Aging, Productivity, and Spillovers: Evidence from emergency department physicians

Hospital care and physician services account for about half of all healthcare spending in the United States, a share that has been rising (CMS, 2012). A key policy question is how to sustainably manage these costs and potentially "bend the cost curve". While there are many policy levers that could be used to reduce cost growth of physician services, little is understood about the fundamentals of this growth. In this project, I study the determinants of a key input of the costs of physician services -- physician productivity. I focus on three salient factors in the productivity of the physician workforce: aging, worker complementarities, and technology adoption.

Using hospital discharge records from New York from 2004 to 2012, I document substantial dispersion in the productivity of ED physicians. The data contain detailed patient- and case-level information which allow me to overcome challenging selection problems in estimating physician productivity, quality of care, and treatment choices. Preliminary results suggest that physicians become less productive later in their careers along a variety of outcomes. The aging of the ED physician workforce over this period can explain a sizable share of cost growth in terms of lost productivity. There are many explanations for these findings: aging physicians may decrease effort or time at work, their skills may depreciate, they may fail to adopt newer more effective technologies, or they may suffer from physical and cognitive decline. My ongoing work differentiates among these explanations using detail on physicians' time at work, procedural orders, and efficiency in treating patients of different severity levels.

Additionally, I explore market mechanisms for mitigating the negative productivity effects of aging. In particular, I focus on human capital spillovers between younger and more experienced physicians. Physician groups may recruit new hires whose training and skills complement or spill over on to existing physicians. Using detail on date and time of each case, I implement a research strategy based on physician mobility across hospitals and within hospitals across shifts to estimate productivity spillovers and training complementarities. I also estimate the contemporaneous influence of different types of coworkers (e.g. characterized by medical school ranking, experience, procedural profile and estimated productivity) on a physician's own orders and outcomes. My preliminary findings show evidence of sizable peer effects in productivity: working with a 10% less productive ED physician on a shift decreases own productivity by 2%. These results have implications for optimal shift-level and hospital-level mixtures of physicians.