Topic: The Role of Physician Behavior in the Allocation of Healthcare Resources in the United States

Speaker: Associate Professor Jeffrey S. McCullough, Department of Health Management and Policy, University of Michigan (Ann Arbor)

Host: Associate Professor Luan Mengna from RIEM, SWUFE

Time: July 16, 2025 (Wednesday) 10:00-12:00

Location: Conference Room 1211, Gezhi Building, Liulin Campus

Organizer: RIEM

Speaker's Profile

Dr. Jeffrey McCullough is an Associate Professor in the Department of Health Management and Policy at the University of Michigan. His research focuses on technology and innovation in health care, with particular emphasis on information technology, pharmaceuticals, and empirical methods. Dr. McCullough's recent research on pharmaceuticals involves the impact of antitrust policies and direct-to-consumer advertising on health outcomes. He has conducted extensive research on the impact of electronic health record systems on quality and efficiency, and his IT‑related work explores how technology facilitates measurement and dissemination of healthcare quality information. Dr. McCullough also applies machine‑learning methods to medical decision‑making, risk adjustment, and incentive design for health‑care delivery, and he studies the underlying causes of gender and group disparities in medical care.

Abstract

    We examine the differences in imaging use across patient groups in emergency care, focusing on diagnostic decisions for patients presenting with acute abdominal pain. Acute abdominal pain is a common condition with high risk if misdiagnosed. Our analysis shows that, despite nearly identical rates of serious conditions across groups, some groups are significantly more likely to receive imaging than others. To investigate the sources of this disparity, we combine detailed electronic health record data from emergency departments in Michigan with features arising from the quasi‑random assignment of patients to physicians to study the role of physician behavior and resource allocation. We find that within the same physician’s practice, missed‑diagnosis rates are broadly similar across groups, indicating that physicians apply comparable clinical judgment when treating similar patients. However, differences in how diagnostic resources are allocated across physicians can result in some groups having fewer opportunities to receive imaging. Preliminary estimates suggest that more balanced allocation of diagnostic resources could substantially reduce this gap, although it may not eliminate it entirely.

    We also assess the influence of patient behaviors (for example, treatment refusal or leaving against medical advice) on imaging rates. Although such behaviors are slightly more common in some groups, they are rare overall and typically occur after imaging is completed. Even under an extreme assumption that all patients who both did not receive imaging and left early had serious conditions, our main conclusions remain robust. Overall, our results suggest that observed differences in imaging use are primarily driven by resource allocation processes in care delivery rather than by physicians applying different diagnostic standards under comparable conditions; remaining differences may reflect variation in clinical presentation or symptom detectability across groups.