Brought by the owners of Medicare provider hospitals ("Hospitals") and the Secretary of Health and Human Services ("Secretary"), these cross-appeals present two issues. First, under the Medicare statute, must the Secretary provide hospitals with retroactive reimbursements to ensure that aggregate outlier payments during any given fiscal year meet minimum statutory targets? And second, has the Secretary adequately explained why, when calculating outlier thresholds for fiscal years 1985-1986, she relied on a 1981 database instead of more contemporaneous records from 1984 Medicare discharges? Finding that Congress had spoken directly and unambiguously to the first question, the district court granted partial summary judgment to the Hospitals. With respect to the second issue, however, the court perceived nothing unreasonable in the Secretary's choice of data, and entered judgment accordingly for the Secretary. Because we disagree with the district court on both points, we now reverse.

Through a "complex statutory and regulatory regime," Good Samaritan Hosp. v. Shalala, 508 U.S. 402, 404 (1993), the Medicare program reimburses qualifying hospitals for the services that they provide to eligible patients. See Social Security Act, Pub. L. No. 89-97, tit. XVIII, 79 Stat. 286, 291 (1965) (codified as amended at 42 U.S.C. §§ 1395-1395ggg (1994 & Supp. III 1997)). From its inception in 1965 until October 1983, Medicare compensated hospitals for the "reasonable costs" of the inpatient services that they furnished. See 42 U.S.C. § 1395f(b). Experience proved, however, that this system bred "little incentive for hospitals to keep costs down" because "[t]he more they spent, the more they were reimbursed." Tucson Med. Ctr. v. Sullivan, 947 F.2d 971, 974 (D.C. Cir. 1991).

To stem the program's escalating costs and perceived inefficiency, Congress fundamentally overhauled the Medicare reimbursement methodology in 1983. See Social Security Amendments of 1983, Pub. L. No. 98-21, § 601, 97 Stat 65, 149. Since then, this new regime, known as the Prospective Payment System ("PPS"), has reimbursed qualifying hospitals at prospectively fixed rates. By establishing predetermined reimbursement rates that remain static regardless of the costs incurred by a hospital, Congress sought "to reform the financial incentives hospitals face, promoting efficiency in the provision of services by rewarding cost/effective hospital practices." H.R. Rep. No. 98-25, at 132 (1983), reprinted in 1983 U.S.C.C.A.N. 219, 351.

Calculating prospective-payment rates begins with determining the "federal rate," a standard nationwide cost rate based on the average operating costs of inpatient hospital services. See 42 U.S.C. § 1395ww(d)(2)(A)-(B); 49 Fed. Reg. 234, 251 (1984). To account for regional variations in labor costs, the Secretary then establishes a wage index that augments the adjusted standardized payment depending on the location of a qualifying hospital. § 1395ww(d)(2)(H), (d)(3)(E). The final variable is an additional weighting factor that reflects the disparate hospital resources required to treat major and minor illnesses. § 1395ww(d)(4). For each of 470 medical conditions--known as diagnosis related groups or "DRGs"--the Secretary assigns particular weights by which the federal rate is to be multiplied. The more complicated and costlier the treatment is, the greater the weight assigned to that particular DRG will be. To calculate the final "DRG prospective payment rate" for a patient discharge, the Secretary takes the federal rate, adjusts it according to the wage index, and then multiplies it by the weight assigned to the patient's DRG. By statutory mandate, the Secretary must publish the weights and values that she will factor into the prospective-payment calculus before the start of each fiscal year. § 1395ww(d)(6).

Despite the anticipated virtues of PPS, Congress recognized that health-care providers would inevitably care for some patients whose hospitalization would be extraordinarily costly or lengthy. To insulate hospitals from bearing a disproportionate share of these atypical costs, Congress authorized the Secretary to make supplemental "outlier payments." During the years at issue in these cross-appeals, the outlier-payment provisions were set forth in four clauses of the Medicare statute. 42 U.S.C. § 1395ww(d)(5)(A)(i)-(iv) (Supp. IV 1986). With the first two clauses, Congress established two classes of outlier payments: day outliers and cost outliers. § 1395ww(d)(5)(A)(i)-(ii). A hospital could qualify for a day-outlier payment if the patient's length of stay exceeded the mean length of stay for that particular DRG by a fixed number of days or standard deviations .s 1395ww(d)(5)(A)(i). Along the same lines, the Secretary would make cost-outlier payments when a hospital's cost adjusted charges surpassed either a fixed multiple of the applicable DRG prospective-payment rate or such other fixed dollar amount that the Secretary established.s 1395ww(d)(5)(A)(ii). In the third clause, Congress provided that outlier payments "shall be determined by the Secretary and shall approximate the marginal cost of care beyond the cutoff point applicable" to the day or cost outlier.s 1395ww(d)(5)(A)(iii).

It is the fourth and final clause, however, that forms the textual nub of the present controversy.s 1395ww(d)(5)(A)(iv). During 1985 and 1986, paragraph (5)(A)(iv) provided:

The total amount of the additional payments made underthis subparagraph for discharges in a fiscal year may notbe less than 5 percent nor more than 6 percent of thetotal payments projected or estimated to be made basedon DRG prospective payment rates for discharges in thatyear.

Id. Traditionally, the Secretary has read paragraph (5)(A)(iv) to mean that at the start of each fiscal year, she must establish the fixed thresholds beyond which hospitals will qualify for outlier payments at levels likely to result in outlier payments totaling between five and six percent of projected DRG payments for that year. In making this estimation, the Secretary first settles on the per-diem outlier payment, which pursuant to § 1395ww(d)(5)(A)(iii), must approximate the marginal cost of care. She then examines historical Medicare-discharge data to determine which thresholds, when multiplied by the per-diem payment rate, would probably yield total outlier payments falling within the fiveto-six-percent range in paragraph (5)(A)(iv). As the Secretary observed during the rulemaking, however, "given the data available, forecasts of probable future outlier payments are inexact." 50 Fed. Reg. 35,646, 35,710 (1985). If it turns out that the Secretary overestimated the mean length of stay for DRGs, the actual total outlier payments at the end of the year may amount to less than five percent of estimated DRGrelated payments. Conversely, underestimating the mean length of stay might produce outlier payments in excess of six percent of estimated total DRG payments.

