OPINION OF THE COURT
Mikoll, J.Petitioners are associations representing the interests of various not-for-profit and county-owned nursing homes located throughout the State. Respondent Department of Health (hereinafter DOH) promulgated in 1985 (effective January 1, 1986) a new method for calculating nursing home Medicaid reimbursement rates (the Long Term Care Case Mix Reimbursement System). Pursuant to this plan each facility prepares "Patient Review Instruments” (hereinafter PRIs) reflecting in detail each patient’s functional dependencies and care requirements (see, 10 NYCRR 86-2.30). The data is then used to place each patient into one of 16 "Resource Utilization Groups” (hereinafter RUGs), which correspond to the severity of the patient’s illnesses and the level of care required. A weighted total of the facility’s patient distribution in the RUGs determines a facility’s "Case Mix Index” (hereinafter CMI). The CMI is then utilized to calculate the facility’s reimbursement rate.
Initial reimbursement rates were based on a State-wide average CMI from PRIs submitted in 1985. DOH, however, concluded in late 1986 that an increase in facilities’ CMIs from 1985 to 1986 resulted from "paper optimization,” that is, more effective completion of PRIs, rather than any actual changes in the level of care rendered patients (see, New York State Assn. of Counties v Axelrod, 78 NY2d 158, 167). DOH then promulgated a new regulation, 10 NYCRR former 86-2.31, providing for a State-wide recalibration of reimbursement rates through a 3.035% across-the-board reduction to the *165new reimbursement rate method. The Court of Appeals ruled that the new regulation was irrational, concluding, inter alia, that there was no justification for DOH’s finding that the increase in CMI was solely attributable to paper optimization and not to any actual changes in levels of patient care (supra, at 167-168).
DOH, following repeal of the original recalibration regulation, promulgated two new recalibration regulations aimed at correcting the paper optimization. The first (10 NYCRR 86-2.31 [a]) applied to rate years 1989 to 1991 (hereinafter the 1989-1991 recalibration regulation). The second (10 NYCRR 86-2.31 [b]) applied to rate years 1992 and onward (hereinafter the 1992 recalibration regulation). Both regulations employed the same methodology to factor out all causes for CMI increase other than paper optimization. The 1989-1991 recalibration regulation was capped at an across-the-board 3.035% reduction contained in the original regulation (see, 10 NYCRR 86-2.31 [a] [1]) while the 1992 recalibration regulation removed the 3.035% cap for subsequent years (see, 10 NYCRR 86-2.31 M).
In a CPLR article 78 proceeding challenging the 1989-1991 recalibration regulation, Supreme Court held that the regulation was both arbitrary and capricious and in violation of the prohibition against retroactive rate making contained in Public Health Law § 2807 (7) (a). On appeal, this Court and the Court of Appeals, without reaching the substantive issue of the rationality of the new rate adjustment, affirmed, ruling that the 1989-1991 recalibration regulation was retroactive rate making (see, Matter of New York Assn. of Homes & Servs. for Aging v Commissioner of N. Y. State Dept. of Health, 195 AD2d 822, affd 84 NY2d 252).
During the legal proceedings involving the challenge to the 1989-1991 recalibration regulation, petitioners commenced the instant CPLR article 78 proceeding asserting five causes of action claiming, inter alia, that the 1992 recalibration regulation is arbitrary and capricious. Supreme Court ruled that the 1992 recalibration regulation lacked a rational basis, but rejected petitioners’ four other causes of action alleging that: (1) the recalibration reduction of the 1992 rates violated the 60-day prior notice requirements of Public Health Law § 2807 (7) (a), (2) the regulation violated the State Administrative Procedure Act and New York Constitution, (3) the Medicaid reimbursement rates calculated by respondents are inadequate and in violation of the Public Health Law, and (4) *166respondents’ actions deprived petitioners’ members of due process. Supreme Court concluded that the 1992 recalibration regulation was "adopted without any reasonable or measured period of empirical analysis” and remanded the matter to respondents for: "a redetermination of the 1992 and future Medicaid reimbursement rates * * * with * * * any recalibration component [to] be determined according to the guidelines set forth * * * in [New York State] Assn. of Counties v. Axelrod, 78 NY2d 158 (1991) and the ruling in this proceeding.”
