Opinion for the Court filed by Senior Circuit Judge McGOWAN.
McGOWAN, Senior Circuit Judge:
In these consolidated cases, several hospitals contest the legality of a rule promulgated by the Department of Health and Human Services 1 (HHS) to reimburse hospitals for malpractice insurance premiums stemming from the treatment of Medicare patients. The District Court, hearing the case on “certified review” from HHS’s Provider Reimbursement Review Board (PRRB), held that the Secretary’s actions were within the scope of her broad discretion. 573 F.Supp. 884 (D.D.C.1983). We remand these cases to the District Court for its consideration of the full rulemaking record, which was not before the District Court, under the guidelines discussed below.2
On March 15, 1979, the Secretary published a notice of proposed rulemaking (NPRM) that expressed her desire to promulgate a new rule, governing Medicare reimbursement of malpractice premiums, to prevent Medicare from paying a “disproportionate” amount of total malpractice premiums. 44 Fed.Reg. 15,744 (1979). The rule then in effect reimbursed a hospital for malpractice premiums in proportion to Medicare patients’ utilization of [113]*113its services during the year. See 42 C.F.R. § 405.452(b)(1) (1978). If, for example, 25% of the services (measured in dollars) rendered by the hospital in a particular year were for Medicare patients, Medicare would reimburse the hospital for 25% of its total malpractice premiums for the' year. The proposed rule, in contrast, was to compensate hospitals for their malpractice premiums in proportion to malpractice awards paid to Medicare patients during the year in question and the four preceding years. If, for example, 40% of a hospital’s malpractice claim dollars actually paid out during the five fiscal years ending 1976 through 1981 were paid to Medicare patients, then the hospital would be reimbursed for 40% of its malpractice premiums in 1981. The proposed rule specified an exception for hospitals that had paid no malpractice claims at all during the previous five years. Such hospitals were to be compensated according to an “actuarial estimate” of the proportion of Medicare losses to be expected in the next year.
Approximately two weeks after the close of comments, HHS promulgated its final rule (“Malpractice Rule”), which differed from the proposed rule only in its treatment of hospitals without claim records from the past five years. See 44 Fed.Reg. 31,641 (1979) (final rule) (codified at 42 C.F.R. § 405.452(b)(1)(h) (1982)). For such hospitals, HHS was to pay a fixed percentage of premiums: the national average percentage of paid malpractice claims attributable to Medicare patients. This “national ratio,” set at 5.1% for the first year, was to be recalculated each year in light of the previous year’s data. The Malpractice Rule also specified that any hospital with records insufficient to determine whether every claim paid during the last five years was to a Medicare patient or a non-Medicare patient was to be reimbursed according to the national ratio.3
The hospitals brought their claims before the PRRB, an entity designed to resolve technical reimbursement disputes. 42 U.S.C. § 1395oo (a) (1982). The PRRB certified that the matter was outside its jurisdiction, thus allowing the hospitals to bypass any review by the Secretary and proceed directly to the District Court. 42 U.S.C. § 1395oo (f)(1).
On the basis of a record that included some but not all of the Department’s rule-making record, the District Court found that, given the Secretary’s broad discretion in administering a complicated statute, her actions comported both with the procedural and substantive requirements of the Administrative Procedure Act and with more specific statutory duties — in particular, with the statute that requires HHS to compensate hospitals for the “reasonable cost” of caring for Medicare patients and which prohibits “cost-shifting” between Medicare and non-Medicare patients, 42 U.S.C. § 1395x(v)(1)(A) (1982) (“Medicare Act”). The hospitals appeal the District Court’s decision.4
[114]*114Under the Administrative Procedure Act, a reviewing court “shall review the whole record or those parts of it cited by a party.” 5 U.S.C. § 706 (1982). We therefore begin with a discussion of the record that should properly have been before the District Court before discussing whether HHS complied with the Administrative Procedure Act and with the Medicare Act.
I
If a court is to review an agency’s action fairly, it should have before it neither more nor less information than did the agency when it made its decision. The Supreme Court’s formulation in Overton Park cautions against both under- and over-inclusiveness in the administrative record before a reviewing court: “[R]eview is to be based on the full administrative record that was before the Secretary at the time he made his decision.” Citizens to Preserve Overton Park, Inc. v. Volpe, 401 U.S. 402, 420, 91 S.Ct. 814, 825, 28 L.Ed.2d 136 (1971) (emphasis added) (footnote omitted); see also Environmental Defense Fund, Inc. v. Costle, 657 F.2d 275, 284 (D.C.Cir.1981). To review less than the full administrative record might allow a party to withhold evidence unfavorable to its case, and so the APA requires review of “the whole record.” 5 U.S.C. § 706 (1982); see S.Rep. 752, 79th Cong., 1st Sess. 28 (1945) (“The requirement of review upon ‘the whole record’ means that courts may not look only to the case presented by one party, since other evidence may weaken or even indisputably destroy that case.”); H.R.Rep. No. 9180, 79th Cong., 2nd Sess. 46 (1946) (same); see also Universal Camera Corp. v. NLRB, 340 U.S. 474, 487-88, 71 S.Ct. 456, 463-65, 95 L.Ed. 456 (1951). To review more than the information before the Secretary at the time she made her decision risks our requiring administrators to be prescient or allowing them to take advantage of post hoc rationalizations, see American Petroleum Institute v. Costle, 609 F.2d 20, 23 (D.C.Cir.1979). We are faced in this case with examples of both an under- and an over-inclusive record, and remand to the District Court so that it may obtain the full administrative record and may prevent any improper supplementation thereof.
