Job market interview
Time: 12:30pm – 13:45 pm
Date: 14th of January 2020
Place: Room 3001
Pieter de VLIEGER (University of Michigan) “Quantifying Sources of Persistent Prescription Behavior : Evidence from Belgium”
Abstract : Physician behavior is highly persistent and difficult to change, yet the drivers of this persistence are not well understood. This paper quantifies the relative importance of physician and patient factors in this persistence, by exploiting a policy mandate introduced in Belgium in 2006 that required primary care physicians (PCPs) to prescribe a minimum percentage of cheap or generic drugs. Using an event study design and administrative data linking 26 million dispensed prescription drugs to 152,000 patients as prescribed by their PCPs, I analyze the change in PCPs’ treatment decisions. I find that PCPs increase the prescription rate of generics without changing the quality of dispensed drugs. This suggests that, unless required otherwise, PCPs exhibit a bias towards prescribing brand name drugs which is evidence of physician factors in explaining persistence. Patient factors, however, also matter. PCPs are much more likely to prescribe a generic to a patient who is prescribed a drug for the first time than to switch a patient from a branded drug, especially when these patients are older and take multiple prescription drugs. This suggests it is costly to switch these patients. Using an instrumental variable approach, I show that changing these patients’ prescriptions indeed comes at the cost of decreased medication adherence, and that physicians take this patient behavior into account. I develop and estimate a structural model to quantify the relative importance of both factors and analyze the impact of policy counterfactuals. I find that patient and physician factors are about equally important. I show that requiring PCPs to always prescribe (or pharmacists to only dispense) generics may decrease overall welfare – despite savings in prescription drug expenditures – unless patient considerations are decreased by 60%. These results highlight the importance of patient factors and behavior when introducing policies that target treatment decisions of PCPs.