FLAUM, Circuit Judge.
In these consolidated appeals, the Secretary of the Department of Health and Human Services (the “Secretary”) seeks reversal of two district court decisions invalidating a regulation promulgated in 1979 under Title XVIII of the Social Security Act, 42 U.S.C. §§ 1395 et seq. (1982) (the “Medicare Act”). The regulation at issue prescribes a new formula for reimbursing Medicare health care providers for malpractice insurance premiums associated with the care of Medicare patients. 42 C.F.R. § 405.-452(a)(l)(ii) (1984) (the “Malpractice Rule”). The district courts below concluded that the Malpractice Rule was invalid because it was arbitrary and capricious and contrary to the Medicare Act and because the Secretary failed to provide an adequate basis and purpose statement. St. James Hospital v. Heckler, 579 F.Supp. 757 (N.D.Ill.1984) (Will, J.); Humana of Illinois, Inc. v. Heckler, 584 F.Supp. 618 (C.D.Ill.1984) (adopting Judge Will’s decision).1 We affirm.
[1463]*1463I.
A. The Malpractice Rule
Under the Medicare Act, a health care provider is entitled to government reimbursement for the lesser of the “reasonable cost” of the services it provides for Medicare patients or its “customary charges” with respect to those services. 42 U.S.C. § 1395f(b)(l) (1982). “Reasonable cost” is defined as “the cost actually incurred, excluding therefrom any part of incurred cost found to be unnecessary in the efficient delivery of needed health services, and shall be determined in accordance with regulations” to be promulgated by the Secretary. 42 U.S.C. § 1395x(v)(l)(A) (1982). The statute further provides that such regulations shall take into account “both direct and indirect costs of providers of services ... in order 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] ____” Id.
With these directives in mind, the Secretary adopted regulations in 1966 that pooled all of a health care provider’s general and administrative (“G & A”) costs together and then allocated them among Medicare and non-Medicare patients on the basis of patient hospital usage (“patient utilization ratio”). 31 Fed.Reg. 14,808 (1966) (codified at 20 C.F.R. Part 405, Sub-part D (1966)). For example, if Medicare patients constituted 30% of the patient charges at a hospital in a given year, then Medicare reimbursed the hospital for approximately 30% of its G & A costs. G & A costs included a variety of overhead expenses, including billing costs, accounting and legal fees, and all types of insurance premiums. This method of allocating indirect costs remained unchanged from 1966 until promulgation of the Malpractice Rule in 1979.
On March 15, 1979, the Secretary issued the proposed Malpractice. Rule for public comment. 44 Fed.Reg. 15,744 (Mar. 15, 1979). The proposed rule removed malpractice insurance premiums from the G & A pool of indirect costs and apportioned them directly based on malpractice loss experience. The stated reason for segregating malpractice insurance premiums from the other G & A costs was that the Medicare program was paying a disproportionate amount of malpractice insurance costs under the old formula. This conclusion was based on “[a] study conducted by an HEW consultant indicating] that malpractice awards for Medicare ... patients are significantly lower in amount than losses for other patient population [sic].” Id. at 15,745.2 The Secretary attributed the low[1464]*1464er awards for Medicare patients to the fact that their income potential and life expectancies are generally less than those of the remainder of the patient population. Id.
The proposed rule removed malpractice insurance premiums from the G & A pool and instead reimbursed those costs by dividing the losses paid to Medicare patients by the total plaid malpractice losses, and then multiplying that ratio times malpractice insurance costs. Id. The ratio was to be calculated for both the cost year at issue and the four preceding years. The Secretary selected a single provider’s five-year malpractice loss ratio as the relevant measure on the theory that “the estimated malpractice losses paid in future periods are closely related to past malpractice losses paid.” Id. Hospital's with no malpractice loss experience for the relevant five-year period were required to obtain an actuarial estimate of Medicare’s share of current malpractice costs from an independent actuary, insurance company, or broker. Id.