The Secretary submitted as evidence before the District Court only an internal memorandum. The plaintiffs submitted a miscellany of documents, many of which might properly appear in an administrative record. See Schedule of Exhibits, Memorandum in Support of Plaintiff’s Motion for Summary Judgment, Walter O. Boswell Memorial Hospital v. Heckler, No. 82-0710 (D.D.C. filed June 24, 1982). Nonetheless, nothing purporting to be the complete administrative record was submitted by either party.
HHS did eventually compile an administrative record for the Malpractice Rule, which it has submitted in other cases and which is in the appellate record as an appendix to an amicus brief. See Brief for Amicus Curiae South Carolina Hospital Association, Walter O. Boswell Memorial Hospital v. Heckler, Nos. 83-2223, 83-2224 & 83-2225, at 4-5 (D.C.Cir. filed Jan. 31, 1984) [hereinafter “Amicus Brief”]. That record comprises eleven volumes, in contrast to the five volumes of documents (many of which are quite properly not included in the administrative record) submitted by the plaintiffs to the District Court.
In Overton Park, the Supreme Court reversed a judgment based only on “litigation affidavits” rather than on any direct [115]*115evidence before the agency. Such affidavits, said the Overton Park Court, “clearly do not constitute the ‘whole record’ compiled by the agency: the basis for review required by § 706 of the Administrative Procedure Act.” Overton Park, 401 U.S. at 419, 91 S.Ct. at 825. Although Overton Park presents an extreme case, since essentially none of the evidence before the agency was before the reviewing court, the District Court here was similarly confronted with large gaps in the facts and opinions before the agency, and thus faced material that did not constitute the “whole record” before the agency at the time of its decision.
The APA does allow review based not only on “the whole record,” but on “those parts of it cited by a party.” One might therefore argue that review could go forward in this case using those parts of the record cited to the District Court. We believe, however, that in the circumstances of this case such an outcome would be fundamentally unfair. For review to go forward on a partial record, we would have to be convinced that the selection of particular portions of the record was the result of mutual agreement between the parties after both sides had fully reviewed the complete record. In that situation, we might naturally assume that the omitted portions did not materially affect either party’s case and, for our own convenience, review the ease on that portion of the record cited by the parties.
In this case, however, it appears that the plaintiffs were unaware of a number of items included in the eleven-volume version of the record compiled by the government after the District Court hearing. See Opposition of Amicus Curiae to DefendantAppellee’s Motion to Strike Appendix C and Related Portions of the Amicus Brief at 3-4 (D.C.Cir. filed Feb. 21, 1984). This asymmetry in information undermines the reliability of a court’s review upon those portions of the record cited by one party or the other. Since there would be no check upon the failure of the agency to disclose information adverse to it, the normal pressures towards inclusion of all relevant material in the record before the court are absent. Some of the documents of which the plaintiffs were apparently unaware are quite critical of the single study that the Secretary cites as a basis for the Malpractice Rule. See Amicus Brief at 17-19. The likelihood that the plaintiffs would have drawn the District Court’s attention to such documents had they been aware of them seems great, and thus the harm to the plaintiffs is significant.5 We cannot under these circumstances accept the citations to the record given by the parties as sufficient to allow the District Court to discharge its duty of review. We therefore remand the case to the District Court for consideration using the eleven-volume administrative record submitted as Appendix B of the amicus brief in this case.6 Cf. Natural Resources Defense Council, Inc. v. Train, 519 F.2d 287, 291-92 (D.C.Cir.1975) (remanding case with incomplete administrative record where District Court was aware that document was missing).
