2016/11/16 by Ahmed M. Alaa, Alaa, Ahmed M., Jinsung Yoon +5
Computer Science · Medicine · #Bayesian Modeling and Causal Inference #FOS: Computer and information sciences #Healthcare Technology and Patient Monitoring #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Machine Learning in Healthcare #Sepsis Diagnosis and Treatment #Time Series Analysis and Forecasting
paper · pdf · doi:10.48550/arxiv.1611.05146
openalex publication_date 2016/11/16 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
Critically ill patients in regular wards are vulnerable to unanticipated\nclinical dete- rioration which requires timely transfer to the intensive care\nunit (ICU). To allow for risk scoring and patient monitoring in such a setting,\nwe develop a novel Semi- Markov Switching Linear Gaussian Model (SSLGM) for the\ninpatients' physiol- ogy. The model captures the patients' latent clinical\nstates and their corresponding observable lab tests and vital signs. We present\nan efficient unsupervised learn- ing algorithm that capitalizes on the\ninformatively censored data in the electronic health records (EHR) to learn the\nparameters of the SSLGM; the learned model is then used to assess the new\ninpatients' risk for clinical deterioration in an online fashion, allowing for\ntimely ICU admission. Experiments conducted on a het- erogeneous cohort of\n6,094 patients admitted to a large academic medical center show that the\nproposed model significantly outperforms the currently deployed risk scores\nsuch as Rothman index, MEWS, SOFA and APACHE.\n