Whether the Secretary's projections prove to be correct will depend, in large part, on the predictive value of the historical data on which she bases her calculations. For fiscal year 1984, the Secretary relied on data culled from the 1981 Medicare Provider Analysis and Review ("1981 MEDPAR") file, a database containing 1.6 million Medicare discharges from 1981. As a product of the old reasonable-cost system, however, the 1981 MEDPAR file obviously did not reflect one of "[t]he most commonly accepted expectation[s] about the PPS at the time of its inception[:] that it would result in shorter stays for Medicare patients." Office of Research & Demonstrations, Health Care Fin. Admin., U.S. Dep't of Health & Human Servs., Pub. No. 03231, Report to Congress: Impact of the Medicare Hospital Prospective Payment System 6-13 (1984). By 1984, however, preliminary data indicated that the mean length of stay for virtually all DRGs had, as anticipated, declined dramatically under PPS. The Secretary, nevertheless, chose to rely again on the 1981 MEDPAR file in setting outlier thresholds for fiscal years 1985-1986.During those years, though the Secretary set thresholds at a level projected to result in outlier payments at or above paragraph (5)(A)(iv)'s five-percent floor, actual outlier payments in 1985 constituted only 3.0 percent of estimated DR Grelated payments and in 1986 they amounted to 4.4 percent.1Given the enormity of the Medicare program, these seemingly modest percentage differences represent substantial sums of money: $241 million for fiscal year 1985 and $101 million for fiscal year 1986.

Because actual outlier payments for fiscal years 1985-1986 amounted to less than five percent of projected DRG-related payments, the Hospitals petitioned the Secretary for retroactive reimbursements to satisfy the difference. According to the Hospitals, paragraph (5)(A)(iv) does more than instruct the Secretary about where she should set outlier thresholds at the beginning of each fiscal year; it affirmatively commands her to recalibrate retroactively the outlier thresholds if, after the fiscal year concludes, actual outlier payments do not equal at least the five-percent statutory target. Moreover, the Hospitals claimed that the Secretary had acted arbitrarily and capriciously by relying on the 1981 MEDPAR file when she forecast outlier thresholds for fiscal years 19851986. The Secretary rejected both claims, and the Provider Reimbursement Review Board authorized the Hospitals to seek expedited judicial review pursuant to 42 U.S.C. § 1395oo(f).

In an opinion and order dated January 20, 1998, the district court granted in part and denied in part both the Secretary's and the Hospitals' cross-motions for summary judgment. In granting a portion of the Hospitals' motion, the court proceeded no further than the first step of Chevron U.S.A. Inc. v. Natural Resources Defense Council, Inc., 467 U.S. 837, 842-43 (1984), concluding that paragraph (5)(A)(iv) unambiguously requires the Secretary to adjust outlier payments retroactively to ensure that total actual outlier payments fall within the statute's five-to-six-percent range. County of Los Angeles v. Shalala, 992 F. Supp. 26, 31-33 (D.D.C. 1998).Holding, however, that the Secretary's decision to favor the 1981 MEDPAR file over the more recent, though preliminary, 1984 data when determining outlier thresholds was "a rational choice between two imperfect databases," the district court granted the Secretary's motion for summary judgment on the Hospitals' claim of arbitrary and capricious agency action. Id. at 34-36.

Having determined that paragraph (5)(A)(iv) required the Secretary to make retroactive outlier payments to the Hospitals, the district court instructed the parties to meet and confer on how to structure the final remedy. On April 30, 1998, based largely on a joint stipulation filed by the parties, the district court entered an order granting judgment to the parties on each of the claims on which they respectively prevailed. While the April 30 order also instructed the Secretary to tender retroactive outlier payments to the Hospitals, it did not specify a sum certain to be paid; rather, the

court left it to the Secretary to calculate the amount owed. Thereafter, the Secretary noted a timely appeal and the Hospitals noted timely cross-appeals.

Before reaching the merits, it is necessary to examine our jurisdiction to entertain these cross-appeals. Section 1291 of the Judicial Code provides that the courts of appeals may review "all final decisions of the district courts of the United States." 28 U.S.C. § 1291 (1994). In the proceedings below, the district court, managing this case as it would any garden variety civil suit, adjudicated not only the respective legal rights of the parties but also took steps toward decreeing a proper remedy. Thus in its January 20, 1998 order, the court resolved the merits of the Hospitals' claims, and with its April 30, 1998 order, directed the Secretary to calculate the amount of outlier payments due to the Hospitals and to make payment accordingly. This latter order has spawned some confusion about our jurisdiction because of the general rule applicable to civil actions that "where assessment of damages or awarding of other relief remains to be resolved," a district court's judgment is not " 'final' within the meaning of 28 U.S.C. § 1291." Liberty Mut. Ins. Co. v. Wetzel, 424 U.S. 737, 744 (1976); see also A & S Council Oil Co. v. Lader, 56 F.3d 234, 238 (D.C. Cir. 1995) (holding that an order establishing liability but referring the issue of damages to arbitration is not final). For it is clear that neither of the district court's orders resolved the precise quantum of payments to be made to the Hospitals.

This rule of finality does not apply here, however, because this is not an appeal from an ordinary civil judgment rendered by the district court. With both of their claims, the Hospitals challenged the Secretary's actions under section 10(e) of the Administrative Procedure Act, 5 U.S.C. § 706(2).As we have often observed, "[w]hen a final agency action is challenged under the APA in district court, if the relevant substantive statute does not provide for direct review in the court of appeals, the district court does not perform its normal role" but instead "sits as an appellate tribunal." PPG Indus., Inc. v. United States, 52 F.3d 363, 365 (D.C. Cir. 1995) (quoting Marshall County Health Care Auth. v. Shalala, 988 F.2d 1221, 1225 (D.C. Cir. 1993)); accord James Madison Ltd. v. Ludwig, 82 F.3d 1085, 1096 (D.C. Cir. 1996) ("Generally speaking, district courts reviewing agency action under the APA's arbitrary and capricious standard do not resolve factual issues, but operate instead as appellate courts resolving legal questions."), cert. denied, 519 U.S. 1077 (1997).Whether it is a court of appeals or a district court, "[u]nder settled principles of administrative law, when a court reviewing agency action determines that an agency made an error of law, the court's inquiry is at an end: the case must be remanded to the agency for further action consistent with the corrected legal standards." PPG Indus., 52 F.3d at 365; see also South Prairie Constr. Co. v. Local No. 627, Int'l Union of Operating Eng'rs, 425 U.S. 800, 806 (1976); SEC v. Chenery Corp., 318 U.S. 80, 94-95 (1943). Once, therefore, the district court held that the Secretary had misinterpreted § 1395ww(d)(5)(A)(iv), it should have remanded to the Secretary for further proceedings consistent with its conception of the statute. Not only was it unnecessary for the court to retain jurisdiction to devise a specific remedy for the Secretary to follow, but it was error to do so. See Ommaya v. National Insts. of Health, 726 F.2d 827, 830 (D.C. Cir. 1984) ("If MSPB relied on incorrect legal grounds, it would be error for this court to enforce without first remanding for agency examination of the evidence and proper fact-finding.") (quoting White v. United States Dep't of the Army, 720 F.2d 209, 210 (D.C. Cir. 1983)). Accordingly, because that was all that the district court had the power to do, we construe its January 20, 1998 order as a remand to the Secretary, and ignore, for jurisdictional purposes, its later order on specific relief.

