Page 85

85

these differences to the payment year. Additionally, as discussed for the CMS-HCC model normalization factors (see Attachment III, Section N), CMS believes that, by using a multiple linear regression methodology, we can more appropriately take into account the significant change in the risk scores observed at the onset of the COVID-19 pandemic, when risk scores dropped significantly due to atypically low utilization. This methodology enables CMS to incorporate the most recent years of data in the risk score trends in order to reflect current risk and project a risk score that best estimates the average Part D risk score in the payment year given the currently available data. As noted in the CY 2025 Advance Notice and Rate Announcement and the CY 2026 Advance Notice, we did not propose the multiple linear regression methodology for RxHCC models for CY 2025 because we did not have the average 2023 Part D risk score available to evaluate whether the multiple linear regression approach was appropriate. For CY 2026, we have the 2023 Part D risk scores and find that the normalization factors calculated with the historical linear slope methodology are not consistent with recent trends. Our analysis showed that the normalization factor for PDPs is 7 percent higher than the 2023 PDP risk score when most recently the average PDP risk score has been decreasing, and the MA-PD normalization factor is 2 percent lower than the 2023 average MA-PD risk score when most recently MA-PD risk scores have been increasing. Comment: Many commenters opposed applying separate normalization factors for PDPs and MA-PD plans and suggested CMS switch back to a single factor like was used prior to CY 2025 instead. Commenters raised a number of concerns with implementing separate normalization factors, including that applying separate factors calculated using the multiple linear regression methodology inappropriately adjusts for coding and population differences between market sectors in a way that is inconsistent with how CMS has described the purpose of normalization before, and that more information and/or time is needed to evaluate the impacts of the IRA and the proposed factors before implementing such a change. Several commenters noted concerns about the disproportionate impacts on specific types of MA plans, such as those serving low-income beneficiaries, and perverse incentives for EGWPs to separate their enrollment into MA-only and PDP products when combined medical and drug benefits would be better. Commenters also suggested that applying separate normalization factors will adjust for differences between market sectors on average and will not capture variation, which may still lead to over- or under-payments for certain plans. A commenter suggested addressing the over- and under-prediction highlighted by CMS through separate models for PDPs and MA-PD plans rather than separate normalization factors. Many commenters who opposed applying separate factors stated that if CMS finalized separate factors, a different methodology should be used, such as an MA-PD normalization factor calculated without special needs plan enrollees, the linear slope methodology implemented for CY 2025, or an alternative that removes what the commenters argue is an additional adjustment for coding differences between PDPs and MA-PD plans from the calculation, thereby resulting in a similar