OPINION OF THE COURT
Chief Judge Wachtler.Petitioner, a hospital authorized to provide Medicaid services, challenges the method by which respondent, the agency charged with administering the Medicaid program in New York State, has determined that overpayments in the amount of $113,771.24 were made to the hospital for Medicaid services provided during a two-year period. At issue is whether respondent was authorized to calculate the amount of overpayment by means of statistical sampling rather than by an individual review of all the cases in the audit period, without a showing that petitioner’s records were inadequate to conduct a case-by-case review. We conclude that it was neither arbitrary and capricious nor in excess of respondent’s author*200ity to employ a statistical sampling method in these circumstances.
I.
Medicaid is a joint Federal-State medical assistance program which reimburses medical providers for services provided to eligible individuals (see, Social Security Act tit XIX, 42 USC § 1396 et seq.; Social Services Law § 363 et seq.). Federal funds are available to pay a percentage of the total amounts spent by states for medical assistance (see, 42 USC § 1396b). To participate in the program and receive Federal funds, states are required to submit for Federal approval a medical assistance plan which conforms to standards set forth in Federal statutes and regulations (42 USC § 1396a). States are required, among other things, to adopt procedures to prevent fraud, abuse, unnecessary or inappropriate use of Medicaid services and excess payments (42 CFR 447.45, 455.13, 456.3, 456.23).
New York State opted to participate in the Medicaid program and the Legislature delegated to respondent, the State Department of Social Services (DSS), the task of developing a plan to submit for Federal approval. To that end, Social Services Law § 363-a (1) provides that DSS "shall submit a plan for medical assistance, as required by title XIX of the federal social security act, to the federal department of health, education and welfare.1 for approval * * * and shall act as the single state agency to supervise the administration of the plan in this state. [DSS] shall act for the state in any negotiations relative to the submission and approval of such plan and it may make such arrangements, not inconsistent with law, as may be required by or pursuant to federal law to obtain and retain such approval and to secure for the state the benefits of the provisions of such law.” In addition, section 363-a (2) provides that DSS "shall make such regulations, not inconsistent with law, as may be necessary to implement this title.”
Pursuant to this grant of authority from the State Legislature, and in accordance with Federal regulations requiring such procedures as part of the State plan, DSS administers a plan which includes mechanisms for detecting and recouping payments to providers for unnecessary or inappropriate services or in excess of the authorized amount (see generally, 18 *201NYCRR parts 517 [audit procedures], 518 [recoupment procedures], 519 [hearing procedures]). Regulations promulgated by DSS require providers to prepare and maintain detailed records reflecting the nature and extent of services furnished and to make them available to DSS for audit (18 NYCRR 504.3 [a], [g]). Audits are conducted on notice to the provider, who may discuss and lodge objections to the audit procedures, audit findings and any proposed action prior to the issuance of a final audit report (18 NYCRR 517.3, 517.5, 517.6). Upon the issuance of a final audit report, the provider is entitled to a hearing to challenge any determination that overpayments have been made (18 NYCRR 517.6, 519.4). The regulation governing hearing procedure provides, in part, that "[a]n extrapolation based upon an audit utilizing a statistical sampling method certified as valid will be presumed, in the absence of expert testimony and evidence to the contrary, to be an accurate determination of the total overpayments made or penalty to be imposed. The [provider] may submit expert testimony challenging the extrapolation by the department or an actual accounting of all claims paid in rebuttal to the department’s proof’ (18 NYCRR 519.18 [g]).2
II.
Respondent’s audit of petitioner’s Medicaid billings covered the period from December 1, 1982 through November 30, 1984, and concerned billings for outpatient services only. These services fell into three categories — emergency room services, ordered ambulatory services (tests and procedures ordered by an outside physician), and laboratory services. The hospital’s Medicaid billings for the audit period consisted of 5,047 emergency room cases, 1,860 ordered ambulatory cases and 2,979 laboratory cases, for a total of 9,886 cases.
