Millett, Circuit Judge:

Several hospitals challenge the methodology that the Department of Health and Human Services used to calculate the "outlier payment" component of their Medicare reimbursements for 2008, 2009, 2010, and 2011. Following this court's decision in Banner Health v. Price , 867 F.3d 1323 (D.C. Cir. 2017) (per curiam)-which upheld the challenged methodology at its inception in 2007-the primary question before us is whether the Department's decision to continue with its methodology after the 2007 fiscal year was arbitrary in light of accumulating data about the methodology's generally sub-par performance. Because the Department had, at best, only limited additional data for 2008 and 2009, and because the 2009 data suggested that hospitals were paid more than expected, the Department's decision to wait a bit longer before reevaluating its complex predictive model was reasonable.

On appeal, the Hospitals also challenge the Department's failure to publish a proposed, but later abandoned, draft rule during the 2003 rulemaking process. As the parties now acknowledge, Banner Health decided this issue in favor of the Department. That prior circuit precedent controls here.

I

A

Congress first established Medicare in 1965 as part of the Social Security Act, Pub. L. 89-97, Title XVIII, 79 Stat. 286 , 291 (1965), as a "federally funded medical insurance program for the elderly and disabled," Fischer v. United States , 529 U.S. 667 , 671, 120 S.Ct. 1780 , 146 L.Ed.2d 707 (2000). In its early years, Medicare paid its claims much like most other insurance providers, reimbursing hospitals for the "reasonable costs" of services provided to Medicare patients. County of L.A. v. Shalala , 192 F.3d 1005 , 1008 (D.C. Cir. 1999).

But over time, that system broke down. The "reasonable cost" payment structure provided little incentive for hospitals to husband their costs. The more they spent, the more they would receive. County of L.A. , 192 F.3d at 1008 . So healthcare costs rose, driving up the costs of the Medicare program. See id.

In 1983, Congress confronted the problem of rising costs. To better align the providers' incentives, it constructed a new "prospective" payment system that reimbursed hospitals based on the average rate of "operating costs [for] inpatient hospital services." County of L.A. , 192 F.3d at 1008 . After adopting this new scheme, the Department of Health and Human Services began to reimburse hospitals "at a fixed amount per patient, regardless of the actual operating costs they incur in rendering [those] services." Sebelius v. Auburn Reg'l Med. Ctr. , 568 U.S. 145 , 149, 133 S.Ct. 817 , 184 L.Ed.2d 627 (2013).

Generally speaking, this reimbursement system operates as follows:

First, the Secretary of Health and Human Services calculates a base payment rate. See 71 Fed. Reg. 47,870 , 47,876 (Aug. 18, 2006) (codified at 42 C.F.R. §§ 412.308 , 412.312 ). This rate contains both a labor and a non-labor cost component. Id. The Secretary then adjusts the labor-related component to account for labor costs in the area where the hospital is located. Id.

Second, the Secretary develops a list of "diagnosis-related groups." 71 Fed. Reg. at 47,876 . These groups encompass numerous related medical diagnoses that the Secretary believes impose a similar cost on the provider hospital. Id. To reflect the average cost of treatment for patients in each diagnosis group, the Secretary establishes a unique "relative weight" for that group. Id.

Third, the base payment rate is multiplied by the relative weight to create a generic payment amount for each diagnosis-related group. 71 Fed. Reg. at 47,876 . That is:

Base Payment Rate × Relative Weight = Generic Prospective Payment

Fourth, qualifying hospitals can receive various payment "add-ons." For example, if a hospital treats a high proportion of low-income patients, it receives a percentage increase in Medicare payments known as the "disproportionate share hospital (DSH) adjustment." 71 Fed. Reg. at 47,876 . Likewise, if the hospital serves as an approved teaching hospital, it can receive a percentage add-on payment, known as the indirect medical education adjustment. Id. Hospitals also can receive additional payments for cases involving the use of new technologies. Id.

