Predictive Analytics for Reducing Patient Re-admission Rates in the Healthcare Sector

Lenka, Aman (2024) Predictive Analytics for Reducing Patient Re-admission Rates in the Healthcare Sector. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (696kB) | Preview
Preview	PDF (Configuration Manual) Download (378kB) | Preview

Abstract

Patient relapses as a referring factor is a major concern in healthcare because of its negative impacts on patient status and financial pressure on healthcare systems. This danger provides predictive modelling with the potential to reduce such risks by flagging patients most at risk. This work investigates the enhancement of machine learning models for hospital readmissions prediction, with Random Forest, XGBoost, neural network models through Grid Search and Bayesian Optimization for hyperparameters tuning.

Preprocessing of the data involved use of categorical variables for encoding, numerical scaling as well as feature engineering using clinical and demographic variables. Comparing the results showed that Bayesian Optimization worked better optimizing the neural networks has a better AUC of 0.65 and accuracy of 0.6174. Grid Search optimization of neural networks provided competitive results (AUC: 0.64, Accuracy: 0.6078). The Random Forest and XGBoost models, despite their robustness and interpretability, yielded lower predictive performance (AUC: 0.60, Accuracy: 0.60 and 0.59, respectively).

The observations at this case show how modern hyperparameter optimization methods add more value to the neural networks in terms of metrics such as precision/recall. Thus, now there are still some issues that remain unsolved: Imbalanced datasets or datasets that have unequal numbers of data points belonging to different classes, where it is difficult to determine the efficiency of the model, and issues related to the interpretability of the identified patterns. This study is important as it demonstrates how deep learning and optimization schemes can be utilized for the formulation of practical intervention strategies in healthcare information system, which presents a viable way of using evidence- based information for the decline on the rate of hospital readmission.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Cosgrave, Noel UNSPECIFIED
Subjects:	Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science R Medicine > Healthcare Industry Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Ciara O'Brien
Date Deposited:	03 Sep 2025 11:20
Last Modified:	03 Sep 2025 11:20
URI:	https://norma.ncirl.ie/id/eprint/8735

Actions (login required)

View Item