Context: Reforming the Medicare Part D program—which provides prescription drug coverage to 49 million beneficiaries—has emerged as a key policy priority.
Methods: The authors evaluate prescription drug claims from a 100% sample of Medicare Part D beneficiaries to evaluate the current spending distribution across different payers for different types of beneficiaries across different benefit phases. They then model how these estimates would change under a proposal to redesign the Medicare Part D standard benefit.
Findings: Spending patterns differ for beneficiaries who do and do not qualify for low-income subsidies. Part D plans face limited liability for total spending under the current standard benefit design, amounting to 36% of total spending for beneficiaries who do not receive low-income subsidies and 28% of total spending for those who do. Proposed reforms would increase plan liability and significantly change the distribution of liability across plans, drug manufacturers, and the federal government.
Conclusions: Though the original goal of the Part D program was to create a market of competing private plans that provide prescription drug coverage to Medicare beneficiaries, the standard benefit design that was included in the original legislation reflected significant political compromises. Reforming the standard benefit design to give plans more skin in the game could significantly affect competition in the market, with differential impact across drug classes and types of beneficiaries.
Reforming the Medicare Part D program has emerged as a key priority among policy makers and the public (Kirzinger et al. 2021). Medicare Part D—created by the Medicare Prescription Drug, Improvement, and Modernization Act (MMA) of 2003—provides prescription drug coverage to 49 million beneficiaries at an annual cost to the federal budget of $105 billion (Medicare Trustees 2021). While the vast majority of Part D beneficiaries are satisfied with their coverage (Morning Consult 2021), the original design of the program was a product of political compromises, and in the face of changes to the prescription drug market over the last two decades, the program is ripe for reform.
Some historical background is useful to put these potential reforms in context. Before the enactment of the MMA, many seniors lacked prescription drug coverage (Hoadley 2006). While there was widespread support among policy makers for coverage expansion, there was considerable debate over whether the federal government should provide the benefit directly and whether all beneficiaries should be eligible or just those of modest means (Oliver, Lee, and Lipton 2004). With the endorsement of the George W. Bush administration, Congress ultimately enacted the MMA, creating Medicare Part D—a government-subsidized but privately administered prescription drug coverage program, in which private plan sponsors set premiums, negotiate prices with manufacturers, establish formularies, and use common management tools to control costs, subject to minimum benefit requirements. The rationale in part was to enable plan sponsors to compete on both cost and quality to provide beneficiaries with a wide selection of plan options and premiums. However, there was also significant concern about whether private insurers would be willing to participate in this program—particularly if they were not permitted to set premiums—since private, stand-alone prescription drug plans were unprecedented, and many plan sponsors expressed concern about financial liability driven by adverse selection. It is somewhat difficult to conceive today, given the robust plan participation that has actually occurred, but this was a significant enough concern that the MMA included provisions for a federally run “fallback” drug plan to be implemented in the event that an inadequate number of private plans were willing to participate in the program (Holtz-Eakin and Book 2013).
An important constraint in the development of Part D was the need to keep its cost below the budget reconciliation threshold of $400 billion over 10 years, which was much less than the estimated $1.85 trillion that Medicare beneficiaries were expected to spend on prescription drugs over that period (Oliver, Lee, and Lipton 2004). Meanwhile, politicians were concerned about backlash if beneficiaries were disappointed with coverage generosity from the newly created program. Ultimately, the benefit design included in the MMA reflected these joint constraints rather than a typical insurance benefit, most notably through a significant gap in coverage in the middle of the benefit design, known as the “doughnut hole” (Oberlander 2007).
In 2006—the first year in which plans were offered—the Part D standard benefit design included a $250 deductible followed by 25% coinsurance up to $2,250 in total drug costs, after which beneficiaries entered the doughnut hole, in which they were responsible for the full cost of their drugs up to $3,600 in out-of-pocket costs (or $5,100 in total drug costs). After that, they reached catastrophic coverage, in which the federal government paid 80% of drug costs through the federal reinsurance program, beneficiaries paid 5% coinsurance, and plan sponsors were responsible for the remaining 15% of spending. This structure enabled many beneficiaries to see some coverage benefit while keeping overall program costs low through the coverage gap. Additionally, the reinsurance program, together with risk corridors, encouraged plan participation by limiting sponsors’ exposure to high financial risk.1 Moreover, to ensure affordability for low-income beneficiaries, the federal government provided additional subsidies to cover their premiums and cost sharing.2 Around one-third of Medicare Part D beneficiaries qualify for these low-income subsidies (LIS).
