Seventy-six hospitals, plaintiffs-appellants here, challenge the Department of Health and Humans Services' ("HHS") implementation of a statutory requirement that the agency adjust hospitals' reimbursements for the costs of administering care to Medicare recipients to reflect "differences in hospital wage levels" across "geographic area[s]." 42 U.S.C. § 1395ww(d)(3)(E)(i).

For more than two decades, HHS has divided the nation into geographic areas for these purposes by adopting the Metropolitan Statistical Areas ("MSAs") formulated by the Office of Management and Budget ("OMB"). Most recently, in 2004, it adopted the version of the MSAs released by OMB in 2003. Compared with previous iterations, the New York City MSA was slightly expanded and now includes certain additional hospitals in northern New Jersey. Because the New Jersey hospitals' wages are somewhat lower, the average wage level in the MSA dropped, along with the wage adjustment for hospitals in that MSA. Plaintiffs allege they will receive $812 million less in reimbursements over the next ten years than they would have under their former wage adjustment.

Plaintiffs argue both that the use of MSAs as proxies for "geographic areas" is an unreasonable interpretation of the Medicare Act and that the agency took improper considerations into account in adopting the MSAs. Applying the analysis required by Chevron U.S.A. Inc. v. Natural Resources Defense Council, Inc., 467 U.S. 837, 104 S. Ct. 2778, 81 L. Ed. 2d 694 (1984), we hold that the term "geographic area" is ambiguous and that the MSAs are a reasonable gap-filler. In addition, we find that the defendant-appellee did not act arbitrarily and capriciously in its 2004 adoption of the latest MSAs. With respect to this issue, the trial court is affirmed in full.

Plaintiffs also challenge defendant's decision in the same 2004 rulemaking to apply at only ten-percent effectiveness in its first year a new reimbursement adjustment that controls for hospitals' decisions to hire more or fewer skilled professionals, on the ground that the data the agency had collected were not of sufficient quality to merit full implementation. We agree with the district court that the agency lacked the statutory authority to do so and also find that the agency acted arbitrarily and capriciously. However, we modify the district court's remedy, which was to order the agency to immediately apply the adjustment in full. Instead, we order the agency to apply the adjustment in full by September 30, 2006, and to have completed all data collection and measurement and any other steps necessary to do so by then. With respect to this issue, the judgment of the trial court is affirmed in part and modified in part.

The Medicare program, established by Title XVIII of the Social Security Act, 42 U.S.C. § 1395 et seq., pays for covered medical services provided to eligible aged and disabled persons. Of relevance to this case, it reimburses hospitals for the cost of serving Medicare beneficiaries. See 42 U.S.C. § 1395f. The Centers for Medicare and Medicaid Services ("CMS") is the agency within the Department of Health and Human Services ("HHS") responsible for administering the Medicare program.2

From the inception of Medicare in 1965 until 1983, hospitals were reimbursed for their actual costs in treating beneficiaries, so long as those costs were reasonable. In 1983, Congress overhauled the reimbursement system, switching to what is known as the Inpatient Prospective Payment System ("IPPS"). See Social Security Amendments of 1983, Pub.L. No. 98-21, 97 Stat. 65 (1983). Under the IPPS, hospitals are not reimbursed for their actual costs, but are instead paid fixed rates for providing specific categories of treatment, known as "diagnosis related groups," or "DRGs." See 42 U.S.C. § 1395ww(d). Separate DRG rates are set for hospitals in urban and rural areas. Id. § 1395ww(d)(2)(G) & (3). The purpose of this switch was to "encourage health care providers to improve efficiency and reduce operating costs." Methodist Hosp. of Sacramento v. Shalala, 38 F.3d 1225, 1227 (D.C.Cir.1994); see also H.R. Conf. Rep. No. 98-25, at 132 (1983), reprinted in 1983 U.S.C.C.A.N. 219, 351.