The Secretary received nearly 600 comments on the proposed rule.3 The comments were submitted by health care institutions, consumers, accounting firms, insurance companies, actuaries, health care consultants, physicians and nurses, and Medicare beneficiaries. All comments received were opposed to the proposed rule and recommended its complete withdrawal. 44 Fed.Reg. 31,641 (June 1, 1979). There were seven major criticisms of the proposed rule: (1) the inequity of removing malpractice premiums from the G & A pool while retaining in the pool other costs incurred principally for Medicare patients, (2) providers’ anticipated cash flow problems due to erratic reimbursement patterns as the paid-loss ratio fluctuates from year to year, (3) the increased likelihood that hospitals would fraudulently treat malpractice claims by Medicare patients, (4) the statistical invalidity of the Westat study, (5) the fact that malpractice insurance protects hospitals’ assets and thus benefits all patients equally, (6) the fact that insurance companies set premiums on the basis of an overall assessment of risk, without allocating the risk among different types of patients, and (7) the complexity and expense of the additional accounting required by the proposed rule.
Notwithstanding the many adverse comments, the Secretary issued the final Malpractice Rule on June 1,1979. 44 Fed.Reg. 31,641 (originally codified at 42 C.F.R. § 405.452(b)(l)(ii)). The only change to the proposed rule was the formula to be used for providers with no malpractice loss experience. Instead of the actuarial estimate required in the proposed rule, the final rule calculated reimbursement for these providers on the basis of the national ratio of malpractice awards paid to Medicare patients to the malpractice awards paid to all patients. Id. at 31,642. This national ratio, set at 5.1% for the first year based on the Westat study, was to be recalculated each year in light of the previous year’s data.4
B. The Hospitals
Plaintiff-appellees St. James Hospital and Humana of Illinois, Inc. (“Humana”) are both health care providers as defined in the Medicare Act. See 42 U.S.C. § 1395x(u) (1982). For its cost year ending in 1980, St. James Hospital’s overall Medicare patient utilization ratio was 30%. Under the new Malpractice Rule, however, Medicare only reimbursed 20.11% of its malpractice insurance premiums. The [1465]*1465amount of denied Medicare reimbursement at issue is $24,159.
On June 3, 1981, St. James sought a hearing on its malpractice reimbursement determination before the Provider Reimbursement Review Board (“PRRB” or the “Board”) pursuant to 42 U.S.C. § 1395oo(a) (1982). The Board held that it had no jurisdiction to review the validity of the Malpractice Rule and therefore, pursuant to statute, certified the question for appeal to the federal district courts. 42 U.S.C. § 1395oo(f)(l) (1982). On appeal, Judge Will held that the Secretary failed to provide an adequate basis and purpose statement for the Malpractice Rule and that the Rule was arbitrary and capricious and not in compliance with the requirements of the Medicare Act. St. James Hospital v. Heckler, 579 F.Supp. at 767. St. James Hospital’s motion for summary judgment was accordingly granted. Id.
The Malpractice Rule’s impact on Humana was even greater than its effect on St. James Hospital. Humana’s Medicare patient utilization ratio was 49.38% for cost year 1980, yet its reimbursement under the Malpractice Rule was zero. Humana received no Medicare reimbursement because it had only paid a single $750 malpractice award in the relevant five-year period and the recipient of that award was a non-Medicare patient. The amount of denied Medicare reimbursement at issue for Humana is approximately $37,000. On appeal to the district court, Judge Ackerman adopted Judge Will’s opinion holding the Malpractice Rule invalid and granted Humana’s motion for summary judgment. Humana of Illinois, Inc. v. Heckler, 584 F.Supp. 618, 619 (C.D.Ill.1984).
II.
In assessing the validity of the Malpractice Rule on appeal, we will address three separate, although somewhat interrelated, questions. The first question is whether the Secretary’s action was substantively “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law,” under the judicial review provisions of the Administrative Procedure Act (“APA”), 5 U.S.C. §§ 551-559, 701 et seq. (1982). 5 U.S.C. § 706(2)(A) (1982). The second inquiry is whether the Rule complies with the procedural requirements of the APA, in particular the statement of basis and purpose required by 5 U.S.C. § 553(c). 5 U.S.C. § 706(2)(D) (1982). The last issue is whether the Rule violates the Medicare Act. Cases invalidating the Malpractice Rule, see supra note 1, have based their decisions variously on some or all of the above rationales. We begin by discussing whether the Secretary violated the APA in promulgating the Rule, first substantively and then procedurally, and then move on to address whether the Rule violates the Medicare Act.