Of course, properly characterizing the district court's order as a remand does not, without more, resolve our jurisdictional quandary, for a "remand order usually is not a final decision."NAACP v. United States Sugar Corp., 84 F.3d 1432, 1436 (D.C. Cir. 1996). We have recognized "an exception to this general rule, however, where the agency to which the case is remanded seeks to appeal and it would have no opportunity to appeal after the proceedings on remand." Occidental Petroleum Corp. v. SEC, 873 F.2d 325, 330 (D.C. Cir. 1989).Animating this principle is a pragmatic concern. Because an agency must conduct its proceedings and render its decision pursuant to the legal standard that the district court articulates in its remand order, "[u]nless another party appeals [the agency's subsequent] decision, the correctness of the district court's legal ruling will never be reviewed by the court of appeals, notwithstanding the agency's conviction that the ruling is erroneous." Id. Here, were the Secretary unable to appeal the district court's decision at this point, on remand she would have to interpret paragraph (5)(A)(iv) as the district court has construed it, and disburse millions of dollars in retroactive outlier payments to various Medicare-provider hospitals. Absent an appeal from that decision by one of the Hospitals, the Secretary would have no opportunity to challenge the legal basis for the disbursements. Our jurisdiction is therefore proper because the Secretary's appeal falls within the exception recognized in Occidental Petroleum. Additionally, vested with jurisdiction to review the Secretary's appeal under § 1291, we may also consider the Hospitals' cross appeal of the district court's grant of summary judgment to the Secretary on their arbitrary and capricious agency-action claim. See United States Sugar Corp., 84 F.3d at 1436.

Turning first to the Secretary's challenge, we face a question of statutory interpretation that the district court resolved in the affirmative: whether, under § 1395ww(d)(5)(A)(iv), the Secretary must make retroactive reimbursements to ensure that aggregate outlier payments during any given fiscal year constitute at least five percent of estimated or projected DRG-related payments. Because we may set aside agency action only if it is "arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law," 5 U.S.C. § 706(2)(A), we accord no particular deference to the judgment of the district court when conducting our review. See HCA Health Servs. v. Shalala, 27 F.3d 614, 616 (D.C. Cir. 1994); Hennepin County v. Sullivan, 883 F.2d 85, 91 (D.C. Cir. 1989) ("Our review proceeds as if this case were an immediate appeal from a decision reached after an administrative hearing on the record."); Biloxi Reg'l Med. Ctr. v. Bowen, 835 F.2d 345, 348-49 (D.C. Cir. 1987).

We initiate statutory analyses of the sort presented here by first asking whether "Congress has directly spoken to the precise question at issue." Chevron U.S.A. Inc. v. Natural Resources Defense Council, Inc., 467 U.S. 837, 842 (1984).For if after "exhaust[ing] the traditional tools of statutory construction," Natural Resources Defense Council, Inc. v. Browner, 57 F.3d 1122, 1125 (D.C. Cir. 1995) (quoting Chevron, 467 U.S. at 843 n.9), we ascertain that Congress's intent is clear, "that is the end of the matter." Chevron, 467 U.S. at 842. But "if the statute is silent or ambiguous with respect to the specific issue, the question for the court is whether the agency's answer is based on a permissible construction of the statute." Id. at 843. Where judicial review is refracted through this analytic prism, the "view of the agency charged with administering the statute is entitled to considerable deference; and to sustain it, we need not find that it is the only permissible construction that [the agency] might have adopted." Chemical Mfrs. Ass'n v. Natural Resources Defense Council, Inc., 470 U.S. 116, 125 (1985).

Contending that the statutory language boasts an ambiguity, the Secretary maintains that she has reasonably construed paragraph (5)(A)(iv) as prescribing a specific methodology that she must follow when setting outlier thresholds at the beginning of each fiscal year. Under the Secretary's interpretation, paragraph (5)(A)(iv) mandates that she must select outlier thresholds which, when tested against historical data, will likely produce aggregate outlier payments totaling between five and six percent of projected or estimated DR Grelated payments. Because, under the Secretary's construction, paragraph (5)(A)(iv) speaks only to how she is to calculate outlier thresholds for the forthcoming year, the Secretary posits that she has no obligation to ensure that actual outlier

payments for the year total five percent of projected DR Grelated payments.

Advocating a results-oriented approach, however, the Hospitals argue, and the district court agreed, that the Secretary's interpretation contradicts the unambiguous meaning of paragraph (5)(A)(iv). The textual lodestar guiding their plain-meaning critique is the single phrase "payments made."According to the Hospitals, by indicating in paragraph (5)(A)(iv) that "the total amount of the additional payments made ... may not be less than 5 percent" of total DRGrelated payments "estimated or projected to be made," 42 U.S.C. § 1395ww(d)(5)(A)(iv) (Supp. IV 1986) (emphasis added), Congress unmistakably meant that the total amount of additional payments actually made during a fiscal year must meet the five-percent target. During years in which the Secretary's chosen thresholds yield outlier payments that fall short of the statutory mark, the Hospitals' interpretation would require the Secretary to bridge the difference by recalibrating outlier variables and making retroactive payments accordingly.

