Deep Learning with Heterogeneous Graph Embeddings for Mortality\n Prediction from Electronic Health Records

2020/12/27 by Tingyi Wanyan, Hossein Honarvar, Wanyan, Tingyi +7 · 1 citation
Computer Science · Health Professions · Medicine · #Artificial Intelligence in Healthcare #Artificial Intelligence in Healthcare and Education #FOS: Computer and information sciences #Machine Learning (cs.LG) #Machine Learning in Healthcare

paper · pdf · doi:10.48550/arxiv.2012.14065

openalex publication_date 2020/12/27 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

Computational prediction of in-hospital mortality in the setting of an\nintensive care unit can help clinical practitioners to guide care and make\nearly decisions for interventions. As clinical data are complex and varied in\ntheir structure and components, continued innovation of modeling strategies is\nrequired to identify architectures that can best model outcomes. In this work,\nwe train a Heterogeneous Graph Model (HGM) on Electronic Health Record data and\nuse the resulting embedding vector as additional information added to a\nConvolutional Neural Network (CNN) model for predicting in-hospital mortality.\nWe show that the additional information provided by including time as a vector\nin the embedding captures the relationships between medical concepts, lab\ntests, and diagnoses, which enhances predictive performance. We find that\nadding HGM to a CNN model increases the mortality prediction accuracy up to\n4 %. This framework serves as a foundation for future experiments involving\ndifferent EHR data types on important healthcare prediction tasks.\n

Cited by

• Exploring a Graph-based Approach to Offline Reinforcement Learning for Sepsis Treatment

Related

• Modeling Irregularly Sampled Clinical Time Series
• Scalable and accurate deep learning with electronic health records
• Integrating Physiological Time Series and Clinical Notes with Deep\n Learning for Improved ICU Mortality Prediction
• All Data Inclusive, Deep Learning Models to Predict Critical Events in the Medical Information Mart for Intensive Care III Database (MIMIC III)
• Dynamic survival prediction in intensive care units from heterogeneous time series without the need for variable selection or pre-processing