While Part D offered a significant coverage expansion and reduction in out-of-pocket expenditures on prescription drugs, the existence of the doughnut hole left many beneficiaries who did not qualify for low-income subsidies with concerns about affordability and poor health outcomes as a result of reduced adherence (Joyce, Zissimopoulos, and Goldman 2013). The Affordable Care Act of 2010 (ACA) (and, subsequently, the Balanced Budget Act of 2018) filled in the doughnut hole through a combination of funding sources. For non-LIS beneficiaries, manufacturers of brand-name drugs are now required to pay 70% of costs incurred in the coverage gap, while plans are responsible for 5% of the cost of these drugs. Additionally, plans are responsible for 75% of generic drug costs, leaving non-LIS beneficiaries with 25% coinsurance on claims incurred in the coverage gap for both branded and generic drugs. In contrast, plans and manufacturers are not liable for coverage gap claims for beneficiaries who qualify for LIS. Instead, the federal government directly pays for the full cost of these drugs, with the exception of nominal beneficiary cost sharing. As a result, while any given plan typically has a mix of LIS and non-LIS enrollees, plan sponsors’ (and manufacturers’) liability for drug spending varies across benefit phase, drug type, and beneficiary LIS status.
In addition to these policy changes, trends in prescription drug spending have evolved quite significantly since the introduction of Part D. Generic entry has occurred for many of the blockbuster brand-name drugs that accounted for the majority of Part D spending in the early years of the program, leading to significant reductions in spending on those drugs. However, higher list prices for newer branded drugs and the introduction of new specialty drugs over the last decade has led to an increasing share of Part D spending occurring among the small share of beneficiaries who reach catastrophic coverage. These patterns have raised concerns about the Part D benefit design along several dimensions. First, while the real value of insurance arises when risk is catastrophic, Part D remains one of the few large insurance programs that does not have an out-of-pocket maximum for beneficiaries, and the number of beneficiaries facing such unlimited liability has increased over time (Trish, Xu, and Joyce 2018). In addition, there is growing evidence that the private plans who administer the program have exploited the program's rules and idiosyncrasies (MedPAC 2015; Trish 2019). For example, filling the doughnut hole did more than reduce out-of-pocket costs; because manufacturer discounts count toward “true out-of-pocket costs,” it also accelerated entry into the catastrophic phase, where government responsibility is highest.3 The net result is that, today, Part D plans are responsible for only about one-third of total prescription drug spending, raising questions about the extent to which they are actually incentivized to effectively manage utilization and spending, particularly among high-cost enrollees (Trish et al. 2019). Moreover, research and real-world experience indicate that this benefit design has propagated other market distortions, including encouraging growth in list prices (coupled with high rebates, or after-the-fact discounts paid from manufacturers to pharmacy benefit managers and/or plans) and strategic bidding of expected reinsurance subsidies, enabling plans to extract excess federal subsidies (Dusetzina et al. 2017, 2021; MedPAC 2015; Trish 2019).
Taken together, these concerns have generated momentum to reform and modernize the Part D standard benefit design through redesign provisions included in policy proposals such as the Build Back Better Act (BBBA), the Prescription Drug Pricing Reduction Act, the Elijah E. Cummings Lower Drug Costs Now Act, and the Inflation Reduction Act. While the details of these specific redesign proposals vary, they tend to share several common themes, including creating a maximum out-of-pocket cap for beneficiaries when they reach the catastrophic spending phase of the program; ensuring that Part D plans are incentivized to promote drugs that offer the most value at the lowest cost—for example, by reducing government subsidies for reinsurance and shifting responsibility for large claims to plans themselves; requiring branded drug manufacturers to subsidize a portion of drug spending throughout the benefit design (or after reaching the catastrophic phase) rather than being limited to the coverage gap; and streamlining the benefit between the deductible and the catastrophic out-of-pocket threshold to eliminate the coverage gap.