A. Arbitrary and Capricious
The scope of review under the “arbitrary and capricious” standard is narrow, for we are not permitted to substitute our own judgment for that of the agency. Motor Vehicle Manufacturers Ass’n v. State Farm Mutual Automobile Insurance Co., 463 U.S. 29, 103 S.Ct. 2856, 2866, 77 L.Ed.2d 443 (1983). The agency is required, however, to “examine the relevant data and articulate a satisfactory explanation for its action including a ‘rational connection between the facts found and the choice made.’ ” Id. 103 S.Ct. at 2866-67. An agency rule may 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.
Id. 103 S.Ct. at 2867. See also Citizens to Preserve Overton Park, Inc. v. Volpe, 401 U.S. 402, 415, 91 S.Ct. 814, 823, 28 L.Ed.2d 136 (1971) (presumption of regularity afforded agency’s decision does not shield it from a “thorough, probing, in-depth review”); City of West Chicago v. United States Nuclear Regulatory Commission, [1466]*1466701 F.2d 632, 648 (7th Cir.1983). There is no contention here that the Secretary relied on factors other than those Congress intended her to consider. The purpose of this review, therefore, is to determine whether the Secretary entirely failed to consider important aspects of the malpractice problem or offered an explanation for the Malpractice Rule that runs counter to the evidence before the agency.
The hospitals’ attack on the Malpractice Rule as arbitrary and capricious focuses primarily on the Westat study, since the Secretary issued the Rule based almost exclusively on the findings in the study. The hospitals argue that the Westat study, and hence the rulemaking record, is so deficient that the Secretary’s implementation of the Rule was arbitrary and capricious.
The Westat study was sponsored by the Department of Health, Education, and Welfare (predecessor agency to the Department of Health and Human Services) to analyze the frequency and type of malpractice claims and awards between July 1, 1976, and October 31, 1976. The study had a very broad scope, its general purpose being to determine the factors influencing the frequency of occurrence of malpractice claims and the size of awards, so that the level and rate of increase of medical malpractice insurance premiums might be reduced in order to achieve the general policy objective of controlling the cost of health care for all patients, not just Medicare patients. Only one of the study’s eight chapters addressed Medicare issues.
Westat’s analysis of Medicare claims as compared to non-Medicare claims led it to conclude that although Medicare paid for 21% of the hospital discharges in 1973, “only a maximum of 13 percent of the malpractice claims closed in 1976 (for incidents, most of which occurred in 1973) were filed by persons for whom Medicaid [sic] paid for their medical care.” In addition to finding that Medicare patients filed fewer claims, the study concluded that “the average award for Medicare patients is considerably smaller than for other groups.” With regard to Medicare reimbursement of malpractice insurance costs, however, Westat only surmised that “[i]f the percentage of Medicare patients involved in medical malpractice claims is proportionately smaller than the percentage of hospitalizations for which Medicare paid, then Medicare may be subsidizing malpractice insurance costs that should actually be covered by other payors” (emphasis in original).
The hospitals first point out the limited scope of the Westat study and argue that the data base was insufficient to justify reliance on the study’s conclusions. As noted above, the study surveyed medical malpractice claims closed during a single four-month period in 1976. Data was acquired from only nine of more than fifty private insurance carriers for the four-month period. It appears, however, that insureds from those nine companies were involved in 84% of all malpractice claims closed by private carriers in 1976. Although claims paid by private insurance carriers thus seem to be adequately represented, the use of data from these private carriers alone means that self-insuring providers (usually hospitals) were excluded from the study, as were federal institutions. A study conducted by the American Hospital Association (cited in the Westat study) revealed that in 1976, approximately 20% of non-federal community hospitals in the United States were either partially or totally self-insured. Claims from the totally self-insured hospitals were not reflected in the Westat study, and those from the partially self-insured hospitals were underrepresented.
The authors of the Westat study identified a second source of bias in addition to the study’s exclusion of self-insured providers — the fact that Westat did not have complete data on the identity of all defendants in multiple defendant cases. Although the Secretary now speculates that because these omissions relate to defendant identity they are not relevant to the Medicare chapter, where the key factor was plaintiff identity, the study’s authors made no such distinctions. In fact, as will be discussed [1467]*1467shortly, the Secretary should have paid particular attention to the study’s warnings on this source of bias because defendant identity was indeed an important factor in this rulemaking.
