fixed-loss threshold rule for HHS "to explain its decision regarding its treatment of certain data—or to recalculate the fixed[-]loss threshold if necessary[.]" Id. at 37 ; see id. at 96–99. The district court explained that it was bound to do so by this court's remand in District Hospital Partners . Id. at 98 (citing Dist. Hosp. Partners , 786 F.3d at 60 ). Otherwise, it rejected the Hospitals' challenges to the FY 2004 fixed-loss threshold. Id. at 99. The district court denied their challenges to the regulations in all other respects. See id . at 37, 67–96.

On remand, HHS elaborated the rationale for calculating the threshold in FY 2004 but made no substantive changes to the regulations. See Remand Explanation, 81 Fed. Reg. 3,727, 3,728 –29 (Jan. 22, 2016); see also Part VI, infra . The district court subsequently granted summary judgment to HHS, concluding that it had provided an adequate explanation for the decision not to exclude the 123 turbo-charging hospitals from the calculations used to set the FY 2004 fixed-loss threshold, and that the Hospitals had failed to identify any flaws in the Remand Explanation that undermined that conclusion or raised issues outside those remaining in the case. Banner Health v. Burwell , 174 F.Supp.3d 206, 208–09 (D.D.C. 2016) (" Banner Health 2016 "). The Hospitals appeal. The 186 hospitals in District Hospital Partners have filed an amicus brief, as the district court there failed to retain jurisdiction upon remand, urging that the Remand Explanation was inadequate and the HHS's failure to correct for all known turbo-chargers when setting the 2004 threshold resulted in the outlier payments to the 186 hospitals being too low.

II.

As a threshold matter, HHS contends that the Hospitals lack standing under Article III of the U.S. Constitution to challenge its failure between 1997 and 2003 to amend the outlier-payment regulations and threshold determinations in response to the turbo-charging phenomenon because any injury they may have suffered would not be redressed by their requested relief. Although questioning the Hospitals' standing on other grounds in the district court, see Banner Health 2015 , 126 F.Supp.3d at 64–67, HHS did not make this specific argument. Nonetheless, "because [it] goes to our jurisdiction, we must consider it." Shays v. FEC , 528 F.3d 914, 922–23 (D.C. Cir. 2008) ; see Cierco v. Mnuchin , 857 F.3d 407, 415–16 (D.C. Cir. 2017). We hold that the Hospitals have Article III standing to pursue their challenges.

To establish Article III standing, the plaintiff must have "suffered an injury in fact" that "is fairly traceable to the challenged action of the defendant" and it must be "likely, as opposed to merely speculative, that the injury will be redressed by a favorable decision." Friends of the Earth v. Laidlaw Envtl. Servs. , 528 U.S. 167, 180–81, 120 S.Ct. 693, 145 L.Ed.2d 610 (2000) (citing Lujan v. Defs. of Wildlife , 504 U.S. 555, 560–61, 112 S.Ct. 2130, 119 L.Ed.2d 351 (1992) ) (internal quotation marks omitted). For purposes of standing, this court is to "assume" that a plaintiff is "correct on the merits," Sierra Club v. EPA , 699 F.3d 530, 533 (D.C. Cir. 2012), and that the court will grant the relief sought, West v. Lynch , 845 F.3d 1228, 1235 (D.C. Cir. 2017) (citing Fla. Aud u bon Soc'y v. Bentsen , 94 F.3d 658, 663–64 (D.C. Cir. 1996) (en banc)). A plaintiff lacks standing, however, if they fail to show that they would benefit under their alternate methodology. See, e.g. , Franklin v. Massachusetts , 505 U.S. 788, 802, 112 S.Ct. 2767, 120 L.Ed.2d 636 (1992) ; Nat'l Law Ctr. on Homelessness & Poverty v. Kantor , 91 F.3d 178, 183 (D.C. Cir. 1996).

HHS maintains that any injury suffered by the Hospitals is not redressable because they would have received the same amount or less in outlier payments had HHS taken the action they propose. HHS has misconstrued the Hospitals' challenge, suggesting that they only seek changes to the cost-to-charge ratios and not to the thresholds, and that even if it were to revise the thresholds, they would have gone up rather than down. Under the Hospitals' theory, however, HHS violated the APA by failing to recognize and respond to turbo-charging in a timely manner. Were HHS to revise the outlier payment and threshold regulations to avoid making "unlawful turbo-charged payments," the Hospitals maintain, such payments could no longer factor into the threshold calculations and the thresholds would be lower. Reply Br. 3 (emphasis added). That the Medicare Act does not require HHS to recalculate thresholds retroactively when actual outlier payments fall above or below the targeted percentage, see Cty. of L.A. , 192 F.3d at 1017–20, does not also mean HHS would not reevaluate its threshold when the premises underlying his or her predictions have been successfully challenged.

III.

The Hospitals challenge a number of the district court's procedural rulings. We conclude that none of these challenges have merit for the following reasons.

A.

The Hospitals first contend the court abused its discretion in refusing to consider as evidence, or, in the alternative, as adjudicatory facts, materials referenced in their summary judgment motion, and in denying their motion to supplement the record of the FY 2004 rulemaking. The court reviews such evidentiary and docket management decisions for abuse of discretion, see Am. Wildlands v. Kempthorne , 530 F.3d 991, 1002 (D.C. Cir. 2008) ; Jackson v. Finnegan, Henderson, Farabow, Garrett & Dunner , 101 F.3d 145, 150, 151 (D.C. Cir. 1996), and finds none.

The district court struck three tables depicting data from the administrative record appended to the hospitals' motion for summary judgment. The court had set a "generous" seventy-page limit for motions, and "caution[ed] the parties that any attempt to subvert the[ ] page limits by including additional briefing in appendices will be rejected, and such appendices will be stricken from the record." Sched. & Proc. Order at 4, No. 10-1638 (July 17, 2014). The Hospitals contend that the tables should not count towards the page limits because they "faithfully reproduced record data." Appellants' Br. 92. The tables, however, compile data from various disparate sources and present it in a simplified manner meant to persuade. In enforcing its warning against briefing through appendices, "the district court exercised its prerogative to manage its docket, and its discretion to determine how best to accomplish this goal." Jackson , 101 F.3d at 151.

The district court also refused to consider certain items as extra-record evidence: (1) congressional testimony by an HHS official, Thomas Scully ("Scully Testimony"); and (2) two briefs submitted by the government in another outlier-related case ("Boca Briefs"). "It is well understood in administrative law that the ‘focal point for judicial review should be the administrative record already in existence, not some new record completed initially in the reviewing court.’ " Tripoli Rocketry Ass'n v. Bureau of Alcohol, Tobacco, Firearms, & Explosives , 437 F.3d 75, 83 (D.C. Cir. 2006) (quoting Envtl. Def. Fund v. Costle , 657 F.2d 275, 284 (D.C. Cir. 1981) ). "Exceptions to that rule are quite narrow and rarely invoked[,] ... primarily limited to cases where the procedural validity of the agency's action remains in serious question, or the agency affirmatively excluded relevant evidence." CTS Corp. v. EPA , 759 F.3d 52, 64 (D.C. Cir. 2014) (internal quotation marks and citations omitted); see Am. Wildlands , 530 F.3d at 1002.

In District Hospital Partners , 786 F.3d at 56, the court affirmed the exclusion of the Scully Testimony, concluding that it did not fall within any of the established exceptions to the rule limiting review to the existing administrative record. To the extent that the Hospitals present a different challenge here, it is no more persuasive. They contend that the testimony would show that HHS changed its approach to turbo-charging in 2003 due to opposition from OMB, and that HHS, in Scully's view, " ‘did not understand why’ it ‘kept missing and missing’ its targets and ‘really never understood the dynamics' of its model." Appellants' Br. 94 (citing Scully Testimony at 4). The already-voluminous record, however, does not "say so little" as to " ‘frustrate judicial review,’ " Dist. Hosp. Partners , 786 F.3d at 56 (quoting Am. Wildlands , 530 F.3d at 1002 ), and the testimony does not constitute "background information" necessary "to determine whether the agency considered all of the relevant factors," Am. Wildlands , 530 F.3d at 1002 (quoting James Madison Ltd. by Hecht v. Ludwig , 82 F.3d 1085, 1095 (D.C. Cir. 1996) ). The Hospitals offer no reason why OMB's involvement in the decision to use traditional notice-and-comment rulemaking is pertinent to any of their challenges, and HHS acknowledges that it found it "surprising" that hospitals were able to manipulate the outlier payment regulations through turbo-charging. Appellee's Br. 12.

The Hospitals' position regarding the Boca Briefs is not any more compelling. The government's position in that litigation as to the proper construction of the outlier provisions in the Medicare Act neither directly contradicts, nor sheds light on, the challenged actions. Cf. Nat'l Res. Def. Council v. EPA , 755 F.3d 1010, 1020–21 (D.C. Cir. 2014). That one of the briefs identified more than one hundred turbo-charging hospitals by name is also irrelevant for effective judicial review. To the extent the Hospitals contend, in the alternative, that these materials are appropriate for judicial notice as publicly available materials, the district court did "take judicial notice of these documents as necessary in resolving this matter[.]" Banner Health 2015 , 126 F.Supp.3d at 62.