Several policy and actuarial analyses have modeled the potential impact of various policy proposals (Hayes, Alcocer, and Yi 2019; O'Neill Hayes 2020; Wouters and Stone 2020). However, these analyses tend to overlook the distinctions between LIS and non-LIS beneficiaries, both in terms of differences in benefit design and spending patterns across these two groups today, as well as the potential implications of policy changes such as those described above that would harmonize the benefit design across these two groups. While previous work has highlighted the differential implications of extending manufacturer-financed discounts to LIS beneficiaries across drug classes (Gottlieb and Ippolito 2019), the implications of benefit redesign across these two groups more generally have been less explored. Here, we provide insight into how spending patterns among LIS and non-LIS beneficiaries groups differ today, and we examine the ways in which these differential spending patterns contribute to changes in the distribution of liability across payers, beneficiaries, manufacturers, and the federal government under the proposed redesign of the Part D standard benefit as included in the BBBA. We note that there are small differences from the redesign parameters included in the newly enacted Inflation Reduction Act, but the general takeaways would be similar.
We use a 100% sample of 2018 Medicare Part D claims accessed through the Centers for Medicare and Medicaid Services (CMS) Virtual Research Data Center. We include beneficiaries enrolled in stand-alone prescription drug plans and Medicare Advantage plans with Part D coverage. We exclude beneficiaries enrolled in employer-group waiver plans (16% of enrollment) because benefit information for these plans is incomplete in our data.
Nearly all Part D plans (accounting for >99.9% of enrollment in our sample) do not actually use the standard benefit design, but instead use formularies with tiered cost-sharing parameters in the initial coverage phase. Additionally, most beneficiaries (62% of our sample overall; 76% of non-LIS) are enrolled in enhanced plans, which provide more generous coverage than the standard benefit design, such as by reducing or eliminating the deductible for generics or all drugs, typically in exchange for a higher premium. To simplify the analysis, update it to 2020 parameters, and make it more consistent across beneficiaries in different plans and across alternative hypotheticals, we first “reprocess” each claim under the 2020 standard benefit design and advance beneficiaries through the benefit phases accordingly. This reprocessing step is important because aggregate spending under plans’ actual benefit design tends to differ from aggregate spending under the simulated standard benefit design, suggesting that a comparison between spending under the proposed standard benefit redesign and actual spending could result in misleading estimates of the impact of such a policy change on the distribution of spending. Part of this difference is because of the high share (62%) of enrollees in enhanced plans, but we find that these differences in actuarial value of actual plans relative to the standard benefit design exist even for basic plans (see appendix). We assume that the total spending on the claim remains constant—that is, there are no changes to pricing or utilization under the current standard benefit design (relative to actual plan benefit design) or under the proposed benefit redesign.
To execute this reprocessing, we assign liability for each claim according to the 2020 standard benefit design, which includes the following parameters for non-LIS beneficiaries:
Initial coverage (up to $4,020 in total drug costs): 25% coinsurance; 75% plan liability
Coverage gap (up to $6,350 in true out-of-pocket spending, including manufacturer-financed discounts): 25% coinsurance; and:
Branded drugs: 70% manufacturers; 5% plan liability
Generic drugs: 75% plan liability
Catastrophic coverage: 5% coinsurance; 15% plan liability; 80% federal reinsurance
The standard benefit design for LIS beneficiaries is the same as that described above for non-LIS beneficiaries except in the coverage gap, where the federal government is responsible for all spending for both generic and branded drugs (through low-income cost-sharing subsidies) except for nominal LIS beneficiary cost sharing. That is, neither plans nor manufacturers face liability in the coverage gap for LIS beneficiaries.
To account for cost sharing paid by LIS beneficiaries (across all phases), we apply low-income cost-sharing rules (which describe the actual out-of-pocket liability rather than the coinsurance parameters described above) according to 2020 CMS cost-sharing parameters based on beneficiary cost-sharing codes included in the data. Actual out-of-pocket liability varies according to LIS beneficiary type, benefit phase, and drug type (branded or generic), but, for example, many LIS beneficiaries face $0 or nominal copayments such as $1.30–$3.60 for generic drugs and $3.90–$8.95 for branded drugs. The remainder of the out-of-pocket liability described in the standard benefit design parameters above is instead attributed to low-income cost-sharing subsidies. We assign beneficiaries to being LIS if they qualify for LIS at any point in the year, and we reprocess all of their claims for the year under the LIS benefit. We drop 8,009 beneficiaries (0.02%) who have one or more claims with nonstandard cost-sharing codes because we cannot identify the specific applicable low-income cost-sharing obligations. We categorize drugs as branded if the brand name field does not match the generic name field on the claim and as generic if they match.
