How Management Made Medicine: The Evolution of Quality Improvement from Industrial Production to Medical AI

Abstract: In contemporary academic health systems, many activities bearing resemblance to biomedical science are classified as quality improvement (QI) rather than clinical research. The goal of QI is typically not the creation of generalizable knowledge, and therefore this work is often exempt from ethical oversight by institutional review boards. Yet in practice, the boundary between QI and clinical research is porous, as medical experts with similar credentials and aims are often involved in both types of activities. This paper considers the bioethical implications of this fuzzy distinction. First, the paper explores how QI originated in mid-twentieth century management science and gained favor with private sector organizations such as Xerox and AT&T. Charting this history forward, I show how QI later became an alternative category of medical activity to clinical research, shepherded by the work of Donald Berwick on the National Demonstration Project. By the 2000s, this movement was formally recognized by the National Academy of Medicine (then the Institute of Medicine), which published a series of reports grappling with the implications of QI for care delivery in modern health systems that engage in continuous learning. Today, many applications of AI and machine learning are classified as QI work, which often suits the needs of model developers but can pose risks to patients and clinicians. The paper argues that the historical blurring of the boundary between QI and clinical research may provide lessons for conceptualizing new infrastructure to provide ethical oversight and monitoring of emergent health technologies in the future.

Keywords: Quality Improvement, Ethical Oversight, AI Governance