This case also presents this court with an overinclusive record on review. The second part of the Overton Park formulation is to limit the record to that information before the Secretary at the time of his decision, see supra p. 792, thus excluding ex post supplementation of the record by either side. The hospitals are guilty of such efforts in this case. The amicus hospital attempted to submit affidavits before this court that summarize a study performed after HHS published the final rule. This court struck the portion of the amicus brief discussing these exhibits. Order in [116]*116Walter O. Boswell Memorial Hospital v. Heckler, Nos. 83-2223, 83-2224 & 83-2225 (D.C.Cir. May 3, 1984); see Association of Pacific Fisheries v. EPA, 615 F.2d 794, 811-12 (9th Cir.1980).7 The appellant hospitals placed evidence before the District Court indicating their actual experience with the Malpractice Rule, which the District Court admitted as relevant in determining the plaintiffs’ standing. Such information is not, however, relevant to determining the validity of the rule. If the parties wish to contest the merits of the case by referring or submitting to the District Court any material that does not appear in the eleven-volume administrative record, they may do so only by joint stipulation.
II
The Administrative Procedure Act requires the Secretary to consider the comments submitted to her regarding a proposed rule and to present a “concise general statement” of the rule’s “basis and purpose.” 5 U.S.C. § 553(c) (1982). Reading these requirements together, this court has set forth the standard for the basis-and-purpose statement:
We do not expect the agency to discuss every item of fact or opinion included in the submissions made to it in informal rule making. We do expect that, if the judicial review which Congress has thought it important to provide is to be meaningful, the “concise general statement of * * * basis and purpose” mandated by [the APA] will enable us to see what major issues of policy were ventilated by the informal proceedings and why the agency reacted to them as it did.
Automotive Parts & Accessories Association v. Boyd, 407 F.2d 330, 338 (D.C.Cir.1968); see Home Box Office, Inc. v. FCC, 567 F.2d 9, 35-36 (D.C.Cir.) (per curiam), cert. denied, 434 U.S. 829, 98 S.Ct. 111, 54 L.Ed.2d 89 (1977).
On its review of the full administrative record, the District Court should pay particular attention to HHS’s basis-and-purpose statement and the comments received on the rule, and decide whether HHS provided sufficient insight into the issues raised, and its reactions thereto, in three instances: its decision that reimbursement under the initial definition of the “general and administrative” cost pool subsidized non-Medicare patients at the expense of Medicare patients; its use of the “Westat study”; and its consideration of a “separate-policies” alternative to the Malpractice Rule.
The Secretary must issue regulations defining reasonable cost in such a way that
the necessary costs of efficiently delivering covered services to individuals covered by [Medicare] will not be borne by individuals not so covered, and the costs with respect to individuals not so covered will not be borne by [Medicare] ____
42 U.S.C. § 1395x(v)(1)(A) (1982). As this court has said, “[t]his provision clearly is meant to prevent the Medicare program from subsidizing non-Medicare related costs, but it equally clearly proscribes Medicare from being subsidized by non-Medicare sources.” St. Mary of Nazareth Hospital Center v. Schweiker, 718 F.2d 459, 462 (D.C.Cir.1983).
Before promulgation of the Malpractice Rule, Medicare reimbursed malpractice insurance expenses as part of the expenses in the “general and administrative” (G & A) cost center. Medicare reimbursed expenses in such cost centers by determining the percentage of the provider’s total medical services rendered to Medicare patients and then paying the provider that percentage of the expenses in the cost center. See 42 C.F.R. § 405.452(b)(1) (1978). For example, if costs in a hospital’s G & A cost center were $100 million for a hospital and the hospital rendered 27% of its total services to Medicare patients, the hospital would have received $27 million from Medicare towards G & A expenses.
Given the prohibition on cost-shifting between Medicare and non-Medicare patients, [117]*117the implicit assumption of the G & A cost-center scheme is that Medicare patients consume general and administrative services in proportion to their total consumption of all services in the hospital. If, for example, Medicare patients consumed proportionately less G & A services, taken as a whole, than total hospital services, Medicare patients would be subsidizing the G & A costs generated by non-Medicare patients, in violation of the Medicare Act.
In this case, the Secretary concluded on the basis of a consultant’s study (the “Westat study”) that Medicare patients were consuming proportionately less of a particular G & A service — the paid malpractice claims made possible by G & A expenditures on malpractice premiums — than they were consuming of total hospital services. 44 Fed.Reg. 31,641 (1979). The Secretary’s Malpractice Rule took malpractice premiums out of the pool of G & A costs and instead reimbursed premiums based on the percentage of claims paid out to Medicare patients. Id. at 31,642. For such an action to be consistent with the Medicare Act’s prohibition of cost-shifting, the Secretary must have implicitly determined that the old G & A pool, taken as a whole, had come to subsidize non-Medicare patients at the expense of Medicare patients. Without such a determination, there would be no legitimate grounds for removing malpractice premiums from the general G & A pool, since the old G & A pool had long been presumed to balance costs fairly between Medicare and non-Medicare. patients. Given no net cost-shifting in the old pool and the removal of Medicare-subsidized malpractice premiums, the new G & A pool would inevitably subsidize Medicare patients at the expense of non-Medicare patients:
HHS in fact set forth the conclusion in its basis-and-purpose statement necessary to support removing Malpractice premiums from the G & A pool:
We believe that the current cost finding and apportionment procedure is generally an equitable method of allocating [non-malpractice] overhead costs. However, because malpractice costs are so significant and the disproportionate allocation of malpractice costs to Medicare is so great, we believe a unique exception is warranted to deal with these costs.