After reprocessing all claims under the standard benefit design, we then reprocess each beneficiary's claims under the alternative benefit design structure proposed under the BBBA. The proposed standard benefit design parameters under the BBBA are:
Initial coverage (up to $2,000 in out-of-pocket spending): 23% coinsurance, and:
Branded drugs: 67% plan liability; 10% manufacturers
Generic drugs: 77% plan liability
Catastrophic coverage: 60% plan liability; and:
Branded drugs: 20% manufacturers; 20% federal reinsurance
Generic drugs: 40% federal reinsurance
While this proposal would be phased in over several years (by which point the deductible and other thresholds may have increased), we model it as if fully implemented and using the 2020 deductible. Though the BBBA does not describe specifically if it would make changes to cost sharing paid by LIS patients, we assume that the current LIS cost-sharing rules would apply (by drug type, phase, and LIS category). Additionally, we note that the BBBA would extend manufacturer discounts for branded drugs to LIS beneficiaries, and we incorporate that into our modeling.
We compare spending by payer and beneficiary ending phase under the current and proposed alternative (BBBA) standard benefit designs, separately for LIS and non-LIS beneficiaries. For non-LIS beneficiaries, we also compute the change in out-of-pocket spending under the proposed redesign parameters compared with their out-of-pocket spending under the 2020 standard benefit design. Finally, we assess the variation in changes in liability across payers under proposed reforms for top-spending drug classes. We categorize drug classes by merging National Drug Codes to a description based on the first three characters of the Hierarchical Ingredient Code (HIC3 class) from the First Databank drug database, and then aggregating categories related to the conditions for which they are indicated.
Our sample includes 37.3 million Part D beneficiaries, with total spending of $116 billion. Non-LIS beneficiaries account for 68% of Part D enrollment but only 48% of total spending (figure 1).
Spending patterns differ between LIS and non-LIS beneficiaries (figure 1). Under the 2020 standard benefit design, nearly half of non-LIS beneficiaries have low total annual spending, ending the year with no claims (8%) or in the deductible (40%). Only 3% of non-LIS beneficiaries reach catastrophic coverage, but 32% of spending by non-LIS beneficiaries occurs in this final phase. In contrast, LIS beneficiaries have a more bifurcated spending distribution; relative to the non-LIS beneficiaries, a higher share of LIS beneficiaries have no claims for the entire year (19%), but a higher share also reaches catastrophic coverage (14%), and 46% of spending by LIS beneficiaries occurs in this final phase.
Under the 2020 standard benefit design, plans are responsible for only 36% of total spending for non-LIS beneficiaries and only 28% of total spending for LIS beneficiaries, and this liability is concentrated primarily in the initial coverage phase, particularly for LIS beneficiaries (figure 2). Beneficiaries (28%), federal reinsurance (27%), and manufacturers (9%) finance the remainder of spending for non-LIS beneficiaries, while the federal government covers nearly all of the remaining liability for LIS beneficiaries through reinsurance (38%) and low-income cost-sharing subsidies (33%).
Relative to the status quo, the most notable change in beneficiaries’ final ending phase under the BBBA redesign proposal is driven by the fact that the coverage gap phase is eliminated and instead combined into the initial coverage phase. The catastrophic coverage threshold is also lower under BBBA relative to the current standard benefit design, so a higher share of beneficiaries would reach this phase (5% of LIS and 19% of non-LIS (figure 3).
Plan liability would increase considerably under the BBBA redesign proposal (56% of total spending for non-LIS beneficiaries and 60% for LIS beneficiaries compared with 36% and 28% under current law, respectively) (figure 4). Notably, the BBBA redesign would significantly increase plan liability for spending that occurs while beneficiaries are in later benefit phases and would significantly decrease federal reinsurance liability (to 8% of total spending for non-LIS beneficiaries and 12% for LIS beneficiaries, compared with 27% and 38% under current law, respectively). Additionally, manufacturer liability for non-LIS beneficiaries would increase (to 11% of total spending, compared with 9% under current law) and be extended to LIS beneficiaries (13% of total spending, compared with 0% under current law).