Of particular significance for this case, the Secretary must adjust DRG payment rates for the relative labor costs in each hospital's geographic area. Accordingly, the base DRG payment rate is divided into two portions: the labor-related costs, which get adjusted for these geographic differences, and the non-labor-related costs, which do not. While the relative proportions of these two cost sources formerly were "estimated by the Secretary from time to time," see 42 U.S.C. § 1395ww(d)(3)(E)(i), for discharges occurring on or after October 1, 2004, Congress has removed the Secretary's discretion and set the labor-cost proportion at sixty-two percent of the base DRG payment. Id. § 1395ww(d)(3)(E)(ii). In relevant part, the Secretary is required to

adjust the [labor-related costs] of the DRG prospective payment rates . . . for area differences in hospital wage levels by a factor (established by the Secretary) reflecting the relative hospital wage level in the geographic area of the hospital compared to the national average hospital wage level. . . . [A]t least every 12 months[,] the Secretary shall update the factor under the preceding sentence on the basis of a survey conducted by the Secretary (and updated as appropriate) of the wages and wage-related costs of [covered] hospitals in the United States . . . . Any adjustments or updates made under this subparagraph for a fiscal year . . . shall be made in a manner that assures that the aggregate payments under this subsection in the fiscal year are not greater or less than those that would have been made in the year without such adjustment. . . .

42 U.S.C. § 1395ww(d)(3)(E)(i).

In other words, the Secretary must, at least once annually, compute a wage factor for each hospital "reflecting" the relative wage level in that hospital's "geographic area," and then apply that factor to the sixty-two percent of the DRG base rate that is attributable to labor costs. These adjustments must be cost neutral, so that any increase in one hospital's wage factor must be offset by a decrease in another's.

From its initial implementation of this law in 1985 through the present, CMS has consistently grouped hospitals into geographic areas by adopting the Metropolitan Statistical Areas ("MSAs") developed by the Office of Management and Budget ("OMB") for use throughout the federal government. It also uses MSAs for other aspects of the reimbursement formula, such as to divide hospitals into urban and rural regions (i.e., hospitals that are part of MSAs and those that are not). See 42 C.F.R. § 412.64(b)(1)(ii)(A) (defining "urban area" as "[a] Metropolitan Statistical Area, as defined by the Executive Office of Management and Budget"). Since the beginning, CMS has acknowledged that MSAs, which were not designed for this specific purpose, are an imperfect proxy for labor markets, particularly with respect to hospitals in rural areas. It has promised to consider alternative methodologies that are based on "objective criteria that will provide more equitable labor market area definitions than the current MSA/non-MSA classifications." See Final Rule: Changes to the Inpatient Hospital Prospective Payment System and Fiscal Year 1986 Rates, 50 Fed.Reg. 35,646, 35,684 (Sept. 3, 1985).

Among the many alternatives that have been proposed in the last two decades are various ways to divide MSAs into their "core" areas, where wages are higher, and "ring" areas.3 CMS has repeatedly rejected these proposals because they lacked objective, easily administrable criteria that could distinguish "core" areas from "ring" areas, and thus implementing such a division would result in equally arbitrary line-drawing. See, e.g., Proposed Rule: Changes to the Inpatient Hospital Prospective Payment System and Fiscal Year 1988 Rates, 52 Fed.Reg. 22,080, 22,094 (June 10, 1987). In doing so, CMS also made assertions in several rulemakings during the 1980s to the effect that urban hospitals, which presumably would be the beneficiaries of a core-ring division, were already doing well enough under the IPPS. See, e.g., id. By 1995, CMS had rejected the core-ring proposal as well as assorted other proposed modifications. It concluded that "there is no clear `best' labor market area option" that would be obviously superior to the MSA system. See Proposed Rule: Changes to the Hospital Inpatient Prospective Payment Systems and Fiscal Year 1996 Rates, 60 Fed.Reg. 29,202, 29,220 (June 2, 1995). Because the industry itself could reach "no consensus" (unsurprisingly, since any modification to a cost-neutral system means much of the industry loses money), and because CMS was less than captivated by any of the alternatives, it decided to simply stay with the MSA system. Id. at 29,219-20. Since then, CMS has seemed to give less serious consideration to adopting these alternatives, which are periodically mentioned during rulemakings or in reports, simply adopting by reference its prior findings. See, e.g., Final Rule: Changes to the Hospital Inpatient Prospective Payment Systems and Fiscal Year 2005 Rates, 69 Fed. Reg. 48,916, 49,027-28 (Aug. 11, 2004) ("2004 Final Rule").