Standing alone, however, the phrase "payments made" hardly conveys a single meaning, much less the one that the Hospitals advance. As it is employed in paragraph (5)(A)(iv), "payments made" is "simply an adjectival phrase, not a verbial phrase indicating the past tense, and hence allows alternative temporal readings." United States Dep't of the Treasury v. FLRA, 960 F.2d 1068, 1072 (D.C. Cir. 1992). It is not unlike the phrase "recognized as reasonable," which the Supreme Court, quoting favorably from our decision in Administrators of the Tulane Educational Fund v. Shalala, 987 F.2d 790 (D.C. Cir. 1993), held "does not tell us whether Congress means to refer the Secretary to action already taken or to give directions on actions about to be taken."Regions Hosp. v. Shalala, 522 U.S. 448, ___, 118 S. Ct. 909, 916 (1998) (quoting Tulane Educ. Fund, 987 F.2d at 796).Evincing the same syntactical equivalence, the phrase "payments made" in paragraph (5)(A)(iv), though certainly capable of accommodating the Hospitals' interpretation, can just as easily be read to reflect Congress's intent to "give directions on actions about to be taken." Id. In other words, instead of embodying a retrospective inquiry into the amount of outlier payments that have been made, the phrase "payments made under this subparagraph" might just as plausibly reflect a prospective command to the Secretary about how to structure outlier thresholds for payments to be made in advance of each fiscal year. Cf. Regions Hosp., 522 U.S. at ___, 118 S. Ct. at 916 ("[T]he phrase 'recognized as reasonable' might mean costs the Secretary (1) has recognized as reasonable ..., or (2) will recognize as reasonable...."). Ultimately, whether the phrase is "recognized as reasonable," "adversely affected," or "payments made," it is difficult to divine with much confidence the pellucid intent of Congress because "[t]he language, in short, is ambiguous." United States Dep't of the Treasury, 960 F.2d at 1072 (describing as ambiguous the phrase "adversely affected"); accord Tulane Educ. Fund, 987 F.2d at 796.

Hoping to stave off judicial review under Chevron's deferential second step, the Hospitals attempt to resuscitate their plain-meaning interpretation by contrasting the two ways in which Congress modified the word "made" in paragraph (5)(A)(iv). When it first appears, "made" is used without modifiers to describe the "total amount of the additional payments made under this subparagraph"; later, the word materializes to indicate that the total amount of outlier payments just described may not be less than five percent "of the total payments projected or estimated to be made" for DRGrelated payments. 42 U.S.C. § 1395ww(d)(5)(A)(iv). Because, the Hospitals reason, Congress employed words of estimation and projection to modify the total amount of DRGrelated payments "to be made" but neglected to do so when describing the total amount of outlier payments "made," it must have intended that total outlier payments actually made during a fiscal year would equal at least five percent of estimated or projected DRG-related payments.2

Whatever logic this internal construction of paragraph (5)(A)(iv) enjoys, to prevent statutory interpretation from degenerating into an exercise in solipsism, "we must not be guided by a single sentence or member of a sentence, but look to the provisions of the whole law." United States Nat'l Bank v. Independent Ins. Agents of Am., Inc., 508 U.S. 439, 455 (1993) (quoting United States v. Heirs of Boisdore, 49 U.S. (8 How.) 113, 122 (1849)). Under Chevron step one, "we consider not only the language of the particular statutory provision under scrutiny, but also the structure and context of the statutory scheme of which it is a part." Illinois Pub. Tele. Ass'n v. FCC, 117 F.3d 555, 568 (D.C. Cir. ), modified, 123 F.3d 693 (1997), cert. denied, ___ U.S. ___, 118 S. Ct. 1361 (1998); accord Conroy v. Aniskoff, 507 U.S. 511, 515 (1993) ("[T]he meaning of statutory language, plain or not, depends on context."); Davis v. Michigan Dep't of Treasury, 489 U.S. 803, 809 (1989) ("[W]ords of a statute must be read in their context and with a view to their place in the overall statutory scheme."). By examining paragraph (5)(A)(iv) in its immediate statutory context, any putative clarity that that provision might arguably have quickly recedes to ambiguity once again.

Preceding paragraph (5)(A)(iv) by two paragraphs, § 1395ww(d)(3)(B) provided during the time relevant to this litigation:

The Secretary shall reduce each of the average standard-ized amounts under subparagraph (A) ... by a proportion equal to the proportion (estimated by the Secretary)of the amount of payments under this subsection based on DRG prospective payment amounts which are additional payments described in paragraph (5)(A) (relating to outlier payments)....

42 U.S.C. § 1395ww(d)(3)(B) (Supp. IV 1986) (emphasis added). Not only does this provision expressly indicate that total outlier payments are to be estimated by the Secretary, but it also employs language that closely parallels the language that later appears in paragraph (5)(A)(iv). In our endeavor to determine whether the "total amount of the additional payments made under this subparagraph" contemplates outlier payments actually made or those estimated to be made, we find it significant that in paragraph (3)(B) Congress provided that the "amount of payments ... which are additional payments described in paragraph (5)(A)" are to be "estimated by the Secretary." § 1395ww(d)(3)(B). Given that in paragraph (3)(B) it had already indicated that the Secretary would estimate the amount of outlier payments described in subparagraph (5)(A), Congress could have reasonably concluded that there was no need to provide expressly in paragraph (5)(A)(iv) that the phrase "payments made" referred to payments estimated to be made. Thus, whatever can be said for Congress's disparate modification of the word "made" in paragraph (5)(A)(iv), when we open the analytic aperture to examine that clause in its proper statutory context, the inherently ambiguous phrase "payments made" becomes no clearer.

Nor does a passing observation in the Conference Report that "the Secretary would be required to provide additional payments for outlier cases amounting to not less than 5 percent, and not more than 6 percent, of total projected or estimated DRG related payments," compel us to adopt the Hospitals' construction of the statute under Chevron step one. H.R. Conf. Rep. No. 98-47, at 189 (1983), reprinted in 1983 U.S.C.C.A.N. 404, 479 (emphasis added). Ambiguous in its own right, this passage, if given the gloss that the Hospitals advance, would chafe against the commentary in the Senate Report. Suggesting that paragraph (5)(A)(iv) reflects the prospective inquiry that the Secretary advocates, the Senate Report provides that the "[t]otal expected payments resulting from this policy will be not less than 5 percent, nor more than 6 percent, of total medicare payments to hospitals." S. Rep. No. 98-23, at 51, reprinted in 1983 U.S.C.C.A.N. 143, 191 (emphasis added). The only conclusion that we can safely draw from these seemingly contradictory passages is that "the little legislative history that exists for [paragraph (5)(A)(iv)] is as ambiguous as the statute itself." Deaf Smith County Grain Processors, Inc. v. Glickman, 162 F.3d 1206, 1212 (D.C. Cir. 1998).

Ultimately, neither the text, structure, nor legislative history of paragraph (5)(A)(iv) illuminates Congress's unambiguous intent. Although the Hospitals' interpretation, and the one that the district court adopted, is plausible, it is not the "only possible interpretation," Sullivan v. Everhart, 494 U.S. 83, 89 (1990), and it certainly is not "an inevitable one."Regions Hosp., 522 U.S. at ____, 118 S. Ct. at 917. Because we find the statute ambiguous, we proceed to assess whether the Secretary's interpretation of paragraph (5)(A)(iv) is "reasonable and consistent with the statutory scheme and legislative history." Cleveland v. United States Nuclear Regulatory Comm'n, 68 F.3d 1361, 1367 (D.C. Cir. 1995).