The district court also denied the hospitals' motion to supplement the record with a comment letter from the Federation of American Hospitals responding to the FY 2004 rulemaking. The letter was made part of the record in the District Hospital Partners litigation. See Dist. Hosp. Partners, LP v. Sebelius , 971 F.Supp.2d 15, 26–28 (D.D.C. 2013)aff'd in part and rev'd in part sub nom. Dist. Hosp. Partners , 786 F.3d at 56. Here, the district court concluded that the letter was too prejudicial given the late hour at which the Hospitals sought its admission. The record shows that the Hospitals had ample time to act before the day briefs were to be filed and that their delay denied HHS the opportunity to treat the comment as part of the administrative record in preparing its motion for summary judgment. The district court "acted within the range of permissible alternatives that were available to it" in denying the motion. Jackson , 101 F.3d at 150.

B.

The Hospitals further contend that the district court erred in denying their motion for leave to amend their complaint to allege that HHS had violated 5 U.S.C. § 553 by failing to disclose data, analysis, and conclusions in the 2003 draft IFR that were adverse to the determinations made in subsequent rulemakings. The district court ruled that the proposed amendment was futile because the Hospitals had failed to show that HHS relied on the draft IFR and its supporting materials in the challenged regulations. Banner Health 2014 , 55 F.Supp.3d at 11–12. Our review is de novo . See In re APA Assessment Fee Litigation , 766 F.3d 39, 55 (D.C. Cir. 2014).

"Under APA notice and comment requirements, ‘among the information that must be revealed for public evaluation are the technical studies and data upon which the agency relies in its rulemaking.’ " Am. Radio Relay League v. FCC , 524 F.3d 227, 236 (D.C. Cir. 2008) (citing Chamber of Commerce v. SEC , 443 F.3d 890, 899 (D.C. Cir. 2006) ) (internal quotation marks and alterations omitted); see Portland Cement Ass'n v. Ruckelshaus , 486 F.2d 375, 393–94 (D.C. Cir. 1973). This "allow[s] for useful criticism," including by enabling commenters "to point out where ... information is erroneous or where the agency may be drawing improper conclusions[.]" Am. Radio Relay League , 524 F.3d at 236 (internal quotation mark omitted); see Chamber of Commerce , 443 F.3d at 900 ; Sierra Club v. Costle , 657 F.2d 298, 398 n.484 (D.C. Cir. 1981).

HHS maintains that the Hospitals' motion to amend is futile, citing National Mining Association v. Mine Safety and Health Administration , 599 F.3d 662, 671 (D.C. Cir. 2010), because the new allegation challenges HHS's decision not to make a midyear adjustment to the FY 2003 threshold. HHS failed to make this argument in the district court, and it is forfeit. See Am. Wildlands , 530 F.3d at 1001. To the extent HHS maintains that a Section 553 claim based on American Radio Relay League conflicts with the APA and Vermont Yankee Nuclear Power Corp. v. Natural Resources Defense Council , 435 U.S. 519, 98 S.Ct. 1197, 55 L.Ed.2d 460 (1978), the court rejected this argument in American Radio Relay League , 524 F.3d at 239–40, stating that it "is not imposing new procedures but enforcing the agency's procedural choice by ensuring that it conforms to APA requirements," id . at 239. The availability of a Section 553 claim remains the law of the circuit. See Allina Health Servs. v. Sebelius , 746 F.3d 1102, 1110 (D.C. Cir. 2014) ; LaShawn A. , 87 F.3d at 1395.

The Hospitals' nonetheless falter in their attempt to bring their challenge under the ambit of American Radio Relay League . There, this court held that the Federal Communications Commission must release unredacted versions of the technical studies and data on which it relied in promulgating a rule. Am. Radio Relay League , 524 F.3d at 240. The court explained the redactions "may contain contrary evidence, inconvenient qualifications, or relevant explanations of the methodology employed" that speak to the weight that should be given the unredacted portions of a study. Id. at 239. The Hospitals seek to equate HHS's "cherry-pick[ing]" portions of the draft IFR with the redaction of a technical study. Appellants' Br. 97. But the court has never suggested that an unpublished draft rule constitutes a "study" for purposes of this doctrine, see Am. Radio Relay League , 524 F.3d at 239, nor does there appear to be reason to do so here. The packaging of facts and conclusions in the draft IFR did not "inextricably b[i]nd" them together in the same manner as a study, which is to be considered as a whole. Id. The draft IFR merely represented one way for HHS to respond to available information. That the published rules referenced some of the same information and analysis as the draft IFR does not mean that HHS relied on the draft IFR, but rather on some of the same underlying material, such as analyses indicating vulnerabilities in the original outlier methodology and findings regarding the 123 hospitals that had been receiving disproportionately high outlier payments. Although HHS's decision to disregard certain information may be challenged as arbitrary and capricious, Section 553 does not require disclosure of materials that were considered and rejected in the course of a rulemaking. See id. at 240 (quoting 1 RICHARD J. PIERCE, JR., ADMINISTRATIVE LAW TREATISE 437 (4th ed. 2002)).

IV.

Having rejected the Hospitals' procedural challenges, we now turn to their challenges to various aspects of the outlier rules for every fiscal year between 1997 and 2007, as well as a midyear rule promulgated during FY 2003. The district court rejected each of those challenges. We address the Hospitals' challenges in chronological order, affirming the district court's grant of summary judgment except with regard to aspects of the FY 2004, 2005, and 2006 rules. We begin here with the challenges to the rules governing FYs 1997 through 2003.

The Hospitals acknowledge that HHS was not aware of turbo-charging before October 2002, when a stock-market analyst wrote an editorial exposing turbo-charging. Their primary challenge to the fixed-loss thresholds for FYs 1997 through 2003 is that the failure to discover and stop turbo-charging was arbitrary and capricious. At the very least, the Hospitals argue, HHS committed an error when calculating the fixed-loss threshold for each of those fiscal years. We reject both claims.

A.

As a preliminary matter, HHS argues that the Supreme Court's decision in Auer v. Robbins , 519 U.S. 452, 117 S.Ct. 905, 137 L.Ed.2d 79 (1997), bars the Hospitals' claim that HHS's failure to uncover turbo-charging was arbitrary and capricious. In Auer , the Secretary of Labor declined to consider whether the agency should amend one of its regulations in response to a Supreme Court decision. The Auer Court held that the agency could not have acted arbitrarily and capriciously in failing to amend its regulation, because nobody had asked the agency to do so. See id. at 459, 117 S.Ct. 905. The Court thus found that it had "no basis" on which to question the Secretary's failure to act. Id. HHS argues that because the Hospitals never asked it to amend its regulations to address turbo-charging, we similarly have no basis to question its failure to do so.

HHS is mistaken. The petitioners in Auer believed a public and well-known Supreme Court decision should have led the agency, on its own, to change one of its regulations. By contrast, the Hospitals here argue that HHS was in possession of non-public information that—in combination with public data—uniquely positioned it to uncover turbo-charging, such that its failure to do so prior to October 2002 was arbitrary and capricious. If the Hospitals are correct, we cannot fault them for failing to petition HHS to address a problem that only it could have known about. We do not read Auer , which says nothing about an agency's obligations when it has access to important information that commenters do not, to foreclose such a claim.

B.

The Hospitals look first to the information that was publicly available, suggesting that the rapid rise of the fixed-loss threshold, coupled with steadily decreasing hospital-level cost-to-charge ratios, should have tipped HHS off to turbo-charging's existence. It is true that those developments show that hospital charges were rising more rapidly than were costs. But intentional charge manipulation is neither the most obvious, nor necessarily the most probable, explanation for what hindsight reveals was actually widespread turbo-charging. The healthcare market is notoriously complex, and a host of other factors—e.g., the changing demographics of the Medicare-insured population, or the introduction of new medical technology and medications—could, as the Hospitals acknowledge, see Reply Br. 37–39, have been responsible for the phenomenon.

But the Hospitals urge that various comments made during past rulemakings also should have tipped HHS off that there was a distinct possibility that turbo-charging was the real culprit. In 1988, for example, HHS adopted a number of the features of the outlier-payment system that enabled turbo-charging. One commenter warned that the rule created "an incentive for hospitals to increase their charges and to manipulate their charge structures" in order to receive lower cost-to-charge ratios. FY 1989 Final Rule, 53 Fed. Reg. at 38,509. And in 1994, another commenter "expressed concern over the use of statewide averages" for hospitals with cost-to-charge ratios three standard deviations below the statewide mean, because that practice "created a clear incentive for hospitals to artificially inflate their gross charges, and circumvent the intent that hospitals only be paid marginal costs for outliers." FY 1995 Final Rule, 59 Fed. Reg. 45,330, 45,407 –08 (Sept. 1, 1994). These comments may have put HHS on notice that the outlier-payment system created incentives, in theory, for hospitals to manipulate their charges, but both comments predated the era of widespread turbo-charging, and neither suggested the practice was anything other than a mere possibility.

In our view, those comments fall far short of demonstrating that HHS should have discovered that the skyrocketing fixed-loss thresholds in subsequent years were actually caused by turbo-charging. Critically, even though the annual increases in the fixed-loss threshold and declines in hospital-specific cost-to-charge ratios were publicly available, the Hospitals can point to no contemporaneous comment that even hinted the underlying cause of those trends was willful charge manipulation. In fact, no commenter asked HHS to investigate whether willful charge manipulation might be to blame. Not until FY 2003 did a commenter even urge HHS to "[r]eevaluate assumptions about cost and charge increases and other factors that influence the outlier projections," noting in particular the lack of up-to-date cost information. Comment Responding to FY 2003 NPRM, 67 Fed. Reg. 31,404 (May 9, 2002) (available at J.A. 638). But although in retrospect HHS realized that the data lag helped to enable turbo-charging, this type of generalized comment was insufficient to put HHS on notice of that specific problem.

