Emergency Department Triage Prediction Of Clinical Outcomes Using Deep Learning Approach

Abstract

Triage Systems Development In Hospitals Is Essential For Emergency Department (ED) To Accurately Differentiate And Prioritize Critically Ill From Stable Patients. It Is To Give Treatment Priority To The Most Urgent Patients. Adama Hospital Is One Of The Largest Referral Hospitals In East Showa, Many Patients, Especially Those With Very Critical Conditions Are Admitted Daily To Hospitals. The Hospital Emergency Department (ED) Is Still Working On The Manual Procedure To Triage The Patients Who Are Even Under Critical Conditions, This Makes The Hospital Difficult To Save Lives And Inefficient In Resource Mobilization. In This Study, We Have Planned To Automate This Process By Building A Deep Learning-Based Model, Which Can Reduce Complexity And Boost The Quality Of Service. We Have Developed A Dataset For Model Development By Gathering Real Data From ACSMC. Our Dataset Comprises 5000 Instances With Twelve Variables, One Of Which Is The Target Variable. We Have Been Applying Multiple Data Preprocessing And Preparation Techniques Like Cleaning And Feature Scaling. Then We Trained Distinct Modes, By Using These Deep Learning Algorithms Such As Long-Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), And Convolutional Neural Network (CNN), For Comparison Of Their Performance Of Prediction Of The Triage Outcome Of Patients. We Have Investigated The Effect Of Algorithms Hybridization Such As CNN With LSTM, In This Case, We Used CNN For Extracting Optimized Features Directly From The Dataset. In Addition, We Also Analyzed Bidirectional Encodings Such As Bi-LSTM And Bi-GRU. With Performance Evaluation Parameters Like Accuracy And F1-Score, The Performance Of The Models Is Assessed Under Stratified 10-Fold Cross-Validation. The Performance Evaluation Findings Showed That The LSTM Model Outperformed Other Models For Prediction. It Achieved An F1-Score Of 83.5 % And An Accuracy Of 83.7 %. At The End, We Recommend A Further Experiment To Improve The Accuracy Of The Work Itself By Gathering Ample Data That Is Efficient For The Deep Learning Model.