Background on the Part D Risk Adjustment Model

Each year, Part D plans submit bids to the Centers for Medicare & Medicaid Services (CMS) that predict the expected costs to provide drug coverage to enrollees in their plans. Plans must adjust their bids based on the health status and demographic characteristics of their enrollees. Plans make this adjustment using CMS’s Part D risk-adjustment model, also known as the Prescription Drug Hierarchical Condition Category (RxHCC) model.

The Part D risk-adjustment model predicts plan liability based on the health status, medical diagnoses,¹ and demographic characteristics (e.g., age, gender, low-income status, institutionalized status) of a plan’s enrollees. CMS notes that the purpose of risk adjustment is to adjust plan payments from CMS to account for enrollee health conditions and risk factors and to incentivize plans “to enroll all beneficiaries, regardless of health status.” Risk adjustment is an important tool to avoid adverse selection in the Part D program. In the absence of adequate risk adjustment for certain high-cost conditions, plans may have incentives to discourage enrollment of beneficiaries with those conditions through narrower formularies or more utilization management (UM).

Aside from cost sharing and premium subsidies for certain low-income enrollees, plans receive two types of payments from CMS: a direct subsidy and a reinsurance payment. The direct subsidy is the plan’s bid, adjusted by the Part D risk model, minus the beneficiary’s premium. The reinsurance payment covers the expected spending for the government’s share of costs for claims in the catastrophic phase of the benefit. For 2023, the National Average Monthly Bid Amount (NAMBA) was $34.71, while the government reinsurance amount was $93.68. Thus, about 27% of Part D drug spending is risk adjusted while the majority (73%) is paid for by the government through reinsurance payments.

Risk-adjusted payments are a factor considered by plans when designing their formularies and benefit designs. While the predictive accuracy of the risk adjustment model may be of interest to plans, it has not received much attention to date from policymakers.² Although researchers who developed the model for CMS assessed the accuracy of the Part D model, the study is based on the Part D benefit that was in effect 10 years ago, and it did not assess predictive accuracy by therapeutic area.

While the Part D risk-adjustment model is intended to incentivize plans to enroll all beneficiaries by appropriately adjusting plan payments to account for enrollment by more costly (or less costly) beneficiaries, concerns persist around the accuracy of the model in adjusting payments for certain conditions. Inaccuracy of the risk adjustment model can lead to plan payments from CMS that do not appropriately cover actual costs and that create incentives for plans to design their formularies and benefit designs to minimize financial losses.

Predictive Accuracy and Limitations of the Current Part D Risk-Adjustment Model

To examine accuracy of the Part D risk-adjustment model, Avalere conducted an analysis comparing 2019 actual Part D plan liability to plan liability as predicted by the risk adjustment model for enrollees in Medicare Advantage Prescription Drug Plans (MA-PDs) and standalone Prescription Drug Plans (PDPs). As shown in Table 1, plan liability is underpredicted by the risk adjustment model for each of the examined conditions. For example, spending is underpredicted by 53% for oncology and by more than 60% for three other therapeutic areas (TAs), including autoimmune conditions, hepatitis C, and multiple sclerosis.

These inaccuracies could impact how plans consider these TAs when setting their formularies, as conditions for which plan liability is underpredicted could result in a financial loss to plans. However, the impact today is likely to be moderate because the risk-adjusted portion of plan liability under the current benefit design accounts for 27% of current spending, as noted above. As discussed in more detail below, the impact of the Part D risk adjustment model on plan payment and in setting plan formularies will be much greater under Part D benefit redesign.

* 2019 plan liability reflects gross liability and does not include rebates.

Several factors could lead to this predictive inaccuracy. Among other limitations, the current Part D risk-adjustment model:

• Assigns disease categories based on medical diagnoses and does not incorporate data reflecting prescription drug utilization or disease severity
• Does not reflect net plan liability (i.e., it does not account for manufacturer rebates)
• Encompasses a limited number of disease groupings, such that some conditions do not have a disease grouping; accordingly, the model does not account for spending associated with rare conditions

Impact of Part D Benefit Redesign on Risk-Adjusted Plan Payments

The Inflation Reduction Act (IRA) redesigns the Part D benefit, including changes to stakeholder liabilities, beginning in 2025 (Figure 1).³

Note: Reflects benefit design for applicable drugs (i.e., brand drugs, biologics, biosimilars). Standard benefit in 2023 applies to non-low-income subsidy (LIS) enrollees only. Benefit design beginning in 2025 applies to both non-LIS and LIS enrollees.

Under Part D redesign, plans will assume additional liability, including:

• In the catastrophic phase: Plan liability in catastrophic will increase from 15% in 2023 to 60% beginning in 2025.
• For spending that currently falls in the coverage gap: Today, plans pay 5% for drug spending in the coverage gap for non-LIS beneficiaries. Beginning in 2025, the coverage gap will be eliminated, and plans will have 65% liability for drug spending that falls in the initial coverage phase (i.e., spending that falls below the new $2,000 cap under Part D redesign) for both LIS and non-LIS beneficiaries equally.
• For LIS beneficiaries: Currently, the Medicare low-income cost-sharing subsidy covers 100% of spending for LIS beneficiaries in the coverage gap. Beginning in 2025, the new benefit design will apply to both LIS and non-LIS enrollees, increasing plan liability to 65% from 0% for LIS enrollees. The increase in liability is due to the new plan liability for LIS in the coverage gap and the fact that a greater share of LIS beneficiaries’ total drug spending falls in the catastrophic phase compared to non-LIS enrollees.

