RIPPLE, Circuit Judge.
This appeal arises out of a Medicare reimbursement dispute between the Board of Trustees of Knox County, d/b/a Good Samaritan Hospital (“Good Samaritan”), and the Secretary of the Department of Health and Human Services (“Secretary”). In this case, the Health Care Financing Administration (“HCFA”) denied Good Samaritan’s request for increased reimbursement under a program aimed at ensuring that large rural hospitals receive Medicare reimbursement similar to that received by their urban counterparts. Good Samaritan then sought review of the HCFA’s decision before the Provider Reimbursement Review Board (“PRRB”), the administrative body charged with hearing Medicare reimbursement disputes between hospitals and their Medicare fiscal intermediaries. The PRRB ruled in favor of Good Samaritan. However, the Secretary later reversed the decision of the PRRB. Good Samaritan then appealed the Secretary’s decision to the district court. The parties filed cross-motions for summary judgment, and the district court entered summary judgment in favor of the Secretary. Good Samaritan now appeals to this court. For the reasons set forth in the following opinion, we affirm the judgment of the district court.
I
BACKGROUND
A.
At the outset, a review of the statutory and regulatory framework is helpful. The Medicare program reimburses hospitals and other health care providers for a portion of the expenses they incur in treating the aged and disabled. Medicare providers are reimbursed by the federal government, usually through private organizations acting as “fiscal intermediaries” under contract with the Secretary.
Initially, the government, through the HCFA, reimbursed Medicare providers based on a retrospective or reasonable cost basis. In order to receive reimbursement under this regime, providers were required to submit several documents to both the HCFA and their fiscal intermediaries detailing the costs incurred in treating Medicare beneficiaries. See 47 Fed.Reg. 43,296, 43,304 (1982). Most notably, providers were required to submit a “Medicare cost report” which contained information concerning the costs the provider incurred treating Medicare beneficiaries during a designated cost reporting period. Id. The information contained in those reports was then used by the fiscal intermediaries to calculate the total amount of Medicare reimbursement to which the hospital was entitled for that cost period. Id. In addition, providers were required to submit detailed billing information for each Medicare inpatient and a narrative description of the services provided to certain randomly-selected Medicare beneficiaries. Id. This information was then compiled by the HCFA and maintained in a statistical file called “MEDPAR.” Id.
Under the reasonable cost regime, providers were reimbursed for the actual costs incurred in treating Medicare beneficiaries as long as those costs fell within certain parameters. Consequently, as hospital costs increased, so too did Medicare reimbursements. As Medicare costs soared in the 1970s and early 1980s, members of Congress became concerned that the reasonable cost reimbursement regime, fueled by the dramatic increase in hospital costs,1 was placing undue strain on the federal budget. There was concern that reimbursement under the reasonable cost regime was tantamount to giving hospitals a blank cheek to cover the cost of care for Medicare beneficiaries because hospitals had no incentive to render [496]*496cost efficient treatment to Medicare beneficiaries.2
Congress modified the reasonable cost regime with the passage of the Tax Equity and Fiscal Responsibility Act of 1982 (“TEFRA”). See Pub.L. No. 97-248, 96 Stat. 324 (1982). In TEFRA, Congress enacted several measures aimed at curbing hospital costs.3 One of these measures was the placement of a cap on the amount of reimbursement allowed each hospital. One factor used in setting this cap was a case-mix index (“CMI”) specific to each hospital. See 42 U.S.C. § 1395ww(a)(l)(B)(i). The CMI is a mathematical representation of the complexity (i.e., the cost) of the average case treated at a given hospital. More specifically, it represents the cost of each hospital’s particular mix of cases in comparison to the cost of the mix of cases at the average American hospital. To determine a CMI for each hospital provider, the Secretary relied upon the information collected under the original reasonable cost regime: the MEDPAR file and the annual Medicare cost reports. Using this information, the Secretary calculated a 1981 CMI for all providers. These CMIs were published in the Federal Register in September 1983.