Id. at 31,642. The precise factual basis for this conclusion, however, remains obscure from the record before the District Court.8 On remand, HHS should direct the District Court to the facts upon which HHS based its conclusion that malpractice premiums are so significant and disproportionate as to bias a once-balanced G & A pool, so that the reviewing court may have some idea whether comments ventilated this important consideration and whether HHS responded to them adequately.
We next consider the Westat study. The saga of the Malpractice Rule began when the Secretary hypothesized that the age differential between Medicare patients and the rest of the patient population9 meant [118]*118that Medicare patients were paying a disproportionate share of the costs of malpractice insurance. Because Medicare patients are older and are more likely to be retired, they are likely to have smaller lifetime economic losses from an injury than younger, working adults. Since economic losses are part of malpractice awards, Medicare patients might receive smaller malpractice awards for a given injury. The Secretary hypothesized that, given the “utilization method” then used for malpractice premiums as part of the G & A pool, see supra p. 790, smaller average awards might mean Medicare patients were paying a disproportionate share of hospitals’ malpractice insurance premiums. She concluded that the Westat study supported this hypothesis sufficiently to enact the Malpractice Rule.
The Westat study was a census of malpractice claims against hospitals and physicians paid by nine insurers during four months of 1976. Westat Study at 2-1, App. at 101. Insureds of these nine companies were involved in 84% of malpractice claims closed in the U.S. during those four months. Id. The study concluded that “[i]n general, the size of [a malpractice] award decreases with age,” id. at 3-1, App. at 110, and that Medicare patients received “considerably” smaller average awards than other groups, id. at 3-2, App. at 111.
The commenters raised numerous criticisms of the study. The single chapter of the study specifically dealing with Medicare patients and malpractice10 itself cautions that the data is useful for inference “only within the time frame for which the data was collected and only to the universe of claims closed by the nine participating companies.” Id. at 2-9, App. at 109. The authors warn that “Conclusions [about injury prevention] must be drawn very cautiously because of the possibility of biased data,” id. at 2-3, App. at 103, and the same cautions would seem to apply to conclusions about malpractice award size.11 See also supra note 5 (discussing remarks of study’s authors in full record but not before District Court). Because the census used only data provided by insurance companies, it excluded self-insured hospitals, which the study estimated to constitute 20% of hospitals in 1976. Id. at 2-2 (citing Tanenbaum & Korsak, AHA Research Capsule: Methods of Hospital Liability Insurance, in American Hospital Association, 1976 Survey of Selected Topics), App. at 102. The data involved was so spotty that it was impossible to determine whether the claiming patient was covered by Medicare in fully one-third of the claims, Westat Study at 2-8, App. at 108, a fact of particular import given Medicare’s final decision to require hospitals either to provide complete data for every claim in the relevant five-year period or use the national ratio.
Perhaps the most significant flaw in the Westat study is the failure of its conclusions to distinguish claims against hospitals from claims against physicians. Although the Malpractice Rule applies exclusively to [119]*119hospitals, only 32% of the claims examined in the Westat study were against hospitals, with 63% of the claims against physicians. Id. at 3-3, App. at 112. In concluding that Medicare was paying a disproportionate share of premiums, the Westat study used “malpractice cases occuring [sic] in hospitals,” id. at 5-5 (emphasis added), not malpractice cases against hospitals. The relevant data therefore included malpractice occurring in hospitals but resulting in a suit against a physician rather than the hospital.
HHS did not attempt to show that the data on suits against physicians for malpractice in hospitals were equally applicable to suits against hospitals. Indeed, there is evidence in the Westat study to the contrary. What the Westat study termed “institutional problems,” which “generally refer[] to causes of injuries that are outside the control of the physician,” id. at 5-15, App. at 129, were the cause of injury in 69% of Medicare patients’ malpractice claims but in only 35% of all patients’ claims, id. at 5-17, App. at 131. The study states that its four other categories for cause of injury “pertain to errors usually made by a physician,” id. at 5-15, App. at 129, and in those categories a higher percentage of non-Medicare patients are injured than of Medicare patients, id. at 5-17, App. at 131. These disproportions could be serious enough to mean hospitals pay out more in claims to Medicare patients than their proportionate consumption of general services.12 It would be incautious to draw such a conclusion from the limited data presented in the Westat study, but it is equally incautious to draw, as HHS did without any discussion of the Westat study’s limits in the basis-and-purpose statement, the opposite conclusion. HHS also used this study to determine, at least for the first year of the rule, the national ratio used to compensate hospitals with nonexistent or incomplete claims records, despite the study’s minimizing the significance of some problems with the data because “it [was] not the explicit purpose of this study to make national yearly estimates.” Id. at 2-5, App. at 105. Whether HHS’s actions, unaccompanied by any defense of the Westat study in the basis-and-purpose statement, was a violation of the Automotive Parts standard is a matter for the District Court to decide on the full administrative record in light of the importance the District Judge adjudges the Westat study to have had in the agency’s decisions.