Because of the above-noted limitations on the scope of the Westat study, the study’s data base consisted of a total of 3,932 medical malpractice claims. Information on the source of payment for medical care was available for only 2,637 (approximately two-thirds) of the claimants. Of those, only 340 claimants, or approximately 12%, were identified as Medicare recipients. With respect to the findings on the size of awards made to Medicare claimants, the pool of information was even smaller. Because less than half of all malpractice claims resulted in awards (for Medicare and non-Medicare claimants alike), award information for Medicare claimants was based on only 166 claims.
The authors of the Westat study cautioned at the outset that readers should keep in mind the limited nature of the study and certain possible sources of bias. With regard to one particular chapter — injury prevention — the authors warned that “[cjonelusions must be drawn very cautiously because of the possibility of biased data.” At least one court has noted that the same cautions would seem to apply to all conclusions drawn from the study. See Boswell v. Heckler, 749 F.2d 788, 796 & n. 11 (D.C.Cir.1984).
In addition to the cautions contained in the study itself, Westat sent a memorandum to the agency on August 17, 1978, noting deficiencies in the Westat study and making recommendations for future research. The memorandum recommended a more precise sample design if “national estimates or ‘representative data’ ” were desired, pointing out that with the sample design used in the 1976 Westat study, “it was virtually impossible to generalize beyond the sample — to annual totals or to national totals, to cite the most elementary limitations.” The authors further stated that “there was no assurance [in the 1976 study] that the sample was in any way representative, of hospitalsf,] of physicians, of injuries, of geographic regions, of claimants.” Notwithstanding this memorandum, the Secretary issued the proposed Malpractice Rule seven months later, to have nationwide effect, based almost exclusively on the findings in the Westat study.
The most significant source of bias in the Westat study with respect to the Malpractice Rule, however, is one not noted in the study itself — the fact that the data base included claims against both hospitals and practitioners.5 Westat based its Medicare analysis on the number of claims filed for injuries occurring in hospitals rather than claims filed against hospitals. The study found that although hospitals were reported as the site of injury in 78% of all claims in the study, however, hospitals were named defendants in only 32%. Thus, Westat’s use. of claims for all injuries occurring in hospitals included a fair number of claims filed against physicians for injuries occurring in hospitals.
Claims against hospitals represented only 32% of the claims in Westat’s data base, and only 20% of claim dollars in the study. Since the Malpractice Rule is based on hospital claims experience only, the study’s inclusion of data on claims against physicians permits reliable conclusions with respect to Medicare claimants only if they file claims against hospitals as opposed to physicians in the same proportion as do non-Medicare claimants. This, however, does not appear to be the case. The Westat study itself indicates that 69% of the [1468]*1468injuries complained of in paid claims filed by Medicare patients were “institutional problems” (including fluid/specimen administration, falls, medical management, and equipment problems), as compared to 35% for all paid claims. This data suggests that Medicare recipients file a much greater proportion of claims against hospitals than do other patient populations, which tend to file more claims against physicians.
Reanalyzing Westat’s data to take account of this potential source of bias produces a startling effect. Westat’s analysis concluded that although Medicare paid for 21% of hospital discharges in 1973, only a maximum of 13% of malpractice claims closed in 1976 were filed by Medicare patients, representing 5.1% of dollars paid. When claims against hospitals only are isolated, however, Medicare patients accounted for 20.2% of claims paid and 17.8% of dollars paid.6 When Westat’s data is further refined to exclude claims against hospital-employed physicians (since these claims are excluded from the Malpractice Rule’s paid-claim ratio), Medicare patients accounted for 22.5% of claims paid and 19.9% of dollars paid. This result seriously undermines the study’s conclusion, heavily relied on by the Secretary in formulating the Malpractice Rule, that the percentage of Medicare patients involved in malpractice claims is substantially smaller than the percentage of hospitalizations (and hence the percentage of indirect G & A costs such as malpractice insurance premiums) paid for by Medicare.
In summary, we agree with the district courts below that the Westat study furnished an inadequate basis for the Malpractice Rule and that the Secretary committed a clear error of judgment in relying on the study as the sole empirical basis for the Rule. See St. James Hospital v. Heckler, 579 F.Supp. at 766. See also Humana of Aurora, Inc. v. Heckler, 753 F.2d 1579, 1583 (10th Cir.1985) (“When an agency adopts a regulation based on a study not designed for the purpose and which is limited and criticized by its authors on points essential to the use sought to be made of it, the administrative action is arbitrary and capricious and a clear error in judgment.”).