Although the TEFRA regime itself was only a modification of the reasonable cost regime, Congress also directed the Secretary to develop a prospective system of Medicare reimbursement by December 31, 1982. See TEFRA, Pub.L. No. 97-248, § 101(a)(1), 96 Stat. 324, 335 (1982). On December 27,1982, the Secretary submitted a report to Congress entitled “Hospital Prospective Payment for Medicare.” See 48 Fed.Reg. 39,753, 39,-754 (1983). In April 1983, Congress voted to substitute the Prospective Payment System (“PPS”) for the reasonable cost regime. See Social Security Amendments of 1983, Pub.L. No. 98-21, § 601, 97 Stat. 65, 149 (1983). Accordingly, since the beginning of hospital fiscal years commencing on or after October 1, 1983, Medicare reimbursement has been administered under the PPS regime.4
Under PPS, hospitals are reimbursed according to prospectively fixed rates.5 These rates are determined by the type of treatment involved in a particular case6 with certain adjustments made to account for the varying costs associated with the location of the hospital (i.e., urban or rural). For example, a provider will be able to determine in advance the amount of Medicare reimbursement it will receive for a coronary bypass with cardiac catheterization as opposed to a coronary bypass without cardiac catheterization.7 Thus, in contrast to hospital reimbursements under the reasonable cost method, hospitals are reimbursed according to the applicable fixed rate regardless of the costs actually incurred. Congress’ purpose in implementing this new regime was to encourage Medicare providers to improve efficiency and reduce operating costs.8
[497]*497Shortly after the implementation of the PPS regime, Congress directed the Secretary to provide for sufficient “exceptions and adjustments” to providers’ Medicare reimbursement to account for “the special needs of regional and national referral centers.” 42 U.S.C. § 1395ww(d)(5)(C)(i). Most notably, Congress directed the Secretary to allow rural hospitals to seek to be designated as a Rural Referral Center (“RRC”) based on criteria established by the Secretary. Id. The term RRC denotes a rural hospital operating with the characteristics and costs similar to those of an urban hospital. Accordingly, hospitals designated as RRCs are reimbursed based on the payment rate for urban areas.
Pursuant to Congress’ directive, the Secretary provided three alternate sets of criteria under which a provider could qualify as an RRC. See 42 C.F.R. § 405.476(g)(l)(i), (ii), (iii) (1984).9 Under the first set of criteria, a hospital qualified as an RRC if it was located in a rural area and had 500 or more beds available for use. See § 405.476(g)(l)(i).10 Under the second of the three tests, a hospital could claim RRC status if it met three criteria: (1) an inpatient population such that at least 50% of its Medicare patients’ were referred from other hospitals or from physicians not on the staff of the hospital; (2) at least 60% of its Medicare patients lived more than 25 miles from the hospital; and (3) at least 60% of all the services it furnished to Medicare beneficiaries were furnished to beneficiaries who live more than 25 miles from the hospital. See § 405.476(g)(l)(ii).
In order to satisfy the third test (the one at issue in this case), a hospital, once again, had to be located in a rural area. It was required to meet or exceed the mandatory discharge criteria set forth at § 405.476(g)(l)(iii)(B). The hospital also had to satisfy one of three criteria pertaining to the composition of the hospital’s medical staff and the nature of its inpatient population. See § 405.476(g)(l)(iii)(C), (D) & (E). Finally, the hospital had to satisfy any one of four alternate CMI criteria: (1) a hospital was eligible for RRC status if its 1981 CMI was equal to or exceeded 1.03;11 (2) a hospital was eligible for RRC status if its 1981 CMI was equal to or exceeded the 1981 median urban CMI for the region in which the provider was located; (3) a hospital was eligible for RRC status if its CMI for the first cost reporting period under PPS was equal to or exceeded 1.1053 (the 1984 national adjusted median urban CMI); (4) a hospital could establish its eligibility for RRC status by showing that its CMI for the first cost reporting period under PPS was equal to or exceeded the 1984 adjusted median urban CMI for the region in which the hospital was located. See § 405.476(g)(l)(iii)(A)(l), (2), (3) &(4).
A hospital that wished to apply for RRC status was to submit its written request directly to the appropriate HCFA regional office. See 49 Fed.Reg. 34,728, 34,743 (1984). A provider had to submit such a request during the quarter immediately preceding the start of the fiscal year with respect to which it wished to claim an RRC adjustment. This case involves Good Samaritan’s request that the HCFA designate it an RRC for its 1985 fiscal year.