Although it is not disputed that the hospital had adequate records for all of the cases subject to the audit, the auditors reviewed only 400 randomly selected cases: 200 emergency room cases, 100 ordered ambulatory cases, and 100 laboratory cases. Based upon a review of the hospital’s medical records for the sampled cases, the auditors determined that the hospital had been overpaid a total of $4,256.76 for the 200 emergency room cases sampled (an average of $21.2838 per case) *202and a total of $341.50 for the 100 ordered ambulatory cases sampled (an average of $3,415 per case). No significant over-payments were found in the laboratory cases. By multiplying the average overpayment by the total number of cases in each category, DSS projected overpayments of $107,419.34 for the 5,047 emergency room cases and $6,351.90 for the 1,860 ordered ambulatory cases — a total of $113,771.24.
After being notified of DSS’s intention to recoup the alleged overpayments, the hospital requested a hearing, asserting, among other things, that the use of statistical sampling was improper. At the hearing, DSS presented expert testimony from a statistician that the random sample reviewed was representative of the universe of cases subject to the audit. He explained the method by which the results of the sample were extrapolated to determine the total amount of overpayments and testified that, in his opinion, the result achieved by this method was the amount that the auditors most likely would have found if they had reviewed the entire universe of cases.
Other witnesses for DSS explained the reasons that claims were disallowed. In the emergency room cases, for example, claims for services rendered during prescheduled visits were disallowed pursuant to regulations defining emergency services as those provided "to the ill and injured who require immediate medical or surgical care on an unscheduled basis” (10 NYCRR 444.19 [a] [1]).
The hospital’s expert did not challenge the validity of the sample used in this case, or the efficacy of sampling techniques in general. He asserted, however, that the audit in this case was flawed because the criteria applied by the auditors were not sufficiently specific and because the auditors’ lack of a medical background did not qualify them to determine whether a given case involved a medical emergency as defined in the regulations.
Following the hearing, the Administrative Law Judge upheld the determination that the hospital had received over-payments, holding that the use of statistical sampling techniques was authorized and that the audit in this case was properly conducted. The ALJ initially reduced the amount of the overpayment by allowing the hospital to offset each disallowance for prescheduled emergency room cases by the amount billable for a routine physician’s office visit. Following a request from DSS’s hearing counsel for clarification of the basis for the offset, the ALJ issued a corrected decision elimi*203nating the offset, which he stated was the product of a proofreading mistake in the original decision.
This CPLR article 78 proceeding ensued. Upon transfer from Supreme Court (see, CPLR 7804 [g]), the Appellate Division, Third Department, annulled the determination and remitted the matter to DSS for further proceedings. The Appellate Division held that DSS’s use of the random sample audit was arbitrary and capricious because adequate records were available to conduct a case-by-case review (169 AD2d 1009). In light of this holding, the Appellate Division did not address the other issues raised by the hospital, including its arguments with respect to the issuance of the corrected decision, and its objections to the criteria used and the qualifications of the auditors. We granted the motion of DSS for leave to appeal3 and now reverse.
III.
Petitioner contends first that DSS exceeded its authority in conducting the random sample audit, notwithstanding the existence of regulations purporting to authorize such methods, because the legislation delegating to DSS the authority to administer the Medicaid program does not expressly or impliedly authorize the use of random sample audits.
It is true, of course, that the Social Services Law does not expressly refer to such audits, but that observation does not materially advance the analysis. The legislation makes no express reference to audits of any kind, but there can be no doubt that, to administer a Medicaid program in conformance with Federal requirements regarding the prevention of fraud and abuse, DSS must have the authority to audit the medical records maintained by providers to determine the legitimacy of claims made by them (see, Matter of Camperlengo v Blum, 56 NY2d 251, 255-256). Such authority is implicit in the designation of DSS as the agency "to supervise the administration of the [Medicaid] plan in this state” (Social Services Law § 363-a [1]). It is well settled that an agency’s powers include not only those expressly conferred, but also those "required by necessary implication” (Matter of City of New York v State of New York Commn. on Cable Tel., 47 NY2d 89, 92). This is especially true where, as here, the Legislature has *204delegated administrative duties in broad terms, leaving the agency to determine what specific standards and procedures are most suitable to accomplish the legislative goals (see, id.; Matter of Levine v Whalen, 39 NY2d 510, 515-516). DSS’s decision to employ statistical sampling in conducting audits falls well within that broad authority. If it is reasonably designed to further the regulatory scheme, it cannot be disturbed by the courts unless it is arbitrary, illegal or runs afoul of the enabling legislation or constitutional limits (see, Matter of Barie v Lavine, 40 NY2d 565, 568-569).
