Does Artificial Intelligence have a role to play in the pursuit of health equity? Three panelists share their perspectives from history, statistics, machine learning, and policy.
Dr. Heather Mattie will talk about algorithmic bias in healthcare, describe where bias can be introduced in the process of creating and implementing prediction algorithms, and review ways to mitigate bias.
Dr. David S. Jones will talk about reconsidering use of race correction in algorithms with this talk titled, "Hidden in Plain Sight -- Reconsidering the Use of Race Correction in Clinical Algorithms".
Dr. Nathaniel Hendrix will talk about equitable translation of clinical AI methods in his talk, titled "Cost-Effectiveness Analysis to Support the Equitable Implementation of Clinical Artificial Intelligence."
Bio: Dr. Heather Mattie is a Lecturer on Biostatistics and co-director of the Health Data Science Master’s program in the Biostatistics department at the Harvard T.H. Chan School of Public Health. She teaches several courses including Introduction to Data Science, Reproducible Data Science, and Data Science II: Deep Learning. Dr. Mattie’s research focuses on the intersection between biostatistics, data science, and network science. Specifically, she has used network science and machine learning to study interactions in communities, as well as the development and application of artificial intelligence in healthcare research. Her research has also involved the notion of algorithmic bias, in terms of an algorithm compounding inequities working against underrepresented or disadvantaged groups in society. Her work has found links between unhealthy weight control behaviors and the use of mobile dating applications, particularly in racial and ethnic minorities. She has developed methods that predict tie strength in a network, which assists in modeling the spread of disease and information. Additionally, her work has examined the potential for artificial intelligence to improve inference from data for care and population health, as well as the challenges related to bias and scalability in such models.
Talk title: Algorithmic bias in health care
Talk abstract: Machine learning algorithms are increasingly being used in public health and medicine with the potential to exacerbate health inequities at scale. Ensuring these algorithms are accurate but also objective and fair is critical. I will introduce algorithmic bias in health care, describe where bias can be introduced in the process of creating and implementing prediction algorithms, and review ways to mitigate bias.
David S. Jones:
Bio: Trained in psychiatry and history of science, David Jones is the Ackerman Professor of the Culture of Medicine at Harvard University. His research has explored the causes and meanings of health inequalities (Rationalizing Epidemics: Meanings and Uses of American Indian Mortality since 1600) and decision making in cardiac therapeutics (Broken Hearts: The Tangled History of Cardiac Care). He is now pursuing three new projects, on the evolution of coronary artery surgery, on heart disease and cardiac therapeutics in India, and on the threat of air pollution to health. He teaches the history of medicine, medical ethics, and social medicine at Harvard College and Harvard Medical School.
Talk title: "Hidden in Plain Sight — Reconsidering the Use of Race Correction in Clinical Algorithms”
Abstract: Many diagnostic tests, risk calculators, and clinical practice guidelines used in medicine in the United States take a patient’s race and ethnicity into consideration. Over the past five years, a furious debate has emerged about whether such practices (1) are evidence-based and (2) are likelier to address or exacerbate pervasive health inequities. Most of the existing tools are based on conventional methods of research and multivariate analysis. I will review the existing tools and their limitations in hopes of identifying areas where big data and medical AI might be able facilitate the development of better clinical tools and guidelines.
Bio: Nathaniel Hendrix is a postdoctoral research fellow in the Department of Global Health & Population at the Harvard T.H. Chan School of Public Health. His past training is in pharmacy and health economics, and the focus of his fellowship is cost-effectiveness methodology, which he studies under the supervision of Stéphane Verguet. His research in artificial intelligence is in three main areas: using conjoint analysis to study stakeholder attitudes towards clinical artificial intelligence; leveraging simulation methods to inform the clinical translation of artificial intelligence-based risk scores in cancer screening; and developing methodologies for performing cost-effectiveness analysis on artificial intelligence-based clinical tools.
Title: Cost-Effectiveness Analysis to Support the Equitable Implementation of Clinical Artificial Intelligence
Abstract: The increasing number of artificial intelligence-based tools approved for use in clinical settings has made translational research on this topic urgently necessary. Methods in cost-effectiveness analysis have been developed to promote equitable decision-making around many health technologies, and further methodological work is under way to ensure that these methods are well-suited to clinical AI. I will discuss equity-centered cost-effectiveness methodologies and the data that health economists need from AI research to perform these analyses on AI-based tools.