B.
Good Samaritan is a 342-bed hospital in Vincennes, Indiana. It offers many services comparable to major urban hospitals, including end-stage renal dialysis services, inpatient and outpatient surgical services and other resource-intensive services. Because Good Samaritan offers such sophisticated services, it must compete with its urban counterparts for employees, medical specialists and other support staff. As an additional result, it incurs costs similar to those of urban hospitals. Good Samaritan has quali[498]*498fied for RRC status for all fiscal years subsequent to 1985.
However, the HCFA denied Good Samaritan’s request to be designated as an RRC for its 1985 fiscal year. The HCFA’s denial of that request is the focus of this case. On December 17, 1984, Good Samaritan submitted a timely request for RRC designation. In its application, Good Samaritan sought to qualify under the third set of criteria set forth at 42 C.F.R. § 405.476(g)(l)(iii) (1984). As noted above, a hospital must meet four requirements to qualify as an RRC under the third test. It is undisputed that Good Samaritan met three of those requirements: It is located in a rural area, satisfied the requisite discharge criteria and satisfied the requisite criteria concerning the composition of its medical staff and the nature of its inpatient population. However, the parties dispute whether Good Samaritan satisfied the remaining criterion.
As we noted earlier, the final criterion requires a hospital to meet one of four CMI criteria in order to qualify for RRC status. See § 405.476(g)(l)(iii)(A). Despite the fact that the Secretary had calculated and published its 1981 CMI as 1.0232, Good Samaritan argued that it met the criterion requiring that a hospital’s 1981 CMI equal or exceed 1.03. See § 405.476(g)(l)(iii)(A)(l). Good Samaritan based this argument on a study of its 1981 Medicare discharges conducted by the Commission of Professional Hospital Activities (“CPHA”), a nationally recognized consulting firm. In that report, the CPHA concluded that Good Samaritan’s 1981 CMI was in fact 1.0637. Good Samaritan contended that the HCFA ought to accept the recalculated 1981 CMI because the CPHA study was more accurate than the Secretary’s calculation. The CPHA recalculation was based on a study of 100% of Good Samaritan’s 1981 Medicare discharges; the Secretary’s calculation, in contrast, was based in large part on the MEDPAR file which included information concerning only 20% of Good Samaritan’s 1981 Medicare discharges.
The HCFA responded to Good Samaritan’s request by a letter dated February 15, 1985. The Agency rejected Good Samaritan’s request for RRC designation for 1985 because the hospital’s published 1981 CMI did not satisfy the minimum requirement specified in § 405.476(g)(l)(iii)(A)(l). The HCFA further stated that it could not accept Good Samaritan’s recalculated CMI based on the Secretary’s policy against accepting such recalculations. That policy was established in 1984 and was announced in the Federal Register: “We cannot allow substitution of the CPHA data in lieu of our published 1981 case-mix index because these data are developed and published by a source totally independent of HCFA.” 49 Fed.Reg. 34,728, 34,744 (1984).
Despite the HCFA’s ruling, Good Samaritan claimed the RRC adjustment and credited itself $1,262,926 on its Medicare cost report for 1985. The fiscal intermediary audited that report and debited the hospital $1,262,926 to disallow the additional reimbursement resulting from RRC status. Good Samaritan then appealed that decision to the PRRB by a letter dated April 18, 1988. The PRRB conducted hearings on Good Samaritan’s appeal in January 1993, and in March 1995 it ruled that the hospital, by virtue of its recalculated 1981 CMI, had satisfied the criteria set forth at § 405.476(g)(l)(iii). On May 26, 1995, the Secretary, acting through the Administrator of the HCFA, reversed the PRRB’s decision.
Good Samaritan then appealed the Secretary’s decision to the United States District Court for the Southern District of Indiana. The hospital’s primary argument before the district court was that the Secretary’s policy of refusing to consider a recalculation of a provider’s 1981 CMI based on a 100% discharge study was arbitrary and capricious. Upon the parties’ cross-motions for summary judgment, the district court entered summary judgment in favor of the Secretary. The district court held that the Secretary “did not act in a manner which was arbitrary and capricious, an abuse of discretion, or contrary to law when she determined that Good Samaritan’s CMI did not satisfy the regulatory requirement.” Board of Trustees of Knox County Hosp. v. Shalala, 959 F.Supp. 1026, 1028 (S.D.Ind.1997).