Although well aware of this controversy,4 Congress has never directed CMS to implement any methodology for dividing core from ring or otherwise deviate in any fundamental way from the MSA system. Instead, Congress has enacted a series of exceptions by which hospitals particularly aggrieved by MSA cut-offs can get some relief by, for example, relocating into other MSAs or having dramatic changes to their wage factors phased in over a period of time. See, e.g., 42 U.S.C. § 1395ww(d)(8)(A) (hospitals formerly classified in an MSA but suddenly "rural," which presumably will have severe reductions in reimbursement as a result, have new wage factors phased in over three years); id. § 1395ww(d)(10) (hospital can petition review board for reclassification if it meets certain criteria established by CMS); id. § 1395ww(d)(8)(B) (rural hospital can get benefit of wage index of adjoining MSA if it shows a certain amount of commuting to and from the MSA).

On December 27, 2000, OMB announced various changes to its methodology for computing MSAs. Of significance for this appeal, it simplified its standards for inclusion of outlying counties, basing inclusion entirely on commuting patterns. Under the new methodology, outlying counties are included if at least 25 percent of their residents work in the core county or at least 25 percent of their own jobs are filled by workers commuting from the core county.5 See Notice of Decision: Standards for Defining Metropolitan and Micropolitan Statistical Areas, 65 Fed.Reg. 82,228, 82,233 (Dec. 27, 2000). OMB stated that, while it was aware that its MSA definitions were used in the administration of a variety of programs, in order to preserve the integrity of its statistical analysis from political influence, it specifically did not take the needs of those programs into account. Accordingly, it warned, the definitions "should not be used to develop and implement. . . nonstatistical programs and policies without full consideration of the effects of using these definitions for such purposes," because they "may or may not be suitable for use in program funding formulas." Id. at 82,228. In particular, while the MSAs are well designed to reflect economic ties between communities, "as measured by commuting," they are not intended for use as a proxy for the urban/rural distinction. Id. at 82,228-29.

On June 6, 2003, OMB published its revised list of MSAs, incorporating information from the 2000 Census and using its new methodology. See OMB Bull. No. 03-04, available at http://www.whitehouse.gov/omb/bulletins/b03-04.html. On that list were 370 MSAs, 49 of them new. Id. at 3. Most previously existing MSAs became smaller. However, the old New York City MSA, which already had included the outlying New York counties of Westchester, Putnam, and Rockland, was expanded to include the New Jersey counties of Bergen, Passaic, and Hudson. Id. at 41. It is unclear whether this occurred as a result of the new methodology or whether it would have occurred in any case due to more recent Census data.

On May 18, 2004, CMS proposed to adopt OMB's new MSAs. See Proposed Changes to the Hospital Inpatient Prospective Payment Systems and Fiscal Year 2005 Rates, 69 Fed.Reg. 28,196 (May 18, 2004). On July 12, 2004, plaintiffs' trade association filed comments in opposition, arguing that hospitals in the New Jersey counties had much lower wages ("only" 117 percent of the national average) and should not be included in the New York City MSA, since their inclusion would trim plaintiffs' wage index from 136 percent of the national average to 133 percent, cutting into plaintiffs' reimbursements at a time when New York City hospitals were struggling financially for other reasons.6 For FY 2005, the association proposed that CMS allow at least its members, and possibly other "similarly situated" hospitals, to have the benefit of their former wage indices. For subsequent years, the association proposed that CMS adopt a scheme that would incorporate both old and new MSAs, treating as partial members of an MSA those counties which were included in either the old version or the new version but not both.

On August 11, 2004, CMS adopted OMB's new MSAs for purposes of the hospital wage index.7 See 2004 Final Rule at 49,026. It noted OMB's cautionary statement about using these definitions for non-statistical purposes, see id. at 49,027, but determined that no proposed alternative was better. CMS observed that commenters had proposed completely inconsistent alternatives, with some emphasizing "expanding existing MSAs" and others calling for "creating smaller units or at least distinguishing segments within larger MSAs." Id. at 49,028. It added that, while many commenters had demonstrated that their proposals would better serve their "specific situations," it could not adopt any alternative without assessing "all of the effects that these proposed revisions might have." Id. In particular, CMS reminded the commenters that it was working within a zero-sum framework, so that, while some proposals might help those hospitals that would otherwise receive lower wage indices under the new MSAs, "hospitals that stand to benefit from the new definitions might experience lesser gains from the proposed revisions." Id. Nonetheless, recognizing the "scope and drastic implications of these new boundaries," and seeking "to buffer the subsequent negative impact on numerous hospitals," CMS decided that, for FY 2005 only, any hospital whose wage index decreased because of the MSA changes would receive a blended index based 50 percent on the newly calculated index and 50 percent on its old index. Id. In addition, because the most significant changes were felt by hospitals formerly part of MSAs and now considered rural,8 CMS permitted those hospitals to retain their former MSA classifications for three years. Id. at 49,033.

