Dynamic Learning
In a dynamic setting, physicians are forward-looking in that they consider the value of “experimentation”. In other words, physicians might try something new in hopes of learning more about the quality of the new treatment, understanding that this knowledge will affect all patients moving forward. We’ll discuss the following topics and papers in this class, all of which allow for experimentation in learning.
Learning with Experimentation
We’ll discuss Crawford and Shum (2005) in this topic. That paper introduces a dynamic matching model where patients learn from prescription experiences about drug effectiveness. A notable aspect of this model is its differentiation of drugs’ symptomatic and curative effects and the endogenization of treatment duration based on drug choices and patients’ recovery probabilities. The study reveals that while drug efficacy varies significantly across patients, learning effectively reduces the costs of uncertainty in pharmaceutical markets.
Learning with Physician Skill
Physicians may be limited in the extent of learning based on their underlying skill level, as studied in (currie2020?). This paper introduces a model that highlights the trade-off between exploiting known treatments with the highest expected value and experimenting with newer treatments that may offer higher potential benefits. The model suggests that a doctor’s diagnostic skill influences this trade-off, with higher skill leading to a greater variety of drug choices and better matching of drugs to individual patients. The study utilizes patient claims data and finds that, generally, a broader drug portfolio used by skilled doctors predicts better patient outcomes, except in cases where the doctor’s decisions contravene established professional guidelines.
Learning with External Factors
Learning might also be impeded or enhanced by other factors, such as insurers. Dickstein (2018) considers this in his paper, where he estimates a dynamic model to examine how physicians, under varying insurance plan structures, learn and update their treatment strategies based on patient feedback. This model uniquely accounts for the role of insurance in influencing both patient access to drugs and the feedback loop essential for physicians’ learning. The paper highlights the interplay between insurance plan design and the efficient provision of experience goods in healthcare, demonstrating how insurance parameters can shape the learning and decision-making process.