The success of the Hospitals' claim thus turns on whether the combination of public and non-public information in HHS's possession put it in a position to discover an illegal practice that had evaded detection by the rest of the industry. The Hospitals point to non-public data showing that, in some fiscal years, a select few hospitals received up to twice as much in outlier payments as they did in non-outlier inpatient payments. No doubt, those figures raise red flags with regard to those particular hospitals. Under the Medicare Act, outlier payments should account only for 5–6% of the average hospital's Medicare-related payments. See 42 U.S.C. § 1395ww(d)(5)(A)(iv). Had HHS adequately overseen the outlier-payment system, the Hospitals argue, it would have known to attribute those aberrational outlier payments to turbo-charging.

Again, we disagree. It is far from obvious that this sporadic and anomalous data should have put HHS on the lookout for a widespread and systematic scheme to defraud Medicare through turbo-charging. Indeed, we fail to see how the mere presence of such aberrational data in HHS's possession should have alerted it to turbo-charging. The charge figures for an individual hospital are but a speck in the vast reams of data that HHS maintains—the offending hospitals constituted only around 2% of all hospitals participating in the Medicare program. Not to mention, despite seeing the same trend in terms of rising thresholds, it did not occur to any regulated party to ask HHS to pore over hospital-level data in search of evidence of willful charge manipulation, as the Hospitals now claim HHS should have done.

With the benefit of 20/20 hindsight, the Hospitals have been able to identify suspicious charge data for individual healthcare providers. But they have failed to convince us that it was arbitrary and capricious for HHS not to have found the cause, in real time, before turbo-charging was brought to its attention in October 2002.

C.

The Hospitals raise another challenge to HHS's fixed-loss thresholds for FYs 1997 through 2003. When the projected cost-to-charge ratio for a hospital in a given year was three standard deviations above or below the statewide average, HHS used that statewide-average figure in its calculations. See Part I.B, supra . The Hospitals argue that the statewide-averages HHS used were outdated and, in fact, higher than the correct averages. If the Hospitals are right, that means HHS was overestimating the outlier payments it would make to those hospitals, which in turn would have led it to set the fixed-loss threshold too high.

However, the Hospitals fail to show that HHS actually made this mistake. Their claim relies solely on the inclusion of year-old statewide averages in certain data files, known as "impact files," containing records of hospital costs and charges that HHS uses when calculating the annual threshold. Because some outdated data was discovered in select impact files included in the administrative record, the Hospitals insist that HHS in fact calculated the annual thresholds based on the wrong statewide averages for every fiscal year between 1997 and 2003.

HHS does not deny that at least some of the impact files in its possession contain outdated data. Rather, HHS argues that although there may have been some out-of-date data in some of the impact files it kept, that data was not used in calculating the fixed-loss threshold. HHS claims that it used files containing the correct, up-to-date statewide averages and backs up this assertion by noting that it reported the up-to-date statewide averages in the same annual rulemakings in which it set the fixed-loss threshold for every fiscal year between 1997 and 2003. There is no dispute that HHS had the correct data in its possession, and we believe the most sensible inference is that HHS used that data. Indeed, for the Hospitals to be correct, it would have to be the case that for seven consecutive years, HHS publicly announced the current (correct) statewide averages, but nonetheless employed outdated statewide averages in its calculations. That strikes us as unlikely, to say the least. It is true that in the many intervening years, HHS has either lost or misplaced the relevant records that could definitively settle which figures it actually employed. But the Hospitals have not provided an adequate basis to overcome the "presumption of regularity" that agency proceedings enjoy. San Miguel Hosp. Corp. v. NLRB , 697 F.3d 1181, 1186–87 (D.C. Cir. 2012).

V.

The Hospitals next argue that HHS's decision not to lower the FY 2003 fixed-loss threshold midyear violated the agency's obligations under both the Medicare Act and the APA. We disagree.

A.

Recall that at the beginning of each fiscal year, HHS sets a fixed-loss threshold that it believes will lead outlier payments to equal 5.1% of non-outlier inpatient payments for that year. See Part I.A, supra . According to the Hospitals, in calculating what threshold will allow HHS to hit its target, HHS must count only the outlier payments it expects to make to non -turbo-charging hospitals. Because HHS made its calculations for FY 2003 using payments it expected to make to all hospitals that year, it necessarily set a threshold that was higher than it would have had it excluded from its model the payments it expected to make to turbo-charging hospitals. Of course, HHS was unaware of turbo-charging at the time it set the FY 2003 threshold. But in the Hospitals' view, as soon as HHS became aware that some of the payments it had been making were going to turbo-chargers, it was statutorily obligated to go back and lower the threshold—in the middle of the fiscal year. Only by lowering the threshold midyear, the Hospitals argue, could HHS actually hit its target.

The Hospitals' argument fails. Most fundamentally, the court held in County of Los Angeles , 192 F.3d at 1019, that HHS is not obligated to make later adjustments in order to hit its target percentage so long as it acts reasonably in setting the fixed-loss threshold at the beginning of each fiscal year. The court explained that this holding stems from the prospective nature of the outlier-payment program, which allows for efficient administration. Indeed, it would be unduly burdensome for HHS to "establish outlier thresholds in advance of each fiscal year, and process millions of bills based on those figures," only to require it to later "recalibrate those calculations, reevaluate anew each of the millions of inpatient discharges under the revised figures, and disburse a second round of payments." Id. (citation omitted). In addition, the prospective nature of the system allows for "certainty and predictability of payment for not only hospitals but the federal government." Id. That certainty and predictability would disappear if the fixed-loss threshold were subject to midyear, or end-of-year, course correction.

The Hospitals also argue that, regardless of County of Los Angeles , HHS opened the door to a challenge simply by responding to comments that asked it to lower the threshold and explaining its decision not to make a midyear change. Not so. As here, when an "agency merely responds to ... unsolicited comment[s] by reaffirming its prior position, that response does not" open the agency's position up to a challenge. Kennecott Utah Copper Corp. v. U.S. Dep't of Interior , 88 F.3d 1191, 1213 (D.C. Cir. 1996). Moreover, an agency does not "reopen an issue by responding to a comment that addresses a settled aspect of some matter, even if the agency had solicited comments on unsettled aspects of the same matter." Id. The FY 2003 threshold was a "settled aspect" of the matter at the time HHS issued the 2003 Outlier NPRM, meaning that there was no "reopening" here. In any event, HHS's explanation was well-reasoned, resting in significant part on its assessment that "[c]hanging the threshold for the remaining few months of the fiscal year could disrupt hospitals' budgeting plans and would be contrary to the overall prospectivity" of the outlier-payment program. 2003 Outlier Final Rule, 68 Fed. Reg. at 34,506. Those were the very concerns that drove our decision in County of Los Angeles .

Moreover, the Hospitals are mistaken that HHS could not lawfully factor in outlier payments to turbo-charging hospitals when determining whether it was likely to hit its 5.1% target. It is true that the Medicare Act provides that a hospital may request outlier payments only when its "charges, adjusted to cost," exceed the outlier threshold. 42 U.S.C. § 1395ww(d)(5)(A)(ii). It is also the case that hospitals' outlier payments are supposed to "approximate the marginal cost of care beyond the [outlier threshold]." Id. § 1395ww(d)(5)(A)(iii).

Although HHS has argued in other litigation that turbo-charging hospitals acted improperly in manipulating their charging practices in order to receive extensive outlier payments, see Boca Raton Cmty. Hosp. v. Tenet Healthcare Corp. , 9:05-cv80183-PAS, ECF No. 49 (S.D. Fla. May 17, 2005), it does not follow that HHS unlawfully issued outlier payments to turbo-chargers. HHS cannot ascertain a hospital's true costs in treating an outlier case until long after it makes the outlier payment. HHS estimates a hospital's charges, adjusted to cost, and marginal cost of care beyond the threshold, using cost-to-charge ration data in its possession at the time of payment. It would be inconsistent with the prospectivity of the outlier-payment program to require HHS to "reevaluate anew each of the millions of" outlier payments it made during a fiscal year because the cost-to-charge ratios it had been employing turned out, due to no fault of its own, to have been too high. Cty. of L.A. , 192 F.3d at 1019. That is the case even if those cost-to-charge ratios were too high as a result of the unlawful efforts of the turbo-charging hospitals. We therefore reject the Hospitals' argument that HHS acted contrary to the Medicare Act by declining to alter the FY 2003 fixed-loss threshold.

B.

In the alternative, the Hospitals argue that HHS's decision not to lower the threshold midyear was nonetheless arbitrary and capricious. The Hospitals advance three distinct concerns with the HHS's course of action. None is persuasive.

First, they complain that HHS should have explained why it did not exclude the 123 turbo-charging hospitals that year from its target calculations. This challenge merely recasts the Hospitals' argument that HHS violated the Medicare Act as a challenge under the APA. Suffice it to say, there was nothing amiss in accounting for outlier payments the Medicare Act permitted it to consider.

The Hospitals also claim that HHS erred in not explaining why it rejected the draft IFR's conclusion that an immediate reduction in the FY 2003 threshold was warranted. But to the extent the Hospitals argue that HHS departed from its prior policy, that claim is squarely foreclosed by our decision in District Hospital Partners . As we explained in that case, the draft IFR "was never ‘on the books' in the first place." Dist. Hosp. Partners , 786 F.3d at 58 (quoting FCC v. Fox Television Stations , 556 U.S. 502, 515, 129 S.Ct. 1800, 173 L.Ed.2d 738 (2009) ). HHS therefore need not explain any departure from what it said in the draft IFR. See id. And to the extent the Hospitals argue that the draft IFR contained obvious alternatives to what HHS did in its final rule, they ignore that HHS issued its final rule in June, four months after the IFR. As HHS explained, circumstances in June differed markedly from those in February: the end of the fiscal year was fast approaching and changing the fixed-loss threshold at that time could have resulted in considerable disruption to hospitals' budgets. 2003 Outlier Final Rule, 68 Fed. Reg. at 34,505. Whatever HHS's obligation was to explain its decision to "depart" from the draft IFR, it met that obligation here.

