Dissertation
Scientific racism without scientific racists: how algorithms launder inequality into medicine
University of Iowa
Doctor of Philosophy (PhD), University of Iowa
Summer 2025
DOI: 10.25820/etd.008069
Abstract
American medicine has been plagued in recent years by two big controversies. One controversy revolves around a ‘race question’: What ought to be the role of race in medicine? The other revolves around an ‘AI question’: What ought to be the role of “artificial intelligence” or statistical prediction in medicine? This dissertation offers a new perspective on both of these controversies through a historical-sociological study of a racialized medical algorithm. I will argue that statistical prediction is a novel channel through which scientific racism is laundered into medicine. Given the endurance of racial health disparities, it is critical to understand the novel channels through which racism is perpetuated in medicine. And given the rapid expansion of statistics-based clinical decision-making, it is critical to investigate what these devices actually do and how they are used.
The algorithm at the center of this study is widely used by clinicians to estimate patients’ kidney function, used for everything from monitoring long-term declines in kidney function, to dosing medications, to defining eligibility for the national kidney transplant registry. From roughly 1999 through 2022, this algorithm included a Black race adjustment, such that Black patients had their kidney function score increased by 15-18%. In 2022, a national task force formally recommended the use of kidney function estimating equations without a Black race coefficient.
The common narrative of the life of this algorithm is framed as one of linear progress: The coefficient was added in the late 1990s due to erroneous, racist ideas about biology and race; the coefficient was removed in 2022 because now we know better. Through interviews with clinicians and analysis of literature and archival documents, this dissertation challenges this narrative in two respects: First, I attempt to show that the source of legitimacy of the race coefficient in the 2000s and 2010s was less biological myths about race, and more statistical procedural norms. Second, I argue that the task force decision to remove the race coefficient, while a beneficial step, does not resolve a deeper contradiction in the use of race and statistics in medicine. That deeper contradiction lies in a way of interpreting data and correlations embedded in statistical procedure, which I name statistical predictionism. Statistical predictionism maps individual-level meaning onto statistical relationships, even when there is no known individual-level causal mechanism to justify doing so. I argue that this way of interpreting statistics is facilitated by a political economy of American healthcare which privileges individual-level causes of and solutions to chronic disease, while obscuring system-level ones. Understood this way, both the race coefficient and the medical algorithm which houses it are novel channels through which racial and economic hierarchies are laundered into medicine. Rather than assuming statistical prediction tools will automatically bring about better medicine, this dissertation argues that we should be critical of the assumptions and goals embedded within these devices.
Details
- Title: Subtitle
- Scientific racism without scientific racists: how algorithms launder inequality into medicine
- Creators
- Hannah Zadeh
- Contributors
- Louise Seamster (Advisor)Dan Hirschman (Committee Member)Freda Lynn (Committee Member)Mike Sauder (Committee Member)
- Resource Type
- Dissertation
- Degree Awarded
- Doctor of Philosophy (PhD), University of Iowa
- Degree in
- Sociology
- Date degree season
- Summer 2025
- DOI
- 10.25820/etd.008069
- Publisher
- University of Iowa
- Number of pages
- xiv, 299 pages
- Copyright
- Copyright 2025 Hannah Zadeh
- Language
- English
- Date submitted
- 07/16/2025
- Description illustrations
- illustrations, tables, graphs, facsimiles
- Description bibliographic
- Includes bibliographical references (pages 242-284).
- Public Abstract (ETD)
American medicine has been plagued in recent years by two big controversies. One controversy revolves around a race question : What ought to be the role of race in medicine? The other revolves around an AI question : What ought to be the role of artificial intelligence or statistical prediction in medicine? This dissertation offers a new perspective on both of these controversies through investigating the past and present of a racialized medical algorithm. By exploring how this algorithm came to be and how it is modernly used, I show how these controversies about the use of race and AI in medicine can both be explained by a powerful and ubiquitous, but largely unacknowledged, way of understanding statistics. This way of thinking about statistics, which I call statistical predictionism, allows researchers to seamlessly conflate statistical averages with individual-level certainties, such that statistical correlations are automatically interpreted as individual-level statistical predictions, even without any known individual-level causal mechanism to justify doing so. Statistical predictionism which quietly dissolves the old adage that correlation is not causation makes possible racialized medical algorithms and risk scores, as well as many emergent big data approaches to medicine, and its growth is facilitated by an American healthcare policy environment which privileges individual-level causes of and solutions to disease, over environmental ones. Identifying and challenging this way of interpreting statistics is important to developing a robust and effective medical science that can benefit all patients and communities. In its current form, statistical predictionism operates as a magic wand which can turn system-level causes of disease into individual-level ones, building a world where risk of disease and death leaches into our soils and watersheds, and then into our bodies, while the statistics say the risk was always in our bodies to begin with.
- Academic Unit
- Sociology and Criminology
- Record Identifier
- 9984948738702771
Metrics
3 File views/ downloads
15 Record Views