HCA Health Services of Kansas, Inc., d/b/a HCA Wesley Medical Center (Wesley), appeals the trial court’s judgment in favor of the Kansas Department of Social and Rehabilitation Services (SRS) for reimbursement of overpaid Medicaid claims. Wesley claims the methodology used by SRS in determining the recoupment amount was improper. SRS cross-appeals the trial court’s determination of the effective date of its policy change requiring written referrals. We affirm in part, reverse in part, and remand.
The Kansas Primary Care Network (KPCN) operates under a waiver issued by the Health Care Financing Administration (HCFA), which is part of the federal Department of Health and Human Services (HHS). Electronic Data Systems Federal (EDS) serves as fiscal agent for SRS in the Medicaid program. At the request of the KPCN, EDS audited a sample of Wesley’s outpatient Medicaid claims from April 1, 1987, through May 31, 1988.
*143An SRS representative testified that because of the large number of claims, EDS could only audit a sample of the claims paid. After examining the records of the sample, EDS calculated an error rate which was then applied to the total dollar amount of claims paid during the period to determine the amount that SRS had overpaid. The results of the audit showed the sample had an error rate of 45.7%. After applying that percentage figure to the total dollar amount of claims paid, EDS determined SRS had overpaid Wesley by $250,561.15.
Wesley challenged EDS’s findings at an administrative hearing where Wesley provided additional documentation on various claims. Although the overpayment amount was reduced to $140,851.55, Wesley appealed, requesting a statutory fair hearing. As a result of stipulations by the parties, the overpayment amount was reduced to $93,371.03 before the fair hearing. The amount was further reduced to $75,606.96 by the fair hearing officer, whose decision was later affirmed by the State Appeals Committee.
Next, Wesley filed a petition for judicial review of the hearing officer’s decision. Finding that the bulletin sent out in December 1985 furnished adequate notice that telephone referrals would no longer be acceptable, the trial court reversed part of the hearing officer’s decision. Following the trial court’s decision, both parties stipulated that the overpayment amount would be $68,753.54. Both parties appealed.
Because judicial review of agency actions are reviewable by us, we must determine from the record if substantial competent evidence existed to support the findings of the trial court. See K.S.A. 77-621(c)(7). We may not reweigh the facts or substitute our own judgment for that of the trial court or the administrative body. Kaufman v. Kansas Dept. of SRS, 248 Kan. 951, 961-62, 811 P.2d 876 (1991). The party challenging the validity of the agency action bears the burden of proof. 248 Kan. 951, Syl. ¶ 4.
Wesley argues the evidence was insufficient to support the decision of both the hearing officer and the trial court concerning services it provided without a KPCN referral. During the audit, EDS employees reviewed the medical charts of each claim in the sample to check for the type of service and for the proper KPCN *144referral if required. As stated earlier, the KPCN operates under a waiver from the federal agency. This waiver allows Kansas to restrict Medicaid patients to a specific provider; however, this restriction is inapplicable in emergency care situations.
During the period audited, the KPCN required that patients be treated either by their KPCN doctor or by another doctor with a referral from their KPCN doctor. The KPCN required that the referral be in writing. With the exception of emergency cases, failure to have the required referral was grounds for SRS to refuse the claim or to recoup monies previously paid on the claim.
CLASSIFICATION OF CLAIMS AS EMERGENCY OR NONEMERGENCY
Wesley argues it should have been allowed to show that bills submitted as nonemergency services were actually for emergency services. Because the services were emergency, no referral was required, and SRS should not have been allowed to recoup the monies paid for these services. Wesley argues that SRS’s determination of what constitutes emergency services is arbitrary because of its shifting definition of the term “emergency services.” Finally, Wesley claims the testimony in the record shows that most of the claims denied by SRS were for emergency services and, therefore, SRS could not deny payment.
The challenged claims can be broken down into three different categories. The first category consisted of claims originally billed by Wesley as nonemergency claims. These claims were paid by EDS as nonemergency services. Because they were claims for non-emergency services, SRS required a written referral from the KPCN doctor to the emergency room doctor before SRS would pay for the services.
The second category consisted of claims originally billed by Wesley as emergency claims for which, at the time of billing, SRS had denied the emergency status and paid as nonemergency services. Because SRS determined the services were nonemergency, a KPCN referral was required.