Finally, the Hospitals take issue with a sentence in the final rule that states that HHS "inflated charges from the FY 2002 Medicare Provider Analysis and Review (MedPAR) file by the 2-year average annual rate of change in charges per case to predict charges for FY 2004." Id. (MedPAR is a database that aggregates the claims submitted by hospitals to HHS.) Both parties agree that HHS should have instead been predicting charges for FY 2003. HHS, noting that this sentence appeared in a section of the rule concerned exclusively with the 2003 fiscal year, assures us that it in fact predicted charges for FY 2003 and that the reference to FY 2004 was, as the district court found, "self-evidently" a typo. Banner Health 2015 , 126 F.Supp.3d at 95 ; see also Appellee's Br. 56–57. Upon reviewing the context in which the statement was made, and considering that the Hospitals provide no reason to doubt HHS's assurances, we accept this explanation.

VI.

The Hospitals next take issue with several aspects of the FY 2004 Final Rule. We first describe the details of this Rule, as well as the 2016 Remand Explanation issued by HHS to elaborate on its decision-making, and then turn to the merits of the Hospitals' challenges to the adequacy of both documents. We agree with the district court that HHS sufficiently justified its decision to anticipate reconciling only 50 turbo-charging hospitals, and therefore to utilize projection cost-to-charge ratios for only that smaller subset of turbo-chargers. See Banner Health 2015 , 126 F.Supp.3d at 99 ; Banner Health 2016 , 174 F.Supp.3d at 208. But we hold that it inadequately explained its failure to exclude turbo-chargers from its calculation of the annual rate of charge inflation.

A.

As previously explained, HHS's June 2003 rulemaking was designed to cure most of the ills that had plagued the outlier-payment system during the turbo-charging era. HHS expected that its three key reforms would substantially diminish, if not wholly eradicate, the practice of turbo-charging. See Part I.B, supra. But that set of reforms introduced a fresh problem: when it came time to set the FY 2004 threshold in August 2003, no outlier payments had been made under the new system. HHS would therefore have to estimate the outlier payments to be made under a framework it had never actually implemented.

For FY 2004, as for all other relevant years, HHS sought a formula that would generate a fixed-loss threshold under which outlier payments would amount to 5.1% of non-outlier inpatient payments. See 42 U.S.C. § 1395ww(d)(5)(A)(iv) ; FY 2004 Final Rule, 68 Fed. Reg. at 45,478. Its chosen methodology ultimately yielded a FY 2004 fixed-loss threshold of $31,000, a modest decrease from the previous year's threshold of $33,560. See id. at 45,477.

HHS's FY 2004 rule divided hospitals into two groups: those it expected would ultimately be subjected to reconciliation, and those it expected would not. See id. at 45,476 –77. HHS estimated its anticipated outlier payments to the latter group of hospitals using a multi-step formula. It first identified the charges billed by each hospital in the FY 2002 MedPAR files. To predict each hospital's charges during the 2004 fiscal year, HHS multiplied the 2002 MedPAR data by 1.268, the two-year average annual rate of change in charges per case from FY 2000 to FY 2002. Lastly, HHS multiplied the product of that calculation by each hospital's cost-to-charge ratio from "the most recent cost reporting year." Id . at 45,476.

HHS also identified "approximately 50" hospitals that it anticipated subjecting to reconciliation "[b]ased on [its] analysis of hospitals that ha[d] been consistently overpaid recently for outliers." Id. For that set of hospitals, HHS calculated projection cost-to-charge ratios that it believed would more accurately capture HHS's net outlier payouts during the 2004 fiscal year, factoring in the amounts it expected to recoup through reconciliation. See id. at 45,477. The Hospitals do not challenge the methodology underlying those adjustments.

The Hospitals do, however, challenge several other components of the methodology HHS employed in setting the FY 2004 outlier threshold. This court has already adjudicated a related challenge to the FY 2004 rulemaking. In District Hospital Partners , 786 F.3d at 60, we held that HHS had inadequately justified the methodology it employed in setting the FY 2004 fixed-loss threshold. Although HHS's 2003 Outlier NPRM had identified 123 hospitals as turbo-chargers, the FY 2004 Final Rule "did not explain how the 50 hospitals [that HHS anticipated subjecting to reconciliation] differed from the 123 ... identified in the [2003] NPRM." Id. at 58. Using projection cost-to-charge ratios for the previously identified 123 turbo-chargers was thus "a significant and obvious alternative" that HHS was required to consider. Id. at 59.

On remand, HHS was tasked with explaining (i) "why [it] corrected for only 50 turbo-charging hospitals ... rather than for the 123 [it] had identified in the [2003] NPRM," and (ii) "what additional measures (if any) were taken to account for the distorting effect that turbo-charging hospitals had on the dataset for the 2004 rulemaking." Id. at 60. To the extent it decided to recalculate the FY 2004 threshold on remand, HHS was further directed to decide "what effect (if any)" this would have on the FY 2005 and 2006 thresholds. Id. The district court in this litigation, taking note of the remand order in District Hospital Partners , likewise remanded the FY 2004 rule "to provide the agency an opportunity to explain further why it did not exclude the 123 identified turbo-charging hospitals from the charge inflation calculation for FY 2004—or to recalculate the fixed[-]loss threshold if necessary." Banner Health 2015 , 126 F.Supp.3d at 98.

In a document issued in early 2016, HHS sought to address the District Hospital Partners court's concerns with the methodological choices it made in setting the FY 2004 threshold. See Remand Explanation, 81 Fed. Reg. at 3,727–29. This Remand Explanation aimed to provide clarification along three axes.

First, HHS addressed why it did not exclude any of the 123 hospitals previously identified as turbo-chargers from the MedPAR files it used to calculate the two-year average annual rate of change in charges per case from FY 2000 to FY 2002. See id. at 3,729. In other words, given that charging practices during the 2000, 2001, and 2002 fiscal years were the product of an age of artificial excess, HHS was tasked with justifying its assumption that the model would accurately predict charge inflation during a fiscal year governed by rules "expected ... to curb turbo[-]charging." Id. The Remand Explanation suggested that HHS believed past charge increases would reliably simulate future growth, because "[t]he outlier final rule was in effect for only part of the interval that our charge inflation estimate was intended to reflect."Id. HHS also reasoned that, because the 123 turbo-chargers could claim outlier payments in FY 2004, "excluding them ... would have introduced a different form of distortion into our simulations, by causing the[m] to disregard the impact of those hospitals." Id.

Second, the Remand Explanation sought to shore up HHS's explanation for why only approximately 50 turbo-chargers were selected as likely candidates for reconciliation. See id. According to HHS, "reconciliation generally would be performed only if a hospital met the criteria we had specified": "[a] 10-percentage point change in the hospital's [cost-to-charge ratio] from the time the claim was paid compared to the [cost-to-charge ratio] at cost report settlement; and receipt of total outlier payments exceeding $500,000." Id. at 3,728–29. HHS claimed to have "identified approximately 50 hospitals that [it] determined likely to meet these criteria in FY 2004," and adjusted those hospitals' projection cost-to-charge ratios accordingly. Id. at 3,729. Not all 123 previously identified turbo-chargers were selected for this treatment, because HHS "did not expect that all of the 123 hospitals discussed in the March 2003 proposed rule would be likely to meet the criteria for reconciliation." Id.

Third, HHS explained why it believed that its decision to project cost-to-charge ratios for only 50 turbo-charging hospitals had no "distorting effect" on the threshold calculation. Id. at 3,727; see id. at 3,728. HHS, in its midyear 2003 rule, switched from using the most recent settled cost report for a hospital when calculating its cost-to-charge ratio to using the hospital's most recent tentatively settled cost report. See 2003 Outlier Final Rule, 68 Fed. Reg. at 34,502. HHS thus urged that its "payment simulations employed cost-to-charge ratios calculated from very recent data ... and did not employ cost-to-charge ratios drawn from older historical data." Remand Explanation, 81 Fed. Reg. at 3,728. That adjustment "reduc[ed] any reason for concern that cost-to-charge ratios drawn from older historical data ... would not reliably approximate the cost-to-charge ratios that would be used to pay FY 2004 claims." Id . HHS saw no reason to adjust projection cost-to-charge ratios for the remaining turbo-chargers, because it "anticipated that implementation of the June 2003 outlier final rule would curb ... turbo[-]charging practices." Id .

B.

In its FY 2004 Final Rule, HHS assumed that the rate at which charges had inflated between FYs 2000 and 2002 would accurately approximate charge growth during a period that included an entire year (FY 2004) in which the anti-turbo-charging reforms would be in effect. The Hospitals contend that HHS's Remand Explanation failed to offer a reasoned basis for making that assumption. We agree.

