Essays on Physician Behavior

Abstract

This dissertation comprises three chapters that study the behavioral economics of physician decision-making. All three chapters examine the common but high-risk clinical scenario of assessing patients in the emergency department (ED) with shortness of breath (SOB) for the risk of pulmonary embolism (PE) in the Veterans Affairs (VA). Chapter 1 studies the availability heuristic, under which the assessment of an event’s probability is influenced by the ease with which such events can be recalled, such as due to the recency of the event. I perform an event study that examines whether a recent diagnosis of a PE was associated with changes in rates of PE testing for subsequent patients, controlling for other PE risk factors like high heart rate. Compared to the 10 days prior to a PE diagnosis, the rate of PE testing increases by approximately 10 percent in the 10 days after. I fail to find statistically significant changes in composition of the patients after PE diagnosis or in the use of two tests unrelated to PE. I also fail to find significant differential changes in testing after PE diagnosis between older and younger physicians, arguing against a Bayesian updating interpretation of my results. Chapter 2 studies how physicians make decisions when communication with patients may be difficult, such as with patients with dementia. I first examine whether physicians were more likely to test patients for PE when they have known PE risk factors. I then examine whether physicians take these factors less into account when evaluating patients with dementia. I find that while each clinical factor examined was associated with an increased probability of testing, physicians took three of the four factors examined less into account for patients with dementia, even though these clinical factors had inconsistent differential associations with subsequent PE for these patients. Chapter 3 studies the anchoring bias, or the focus on one—often an initial—piece of information. I examine, for patients with congestive heart failure (CHF), how the mention of this diagnosis in the patient record field “Patient Visit Reason,” which is entered by a nurse or administrator prior to the physician first interacting with the patient, influences PE testing. I find that, conditional on CHF severity and PE risk factors, that the mention of CHF in the “Patient Visit Reason” is associated with over a third less testing for PE, a longer time to PE testing for those tested, and increased use of blood tests for exacerbations of CHF.