Aggregate changes in liability by payer under the proposed BBBA redesign provisions are significant for plans, manufacturers, and the federal government (figure 5). Plan liability would increase by $31.3 billion, with 63% of that change arising from increased liability for LIS beneficiaries. There is a considerable reduction in federal spending on reinsurance ($26.2 billion), in addition to reductions in federal spending on low-income cost-sharing subsidies amounting to $11.4 billion. Manufacturer liability would increase by $8.6 billion, with 88% of that increase being driven by LIS beneficiaries.
Reductions in aggregate patient out-of-pocket spending under the BBBA redesign proposal are modest at $2.1 billion, assuming no behavioral changes (figure 5). About half of non-LIS patients would see some reduction in out-of-pocket spending under the BBBA redesign provisions, driven by modestly reduced coinsurance in the initial coverage phase (figure 6). Additionally, 4% of non-LIS patients would save more than $500 out of pocket annually relative to the status quo, driven by the BBBA's cap on annual out-of-pocket spending.
The BBBA redesign provisions’ impact on changes in plan liability, manufacturer liability, and average (non-LIS) patient out-of-pocket spending vary widely across drug classes (table 1). For example, plan liability as a share of spending would increase by 43 percentage points for cancer drugs (from 18% of total spending on cancer drugs to 60%) and by 15 percentage points for anticoagulants (from 45% to 61%). This variation in the changes in plan liability are driven by the fact that plan liability as a share of total drug spending varies considerably across classes under the current standard benefit design, with plan liability tending to be lower for drugs that are more commonly taken among beneficiaries in catastrophic coverage and/or by LIS beneficiaries. Similarly, the increase in manufacturer-financed discounts as a share of total spending also varies considerably, for example with manufacturer liability increasing 17 percentage points for RA drugs (from 2% to 19%) compared with a decrease of 1 percentage point for anticoagulants (from 11% to 10%). The impact on average non-LIS annual patient out-of-pocket spending (among users and restricted to spending on drugs in that class) varies considerably across classes, with considerable decreases for MS drugs (-$3,454) and hepatitis C drugs (-$3,084) compared with negligible decreases (-$5) for mental health drugs.
Despite the policy goals of creating a market-based program where competing private plans would deliver on value to beneficiaries and taxpayers, over time the prescription drug market has evolved in such a way that plans actually have quite limited liability for total spending in the Part D program. This is particularly true for LIS beneficiaries, who account for the majority of spending, despite being only one-third of enrollment. Policy changes to fill in the doughnut hole—while reducing out-of-pocket burden for many non-LIS beneficiaries—exacerbated the differences in benefit design between LIS and non-LIS beneficiaries for plans, manufacturers, and the federal government. Moreover, differences in how drug spending is distributed across benefit phases—for both LIS and non-LIS beneficiaries—has resulted in significant variation in who bears the brunt of financial risk for different drugs. This has created heterogeneity in the competitive pressure that plans face in terms of negotiating drug prices with manufacturers and has led to other market distortions, such as an emphasis on preferring to cover drugs with higher list prices and higher rebates rather than negotiating for lower list prices. While beneficiaries and taxpayers may benefit from these rebates in the form of lower premiums, these distortions disproportionately harm beneficiaries who take these highly rebated drugs by imposing higher cost sharing on them (Lakdawalla and Li 2021).
Current proposals provide an opportunity to improve competition in the Part D market and give plans stronger incentives to effectively manage spending and utilization across all beneficiaries. However, plans will likely also need additional options to do so. For example, policies have been proposed to give plans more flexibility regarding coverage for drugs in protected classes and around providing modest cost-sharing flexibility for LIS beneficiaries, which could be particularly important given the significant changes to plan liability for LIS beneficiaries that we found would occur under the redesign proposals we studied (MedPAC 2020). Moreover, given the significant increases in plan liability described in these proposals, it is important to note that, absent changes to the competitive bidding system, the federal government would redirect much of the savings derived from the reinsurance program to increase direct subsidies to plans to offset what would otherwise be considerable premium increases. Indeed, the BBBA also included provisions to increase premium subsidies to 76.5% of program spending (from 74.5% under current law).