Petitioner has made no effort to demonstrate that statistical sampling methods generally are arbitrary, unfair or unreliable.4 The use of such methods has been upheld by this Court in other contexts (see, Matter of Grant Co. v Joseph, 2 NY2d 196, cert denied 355 US 869; Matter of Inwood Post No. 581, Am. Legion v State Bingo Control Commn., 22 AD2d 884, affd no opn 17 NY2d 699) and courts in other jurisdictions have approved their use in this or closely analogous contexts (see, Chaves County Home Health Serv. v Sullivan, 931 F2d 914 [DC Cir] [Medicare]; Illinois Physicians Union v Miller, 675 F2d 151 [7th Cir] [Medicaid]; State of Georgia, Dept. of Human Resources v Califano, 446 F Supp 404 [ND Ga] [Medicaid]; see also, Michigan Dept. of Educ. v United States Dept. of Educ., 875 F2d 1196 [6th Cir] [audit of State’s vocational rehabilitation expenditures]).
We agree with those courts that, in view of the vast number of claims subject to audit, it is not unreasonable for a supervising agency, such as DSS, to use statistical samples to establish that overpayments have been made and to estimate their total amount. We emphasize that, as the regulation governing the use of statistical sampling dictates, the provider, who at all times bears the burden of proving entitlement to the Medicaid funds, must be given a fair opportunity to challenge the accuracy of the estimate by attacking the reliability of the methods or standards employed, as petitioner did in this case, or by conducting and submitting a complete audit, which petitioner chose not to do (18 NYCRR 519.18 [g]).
Relying on Tax Law § 1138 (a), however, petitioner contends *205and the Appellate Division held that, even if random sample audits are permissible, it was arbitrary and capricious to use that method in these circumstances. Tax Law § 1138 (a) provides that the Tax Commission shall determine the amount of sales tax due "from such information as may be available [and] [i]f necessary, the tax may be estimated on the basis of external indices”. This provision has been construed by the Appellate Division, Third Department, as precluding the use of a test period audit unless the taxpayer’s records are inadequate to permit a complete audit (see, Names in The News v New York State Tax Commn., 75 AD2d 145, 147; Matter of Mohawk Airlines v Tully, 75 AD2d 249, 250-251; Matter of Chartair, Inc. v State Tax Commn., 65 AD2d 44, 46).
By analogy to those sales tax cases, the Third Department has held, as in this case, that it is arbitrary and capricious for DSS to conduct a statistical sampling audit of Medicaid billings if the provider’s records are available for a complete audit (see, Matter of Graziosi v New York State Dept. of Social Servs., 167 AD2d 793, 794-795; Matter of Allen v Commissioner of Social Servs., 116 AD2d 35, 38). The First and Second Departments, on the other hand, have accepted the use of statistical sampling in Medicaid audits, apparently without regard to the adequacy or availability of the provider’s records (see, Matter of Metzies Shoe Brooklyn N. Y. Corp. v New York State Dept. of Social Servs., 151 AD2d 675 [2d Dept]; Woldeyohannes v Webb, 96 AD2d 493 [1st Dept]; Matter of Sunset Taxi Co. v Blum, 73 AD2d 691 [2d Dept]).
This Court has not passed on the validity of the Third Department’s construction of Tax Law § 1138 (a) and we do not do so now.5 Assuming, however, that the Tax Law indeed *206limits the Tax Commission’s use of test period audits to cases in which a complete audit is not possible, we do not believe that a legislative preference for complete audits in the sales tax context establishes a general rule applicable to all administrative agencies with audit responsibilities. If the Legislature wanted to create a limitation of general applicability, it is unlikely that it would do so by expressing it in a specific, narrow context and hoping that it would spread by implication to other contexts. The Legislature does not ordinarily express its will by such circumlocution.