On appeal respondents contend that petitioners have not made a compelling showing that the 1992 recalibration methodology is without rationality. Respondents claim that the methodology chosen was the product of careful reasoning and empirical analysis. Further, respondents assert that it is not sufficient to show that the methodology utilized is imperfect or that a better methodology could have been devised. They also argue that other factors that, according to Supreme Court and petitioners, could have brought about increased CMIs are accounted for in the new methodology. We disagree.
In our view, petitioners have met their burden of proving that the 1992 recalibration regulation is without a rational basis and is therefore null and void. The judgment of Supreme Court should be modified, however, by reversing so much thereof as directed respondents, on remand, to consider any recalibration component in recomputing petitioners’ affected rates.
The Court of Appeals in New York State Assn. of Counties v Axelrod (78 NY2d 158, supra) annulled the original recalibration regulation, finding that it was irrational because of (1) the arbitrariness of the percentage figure selected for reimbursement reduction—3.035%—absent any documentation that there was even an average or mean increase of 3.035% in CMI "attributable solely to improved PRI reporting compliance experience” (supra, at 168), and (2) the discriminatory and disparate impact of an across-the-board percentage reduction, even to facilities having no increase in CMIs or increases demonstrably attributable to higher levels of patient care (supra, at 168). Upon review of the record it appears that petitioners have demonstrated that respondents’ 1992 recalibration regulation (10 NYCRR 86-2.31 [b]) does not overcome the deficiencies the Court of Appeals noted in regard to the original recalibration regulation.
*167According to respondents the increase in CMIs from one assessment period to the next results from four factors: (1) changes in the composition of patients due to new admissions, (2) changes in the composition of patients due to death or discharge, (3) changes in the conditions or care needs of patients, and (4) paper optimization.
Respondents arrived at the 1992 recalibration regulation by comparing 1985 patient data with 1988 patient data and found that the greatest increase in CMI was in the first year of implementation of the system. Respondents consequently sought to employ an expanded data set to assess the change in CMI due to paper optimization. Respondents claim that this comparison was appropriate because of the stabilization of the increase in CMI during that time. The methodology used called for the utilization of "closed matched sets” of patients, i.e., the data comparison used data only for patients who were in the same facility in both the 1985 and 1988 patient assessments. These matched sets were used so as to exclude changes in patient mix due to new admissions and those due to death and discharges (the first two factors for increased CMI).
To account for increases in CMI between 1985 and 1988 attributable to increased patient care due to aging and deteriorization over time (the third factor for increased CMI), respondents relied on a length of stay (hereinafter LOS) adjustment. The LOS adjustment is determined by ascertaining a facility’s average CMI in 1985 (using the 1985 State-wide average CMI for each LOS group for matched set patients and comparing that to the facility’s average CMI in 1988 (using the 1985 State-wide average CMI for each LOS group) (see, 10 NYCRR 86-2.31 [b]). The final recalibration adjustment (due to paper optimization) is then arrived at by annualizing the difference between the actual percentage increase reported and the percentage increase attributable to LOS.
Supreme Court properly found DOH’s assertion that Statewide LOS statistics could properly predict actual changes in patient condition flawed because of DOH’s failure to perform any empirical study to ascertain what portion, if any, of the 1985 to 1988 CMI increase was due to actual changes in the condition of patients. The actual CMI increase cannot be measured accurately because of the inability to reobserve the actual past condition of a patient and accurately compare it with his or her medical record. Admittedly, the LOS relationship to acuity cannot be statistically validated. Additionally, *168DOH stated in its 1982 Residential Health Care Facility Reimbursement Methodology handbook that average LOS is inappropriate as a predictor of actual patient service needs. Some surrogate measure is required but is not present here.
Moreover, there are fundamental flaws in the methodology, which result in the generation of recalibration percentages that are unrelated to the paper optimization for which they are intended to compensate. In our view, petitioners have met their burden of proving the regulation irrational by demonstrating that the LOS adjustment, which is utilized within the recalibration methodology as a means of approximating the amount of change in the matched set patients’ reported CMI that is attributable to "actual change in patient conditions, care needs and resource utilization” (New York State Assn. of Counties v Axelrod, 78 NY2d 158, 168, supra), does not, in fact, do so.
