A0877
Title: Prediction of relapse in pediatric chronic disease using compound medical records
Authors: Jiasheng Shi - The Chinese University of Hong Kong (Hong Kong) [presenting]
Abstract: In pediatric chronic disease studies, the recorded data often presents compound data features, e.g., longitudinal data with fragment medical records and time series medical records interlaced with each other. These compound data features provide ample information but require a hybrid statistical analyzing procedure. To address this difficulty, a novel formulation is proposed for medical data with such a compound data structure and an efficient algorithm is proposed to tackle the extrapolation and clustering loop within the overall estimation procedure. An application to pediatric ulcerative colitis chronic disease is presented for the estimation and relapse risk prediction, which is particularly useful for patients' disease management profiles.
