Forecasting Patient Discharge Volume in Northern Region Government Hospitals of Malaysia Using Machine Learning Techniques

Le résumé fourni par la source

The Malaysian public healthcare system is under growing strain, which is driven by an aging population and the rising burden of chronic non-communicable diseases (NCDs). These shifts have placed added pressure on hospitals, particularly in handling inpatient discharges. To respond to this challenge, the study set out to build a machine learning-based forecasting model that can predict discharge volumes. Hospital Taiping, one of the key government hospitals in northern Malaysia, was selected as the case study. Monthly inpatient discharge data from 2019 to 2023 formed the foundation for building predictive models. To strengthen the dataset, additional temporal features were engineered, including lagged discharge counts, rolling averages and seasonal indicators. Three machine learning models, namely Random Forest (RF), XGBoost and Support Vector Regression (SVR), were then trained and fine-tuned through Bayesian Optimization. Model performance was evaluated using coefficient of determination (R²), Root Mean Square Error (RMSE), Mean Absolute Scaled Error (MASE) and Mean Absolute Percentage Error (MAPE). The best-performing model was prospectively validated with actual discharge data from 2024. The Random Forest model demonstrated the lowest RMSE of 177.85, a MAPE of 3.23% and the highest R² of 0.43. Prospective validation using 2024 data confirmed its robustness in forecasting inpatient discharges. The study developed and validated a machine learning model to forecast inpatient discharge volumes in government hospitals across the Northern Region of Malaysia. It also contributes to both methodology and practical application by providing a forecasting framework that supports more informed hospital planning and resource management.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.