ira与medicare d部分风险调整的潜在互动：模型考虑

August 2023 Potential IRA Interactions with Medicare Part D Risk Adjustment2August 2023Model ConsiderationThe 2023 RxHCC model is the most recent calibration,2F3which uses 2018 Medicare Fee-For-Service and MA encounter data for diagnosesand 2019 Prescription Drug Event expenditure data (i.e., using 2018 medical diagnoses to predict 2019 prescription drug costs). In otherwords, CMS determines 2023 direct subsidy payments using a model calibrated to claim experience from four to five years prior. WhileCMS accounts for cost trends, this lag does not capture other factors that may influence claim expenditures, such as new drug launches,patent expirations, or other market shifts between the calibration year and the payment year.AnalysisWhile CMS’ specific calibration methodology is not publicly known, CMS has historically made explicit adjustments to its calibrationmethodology to reflect material changes to the Part D program (such as the closing of the coverage gap). Given the materiality of theIRA’s impact on plan liability, we expect CMS to make some adjustments to the RxHCC model. However, the timing and details of suchupdates are unknown at this time.This paper discusses several elements for CMS to consider when updating the RxHCC model to reflect the IRA’s provisions. To assesstheir importance, we examine how these elements could impact plan payments prior to CMS making a model adjustment. This discussionis intended to illustrate the materiality of the IRA’s impact on plan costs and does not imply that we expect the current RxHCC model willbe in place in 2025 without a material calibration.Receiving guidance on planned model updates with enough time to react is of critical importance to plans as they make strategic decisionsfor 2025 given the material impact of risk adjustment on plan revenue.As with all risk scoring models, the current RxHCC model has certain limitations. The IRA benefit re-design represents asignificant structural change to the Part D benefit and necessitates model adjustments to account for the new structure.However, even with an appropriate re-calibration, the model will still maintain many of its current limitations.The Significance of the Direct SubsidyIn the Part D program, plan revenue is comprised of the following components:Figure 1: Plan revenue calculationUsing 2019 and 2020 claims and risk score data for FFS and MA beneficiaries with Part D coverage from the CMS 100% ResearchIdentifiable (RIF) dataset, we created a comparison of plan paid amounts3F4to risk scores to understand which conditions are over- orunder- predicted under various scenarios. We compared the relativity of plan paid amount gross of rebates to the market to the relativityof risk score to the market (“paid-to-risk score ratio”). A ratio less than 1.0 indicates the RxHCC model over-predicts plan costs, while aratio greater than 1.0 indicates an under-prediction of plan costs.32023 Medicare Advantage and Part D Rate Announcement Fact Sheet | CMS4Throughout this paper “plan paid” or “plan liability” refer to gross drug costs less beneficiary cost sharing, coverage gap discount payments (status quo) / manufacturerdiscount payments (IRA), federal reinsurance, and low income cost sharing subsidies. This amount is gross of manufacturer rebates unless explicitly noted. Costs are trendedto 2025 and reflect actual benefit designs for status quo scenarios and are adjudicated under the IRA defined standard design for IRA scenarios. Potential IRA Interactions with Medicare Part D Risk AdjustmentModel ConsiderationFigure 2 displays an example of an over-predicted RxHCC:Figure 2: Example plan-to-risk score ratio calculationComponentRxHCC risk scorerelativityActual plan liabilityfor beneficiaries with thisRxHCCAverage plan liability for the market(across all conditions)Relativityofactual plan liabilityto the market averagePaid-to-risk score ratioIn this hypothetical example, the actual plan liability for a beneficiary with this RxHCC is $2,000, compared to an assumed average liabilityof $1,000 for the market across all conditions. In other words, the plan’s liability for a beneficiary with this RxHCC is 2.0 times higher thanthe market average. However, the risk score is 2.25 times higher than the market average, and therefore, greater than the ratio of actualplan liability to the market average, which results in a paid-to-risk score ratio of less than 1.0 (0.89). As noted above, a paid-to-risk scoreratio below 1.0 indicates the RxHCC model over-predicts plan costs, such that the plan would be compensated more than necessary tocover actual claim costs.Please note, this analysis focuses on a comparison between relative risk scores and relative costs before and after the IRA and not ontotal plan compensation. Because we anticipate the direct subsidy to increase considerably, risk scores will become a significantly morematerial component in the plan revenue equation.It is important to keep in mind that "over / under prediction