Team Formation and Performance: Evidence from Healthcare Referral Networks
Leila Agha (Dartmouth College)
Keith Ericson (Boston University)
Kimberley Geissler (University of Massachusetts)
James Rebitzer (Boston University)

Abstract : How does team structure affect productivity? We address this question with an application to healthcare by examining the teams that primary care physicians (PCPs) assemble when they refer patients to specialists. Our theoretical model analyzes how PCPs trade off costly coordination against beneficial specialization, predicting that coordination improves when PCPs concentrate their referrals within a smaller set of specialists. Empirically we find that patients of PCPs with concentrated referrals have lower healthcare costs. This effect exists for commercially insured and Medicare populations; is statistically and economically significant; and holds under identification strategies that account for unobserved patient and physician characteristics.

Learning on the Job: Evidence from Physicians in Training
David Chan (Stanford University)

Abstract : Learning on the job presents a tradeoff in making team decisions: Workers with less knowledge have less to contribute to team decisions, yet in order to learn, they may need to have an experiential stake in decision-making. This paper studies learning and influence in team decision-making among physician trainees. Exploiting a discontinuity in relative experience, I find reduced-form evidence of influence due to seniority between trainees. I specify a simple structural model of Bayesian information aggregation and define a benchmark of static efficiency that allocates influence to make the best decision using knowledge at hand. The vast majority of learning occurs only after trainees are senior and can influence decisions. Influence is approximately efficient between trainees, but trainees exert much more influence than is statically efficient relative to their supervisors, possibly because such influence contributes to experiential learning.

Small Area Variations and Physician Decision Making: the Case of Depression
Janet Currie (Princeton)
Bentley MacLeod (Columbia)

Abstract : A large literature on "small area variations" in medical care documents large differences in practice style across areas. It is generally assumed that these variations indicate inefficient provision of care. Using the case of treatment for depression, we show that there is as much variation within areas as across areas. We develop a model of physician behavior which demonstrates that doctor's propensity to experiment varies with individual characteristics of physicians, and that some amount of experimentation (and therefore error) is likely to be necessary in order to match the optimal treatment to each patient.

Physician Practice Style and Healthcare Costs: Evidence from Emergency Departments
Gautam Gowrisankaran (University of Arizona)
Keith Joiner (University of Arizona)
Pierre-Thomas L├ęger (University of Illinois, Chicago)

Abstract : We examine the impact of emergency department (ED) treatment on future healthcare costs and outcomes. We postulate that ED physicians may affect patient outcomes through acumen in diagnosis and appropriate disposition conditional on diagnosis. We make use of a unique dataset and quasi-experiment on patients who seek treatment at an ED in Montreal, Canada. Physicians there rotate across shifts between simple cases and difficult cases, implying that the assignment of patients will be quasi-random across physicians in an ED. We examine how the initial assignment of ED physician affects the use of resources (what we call physician practice style) during the initial ED visit and the patient outcomes (what we call physician skills). We consider three serious conditions, that present frequently in the ED, angina, appendicitis, and transient ischemic attacks. We find that physician practice style and outcomes (from ED revisits to hospitalizations) vary across ED physicians practicing in the same ED. We also find a strong correlation of physician practice style across the three illness categories. Similarly, we find that physician skills correlate positively across outcomes and (weakly) across illness categories. Our results also suggest that physicians associated with costly practice styles are not associated with better outcomes. Finally, our results suggest that differences in outcomes when treating patients who may be suffering from angina or TIA are driven by diagnostic ability.