The third category consisted of claims originally billed and paid as emergency claims but later, during the post-audit review, de*145termined to be for nonemergency services. SRS sought to recoup overpayment for these claims.
Wesley first argues that it was denied due process because it was prevented from proving that certain claims originally billed as non-emergency were actually emergency services and, therefore, did not need a KPCN referral. Wesley also claims it should be allowed to challenge claims originally billed as emergency but denied by SRS and paid as nonemergency.
The hearing officer ruled that according to K.A.R. 30-7-30 (revoked July 1, 1989) and K.A.R. 30-7-68, Wesley could have requested a fair hearing challenging SRS’s reclassification of Wesley’s claims within 30 days of receiving payment for the claims. Because it failed to challenge either its own billing classification or SRS’s reclassification, Wesley was precluded from trying to challenge SRS’s reclassification of those claims at the administrative proceeding.
In discussing the basic elements of procedural due process, our Supreme Court stated: “The essence of due process is protection against arbitrary government action. The basic elements of procedural due process of law are notice and an opportunity to be heard at a meaningful time in a meaningful manner." Joe Self Chevrolet, Inc. v. Board of Sedgwick County Comm’rs, 247 Kan. 625, 630, 802 P.2d 1231 (1990).
K.S.A. 75-3306 furnishes a fair hearing for any interested party to appeal a final action of SRS. Under K.A.R. 30-7-30, a request for a fair hearing will be dismissed if filed more than 30 days from the date of the agency action. K.A.R. 30-7-30 was replaced by K.A.R. 30-7-68, which states a fair hearing must be requested in writing within 30 days from the date of the order or notice of action.
K.A.R. 30-5-59(a)(10) requires that, as a condition of participating in the Medicaid program, a provider must accept as full payment, subject to audit, the amount paid by the agency for covered services. K.A.R. 30-5-60(a)(6) states that SRS may terminate the provider’s participation in the program for a pattern of submitting inaccurate billings. Therefore, Wesley had a duty to submit bills that accurately represented the services it furnished to its patients.
*146The previously cited regulations furnished ample opportunity for Wesley to challenge an earlier SRS determination. If Wesley submitted an erroneous bill and failed to correct its own error, it should not now, as a result of a post-audit review, be allowed to recharacterize the type of services furnished to match the documentation contained in the file. Similarly, if Wesley disagreed with SRS’s reclassification of a claim, Wesley should have sought administrative relief at that time. Accordingly, the hearing officer was reasonable in concluding that Wesley should be precluded from challenging its own billing classification of the services it provided.
Wesley also claims SRS’s definition of the term “emergency services” fluctuated over time and SRS’s denial of claims based upon this definition was arbitrary. Wesley argues that only the treating physician is qualified to determine if the service was an emergency service. Therefore, Wesley contends it should be allowed to challenge SRS’s earlier reclassification of claims from emergency to nonemergency.
Although the record shows the definition of emergency services did fluctuate over time, Wesley had an opportunity to challenge SRS’s denial of claims based upon those definitions. Because an agency’s interpretation of its own rules and regulations is given great deference, we conclude that Wesley failed to show the decisions of the hearing officer and the trial court were arbitrary or capricious.
TELEPHONE REFERRALS
Wesley next argues that telephone KPCN referrals should have been adequate because SRS had determined that they were acceptable before and after the post-audit period. Before the audit period, SRS had allowed providers to document verbal referrals from KPCN doctors with a notation in the medical file. SRS, however, soon discovered the practice was unreliable because it could not document the fact the referrals were actually made, and SRS changed the policy to require written referrals.
In November 1985, SRS changed the policy effective January 1986 and notified the providers by Bulletin 85-6. The bulletin reads in relevant part: “The recipients are restricted to services provided *147by the physician case manager and may not, except in emergencies, receive any medical services from another provider without a written referral from the physician case manager.”
In a supplement to the Medicaid/Medikan Provider Manual, dated May 1987, SRS gave a more specific direction concerning written referrals. The revision reads in relevant part: “Non-case manager providers . . . must receive a written referral from the patient’s case manager before any service provided to the patient is considered for reimbursement. Failure to receive a referral results in the denial of payment for the service(s) rendered.”
