Prior Authorization Reduces Net Costs of Medicare Part D

Administrative costs are estimated to make up between 20 and 34 percent of US health care expenditures, roughly 1–4 percent of GDP. Academic and policy discussions generally characterize these costs as pure waste and a major explanation for why the US spends more on health care than other OECD countries. However, these discussions tend to overlook the fact that half of administrative effort is spent on limiting utilization and spending. Thus, eliminating administrative expenses may increase costs in some cases.

In new research, Zarek Brot-Goldberg, Samantha Burn, Timothy Layton, and I use data from Medicare Part D to examine this trade-off by studying the most common form of administrative hassle associated with prescription drugs: prior authorization restrictions. Under prior authorization, an insured patient is only granted coverage for a drug if their insurer approves their provider’s request. In Medicare Part D, in 2015, roughly 4 percent of prescription fills required prior authorization, but these fills made up 20 percent of net drug spending.

Here are the key takeaways from our new working paper, “Rationing Medicine Through Bureaucracy: Authorization Restrictions in Medicare”:

• Prior authorization in Medicare Part D reduces the use of a prior authorized drug by 25 percent and overall Part D spending by 3 percent.
• Savings from prior authorization vastly exceed the overhead costs of administering it, by about a factor of 10.
• Efforts to lower administrative spending should consider trade-offs with other costs.

Using randomization to quantify the effects of prior authorization: Despite the prevalence of prior authorization restrictions, few credible estimates exist on how much these restrictions deter care, reduce spending, or impact health. This is largely because prior authorization is not randomly used, making it hard to isolate its effects from other factors. We overcome this hurdle by focusing on a setting where prior authorization is effectively randomized: low-income subsidy (LIS) beneficiaries in Medicare Part D, who are typically randomly assigned to drug plans. This means that enrollees are randomly exposed to prior authorization for a drug because each plan imposes different prior authorization requirements across different sets of drugs.

For each drug, we compare LIS beneficiaries randomly assigned to a plan that put a prior authorization restriction on their drug to beneficiaries randomly assigned to a plan that did not.

Prior authorization affects utilization: We show that prior authorization moves the needle substantially in terms of deterring care: beneficiaries who randomly faced prior authorization restrictions for a drug were 26.8 percent less likely to ever fill a prescription for that drug in the following year. Effects were larger for non-white and older beneficiaries. They were somewhat smaller for beneficiaries who had previously used the drug under a different plan, as well as for drugs in high-value therapeutic categories such as antidepressants and antidiabetic agents.

Cost savings from prior authorizing a drug are partially offset by shifting to cheaper substitutes: A key goal of prior authorization is to shift beneficiaries away from high-cost drugs to lower-cost clinical substitutes. Any substitution to unrestricted drugs that occurs in practice will at least partially offset savings from the reduction in use of restricted drugs. Among patients that shifted away from a drug due to prior authorization, an estimated 46.2 percent of them instead take a therapeutic substitute, with the remaining 53.8 percent taking no drug at all. The total reduction in Part D drug spending due to prior authorization is $95.88 per beneficiary-year, made up of a $111.57 reduction in spending on restricted drugs, and a $15.69 increase in spending on (often much cheaper) substitute drugs.

Prior authorization is likely worth the cost: Is this spending reduction worth the administrative costs associated with prior authorization? To answer this question, we compare spending reductions due to prior authorization with the cost of prior authorization requests (ranging from $11–$33) and request rejection rates (ranging from 0–15 percent). Results show that the administrative costs of the status quo use of prior authorization are much smaller than their spending reductions (specifically, about 6–18 percent the size).

A natural question that arises is how much prior authorization, and the accompanying shift in drug utilization, harms patients. Unfortunately, despite looking at hundreds of thousands of LIS beneficiaries, we lack the statistical power to precisely estimate how prior authorization affects patient health outcomes like mortality. However, we can quantify how much LIS beneficiaries would have otherwise been willing to pay for these foregone prescriptions out-of-pocket, as a proxy for how much patients value those prescriptions. Our model suggests that patients may value the forgone drugs less than the cost savings to plans, but this result is sensitive to various assumptions about patient and doctor behavior.

Our evidence clearly shows that prior authorization has an important effect on drug spending—a fact that should be central to policy discussions regarding drug markets and health care markets more broadly.