In marking off the metes and bounds of our review under the second step of Chevron, we accord particular deference to the Secretary's interpretation of paragraph (5)(A)(iv) "given the tremendous complexity of the Medicare statute." Appalachian Reg'l Healthcare, Inc. v. Shalala, 131 F.3d 1050, 1054 (D.C. Cir. 1997); accord Methodist Hosp. v. Shalala, 38 F.3d 1225, 1229 (D.C. Cir. 1994). The Hospitals, however, urge us not to defer in any way to the Secretary's interpretation of paragraph (5)(A)(iv) because, they contend, that provision does not delegate interpretative authority to the Secretary but explicitly limits her discretion. The problem with this argument, of course, is that it assumes the truth of the proposition that we just rejected. Were paragraph (5)(A)(iv) truly "an explicit limitation on the Secretary's discretion," Br. of Hosps. at 40, there would be no need to analyze the provision under Chevron step two. While paragraph (5)(A)(iv) is certainly designed to regulate the Secretary's discretion to some extent, as we have already concluded, the precise contours of that provision are hardly explicit but are instead ambiguous. Nor is it problematic, as the Hospitals suggest, that Congress did not expressly delegate interpretative authority to the Secretary in paragraph (5)(A)(iv). Deference to agency interpretation is warranted "when Congress has left a gap for the agency to fill pursuant to an express or implied delegation of authority to the agency." Chevron, 467 U.S. at 843-44 (internal quotations omitted). Where, as here, Congress enacts an ambiguous provision within a statute entrusted to the agency's expertise, it has "implicitly delegated to the agency the power to fill those gaps." National Fuel Gas Supply Corp. v. FERC, 811 F.2d 1563, 1569 (D.C. Cir. 1987); see also Chevron, 467 U.S. at 843-44.

Equally untenable is the Hospitals' argument that the Secretary's interpretation is not entitled to deference because she did not adopt it through either formal rule making or adjudication. But as our precedents make clear, "an agency need not promulgate a legislative rule setting forth its interpretation of a statutory term for that term to be entitled to deference." Association of Bituminous Contractors, Inc. v. Apfel, 156 F.3d 1246, 1251-52 (D.C. Cir. 1998). In fact, "[e]ven if the legal briefs contained the first expression of the agency's views, under the appropriate circumstances we would still accord them deference so long as they represented the agency's 'fair and considered judgment on the matter.' "United Seniors Ass'n v. Shalala, 182 F.3d 965, 971 (D.C. Cir. 1999) (quoting Auer v. Robbins, 519 U.S. 452, ____, 117 S. Ct. 905, 912 (1997)); see National Wildlife Fed'n v. Browner, 127 F.3d 1126, 1129 (D.C. Cir. 1997) ("The mere fact that an agency offers its interpretation in the course of litigation does not automatically preclude deference to the agency."); Tax Analysts v. IRS, 117 F.3d 607, 613 (D.C. Cir. 1997).

There is no reason to suspect that the Secretary's interpretation of paragraph (5)(A)(iv) embodies anything other than her fair and considered opinion. In a final rule published on January 3, 1984, the Secretary articulated the same interpretation of paragraph (5)(A)(iv) that she has pressed before both us and the district court. See 49 Fed. Reg. 234, 265 (1984). With that rule, she not only observed that under her interpretation "there is no necessary connection between the amount of estimated outlier payments and the actual payments made to hospitals for cases that actually meet the outlier criteria," id., but she also admonished Medicare providers that, in the event she overestimated the amount of outlier payments, she would "not adjust the DRG rates to compensate hospitals for funds that were not actually paid for outlier cases." Id. at 266. Even if, as the Hospitals complain, the final rule failed to provide a "cogent explanation" of the policies undergirding the Secretary's interpretation, the fact remains that for the past fifteen years, the Secretary has never wavered from that interpretation. We are confident that the interpretation of paragraph (5)(A)(iv) under review embodies the Secretary's "fair and considered judgment on the matter." Auer, 519 U.S. at ____, 117 S. Ct. at 912. It, accordingly, demands deference from the judiciary.

Having settled on the scope of our review, we have no difficulty concluding that the Secretary has advanced a reasonable interpretation of paragraph (5)(A)(iv). Congress established outlier payments because it recognized that "there will be cases within each [DRG] that will be extraordinarily costly to treat ... because of severity of illness or complicating conditions, and [will] not [be] adequately compensated for under the DRG payment methodology." S. Rep. No. 98-23, at 51 (1983), reprinted in 1983 U.S.C.C.A.N. 143, 191. But as the term "outlier" suggests, these payments were not intended to subsidize hospitals simply for treatments, which in absolute terms, were extraordinarily costly or lengthy. Rather, Congress directed the Secretary to provide "additional payments for cases which are extraordinarily costly to treat, relative to other cases within the DRG." Id. (emphasis added); accord H.R. Rep. No. 98-25, at 134-35 (1983), reprinted in 1983 U.S.C.C.A.N. 219, 353-54 ("Your Committee is concerned that under the prospective payment system, there will be cases within each [DRG] that will be extraordinarily costly to treat, relative to other cases within that DRG....").Thus the House version would have required the Secretary to tender outlier payments only when a patient's length of stay exceeded by thirty days the average length of stay for cases in that same DRG, see H.R. Rep. No. 98-25, at 135, reprinted in 1983 U.S.C.C.A.N. at 354, while the Senate bill, which the Conference Committee adopted, authorized outlier payments for cases in which a patient's length of stay eclipsed the mean length of stay for discharges within that same DRG by a fixed number of days or standard deviations. See 42 U.S.C. § 1395ww(d)(5)(A)(i) (Supp. IV 1986); S. Rep. No. 98-23, at 51, reprinted in 1983 U.S.C.C.A.N. at 191; H.R. Conf. Rep. No. 98-47, at 188 (1983), reprinted in 1983 U.S.C.C.A.N. 404, 478.

The Secretary's interpretation of paragraph (5)(A)(iv) evinces far greater fidelity to Congress's conception of outlier payments than does the view espoused by the Hospitals. Under the Secretary's reading of the statute, if it turns out that actual outlier payments do not meet the five-percent target at the end of the fiscal year, it is because the lengths of stay for DRGs in that year proved to be shorter than the historical averages reflected in the data on which the Secretary based her threshold calculations. Whether it is because hospitals became more efficient or a miracle drug was introduced during the year, the shorter lengths of stay mean that there were fewer extraordinarily costly cases during the year. In other words, there were fewer outliers--and therefore, fewer outlier payments needed to be made.

