2019/12/19 by Alexis Bellot, Mihaela van der Schaar, Bellot, Alexis +1 · 1 citation
Computer Science · Health Professions · Mathematics · Medicine · #Applications (stat.AP) #Artificial Intelligence in Healthcare #Bayesian Methods and Mixture Models #FOS: Computer and information sciences #Liver Disease Diagnosis and Treatment #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Statistical Methods and Inference
paper · pdf · doi:10.48550/arxiv.1912.09086
openalex publication_date 2019/12/19 · openalex created_date 2022/07/26 · openalex updated_date 2026/07/28
Analyzing electronic health records (EHR) poses significant challenges\nbecause often few samples are available describing a patient's health and, when\navailable, their information content is highly diverse. The problem we consider\nis how to integrate sparsely sampled longitudinal data, missing measurements\ninformative of the underlying health status and fixed demographic information\nto produce estimated survival distributions updated through a patient's follow\nup. We propose a nonparametric probabilistic model that generates survival\ntrajectories from an ensemble of Bayesian trees that learns variable\ninteractions over time without specifying beforehand the longitudinal process.\nWe show performance improvements on Primary Biliary Cirrhosis patient data.\n