Finally, an agency must consider “reasonably obvious alternative ... rules and explain its reasons for rejecting alternatives in sufficient detail to permit judicial review.” NAACP v. FCC, 682 F.2d 993, 998 (D.C.Cir.1982); see also Home Box Office, Inc. v. FCC, 567 F.2d 9, 35 (D.C.Cir.1977) (opportunity to comment should not be “meaningless”), cert. denied, 434 U.S. 829, 98 S.Ct. 111, 54 L.Ed.2d 89 (1977). As discussed below, see infra pp. 802-803, one alternative to the Malpractice Rule, a “separate-policies” rule, has several desirable characteristics. The basis-and-purpose statement did not mention this alternative. Indeed, the basis-and-purpose statement does not mention any alternatives to the Malpractice Rule except for the status quo ante. The District Court in its examination of the full record must determine if HHS’s [120]*120failure to discuss the separate-policies option, or some other alternative, violated the mandates of the Administrative Procedure Act.
Ill
We now consider the standards the District Court should use to decide if the Secretary’s actions were arbitrary and capricious under the Administrative Procedure' Act, and if her actions violated the Medicare Act. The two questions are closely linked.
The Supreme Court has recently reiterated the governing standard:
Normally, an agency rule would be arbitrary and capricious if 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 Manufacturers Association of the United States v. State Farm Mutual Automobile Insurance Co., 463 U.S. 29, 103 S.Ct. 2856, 2867, 77 L.Ed.2d 443 (1983). HHS has not here relied on factors Congress did not intend it to consider. The District Court should, however, closely examine the full administrative record for the-other administrative excesses described above.
There are two important aspects of the problem that HHS may have failed entirely to have considered in a reasoned fashion. The first is whether the pre-Malpractice Rule G & A pool resulted in Medicare patients’ subsidization of non-Medicare patients. The second is whether HHS considered a particularly plausible alternative to the Malpractice Rule. Intermingled with these claims is the assertion that the explanation for the Malpractice Rule runs counter to the Westat study’s evidence before the agency, and the assertion that the Malpractice Rule is based on an implausible interpretation of the Medicare statute or is intrinsically incapable of fulfilling its stated goals.
The Malpractice Rule is a “unique exception” to the long-standing agency practice of lumping a wide variety of costs together into the “G & A” cost center. Final Rule, 44 Fed.Reg. 31,641, 31,642 (1979). Exceptions to long-standing rules are clearly permissible. Agencies like HHS
are not expected to cast their rules in stone; “they are supposed, within the limits of the law and of fair and prudent administration, to adapt their rules and practices to the Nation’s needs in a volatile, changing economy.”
Advanced Micro Devices v. CAB, 742 F.2d 1520, 1541 (D.C.Cir.1984) (quoting American Trucking Associations, Inc. v. Atchison, T. & S.F. Ry. Co., 387 U.S. 397, 416, 87 S.Ct. 1608, 1618, 18 L.Ed.2d 847 (1967)). Although the ratemaking agencies are perhaps more intimately involved in purely economic considerations, HHS should also take changes in costs into account, as it did here. “An administrative agency ... is not bound to rigid adherence to precedent. It may switch rather than fight the lessons of experience.” Office of Communication of the United Church of Christ v. FCC, 590 F.2d 1062, 1069 n. 26 (D.C.Cir.1978) (quoting New Castle County Airport Commission v. CAB, 371 F.2d 733, 735 (D.C.Cir.1966) (footnote omitted), cert. denied, 387 U.S. 930, 87 S.Ct. 2052, 18 L.Ed.2d 991 (1967)). The question, however, is whether HHS has here based its actions on lessons of experience or mere expedience.