Under the Hospitals' interpretation, however, regardless of actual costs or inpatient lengths of stay during a fiscal year, Medicare providers are guaranteed a substantial and fixed sum of outlier payments. As they read the statute, even during a fiscal year in which the length of stay for every inpatient discharge in every DRG in every hospital equaled or exceeded by just a day the mean length of stay for each respective DRG, the Secretary would nonetheless have to reward hospitals with additional "outlier" payments totaling five percent of the entire DRG budget. One need not be well versed in the discipline of statistics to recognize that such insignificant deviations from the mean do not constitute outliers. To sanction the Hospitals' interpretation of paragraph (5)(A)(iv) would not only require us to assume that Congress did not appreciate the meaning of outlier--a term, it should be noted, that appears throughout both the legislative history and the text of the Medicare statute--but it would also transform the character of the outlier-payment regime from a system intended to insulate hospitals from aberrational and extraordinary costs into nothing more than an entitlement program for Medicare providers. Such was hardly Congress's intent, for if anything, Congress indicated that it was "equally concerned that adjustments may be required for cases which have an unusually short length of stay or which are significantly less costly than the DRG payment." H.R. Conf. Rep. No. 98-47, at 478, reprinted in 1983 U.S.C.C.A.N. at 478 (emphasis added); see also H.R. Rep. No. 98-25, at 135, reprinted in 1983 U.S.C.C.A.N. at 354 ("The Secretary would be required to study ... the appropriateness of, and necessity for, adjustments in payment rates for extremely short lengths of stay within a DRG...."). Proposals like these reflect a reluctance to reimburse Medicare providers at rates grossly disproportionate to the cost of treatment. We find it unlikely that Congress nevertheless would have wanted hospitals to reap additional compensation over and above the standard DRG payment where treatment costs for a particular discharge were not extraordinarily costly relative to the mean costs for that DRG.

Moreover, compared to the Hospitals' interpretation of paragraph (5)(A)(iv), the Secretary's reading better harmonizes each of the four clauses in paragraph (5)(A). As did the district court, the Hospitals struggle to reconcile their conception of the fourth clause with the language of the third, which provides that the amount of each outlier payment "shall approximate the marginal cost of care beyond the cutoff point applicable" for each DRG. 42 U.S.C. § 1395ww(d)(5)(A)(iii) (Supp. IV 1986). Under the Hospitals' construction of the statute, outlier payments might bear little relationship to the marginal cost of care. At the end of each fiscal year, if actual outlier payments fell short of the five-percent target, the Secretary would be required retroactively to supplement those payments to satisfy the difference. Depending on how great that initial disparity was, by the time that these curative outlays were made, the newly computed outlier payments might not approximate anything close to the marginal cost of care as paragraph (5)(A)(iii) mandates. By contrast, outlier payments under the Secretary's interpretation will always approximate the marginal cost of care because when determining where to set outlier thresholds for DRGs at the beginning of each fiscal year, she directly factors the marginal cost of care into her calculus.

Echoing the district court's holding, the Hospitals discount paragraph (5)(A)(iii) as merely a "guideline" while contending that paragraph (5)(A)(iv) operates "as a limitation on the Secretary's discretion." County of Los Angeles, 992 F. Supp. at 32. Based on this view, the Secretary is supposed to set outlier thresholds at the beginning of each year "at marginal cost and then, when actual outlier data is known, adjust[ ] the final payments to ensure that the Secretary has met her statutory obligation to the providers." Id. at 31. Why this parsing of the statutory language is more reasonable than that of the Secretary's--much less compelled as an unambiguously plain reading of the provision, as the Hospitals have urged--is not at all clear. After all, paragraph (5)(A)(iii) employs mandatory language of the sort not normally used if all that Congress intended to do was to offer a discretionary guideline for the Secretary to follow. See § 1395ww(d)(5)(A)(iii) ("The amount of such additional payment under clauses (i) and (ii) shall be determined by the Secretary and shall approximate the marginal cost of care....") (emphasis added). Nevertheless, even were the Hospitals' synthesis of the third clause into the remainder of paragraph (5)(A) plausible, it would not be enough to impugn the otherwise reasonable interpretation that the Secretary has advanced since "we need not find that it is the only permissible construction that [the Secretary] might have adopted but only that [her] understanding of this very 'complex statute' is a sufficiently rational one to preclude a court from substituting its judgment for that of [the Secretary]."Young v. Community Nutrition Inst., 476 U.S. 974, 981 (1986) (internal quotation omitted).

Moreover, the Secretary's interpretation avoids the substantial administrative burden attendant with the Hospitals' vision of paragraph (5)(A)(iv). It strains credulity to assume that Congress would have directed the Secretary to establish outlier thresholds in advance of each fiscal year, see § 1395ww(d)(3)(B), (d)(6), and process millions of bills based on those figures, only to have her at the end of the year recalibrate those calculations, reevaluate anew each of the millions of inpatient discharges under the revised figures, and disburse a second round of payments. As we have held in an analogous context, "[u]nder these circumstances, retroactive corrections would cause a significant, if not debilitating, disruption to the Secretary's administration of the already complex Medicare program." Methodist Hosp., 38 F.3d at 1233. Nor is this administrative process rendered less awkward and unwieldy if, as the Hospitals suggest, the Secretary actively monitors outlier payments and adjusts the thresholds as the fiscal year unfolds. Apart from the tremendous resources that would be required to maintain such a vigilant watch over a program as expansive as Medicare, intermittently modifying outlier thresholds at various times during the year would mean that different hospitals would likely receive different outlier reimbursements for the same treatments based on nothing more than the fortuity of when they treated a patient.

By the same token, this uncertainty and fluidity in outlier payment amounts under the Hospitals' interpretation lead us to the final virtue of the Secretary's construction. One of the touchstones of the Prospective Payment System, as its name suggests, is prospectively determined reimbursement rates that remain constant during the fiscal year. In setting, prior to each fiscal year, fixed outlier thresholds and per-diem reimbursement rates that are not later subject to retroactive correction, the Secretary promotes certainty and predictability of payment for not only hospitals but the federal government--concerns that played a prominent role in Congress's decision to adopt PPS. See H.R. Rep. No. 98-25, at 132, reprinted in 1983 U.S.C.C.A.N. at 351 ("The bill is intended to improve the medicare program's ability to act as a prudent purchaser of services, and to provide predictibility [sic] regarding payment amounts for both the Government and hospitals."). To be sure, we have previously speculated that "the real linchpin of the [PPS] system may not be that the exact reimbursement figure is known in advance, but rather may be that the hospital knows that nothing it does in providing services will lead to a higher reimbursement level."Georgetown Univ. Hosp. v. Bowen, 862 F.2d 323, 330 (D.C. Cir. 1988) (Georgetown II). Yet while we, therefore, have recognized that retroactive corrections may not ultimately undermine PPS, we have emphasized that that "does not establish that a prospective-only policy is unreasonable."Methodist Hosp., 38 F.3d at 1232.