Fifth, even with those add-ons, Congress recognized that healthcare providers would encounter patients with needs well outside the norm. County of L.A. , 192 F.3d at 1009 . To account for those abnormally costly cases and to protect against large financial losses for hospitals, the statute permits hospitals to request additional "outlier payments." See 42 U.S.C. § 1395ww(d)(5)(A)(ii). Hospitals may seek such payments where "charges, adjusted to cost, * * * exceed the sum of the applicable [diagnosis-related group] prospective payment rate plus any amounts payable under [the payment adjustment provisions] plus a fixed dollar amount determined by the Secretary." Id. In other words, hospitals are eligible for outlier payments where:

Charges, Adjusted to Cost > (Generic Prospective Payment + Any Payment Adjustments + Additional Buffer Amount (set by the Secretary))

Any cost-adjusted charges above the applicable threshold are eligible for outlier compensation. Charges below the threshold are not. For that reason, the latter half of the above formula-the generic prospective payment, adjustments, and additional buffer (or "outlier threshold")-is collectively referred to as the "fixed-loss cost threshold." 72 Fed. Reg. 47,130 , 47,417 (Aug. 22, 2007) (codified at 42 C.F.R. pt. 412). The Department can control the threshold (and thus the number of cases eligible for outlier payments) by adjusting the additional buffer amount up or down at the start of the year.

The first figure in the formula, "charges, adjusted to cost," represents the estimated cost of care for the patient at issue. Since the Department will not know the hospital's actual cost of care at the time of payment, it can only estimate the hospital's costs using historical information about the hospital's past costs in relation to its prior charges. 1 The Department estimates costs as follows:

Historical Costs Charges, Adjusted to Cost = Actual Charges × __________________ Historical Charges

See 53 Fed. Reg. 38,476 , 38,503 (Sept. 30, 1988).

The final piece of this formula-historical costs/historical charges-is known as the hospital's "cost-to-charge ratio." It reflects the percentage of that hospital's charges attributable to actual costs. To illustrate: If a hospital submits a bill for $1,000, the Department will look to see whether the hospital's estimated costs (or, as the Department refers to them, "charges, adjusted to cost") exceed the fixed-loss cost threshold. To do so, it will first need to know the costs embedded in that $1,000 charge. That is where the cost-to-charge ratio enters in. If the hospital charged $500 for this procedure in prior years, and its costs were $375, the hospital would have a cost-to-charge ratio of $375/$500 or .75. Put another way, in the past, 75% of the hospital's charges reflected its costs of care. Knowing that, the current costs of care can be estimated as follows:

$1,000 × .75 = $750

In this instance, the hospital's "charges, adjusted to costs" would be $750, and that number can be weighed against the fixed-loss cost threshold for that patient's diagnosis-related group to determine whether the hospital should receive an outlier payment.

Finally, the statute provides that the total outlier payment for a given hospital "shall be determined by the Secretary and shall * * * approximate the marginal cost of care beyond the [applicable] cutoff point." 42 U.S.C § 1395ww(d)(5)(A)(iii). To implement this objective, the Department currently pays 80% of all costs above the applicable threshold. 42 C.F.R. § 412.84 (k).

Continuing the previous example: If the fixed-loss cost threshold was $500, but the estimated cost of a patient's care was $750, the hospital would be eligible for an outlier payment of $200 (or 80% of $250, the amount falling above the $500 threshold). If, however, the threshold was $1,000, the hospital would receive no payment at all.

B

Over the years, the Department has taken various approaches to the cost-to-charge ratio data used to calculate the cost-adjusted charges in this formula. Originally, it employed a nationwide ratio for all hospitals. In the late 1980s, it shifted to hospital-specific ratios that more accurately reflected the costs at a given facility. 53 Fed. Reg. at 38,503, 38,507-38,509 (codified at 42 C.F.R. pts. 405, 412, 413 & 419). But even after making that change, the Department had difficulties aligning its projections with actual costs of care. Prior to 2003, it used cost and charge data from the "latest available settled cost report" without any forward projections. 68 Fed. Reg. 34,494 , 34,495 (June 9, 2003) (codified at 42 C.F.R. pt. 412). 2 But cost reports take several years to settle. And that time lag generated opportunities for abuse. Hospitals could manipulate their outlier payments by inflating current charges so that the historic cost-to-charge ratio employed to calculate outlier payments did not reflect the hospital's true costs. In those situations, the hospital's cost-to-charge ratio would overstate actual costs, resulting in an inflated cost estimate for the current year's claims. Id. at 34,496 .