The hearing officer affirmed SRS’s determination that a written referral was necessary during the audited period and that a notation of a telephone referral was not sufficient. The trial court determined that Bulletin 85-6 was insufficient notice that SRS had implemented a major policy change and that the notice was not effective until the 1987 bulletin was issued. The trial court also determined that the denial of pre-1987 claims, based on the files containing only documentation of verbal referral, was arbitrary and capricious.
Because SRS did not have a valid reason for changing the policy and then changing it back, Wesley contends it should not be bound by the requirement. But Wesley fads to cite any regulation or case law which supports its argument. SRS’s stated reason for requiring a written referral was to insure the referrals were actually being made. Because Wesley has failed to show SRS’s motivation was contrary to its stated purpose of insuring the referrals were actually made, we agree with the trial court and conclude that Wesley should be bound by the written referral requirement.
TREATMENT BY NON-KPCN RESIDENTS WITHOUT WRITTEN REFERRALS
Wesley maintains an accredited graduate medical education program where medical residents are assigned to Wesley as part of their residency program. These residents work under the supervision of physicians at the hospital.
Several of the claims denied for lack of a KPCN referral involved treatment by residents assigned to the Wesley clinic. Three KPCN *148doctors were supervisors of residents at the clinic during the audit period. SRS denied the claims because there were no referrals from the KPCN doctors to the residents. Wesley argues these residents were acting on behalf of their supervising doctors and, therefore, a referral was not necessary. Wesley contends such a practice is reasonable because the SRS policies do not expressly forbid it.
The KPCN agreement, which each KPCN doctor signs, provides for a list of doctors who are designated to cover for the KPCN doctor. No KPCN referral is needed for the designated doctor, and the designated doctor can refer to other doctors. Wesley s KPCN doctors, however, failed to list the medical residents as covering physicians for them.
The hearing officer affirmed SRS’s denial of the claims because the purpose of the KPCN is to make sure the KPCN doctor is monitoring the delivery of services. The hearing officer determined the purpose would not be achieved if a resident could render medical services without ever consulting with the KPCN doctor. The trial court agreed, concluding the referral system is necessaiy and consistent with the intent of the program. We agree. Finally, SRS’s policy appears to allow the KPCN doctor to name an ample number of covering physicians. Therefore, Wesley has failed to show the policy requirement is arbitrary or capricious.
SIGNATURE STAMPS
Wesley next argues the trial court erred when it determined SRS could require doctors to sign orders rather than use a signature stamp. SRS sought to recoup monies on several claims because the doctors’ orders were stamped with a signature stamp rather than signed. If a hospital is accredited by the Joint Commission on Accreditation of Health Care Organizations (JCAHCO), the hospital is deemed to meet all of the Medicare regulations. 42 C.F.R. § 488.5 (1988). Because Wesley’s policy of allowing doctors to use signature stamps is in accordance with JCAHCO regulations, Wesley argues SRS’s signed order requirement is arbitrary and capricious.
The testimony of SRS’s witnesses reflect that under many circumstances signature stamps or computerized signatures are con*149sidered acceptable signatures under the Medicaid program. Apparently, SRS is taking the position signature stamps are unacceptable in the context of physician’s orders. Consequently, the agency and the trial court relied upon Kansas Medicaid/Medikan Bulletin 84-1 dated March 1984, which stated, “The Division of Medical Programs will accept as documentation only orders which are signed by a physician.”
Although the rule requires signed orders from an attending physician, it does not define what is an acceptable signature. The rule says nothing about the use of signature stamps. Nothing in the regulation would put a Medicaid provider on notice that the use of a signature stamp is unacceptable with respect to physician’s orders. As our Supreme Court concluded in Riverside Hospital, Inc. v. State Department of SRS, 248 Kan. 609, 612-13, 808 P.2d 1348 (1991), providers are entitled to notice before Medicaid policy can be implemented. Therefore, we conclude that the aforementioned bulletin was inadequate to give notice that Wesley’s controlled use of stamped signatures, which had been patterned after JACHCO’s policy, was unacceptable. Furthermore, we conclude that SRS’s action for recoupment, based on Wesley’s controlled use of stamped signatures, is arbitrary and capricious.