When HHS set its FY 2004 threshold on August 1, 2003, it was well aware that scores of hospitals in recent years had "inappropriately maximiz[ed] their outlier payments" and "caused the threshold to increase dramatically." 2003 Outlier Final Rule, 68 Fed. Reg. at 34,496. That exploitative practice prompted HHS to reassess its entire framework for making outlier payments. HHS believed that its 2003 reforms, once in place, would "greatly reduce the opportunity for hospitals to manipulate the system to maximize outlier payments." Id. at 34,503. It declared that it was "essential to eliminate those effects as soon as possible," id. at 34,497, in order to avoid underpaying "hospitals that ha[d] already been harmed by the inappropriate redistribution" of outlier payments, id. at 34,499.

Yet its decision to project future charges using turbo-charging-infected data foreseeably "allow[ed] the effects of this inappropriate redistribution ... to continue into the future." Id. at 34,497. In setting the FY 2004 threshold, HHS sought to predict hospitals' charging behavior for a full year after the structural causes of turbo-charging had been eradicated. HHS nonetheless multiplied the 2002 MedPAR charge data by the rate at which charges had inflated between FYs 2000 and 2002, without removing charge data attributable to the 123 hospitals it had identified as turbo-chargers. See FY 2004 Final Rule, 68 Fed. Reg. at 45,476. The 2004 Rule gave no indication as to why HHS thought it rational to use turbo-charging-infused data to forecast charges for a year in which hospitals' "opportunity ... to manipulate the system" had been "greatly reduce[d]." 2003 Outlier Final Rule, 68 Fed. Reg. at 34,503.

The two reasons HHS provided in its Remand Explanation likewise fail to explain why it believed "past charge growth would still be a satisfactory basis for estimating more recent charge growth." 81 Fed. Reg. at 3,729. The first was that "[t]he outlier final rule was in effect for only part of the interval that our charge inflation estimate was intended to reflect." Id. But even if it may have been reasonable to expect turbo-charging-era levels of charge inflation to persist during FY 2003, HHS repeatedly intimated that it expected charging practices to return to normal in FY 2004. And even if HHS could have reasonably predicted "future significant charge growth due to other reasons," Appellee's Br. 61, it never identified what those reasons might be, or why they would likely lead to turbo-charging-era levels of charge inflation during FY 2004.

The Remand Explanation's second reason for including turbo-chargers' data is equally unsatisfying. HHS submits that, because "[t]he 123 hospitals were not excluded from claiming outlier payments" during FY 2004, "excluding them from our simulations would have introduced a different form of distortion into our simulations, by causing the simulations to disregard the impact of those hospitals." Remand Explanation, 81 Fed. Reg. at 3,729. That alleged symmetry is illusory. There is no necessary linkage between, on one hand, the hospitals whose data are used to calculate the two-year average annual rate of change in charges per case, and, on the other hand, the hospitals that are eligible to receive outlier payments during an upcoming fiscal year. Nothing would have precluded HHS from calculating the industry-average rate of charge inflation after removing a series of warped data points, while still accounting for the reality that all hospitals remained eligible to collect outlier payments during the upcoming fiscal year.

It was entirely predictable that including turbo-charged data would lead to a charge-inflation projection that greatly exceeded the actual rate of charge inflation during FY 2004, as in fact actually happened. FY 2006 Final Rule, 70 Fed. Reg. at 47,494. HHS thus overlooked an important consideration in attempting to "ensure that [its] simulated FY 2004 payments would match up as closely as possible with how FY 2004 claims would actually be paid." Remand Explanation, 81 Fed. Reg. at 3,728. As a result, we hold that HHS acted arbitrarily and capriciously in failing to exclude charge data for the 123 historical turbo-chargers from its FY 2004 charge-inflation calculation.

C.

The Hospitals next contend that HHS, in setting the FY 2004 fixed-loss threshold, failed to offer a reasoned explanation for adjusting the projection cost-to-charge ratios of only 50 turbo-chargers in order to account for the possibility of reconciliation. Although the issue is a close one, we affirm the adequacy of HHS's explanation.

HHS has repeatedly reiterated its belief that the June 2003 outlier rule would "curb the turbo[-]charging practices that had caused rapid increases in charges." Id. at 3,729. Yet HHS acknowledged that the time lag between issuing outlier payments and tentatively settling hospitals' cost reports left a large enough window for a hospital's true cost-to-charge ratio to vary from the cost-to-charge ratio used to make outlier payments during a fiscal year. See id . at 3,728. In its FY 2004 Final Rule, HHS announced that fiscal intermediaries would reconcile outlier payments if, once the intermediary had settled a hospital's FY 2004 cost report, it found that the hospital's actual cost-to-charge ratio during the period was "substantially different" from that used to make outlier payments to it during the fiscal year. 68 Fed. Reg. at 45,476. HHS explained that it expected to subject 50 hospitals' payments to reconciliation, "[b]ased on [its] analysis of hospitals that ha[d] been consistently overpaid recently for outliers." Id.

As far as the District Hospital Partners court could tell, however, all 123 previously identified turbo-chargers—not just the 50 hospitals selected as candidates for reconciliation—had been drastically "overpaid recently." 786 F.3d at 58 (quoting FY 2004 Final Rule, 68 Fed. Reg. at 45,476 ). According to HHS itself, all of the turbo-charging hospitals had initiated "dramatic increases in charges." 2003 Outlier NPRM, 68 Fed. Reg. at 10,424. The court therefore remanded for HHS to explain why it selected only 50 turbo-chargers as likely candidates for reconciliation, and thus for adjustment of their projection cost-to-charge ratios. District Hospital Partners , 786 F.3d at 59.

On remand, HHS claimed it had identified a past turbo-charging hospital as a candidate for reconciliation if it expected that the hospital would satisfy the following two conditions for FY 2004: (i) "[a] 10-percentage point change in the hospital's [cost-to-charge ratio] from the time the claim was paid compared to the [cost-to-charge ratio] at cost report settlement;" and (ii) "receipt of total outlier payments exceeding $500,000 during the cost reporting period." Remand Explanation, 81 Fed. Reg. at 3,728 –29. Though HHS acknowledged "it was difficult to project which hospitals would be subject to reconciliation of their outlier payments using then-available data," it anticipated that only approximately 50 turbo-charging hospitals would end up satisfying those criteria. Id. at 3,728.

The Hospitals object to the adequacy of the Remand Explanation. They contend that HHS's contemporaneous statements belie any articulation of a rule-like formula for determining which hospitals' projection cost-to-charge ratios to adjust. Although the Remand Explanation categorized the above two conditions as "criteria we had specified for reconciliation," id. , the Hospitals argue that the June 2003 outlier rule introducing those conditions offered them only as tentative guidelines for fiscal intermediaries to consider when deciding whether to engage in reconciliation. In the Hospitals' view, HHS cannot have actually employed those criteria to identify the 50 reconciliation candidates.

In response, HHS points to the language of the 2003 Outlier Final Rule. According to the Remand Explanation, that Rule "instructed our contractors to put a hospital through outlier reconciliation" if it complied with both purported requirements. Id. As a claim about the Outlier Rule's instructive force, that statement is problematic. The Rule did recite the criteria quoted above, but it prefaced them by saying that "we are considering instructing fiscal intermediaries to conduct reconciliation [in that manner]." 2003 Outlier Final Rule, 68 Fed. Reg. at 34,503 (emphasis added). In the same document, HHS explicitly disclaimed any pretense of finality in the proposed formula: "We intend to issue program instructions to the fiscal intermediaries that will provide specific criteria for identifying those hospitals subject to reconciliation ... for FY 2004." Id. at 34,504. There is no record evidence that HHS followed up on that intention in any way.

Making matters worse, the text of the FY 2004 Final Rule recites a reconciliation policy that, while not technically inconsistent with the two-part formula identified by the Remand Explanation, is expressed in a significantly less rule-like fashion. In short, the FY 2004 Final Rule explained that reconciliation would occur when "a hospital's actual ... cost-to-charge ratios are found to be substantially different from the cost-to-charge ratios used during that time period to make outlier payments." 68 Fed. Reg. at 45,476. If the Remand Explanation accurately captures the 2004 Rule's methodology, the formula's absence from the Rule itself is a surprising omission.

Certain statements in later rules' descriptions of the reconciliation process also tend to undermine the Remand Explanation's account of definite reconciliation criteria. The FY 2005 Final Rule, for example, stated that reconciliation would occur when the cost-to-charge ratios from hospitals' final settled cost reports "are different than" those appearing on tentatively settled cost reports. 69 Fed. Reg. at 49,278. And as late as the FY 2006 Final Rule, HHS suggested that hospitals qualified for reconciliation if their cost-to-charge ratios "fluctuate[d] significantly," but noted that it still "plan[ned] on issuing instructions to fiscal intermediaries" to "detail the specifics of reconciling outlier payments." 70 Fed. Reg. at 47,495.

Still, enough record evidence exists to sustain the Remand Explanation's description of how HHS settled upon expecting to reconcile only approximately 50 turbo-charging hospitals. We find HHS's contemporaneous characterization of two comments on its proposed FY 2004 Rule to be particularly significant. Per HHS, those comments expressed concern with "the criterion in the final rule on outliers that specifically addressed our policy on reconciliation (that if a hospital's cost-to-charge ratio changed by 10 or more percentage points, a hospital would be subject to reconciliation)." FY 2004 Final Rule, 68 Fed. Reg. at 45,477. HHS is thus on record for FY 2004 as characterizing one component of the Remand Explanation as "our policy." The Rule also acknowledged commenters' request that HHS "modify the trigger for outlier reconciliation by promulgating a scale of cost-to-charge ratios rather than a constant amount." Id. (emphasis added). We doubt that HHS would have implicitly ratified that description of its methodology unless reconciliation were to be presumptively "trigger[ed]" upon the satisfaction of "constant" conditions.