In 2000, Congress directed CMS to refine its survey of hospital wages in geographic areas by controlling for differences in hospitals' occupational mixes.9 See Medicare, Medicaid, and SCHIP Benefits Improvement and Protection Act of 2000, Pub.L. No. 106-554, § 304(c), 114 Stat. 2763, 2763A-495 ("2000 Bill"). Specifically, Congress instructed the Secretary of HHS to "provide for the collection of data every 3 years on occupational mix for employees of each [covered] hospital . . . in the provision of inpatient hospital services, in order to construct an occupational mix adjustment in the hospital area wage index."10 Id. § 304(c)(1). As codified, CMS's instruction is to "measure the earnings and paid hours of employment by occupational category and [to] exclude data with respect to the wages and wage-related costs incurred in furnishing skilled nursing facility services." 42 U.S.C. § 1395ww(d)(3)(E)(i). In uncodified language that is a subject of the instant controversy, Congress added:

By not later than September 30, 2003, for application beginning October 1, 2004, the Secretary shall first complete (A) the collection of data [on occupational mix]; and (B) the measurement [of earnings and paid hours of employment by occupational category].

2000 Bill, § 304(c)(3).

By the summer of 2001, CMS had promulgated a final rule as to how it would collect these data and expressed its intention to survey hospitals during the 2002 calendar year. See Final Rules: Changes to the Hospital Inpatient Prospective Payment Systems and Rates and Costs of Graduate Medical Education, 66 Fed.Reg. 39,828, 39,860-39,862 (Aug. 1, 2001). Nonetheless, for reasons that are unclear, not until September 19, 2003 did CMS publish a final notice of intent to collect data. See Agency Information Collection Activities, 68 Fed.Reg. 54,905-01 (Sept. 19, 2003). Data collection began in January 2004 and was completed in April 2004. See 2004 Final Rule at 49,034.

In the final rule being challenged here, CMS stated that it lacked full confidence in its data for several reasons, including: (1) Hospitals were surveyed for a short (and possibly unrepresentative) period of time, whereas in future years CMS hoped to chart the occupational mix over the full year. Id. at 49,036. (2) There was no time to audit the data. Id. at 49,037. (3) Because this was the first time the data were collected, CMS had no baseline against which to evaluate them for potential errors. Id. (4) The results were surprising; the occupational mix adjustment was expected to benefit rural hospitals, but instead one-third of rural hospitals would see their wage indices decrease while "several large academic medical centers" ended up benefitting, a result that was "counterintuitive" and that some commenters believed was due "to errors in the data." Id. at 49,047-48. In light of its lack of confidence in its data, CMS decided to apply the occupational mix adjustment to only ten percent of the wage index for FY 2005, an action that was supported by a "majority of commenters." Id. at 49,052.

The following year, rather than conduct a new survey (an action which, it noted, the statute only compels it to take once every three years), CMS used largely the same data. See Final Rule: Changes to the Hospital Inpatient Prospective Payment Systems and Fiscal Year 2006 Rates, 70 Fed.Reg. 47,278, 47,365 (Aug. 12, 2005). While all hospitals were permitted to update their submissions, they were not required to do so, and the vast majority did not. Id. at 47,376. Therefore, because CMS had the same concerns about the robustness of its data, it continued to apply the occupational mix adjustment to only ten percent of the wage index. Id.

Two months later, CMS proposed to collect data during 2006, for use in constructing an occupational mix adjustment for FY 2008. See Proposed Collection, 70 Fed. Reg. 60,092 (Oct. 14, 2005). Until then, CMS presumably intends to continue using the same data, and following the same policy of applying the adjustment to only ten percent of the wage index.

On November 1, 2004, plaintiffs filed this action in the Southern District of New York pursuant to the Administrative Procedure Act ("APA"), 5 U.S.C. § 701 et seq., as well as the judicial review provision of the Medicare Act, see 42 U.S.C. § 1395oo(f)(1), challenging CMS's adoption of the new MSAs and its decision to implement the occupational mix adjustment at only ten-percent effectiveness. On March 1, 2005, in a thoughtful opinion, the district court (McKenna, J.) granted defendants summary judgment with respect to the "geographic areas" issue, finding that this term was ambiguous and that the use of MSAs to fill that gap was reasonable. It concluded that, because the statutory scheme requires hospitals' costs to be averaged, hospitals with somewhat dissimilar costs will inevitably be lumped together, and plaintiffs had made no showing that the level of "inaccuracy" created by this particular grouping went beyond what the statute permits.