To use another example: If a hospital charged $2,000 for a certain procedure ($1,500 of which reflected the hospital's costs), it would have a historic cost-to-charge ratio of 75%. If the hospital wanted to increase its chances of obtaining an outlier payment, it could simply increase the charge for that procedure to $4,000. Using the now outdated cost-to-charge ratio, the Department would calculate the hospital's estimated costs as $3,000 ($4,000 * .75), though, in reality, the costs were likely much closer to the original $1,500.

To make matters worse, prior to 2003, if a hospital's cost-to-charge ratio fell outside a specified window, the Department would substitute a statewide cost-to-charge ratio in lieu of the hospital-specific ratio. 68 Fed. Reg. 10,420 , 10,424 (March 5, 2003). That compounded the benefit of charge inflation because hospitals could reap the rewards of inflated charges in the short term without feeling the effects of a deflated cost-to-charge ratio in subsequent payment periods.

Unsurprisingly, this approach led to rampant inflation in hospital charges, a problem that came to be known as "turbo-charging." Banner Health , 867 F.3d at 1333 . The Department later identified 123 hospitals that had engaged in the practice of turbo-charging starting in the 1990s. 68 Fed. Reg. at 10,423.

In 2003, when the turbo-charging problem came to light, the Department issued a regulation that addressed the problem in two key ways. First, the Department sought to close the gap between cost-to-charge ratios and current costs by using more recent data-specifically by permitting the use of "either the most recent settled or the most recent tentative settled cost report, whichever is from the later cost reporting period." 68 Fed. Reg. at 34,499. Second, the Department reserved the right to recalculate a hospital's eligibility using actual cost data at the time of settlement. Id. at 34,501. Through this process, known as reconciliation, the agency could claw-back undue outlier payments. Id.

At that same time, the Department also considered immediately adjusting the 2003 outlier threshold, which had been set at the beginning of the year, to account for the effect of the reforms on 2003 outlier payments. To that end, the Department drafted a rule proposing to decrease the existing outlier threshold for the remainder of the 2003 fiscal year. The Department ultimately abandoned that effort, opting instead to maintain the existing threshold until the year's end to allow rates to settle. 68 Fed. Reg. at 34,506. So the draft rule was never published. The Hospitals later obtained a copy of the draft through a Freedom of Information Act request.

C

To add complexity to the complexity, the Medicare statute also limits the total amount of all outlier payments the Department can make in a given fiscal year-setting both a floor and a ceiling. Under the Act, the "total amount of" outlier payments made in a fiscal year "may not be less than 5 percent nor more than 6 percent of the total payments projected or estimated to be made based on [diagnosis-related group] prospective payment rates for discharges in that year." 42 U.S.C. § 1395ww(d)(5)(A)(iv). Of course, that requires the Department to estimate certain numbers at the start of the year before it has actual claims information. Because the statutory target is tied to "projected or estimated," not actual, payments, id ., the Department has interpreted the statutory directive to mean that the fixed loss threshold must be set at a level that, "when tested against historical data, will likely produce aggregate outlier payments totaling between five and six percent of projected or estimated [diagnosis-related group] payments." County of L.A. , 192 F.3d at 1013 (emphasis added); see e.g. , 50 Fed. Reg. 35,646 , 35,710 (Sept. 3, 1985). For every year since 1989, the Department has aimed to set the fixed-loss cost threshold so that the total outlier payments will be 5.1% of all Medicare payments, or:

Total Outlier Payments = 5.1% (Total Projected Medicare Payments)

But, alas, this is a predictive enterprise. The Department must set the outlier threshold at the start of each year before it knows how many hospitals will actually have outlier patients. In other words, the agency must estimate the number of outlier cases for the upcoming year and set a threshold that it believes will result in outlier payments of 5.1%.