Nor do we accept petitioner’s argument that the reference to test period audits in section 1138 (a) is a grant of authority to conduct such audits that would not otherwise exist and that, since there is no similar reference in the Social Services Law, DSS has not been so empowered by the Legislature. The argument proves too much. If it were true, then DSS would have no authority to employ statistical sampling techniques even when a provider’s records are inadequate to allow a complete audit. Such a rule would frustrate efforts to recoup overpayments and would undermine the agency’s ability to comply with Federal requirements concerning the prevention of fraud and abuse. In our view, to the extent that Tax Law § 1138 authorizes test period audits only when inadequate records exist, it merely expresses a limitation on the agency’s authority to conduct audits. The absence of a similar reference in the Social Services Law establishes, not an absence of authority to conduct statistical sampling audits in any circumstances, but rather an absence of a similar limitation on the DSS’s authority to conduct such audits.
Finally, we reject petitioner’s contention that the determination to use indirect audit methods is a "uniquely legislative function” that cannot be made by an administrative agency without violating constitutional separation of powers principles (Boreali v Axelrod, 71 NY2d 1, 12). The agency’s choice of methods for detecting and valuating overpayments— tasks specifically imposed on the State by Federal regulations and assigned to DSS by the State Legislature — is an instance of an agency merely "fill[ing] in the details of broad legislation describing the over-all policies to be implemented” (id., at 13). In doing so, DSS was implementing legislative policy, not making legislative policy, and was thus acting within its proper sphere (see, Matter of New York State Health Facilities Assn. v Axelrod, 77 NY2d 340, 348).
*207In sum, we conclude that the authority for DSS to conduct Medicaid audits based upon statistical sampling is implicit in the general grant of authority to supervise the administration of the Medicaid program in this State and that such authority is not limited to cases in which the inadequacy of a provider’s records preclude a complete audit.
Accordingly, the order of the Appellate Division should be reversed, with costs, and the matter remitted to the Appellate Division, Third Department, for consideration of issues raised but not passed upon on the appeal to that court.
Judges Simons, Kaye, Alexander, Titone, Hancock, Jr., and Bellacosa concur.
Order reversed, etc.
. The relevant Federal agency is now the Department of Health and Human Services.
. This regulation became effective on June 6, 1988, the date the administrative hearing in this case began. A substantially similar regulation (18 NYCRR former 515.14 [b] [3]) was previously in effect.
. Although the Appellate Division’s order remitting the matter to DSS is nonfinal, it is appealable to this Court pursuant to CPLR 5602 (a) (2) (see, Matter of F. J. Zeronda, Inc. v Town Bd., 37 NY2d 198, 200-201).
. Petitioner does claim, however, as it did during the administrative hearing, that the methods used in this particular audit rendered the results unreliable. That claim, which relates to whether the administrative determination was supported by substantial evidence, must be considered by the Appellate Division on remittal.
. In the only case to reach us arguably presenting the opportunity to consider whether Tax Law § 1138 (a) limits the use of test period audits, the Third Department had ruled that a test period audit was permissible because the taxpayer’s records were inadequate (see, Matter of Markowitz v State Tax Commn., 54 AD2d 1023, affd 44 NY2d 684). That case, therefore, did not require us to determine whether such an audit would be permissible if adequate records existed. Matter of Grant Co. v Joseph (2 NY2d 196 involved a provision of the Administrative Code of the City of New York similar to Tax Law § 1138 (a). In that case, too, the taxpayer’s records were inadequate and we upheld the use of a test period audit for sales tax purposes. Dicta in that case runs counter to the Third Department’s strict approach to the use of test period audits. We stated: "Even when an item-by-item audit is possible, and it was not here, there is no inflexible requirement that the comptroller resort to it. * * * It is sufficient if the method adopted is reasonably calculated to reflect the taxes due.” (2 NY2d, at 206, supra.)