| Abstract: |
Spending on prescription drugs (Rx) represents one of the fastest growing
components of U.S. healthcare spending, and has coincided with an expansion of
pharmaceutical promotional spending. Most (83%) of Rx promotion is directed at
physicians in the form of visits by pharmaceutical representatives (known as
detailing) and drug samples provided to physicians’ offices. Such promotion
has come under increased public scrutiny, with critics contending that
physician-directed promotion may play a role in raising healthcare costs and
may unduly affect physicians’ prescribing habits towards more expensive, and
possibly less cost-effective, drugs. In this study, we bring longitudinal
evidence to bear upon the question of how detailing impacts physicians’
prescribing behaviors. Specifically, we examine prescriptions and promotion
for a particular drug class based on a nationally-representative sample of
150,000 physicians spanning 24 months. The use of longitudinal physician-level
data allows us to tackle some of the empirical concerns in the extant
literature, virtually all of which has relied on aggregate national data. We
estimate fixed-effects specifications that bypass stable unobserved
physician-specific heterogeneity and address potential targeting bias. In
addition, we also assess differential effects at both the extensive and
intensive margins of prescribing behaviors, and differential effects across
physician- and market-level characteristics, questions which have not been
explored in prior work. The estimates suggest that detailing has a significant
and positive effect on the number of new scripts written for the detailed
drug, with an elasticity magnitude of 0.06. This effect is substantially
smaller than those in the literature based on aggregate information,
suggesting that most of the observed relationship between physician-directed
promotion and drug sales is driven by selection bias. Qualitatively consistent
with the literature, we find that detailing impacts selective brand-specific
demand but does not have any substantial effects on class-level demand.
Results also indicate that most of the detailing response may operate at the
extensive margin; detailing affects the probability of prescribing the drug
more than it affects the number of prescriptions conditional on any
prescribing. We draw some implications from these estimates with respect to
effects on healthcare costs and public health. |