IS THE KPCN CONTRARY TO FEDERAL LAW?
Wesley argues SRS’s implementation of the KPCN conflicts with the requirements imposed on hospitals by the Medicaid program. To support its claim, Wesley refers to 42 U.S.C. § 1395dd (1988). This statute, which is referred to as the anti-dumping provision of Medicare law, requires a program hospital to furnish an appropriate medical screening exam to determine if an emergency condition exists. The statute also imposes sanctions upon hospitals that fail to treat individuals with emergency medical conditions.
Wesley claims implementation of the KPCN conflicts with the federal mandate in two ways. First, SRS defines the term emergency more narrowly than does the federal government. Second, SRS’s implementation of its KPCN denies payment for the screening required by the federal statute.
*150Both the hearing officer and the trial court rejected Wesley’s claims. The hearing officer ruled that “anti-dumping legislation” imposed no requirement on SRS to define “emergency” in the same manner for purposes of reimbursement. The trial court noted HCFA’s response to a letter HCFA received from Wesley requesting a clarification on this issue. HCFA stated Wesley was required to do a screening to determine if an emergency existed. If no emergency existed, die emergency room was under no obligation to provide service, and it should refer die patient to the KPCN doctor.
William Headen, director of Health Financing Resources for Wesley, testified he believed the State definition of emergency was superseded by federal law. He admitted, however, the definitions in die State regulation and the definition contained in the federal statute were very similar. He testified that once the screening was completed, the medical care_ remaining to be provided was generally limited.
Dr. Caliendo, director of the emergency department at Wesley, testified Wesley did not do screening exams, but rather treated anyone who came to the emergency room.
The first problem with Wesley’s claim is that it fails to cite any federal regulation which requires Medicaid to pay for these screening exams.
Second, die testimony of Wesley’s own employees concedes the definitions of emergency services in the state and federal regulations are substantially similar. Even if the definitions are different, Wesley appears to be disregarding the definitions by adopting a policy that everyone coming to the emergency room will be treated whether or not an emergency actually exists. Wesley is free to implement this policy, but there is nothing in the state or federal laws that requires SRS to reimburse Wesley for the services unless it meets the statutory guidelines. Therefore, the decision of the trial court is supported by substantial competent evidence.
Wesley next argues the waiver the KPCN operates under is no longer valid because SRS had changed certain KPCN policies without the approval of HCFA. The trial court rejected this claim, stating that if Wesley believed a violation of the federal waiver occurred, it should have addressed those concerns to the appropriate *151federal agency. We agree. The waiver is valid as long as Kansas does not restrict eligible recipients access to emergency services or access to medically necessary primary care services, and HCFA has never found the KPCN to be out of compliance with the waiver.
SAMPLING
Wesley next argues the hearing officer and the trial court both erred in relation to the overpayment determination. Wesley s argument on this issue is divided into several subissues. First, Wesley challenges SRS’s authority to use sampling as a method to determine the amount of overpayment for a large number of claims. Wesley argues SRS cannot use this method because no statute or regulation exists authorizing the use of sampling methodology in such cases and because the rulemaking procedure prescribed in K.S.A. 77-415 et seq. was not followed. Finally, Wesley argues the methodology used by SRS is unreliable. SRS counters by claiming the use of sampling is authorized by federal regulation and has been found to be valid by case law.
ARE SAMPLE RESULTS SUFFICIENTLY RELIABLE TO PROVIDE A REBUTTABLE PRESUMPTION OF OVERPAYMENT?
The federal government subsidizes Medicaid through the state, paying for medical services to individuals that meet certain eligibility requirements. In order to obtain federal financing, a state must comply with federal statutes and regulations. 42 U.S.C. § 1396a (1988). Under K.S.A. 39-701 et seq. and K.S.A. 39-708c, SRS is responsible for the administration of the public welfare program in Kansas. As the state Medicaid agency, SRS is required by federal law to have procedures for auditing the Medicaid program. Riverside, 248 Kan. at 614.