Compounding the statistical limitations of the study itself, moreover, the Secretary promulgated the Rule without investigating several other important aspects of the malpractice problem, primarily the extent to which any disproportionality in the amount paid on malpractice claims filed by Medicare recipients affected the malpractice insurance costs of providers. The public comments received by the Secretary uniformly criticized the Malpractice Rule’s use of actual loss experience as the sole measure of reimbursement for malpractice insurance premiums. Several insurance companies commented that the proposed rule was “illogical” because “past malpractice award experience for a particular patient group does not fairly reflect current malpractice costs attributable to the group.” The insurance carriers noted that the cost of insurance depends on factors such as the size of the provider, the number of patients it services, the types of services it provides, and other factors in addition to past claim experience. Other commenters pointed out that administrative expenses such as risk management services, acquisition costs of policy issuance, accounting services, agents’ commissions, licensing fees and taxes, unallocated claims expense, and the general administrative overhead expenses of the insurer account for anywhere from 30% to over 50% of premium costs.
The Secretary now belatedly attempts to justify the Malpractice Rule’s use of loss experience to reimburse malpractice premium costs by citing to some commenters’ statements that despite the various other factors considered when setting premium rates, loss payment is the largest single element of the malpractice [1469]*1469premium. The agency itself, however, made no attempt to examine the relationship between actual loss experience and premium costs before promulgating the Malpractice Rule, apparently assuming that the relationship was strong enough to support the Rule’s use of loss experience as the sole measure of premium costs.7 The Secretary now contends that a few public comments, in the course of criticizing the Malpractice Rule’s reliance on this very relationship, confirm that assumption and that the relationship was therefore adequately established in the rulemaking record. This is clearly an insufficient basis upon which to rest a regulation with such substantial economic effects on health care providers. We therefore find that even if the Westat study was statistically valid and its findings reliable, promulgation of the Malpractice Rule was arbitrary and capricious because the Secretary completely failed to ascertain the extent of the correlation, if any, between actual malpractice losses and premium costs.
The Secretary failed to investigate another aspect of the malpractice problem as well. In deciding to remove malpractice insurance costs from the G & A pool, the Secretary neglected to first determine whether the existing pool adequately compensated for any imbalance in the allocation of malpractice costs. This oversight will be more fully discussed in section C, infra.
In sum, we find that the Secretary entirely failed to consider several important aspects of the malpractice problem before promulgating the Malpractice Rule, and that the explanation given for the Rule is not adequately supported by the Westat study, the only statistical evidence before the agency. We therefore find the Rule arbitrary and capricious.
B. Inadequate Basis and Purpose Statement
The Malpractice Rule is also challenged as violating the procedural requirements of the APA, particularly the requirement that “[ajfter consideration of the relevant matter presented [through notice and comment], the agency shall incorporate in the rules adopted a concise general statement of their basis and purpose.” 5 U.S.C. § 553(c) (1982).8 Although the agency need not respond to every fact or contention in the comments submitted, the basis and purpose statement must identify “what major issues of policy were ventilated by the informal proceedings and why the agency reacted to them as it did.” Automotive Parts & Accessories Ass’n v. Boyd, 407 F.2d 330, 338 (D.C.Cir.1968).
We find that the Secretary’s basis and purpose statement did not adequately respond to the criticisms raised in the many adverse comments submitted on the Malpractice Rule in several respects. First and foremost, the Secretary made no attempt to respond to comments that the Westat study was statistically unreliable and that its own authors had cautioned against its use in the manner contemplated by the Secretary. The Secretary’s post hoc attempts to respond to those comments on appeal are insufficient and will not remedy the failure to do so in the basis and purpose statement. See Motor Vehicle Manufacturers Ass’n v. State. Farm Mutual Automobile Insurance Co., 463 U.S. 29, 103 S.Ct. 2856, 2870, 77 L.Ed.2d 443 (1983).
