Seminar Title: Generalizability Challenges of Risk Prediction Models Based on Multi-institutional EHR Database
Dr. Hyungrok Do
Postdoctoral Fellow
NYU Grossman School of Medicine
Wednesday, February 8, 2023, In-Person, SERC-117 (Auditorium) at 12:10PM
Abstract:. As machine learning and statistical methods are increasingly being used to support and perform crucial decision making tasks, prediction models are required to be robust and generalizable in real-world scenarios. Particularly, as a multi-institutional database develops, we start to encounter new kinds of difficulties in building predictive models. In this talk, I illustrate the difficulties of building a robust and generalizable risk prediction model when the data have been collected from multiple institutions and introduce an algorithmic way to build a generalizable and robust prediction model.
Bio: Dr. Hyungrok Do is a postdoc research fellow at the division of biostatistics at NYU School of Medicine. His PhD is from the department of industrial management engineering at Korea University. His research interests include algorithmic fairness, domain generalization, and sample selection bias correction and their applications to healthcare problems.
For questions: Dr. Aziz Ezzat (aziz.ezzat@rutgers.edu) or Dr. Myong Kee Jeong (mjeong@soe.rutgers.edu)