As a result, to compute an appropriate outlier threshold, the Department must estimate the total outlier costs for all hospitals for the upcoming year. In practice, the Department uses a formula similar to the one it uses to calculate actual outlier payments, inputting projected and historical cost and charge information to fill in the gaps. More specifically, the agency takes historical charges and projects them forward to reflect the upcoming year's charges. The Department then takes the historical cost-to-charge ratio from the most recent year available and projects those figures forward to predict current cost-to-charge ratios. 3 The basic formula is as follows:

Total Charges, Adjusted to Costs = Projected Charges × Projected Cost Ratio

In 2007, the Department refined its methodology for projecting historical cost-to-charge ratios forward to the current year. 71 Fed. Reg. at 48,149 . In the aftermath of the 2003 turbo-charging reforms, cost-to-charge ratios had fallen. As a result, the Department's efforts to hit its 5.1% target had begun to land consistently short. Commenters, including many hospital providers, asked the Department to adopt an "adjustment factor" to update the cost-to-charge ratio formula based on the current downward trend in hospital costs. Id. at 48,150 . The Department agreed. It proposed the following formula to estimate costs for purposes of determining the outlier threshold:

Total Charges, Adjusted to Costs = (Historical Charges × Charge Inflation Factor) × | Historical Costs Cost Inflation Factor | |__________________ × _______________________| |Historical Charges Charge Inflation Factor|

See id. at 48,149 .

This refined formula relied upon two critical metrics to predict future costs and charges from the available historical information: the cost inflation factor and the charge inflation factor.

The latter was relatively simple. The agency would update historical charge data using the average annualized rate of change in charges per case. In other words:

Charge Inflation Factor = Average Rate of Annual Change in Charges per Case

71 Fed. Reg. at 48,149 .

The cost inflation factor, however, was more complex. It factored in both hospital-specific cost inflation and general inflation as measured by the change in a standard market basket of goods and services. At the highest level, the formula is as follows:

Cost Inflation Factor = (Average Annual Hospital Cost Inflation for three years prior × Annual Inflation of Market Basket for the most recent year available)

And within this formula:

Annual Hospital Cost Inflation = Annual Change in Costs per Discharge _____________________________________ Annual Market Basket Increase

To put this all together in a more concrete setting, the Department would calculate the cost inflation factor for purposes of the 2008 threshold using historical data as follows:

2008 Cost Inflation Factor = | |2004 to 2005 Change in Costs per Discharge | | | |__________________________________________+| | | | 2005 Market Basket Increase | | | | | | | |2003 to 2004 Change in Costs per Discharge | | | |__________________________________________+| | | | 2004 Market Basket Increase | | | | | | | | 2002 to 2003 Change in Costs per Discharge| | | | __________________________________________| | | | 2003 Market Basket Increase | | | |_____________________________________________| | | | 3 | | | × 2006 Market Basket Increase

See 71 Fed. Reg. at 48,150 (outlining this process for the 2007 threshold).

Once the agency calculates both the cost inflation and charge inflation factors, it then divides the cost inflation factor by the charge inflation factor to obtain the "adjustment factor." This adjustment factor is then multiplied by the historical cost-to-charge ratio to obtain an updated, or projected, cost-to-charge ratio for that year.

Historical Costs Cost Inflation Factor __________________ × _______________________ Historical Charges Charge Inflation Factor = Projected Cost Ratio

D

Many providers supported the downward adjustment in the cost-to-charge ratios, but not the "magnitude of that adjustment." Banner Health , 867 F.3d at 1355 . Several hospitals then challenged the 2007 outlier payment methodology. This court rejected those claims, holding that the Department's new formula was not arbitrary or capricious in light of the information available to the Department. Id. at 1355-1356 (finding "many non-arbitrary reasons for predicting that costs and charges * * * will not continue on their current trajectories").

