Multimodal Attention-based Deep Learning for Emergency Triage with Electronic Health Records

2026-07-18

AI Summary

This study addresses the challenge of accurately prioritizing patients in emergency triage by integrating unstructured chief complaints with structured vital signs. The authors propose a novel multimodal deep learning architecture that, for the first time, applies self-attention mechanisms to emergency triage, enabling joint modeling of textual and tabular data from electronic health records. This approach effectively captures both local and global dependencies among heterogeneous features. Evaluated on an emergency department dataset comprising 11,102 patient encounters, the model achieves relative improvements of 1.95% in accuracy, 2.49% in F1 score, and 1.41% in ROC AUC over baseline methods, demonstrating significantly enhanced capability in interpreting complex clinical information and supporting triage decisions.

Abstract

Accurate emergency triage decision is critical to avoid clinical deterioration, morbidity, and mortality. Machine learning-based triage system involves acquiring the main presenting complaint in text form and assessing vital signs in numerical data, enabling an automated and efficient analysis of patient information for timely and accurate prioritization of medical attention. However, modelling the intricacies of both data types requires a comprehensive understanding of the temporal structure and dependencies within the data. Thus, the aim of this study is to propose a multimodal deep learning architecture that can effectively handle both tabular and textual data. Furthermore, the proposed model exploits self-attention to to capture both local and global relationships between the features. A dataset consisting of 11,102 triage data collected from emergency department of Hospital Universiti Sains Malaysia is used for model development and validation. The proposed model demonstrated an increase of 1.95% in accuracy, 2.49% in F1-score, and 1.41% in ROC AUC compared to the baseline model. The experimental results demonstrated the potential of the proposed model in predicting triage decisions.

Problem

• emergency triage
• multimodal data
• electronic health records
• text and tabular data
• clinical decision support

Innovation

• multimodal deep learning
• self-attention
• emergency triage
• electronic health records
• text-tabular fusion