Introduction
Pharmaceuticals are a critical component of modern medical care and have contributed
substantially to health and quality of life. They are also expensive, accounting for
≈10% of US health care costs.1 By the end of this year, Medicare beneficiaries are
expected to spend $343 billion on medications, and drug spending is projected to grow
faster than both the inpatient and outpatient care categories.2 Commercial and public
payers use a range of mechanisms to constrain these costs, including formularies,
tiered copayments, and preauthorization. These coverage policies may have potent impacts
on medication usage, and their effects deserve close scrutiny. In the current issue
of the journal, Li et al examine the impact of the Medicare prescription drug program
on the use of statin medications.3 The researchers exploit the unusual design of the
Medicare program to assess how changing copayments impact both therapeutic substitution
and medication discontinuation.
At its inception, the Medicare program was designed to provide catastrophic coverage
to hospitalized elderly Americans. As beneficiary survival increased over time, the
emphasis shifted toward management of chronic disease.4 Yet, for its first 38 years,
the program included no prescription drug benefit. The Medicare Modernization Act
of 2003 established the Medicare prescription drug program (known as Medicare Part
D), and coverage began in 2006.5 The drug program added a major new entitlement for
Medicare beneficiaries, the cost of which was shared between beneficiaries and taxpayers.
Even so, the plan struggled to balance the desire for comprehensive coverage with
the need for cost containment.6 As a compromise, comprehensive coverage was established
for indigent patients, while most beneficiaries faced significant coverage gaps. After
meeting the $250 deductible, participating nonindigent patients were responsible for
25% of drug costs up to $2250, but then were responsible for all costs until they
reached the “catastrophic” limit of $5100 (Table 1). This coverage gap between $2250
and $5100 became colloquially known as the “donut hole.” Although many seniors benefited
financially from the program, its unusual design imposed predictable financial shocks
for patients with multiple chronic conditions.
Table 1
Medicare Part D 2006 Program Design for Patients With Incomes Greater Than 135% of
the Federal Poverty Line
Drug Expenditure ($)
Patient (%)
Medicare (%)
0 to 250
100
0
251 to 2250
25
75
2251 to 5100
100
0
Over 5100
5
95
Using a 5% random sample of administrative claims drawn from the Medicare Chronic
Condition Warehouse in 2006, the researchers identify patients taking branded Lipitor
(atorvastatin) and Crestor (rosuvastatin) at the beginning of 2006 and assess whether
they converted to a generic alternative or discontinued treatment. In a difference‐in‐differences
model, their analysis compares patients who were subject to generic‐only gap coverage
(study group) to propensity‐matched, low‐income subsidy patients (LIS; control group).
They find that study patients decreased branded statin use by 12%, and that these
declines were only partially offset by new generic statin prescriptions. Relative
to controls, the coverage gap was associated with reductions in mean monthly 30‐day
fills of any statin (−0.18; CI, −0.23, 0.13) and any lipid‐lowering drug (−0.17; CI,
−0.22, −0.12). The researchers conclude that increased copayments caused some patients
to switch from branded to generic statins, whereas others discontinued them altogether.
The study has a number of limitations. First, to be included in the analysis, all
patients spent more than $2250 on medications in 2006, suggesting that they were sicker
than excluded patients. Patients may respond differently to copayments depending on
their health status. Second, the researchers appropriately used difference‐in‐differences
methods for the study, but they use propensity scores to match patients who were subject
to the coverage gap with LIS patients who were not. Although they incorporate area‐level
median household income as a matching criteria, by definition, LIS patients (less
than $12 123 for individuals and $16 362 for a couple) have incomes well below their
area median household income ($30 387).6 The study and control groups are highly likely
to respond differently to copayment changes.7 As a sensitivity analysis, the researchers
also compare the study group to non‐LIS patients who were not subject to the coverage
gap and find similar results. However, non‐LIS (wealthier) patients who were not subject
to the coverage gap voluntarily enrolled in the most costly plans and may have also
responded differently to copayment changes. Finally, patients subject to the coverage
gap may have elected to purchase branded statins internationally, but were unable
by statute to use Medicare prescription drug coverage to pay for them. Administrative
claims only document prescriptions for which a bill was paid, and the impact of international
purchasing is unknown.
