Surgeon Perception of Risk and Benefit in the Decision to Operate Under Uncertainty

For decades, researchers have noticed widespread variation in the use of surgical procedures across the United States. Yet despite extensive study, the underlying reasons for this variation remains poorly understood since it seems unrelated to traditional measures of supply and demand, such as the prevalence of disease, the number of available physicians, or measurable differences in patient preference. Instead, some researchers have attributed the persistently unexplained variation to differences in physician judgment on the role of surgery. However, the precise composition of this judgment, how it might differ from one physician to the next, and how such judgment translates into clinical decision-making remain unknown. In this dissertation, I explore the variation in surgeons' decisions to operate by examining how they perceive the risks and benefits of two opposing treatment options: to operate or not to operate. Using data collected from a national study where surgeons provide their opinions on a series of detailed clinical vignettes, I document wide variation both in how surgeons perceive these risks and benefits and in their associated clinical decisions. I then go on to demonstrate that the variation in perception is closely aligned with surgeons' clinical decisions, thus revealing the substantial consequences of differences in perception. In fact, accounting only for differences in surgeons' perceptions of the risks and benefits of operating and not operating, I was able to explain a total of 36% of the variation in surgeons' clinical decisions. The dissertation then examines whether surgeons' perception of risk and benefit are associated with their experience or general attitude towards risk. Finally, I examine the results of an experiment where surgeons were, by random assignment, provided data from a surgical risk calculator. Although exposure to risk calculator data strongly influenced surgeons' reported perception of risk, it did not alter their reported likelihood of recommending an operation. One possible reason for this latter finding is that the risk calculator also appeared to influence surgeons' judgments of non-operative risks and benefits, despite the fact that the calculator presented no data on these outcomes. Presented together with a detailed conceptual model, the results of this dissertation offer a useful framework for the future study of surgical decision-making and to promote the idea that clinical decisions can be both data-driven and patient-centered.