The district court then granted plaintiffs summary judgment with respect to the occupational mix adjustment issue, finding that Congress had unambiguously mandated that the agency implement the adjustment in full by October 1, 2004. This appeal followed, with defendant cross-appealing with respect to the occupational mix adjustment.

At the outset, we describe the scope of our review. On appeal from a grant of summary judgment in a challenge to agency action under the APA, we review the administrative record and the district court's decision de novo. City of New York v. Shalala, 34 F.3d 1161, 1166 (2d Cir.1994).

With respect to each challenged action, we begin by reviewing the agency's construction of the statute at issue. We do so by applying the familiar two-step process of statutory interpretation set forth in Chevron U.S.A. Inc. v. Natural Resources Defense Council, Inc., 467 U.S. 837, 104 S. Ct. 2778, 81 L. Ed. 2d 694 (1984). Under Chevron, the first question is "whether Congress has directly spoken to the precise question at issue"; if so, our inquiry is at in end. Id. at 842-43, 104 S. Ct. 2778. If there is silence or ambiguity in the statute on the question, then the agency has discretion in its implementation, and we ask only if the construction it has given the statute is reasonable.11 Id. at 843, 104 S. Ct. 2778.

Assuming the agency's action was authorized by statute, we then ask whether it was "arbitrary, capricious, [or] an abuse of discretion," 5 U.S.C. § 706(2)(A), or "unsupported by substantial evidence," id. § 706(2)(E). Such a finding, which is required to overturn an agency action, can be made only where the agency "has relied on factors which Congress has not intended it to consider, entirely failed to consider an important aspect of the problem, offered an explanation for its decision that runs counter to the evidence before the agency, or is so implausible that it could not be ascribed to a difference in view or the product of agency expertise." Motor Vehicle Mfrs. Ass'n of the U.S., Inc. v. State Farm Mut. Auto. Ins. Co., 463 U.S. 29, 43, 103 S. Ct. 2856, 77 L. Ed. 2d 443 (1983); see also City of New York, 34 F.3d at 1167.

Our task, then, is limited. We have no license to substitute our policy judgment for that of the agency, but only to overturn actions that are not authorized by statute or that are arbitrary or capricious. See Citizens to Preserve Overton Park, Inc. v. Volpe, 401 U.S. 402, 416, 91 S. Ct. 814, 28 L. Ed. 2d 136 (1971).

We first review the defendant's use of OMB's MSAs. In relevant part, the Medicare statute directs defendant to:

adjust the proportion, (as estimated by the Secretary from time to time) of hospitals' costs which are attributable to wages and wage-related costs . . . for area differences in hospital wage levels by a factor (established by the Secretary) reflecting the relative hospital wage level in the geographic area of the hospital compared to the national average hospital wage level.

42 U.S.C. § 1395ww(d)(3)(E)(i).

CMS's task is unambiguous: to calculate a factor that reflects geographic-area wage-level differences, and nothing else. We reject defendant's contention that this provision, or any other in the Medicare Act, confers upon him the discretion to take into account all sorts of unrelated policy considerations, such as whether certain hospitals receive unwarranted advantages from other provisions of the Medicare reimbursement scheme.12

At the same time, as plaintiffs conceded at argument, the statute leaves considerable ambiguity as to the term "geographic area," which, based only on the literal language of the provision, could be as large as a several-state region or as small as a city block. CMS's discretion in interpreting this ambiguous term is cabined by the need to fulfill two somewhat contradictory policies expressed by the text of the provision and the legislative history of the IPPS: (1) the geographic areas must be small enough to actually reflect differences in wage levels and, (2) each geographic area must include enough hospitals that their costs can be meaningfully averaged and individual hospitals do not get reimbursed for their own actual costs. In balancing these two considerations, the agency has considerable discretion. Moreover, even after determining the scale of each geographic area, lines must be drawn between areas that inevitably will be contested and may seem arbitrary; once again, the statute is silent as to how this process is to take place, leaving the agency with broad discretion.