Commenters also criticized the Malpractice Rule on the ground that the accepted insurance industry practice is to calculate premiums on the basis of occupied beds, regardless of the Medicare status of the patient. Thus, the commenters noted, the Malpractice Rule’s formula for reimbursing premium costs bears no relation to the actual premium costs incurred by the hospi[1470]*1470tais on behalf of Medicare recipients. In response to these criticisms, the Secretary stated: “We are aware that insurance companies generally do not determine insurance rates for malpractice insurance based upon the financial status of the patients. However, information about malpractice loss experience by type of patient (i.e., Medicare) is readily available.” 44 Fed. Reg. 31,642 (June 1, 1979). Besides being nonresponsive, this statement ignores the fact that in the Secretary’s own Westat study, the insurance companies had no information on the Medicare status of the claimant for a full one-third of the claims filed.
The opportunity under the APA to comment on proposed rules is “meaningless unless the agency responds to significant points raised by the public.” Home Box Office, Inc. v. FCC, 567 F.2d 9, 35-36 (D.C.Cir.), cert. denied, 434 U.S. 829, 98 S.Ct. Ill, 54 L.Ed.2d 89 (1977). We agree with the district courts’ conclusion that the Secretary’s basis and purpose statement failed to respond to many significant points made by the public in opposition to the Malpractice Rule. The Secretary’s statement of basis and purpose provided no indication of why criticisms of the Westat study were deemed to be invalid, and failed to give a reasoned response to other criticisms of the Rule. We therefore conclude that the Secretary failed to comply with section 553(c) of the APA.
C. Violation of the Medicare Act
The last question for review is whether the Malpractice Rule is contrary to provisions of the Medicare Act. Although the Secretary is ordinarily given a great deal of deference in interpreting and implementing the laws that she is directed to administer, see Heckler v. Campbell, 461 U.S. 458, 103 S.Ct. 1952, 1957, 76 L.Ed.2d 66 (1983) (involving Secretary’s interpretation of Social Security benefit provisions); Batterton v. Francis, 432 U.S. 416, 425, 97 S.Ct. 2399, 2405, 53 L.Ed.2d-448 (1977) (same), a lesser degree of deference is required when reviewing the Secretary’s actions under the Medicare Act’s reimbursement provisions. See Northwest Hospital, Inc. v. Hospital Service Corp., 687 F.2d 985, 990-91 (7th Cir.1982) (Medicare statute specifically circumscribes Secretary’s discretion to define ‘reasonable costs’); St. John’s Hickey Memorial Hospital, Inc. v. Califano, 599 F.2d 803, 813 (7th Cir.1979) (Secretary’s authority under Medicare Act significantly more circumscribed than authority delegated in Social Security benefit cases); Mt. Carmel Mercy Hospital v. Heckler, 581 F.Supp. 1311, 1313-15 (E.D. Mich.1983) (Secretary’s discretion limited in this area). The regulations promulgated under the Medicare Act “must ... take into account both direct and indirect costs and must avoid the result of shifting costs to non-Medicare patients.” St. John’s Hickey Memorial Hospital, 599 F.2d at 813 n. 17 (emphasis added).
The first provision of the Act at issue is the prohibition against shifting Medicare costs to non-Medicare patients and vice versa. 42 U.S.C. § 1395x(v)(l)(A) (1982); see also Northwest Hospital, Inc. v. Hospital Service Corp., 687 F.2d 985, 989 (7th Cir. 1982). As mentioned previously, the Secretary’s mechanism for reimbursing providers’ indirect costs has consistently been to pool them together as G&A costs and then to reimburse a percentage of those costs on the basis of the provider’s Medicare patient utilization ratio. This regulatory scheme was not premised on a belief that all G & A costs applied to all of a hospital’s patients equally, but rather on the assumption that some costs would be disproportionately allocated in favor of Medicare patients while others would be disproportionately allocated in favor of non-Medicare patients. Pooling all of the various costs together was believed to be the fairest and most administratively feasible way of allocating indirect costs among Medicare and non-Medicare patients. In addition, although there may have been cost-shifting with respect to a certain item (i.e., billing expenses associated predominately with the Medicare program), the G&A pool as a whole was thought to fairly allocate indirect costs among Medicare and non-Medicare popula[1471]*1471tions such that neither sector impermissibly paid the costs of the other. The Secretary pulled malpractice insurance premiums out of the G&A pool based on two findings: first, the cost of malpractice insurance skyrocketed in the 1970’s, and second, the Westat study led the Secretary to believe that Medicare was paying more than its fair share of that cost. See 44 Fed.Reg. 31,642 (June 1, 1979). The problem with this analysis is the findings that were not made.
