Drug Plan Design Incentives Among Medicare Prescription Drug Plans
Published Online: July 21, 2014
Haiden A. Huskamp, PhD; Nancy L. Keating, MD, MPH; Jesse B. Dalton, MA; Michael E. Chernew, PhD; and Joseph P. Newhouse, PhD
There are 2 types of Medicare Part D prescription drug plans. In 2010, approximately 17.7 million Medicare beneficiaries enrolled in a stand-alone prescription drug plan (PDP), continuing to receive Parts A and B benefits through the traditional Medicare program.1 Approximately 9.9 million beneficiaries enrolled in a Medicare Advantage-Prescription Drug plan (MA-PD), receiving Parts A and B benefits through the qualified Medicare Advantage (MA) plan that sponsored their MA-PD. Due to differences in the scope of benefits covered under PDPs versus MA/MA-PDs, these 2 types of plans face different incentives for drug plan design. Because of the integration of drug and nondrug coverage in MA plans and MA-PDs, MA-PDs have an incentive to provide relatively generous coverage for medications for which the incremental costs of doing so are offset by decreased medical spending over the period of expected enrollment.2 In contrast, PDPs, which are responsible only for drug expenditures, do not reap any efficiency gains with respect to Parts A and B services that may result from more generous drug coverage, and thus have lower incentives relative to MA-PDs to encourage use of drugs that may result in offsets for nondrug expenditures.
Both MA-PDs and PDPs have a choice of offering a plan with the defined standard benefit, a benefit that is actuarially equivalent to the defined standard benefit and consistent with CMS regulations governing coverage, or enhanced benefits. The defined standard benefit for 2013 has a deductible of $325. For spending between $326 and $2970, the plan pays 75% and the enrollee pays 25%. For expenditures between $2970 and $6733.75—known as the coverage gap—enrollees are responsible for a larger share, with enrollees paying 79% of generic drug expenditures and 47.5% of brand drug expenditures in the gap. For expenditures exceeding the $6733.75 catastrophic coverage limit, Medicare pays 80%, the plan pays 15%, and the enrollee pays 5%.
Previous studies have documented average aggregate (ie, aggregated across all drugs) differences in use of cost sharing and utilization management for PDPs relative to MA-PDs, with MAPDs typically offering more generous coverage.1,3 For example, in 2012, PDP enrollees faced some form of utilization management (ie, prior authorization [PA] requirements, step therapy requirements, or quantity limits) for 36% of drugs listed on the plan’s formulary while MA-PD enrollees faced utilization management for 31% of formulary drugs.1 Also, MA-PDs had a lower average premium, were more likely to have a zero deductible, and were more likely to provide additional benefits in the coverage gap than PDPs.1,4
Little is known about how MA-PDs and PDPs make decisions about coverage and utilization management requirements for specific medications, and data on aggregate differences in coverage levels for the 2 types of plans like the data described above do not shed much light on this complex decision-making process. To begin to explore how MA-PDs and PDPs may be responding to their different incentives around drug benefit design, we compared 2012 PDP and MA-PD average coverage, utilization management, and copayment requirements for specific drugs in several classes of medications used commonly among Medicare beneficiaries.
To identify MA-PDs and PDPs, we used the January 2012 CMS Prescription Drug Plan Formulary and Pharmacy Network Files (hereafter, “CMS formulary files”). These files contain information on plan characteristics as well as National Drug Code (NDC)-level information on formulary coverage, PA and step therapy requirements, and copayment requirements for all MA-PDs and PDPs operating in January 2012, with the exception of employersponsored MA-PDs. We excluded 125 private fee-for-service MA plans, as well as 506 special needs plans, 47 cost plans (ie, plans that Medicare contracts with a cost-reimbursement basis under Section 1876 of the Social Security Act), 61 plans serving US territories, 28 plans with no enrollment as of January 2012, and 1 plan for which formulary information was not available. After these exclusions, there were 1035 PDPs and 1512 MA-PDs in our sample.
Next, we linked the CMS formulary files for these plans with January 2012 plan enrollment information from the MA/Part D Contract and Enrollment Data available on CMS’ website.