This study adds to a body of literature that finds that patients respond to financial
incentives.1 Prescription drug insurance coverage generally increases use of prescription
drugs and decreases cost‐related nonadherence.8, 9 Conversely, cost‐sharing policies
may decrease pharmaceutical expenditures, but may have unintended consequences on
the use of other health services and on care outcomes.10 It is precisely the unintended
consequence of a payment policy decreasing adherence of a proven beneficial treatment
that Li et al have studied with statin prescriptions. The benefits of statins in preventing
cardiovascular events are convincingly demonstrated in multiple, large, randomized,
controlled trials11 and supported epidemiologically by observational decreases in
cardiovascular events concordant with decreases in population cholesterol.12 Moreover,
the abrupt discontinuation of statins after myocardial infarction may have detrimental
effects beyond simply the absence of statins.13 Li et al have linked a policy feature,
namely, the increase in copayments for branded statins, to an overall decrease in
statin prescriptions even when a generic alternative was available.
Although Li et al effectively use a natural experiment to show that payment policy
can adversely impact evidence‐based medication adherence, a number of subsequent developments
have decreased the impact of the Medicare Part D design. First, coverage revisions
have substantially decreased financial shocks within the donut hole (see Table 2).14
Second, drug cost transparency is much greater today for both patients and providers.
Ambulatory electronic health record adoption has increased from around 25% in 2006
to greater than 75% in 2014,15 and many systems display insurance copayment tier levels
at the point of prescribing. Patients and their providers are now better able to make
informed and shared decisions in the exam room about medications that fit the patient's
medical and financial needs.
Table 2
Medicare Part D 2017 Program Design for Patients With Incomes Greater Than 135% of
the Federal Poverty Line14
Drug Expenditure ($)
Patient (%)
Medicare (%)
Discount (%)
0 to 400
100
0
0
401 to 3700
25
75
0
3701 to 4950
40
10
50
Over 5100
5
95
0
Drugs make an increasingly important contribution to health care outcomes, and drug
coverage policies aim to contain costs while improving value. Ideally, coverage decisions
should discourage low‐value drug prescribing, encourage selection of low‐cost drug
options where they exist, and encourage evidence‐based prescribing. However, policies
are written to apply to all patients, and individual patient circumstances may vary.
Although conversion of branded to generic formulations of the same drug makes sense
in nearly all circumstances, the therapeutic substitution to another drug in the same
class, as promoted by tiered copayments, is more problematic. In particular, statin
drugs vary substantially in potency, a factor explicitly recognized in current treatment
guidelines.16 They are also metabolized differently, have different half‐lives, and
vary in chemical characteristics, such as lipophilia. These drug reimbursement policies
may increase drug‐drug interactions and prompt safety concerns.17 Statin side effects
are also unfortunately common and may be idiosyncratically specific to individual
drugs.18
Although tiered pharmacy pricing coverage is commonplace, the design of the original
Part D program placed the greatest financial burden on the sickest patients, and in
the face of sharp copayment increases, some patients choose to forgo statins altogether.
Notably, a recent economic simulation based on the MI‐FREE study suggests that the
opposite approach may actually be cost saving.19 The study found that following a
myocardial infarction, provision of free evidence‐based medications, including statins,
would increase survival by 0.14 quality‐adjusted life years and decrease overall per‐patient
costs by $4011. They project that the policy would save $2 billion annually if nationally
implemented.
Given the very legitimate variations in the needs of individual patients and the fact
that many patients will ultimately defer to their clinicians' judgment, policies directed
to the prescriber may be more appropriate. Prescribers are better positioned to make
informed trade‐offs between costs and outcomes than many patients are. Immediate and
effortless availability of patient‐specific drug formulary options and pricing is
an important step toward achieving greater value. Efforts to hold prescribers accountable
for generic prescribing rates, overall pharmacy costs, and patient outcomes at the
population level also have merit. Population‐level accountability avoids the administrative
inefficiencies of preauthorization programs and allows prescribing flexibility for
individual patients. This approach is increasingly prominent in emerging payment reforms,
including the landmark Medicare Access and Children's Health Insurance Program (CHIP)
Reauthorization Act legislation.20
The growth of US health care spending is unsustainable, and effective solutions are
needed to improve value. Drug spending is increasingly contributing to the cost of
care, and policies are needed to manage pharmacy utilization. Whereas large patient
pharmacy copayments almost certainly impact drug utilization, they may be counterproductive
if they decrease use of evidence‐based therapies. Policy makers should pay careful
attention to program designs to avoid unintended consequences like those demonstrated
in the study by Li et al.
Disclosures
None.