Deep Learning-Based Patient Classification in Emergency Department

Abstract

This research investigates the application of deep learning to improve patient classification and triage in emergency departments (EDs). The project aims to develop a model that accurately predicts patient acuity levels based on readily available electronic health record (EHR) data, reducing wait times and improving resource allocation. Using a convolutional neural network (CNN) trained on a large EHR dataset, the study demonstrates improved classification accuracy compared to traditional methods. The resulting system holds promise for optimizing ED workflow and enhancing patient care. Future work will focus on model robustness and validation in diverse ED settings.

Introduction

Emergency departments (EDs) frequently face overcrowding and long wait times, leading to compromised patient care and increased costs. Accurate and efficient patient triage is crucial for optimizing resource allocation and ensuring timely treatment. Current triage systems, often based on nurse-led assessments, are subjective and prone to error. Deep learning offers a potential solution by enabling automated, data-driven classification of patients based on a wide range of clinical parameters extracted from EHR data. However, the development and validation of robust deep learning models for this application remain a significant challenge, due to the complexity of medical data and the need for explainable AI.

Objectives

• Develop a deep learning model capable of accurately classifying ED patients into acuity levels based on EHR data.
• Evaluate the performance of the developed model compared to existing triage methods.
• Demonstrate the feasibility and potential benefits of the proposed system in improving ED workflow and patient outcomes.