Page 73

73

the methodology for CY 2026 relative to CY 2025. With the statute in mind, CMS considers each year what information to provide as part of our proposals, and because no changes to the MA coding pattern adjustment were proposed for CY 2026, the underlying data and analyses were not provided. Section N. Normalization Factors for the CMS-HCC Risk Adjustment Models Comment: A few commenters expressed support for the methodology as proposed. A commenter believed that the proposed methodology better captures more recent demographic changes in the Medicare population. Another stated that the proposed regression factors fit actual FFS risk scores reasonably well, including for post-COVID years. Response: CMS appreciates the support of the commenters. CMS is finalizing the methodology for the normalization factors for the CMS-HCC and CMS-HCC ESRD risk adjustment models as proposed using the multiple linear regression methodology and 2020 to 2024 average FFS risk scores. Comment: Many commenters were opposed to the multiple linear regression methodology that CMS proposed to continue using for CMS-HCC models. Some commenters supported using a multiple linear regression method in general but suggested alternative methods such as using different years of data in the trend or applying the COVID flag to different years. Many alternative methods were suggested by commenters including other regression methods and the method CMS historically used to calculate normalization factors for the CMS-HCC models prior to CY 2025. Most of these commenters believed the Part C normalization factors proposed for CY 2026 are overstated because they put too much weight on the rebound in risk scores from 2021 to 2022. These commenters pointed out that alternative methods would result in lower normalization factors than the current proposed method. Response: CMS thanks the commenters for their feedback and for suggestions for alternative approaches to calculating the normalization factors. By using a multiple linear regression methodology, we can most 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 FFS risk score trend in order to reflect current risk and to project a risk score that best estimates the average FFS risk score in the payment year given the currently available data. We believe that the proposed multiple regression approach is reasonable given the variability in the observed data and the uncertainty inherent in any forecast. As stated above, CMS received a significant number of alternative recommendations from commenters about how to calculate Part C normalization factors. There was not an industry-wide consensus; the recommended alternatives were varied, sometimes conflicting, and produced different normalization factors with varying degrees of magnitude. Alternative methods suggested by commenters resulted in normalization factors that ranged from 4 percent higher