[499]*499II
DISCUSSION
We review the district court’s decision in this ease de novo. See Hinsdale Hosp. Corp. v. Shalala, 50 F.3d 1395, 1399 (7th Cir.1995). However, our review of the Secretary’s decision is limited. Id. Judicial review of the Secretary’s decision is governed by 42 U.S.C. § 1395oo(f), which incorporates .the standard of review of the Administrative Procedure Act. Id. Under the APA, a court may set aside an agency determination as unlawful only if it determines the agency action to be “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with the law.” 5 U.S.C. § 706(2)(A). Moreover, “[w]e must give substantial deference to the agency’s interpretation of its own regulations.” Thomas Jefferson Univ. v. Shalala, 512 U.S. 504, 512, 114 S.Ct. 2381, 2386, 129 L.Ed.2d 405 (1994). Indeed, as the Supreme Court noted, “[t]his broad deference is all the more warranted when, as here, the regulation concerns ‘a complex and highly technical regulatory program,’ in which the identification and classification of relevant ‘criteria necessarily require significant expertise and entail the exercise of judgment grounded in policy concerns.’ ” Id. (quoting Pauley v. BethEnergy Mines, Inc., 501 U.S. 680, 697, 111 S.Ct. 2524, 2534, 115 L.Ed.2d 604 (1991)).
Good Samaritan urges us to reverse the Secretary’s decision on the basis of three separate arguments.
1.
First, Good Samaritan contends that the Secretary’s interpretation of the RRC statute is arbitrary, capricious and contrary to law. Specifically, the hospital asserts that the Secretary’s interpretation of the RRC statute is inconsistent with its statutory language and basic purpose and that the RRC statute grants Good Samaritan the specific right to challenge the Secretary’s calculation of its 1981 CMI.
At the time of Good Samaritan’s application for RRC status, the statute read in relevant part:
The Secretary shall provide for such exceptions and adjustments to the payment amounts established under this subsection ... as the Secretary deems appropriate to take into account the special needs of regional and national referral centers (including those hospitals of 500 or more beds located in rural areas). A hospital which is classified as a rural hospital may appeal to the Secretary to be classified as a rural referral center under this clause on the basis of criteria (established by the Secretary) which shall allow the hospital to demonstrate that it should be so reclassified by reason of certain of its operating characteristics being similar to those of a typical urban hospital located in the same census region....
42 U.S.C. § 1395ww(d)(5)(C)(i) (1985). The language of the statute clearly directs the Secretary to establish the criteria upon which RRC determinations would be based. Pursuant to this statutory mandate, the Secretary issued regulations setting forth three alternate sets of criteria under which a provider could qualify as an RRC. See 42 C.F.R. § 405.476(g)(l)(i), (ii) & (iii) (1984).
Good Samaritan elected to make its appeal for RRC status under the third set of criteria promulgated by the Secretary. Under that test, one of the criteria a provider seeking RRC status must meet is a minimum CMI. In fact, under that criterion, a provider has a choice of four alternate CMI criteria. In its request for RRC status, Good Samaritan asserted that it met the first of the CMI criteria — a 1981 CMI greater than 1.03. In support of this assertion, Good Samaritan offered a private study of 100% of its 1981 Medicare discharges which concluded that Good Samaritan’s 1981 CMI was 1.0637. The Secretary, however, refused to accept Good Samaritan’s calculation of its 1981 CMI. Instead, the Secretary relied on her own calculation of Good Samaritan’s 1981 CMI which fell below the 1.03 threshold; accordingly, the Secretary denied Good Samaritan’s application for RRC status.
[500]*500The Secretary’s refusal to accept Good Samaritan’s calculation of its 1981 CMI was based on her well-established policy of refusing to consider a provider’s independent studies and relying on her own published calculation of a provider’s 1981 CMI when making RRC determinations. The Agency announced this policy and the reasons behind it in the Federal Register on August 31, 1984:
We do not believe that hospitals should be allowed to substitute other criteria for the one we published in the NPRM [notice of proposed rulemaking]. We selected the 1981 case-mix index for this criterion because it represents the most current published data available....
