A0284
Title: Bayesian electronic health record analysis and its application in Cirrhosis progression
Authors: Xiaodan Fan - The Chinese University of Hong Kong (Hong Kong) [presenting]
Abstract: The high missingness in electronic health records (EHR) often hinders the efficient exploitation of its rich medical information. The high missingness is addressed in the context of disease progression study by proposing a Bayesian inhomogeneous hidden Markov model. The model is highly interpretable and is capable of efficiently predicting the latent states and performing imputation based on incomplete observations. The proposed model is validated by simulation studies and application to the identification of crucial signals and the prediction of disease progression trajectories.