There is no question that MSAs are, literally, "geographic areas," and thus their use complies with the language of the statute. Furthermore, their use comports with the two purposes set out above. Because MSAs are based on commuting patterns into and out of the central county, hospitals in each MSA presumably compete in the same labor market, and so it is likely that their wages bear at least rough similarity. On the other hand, each MSA provides a large enough pool of hospitals to allow cost averaging. We conclude that the use of MSAs to fill the gap left by the ambiguous term "geographic areas" is reasonable. In doing so, we express no opinion as to whether any alternative interpretation would have been "better," as we are not empowered to set aside a reasonable interpretation on that basis. See Chevron, 467 U.S. at 843 n. 11, 104 S. Ct. 2778; see also San Bernardino Mountains Cmty. Hosp. Dist. v. Sec'y of HHS, 63 F.3d 882, 889 (9th Cir.1995).

While this appears to be the first published opinion involving the use of MSAs to give definition to the statutory term "geographic area," we note that two of our sister circuits have upheld the defendant's use of MSAs to determine which hospitals are in urban or rural areas. See San Bernardino, 63 F.3d at 889; Marshall County Health Care Auth. v. Shalala, 988 F.2d 1221, 1227 (D.C.Cir.1993). San Bernardino concluded: "[A]lthough the definition of a Metropolitan Statistical Area was not designed and is not perfectly tailored for determining which hospitals should qualify, . . . it is an established and uniform standard that generally coincides with the definition of an urban area." 63 F.3d at 889. Similarly, we find that an MSA, while not designed for this statutory scheme and inevitably not "perfectly tailored" for it, generally coincides with the statutory purposes and is thus a reasonable proxy for the "geographic area" that the statute charges CMS with defining.

Finally, we observe that CMS has now used MSAs to fill this statutory gap for more than two decades without any action from Congress suggesting disapproval. The fact that defendant has now adopted an MSA-based wage index multiple times certainly counsels deference to its decision to do so this time. Cf. Good Samaritan Hosp. v. Shalala, 508 U.S. 402, 417, 113 S. Ct. 2151, 124 L. Ed. 2d 368 (1993). In addition, this is a clear instance of Congress acquiescing in the agency's interpretation. Congress's inaction cannot be ascribed to lack of interest or knowledge, since it has received multiple reports on the issue and has frequently tinkered with the statutory scheme in question. "[O]nce an agency's statutory construction has been fully brought to the attention of the public and the Congress, and the latter has not sought to alter that interpretation although it has amended the statute in other respects, then presumably the legislative intent has been correctly discerned." United States v. Rutherford, 442 U.S. 544, 554 n. 10, 99 S. Ct. 2470, 61 L. Ed. 2d 68 (1979) (internal quotation marks omitted).

Having concluded that the Medicare statute authorizes an MSA's use as a proxy for a "geographic area," we now ask whether this policy choice was arbitrary or capricious. Before undertaking this inquiry, we pause to note that an agency's burden of supplying a "reasoned analysis" justifying its policy is lower where, as here, an agency is continuing a long-standing policy compared to where the agency is suddenly changing that policy.13 See State Farm, 463 U.S. at 42, 103 S. Ct. 2856; see also Atchison, Topeka & Santa Fe Ry. Co. v. Wichita Bd. of Trade, 412 U.S. 800, 808, 93 S. Ct. 2367, 37 L. Ed. 2d 350 (1973) ("There is . . . at least a presumption that [the agency's mandate from Congress] will be carried out best if the settled rule is adhered to."); Torrington Extend-A-Care Employee Ass'n v. NLRB, 17 F.3d 580, 589 (2d Cir.1994) ("When an agency has committed itself to a settled course of behavior, a presumption in favor of that course arises."). For this reason, we reject at the outset plaintiffs' arguments based on defendant's decision not to extensively study the suitability of MSAs for this regulatory scheme in its 2004 Final Rule. An agency reaffirming its long-standing policy need not analyze all aspects of that policy as if adopting it for the first time, but rather must only consider specific objections made to its continuation of that policy.14

Similarly, we reject plaintiffs' argument that defendant's past consideration of improper factors, without more, renders its decision to adopt the 2004 Final Rule arbitrary and capricious. We are not reviewing those earlier rulemakings, but rather the 2004 Final Rule. The use of MSAs has become a settled agency policy that is entitled to a presumption of regularity.15 That is not to say that the agency's reasoning in continuing that policy has been immunized from review; rather, it means that plaintiffs have the burden of demonstrating that the agency took illegitimate factors into account in formulating policy and rejecting proposed alternatives in its 2004 rulemaking. The fact that the agency has acted improperly in the past is relevant to our inquiry, but plaintiffs still must show that the agency continued to do so.