In Methodist Hospital, we found the Secretary's decision not to recalculate retroactively the DRG wage index to be reasonable, in part, because the Secretary's prospective policy advanced the principles of PPS.3 With language applicable to the present case, we held:While retroactive adjustments might leave the "linchpin"of PPS intact, that does not mean that a prospective-only policy would not further, to some degree, the overall goals of PPS.... [H]ospitals, like other businesses, do make projections about future costs and service levels based on their experience and historical patterns. To the extent that the Secretary's prospectivity policy permits hospitals to rely with certainty on one additional element in the PPS calculation rate ... the Secretary could reasonably conclude that it will promote efficient and realistic cost-saving targets.

Id. The same, quite reasonably, can be said of the Secretary's interpretation of paragraph (5)(A)(iv). Under her construction of the statute, at the outset of each fiscal year, hospitals know the point beyond which a patient's length of stay will trigger outlier payments and the corresponding rate at which they will be reimbursed for each day beyond the threshold. A less determinate policy would not only deprive hospitals of the ability to make accurate projections about outlier payments for the forthcoming year but also threaten them at the end of each year with the prospect of actually having to forfeit a portion of those payments to the Secretary; for as the Hospitals concede, under their interpretation, Medicare providers collectively would be bound to repay any portion of outlier payments that exceeded six percent of estimated DRG-related payments. See Br. of Hosps. at 3132. We conclude that the Secretary has advanced a reasonable interpretation of an ambiguous statutory provision, and, therefore, reverse the judgment of the district court with respect to the Secretary's appeal.

Because outlier thresholds are measured by their distance from the mean length of stay, accurately forecasting outlier payments depends, in large part, on how closely the mean length of stay reflected in the Secretary's historical data reflects the actual average length of stay for that particular DRG. When setting outlier thresholds for fiscal years 19851986, the Secretary relied on the 1981 MEDPAR file, a database containing patient-specific data for a random sample of 20 percent of all Medicare-hospital discharges during 1981.Compiled during an era when Medicare still reimbursed hospitals under the reasonable-cost system, the 1981 MEDPAR file could not have predicted how, under PPS, the average length of stay for virtually all DRGs would eventually decline dramatically. The Hospitals observe, however, that by July 27, 1984, the Secretary had already collected data from 2.5 million discharges under PPS that indicated that the average length of stay for all DRGs had declined from 9.5 days to 7.5 days under the new payment methodology. Pursuant to section 10(e) of the APA, 5 U.S.C. § 706(2)(A), the Hospitals claim that the Secretary acted arbitrarily and capriciously when she calculated outlier thresholds for 1985-1986 based on the 1981 MEDPAR file instead of the preliminary 1984 data and failed to explain adequately her decision. Rejecting the Hospitals' claim, the district court agreed with the Secretary that her decision to use the 1981 MEDPAR file over the more contemporaneous but preliminary 1984 data "was a rational choice between two imperfect databases."County of Los Angeles, 992 F. Supp. at 36.

Under the APA, we may set aside agency action found to be "arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law." 5 U.S.C. § 706(2)(A). Foreclosed from substituting our judgment for that of the agency, we do not set aside agency action lightly. See Motor Vehicles Mfrs. Ass'n v. State Farm Mut. Auto. Ins. Co., 463 U.S. 29, 43 (1983); Petroleum Communications, Inc. v. FCC, 22 F.3d 1164, 1172 (D.C. Cir. 1994). Nevertheless, we intervene to ensure that the agency has "examine[d] the relevant data and articulate[d] a satisfactory explanation for its action." State Farm, 463 U.S. at 43. "Where the agency has failed to provide a reasoned explanation, or where the record belies the agency's conclusion, we must undo its action." BellSouth Corp. v. FCC, 162 F.3d 1215, 1222 (D.C. Cir. 1999) (citation and quotation omitted).

The only contemporaneous explanation that the Secretary offered for using the 1981 MEDPAR file consisted of two sentences in the Federal Register: "Based upon outlier and DRG payment data received through July 27, 1984, there is no evidence to suggest that total outlier payments are below the levels intended. Therefore, as discussed above, we are continuing to set the outlier thresholds on the basis of the 1981 MEDPAR data." 49 Fed. Reg. 34,728, 34,769 (1984) (emphasis added). We agree with the Ninth Circuit, which recently considered this same issue, that the Secretary's "explanation that there was no evidence of an outlier shortfall was simply not supported by the record before her and did not explain her failure to use the more recent data." Alvarado Community Hosp. v. Shalala, 155 F.3d 1115, 1122 (9th Cir. 1998). Data that the Secretary possessed as late as July 27, 1984, indicated that the average length of stay for practically all DRGs had declined considerably under the nascent PPS program. More concretely, at that point during the fiscal year, outliers constituted only 1.9 percent of total PPS discharges instead of 5.0 percent as predicted. And while these conclusions were drawn from preliminary data, that data reflected 2.5 million patient discharges under PPS; the 1981 MEDPAR file, by contrast, contained 1.6 million discharge records. Failure to account for this trend is all the more perplexing insofar as the Secretary herself had anticipated that the average length of stay for DRGs would decline under PPS. In 1984 she observed that "[t]he most commonly accepted expectation about the PPS at the time of its inception was that it would result in shorter stays for Medicare patients.... [R]educed length of stay was to be one of the major vehicles through which hospital costs were to be controlled under the PPS." Office of Research & Demonstrations, Health Care Fin. Admin., U.S. Dep't of Health & Human Servs., Pub. No. 03231, Report to Congress: Impact of the Medicare Hospital Prospective Payment System 6-13 (1984). At bottom, for the Secretary to say that she had "no evidence to suggest that total outlier payments [were] below the levels intended," 49 Fed. Reg. at 34,769, runs "counter to the evidence before the agency" and "is so implausible that it could not be ascribed to a difference in view or the product of agency expertise." State Farm, 463 U.S. at 43.