Even assuming that Medicare paid a disproportionate amount of malpractice insurance costs when they were included in the G&A pool, the Secretary seems to have forgotten that the G&A pool was designed to accommodate such imbalances. What is missing in the Secretary’s promulgation of the Malpractice Rule is the necessary starting premise: that the G&A pool including malpractice insurance costs, taken as a whole, impermissibly shifted costs attributable to non-Medicare patients to the Medicare program. See Boswell v. Heckler, 749 F.2d at 795. If the starting point was otherwise, the removal of a specific cost that was disproportionately attributable to non-Medicare patients would inevitably result in an imbalanced pool, in violation of the Act. Id. The Secretary did not explicitly make this essential preliminary finding.
One could argue that in instigating this rulemaking at all, the Secretary must have made an implicit finding that the then-existing method of allocating indirect costs was inequitable. The problem with this argument, however, is that the Secretary had no factual support for that finding. Although the Secretary had support for the finding that malpractice insurance premiums increased dramatically during the 1970’s, for example, no attempt was made to determine whether and to what extent other indirect costs in the G&A pool, such as legal fees, also increased over the same time period. See Boswell v. Heckler, 749 F.2d at 795 n. 8. The Secretary therefore had no information indicating the amount of change in the percentage of the G&A pool representing malpractice insurance costs. Without that information, it was impossible to determine whether the G&A pool as a whole impermissibly shifted costs from the non-Medicare to the Medicare populations. Therefore, the Secretary could not have made the necessary finding on the information before the agency. See Alexandria Hospital v. Heckler, 586 F.Supp. 581, 589 (E.D.Va.1984) (neither Westat report nor any evidence in rulemaking record supports Secretary’s conclusion that supposed disproportionate allocation of malpractice costs is so great as to defeat the operation of the G&A pool’s averaging principle).
The Secretary’s action in removing malpractice insurance costs from the G&A pool is particularly suspect in light of the agency’s long-standing resistance to providers’ attempts to isolate other costs disproportionately attributable to Medicare and thus under-reimbursed under the pooling system. See PRRB Decision No. 79-D53, 1979-2 Medicare & Medicaid Guide (CCH) ¶ 30,086, at 9707 (“Program policy has consistently maintained that it is impractical to isolate discrete items of [G&A] cost between Medicare and non-Medicare beneficiaries and has posited that all such discrete items making up total [G&A] costs will ‘average out.’ ”); Good Luck Nursing Home, Inc. v. Califano, 1979-1 Medicare & Medicaid Guide (CCH) ¶ 29,450, at 9287 (D.D.C.1978) (finding arbitrary and capricious the PRRB’s refusal to remove from the G&A pool certain legal and accounting fees incurred solely as a result of provider’s participation in Medicare program); PRRB Decision No. 76-D74, 1976 Medicare & Medicaid Guide (CCH) U 28,300, at 10,-609-10 (refusing to remove from G&A pool certain extraordinary legal fees attributable solely to Medicare program); Provider Appeal Decision No. 00-75-04, 1975 Medicare & Medicaid Guide (CCH) 27,361, at 9589 (refusing to remove from the G & A pool certain expenses incurred in preparing Medicare cost reports); Intermediary Letter No. 234 (June 2, 1967) (including in G&A pool all costs of Medicare utilization review, including compensation paid to physicians for their services on utilization re[1472]*1472view committees). The agency has long believed that
[t]he requirement in the law that our costs are not to be borne by others is construed to refer to total costs, not to any one cost. In our cost finding procedures ... it is presumed 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. In this manner, the program is presumed to bear its proper share of the total allowable costs.
Intermediary Letter No. 234 (June 2, 1967). The agency continues to hold this view with respect to G&A costs other than malpractice insurance premiums. See PRRB Decision No. 82-D46,1982 Medicare & Medicaid Guide (CCH) ¶ 31,810, at 9073.
Although an agency is not rigidly bound to its own precedent, the presumption is against changes in established policy that are not justified by the rulemaking record. See Motor Vehicle Manufacturers Ass’n v. State Farm Mutual Automobile Insurance Co., 463 U.S. 29, 103 S.Ct. 2856, 2866, 77 L.Ed.2d 443 (1983). We conclude that the Secretary violated the cost-shifting prohibition of the Medicare Act by removing malpractice insurance costs from the G&A pool without first determining, on the basis of an adequate factual record, that the existing system itself inequitably shifted costs and that removal of malpractice costs from the pool would rectify the inequity.
