Page 79

79

not using these drugs. Another commenter agreed that incorporating MFPs would ensure accuracy of the model but encouraged CMS to blend this model with the CY 2025 RxHCC model to limit volatility of risk score changes. Response: CMS thanks the commenters for their support and, for CY 2026, is finalizing the updates that reflect the CY 2026 Part D benefit, including MFPs, for both RxHCC models (i.e., the model being implemented for non-PACE organizations and the model being implemented solely for PACE organizations), as proposed. We agree that if the models did not use the agreed- upon MFPs, inaccurate estimates of relative plan liability for CY 2026 would likely result. We believe that a phase-in of the proposed updated RxHCC risk adjustment models would not be appropriate because the payments to Part D sponsors in CY 2026 would less accurately reflect the expected changes in relative plan costs under the 2026 Part D benefit design. Comment: Nearly all commenters who commented specifically on the proposal to update the underlying data used in the model for non-PACE, and in part for PACE, supported the proposal to use diagnoses from 2022 FFS claims and MA encounter data and gross drug costs from 2023 PDEs, as well as to update the denominator year to 2023. These commenters believed that more recent data is likely to better reflect cost and utilization patterns expected in 2026, with one commenter requesting that CMS update the model to reflect 2025 data as soon as possible to reflect beneficiary diagnoses and expenditures under the IRA. A commenter expressed concern about the use of 2022 diagnoses to calibrate the model, saying that diagnoses from that year were deflated due to the ongoing impacts of the COVID-19 pandemic on utilization. Response: While we acknowledge the commenter’s concern with using diagnoses from 2022 dates of service, we continue to believe that the value of more recent data outweighs concerns about any potential continued impact of the pandemic on utilization in 2022 and are finalizing the data year update for the RxHCC model being finalized for non-PACE organizations (and in part for PACE organizations) that is calibrated on 2022 diagnoses and 2023 expenditure data. The RxHCC model is constructed such that spending associated with unreported diagnoses is instead reflected in demographic coefficients or in correlated RxHCC coefficients (such as for comorbidities), even if the model is calibrated on data with fewer diagnoses than is typical. Comment: Several commenters expressed concern with reflecting MFPs in the calibration, suggesting that the data not be included at all or that MFP incorporation be delayed until the model makes structural changes to incorporate rebates or until the effects of MFPs on enrollee utilization and formulary design can be studied further. A few of these commenters specifically commented that substituting gross drug costs with MFPs without accounting for prior rebate data could result in distorted plan liability estimates, with one commenter requesting that CMS clarify whether direct and indirect remuneration (DIR) allocated to drugs with substituted MFPs was also removed from the calibration, while others said that CMS has not had time to assess the impact of MFPs on Part D spending and liability, with one saying that reduced cost-sharing for