Decomposing Volume’s Impact on Performance: Lessons from Kidney Transplantation

The correlation between accumulated experience and greater productivity has been widely documented in a range of fields, from airframe production to medical care, and substantial effort has been devoted to unpacking the mechanism by which these improvements occur. In markets for credence goods (such as medical care), suppliers act as not only as providers but also as expert advisors. They diagnose a buyer’s need for that service relative to potential alternatives before providing the service in question. This dual role of the supplier introduces a complication in learning-by-doing theory of the volume-outcome relationship: the observed improvements may result from better decision-making (i.e. diagnosis or selection of treatment) or better execution of treatments. However, these two channels are extremely difficult to separate, in healthcare as well as other settings. Much of the decision-making input to medical care accrues not at the time when a treatment is realized, but along the chain of encounters and referrals that directs a patient to a given procedure.

This paper aims to explore an area of professional decision making – the decision by a kidney transplant team to accept or turn down an offered kidney – where the clinical decision and surgical performance are more clearly delineated and separately observable. Kidney transplantation provides a unique opportunity to explore the decision-making component of the volume outcome relationship: patients are added to the transplant waitlist after meeting clearly-defined diagnostic criteria; they face a common set of potential treatments; and the rich data on potential transplant recipient and donor health allows for detailed risk adjustment. In this setting, detailed records on what organs are accepted or turned down allow for assessment of provider decision-making, as well as of their post-surgical outcomes.

We find that greater center-level volume and cumulative experience are predictive of lower rates of organ offer acceptance throughout the distribution of organ quality. To address the question of center performance, we then evaluate organ offer responses along two dimensions: whether the potential recipient accepted another organ from a healthier donor within one year of a refusal; and whether the potential recipient died or was removed from the transplant waitlist for medical reasons within one year of a refusal. We find that larger centers perform worse on both dimensions, with lower rates of a better organ being accepted in the following year, and higher risk-adjusted rates of death or medical removal. Assessing how volume impacts execution, we explore post-surgical outcomes (death or failure of the transplanted kidney), and find evidence of reduced rates of post-transplant adverse events within one year at larger transplant centers. This tension between improved execution and reduced decision quality implies that practice may not make perfect in complex medical decision-making. Experience may need to be supplemented with decision supports or other tools to improve outcomes.