A second way in which the Malpractice Rule violates the Medicare Act is its failure to recognize that malpractice insurance protects against the risk of future loss. Even if a provider has never incurred any actual malpractice losses, for example, it must still purchase malpractice insurance because of the risk that losses will be incurred in the future. The Provider Reimbursement Manual recognizes this reality and thus provides that hospitals are expected to maintain an adequate insurance program to protect against likely losses, particularly losses so great that their financial stability would be threatened. See also PRRB Decision No. 82-D89, 1982 Medicare & Medicaid Guide (CCH) ¶ 31,987, at 9814 (malpractice insurance carried by provider to protect its assets). Providers may also be required to insure against malpractice loss under state law. The effect of the Malpractice Rule, however, is to reimburse these necessary costs only to the extent that the provider has incurred claim loss in the past. This ignores the fact that insurance premiums must be paid regardless of whether actual losses are incurred. See Mt. Carmel Mercy Hospital v. Heckler, 581 F.Supp. at 1315-16. The carrying of malpractice insurance must be deemed a reasonable cost, and thus reimbursable by Medicare, regardless of whether a hospital has paid one dollar or one million dollars in malpractice claims over the relevant five-year period. To the extent that the Malpractice Rule does not reimburse these costs, therefore, it violates the Medicare Act’s mandate that providers are entitled to government reimbursement for the “reasonable cost” of the services they provide for Medicare patients. 42 U.S.C. § 1395f(b)(l) (1982).
The fact that the Secretary promulgated the Malpractice Rule on the basis of a biased and statistically questionable study and without adequately investigating several key aspects of the malpractice problem, as discussed in section A, supra, is closely tied to our finding that the Rule violates these two provisions of the Medicare Act. Because of the deficiencies in the rulemaking process, reimbursement under the Malpractice Rule seems to bear no reasonable relation to the portion of a provider’s actual malpractice insurance costs attributable to Medicare patients. As one court has stated, the Malpractice Rule prescribes a formula “which seemingly guarantees that the statutory mandate will inevitably be violated in most cases.” Abington Memorial Hospital v. Heckler, 576 F.Supp. 1081, 1089 (E.D.Pa.1983), aff'd, 750 [1473]*1473F.2d 242, 243 (3d Cir.1984) (adopting district court’s opinion). See also Mt. Carmel Mercy Hospital v. Heckler, 581 F.Supp. at 1317 (“merely fortuitous” whether a hospital is properly reimbursed for its malpractice insurance costs under the Malpractice Rule). When reimbursement bears no reasonable relation to actual costs, as in Humana’s case (over $85,000 spent for malpractice insurance premiums in cost year 1980 but nothing reimbursed), the Act is violated both because of impermissible cost-shifting and because, in cases of under-reimbursement, the provider is denied reimbursement for the reasonable cost of a service it provides for Medicare patients.
In sum, we agree with Judge Will’s conclusion that
the Secretary is not, of course, limited to any one method of reimbursement and may use different methods for different costs and circumstances. This the statute clearly allows. 42 U.S.C. § 1395x(v)(l)(A). But when, as here, the chosen method neither assures that Medicare costs will not be borne by non-Medicare patients nor that hospitals will be reimbursed the reasonable costs of their services ... then the regulation does not carry out the mandate of the Medicare statute and is invalid.
St. James Hospital v. Heckler, 579 F.Supp. at 767.
III.
We conclude that (1) the Secretary’s promulgation of the Malpractice Rule was arbitrary and capricious and an abuse of discretion, (2) the Secretary failed to provide an adequate basis and purpose statement for the Rule, and (3) the Rule violates certain provisions of the Medicare Act. In reaching this result, we are not unmindful of the deference afforded the Secretary when reviewing administrative decisions such as the one at issue in this case. However permissible, and even laudable, might be the Secretary’s desire to ferret out unnecessary expenditures in the Medicare program, however, there are certain statutory requirements that Congress has determined must be met before the Secretary can take steps to achieve that goal. Because those requirements were not met with respect to the Malpractice Rule, we must set aside the Rule and affirm the district courts’ grant of summary judgment to the plaintiff hospitals.9