HHS has frequently battled against removing from the G & A pool particular costs subsidized by non-Medicare patients, such as the expense of completing reports required only by Medicare, on the grounds that the pool taken as a whole fairly reflects G & A expenses. See Prov.App.Dec. No. 00-75-04 [1975 Transfer Binder] Medicare & Medicaid Guide (CCH) ¶ 27,361, at 9589; see also Good Luck Nursing Home v. Califano, 3 Medicare & Medicaid Guide (CCH) ¶ 29,450 (D.D.C.1978) (legal and ac[121]*121counting fees). HHS has resisted such efforts on the plausible grounds that “where a particular cost might be allocated disproportionately to or from the program, there will be other costs disproportionately allocated in the other direction which will compensate for the first cost.” Intermediary Letter No. 234 (June 2, 1967), App. at 31. The Malpractice Rule removes from the G & A pool a particular G & A cost assertedly subsidized by Medicare patients. It would be arbitrary and capricious for HHS to bring varying interpretations of the statute to bear, depending upon whether the result helps or hurts Medicare’s balance sheets, without some showing that changes in the underlying balance of G & A expenses, taken as a whole, warranted breaking out malpractice premiums from the G & A pool.13
The substantive disputes in this case stem chiefly from the conflict between compensating actual costs and avoiding cost-shifting between Medicare and non-Medicare patients. Providers of medical services to patients covered by the Medicare program are entitled to reimbursement for the “reasonable cost” of those services, 42 U.S.C. § 1395f(b) (1982), defined as “the cost actually incurred, excluding” expenditures not necessary to the efficient provision of health care, id. § 1395x(v). As described supra at p. 794, the same section of the statute also requires the Secretary to define reasonable cost so as to prevent cost-shifting between Medicare and non-Medicare patients. 42 U.S.C. § 1395x(v)(1)(A) (1982).
HHS requires hospitals to maintain “an adequate insurance program.” Provider Reimbursement Manual § 2160.A, Medicare and Medicaid Guide (CCH) U 5999X. Given the size and frequency of malpractice awards, one can reasonably assume that malpractice insurance is necessary for “adequate” coverage, and thereby infer that HHS considers malpractice insurance for Medicare patients necessary and thus compensable. The question is not whether to reimburse the reasonable costs of covering malpractice awards to Medicare patients, but how to define which costs are reasonable.
The hospitals argue that their reasonable costs are their actual costs, and that their actual costs for a Medicare patient are the additional malpractice premiums that treating another Medicare patient causes the hospital to pay. Insurance companies currently calculate malpractice premiums based on the “exposure” of the hospital to claims; that exposure depends upon the hospital’s own claim record, the claim records of other hospitals in the area, and the quantity of services the hospital provides. Insurers do not currently distinguish between Medicare and non-Medicare patients.
The hospitals assert that, even if the Secretary could trust the Westat study’s conclusion that Medicare patients received proportionately less in claims than they consume of total hospital services, the structure of malpractice insurance would prevent Medicare patients from paying disproportionately high premiums whether or not they received disproportionately low claims. To the extent that malpractice premiums do not reflect a hospital’s loss record, the lower losses in hospitals with many Medicare patients compared to those with few Medicare patients will not be reflected in lower malpractice premiums. And to the extent that calculation of exposure depends on services rendered but does not distinguish Medicare from non-Medi[122]*122care patients, hospitals incur identical costs for a Medicare and non-Medicare patient whatever average claim éach might generate.
Since the old rule reimbursed hospitals for malpractice premiums simply on the basis of utilization of general hospital services, it compensated hospitals without direct regard to the lower claims assertedly generated by Medicare patients, and was thus consistent with the practices of the insurance industry. The Malpractice Rule, however, would, on average, reimburse a hospital with Medicare patients less money even though it paid no less in insurance premiums than before the Malpractice Rule. Non-Medicare patients would have to make up the difference. The hospitals claim that the rule therefore violates the Medicare Act’s provisions that require Medicare to compensate hospitals for their “reasonable costs” and that prohibit cost-shifting between Medicare and non-Medicare patients.
Given that malpractice insurance premiums are not calculated purely according to the loss record of a hospital, and assuming for the moment that Medicare patients did receive less under the old rule in malpractice claims than HHS paid out in premiums, either HHS or hospitals with Medicare patients will suffer no matter what the rule. If Medicare uses the old rule, then HHS pays out more in premiums than Medicare patients receive in claims. This arguably conflicts with HHS’s mandate not to subsidize the costs of non-Medicare patients. If HHS compensates the hospitals under the new rule, then the hospital pays out more in premiums for Medicare patients than it receives as reimbursement from Medicare.14 This arguably violates HHS’s duty to compensate hospitals for their “actual costs.”
To some extent, therefore, the Secretary is simply faced with conflicting statutory provisions. Although an agency must “give effect to the unambiguously expressed intent of Congress,” and is owed no deference by a court if Congress has expressed such an intent, Chevron, U.S.A., Inc. v. Natural Resources Defense Council, Inc., — U.S. -, 104 S.Ct. 2778, 2781-82, 81 L.Ed.2d 694 (1984) (footnote omitted), Congress here has expressed no such unambiguous intent. The words of the statute — the requirement to compensate “actual costs” and the prohibition against cost-shifting — conflict in their application to malpractice premiums because of the way insurance companies determine those premiums.