HHS's FY 2005 Final Rule spoke in similar terms. It observed that "the majority of hospitals' cost-to-charge ratios will not fluctuate significantly enough ... to meet the criteria to trigger reconciliation of their outlier payments." 69 Fed. Reg. at 49,278 (emphasis added). That phrasing immediately followed an assertion that "[r]econciliation occurs when hospitals' cost-to-charge ratios at the time of cost report settlement are different than the tentatively settled cost-to-charge ratios used to make outlier payments." Id. (emphasis added). So HHS evidently saw no incompatibility between using broad, general language to describe its approach to reconciliation, and nonetheless using predetermined "criteria" to "trigger" the recoupment of excess outlier payments.

To be sure, the Remand Explanation does not dovetail seamlessly with HHS's contemporaneous statements. But for the reasons just given, we find a sufficient basis to conclude that HHS has explained, and justified, the approach it took in predicting which hospitals would be subject to reconciliation upon settlement of FY 2004 cost reports. The Hospitals' further criticism that HHS failed to produce the exact data underlying its identification of the 50 hospitals is similarly unavailing. As we have previously noted, agency proceedings enjoy a "presumption of regularity." San Miguel Hosp. Corp. , 697 F.3d at 1186–87. In these limited circumstances, in which we are satisfied that the agency employed a reasonable methodology in a rulemaking that concluded well over a decade ago, we will not remand simply because HHS is unable to reproduce the exact data and calculations in question.

D.

The Hospitals lodge one further arbitrary-and-capricious challenge to the FY 2004 Final Rule. They contend that it was irrational for HHS to "appl[y] a charge[-]inflation factor" when predicting hospital charges for the 2004 fiscal year "without adjusting [hospitals' cost-to-charge ratios]," as well. Appellants' Br. 49 (emphasis omitted). Notwithstanding the considerable deference agencies typically receive when analyzing data and projecting trends, see District Hospital Partners , 786 F.3d at 60, we agree with the Hospitals.

Following the discovery of turbo-charging, HHS amended its regulations to provide that the "cost-to-charge ratios applied at the time a claim is processed are based on either the most recent settled cost report or the most recent tentative[ly] settled cost report, whichever is ... latest." 42 C.F.R. § 412.84(i)(2). That means that when a new tentatively settled cost report is released for a hospital at some point during each fiscal year, HHS begins applying an updated cost-to-charge ratio for the hospital calculated from that report rather than the projection cost-to-charge ratio it used when generating the fixed-loss threshold. As HHS concedes, therefore, projection cost-to-charge ratios "apply for [only] part of the fiscal year." Appellee's Br. 59.

Of course, HHS's "ability to identify true outlier cases is dependent on the accuracy of the cost-to-charge ratios." 2003 Outlier Final Rule, 68 Fed. Reg. at 34,501. In its Remand Explanation, HHS asserted that basing projection cost-to-charge ratios on the most recent tentatively settled cost reports "reduced any reason for concern" that those cost-to-charge ratios "would not reliably approximate the cost-to-charge ratios that would be used to pay FY 2004 [outlier] claims." 81 Fed. Reg. at 3,729. HHS thus found no reason to adjust its projection cost-to-charge ratios to account for known or foreseeable trends in hospital charge or cost levels (or to explain its decision not to do so). HHS is mistaken: once it decided to use a charge-inflation methodology, it ostensibly needed to adjust its projection cost-to-charge ratios downward, as well.

Recall that prior to FY 2003, HHS had employed a cost-inflation methodology to predict the outlier payments it would make in an upcoming fiscal year. See Part I.B, supra . Because an outlier payment aims to approximate a hospital's costs (not its charges), "the relevant variable" for outlier-payment purposes "is [a hospital's] estimated ‘costs' for a given case." FY 1997 NPRM, 61 Fed. Reg. 27,444, 27,497 (May 31, 1996). HHS's avowed basis for switching to a charge-inflation methodology in FY 2003 was that "charges ha [d] been growing at a much faster rate than recent estimates of cost growth," leading it to consistently underestimate the outlier payments it would end up making in a given fiscal year. FY 2003 Final Rule, 67 Fed. Reg. at 50,124.

Charging practices inform threshold calculations in another way, as well: they supply the denominator of projection cost-to-charge ratios. But, despite HHS's awareness that, "[o]ver time, cost-to-charge ratios will reflect the differential increase in charges" relative to costs, id. , HHS made no effort to account for this differential with respect to its projection cost-to-charge ratios. When HHS set the FY 2004 threshold in August 2003, it knew it would end up using updated cost-to-charge ratios to make outlier payments as tentatively settled cost reports came in for each hospital. Yet it did not account for the fact that those updated cost-to-charge ratios were likely to be considerably lower than its projection cost-to-charge ratios, thereby leading many hospitals to be underpaid for outlier cases.

We fail to understand why HHS expected charges to inflate more rapidly relative to costs for some purposes but not for others. The act of using a charge-inflation methodology represented HHS's considered view, irrespective of what it might have otherwise been required to predict, that charges would continue to rise more quickly than costs.

For those reasons, we conclude that HHS's approach was "internally inconsistent and inadequately explained." District Hospital Partners , 786 F.3d at 59 (quoting Gen. Chem. Corp. v. United States , 817 F.2d 844, 846 (D.C. Cir. 1987) ). HHS's broad discretion in this area "is not a license to ... treat like cases differently." Cty. of L.A. , 192 F.3d at 1023 (quoting Airmark Corp. v. FAA , 758 F.2d 685, 691 (D.C. Cir. 1985) ) (alteration in original). We hold that HHS acted arbitrarily and capriciously in failing to adequately explain why it did not adjust its projection cost-to-charge ratios downward. VII.

The Hospitals also raise a number of challenges to the FY 2005 Final Rule. We agree with the district court in rejecting the first two of the Hospitals' challenges, but again conclude, as with FY 2004, that HHS failed to adequately explain its decision not to adjust its projection cost-to-charge ratios downward.

A.

For FY 2005, HHS employed substantially the same methodology for calculating the outlier threshold that it had used in FY 2004, with some notable modifications. See FY 2005 Final Rule, 69 Fed. Reg. at 49,277 –78. Under its revised formula, HHS first identified the charges listed in its FY 2003 MedPAR files. To inflate the FY 2003 charges to FY 2005, HHS multiplied the 2003 MedPAR data by 1.1876, the average rate of change in charges per case from the first half of FY 2003 to the first half of FY 2004, compounded over two years. It then multiplied the product of that calculation by the cost-to-charge ratios generated from hospitals' most recent tentatively settled cost reports to estimate each hospital's costs for FY 2005. Using those estimates, HHS determined that a threshold of $25,800—down from FY 2004's figure of $31,000—would yield outlier payments totaling 5.1% of all non-outlier inpatient payments. See id. at 49,278.

The dataset HHS used to generate this charge-inflation figure relied in part on a half year's worth of turbo-charged data. HHS itself acknowledged the "exceptionally high rate of hospital charge inflation that is reflected in the data" for the 2003 fiscal year, and admitted that its regime-spanning projection would be imperfect. Id. at 49,277. Yet HHS indicated that it "strongly prefer[red] to employ actual data rather than projections" in calculating the outlier threshold, and believed it "optimal to employ comparable periods" in determining the annual rate of change. Id .

HHS again defended its continued practice of inflating past years' MedPAR data by the rate of charge inflation (rather than cost inflation) by noting that "the basic tendency of charges to increase faster than costs is still evident." Id. As with FY 2004, HHS also declined to adjust its historically derived projection cost-to-charge ratios to account for the likelihood that charges would continue to inflate more quickly than costs. That was because HHS had "already taken into account the most significant factor in the decline in cost-to-charge ratios," by using "the most recent tentatively settled cost report[s]." Id. And lastly, HHS omitted any effects of reconciliation from its FY 2005 threshold calculation. It did so principally because it predicted that "the majority of hospitals' cost-to-charge ratios will not fluctuate significantly enough" to trigger reconciliation. Id. at 49,278.

In District Hospital Partners , we rejected an arbitrary-and-capricious challenge to HHS's decision not to "eliminate the turbo-charging hospitals from [the 2003 MedPAR] dataset" used in conjunction with FY 2004 MedPAR data to determine the charge-inflation figure employed to set the FY 2005 threshold. 786 F.3d at 61. The court concluded that it would have made "little sense to remove turbo-charging hospitals from th[e] half of the dataset" corresponding to FY 2003 "without making similar adjustments to the other half of the dataset" corresponding to FY 2004. Id. But there was "no need to modify the 2004 data because that information was collected" after HHS's anti-turbo-charging measures had taken effect. Id. According to District Hospital Partners , HHS "sensibly opted for a simpler approach that did not entail piling projections atop projections": it "reasonably left both halves unaltered." Id. HHS's methodology was necessarily imperfect, to be sure. But "imperfection alone does not amount to arbitrary decision-making." Id.

Notwithstanding our decision in District Hospital Partners , the Hospitals challenge the FY 2005 threshold as arbitrary and capricious in three separate respects. The district court rejected each argument. It followed District Hospital Partners in concluding that HHS could permissibly inflate FY 2003 charges forward two years using charging data partially stemming from the turbo-charging period. Banner Health 2015 , 126 F.Supp.3d at 100. The court also detected no problem with HHS's decision not to account for reconciliation, especially since the outlier rule had "obviated much of the need for [it]." Id. at 101. The court further held that HHS had given a "cogent explanation" for opting not to project a downward trend for its cost-to-charge ratios. Id. at 100.

