UC Berkeley

Consumers do not purchase prescription drugs in a standard marketplace setting; instead, they rely on physicians to select an appropriate drug on their behalf. This potential agency problem is amplified by the fact that different consumers pay different prices for the same drug, depending on the copayments required by their insurance plan. There is a prevalent public concern that physicians are overly influenced by pharmaceutical company promotion, but little is actually known about how they choose which drugs to prescribe.

This dissertation investigates the extent to which agency and information problems affect prescribing, and consequently, patient outcomes. I use individual-level data on prescription drug purchases by employees and retirees in twenty-nine Fortune 500 firms from 2003-2007 to construct a sample of patients receiving first-time prescriptions for chronic drugs. In

the first two chapters, I estimate how initial prescriptions respond to three factors of patient utility: the copays set by individual health plans, large-scale copay shocks induced by patent expirations, and the predicted price-sensitivity of an individual patient. In the third chapter, a smaller sample with physician identifiers is used to measure the range of physician prescribing (number of drugs used) within a class, and its impact on patient outcomes.

In Chapter 1, I study the responses of physicians and patients to variation in the cost of drugs, and assess the welfare and health consequences of asymmetric and imperfect information in the prescription drug market. I focus on statins (cholesterol-lowering drugs) which are currently the most prescribed category of prescription drugs in the United States. Demand for drugs that treat chronic conditions depends on the initial prescriptions written by a physician, and on the subsequent decisions of patients to continue the prescription or stop. I show that the continuation decision is relatively sensitive to co-payment prices. Initial prescriptions, by comparison, are relatively insensitive to co-payment prices, suggesting that physicians either don't know the prices their patients are paying, or fail to take prices into consideration. I use the event of the highly publicized expiration of the patent for Zocor (simvastatin) to test between these explanations. Insurance plans have much lower co-pays for off-patent drugs: my analysis suggests that physicians are aware of this fact, and substantially increased prescriptions for Zocor and its generic equivalents following the patent expiration. Interestingly, the increases were larger for lower-income and healthier patients, suggesting that physicians correctly perceive the adherence elasticity of their patients and adjust their initial prescriptions accordingly, but only in response to a large and universal price change.

In Chapter 2, I study the prescribing responses to ten patent expirations occurring between 2004 and 2007 in four drug classes: antidepressants, statins, calcium channel blockers, and beta blockers. Four of the patent-losing drugs (including Zocor) experienced significant increases in prescribing rates, while three experienced statistically significant decreases. Understanding what drives this variation can inform how pharmaceutical advertising, health plans, and patient costs affect physician decisions. I identify two factors that explain much of the variation in these responses: the size of the copay drop upon expiration (i.e. the difference in copays of the brand and generic versions of the drug), and the current prevalence of generic prescribing in the drug class. Results suggest that physicians are more likely to increase their prescribing of a drug, after it becomes available as a generic, if it previously had a higher copay, on average. However, there is a baseline tendency to reduce prescribing of a patent-losing drug, likely driven by the cessation of its advertising, and this tendency grows stronger with the existing rate of generic prescribing in a class.

In Chapter 3, which is coauthored with Geoffrey Joyce and Neeraj Sood, we measure the range of physician prescribing within the ten most prevalent therapeutic classes, the factors affecting the broadness of this range, and its impact on patient outcomes. Physicians prescribe more broadly than commonly perceived. In 8 of 10 classes, the median physician prescribes at least 3 different drugs despite the small number of initial prescriptions observed per doctor (median=7). Physicians treating patients with a greater range of comorbid conditions and varied formulary designs prescribe a broader range of drugs within a class. Though narrow prescribers are more likely to prescribe highly advertised drugs, few physicians prescribe these drugs exclusively. Narrow prescribing has modest effects on medication adherence and out of pocket costs in some drug classes.