The basic tenet of the prospective payment system is that the rates paid to hospitals are determined prospectively and are based on the best data available at the time. Thus, a hospital knows in advance what its payment amounts will be. In addition, revisions to case-mix index values, whether upward or downward, would upset the budget neutrality adjustment. Therefore, we have denied any hospital the right to request review of its 1981 case-mix index value based on 100 percent of its cases. This policy holds whether the request for review is for purposes of meeting the referral center criteria or for determining the hospital-specific portion of a hospital’s payment rate.
We cannot allow substitution of the CPHA data [the type of study submitted by Good Samaritan in this case] in lieu of our published 1981 case-mix index because these data are developed and published by a source totally independent of HCFA. We have no way to verify either the raw data submitted by the hospital to CPHA or CPHA’s processing of the data except by performing a 100 percent review ourselves.
49 Fed.Reg. 34,728, 34,743-44 (1984).
The Agency’s reasoning is persuasive. Indeed, the 1981 CMIs provide the Secretary and the HCFA with a uniform criterion for judging which providers qualify for RRC status. In this regard, it is important to note that a provider’s CMI is a relative measure which is arrived at by comparing each provider’s case mix to national averages. Accordingly, an alteration in one provider’s CMI disrupts the entire scheme. Moreover, if one provider includes different data in its CMI calculation or employs a different methodology than that used by other providers, the CMI loses its utility as a comparative benchmark and could no longer serve as an equitable means of determining which providers should be given RRC status. Thus, the Secretary has a substantial interest in ensuring that each provider’s CMI is calculated in an accurate and uniform fashion. The published 1981 CMIs allow the Secretary to satisfy that interest without requiring her to scrutinize each hospital’s independent calculation of its CMI. In this way, the Secretary’s use of her own published calculation of a provider’s 1981 CMI as a criterion for RRC status serves interests of accuracy, uniformity and administrative convenience. We therefore conclude that the Secretary’s policy of relying solely on her own calculation of a provider’s 1981 CMI is not arbitrary and capricious.
2.
Good Samaritan’s second contention on appeal is that the Secretary’s policy not to allow providers to submit their own independent studies to demonstrate that they meet the requirements for RRC certification is, in fact, a substantive rule rather than an interpretive rule. Therefore, Good Samaritan submits, the Secretary was required to promulgate her policy through the notice and comment procedures of the APA. Because she did not follow those procedures, Good Samaritan argues that her policy is invalid and unenforceable.
The APA mandates that an agency follow its notice and comment procedures when promulgating a substantive rule. See Hoctor v. United States Dep’t of Agric., 82 F.3d 165, 167 (7th Cir.1996); Alabama Tissue Ctr. v. Sullivan, 975 F.2d 373, 377 (7th Cir.1992). A substantive rule is one “ ‘which ereate[s] law, usually implementary to an existing law.’” Alabama Tissue Ctr., 975 F.2d at 377 (quoting Indiana v. Sullivan, 934 F.2d 853, 856 (7th Cir.1991)). On the other hand, an agency is not bound by the APA’s [501]*501procedural requirements when announcing “interpretive rules, general statements of policy, or rules of agency organization, procedure, or practice.” See 5 U.S.C. § 553(b)(3)(A). An interpretive rule is a statement “‘as to what the administrative officer thinks the statute or regulation means.’ ” Alabama Tissue Ctr., 975 F.2d at 377 (quoting Indiana, 934 F.2d at 856). In this case, the Secretary’s policy of not allowing providers to submit their own independent studies to demonstrate that they qualify as an RRC represents the Secretary’s interpretation of her regulation establishing the RRC criteria and simply explains how the HCFA will apply it. We therefore conclude that the Secretary’s policy is an interpretive rule and that she was not required to follow the APA’s notice and comment procedures in its adoption.12
3.
Good Samaritan’s final contention is that the Secretary has applied her policy in an arbitrary and capricious manner. Specifically, Good Samaritan submits that the Secretary has allowed one provider, Providence Hospital (“Providence”), to qualify for RRC status on the basis of an independent recalculation of its CMI, but has denied Good Samaritan the opportunity to qualify for RRC status on the basis of the same type of evidence. Therefore, Good Samaritan argues that it has a right to equal treatment under the RRC statute. It urges this court to order the Secretary to extend that equal treatment to Good Samaritan by accepting its recalculated 1981 CMI as evidence that it qualifies for RRC status.