Plaintiffs have made only two allegations that are cognizable within this framework. First, they claim that the agency showed favoritism to rural hospitals, in that it: (1) did not adopt a portion of OMB's new methodology (the introduction of Micropolitan Statistical Areas) that would affect only rural hospitals, whereas it adopted without modification the new MSAs, which apply to urban hospitals; and (2) permitted newly "rural" hospitals that were formerly included in MSAs to retain the wage indices of their former MSAs for three years, whereas other hospitals that saw their wage indices drop as a result of the new MSAs received no such "hold-harmless" allowance. However, in both cases the agency offered reasonable justifications. With respect to the decision not to adopt the Micropolitan Statistical Areas, the agency reasonably concluded that some such areas would include so few hospitals as to constitute a partial or complete reversion to reimbursement for actual costs, a result incompatible with Congress's intent in enacting the IPPS. As to the "hold-harmless" allowance, such dispensations must be given infrequently in this cost-neutral system, since each time the agency holds any hospital harmless, every other hospital sees its wage index drop slightly. The defendant reasonably held harmless for three years those hospitals that would see the most drastic cuts to their wage indices, while declining to do so for hospitals that saw much less dramatic (if still significant) changes. Because plaintiffs and the formerly rural hospitals are not similarly situated, it was not arbitrary or capricious for CMS to treat them dissimilarly in this respect. Cf. Marshall County, 988 F.2d at 1224 ("Were the Secretary arbitrarily to grant an exception for some hospitals and not for others identically situated, one could expect a successful challenge.") (emphasis added).

Second, plaintiffs claim that defendant did not seriously consider proposed alternatives in its 2004 rulemaking, because it improperly required any alternative to achieve "consensus" in the industry. However, CMS never stated that such consensus was required; rather, it simply observed that the commenters took sharply inconsistent positions as to how the MSAs should be modified, if it all. It also reminded commenters that it had to take into account all of the effects proposed revisions would have, not simply the effects on the commenters. Neither of these observations is inappropriate or indicative of a failure to seriously consider proposals before it. Nor have plaintiffs demonstrated to us that their proposal (the only one for which the record contains enough detail for us to evaluate) was so superior in fulfilling the agency's mandate that the agency's failure to enact it must have been arbitrary and capricious.16

Finally, we would think that, despite the plaintiffs' fears that the agency favors rural hospitals, the consistent use of MSAs has dampened any such favoritism. MSAs, after all, are constructed through an objective methodology, deliberately without reference to any political considerations, by a different agency that has no involvement in this rulemaking. Plaintiffs (in an argument that is somewhat inconsistent with their claim that the use of MSAs is the result of political favoritism) suggest that the agency's reluctance to engage in such a potentially divisive rulemaking without evidence that the result would be substantially superior to the use of MSAs is impermissible. We disagree. All other things being equal, it is rational and permissible for an agency to adopt an already extant measure that uses objective criteria rather than assuming the task of creating its own measure in the face of a sharply divided regulated industry that has not proposed any clearly superior alternatives. See San Bernardino, 63 F.3d at 889 (noting that MSAs have advantage of being an "established and uniform standard"). Under these circumstances, the agency's continued use of MSAs was not arbitrary and capricious.

In an uncodified section of the 2000 Bill, Congress instructed the agency as follows: "By not later than September 30, 2003, for application beginning October 1, 2004, the Secretary shall first complete (A) the collection of data [on occupational mix]; and (B) the measurement [of earnings and paid hours of employment by occupational category]." 2000 Bill § 304(c)(3).

This provision requires interpretation as to what action was required of the agency on October 1, 2004. Indeed, by its literal terms, it does not command the Secretary to actually apply the adjustment on that date at all, but rather to complete various preparations to do so by September 30, 2003. However, defendant does not read the provision as imposing no obligation to apply the adjustment at all on October 1, 2004, but rather as imposing an obligation only to "begin" applying it, in whatever limited force the agency decides is appropriate. Plaintiffs, as well as the district court, construe the provision to require full implementation of the adjustment on October 1, 2004 under any and all circumstances. We cannot agree with either interpretation.