Federal law requires that the Medicaid agency implement a surveillance and utilization control program to safeguard against unnecessary or inappropriate use of Medicaid services and against excess payments. 42 C.F.R. § 456.3 (1993). Federal law also requires that, as part of the utilization and control process, the agency have procedures which, on a sample basis, evaluate the need, qual*152ity, and timeliness of the Medicaid services. 42 C.F.R. § 456.22 (1993). Finally, federal law directs that the agency have a post-payment review process. 42 C.F.R. § 456.23 (1992). Kansas law authorizes SRS to withhold payments, which are otherwise proper, when it has determined that the provider to whom payments are to be made has been overpaid. K.A.R. 30-5-61a(l).
In Riverside, 248 Kan. at 615, the Kansas Supreme Court recognized that other jurisdictions had upheld the validity of sampling and extrapolation techniques for determining overpayment amounts. The court, however, declined to rule on the propriety of using such methods in Kansas because the parties neglected to raise the issue at the administrative level. Therefore, the Riverside court determined the trial court was precluded from considering whether the use of such a method was proper. 248 Kan. at 617-18.
Even though the Riverside court declined to specifically rule on the issue, it did favorably cite Illinois Physicians Union v. Miller, 675 F.2d 151 (7th Cir. 1982). The Miller court concluded the use of statistical sampling and extrapolation provided a valid basis for making findings of fact in the context of Medicaid reimbursement. The court further stated the use of statistical samples to audit claims was reasonable because the results merely created a rebuttable presumption. The court determined this process to be reasonable when a large number of claims rendered a claim-by-claim review a practical impossibility. The court also concluded it was reasonable to place the burden on the challenging party to present evidence to rebut the results of the statistical sample. 675 F.2d at 155.
In large part, the Miller court relied upon State of Ga., Dept. of Human Resources v. Califano, 446 F. Supp. 404 (N.D. Ga. 1977). The Califano court stated:
“Projection of the nature of a large population through review of a relatively small number of its components has been recognized as a valid audit technique and approved by federal courts in cases arising under Title IV of the Social Security Act. . . . Moreover, mathematical and statistical methods are well recognized as reliable and acceptable evidence in determining adjudicative facts. . . . However, to find that statistics may be admitted as evidence of a proposition *153is not to say that the statistical model will always be conclusive. The weight which must be given to such statistical evidence is necessarily one which must be considered by the fact finder in light of the practical difficulties in obtaining a claim-by-claim review.” 446 F. Supp. at 409-10.
Based on the language in Riverside, it is clear the Kansas Supreme Court would approve of the use of sampling and extrapolation to establish the amount of overpayment in these types of cases. See 248 Kan. at 615-16. Therefore, it was proper for SRS to calculate the amount of overpayments made to Wesley by applying the error rate of the sample to the total amount of claims for the time period audited.
DID THE AGENCY’S POLICY ESTABLISHING THE USE OF SAMPLE RESULTS VIOLATE KANSAS STATUTES RELATING TO RULEMAKING POLICY?
Wesley next argues that the agency’s sampling policy and procedures were adopted without complying with the rulemaking procedures of K.S.A. 77-415 et seq. K.S.A. 1992 Supp. 77-415(4) states:
“ ‘Rule and regulation,’ ‘rule,’ ‘regulation’ and words of like effect mean a standard, statement of policy or general order, including amendments or revocations thereof, of general application and having the effect of law, issued or adopted by a state agency to implement or interpret legislation enforced or administered by such state agency or to govern the organization or procedure of such state agency. Every rule and regulation adopted by a state agency to govern its enforcement or administration of legislation shall be adopted by the state agency and filed as a rule and regulation as provided in this act.”
In Union of American Physicians & Dentists v. Kizer, 223 Cal. App. 3d 490, 272 Cal. Rptr. 886 (1990), the California court concluded that the California Department of Health Service’s policy of using sampling and extrapolation methods for determining over-payments under the Medi-Cal Act were invalid. The court determined the policies were illegal underground regulations because the agency had failed to follow rulemaking procedures established by the California Administrative Procedure Act. The court also specifically held that 42 C.F.R. § 456.22 did not mandate the use of sampling for overpayment determinations. 223 Cal. App. 3d at 499.
*154Nevertheless, SRS’s policy setting out its sampling and methodology procedure appears to be a statement of general policy and application, implemented to enforce the federal and state legislation dealing with post-audit reviews.
