Page 81

81

expressing a belief that the model results in insufficient payment under the Part D redesign when plans take on more liability in the catastrophic phase. One additional commenter recommended that CMS incorporate a high-cost threshold into the model to account for outlier expenditures. Response: CMS recognizes the commenters’ concerns. We note that HCC-based risk adjustment models, including both the CMS-HCC and RxHCC models, are intended to predict expected relative expenditures across key subgroups of beneficiaries. As measured by our predictive ratios, the RxHCC model does well at predicting across levels of risk, including enrollees in the highest deciles of predicted risk, meaning beneficiaries broken out into groups based on their predicted drug costs. The models are not intended to predict the costs of individual beneficiaries, nor are they intended to have any influence on drug prescribing and uptake. Overall expected costs for a plan’s expected enrolled population are reflected in the bid, and the risk adjustment model is intended to ensure that the payments to the plan adequately reflect its expected relative cost, compared to the national average. CMS regularly updates the diagnostic and expenditure data underlying the RxHCC model so that the model is calibrated based on the most recent data available, which at this time is 2022 diagnoses and 2023 expenditures. For CY 2026, we are finalizing a model that uses those data years to reflect more recent utilization and cost patterns. Comment: Some commenters expressed concern about the model’s impact on low-income beneficiaries, including those enrolled in SNPs. A commenter believed that the proposed model’s risk scores for low-income beneficiaries did not account for the expected plan liability for these beneficiaries compared to non-low income beneficiaries. This commenter further stated their belief that because low-income beneficiaries generally have zero or minimal cost sharing, and many use protected class drugs, plans with high proportions of low-income or dually eligible beneficiaries have few methods for steering beneficiaries toward cost-efficient drugs. Response: CMS appreciates the commenters’ concerns about the impact of the model on low- income beneficiaries, including those in SNPs. While plan liability is increasing across all groups of beneficiaries, the average Part D risk score remains at 1.0, reflecting the average enrollee across the entire Part D market. Because the average Part D risk score remains 1.0, even if plan liability increases, risk scores may change in order to reflect how plan liability changes relative to the new overall average. Since the goal of risk adjustment is to ensure that payments to plans for beneficiaries who are expected to cost relatively more than average are higher than for beneficiaries who are expected to cost less than average, the role of the risk adjustment model is to ensure that the payments reflect these relative differences. We note that the diagnoses and costs of all low-income beneficiaries, including those enrolled in SNPs, are included in the RxHCC model calibration and accounted for with separate low-income segments, so any unique patterns of costs and utilization due to zero or minimal cost sharing among this group of beneficiaries will be reflected in the model relative factors. Further, we