We think that the provision can only be fairly read to contemplate application in full on October 1, 2004. Congress need not explicitly tell an agency to implement a program "in full," any more than it need tell the agency to do the job "competently"; any reasonable reader trying to do justice to Congress's intent will infer that meaning. Nor is a call for application "beginning on" a date certain reasonably interpreted as a grant of discretion to the agency to apply the adjustment on that date at whatever limited strength the agency believes is appropriate. There is no ambiguity in the statute on this point, and so the defendant's interpretation does not receive deference under Chevron.

Moreover, the provision does not simply require application on October 1, 2004. Rather, it commands the agency, without the least hint of ambiguity, to first complete the collection and measurement of data on September 30, 2003, "for application beginning October 1, 2004." Here, the agency failed to complete the predicate task of adequately collecting and measuring data in time to competently apply the adjustment on schedule, and now cites that very failure as an excuse for less than full compliance with its second task. We do not think this is a reasonable interpretation of Congress's intent.

Not only are the agency's actions violative of the statute, but they are arbitrary and capricious. CMS simply asserts that its data justify implementation at precisely ten percent effectiveness, with no explanation given as to why ten percent was chosen instead of, e.g., twenty percent or fifty percent. In addition, the agency has not accounted for its continuing failure to comply with Congress's unambiguous mandate that the agency complete the necessary data collection and measurement by September 30, 2003. Without any explanation, the agency did not even begin its data collection until September 19, 2003, although it appeared to be ready two years earlier. When the data it then gathered in rushed fashion predictably proved inadequate to permit full implementation on schedule, the agency simply stated its intent to do better the next time it believed it was required to collect data, three years later, and until then apparently intends to continue applying the adjustment at greatly limited effectiveness. Under these circumstances, it was arbitrary and capricious for the agency not to return to data gathering immediately, or at least explain why it is not doing so, rather than proceed as if it had successfully completed the initial data gathering.17 While the Medicare Act, as amended by the 2000 Bill, only requires the agency to gather this data every three years, it does not preclude the agency from doing so more frequently. See 42 U.S.C. § 1395ww(d)(3)(E) (requiring Secretary to measure earnings and paid hours of employment "[n]ot less often than once every 3 years").

Having easily concluded that the agency's actions were in violation of law, we move to the more difficult task of formulating a remedy. Obviously, at this late date the agency cannot be ordered into compliance with the schedule set by Congress. We therefore recognize at the outset that any remedy we devise is necessarily imperfect. However, we must endeavor, to the extent possible, to honor Congress's intent as well as the language of the statute.

The district court simply granted plaintiff its requested relief and ordered CMS to immediately apply the occupational mix adjustment in full based on the data it has already collected.18 While this approach holds some appeal, in that it would prevent any further agency delay, we think it is not what Congress would have wanted. Congress mandated a two-step schedule, under which the agency would first collect and measure data by September 30, 2003, thus assuring that, by the time the agency was to apply the adjustment on October 1, 2004, there could be no issues regarding the quality of the data. Here, the agency still has failed to complete the first step in any satisfactory way. By its own admission, the data it has collected are unreliable and yield results that are contrary to those Congress expected. We think immediate application of the adjustment using such flawed data not only would result in irrational policy (and almost certainly damage some of the intended beneficiaries of this adjustment, through no fault of their own) but would contravene Congress's purpose in setting up this two-step schedule.

Accordingly, we instead order the agency to immediately return to the first step and collect data that are sufficiently robust to permit full application of the occupational mix adjustment. All data collection and measurement and any other preparations necessary for full application should be complete by September 30, 2006, at which time we instruct the agency to immediately apply the adjustment in full. This date falls three years after the agency's initial deadline for data collection; had CMS complied with Congress's dictate, it would have been required to complete its second round of data collection by that date. Although we cannot undo the past and remedy the agency's failure to comply with its statutory obligations through one full round of data collection, we can order the agency to follow the schedule originally anticipated by Congress from this date forward.19

We have considered the parties' remaining arguments and find them to be without merit. For the foregoing reasons, the judgment of the district court with respect to the geographic area issue is affirmed. With respect to the occupational mix adjustment issue, the judgment of the district court is vacated, and this case is remanded to the district court to enter judgment in accordance with this opinion.