In her brief, the Secretary now contends that what she meant by "no evidence" was "no reliable evidence." To bolster this specific claim and her broader argument that the 1984 data were too suspect and incomplete to make accurate outlier projections, the Secretary appended to her summary judgment motion in the district court an affidavit from Rose Connerton, an official with the Health Care Financing Administration ("HCFA") who helped develop the outlier thresholds for 1985-1986. Essentially, the Connerton affidavit claims that the 1984 data were not complete and did not represent a random sample of cases, that because they were based on a partial year they would not reflect seasonal and regional variances, and that any analysis drawn from them would be skewed.4 See J.A. 90 (Aff. of Connerton pp 10, 12, 15). The Hospitals contend that the Connerton affidavit, having surfaced for the first time during litigation, is an impermissible post-hoc rationalization that the district court should have stricken. See SEC v. Chenery Corp., 318 U.S. 80, 87-88 (1943); Reeve Aleutian Airways, Inc. v. United States, 889 F.2d 1139, 1144 (D.C. Cir. 1989). Indeed, faced with a similar affidavit from Connerton, the Ninth Circuit held that the district court erred in considering it. See Alvarado Community Hosp., 155 F.3d at 1124. Ultimately, we need not reach this question, for even were we inclined to accept everything in the Connerton affidavit, we would still remand to the Secretary for a more adequate justification for her database selection.

"A long line of precedent has established that an agency action is arbitrary when the agency offer[s] insufficient reasons for treating similar situations differently." Transactive Corp. v. United States, 91 F.3d 232, 237 (D.C. Cir. 1996); see also State Farm, 463 U.S. 29, 57 (1983); Airmark Corp. v. FAA, 758 F.2d 685, 691-92 (D.C. Cir. 1985); Local 777, Democratic Union Org. Comm. v. NLRB, 603 F.2d 862, 872 (D.C. Cir. 1978). Although maligning the 1984 data as too unreliable to calculate outlier thresholds for fiscal years 19851986, the Secretary nonetheless used those same data on August 31, 1984, to reduce across-the-board all 470 DRG weighting factors by 1.05 percent. See 49 Fed. Reg. 34,728, 34,770-71 (1984). Such an adjustment was necessary, the Secretary noted at the time, because "[t]he emerging experience under the prospective payment system"--an experience gleaned from the preliminary 1984 data--revealed that the different incentives that hospitals faced under PPS were producing unexpected distortions. See id. In making this correction, the Secretary expressly endorsed the reliability of the 1984 data: "To date, we have now analyzed 2.5 million discharges under the prospective payment system, which fully reflect the effect of those incentives, and we believe this affords us a better measure of the effect of coding improvements in the average case mix." Id. at 34,771. Moreover, in responding to a question about the legitimacy of the preliminary data during a 1984 congressional oversight hearing, the HCFA Administrator responded that "[o]ur sample now is based on approximately 50 percent of all of the claims or the admissions that we had projected for this year. We think that's a fairly representative sample." Adjustments in Medicare's Prospective Payment System: Hearing Before the Subcomm. on Health of the Comm. on Fin., 98th Cong. 62 (statement of Mr. Davis, Administrator, HCFA). In sum, the Secretary has inadequately explained why the 1984 data were suitable for one significant calculation but unreliable for another. Her sole justification is that preliminary data may be used to make across-the-board adjustments, as was done to reduce all DRG weighting factors by 1.05 percent, but that they may not be used for setting outlier thresholds because a unique standard deviation must be calculated for each of the 470 DRGs. What renders this explanation inadequate is that DRG weighting factors, like outlier thresholds, are ordinarily determined on a DRG-by-DRG basis. Indeed, the very purpose of a DRG weighting factor is to reflect the different costs of treating minor and major illnesses; to do so, each DRG must be assigned its own unique weight based on the cost and complexity of treatment peculiar to that DRG. The Secretary's proffered distinction is thus not reasonable. She may in her discretion, of course, rely on preliminary data to make an across-the-board adjustment to variables that ordinarily are determined on a case-by-case basis. But when she does so, she must be prepared to explain why she cannot also use that data to make a similar adjustment to variables that are also typically calculated on an individual basis. As broad as her discretion is, it "is not a license to ... treat like cases differently." Airmark Corp., 758 F.2d at 691; accord Teva Pharms., USA, Inc. v. FDA, 182 F.3d 1003, 1010-11 (D.C. Cir. 1999); Transactive Corp., 91 F.3d at 237; Local 777, 603 F.2d at 872.

This case must therefore be remanded to the Secretary to allow her either to recalculate outlier thresholds for fiscal years 1985-1986 or to offer a reasonable explanation for refusing to use the 1984 data in setting outlier thresholds during those years. In reaching this conclusion, we necessarily part ways with the Ninth Circuit, which, in Alvarado Community Hospital, chose not to remand to the Secretary, but instead ordered her to adjust outlier thresholds for fiscal year 1985 based on final 1984 data. Alvarado Community Hosp., 155 F.3d at 1125. As the Supreme Court has instructed, however, where "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." Florida Power & Light Co. v. Lorion, 470 U.S. 729, 744 (1985); see also Dunlop v. Bachowski, 421 U.S. 560, 574-75 (1975) ("Where the statement inadequately discloses his reasons, the Secretary may be afforded opportunity to supplement his statement."), overruled on other grounds by Furniture & Piano Moving v. Crowley, 467 U.S. 526, 549-50 n.22 (1984). We find no reason to depart from that course here. While we have identified significant inconsistencies and gaps in the Secretary's rationale for using the 1981 MEDPAR file, bedrock principles of administrative law preclude us from declaring definitively that her decision was arbitrary and capricious without first affording her an opportunity to articulate, if possible, a better explanation. See Bechtel v. FCC, 10 F.3d 875, 887 (D.C. Cir. 1993);Philadelphia Gas Works v. FERC, 989 F.2d 1246, 1251 (D.C. Cir. 1993); Sullivan Indus. v. NLRB, 957 F.2d 890, 905 n.12 (D.C. Cir. 1992); Tex Tin Corp. v. EPA, 935 F.2d 1321, 1324 (D.C. Cir. 1991); see also Checkosky v. SEC, 23 F.3d 452, 463 (D.C. Cir. 1994) (Silberman, J., concurring) (citing some of the "many instances where we have remanded to an agency for a better explanation before finally deciding that the agency's action was arbitrary and capricious"). Because we fail to perceive any "rare circumstances" that would warrant a break with established administrative practice, we adhere to the proper course of remanding this matter to the Secretary.

For the foregoing reasons, we reverse the judgment of the district court with respect to the Secretary's appeal, and remand with instructions to enter judgment in the Secretary's favor. As for the Hospitals' cross-appeal, we reverse the judgment of the district court, and instruct it to remand the case to the Secretary for further proceedings consistent with this opinion.

So ordered.