The following year, the Department used the same methodology to calculate the 2008 outlier threshold. See 72 Fed. Reg. at 47,417 . The result was a proposed fixed loss threshold of the prospective payment rate plus any cost adjustments plus $23,015. Id.

Several commenters thought that the buffer amount was too high. They noted that outlier payments had been, by their calculation, only 4.63% of overall 2007 payments. They urged the Department to adopt a simplified cost inflation factor based on the actual rate of change in hospital costs-the same method already being used to estimate projected charges. Commenters also faulted the agency for not using more recent cost-to-charge ratio data, and suggested applying the cost-to-charge ratio adjustment factor over different periods of time (longer or shorter than one year) based on individual hospital's fiscal calendars. Others simply urged the Department to lower the threshold without providing an alternative approach.

The Department rejected all of those proposals. 72 Fed. Reg. at 47,418 . By the time of the final rule, the Department estimated that outlier payments for 2007 had been 4.6% of overall payments. Id. at 47,420 . That fell far short of the 5.1% payment target. But because cost reports had not been settled for 2007, 4.6% was still just an estimate of what the actual 2007 percentage would be. Id. ("This estimate is based on simulations[.]"). For that reason, the Department declined to alter its methodology, reasoning that its chosen cost inflation factor was "more accurate and stable than the commenter's * * * because it takes into account the costs per discharge and the market basket percentage increase[.]" Id. at 47,418 . As it had in prior years, the Department also determined that its methodology would not account for any potential payment claw-backs made during the reconciliation process because those would be too "difficult to predict" and because the amounts are generally too small to materially affect the predictive assumptions. Id. at 47,419 . The Department, however, did update the data for its final rule, a change that resulted in a lower buffer amount of $22,635. Id.

That process was largely repeated in 2009. See 73 Fed. Reg. at 48,763. The Department proposed a fixed loss threshold of the prospective payment rate plus any cost adjustments plus $21,025, and ended with an updated, final buffer of $20,185. 73 Fed. Reg. 23,528 , 23,711 (April 30, 2008) ; 73 Fed. Reg. at 48,766 .

Once more, commenters challenged the cost inflation factor, calling it "unnecessarily complicated." 73 Fed. Reg. at 48,764 . They again urged the agency (i) to use more recent, historical, and industry-wide rates of change; (ii) to vary the cost-to-charge ratio adjustment factor to more or less than one year; and (iii) to apply more recent hospital data to its calculations. Id.

As before, the Department rejected the suggestions, largely reiterating the reasons it had provided in 2008. See 73 Fed. Reg. at 48,763 . The Department now estimated 2007 outlier payments as 4.6% of final payments, and it projected 2008 payments to be 4.7%. Despite widely missing the 5.1% mark again, the Department maintained its approach to cost inflation, asserting that the use of "the market basket in conjunction with the cost per discharge takes into account two sources [of] potential cost inflation and ensures a more accurate and stable cost adjustment factor." Id. at 48,764 .

For 2010, the agency again employed the same formula. 74 Fed. Reg. at 44,007. Using that methodology, the agency proposed a fixed loss threshold of the prospective payment rate plus any adjustments plus $24,240, a 21% increase from the previous fiscal year. With updated data, the Department later arrived at a final buffer of $23,140.

This time, the Department's estimates appeared slightly more promising. By the final rulemaking, the Department estimated that 2008 outlier payments had been 4.8% of final payments, but 2009 outlier payments had been 5.4% of final payments-meeting and even exceeding the 5.1% target.

Nevertheless, the proposed increase in the buffer amount prompted renewed protest. See 74 Fed. Reg. at 44,007. Commenters could not understand why-when the agency had met its target in 2009-there should be any change to the threshold amount. Id. at 44,009. Others accused the agency of purposefully erring on the low end of the 5% to 6% target. Id. Another asked the agency to make a mid-year adjustment if it appeared that the existing threshold would not result in payments in the 5% to 6% range. Id. Still others repeated the requests to use more recent data in the final rule and to account for reconciliation. Id. at 44,009-44,010.