B.

The Hospitals first repeat an argument they made with respect to FY 2004—that HHS acted arbitrarily in including turbo-charged data in its FY 2005 charge-inflation formula. That contention is squarely foreclosed by District Hospital Partners .

The Hospitals seek to escape the holding of District Hospital Partners by contending that the court in that case labored under a factual misimpression. The court stated that HHS "derived the charge[-] inflation factor [employed in the FY 2005 rulemaking] from the cost-to-charge ratios for individual hospitals"—in other words, that the two came from the same dataset. Dist. Hosp. Partners , 786 F.3d at 57 n.6. The Hospitals are correct in questioning that account of HHS's methodology. As we explained above, HHS calculated the charge-inflation factor—the average rate of change in charges per case—solely from MedPAR data. Although HHS then multiplied that charge-inflation factor by hospitals' cost-to-charge ratios, it did not derive the former from the latter.

Even so, the Hospitals have not identified any way in which that apparent misapprehension affected District Hospital Partner 's analysis. Nothing in District Hospital Partner 's logic turned on which dataset HHS employed to generate its charge-inflation factor, and the minor factual inaccuracy identified by the Hospitals does not strip that decision of its controlling force.

We further note that there is no inconsistency between our FY 2004 and FY 2005 holdings pertaining to the inclusion of turbo-chargers' data in HHS's charge-inflation calculation. The MedPAR data used in calculating the FY 2004 threshold entirely predated HHS's anti-turbo-charging reforms. In contrast, the District Hospital Partners court's FY 2005 analysis hinged on the fact that fully half of the charge-inflation dataset "was not infected by turbo-charging" because it "came after the effective date of the outlier correction rule." Id. at 61. Excluding turbo-chargers from only the first half of the dataset would have prevented HHS from "employ[ing] comparable periods in determining the rate of change from one year to the next." Id. (quoting FY 2005 Final Rule, 69 Fed. Reg. at 49,277 ) (internal quotation marks omitted). Alternatively, excluding turbo-chargers from both halves of HHS's charge-inflation dataset would have required HHS to ignore undisputedly valid charge-inflation data from FY 2004 for more than one hundred hospitals. We therefore reject the Hospitals' first arbitrary-and-capricious challenge to the FY 2005 Final Rule.

C.

The Hospitals also challenge HHS's decision not to account for future reconciliation in setting the FY 2005 threshold. We find that HHS had no obligation to adjust any hospitals' projection cost-to-charge ratios in anticipation of recouping overpayments following the 2005 fiscal year.

The Hospitals contend that, even though the three key anti-turbo-charging reforms made in the June 2003 outlier rule had been in effect for well over a year, it was arbitrary and capricious for HHS not to forecast that particular hospitals would continue collecting outlier payments significantly higher than their actual costs. That contention is unsupportable. The parties agree that the 2003 Outlier Final Rule "greatly reduce[d] the opportunity for hospitals to manipulate the system to maximize outlier payments." 68 Fed. Reg. at 34,501. The District Hospital Partners court credited the rule with having "corrected the flaw in the outlier payment system that created the opportunity-and incentive-to turbo-charge." 786 F.3d at 61. It also observed that, once the outlier rule was implemented, "the specter of turbo-charging was nil." Id. at 62. Those statements cannot be squared with a putative obligation to forecast a program of reconciliation for FY 2005 outlier payments, even if HHS's contractors ultimately flagged a number of payments as reconcilable. HHS therefore did not act arbitrarily in predicting that FY 2005 charging practices "w[ould] not fluctuate significantly enough" to justify accounting for reconciliation in setting the FY 2005 threshold. FY 2005 Final Rule, 69 Fed. Reg. at 49,278.

D.

As with the FY 2004 rule, the Hospitals again object to HHS's decision not to adjust hospitals' projection cost-to-charge ratios downward in FY 2005. For the reasons given above, see Part VI.D, supra , we agree that HHS was obligated to explain why it employed projection cost-to-charge ratios that did not reflect its prediction that charges would increase more quickly than costs in FY 2005.

VIII.

The Hospitals next take issue with the FY 2006 Final Rule. We agree with the Hospitals with respect to one of their challenges, but not the other, for reasons previously explained.

A.

To calculate the FY 2006 fixed-loss threshold, HHS used essentially the same methodology as in FY 2005, though with the benefit of more current data. That formula generated a FY 2006 fixed-loss threshold of $23,600, which was $2,200 less than the year before. See FY 2006 Final Rule, 70 Fed. Reg. at 47,494.

In the rule, HHS acknowledged commenters' complaints that it paid out less than its 5.1% target in FYs 2004 and 2005. See id. But it noted that FY 2006 would be the first year in which all three projection variables—MedPAR data, cost-to-charge ratios, and charge-inflation data—would entirely postdate the June 2003 anti-turbo-charging reforms. See id. at 47,494 –95. HHS again declined to adjust its projection cost-to-charge ratios. Id . at 47,495. Lastly, HHS declared that the prospect of reconciliation was so remote as not to be worth considering in setting the FY 2006 threshold. Because of the June 2003 outlier rule, HHS believed that "cost-to-charge ratios will no longer fluctuate significantly and ... few hospitals, if any, will actually have [their] ratios reconciled upon cost report settlement." Id.

In District Hospital Partners , this court rejected an arbitrary-and-capricious challenge to the FY 2006 Final Rule, finding it to be "plainly reasonable." 786 F.3d at 62. That holding, though, was tethered to the fact that "there was no need to account for turbo-chargers" when inflating charges, because "all of the charge data for the 2006 rule was collected with the outlier correction rule in effect." Id. at 62–63. District Hospital Partners did not foreclose all possible challenges to the FY 2006 threshold.

The Hospitals challenge the FY 2006 Final Rule as arbitrary and capricious on two familiar grounds. The district court rejected both arguments. It first concluded that the 2006 Rule "adequately justifie[d] the agency's decision not to adjust" its projection cost-to-charge ratios, especially since the projection cost-to-charge ratios were "actually used, for some portion of the fiscal year, to calculate outlier payments." Banner Health 2015 , 126 F.Supp.3d at 103. And the court deemed "certainly adequate" HHS's "thorough explanation" for why it chose not to account for potential reconciliation in calculating the FY 2006 threshold. Id. at 104.

B.

We affirm the district court on the latter ground, but not the former. For the same reasons explained above, HHS need not have accounted for the possibility of reconciling excess payments, because its anti-turbo-charging measures had by then been in effect for years. See Part VII.C, supra. And as with FYs 2004 and 2005, we again hold that HHS acted arbitrarily and capriciously in failing to explain why it assumed that charges would increase faster than costs throughout FY 2006 for some purposes, but not for others. See Part VI.D, supra .

IX.

Finally, the Hospitals challenge two aspects of the FY 2007 Final Rule. The Hospitals' primary objection targets the methodology HHS employed to adjust its projection cost-to-charge ratios downward when calculating the fixed-loss threshold. We describe this methodology in further detail below, and then address the merits of the Hospitals' challenges, finding both challenges lacking.

A.

In FY 2007, HHS finally attempted to account for the effect of declining cost-to-charge ratios on its efforts to make total outlier payments in a fiscal year equal 5.1% of non-outlier inpatient payments. See FY 2007 Final Rule, 71 Fed. Reg. at 48,150 –51. Because hospital charges during this period consistently increased faster than costs, most hospitals saw their cost-to-charge ratios drop each time HHS's relevant database, known as the "Provider Specific File," was updated with a new tentatively settled cost report for the hospital. Once those lower, updated cost-to-charge ratios became available, fiscal intermediaries immediately began using them "to calculate the outlier payments" hospitals would receive. Id. at 48,150. That differential between projection cost-to-charge ratios and payment cost-to-charge ratios contributed to HHS setting the fixed-loss threshold too high to hit its 5.1% target in FYs 2004, 2005, and 2006. To remedy that problem in FY 2007, HHS decided to apply an "adjustment factor" to the cost-to-charge ratios in the Provider Specific File when predicting the outlier payments it anticipated making during the upcoming fiscal year. Id.

HHS's adjustment factor represented the following quotient: the rate at which it predicted hospital costs in the Provider Specific File would increase as a result of updated cost reports coming in midyear, divided by the rate at which it predicted hospital charges in the file would increase as a result of those updates. See id. Because of the considerable time lag between a hospital's rendering a service and HHS's learning the hospital's costs for the service, however, the adjustment factor reflected costs and charges for medical services already rendered. Each time the Provider Specific File was updated with a hospital's most recent tentatively settled cost report, HHS would match that cost data with the charges billed by the hospital for the corresponding medical services. Because "it takes approximately 9 months for fiscal intermediaries to tentatively settle a cost report from the fiscal year end of a hospital's cost reporting period," much of the data in the Provider Specific File at the time of the FY 2007 rulemaking dated back to FYs 2004 and 2005. Id.

As previously noted, HHS learns a hospital's charges for a given service soon after it is performed. HHS therefore possessed fairly reliable information, when generating its adjustment factor, about charge inflation in the years leading up to the FY 2007 outlier rule. To estimate how significantly charges in the Provider Specific File would rise as a result of updates during the 2007 fiscal year, HHS looked to the "average annualized rate-of-change in charges-per-case" between the first two quarters of FY 2005 and the first two quarters of FY 2006. Id. That basic methodology appeared elsewhere in the FY 2007 Final Rule, see id. , and it also mirrored the approach HHS had employed for the past several years to predict hospital charges in an upcoming fiscal year. The resulting figure then served as the denominator of the adjustment factor.

