Conflict of Interest or Valuable Information? The Impact of Pharmaceutical Industry Payments on Physician Prescribing Behavior

Background: Do we really need to be concerned about all types of financial relationships between pharmaceutical companies and physicians? While we know that many of these financial relationships are correlated with “nonrational prescribing behavior,” this correlation is not identified: physicians often assert that they accept payments from the companies whose drugs they prefer and prescribe anyway, and thus the correlation is not indicative of influence (Steinman, Shlipak, McPhee, 2001; “Doctors Often Receive Payments from Drug Companies.” Talk of the Nation, NPR. 9/13/2011). Particularly in the case of speaking fees, physicians who already prefer the company’s drug will naturally make better spokespeople, even in the absence of influence from a financial incentive. Thus, physicians may simply be paid for expressing their previously held preferences, and these payments need not necessarily have any persuasive effect.

Contributions: The presence or absence of a persuasive effect from payments has been difficult to assess empirically, as pharmaceutical industry payments to physicians typically coincide with changes in their product offerings—companies are more likely to pay physicians to promote new drugs, and less likely to make such payments after the introduction of generics. Thus, it has been impossible to estimate the effect of payments separately from any informational effect due to the change in product offerings. This study fills this gap by making use of a natural experiment to rule out any informational effects, in conjunction with detailed prescription data to understand the impact of financial incentives on prescribing behavior.

Data: I employ New Jersey Department of Health billing data for all inpatient and outpatient hospital discharges from 2009-2010. This dataset identifies charges for branded or generic drugs administered to the patient before discharge. Physician payment data is compiled by ProPublica.

Methodology: I employ a differences-in-differences approach using a natural experiment: GlaxoSmithKline stated their goal to ultimately reduce their financial contracts with physicians by 50%, using fewer speakers more often; between 2009 and 2010 this led to a 15% decline in physician payments, unrelated to any change in their product offerings. This enables me to compare the prescribing behavior of physicians who have never received payments; who currently receive payments; and who have their payments cut off.

Findings: I find that physicians who have their payments reduced are not significantly different from physicians who received no payments at all. In contrast, physicians who had their payments cut off increase their prescriptions—both branded and generic—consistent with evidence that physicians with more experience have more concentrated prescribing behavior. Introducing a shock to their preferences by removing the persuasive payment induces them to experiment with and thereby learn about alternative treatments for the same types of diagnoses.

Conclusions: This analysis unambiguously identifies the presence of a strictly persuasive effect of receiving payments from pharmaceutical companies on physicians’ choice of branded versus generic drugs. This effect on prescribing is neither informational (i.e., a change in physician prescribing due to new information) nor a reflection of baseline physician preferences.