For its part, the Department insisted that it had "use[d] the most recent data available to set the outlier threshold." 74 Fed. Reg. at 44,009. It rejected the mid-year course correction because such adjustments would be "extremely difficult or impracticable (if not impossible) to administer." Id. (incorporating 70 Fed. Reg. 47,278 , 47,495 (Aug. 12, 2005), the Department's response to a similar request made in the 2006 rate proposal). And it denied all other suggestions for the same reasons given in prior rulemakings. See id. at 44,010 .

Fiscal year 2011-the last at issue in this case-proved no different. See 75 Fed. Reg. 50,042 , 50,427 (Aug. 16, 2010). The Department proposed a buffer of $23,970. 75 Fed. Reg. at 24,069 . And it ultimately settled on $23,075 as the final amount. 75 Fed. Reg. at 50,430 . The complaints from hospitals and the answers from the Department were repeated. In addition, using updated data, the Department calculated 2009 outlier payments to be 5.3% of final payments, and 2010 outlier payments to be 4.7% of final payments. Commenters took issue with the Department's 2009 estimate, arguing that its calculations indicated outlier payments of only 4.9% for the 2009 year.

The following chart summarizes the key data points from each rulemaking:

2008 2009 2010 2011 Proposed Fixed-Loss $23,015 $21,025 $24,240 $23,075 Threshold Final Fixed-Loss $22,635 $20,185 $23,140 $23,075 Threshold Agency's Target 5.1% 5.1% 5.1% 5.1% Agency's 4.8% 5.3% 4.7% 4.8% Reported Estimate Hospitals' 4.6% 4.9% N/A N/A Estimate Shortfall -0.3% +0.2% -0.4% -0.3% (Agency Estimate) Shortfall -0.5% -0.2% N/A N/A (Hospital Reported Estimate)

E

There is yet one final piece of this byzantine process that bears a bit of explanation. When a hospital seeks Medicare payments from the Department, it must first submit its request to a fiscal intermediary-that is, a contracted entity to which the Department has delegated payment determinations. See 42 U.S.C. §§ 1395kk-1, 1395oo(a). If the hospital is unsatisfied with the intermediary's final determination, it may appeal the decision to the Provider Reimbursement Review Board. Id. In the normal course, the Board would review the claim, and the hospital would retain the right to seek "judicial review of any final decision of the Board." 42 U.S.C. § 1395oo(f)(1).

However, if the hospital's claim "involves a question of law or regulations relevant to the matters in controversy * * * [that the Board] is without authority to decide," the hospital can ask the Board to allow it go directly to district court. 42 U.S.C. § 1395oo(f)(1) ; 42 C.F.R. § 405.1842 . If the Board agrees, it will certify the question for immediate judicial review. Id. Only the hospital can challenge the Board's certification determination. See Allina Health Servs. v. Price , 863 F.3d 937 , 941-942 (D.C. Cir. 2017).

F

Several acute care hospitals have challenged the outlier payments received in 2008, 2009, 2010, and 2011. They allege that the Department's methodology for determining the outlier threshold during this period was arbitrary and capricious, pointing in particular to the consistent underpayments in each of the relevant years and the failure to account for the possibility of reconciliation. They also object to the Department's failure to publish the proposed but ultimately not-adopted 2003 draft rule, which, in their view, contains much-needed ammunition to show that the Department should have updated its methodology in these later years.

Because the Hospitals challenged the legality of the applicable outlier thresholds, they requested expedited access to judicial review from the Board. The Board granted many of those certification requests, either initially or on reconsideration. With respect to some Hospitals, however, the Board concluded that it lacked jurisdiction to grant expedited review because those Hospitals had failed to comply with the required procedures for filing their cost reports. See e.g. , Ex. 1 PRRB Decisions, Lee Mem'l Hosp. v. Sebelius , No. 1:13-cv-00643 (D.D.C. Jan. 10, 2014), ECF No. 22-1 (citing 42 C.F.R. § 405.1835 (a)(1)(ii) (2013) ). For that subset of Hospitals, the Board dismissed the cases for lack of jurisdiction without deciding the certification question.