With regard to costs, by contrast, HHS does not know a hospital's true costs for a given service at the time it is rendered. At the time it set the fixed-loss threshold, therefore, HHS did not possess actual cost-per-case data as to how updates to the Provider Specific File would affect hospital-specific cost-to-charge ratios during FY 2007. Instead, HHS decided to model the rate at which it believed costs per case had inflated between FYs 2004 and 2005.

To do so, HHS looked first to the FY 2005 "market basket percentage increase." Id . The "market basket measures the pure price change of [goods and services] used by a provider in supplying healthcare services," such as the wages a hospital will need to pay its employees and the cost of the equipment it will need to buy to furnish medical services. Market Basket Definitions and General Information , CTR. FOR MEDICARE & MEDICAID SERVS. , https://www.cms.gov/ Research-Statistics-Data-and-Systems/Statistics-Trends-and-Reports/MedicareProgramRatesStats/Downloads/info.pdf (last visited Aug. 7, 2017). Because the market basket represents the cost of the various inputs that go into a hospital's efforts to provide care, the Medicare Act uses the annual "market basket percentage increase" for such purposes as estimating a hospital's "expected costs" in an upcoming fiscal year. 42 U.S.C. § 1395ww(b)(2)(D). (HHS had also used the market basket percentage increase in FY 2004 when modeling projection cost-to-charge ratios for the 50 hospitals it anticipated subjecting to reconciliation. FY 2004 Final Rule, 68 Fed. Reg. at 45,477.) In cooperation with an economic forecasting firm, HHS can determine "the final updated market basket increase" for a given fiscal year in advance of receiving hospitals' tentatively settled cost reports for the same period. FY 2007 Final Rule, 71 Fed. Reg. at 48,150.

To estimate the rate of cost inflation between FYs 2004 and 2005, HHS first calculated the average relationship between the market basket percentage increase and industry-wide cost-per-case inflation across three consecutive periods: FYs 2001 to 2002, 2002 to 2003, and 2003 to 2004. HHS then multiplied that three-year average by the market basket percentage increase for FY 2005 to approximate the inflation in costs per case between FYs 2004 and 2005. Id. Put simply, HHS used the known increase in the cost of the market basket as a proxy for measuring the increase in costs per case. The resulting figure then served as the numerator of the adjustment factor. By dividing that numerator by the denominator discussed above, HHS generated its ultimate adjustment factor of -0.27%. Id.

B.

The Hospitals applaud HHS's decision to adjust its projection cost-to-charge ratios downward when setting the FY 2007 fixed-loss threshold. They challenge, however, the magnitude of that adjustment. Specifically, the Hospitals note that HHS's own pronouncements indicated that cost-to-charge ratios in the Provider Specific File had declined by approximately 2% in the previous twelve months. From this, the Hospitals infer that HHS acted arbitrarily and capriciously in applying an adjustment figure (-.27%) that diverged so substantially from the recent historical trend. Instead, the Hospitals suggest, HHS should have adopted the "obvious alternative" approach of assuming that cost-to-charge ratios would continue declining at the same rate as in recent years. Appellants' Br. 73.

The Hospitals' challenge in this respect falls short. As an initial matter, it is unclear that HHS's adjustment factor is as inconsistent with the historical trend as the Hospitals suggest. By the Hospitals' own account, "the average annual drop in the national average [cost-to-charge ratio] was 4.8%" between fiscal years 2001 and 2005, whereas HHS had noted only a "one-year actual decline ... of 2%" during the twelve-month period leading up to the FY 2007 rulemaking. Appellants' Br. 74–75 (citing FY 2007 Final Rule, 71 Fed. Reg. at 48,151 ; FY 2007 NPRM, 71 Fed. Reg. 23,996, 24,150 (April 25, 2006) ) (emphasis omitted). The Hospitals fail to explain why modeling a further 1.5% slowdown was so inconsistent with the recent record trend as to cast the adjustment factor into doubt. There may well be many non-arbitrary reasons for predicting that costs and charges in a particular industry will not continue on their current trajectories. That is particularly so in the circumstances presented here, involving fundamental changes to the outlier-payment system to eradicate rampant charge inflation.

More importantly, the Hospitals fail to identify any meaningful concerns with the methodology HHS employed beyond the fact that it generated a lower adjustment factor than they would have preferred. The Hospitals first criticize HHS for using cost data from FYs 2001 through 2005 in generating its cost-inflation figure, given its expressions of skepticism elsewhere in the FY 2007 Final Rule about the accuracy of pre-FY 2004 data. That critique mischaracterizes HHS's position on data from the turbo-charging era. It is true that HHS was hesitatant to make predictive judgments using charge data from that period, during which a small number of hospitals engaged in wildly inflated billing practices. See, e.g. , FY 2007 Final Rule, 71 Fed. Reg. at 48,149 ("[W]e believe that charge data from FY 2003 may be distorted due to the atypically high rate of hospital charge inflation during FY 2003."). But the Hospitals provide no reason to think that hospital costs during the turbo-charging era were similarly unreliable, and they point to no evidence suggesting that HHS mistrusted (or should have mistrusted) cost reports from that period.

Beyond that, the Hospitals merely criticize HHS for "concoct[ing] a complex formula in an attempt to model the most-recent one-year decline in [cost-to-charge ratios] nationally," rather than "basing its adjustment factor on the actual relevant record trend." Appellants' Br. 73–74 (emphases omitted). But an agency does not act arbitrarily and capriciously simply by engaging in the enterprise of predictive modeling. And the Hospitals have provided no reason to doubt that the market basket percentage increase correlated reasonably well with cost-per-case inflation. In fact, in various provisions in the Medicare Act, Congress itself specified the use of the market basket percentage increase as a measure of cost inflation. See, e.g. , 42 U.S.C. § 1395ww(b)(2)(C)–(D). HHS apparently concluded that employing known information about hospital costs during the period in question—in the form of the market basket percentage increase—would more likely result in an accurate estimate of cost-per-case inflation than would simply extrapolating from historical patterns. The Hospitals give us no persuasive basis for questioning that judgment.

To the extent the Hospitals separately object to the alleged complexity of HHS's model, that objection is without merit. A model's complexity, by itself, reveals little about its rationality. When an agency seeks "to measure actual cost changes experienced by [an] industry," there will typically be "[m]any methods ... available" for doing so. Ass'n of Oil Pipe Lines v. FERC , 281 F.3d 239, 241 (D.C. Cir. 2002) (internal quotation marks omitted). Accordingly, "courts routinely defer to agency modeling of complex phenomena." Appalachian Power Co. v. EPA , 249 F.3d 1032, 1053 (D.C. Cir. 2001). We will not remand simply because HHS declined to explain the complexity of its chosen formula relative to available alternatives.

We acknowledge that HHS's explanation of its FY 2007 methodology may not have been a model of clarity. But while we may not "supply a reasoned basis for the agency's action that the agency itself has not given," Ark Initiative v. Tidwell , 816 F.3d 119, 127 (D.C. Cir. 2016) (quoting SEC v. Chenery Corp. , 332 U.S. 194, 196, 67 S.Ct. 1760, 91 L.Ed. 1995 (1947) ), we must "uphold a decision of less than ideal clarity if the agency's path may reasonably be discerned," id. (quoting Motor Vehicle Mfrs. Ass'n v. State Farm Mut. Auto. Ins. Co. , 463 U.S. 29, 43, 103 S.Ct. 2856, 77 L.Ed.2d 443 (1983) ). Because we have been able to discern HHS's path, and because the Hospitals fail to show why it should not receive the deference typically accorded in this context, we reject the Hospitals' challenge to the FY 2007 adjustment factor.

C.

Finally, the Hospitals again take issue with HHS's decision not to account for the possibility of reconciliation in setting the fixed-loss threshold. For the reasons explained above, see Part VII.C, supra , HHS was under no obligation to do so. We therefore reject this final challenge to the FY 2007 outlier rule, as well.

X.

The Supreme Court has explained that "[i]f the record before the agency does not support the agency action, ... the proper course, except in rare circumstances, is to remand to the agency for additional investigation or explanation." Fla. Power & Light Co. v. Lorion , 470 U.S. 729, 744, 105 S.Ct. 1598, 84 L.Ed.2d 643 (1985). In such circumstances, the agency must first be "afford[ed] ... an opportunity to articulate, if possible, a better explanation." District Hospital Partners , 786 F.3d at 60 (quoting Cty. of L.A. , 192 F.3d at 1023 ). We follow that usual course here. On remand, HHS will be given a chance to remedy the explanatory deficiencies identified above.

We note that HHS has already attempted to rectify one of the flaws identified in District Hospital Partners. As explained above, we are unconvinced by the Remand Explanation's stated reasons for having used vastly unrepresentative charge data to inform the FY 2004 charge-inflation factor. In these highly unusual circumstances—in which an apparent methodological flaw that HHS has not yet had an opportunity to explain must be addressed before it could reasonably recalculate various fiscal years' fixed-loss thresholds—we remand both successfully challenged aspects of the FY 2004 Final Rule to be considered in the same posture. If HHS is again unable to supply a satisfactory explanation for including the turbo-charged data, that portion of the 2004 Rule will be subject to vacatur.

Accordingly, for the foregoing reasons, we reverse the district court's grant of summary judgment with respect to the successfully challenged aspects of the FY 2004, 2005, and 2006 Final Rules, and remand for further proceedings consistent with this opinion. In all other respects, we affirm the district court's grant of summary judgment in favor of HHS.