The legislative history is not much more helpful. In mandating the reimbursal of “reasonable costs,” Congress was concerned for the most part with allowing every provider to be fairly compensated despite significant variations in the quality and intensity .of care. H.R.Rep. No. 213, 89th Cong., 1st Sess., at 31-32 (1965), U.S. Code Cong. & Admin.News 1978, p. 1943; S.Rep. No. 404, 89th Cong., 1st Sess., at 35-36 (1965). Nonetheless, the legislative history emphasizes that “costs” include not only direct medical services but items such as depreciation on buildings, interest on capital debts, and continuing education for medical personnel. H.R.Rep. No. 213, at 32; S.Rep. No. 404, at 36. The only specific mention of actual costs suggests that compensating actual costs is designed to help avoid cost-shifting:
[T]he objective, whatever method of computation is used, will be to approximate as closely as practicable the actual cost (both direct and indirect) of services rendered to the beneficiaries of the program so that under any method of determining costs, the costs of services of individuals covered by the program will not be borne by individuals not covered, and the costs of services of individuals not covered will not be borne by the program.
H.R.Rep. No. 213, at 32 (emphasis added); S.Rep. No. 404, at 36 (same). The means of using actual costs for' compensation is [123]*123thus to some extent impliedly subordinate to the end of avoiding cost-shifting, an interpretation that supports the agency’s action here.
Given that the statutory language is ambiguous in this instance and that the legislative history does not conclusively resolve the point, we can only conclude that the conflict presented in this case was not clearly contemplated by Congress. We therefore owe the agency some deference. And in these circumstances, “where the decision under review involves an agency’s interpretation of its own regulations, forming part of a complex statutory scheme which the agency is charged with administering, the arguments for deference to administrative expertise are at their strongest.” Psychiatric Institute of Washington, D.C., Inc. v. Schweiker, 669 F.2d 812, 813-14 (D.C.Cir.1981) (per curiam). Indeed, because Congress in the Medicare Act explicitly delegated to the Secretary the authority to promulgate regulations regarding reasonable costs, the District Court cannot reverse the Secretary unless her interpretation of the statute was arbitrary and capricious. Chevron, U.S.A., 104 S.Ct. at 2782. The question of whether the Secretary violated the APA therefore merges with the question of whether she violated the Medicare Act.
We also note that the Medicare Act plainly contemplates instances in which the Secretary does not agree with an insurer’s definition of actual costs. The Secretary is authorized to promulgate regulations defining reasonable costs, and, in doing so, “shall consider, among other things, the principles generally applied by ... established prepayment organizations.” 42 U.S.C. § 1395x(v)(1)(A) (1982) (emphasis added). We do not, therefore, find arbitrary and capricious the Secretary’s interpretation of the statute so as to deny hospitals some of their premium costs. After all, paying the percentage of premiums reflecting losses from Medicare patients would in the long run fairly compensate the insurance companies for their expenses, and thus would fairly reimburse the hospitals for necessary expenditures.
The hospitals raise some questions about the Malpractice Rule’s short-run effects, however. They assert that it is arbitrary and capricious because it leads to “bizarre results,” and construct several examples in which applying the Malpractice Rule does not appear plausibly to serve its goals. See Brief of Appellants at 8-10 (filed Feb. 1, 1984). Consider a hospital that pays $100,000 each year in premiums. The hospital’s patients are 80% Medicare and 20% non-Medicare. A Medicare patient successfully sues the hospital in 1980 for a malpractice claim of $100. The hospital pays no other claims between 1980 and 1984. Each year from 1981 to 1985, Medicare will, under the Malpractice Rule, pay all of the hospital’s malpractice premiums. In 1985, a non-Medicare patient successfully sues for $1,000,000. The hospital pays no other claims between 1985 and 1989. Each year from 1986 to 1990, Medicare will not pay any of the hospital’s malpractice premiums.
This example demonstrates the Malpractice Rule’s statistical nature. Because the Malpractice Rule uses claims paid out as the basis for compensation even though hospitals pay premiums more frequently than claims, the Malpractice Rule accurately divides premiums between Medicare and non-Medicare patients only to the extent that the hospital has had the “average” loss experience that one would expect from looking at its mix of patients. The hospital in the example, however, never has the “average” loss experience — which would, assuming that Medicare patients do generate fewer dollars of malpractice claims per dollar of service rendered to them than non-Medicare patients, be something less than 80% of its malpractice claims paid out to Medicare patients and something more than 20% of claims paid to non-Medicare patients. Instead, its experience each year is at one of many extremes — all claims paid to Medicare patients, all claims paid to non-Medicare patients, or no claims paid at all. In different years, therefore, the Malprac[124]*124tice Rule reimburses the hospital for all of its expenses or none of its expenses, even though the hospital’s mix of Medicare and non-Medicare patients remains the same.