At the outset, we cannot accept the Secretary’s view that official correspondence memorializing this arrangement with Providence is inadmissible under Federal Rule of Evidence 408 because it is evidence of a settlement. It is not at all clear that the Secretary’s decision on the Providence matter constituted a “settlement.” However, even assuming arguendo that such a characterization is appropriate, the plain wording of the rule and the established case law make clear that the admission of this evidence is not precluded. As the record before the district court demonstrates, Good Samaritan offered evidence of this instance solely to dispute the Secretary’s assertion that she had administered the program even-handedly. The evidence, therefore, was not tendered “to prove liability for or invalidity of the claim or its amount.” Fed.R.Evid. 408. To the extent that the submission might be considered “[e]vidence of conduct or statements made in compromise negotiations,” it is admissible because it was offered for “another purpose” — to rebut the assertion of the Secretary.13
Turning to the merits of Good Samaritan’s final contention, we note that “[a]s [502]*502a general matter, ... the case for judicial deference is less compelling with respect to agency positions that are inconsistent with previously held views.” Pauley v. BethEnergy Mines, Inc., 501 U.S. 680, 698, 111 S.Ct. 2524, 2535, 115 L.Ed.2d 604 (1991); see also Bowen v. Georgetown Univ. Hosp., 488 U.S. 204, 212-13, 109 S.Ct. 468, 473-74, 102 L.Ed.2d 493 (1988). “When an agency begins to grant exceptions in certain cases ... interests represented in other eases can ask a court to review the denial of an exception as arbitrary in light of the agency’s past practice.” Basic Media, Ltd. v. FCC, 559 F.2d 830, 833 (D.C.Cir.1977); see also INS v. Yang, — U.S.-,-, 117 S.Ct. 350, 353, 136 L.Ed.2d 288 (1996) (“Though the agency’s discretion is unfettered at the outset, if it announces and follows — by rule or by settled course of adjudication — a general policy by which its exercise of discretion will be governed, an irrational departure from that policy (as opposed to an avowed alteration of it) could constitute action that must be overturned as 'arbitrary, capricious, [or] an abuse of discretion’ within the meaning of the Administrative Procedure Act, 5 U.S.C. § 706(2)(A).”).
Good Samaritan has the burden of demonstrating that the Secretary has administered the program in an arbitrary and capricious manner. See Sierra Club v. Marita, 46 F.3d 606, 619 (7th Cir.1995). In an effort to shoulder that burden, it presents only one instance in which the Secretary has recalculated a provider’s 1981 CMI for the purpose of determining whether that provider qualifies for RRC status.14 In that case, as we have already noted, Providence sought to be classified as an RRC for the 1985 fiscal year based on the same set of criteria under which Good Samaritan sought that status. Like Good Samaritan, Providence’s published 1981 CMI fell below the 1.03 threshold. However, Providence argued that its 1981 CMI was in fact 1.0691 and submitted to the HCFA a study conducted by the CPHA as support for this contention along with the underlying data used in that study (i.e., information concerning 100% of the hospital’s Medicare discharges). Although the record is unclear, it appears that HCFA used the underlying data submitted by Providence to recalculate Providence’s 1981 CMI and found that provider’s CMI to be 1.0482. It appears that the new calculation was based on 100% of Providence’s Medicare costs in the year in question. As a result of this recalculation, HCFA directed the fiscal intermediary in charge of Providence’s Medicare reimbursement to notify Providence that it qualified as an RRC for the 1985 fiscal year.
Because the Secretary recalculated Providence’s 1981 CMI, Good Samaritan argues that she must also recalculate its 1981 CMI based on a study of 100% of its 1981 Medicare discharges. We do not believe that Good Samaritan has met its burden of establishing that the Secretary has administered the program in an arbitrary and capricious manner. The singular deviation which Good Samaritan presents hardly allows the conclusion that the Secretary’s overall administration of the program is arbitrary and capricious.
Conclusion
For the reasons discussed in the foregoing opinion, we affirm the judgment of the district court.
AFFIRMED.