Moreover, the Oregon Court of Appeals has determined that a state agency does not need to follow rulemaking procedure to use a sampling method because the methodology is merely one means of quantifying data for evidentiary purposes. See OMAC v. Mental Health Div., 96 Or. App. 528, 774 P.2d 1113, rev. denied 308 Or. 405 (1989). The OMAC court held the hearing officer was free to hear rebuttal evidence and could accept the evidence found to be most persuasive.
The Oregon position appears to be the more logical. Because SRS’s methodology merely creates a rebuttable presumption of overpayment, it does not appear to meet the statutory definition of a rule or regulation. Therefore, SRS’s methodology is simply a way of calculating the overpayment, and the provider is free to challenge the amount or methodology used in determining the amount.
WERE THE SAMPLE RESULTS USED TO DETERMINE OVERPAYMENT SUPPORTED BY SUBSTANTIAL COMPETENT EVIDENCE?
Wesley argues the sampling methodology used was flawed to the extent that the results were unreliable. Wesley divides this issue into several subissues. First, the sample was not drawn from claims selected at random. Second, a stratified sample, rather than a simple random sample, should have been used. Third, the sample size was too small to provide a reliable indication of the error rate of the entire population of claims.
The record shows that EDS followed the sampling methodology set out in “Procedure II,” which was contained in a SRS policy document dated January 1,1987. Based on the original number of unit recipients of 4,703, EDS referred to a table in Arkin’s Handbook of Sampling for Auditing and Accounting to determine the sample size should be 207. According to SRS policy, this sample *155size was to produce a result that was accurate to within plus or minus 4% at a 95% confidence level.
Next, EDS had its computer sort the population of claims by recipient identification number. The computer was then programmed to start selecting records beginning at record 3 and then to select every 22nd record. The selected records were then placed in alphabetical order, and a report was generated. EDS then audited the medical files of the recipients listed in the report.
Wesley argues that placing the population in identification order does not comply with Procedure II, which calls for placing the population in random order before selecting the sample. A SRS representative, however, testified that each recipient number is unique to the individual and does not necessarily represent any chronological order.
The hearing officer found that Wesley had failed to meet its burden to show that the sample was not a random sample. We agree. Absent a showing of an arbitrary disregard of undisputed evidence or an indication of bias, passion, or prejudice, this negative finding cannot be disturbed. See Mohr v. State Bank of Stanley, 244 Kan. 555, 567-68, 770 P.2d 466 (1989).
Wesley next argues that EDS failed to follow SRS guidelines, which called for use of a stratified sampling method when heterogeneous populations are involved. The record shows that SRS did have a policy, titled “Procedure III,” to utilize a stratified sampling procedure when the sample population was heterogeneous. According to EDS, however, Procedure III had never been used, and EDS did not have the ability to implement the testing procedure. The procedure was allegedly put in place for future use.
Michael Intriligator, Ph.D., testified on behalf of Wesley that, because of the heterogeneous nature of the population being sampled, a stratified sampling method should have been used. He testified that Procedure III should have been used because of the difference in types of medical procedures, the types of providers of service, and the differing places of treatment. He also stated that there were significant differences in the average number of disallowed claims between emergency room and nonemergency room claims. It was his opinion that these distinctions were not reflected *156in the simple random sampling technique used by EDS. He implied the EDS results were flawed because EDS did not use a large enough sample given the heterogeneous nature of the population sampled and because it used Procedure II instead of Procedure III.
Dr. Intriligator further testified the sample results were unreliable because EDS had failed to reach its policy goal of an accuracy rate of plus or minus 4% with a confidence level of 95%. In other words, based on his analysis, SRS was not 95% sure that it had determined the true amount of overpayment within plus or minus 4%.
Dr. Intriligator testified he ran a computer simulation to test EDS’s sampling techniques. He ran the simulation 100 times. The simulation showed an accuracy rate of plus or minus 5% and a confidence value of 76%. He then ran the simulation 900 additional times to confirm the findings of the first 100 simulations. It was his opinion that SRS had failed to meet the standards set out in its own policies.