To characterize the generosity of drug coverage in MA-PDs versus PDPs, we focused on 6 medication classes used frequently by Medicare beneficiaries for chronic medical conditions and for which at least one generic alternative was available in the class. Because drug plans generally cover all or almost all generic medications generously due to their relatively low cost, we focused analyses on brand name medications. Two of the 6 classes studied (antidepressants and antipsychotics) are Part D “protected” classes, meaning that plans must cover at least 1 formulation of every drug. The other 4 (Alzheimer’s disease medications, angiotensin receptor blockers [ARBs], bisphosphonates, and statins) are nonprotected classes, meaning that plans must cover at least two drugs in the class. Plans are permitted to use utilization management requirements like PA or step therapy to influence medication use in all classes, both protected and non-protected.
We examined 3 primary measures of coverage generosity: 1) formulary coverage; 2) use of PA and/or step therapy, conditional on formulary coverage; and 3) average copayment required for a 30-day prescription. To assess average formulary coverage, we first calculated coverage percentages for each formulation of each drug, collapsing over different NDCs for the same formulation. We expected plan coverage decisions for brand drugs to differ based on whether a generic equivalent of the brand drug was available. We also thought decisions about drugs might differ for brand combination medications (ie, medications that combine two or more molecular entities). Thus, we divided drugs into 3 categories for each class: brand drugs with a generic equivalent available, brand drugs with no generic equivalent available, and combination drugs (of the 19 combination products studied, all but 1 were brands without generic equivalents).
Within each drug class and category, we weighted the coverage percentages by plan enrollment and drug market share within the class, and then calculated the mean coverage percentage for each class and category for PDPs versus MA-PDs. We used 2011 Truven Health MarketScan data, which includes prescription drug claims for over 50 million individuals with employer-sponsored private insurance, to calculate drug market share. By weighting on the basis of drug market share, we “down-weight” observations for rarely-used drugs.
To test for a statistically-significant difference in average coverage between PDPs and MA-PDs, we calculated P values from t tests of the difference in the mean number of drugs covered for drug categories with more than 1 drug and P values from t tests of the difference in proportion covered for categories with a single drug, as well as an F-test of the null hypothesis that all effects are zero. We followed the same process for use of PA/ step therapy.
To calculate the average co-payment requirements for a 30-day prescription, we used plan-level information on the copayment required for each specific NDC from the CMS formulary files. For plans that required coinsurance rather than copayments for a given medication, we instead applied the plan’s coinsurance rate for the initial coverage period (ie, covering expenditures after the deductible is met and before the coverage gap is entered) for a specific NDC from the CMS formulary files to NDC level data on average total price (ie, average of plan plus patient payments per 30-day prescription, not accounting for any rebates negotiated between the plan and the pharmaceutical manufacturer) from the 2011 Truven Health MarketScan data. As for the coverage measures described above, we weighted observations by plan enrollment and drug market share. We calculated P values from t tests of the difference in weighted mean copayment requirements for each drug class and category, as well as an F-test of the null hypothesis that all effects are zero.
Finally, we calculated the percentage of MA-PD plans and the percentage of PDPs that covered all brand drugs with no generic equivalents in each class, weighting by plan enrollment but not drug market share, to ensure that we captured differences in covering all formulations, including those that are rarely used.
Brand name drugs with generic equivalents. Generics accounted for the vast majority of prescriptions filled for drugs with generic equivalents. The average generic market share for these medications ranged from 90.8% for Alzheimer’s medications to 99.7% for statins. On average, MA-PDs were significantly more likely than PDPs to cover brand drugs with generic equivalents in all 4 non-protected classes (Alzheimer’s disease drugs, ARBs, bisphosphonates, and statins) (Table 1).
On average, PDPs covered only 2.5% to 4.6% of brands with generic equivalents for statins, bisphosphonates, and ARBs, and covered 53.5% of brand Alzheimer’s drugs with generic equivalents. MA-PDs covered significantly more of these drugs on average in all classes (12.5% more for Alzheimer’s drugs to 25.6% more for bisphosphonates, all P <.0001). For the 2 protected classes, coverage of antipsychotic brands with generics tended to be more generous for MA-PDs vs PDPs, although this did not reach statistical significance (P = .052). There were no statistically significant differences between PDPs and MA-PDs for antidepressants.
PDF is available on the last page.