We use clinical data on over 240,000 surgeries and quasi-experimental methods to examine how physicians respond to the surprise release of a performance "report card". Such feedback interventions are commonly used to encourage physicians to improve performance, yet show limited evidence of success. Our results show that these limited effects mask heterogeneous behavioral responses to feedback valence. In particular, physicians improve more from positive feedback than from negative feedback, with negative feedback even reducing performance for a non-trivial share of patients. Experiments with laypersons replicate these results and show that struggles with negative feedback can be mitigated by giving incentives directly tied to improvement and by adding qualitative information that helps individuals interpret past performance. These results are consistent with behavioral models that suggest cognitive and emotional difficulties limit how well individuals use negative feedback. Thus, feedback interventions in healthcare should be carefully designed to mitigate these counterproductive behavioral responses.

Power – the asymmetric control of valued resources — affects most human interactions. Though power is challenging to study using real-world data, a unique dataset lets us do so within the critical context of doctor-patient relationships. Using 1.5 million quasi-random assignments in US military emergency departments, we examine how power differentials between doctor and patient (measured using differences in military ranks) affect physician behavior. We find that power confers nontrivial advantages: “high-power” patients (who outrank their physician) receive more resources and have better outcomes than equivalently-ranked “low-power” patients. Patient promotions even increase physician effort. Furthermore, low-power patients suffer if their physician concurrently cares for a high-power patient. Doctor-patient concordance on race and sex also matters. Overall, power-driven variation in behavior can harm the most vulnerable in healthcare.

Clinical decisions made in the delivery setting are often made under high pressure, great uncertainty, and have serious consequences for mother and baby. Theories of decision-making suggest that individuals in such settings may resort to using “heuristics”, or simplified decision-rules, to aid complex decision-making. This study investigates whether physicians’ delivery-mode decisions (i.e., when to perform a vaginal vs. a cesarean) are influenced by such a heuristic. Electronic health record data spanning 86,000 deliveries suggests that, if the prior patient had complications in one delivery-mode, the physician will be more likely to switch to the other -- and likely inappropriate -- delivery-mode on the subsequent patient, regardless of patient indication. There is evidence that this heuristic has small, suboptimal effects on patient health.

We hypothesize that deepening resource scarcity results in rationing on the basis of group identity in settings with underlying discrimination. We provide evidence of such race-based rationing in a high-stakes setting: health care. Using detailed, time-stamped data on 107,000 patient admissions to a large health system, we find that in-hospital mortality increases for Black, but not White, patients as hospitals reach capacity (a state of resource scarcity likely to trigger or exacerbate biases in decision-making). As a mechanism, we identify rationing by wait time, documenting that sick Black patients wait longer for care than healthy White patients at every capacity level, likely because of systematic misevaluation of medical need. Text analysis of unstructured provider notes reveals differential rationing of provider effort by race as another potential mechanism. Together, these findings demonstrate important linkages between three key economic concepts: scarcity, discrimination, and rationing.