Both the dismissed and certified Hospitals filed suit in district court. The government conceded that the Board erred in dismissing some of the cases for lack of jurisdiction. In light of that concession, the district court held that the Board had jurisdiction, and that the court could likewise exercise its own jurisdiction under 42 U.S.C. § 1395oo(f)(1).

The district court subsequently granted summary judgment for the Department, concluding that its approach to calculating the outlier threshold was not arbitrary or capricious. The Hospitals moved for reconsideration, and the district court denied the motion. The Hospitals now appeal.

II

A

We start, as we must, with jurisdiction. The Medicare Act specifies that "[n]o findings of fact or decision of the [Secretary] shall be reviewed by any person, tribunal, or governmental agency" except as the Medicare Act itself provides jurisdiction. 42 U.S.C. § 405 (h). The relevant source of jurisdiction in this case is 42 U.S.C. § 1395oo(f). That provision allows providers to seek review of a final decision of the Provider Reimbursement Review Board and to seek expedited judicial review where the Board lacks "authority to decide" a question of law relevant to the matter at hand. Id .

As noted, for the majority of the plaintiff Hospitals, the Board granted expedited review on the ground that it lacked authority to override the outlier regulations. The district court properly exercised jurisdiction over those claims. See Allina Health Servs. , 863 F.3d at 941-942 .

As for the plaintiff Hospitals over which the Board declined to exercise jurisdiction, the question is more complicated. While the Secretary has since disavowed the Board's procedural objection to the claims in that case, that leaves unanswered whether the district court could proceed without first remanding for either a final decision or certification for expedited review from the Board.

We need not resolve that jurisdictional quandary because there are Hospitals with valid Board certifications of expedited review for each of the years at issue, and only non-individualized injunctive relief is sought. We accordingly proceed to the merits on a clean jurisdictional slate.

B

Under the Administrative Procedure Act, we may only set aside agency action if it is "arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law." 5 U.S.C. § 706 (2)(A). In making that assessment, we must ensure that the agency has "examine[d] the relevant data and articulate[d] a satisfactory explanation for its action including a rational connection between the facts found and the choices made."

Motor Vehicle Mfrs. Ass'n v. State Farm Mutual Auto. Ins. Co. , 463 U.S. 29 , 43, 103 S.Ct. 2856 , 77 L.Ed.2d 443 (1983) (citation and internal quotation marks omitted).

The Hospitals argue that the cost-projection methodology used by the Department to set the annual outlier thresholds is arbitrary and capricious for three reasons. First, they object to the Department's failure to publish the 2003 draft rule, which they allege deprived them of useful information in the subsequent rulemakings. Second, they challenge the Department's failure to account for the possibility of reconciliation claw-backs in setting the 2008, 2009, 2010, and 2011 thresholds. Third, they object to the Department's continued use of its cost-inflation methodology in the face of repeated underpayments and the availability of a simpler formula, which the Department was already using to calculate inflation in hospital charges.

The first two challenges are foreclosed by circuit precedent. See Banner Health , 867 F.3d at 1337, 1356 . As for the third, while the Hospitals' frustration with the Department's frequently off-target calculations is understandable, the methodology has not sunk to the level of arbitrary or capricious agency action.

1

Our decision in Banner Health , which involved a similar challenge to the 1997 through 2007 outlier payment rates, disposes of the Hospitals' procedural challenge regarding the Department's unpublished draft rule. There, this court held that the Department did not err in failing to disclose the 2003 draft rule because it had not relied on it in crafting its final rule. Banner Health , 867 F.3d at 1337 . 4

Banner Health also largely answers the Hospitals' argument that the Department had to factor reconciliation claw-backs into its threshold predictions. Banner Health rejected that exact same challenge to the 2005 outlier thresholds. 867 F.3d at 1351-1352, 1356 . We held that the Department "was under no obligation" to "account for the possibility of reconciliation in setting the fixed-loss threshold." Id. at 1356 ; see also District Hosp. Partners, L.P. v. Burwell , 786 F.3d 46 , 61 (D.C. Cir. 2015) (concluding that the 2003 reforms "corrected the flaw in the outlier payment system that created the opportunity-and incentive-to turbo-charge," and thus the need for large reconciliations).

