Predicting Hospital Admission by Adding Chief Complaints Using Machine Learning Approach

Abstract

Overcrowded conditions in emergency departments (EDs) have increased patients’ waiting time, while the variety of patient afflictions has caused difficulties in the allocation of medical resources. Therefore, the ability to predict a patient’s hospital admission at the time of triage could allocate medical resources to patients who go to EDs in urgent need of immediate care. Using a dataset from the MacKay Memorial Hospital in Taipei (Taiwan), which contains 177,038 valid records collected from 2009 to 2010 in this research, we aim to have on hand chief complaints (CCs), demographic data, administration information and clinical information at the triage stage to predict the probability of a patient’s hospital admission. Firstly, we select terms from the CCs to predict which patients may require eventual hospitalization. We then integrate the selected terms with several algorithms to predict the probability of patient admissions. Accordingly, this research includes a series of machine learning processes, such as data preprocessing for structure data and CC data, imbalanced data processing, models construction by logic regression, neural networks, random forest, XGBoost, and model evaluation. The research results show that the ensemble learning approach, XGBoost, can achieve 0.88, and 0.76 in terms of accuracy and AUC respectively. The results show that triage, fever status, age, and terms extracted from the CCs are important attributes to predict if patients should be hospitalized. The results of this study will provide a reference approach in the field of emergency hospital admissions prediction and help hospitals improve resource allocation in emergency rooms.