In its recent notice of proposed rulemaking setting policy for Medicare Advantage (MA) and the Prescription Drug Program (PDP) for calendar year 2020, CMS announced that it would establish extrapolation as a method to be used in risk adjustment validation (RADV) audits, and further, that it would not make any adjustments to account for errors in Medicare fee for service data in determining recovery amounts.

CMS uses a risk adjustment process to modify MA plan payments to better reflect the relative risk of each plan’s enrollees. Payments to each MA plan are adjusted based on risk scores that reflect enrollees’ health status (categorized into Hierarchical Condition Categories (HCCs)) and demographic characteristics derived from member claims data. To counteract incentives that a plan might have to over-report enrollee diagnoses, CMS emphasizes that all diagnoses submitted to enhance risk must be documented in a medical record, and uses RADV audits to ensure that medical record documentation exists, and thus, that payments to MAOs accurately reflect the level of risk assumed.

CMS is using the proposed rule as an opportunity to announce that it “intends to recover overpayments based on extrapolated audit findings through the use of statistically valid random sampling techniques.” That is, CMS will perform RADV audits of a sample of claims, will aggregate discrepancies to determine an overall payment error rate, and will extrapolate the rate to calculate a contract-level payment error estimate. The use of extrapolation would represent a significant expansion of CMS’s auditing practices.

In describing the revised audit methodology, CMS discusses the audit methodology that it has been using, which was initiated in a notice issued in 2010 and then finalized on February 24, 2012. (Notice of Final Payment Error Calculation Methodology for Part C Medicare Advantage Risk Adjustment Data Validation Contract-Level Audits). CMS notes that its intent has always been to use extrapolation in RADV audits. The 2012 notice addressed public comments to the 2010 announcement that raised concerns related to the use of extrapolation, and described the sampling and extrapolation methods CMS intended to use beginning with payment year 2011.

The proposed rule explains that although CMS used this methodology for audits in 2011, 2012, and 2013, it has not used it to seek contract-level recoveries. The announcement in the proposed rule “provid[es] additional notice and again welcom[es] public input on the agency’s methodology for calculating a contract-level payment error in RADV audits . . . .” CMS is considering applying its expanded audit and repayment policy as far back as payment year 2011.

CMS also explained why it is proposing not to apply a FFS adjuster to determine recovery amounts in a RADV audit, even though the 2012 notice called for the application of an adjustment factor as an offset in the MA organization’s favor to the preliminary recovery amount determined under the methodology. The FFS adjuster would “account[] for the fact that the documentation standard used in RADV audits to determine a contract’s payment error (medical records) is different from the documentation standard used to develop the [MA] risk-adjustment model (FFS claims).” CMS stated in the notice that it withdrew this element of the repayment determination methodology because the agency conducted a study (beginning in 2012) to evaluate the impact of errors in its FFS claims data and concluded that the “study suggests that diagnosis error in FFS claims data does not lead to systematic payment error in the MA program . . . .”

In brief, the study evaluated “claims level” inaccuracy based on a review of medical records associated with more than 8000 claims, and then looked at additional data to determine what CMS called beneficiary level inaccuracy. Even though CMS acknowledged that claims level inaccuracy can be 40 percent or higher, it concluded that the beneficiary level error associated with each HCC was in fact much lower, with a median error rate of 2 percent. From this study, CMS concluded that “it appears that diagnosis error in FFS claims data does not lead to systematic payment error in the MA program,” and thus declined to include an FFS adjuster in any final RADV payment error methodology. The executive summary of this study can be found here.

The proposed rule is preceded by a September 2018 decision by the U.S. District Court for the District of Columbia invalidating CMS’s MA overpayment regulations, largely on the basis that those regulations failed to account for error in fee for service data, and thus failed to comply with the statutory requirement that payments to MAOs must be actuarially equivalent to the fee for service program. CMS indicated in a footnote in the preamble to the proposed rule that the agency is “aware of the district court’s recent ruling,” but appears to believe that the results of the study discussed above support its conclusion that adjustment for fee for service errors is not required to achieve actuarial equivalence under its RADV audit methodology. This study was not part of the record in the litigation.

Finally, the proposed rule briefly indicated that CMS is considering expanding MA organizations’ RADV appeal rights in light of the potentially enormous implications of the use of extrapolation in RADV audits.

CMS specifically solicited comments on the use of extrapolation in RADV audits, the methodology published in the 2012 notice, expansion of MA organizations’ appeal rights, whether to apply the methodology as far back as 2011, and the findings and conclusions following CMS’s study regarding the FFS adjuster. To be assured consideration, comments are due to CMS no later than December 31, 2018.