A0603
Title: Some new advances in similarity-based patient-outcome predictive modeling
Authors: Joel Dubin - University of Waterloo (Canada) [presenting]
Abstract: Earlier work has shown that similarity-based predictive models can improve upon predictive performance, as compared to using the entire training data to help build models, particularly regarding model discrimination for binary responses. The focus is on the similarity-based modeling for joint consideration of model calibration and discrimination, as well as for dynamic prediction models. Properties of the methods are investigated in comprehensive simulation studies, and the methods are demonstrated through a separate analysis of a publicly available intensive care unit (ICU) database.
