Date: Wednesday, April 21st, 2021
9:00 am – 10:00 am Pacific Time
12:00 pm – 1:00 pm Eastern Time
Location: Weekly Seminar, Zoom
Title: Fairness Considerations in Machine Learning Models – the Experience of a Large Healthcare Organization
This talk will focus on various fairness considerations in both the development and the application of machine learning based prediction models, as experienced in Clalit Health Services, a large healthcare organization in Israel.
The application and performance of an algorithm for improving sub-population calibration in predictive models will be presented. The algorithm is meant to address concerns regarding potential unfairness of the models towards groups which are underrepresented in the training dataset and thus might receive uncalibrated scores. The algorithm was implemented on widely used risk models, including the ACC/AHA 2013 model for cardiovascular disease and the FRAX model for osteoporotic fractures.
This algorithm also played a major role in the development of a COVID-19 mortality risk prediction model at a time when individual level data of COVID-19 patients was not yet available in Israel. The development process for this predictor will be presented. The resulting predictor was widely used within Clalit Health Services for prevention purposes, with an intention to notify high-risk members of their increased risk for mortality should they get infected, for prioritization of COVID-19 RT-PCR tests, and for treatment decisions in confirmed cases.
This talk is based on several joint works with Guy Rothblum, Gal Yona, Uri Shalit, Eitan Bachmat and Ran Balicer among others.
Noa Dagan holds an MD and an MPH from the Hebrew University, and a Ph.D. in Computer Science from Ben-Gurion University. She is currently a postdoctoral fellow in the Department of Biomedical Informatics (DBMI), Harvard Medical School. Noa is the director of data and AI-driven medicine at the Clalit Research Institute – the research institute of Israel’s largest healthcare organization, insuring and treating over 50% of the Israeli population.
Noam Barda holds an MD from Tel-Aviv University, a PhD co-advised in public health and computer science from Ben-Gurion University and a BSc in computer science from the Open University. He is a postdoctoral fellow in the Department of Biomedical Informatics (DBMI) at Harvard Medical School. He is the head of epidemiology and research at Clalit Research Institute.