@InProceedings{Chaikalis_WC2025,
author="Chaikalis, Nikolaos and Christodoulou, Theodoros and Kalitsi, Georgia and Tzatzimaki, Aikaterini and Kaldoudi, Eleni and Drosatos, George",
editor="Colyer, Christopher and Dell'Oro, Mikaela and Forster, Jake and Pope, Kenneth and Langman, Lynne",
title="Machine Learning-Based Prediction of Hospital Length-of-Stay and Patient Survival",
booktitle="Proceedings of IUPESM World Congress on Medical Physics and Biomedical Engineering XXVII (IUPESM WC 2025)",
year="2026",
publisher="Springer Nature Switzerland",
address="Cham",
pages="551--574",
series="IFMBE",
volume="140",
doi="10.1007/978-3-032-20291-8_45",
abstract="This study tackles two vital predictive challenges in hospitalised patients with severe illness: (i) Outcome Type (discharge versus death) and (ii) hospital Length-of-Stay. Leveraging a real-world dataset from the RegulaRN database comprising 13,145 patient records with 20 features spanning Request, Admission, Patient, and Clinical categories, we formulate these tasks as classification and regression problems, respectively. We evaluate several state-of-the-art Machine Learning models including eXtreme Gradient Boosting, Histogram Gradient Boosting and Random Forest, alongside Deep Neural Networks such as Multi-Layer Perceptrons. Classification models demonstrate strong performance, achieving F1-score up to 0.8, while regression models exhibit limited predictive accuracy with Mean Absolute Errors around 9--10 days. These results highlight the promise and current limitations of Machine Learning and Deep Learning in enhancing clinical decision support, underscoring the need for richer data and advanced modelling strategies in future research to optimise patient care.",
isbn="978-3-032-20291-8"
}