The example is, however, only an example. Because malpractice claims may not be frequent enough to reflect at each moment the long-run average for a particular hospital, the Malpractice Rule will in some cases produce superficially peculiar results. The record before the District Court seems somewhat lacking in a discussion of how often such results might occur. But in the abstract, the averaging method of the Malpractice Rule will by definition result in a fair division of the long-run average Malpractice premiums attributable to Medicare and to non-Medicare patients. That is a fair formulation of HHS’s statutory mandate. That some hospitals may suffer in the short run — and some, it should be noted, will benefit greatly — does not make the Malpractice Rule invalid.
There is, however, a way to achieve HHS’s goal more predictably than does the Malpractice Rule. Insurance companies could create separate pools of risk for Medicare and non-Medicare patients. This method would soon allow insurance companies to use their extensive experience in evaluating risks to make a more precise estimate of the costs attributable to each set of patients and to charge them different premiums. Medicare could then reimburse hospitals completely for the premiums paid for Medicare patients, and reimburse hospitals not at all for the premiums paid for non-Medicare patients. There would be no cost shifting, and hospitals would be compensated for their actual costs of treating Medicare patients.
The advantages of a scheme of separate policies over the Malpractice Rule are several. First, it moves the risk of statistical fluctuations in claims paid from the hospitals, which are poorly equipped to deal with such risks, to the insurance companies, which exist precisely to absorb such risks. A hospital would be fully reimbursed for a steady premium while its payouts fluctuated, instead of being reimbursed for a potentially unpredictable percentage of its premiums. Second, this same transfer of risk means insurance companies rather than Medicare will calculate the proper premium for hospitals without extensive claims records. If it proves commercially worthwhile to the insurers to use a more focused measure of risk than a national ratio, they will do so. Finally, it gives insurance companies an incentive to determine more precisely the difference in risks presented by Medicare and non-Medicare patients, just as insurance companies have an incentive to determine how risky various types of drivers are to insure.
After issuing the Malpractice Rule, however, Medicare specified in its Provider Reimbursement Manual that it would not “recognize commercial insurance or a separate self-insurance fund only for Medicare (or Medicaid) beneficiaries for malpractice.” Provider Reimbursement Manual, supra, § 2163.2B, App. at 40. Instead, HHS would aggregate the premiums for separate policies and compensate the hospitals for the percentage of the total premiums according to the usual claims-percentage method of the Malpractice Rule, effectively removing any incentive for anyone to use separate policies.
This prohibition seems a curious one in light of the Secretary’s expressed intention in the basis-and-purpose statement to “reimburse Medicare providers on a basis more closely related to actual malpractice experience,” 44 Fed.Reg. 31,641 (1979), since separate policies would depend on Medicare losses rather than Medicare utilization of total services. The prohibition of separate policies seems especially odd in light of HHS’s desire to avoid paying a disproportionate amount of malpractice costs, id., since separate policies would not only result in a more focused attribution of cost but would smooth out the statistical variations to which the hospital might otherwise be exposed. The separate-policies [125]*125option would also clearly result in hospitals’ receiving their actual costs.
The agency did not respond in its basis- and-purpose statement to commenters who proposed this option. See supra p. 797. HHS now advances two post hoc grounds for its prohibition of separate policies. First, it warns that the higher administrative costs of separate policies would outweigh the benefits.15 Yet no investigation into such costs appears to have been undertaken. Second, HHS raises the spectre of hospital-insurer collusion. Tape of Oral Argument of June 4, 1984, Walter 0. Boswell Memorial Hospital v. Heckler, Nos. 83-2223, 83-2224 & 83-2225 (D.C.Cir.). While this possibility may be worthy of investigation,16 the record before the District Court does not reveal even a cursory examination by HHS of its likelihood. In any case, the post hoc rationalizations of counsel are no substitute for consideration or reasoning during rulemaking. See Burlington Truck Lines, Inc. v. United States, 371 U.S. 156, 168-69, 83 S.Ct. 239, 245-46, 9 L.Ed.2d 207 (1962).
We cannot, of course, substitute our judgment for the agency’s. Overton Park, 401 U.S. at 416, 91 S.Ct. at 823. We do not express any opinion as to the comparative merits of the Malpractice Rule and the separate-policies option. We do, however, note that agencies must consider reasonably obvious alternatives to their proposed rule, and that agencies should not prohibit initiatives of regulated entities that further the agency’s expressed goals without offering some plausible rationale. The District Court, upon review of the full administrative record, must decide whether HHS has discharged these responsibilities.
Finally, reliance on a single study criticized extensively by its authors in its particular application can obviously be an arbitrary and capricious action. See Almay, Inc. v. Califano, 569 F.2d 674 (D.C.Cir.1977). There is the possibility in this case that HHS offered an explanation for its action contrary to the evidence before it. The District Court, however, did not have before it all the authors’ criticisms. See Amicus Brief at 17. The District Court should carefully evaluate whether HHS’s reliance on the Westat Study was arbitrary and capricious in light of these criticisms, as well as of other criticism and support of the use of the study contained in the full administrative record.
It is so ordered.