Exploring Prediction Models for Hospital Patient Discharge Turnaround Time: A Comparative Study
DOI:
https://doi.org/10.37934/araset.56.2.196205Keywords:
Hospital discharge, Turnaround time, Prediction model, XGBoostAbstract
Patient Discharge Turnaround Time has a significant impact on healthcare quality. However, improving this process has been a challenge. This study focuses on applying predictive modelling techniques to analyse patient discharge patterns in hospital. The goal is to develop a model that can predict timely patient discharge based on various features. The dataset used is derived from the Hospital Operation Management and Information System (HOMIS) of Cotabato Sanitarium and General Hospital, situated at Maguindanao, Philippines encompassing comprehensive data from admission to discharge. The process involves retrieving data from the database, followed by data cleaning and preparation to ensure quality. Feature engineering is also performed to extract additional information. Several supervised and unsupervised predictive modelling algorithms are employed, and performance metrics are used to evaluate the models. Results indicate that XGBoost achieves the highest performance, with an AUC score of 0.8207 and an accuracy rate of 0.7521. The hour of the discharge order emerges as the most significant predictor for timely discharge. This study demonstrates the application of predictive modelling in healthcare, particularly in predicting patient discharge turnaround time, contributing to existing knowledge that aims to enhance health outcomes using machine learning techniques.