That conclusion applies with equal force to the 2008 through 2011 outlier thresholds. As in Banner Health , the Department reasonably concluded "that [the] charging practices would not fluctuate significantly enough to justify accounting for reconciliation[.]" 867 F.3d at 1352 (alterations, citation, and internal quotation marks omitted); see e.g. , 72 Fed. Reg. at 47,419 (rejecting the need to consider reconciliation for the 2008 threshold for these same reasons). Nothing in the current record supports a different answer here. The Office of the Inspector General's 2012 Report, to which the Hospitals direct us, does not indicate that these payments occur with more regularity now than the Department suggested at the time of Banner Health .

Finally, Banner Health sanctioned the agency's 2007 methodology for calculating cost inflation, at least to the extent that the Hospitals challenge its facial validity. See 867 F.3d at 1356 ("[T]he Hospitals have provided no reason to doubt that the market basket percentage increase correlated reasonably well with cost-per-case inflation."). On top of that, just as Banner Health recognized, there is good reason to believe that a model that accounts for macro-level change may better predict future costs than one that does not. Id.

The Hospitals' proposal of a simpler method does not make the Department's method arbitrary. "A model's complexity, by itself, reveals little about its rationality." Banner Health , 867 F.3d at 1356 . In the wake of the turbo-charging scandal and persistent cost containment problems, it was not unreasonable for the Department to be wary of an industry-specific inflation metric. Plus, as the Department explained, a method accounting for general inflation "is more accurate and stable" than the industry-specific alternative. 72 Fed. Reg. at 47,418 . 5

2

That leaves one final question: Did the Department act arbitrarily in maintaining its cost inflation methodology after the 2007 fiscal year in the face of numerous underpayments? In light of the short pattern of missed targets, the limited and inconsistent data available to the Department at the time of its rulemakings, and the lengthy time lag in finally determining the actual payments made for a preceding year, we conclude that it did not.

To be sure, a methodology used for prediction "can look more arbitrary the longer it is applied." American Petroleum Inst. v. EPA , 706 F.3d 474 , 477 (D.C. Cir. 2013). But that line was not crossed for the years at issue here. In 2008, the Department had only one year's worth of yet unsettled data to test the validity of its new model. By the time the Department had to develop its 2008 outlier payment amount, it had not definitively settled on the 2007 results, a cost-resolution process that can take several years. The tentative 2007 results were not so far off base as to suggest the need for immediate abandonment of the newly adopted system. Even by 2009, the Department had only a slightly more settled 2007 estimate and a tentative number for 2008. The time-lag inherent in obtaining accurate payment data underscores the reasonableness of the Department's deliberative and cautious approach to evaluating the operation and accuracy of its methodology.

Putting some proof in the pudding, the Department's calculations indicated that it not only met, but exceeded, the 5.1% mark in 2009, resulting in payments to the Hospitals that exceeded the Department's 5.1% target. Even considering the Hospitals' contrary estimation of 4.9%, the 2009 payments nearly reached the Department's intended target. For that reason, we cannot conclude that the Department acted arbitrarily in continuing to employ its predictive model for the next two years while accumulating additional data points.

* * * * *

For all its complexity and labyrinthine mathematical formulae, this case turns on a simple concept: Some things take a bit of time to sort out. The Department's efforts to predict Medicare costs for patients across the Nation each fiscal year is fraught with variables, estimates, and uncertainties. The Medicare statute recognizes that difficulty by requiring the Department to model results that fall between 5% and 6% of total projected payments, without mandating that the Department actually hit the bullseye each year. See 42 U.S.C. § 1395ww(d)(5)(A)(iv). Though the Department has an obligation to act reasonably and to account for the actual results of its decisions, the need for time both to study the results and to determine how to improve accuracy must inform any evaluation of the appropriateness of the Department's actions in these years.

For those reasons, we affirm the judgment of the district court.

So ordered.