Analysis of Change in Liability for Specific Therapeutic Areas Under Benefit Redesign

To quantify the impact of benefit redesign on specific TAs, Avalere analyzed the change in plan liability for 11 TAs for MA-PDs and PDPs under the new Part D benefit parameters enacted in the IRA (Table 2). Plan liability will increase by more than 200% for enrollees taking a medication in 4 out of the 11 TAs (autoimmune conditions, HIV, hepatitis C, and multiple sclerosis). For six of the remaining TAs, plan liability will increase by more than 100%. In two of the most highly utilized TAs analyzed–anticonvulsants and asthma/ chronic obstructive pulmonary disease (COPD)–plan liability will increase by 116% and 102%, respectively.

* Plan liability reflects gross liability and does not include rebates.

Due to the increase in plan liability throughout the Part D benefit, the portion of plan payments financed through risk-adjusted payments will increase substantially. Avalere projects that the NAMBA–which is risk adjusted–will account for 84% of total drug spending in 2025 compared to 27% in 2023. As plan bids increase under benefit redesign, risk adjustment for a given condition will affect a much larger dollar value than it does today. Therefore, if the risk-adjustment model underestimates spending for enrollees with specific diseases, the potential financial losses to plans could be larger under redesign than under the existing benefit structure.

Inaccuracies in the risk-adjustment model between actual costs and predicted costs could create misaligned incentives for plans. That is, given inaccuracies in the model, plans may choose formulary and benefit strategies that mitigate potential financial losses for conditions that may be consistently underpaid by the risk-adjustment model. In addition, plans may look to implement formulary and benefit designs that could be more favorable to conditions that receive more adequate payment adjustments under the risk-adjustment model. As shown above in Table 1, these challenges exist today but could be exacerbated further due to the increase in plan liability shown in Table 2 that will accompany redesign.

As plan liability and share of spending throughout the benefit shifts under benefit redesign, the conditions and beneficiaries for whom risk adjustment is inaccurate could also change. Increased plan liability under benefit redesign could lead to changes in plan formulary and benefit designs, shift contracting strategies with manufacturers, and impact patient access to medicines. According to the actuarial firm Milliman, plans “might consider narrowing their formulary coverage to help mitigate some of the plan liability increases.” Other potential changes cited by Milliman include “new UM strategies, such as additional step therapy edits or prior authorization controls.”

Conclusion

Avalere’s analysis shows that costs are underpredicted under the current Part D risk-adjustment model for enrollees taking medicines in all 11 of the TAs analyzed. This analysis shows that costs are underpredicted by more than 60% for three of these TAs. Due to changes from the IRA, a much larger portion of drug spending will be risk adjusted in the future–an estimated 84% compared to 27% of drug spending subject to risk adjustment in 2023. As such, the accuracy of the risk-adjustment model could play an even larger role in plans’ formulary decision making and impact plan contracting with manufacturers. As CMS works to update many aspects of the Part D program to implement the IRA, risk adjustment is one focal point that could help ensure adequate plan payments and that appropriate incentives are in place for plans to enroll beneficiaries regardless of health status.

Funding for this research was provided by the Pharmaceutical Research and Manufacturers of America (PhRMA). Avalere retained full editorial control.

To learn more about how changes to the risk-adjustment model could affect your business, connect with us.

Methodology

Avalere used 2019 Medicare Part D Drug Event (PDE) data and 2018 Medicare Fee-For-Service claims, governed by a research-focused data use agreement with CMS to characterize the increase in plan liability under the IRA. TAs were identified from Medispan data by mapping the PDE data by National Drug Code (NDC) to the Generic Product Identifier-2 level groupings. Avalere’s analysis reflects diagnoses and claims from a 20% random sample of Medicare Part D enrollees. The analysis also reflects diagnoses and claims from enrollees contained in Inovalon’s MORE² Registry®. This database contains about 20% of MA enrollees. The number of beneficiaries identified as filling prescriptions in each TA may not represent 20% of the population taking products in these TAs.

To estimate plan liability, Avalere simulated the Part D benefit both under current law and under the IRA benefit parameters (based on the benefit that will be in effect in 2025), as follows:

1. Identify all beneficiaries taking at least one of the products from the 11 therapeutic areas in the 2019 PDE data
2. Array each beneficiary’s Part D claims in the PDE file chronologically based on dates of service
3. Identify formulary coverage from the formulary file for each claim based on NDCs of the PDE
4. Determine tier placement and cost sharing for each drug on formulary based on the plan characteristics file

Notes

1. Under the Part D risk adjustment model, medical diagnoses are sorted into groups known as RxHCCs.
2. CMS has not published an analysis of the accuracy of the Part D RxHCC model, although the agency has published such analyses of the model used for Medicare Advantage payments (the CMS-HCC model).
3. The IRA also implements an out-of-pocket cap at the catastrophic threshold beginning in 2024 before transitioning to the new benefit design in 2025. In the calendar year 2025 Advance Notice, CMS did not propose updating the Part D risk-adjustment model for the 2024 plan year.