2020/11/02 by Jonathan Heitz, Heitz, Jonathan, Joanna Ficek +9
Computer Science · Decision Sciences · Mathematics · Medicine · #Artificial intelligence #Computer science #Data mining #Econometrics #FOS: Computer and information sciences #Forecasting Techniques and Applications #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Machine Learning in Healthcare #Machine learning #Mathematics #Parametric model #Parametric statistics #Probabilistic logic #Robustness (evolution) #Sepsis Diagnosis and Treatment #Statistics #cs.LG #stat.ML
paper · pdf · doi:10.48550/arxiv.2011.00865
9 pages, 6 figures
arxiv created 2020/11/02 · openalex publication_date 2020/11/02 · arxiv updated 2020/11/03 · openalex created_date 2025/10/10 · openalex updated_date 2026/08/06
Dynamic assessment of mortality risk in the intensive care unit (ICU) can be\nused to stratify patients, inform about treatment effectiveness or serve as\npart of an early-warning system. Static risk scoring systems, such as APACHE or\nSAPS, have recently been supplemented with data-driven approaches that track\nthe dynamic mortality risk over time. Recent works have focused on enhancing\nthe information delivered to clinicians even further by producing full survival\ndistributions instead of point predictions or fixed horizon risks. In this\nwork, we propose a non-parametric ensemble model, Weighted Resolution Survival\nEnsemble (WRSE), tailored to estimate such dynamic individual survival\ndistributions. Inspired by the simplicity and robustness of ensemble methods,\nthe proposed approach combines a set of binary classifiers spaced according to\na decay function reflecting the relevance of short-term mortality predictions.\nModels and baselines are evaluated under weighted calibration and\ndiscrimination metrics for individual survival distributions which closely\nreflect the utility of a model in ICU practice. We show competitive results\nwith state-of-the-art probabilistic models, while greatly reducing training\ntime by factors of 2-9x.\n