SRS’s expert, Frank Webb, agreed that the stratified sampling technique should be used for heterogeneous populations; however, he was unsure whether the population was heterogenous. He testified that although a population appears heterogeneous and that experts could disagree as to whether a stratified sample should have been used, the use of a stratified sample is not automatically required. He failed to state if he made any tests to determine if the population was homogenous or heterogenous.
Webb also admitted that EDS’s sampling method did not meet the stated accuracy goals; however, he felt the sampling was close enough. Nevertheless, he made no attempt to determine the actual confidence level or error rate in the EDS sample.
. The hearing officer ruled Wesley had failed to sustain its burden to prove EDS’s calculations were unreliable. The hearing officer conceded SRS failed to comply with its own policy concerning accuracy and confidence level. But because Wesley failed to present evidence on 100% of the claims at issue which showed SRS’s calculations were wrong, and because it failed to do an audit using a *157stratified sample, the hearing officer ruled Wesley had failed to show the method used by SRS was unreliable.
The trial court determined that although Wesley offered a more accurate sampling methodology, the procedures used by EDS, with SRS’s approval, were not arbitrary, capricious, or otherwise invalid.
The resolution of this issue is troubling. First, SRS had an established policy designed to provide substantial competent evidence of the amount of overpayment in these types of situations. SRS, however, failed to comply with its own procedures.
Depending on the type of population being audited, the policy sets out three different procedures for calculating the amount of overpayment. Procedure II is used for homogeneous populations, and Procedure III is used for heterogeneous populations. According to the policy, differing procedure codes, places of services, and performing providers are characteristics indicating heterogeneity. But neither SRS nor EDS developed any method for using Procedure III. Inasmuch as SRS failed to develop a method for testing a heterogeneous population, one could safely say SRS would never determine a population to be heterogeneous.
Second, Procedure II, as set out in SRS policy, calls for the statistical sample to yield a confidence rate of 95% with an accuracy rate of plus or minus 4%. Wesley, however, presented evidence that SRS’s sampling method failed to achieve a precision rate of plus or minus 4% at a confidence level of 95%. In addition, SRS’s own expert agreed that SRS had failed to meet the policy standards.
Moreover, K.S.A. 77-621(c)(5) allows a trial court to grant relief from an agency’s action if the agency has engaged in an unlawful procedure or has failed to follow prescribed procedure. In addition, K.S.A. 77-621(c)(8) allows a trial court to grant relief from an agency’s action if the agency action is otherwise unreasonable, arbitrary, or capricious.
SRS’s sampling methodology was clearly inadequate to comply with its own policy. Because the sampling methodology applied by SRS was inconsistent with its own policy, both with respect to the procedure used and the accuracy level achieved, we reverse the decision of the trial court. Also, SRS’s failure to determine if the population was homogeneous or heterogeneous before deciding *158which sampling procedure to use was unreasonable, arbitrary, and capricious.
CROSS APPEAL
On its cross-appeal, SRS argues the trial court erred in determining the effective date of its policy change requiring written referrals. The trial court determined the effective date of the policy change was governed by Riverside Hospital, Inc. v. State Department of SRS, 248 Kan. 609, 808 P.2d 1348 (1991). In Riverside, the court determined the bulletin issued in December of 1985 was insufficient to put the hospital on notice of the policy change of requiring written referrals from KPCN doctors and refusing verbal referrals in the future. 248 Kan. at 612-13. The bulletin at issue in Riverside is the same bulletin sent to Wesley.
SRS argues the court in Riverside only had the first page of the bulletin and did not know the bulletin was actually three pages long and was attached to a revised Medicaid/Medikan manual. SRS further argues it furnished Wesley with the entire notice and this notice adequately informed Wesley of the policy change.
A review of the Riverside decision does not support SRS’s assertion that the court was only presented with a portion of the bulletin. The Riverside court reviewed Bulletin 85-6 and affirmed the trial court’s decision that the language in the notice was insufficient to put the hospital on notice of the policy change. The decision fails to indicate how many pages were contained in the bulletin or if the bulletin was or was not attached to a revised manual.
Therefore, the record contains ample evidence to support the trial court’s ruling. Accordingly, the decision of the trial court concerning the effective date of the policy is affirmed.
The judgment is affirmed in part, reversed in part, and remanded to the trial court to determine the proper amount of recoupment.