Our findings provide insight into the current state of the Part D market as well as changes that would occur under proposed reforms. The current distribution of beneficiaries’ final ending phase is rather striking, particularly the significant share of enrollees with no claims or low enough annual spending to end the year in the deductible. While actual benefit designs may have a reduced deductible or no deductible in practice, this general pattern suggests that the majority of beneficiaries will be relatively unaffected by benefit design changes, though they could still be affected by secondary effects such as reduced drug prices through stronger negotiation. Additionally, beneficiaries may face increased access concerns if reforms lead plans to more strongly manage utilization, and those differences might be more striking among classes where plans gain significant financial liability relative to the status quo. Alternatively, redesign provisions could be coupled with other reforms that could reduce beneficiary out-of-pocket spending—such as basing cost sharing on net prices rather than list prices or capping monthly out-of-pocket spending—or reforms that would affect list prices more directly, such as through federal negotiation or inflation rebate penalties (Dusetzina and Oberlander 2019; Trish, Kaiser, and Joyce 2020).
The Part D redesign provisions included in the recently enacted Inflation Reduction Act are similar to those modeled here, albeit with slightly higher patient cost sharing and slightly lower plan liability in the initial coverage phase. Nonetheless, the general takeaways would hold under these alternative parameters, particularly the key insight that the changes in liability across payers created by these redesign proposals are driven primarily by liability for LIS beneficiaries and the distribution of drug spending across different phases in the benefit design.
Our study is limited in that we modeled changes to the standard benefit design, though in practice plans offer alternative coverage packages. That is, we first assumed that utilization would not change if beneficiaries had faced the current-law standard benefit design rather than their actual plan benefit design. Beneficiary utilization may in fact differ, though in practice these actual benefit designs are intended to be actuarially equivalent to the standard benefit design. However, utilization patterns of enrollees in enhanced plans may have been different under the standard benefit design, and we do not model those potential differences here.
Additionally, we assumed no behavioral changes under the proposed redesign, though it is possible that drug prices and/or beneficiary utilization could change under such reforms. That is, we also assumed that utilization again would not change under proposed standard benefit redesign parameters (or the actual benefit design that would result from those parameters). However, it is possible (indeed, perhaps likely) that these reforms would have utilization effects, which is an area ripe for future exploration.
Though the goal of the MMA was to create a market of competing private plans providing prescription drug coverage to Medicare beneficiaries, the standard benefit design that was included reflected significant political compromises. Subsequent policy changes helped to relieve out-of-pocket burden for beneficiaries, but doing so exacerbated differences in liability for prescription drug spending across beneficiaries who do and do not receive low-income subsidies. Together with the evolution of the prescription drug market, this resulted in a scenario where plans actually face quite limited liability for drug spending in the program. Reforming the standard benefit design to give plans more skin in the game could improve plans’ incentives to manage utilization and spending across the full spending distribution for all beneficiaries, but the impact on beneficiary access to drugs should be monitored.
This research was supported by the Leonard D. Schaeffer Center for Health Policy and Economics.
The federal reinsurance program protects plans against the risk of high-spending individuals by subsidizing 80% of spending among beneficiaries while in catastrophic coverage. The risk corridors program limits plans’ overall losses or profits by subsidizing a portion of higher-than-expected aggregate costs (or recovering a portion of excess profits from lower-than-expected aggregate costs).
Part D enrollees with low incomes (less than 150% of the federal poverty level [FPL]) and modest assets qualify for premium and cost-sharing assistance. Full low-income subsidies are available for Medicare beneficiaries who are dually eligible for Medicare and Medicaid, who receive Supplemental Security Income, or who qualify for a Medicare Savings Program as well as for nonduals with incomes below 135% FPL and modest assets. Individuals with incomes between 135%–150% FPL and/or higher assets may qualify for partial subsidies.
As described above, manufacturers are responsible for 70% of the costs of brand-name drugs taken by non-LIS beneficiaries while in the doughnut hole. Though these costs are not actually paid by the beneficiary, they are defined to count toward “true out-of-pocket costs,” a measure intended to capture the amount a beneficiary must pay before reaching catastrophic coverage. The concept of true out-of-pocket costs is meant exclude (uncommon) payments from supplemental coverage from counting toward this measure, and it predates the existence of these manufacturer discounts.