Because the group of patients from which the LOS adjustment is taken (all patients, in all facilities in the State, in 1985) differs, in significant ways, from the matched sets, the key premise upon which the LOS adjustment calculation is grounded—that the latter group will manifest a similar course of deterioration, or require a similar level of care after the same period of residence, as the former—is manifestly unsound.
Two examples are illustrative. First, the methodology is premised on an assumption that the needs of patients admitted in 1983, 1984 and 1985 will be comparable to the needs of patients admitted in 1980, 1981 and 1982, after the same period of residence. There is no evidence, however, that these two groups of patients had like needs, on average, at admission. In view of the fact that certain measures were adopted in the early 1980’s, including the use of high- and low-intensity classification to calculate reimbursement rates, which may have encouraged the admission of a greater number of sicker patients in the later years, it is folly to compare the level of care required by these two groups. Yet, the recalibration regulation does this by, for example, comparing the needs of all patients admitted in 1984, after one year, with the needs of all patients admitted in 1981, after four years, and using that difference as a measure of the expected increase for the matched set patients over the three-year period from 1985 to 1988.
Another, similar defect arises from the fact that the *169matched set patients are, by definition, those who remained in a facility for at least three years, while the data utilized to determine how much their health was "expected” to decline during those three years is drawn from all patients, including those who left the facility shortly after admission. This is critical because, as the record demonstrates, the rates of death and readmission to a hospital are substantial, particularly in the first three months after admission. Thus, the fact that the collective CMI of patients who have remained in a facility for three to four years is lower than that of newly admitted patients does not indicate—as respondents contend—that individual patients are apt to improve over time; the difference more likely results from the documented fact that many of the sickest patients leave the facility a short time after admission, never influencing the averages of the groups characterized by a longer period of stay. Consequently, to the extent that the methodology predicts that the CMI of some of the matched set patients should decrease over time, reflecting improvement in their collective condition—and assigns any increase at all in CMI to paper optimization—it cannot be said to have a sound basis.
Inasmuch as DOH has once again reduced petitioners’ reimbursement rates by arbitrary (albeit facility specific) percentages, without any justification for concluding that those percentages even roughly approximate the amounts by which the facilities’ CMIs increased due to factors other than actual changes in patient condition, the resulting regulation must be annulled (see, New York State Assn. of Counties v Axelrod, supra, at 167-168).
Additionally, in our view the dissent’s reliance on the reasoning of the dissent in Matter of Jewish Home & Infirmary v Commissioner of N. Y. State Dept. of Health (84 NY2d 252, 275-276 [Levine, J., dissenting]) and the more recent Court of Appeals decision in Matter of Consolation Nursing Home v Commissioner of N. Y. State Dept. of Health (85 NY2d 326) is misplaced.
As previously demonstrated, respondent Commissioner of Health has not advanced any measurable rational basis for issuance of the recalibration regulation in dispute. The dissent’s suggestion that the recalibration regulation represents a rational application of broader judgmental considerations by the Commissioner, based on the expertise and the experience of the agency he heads, fails. Respondents’ failure to substantiate the extent of any measurable paper optimization pro*170vides no basis for reducing the payments due facilities for patient care and is inconsistent with statutory goals. Only unnecessary costs are not reimbursable. Thus, the action of respondents in making the reduction for unproven paper optimization is in conflict with statutory goals, beyond respondents’ authority and an abuse of discretion (see, e.g., Matter of Jewish Mem. Hosp. v Whalen, 47 NY2d 331, 343; see also, Matter of Meadowbrook Nursing Home v Axelrod, 132 AD2d 837, 839).
Petitioners’ argument that Supreme Court erred in remanding the matter for redetermination of its members’ 1992 Medicaid reimbursement rates without directing respondents to recalculate the rates without any recalibration adjustment, in view of Public Health Law § 2807 (7) (a), is meritorious. The notice requirements of Public Health Law § 2807 (7) (a) prohibit retroactive rate making (see, Matter of Jewish Home & Infirmary v Commissioner of N. Y. State Dept. of Health, 84 NY2d 252, 260, supra).
We do not reach petitioners’ other arguments regarding the rationality of the 1992 recalibration regulation.