Modelling and Predicting Medical Outcomes for Intensive Care Patients via Network Mechanisms and Machine Learning

Aug 4, 2026

Upload by Hazel

The intensive care unit (ICU) delivers critical life support to patients with diverse care needs, backgrounds, and demographics, making assessment and treatment evaluation highly complex and affecting medical outcomes. ICU monitoring generates extensive patient data, and machine learning has been explored to support clinical decisions, but gaps remain due to data quality issues, patient heterogeneity, and difficulties integrating medical and computer science knowledge. This thesis proposes a framework combining network representations with non-parametric community detection. The approach is applied to multimorbidity profiling and in-hospital mortality prediction, showing robust structures under missing data, automatic incorporation of heterogeneity, flexible hierarchical modeling, reduced parameter selection, and interpretable visual delivery.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages160

Reading Time6 h 43 min

AccessFree to download

Abstract

Lay Summary

Acknowledgements

This thesis has been submitted in fulfilment of the requirements for a postgraduate degree (e. g. PhD, MPhil, DClinPsychol) at the University of Edinburgh. Please note the following terms and conditions of use: • This work is protected by copyright and other intellectual property rights, which are retained by the thesis author, unless otherwise stated. • A copy can be downloaded for personal non-commercial research or study, without prior permission or charge. • This thesis cannot be reproduced or quoted extensively from without first obtaining permission in writing from the author. • The content must not be changed in any way or sold commercially in any format or medium without the formal permission of the author. • When referring to this work, full bibliographic details including the author, title, awarding institution and date of the thesis must be given. Modelling and Predicting Medical Outcomes for Intensive Care Patients via Network Mechanisms and Machine Learning Jorge Alejandro Gaete Villegas Doctor of Philosophy Artificial Intelligence Applications Institute School of Informatics University of Edinburgh 2024 Abstract The intensive care unit (ICU) provides critical life support to a diverse group of patients with varying care needs, medical background, and demographics. This heterogeneity significantly increases the complexity of care, impacting the assessment of patients, the efficacy of treatments, and ultimately, the medical outcomes of these patients. To support medical decisions, ICU patients are constantly monitored, making available a wealth of patient information. Recent work has extensively explored the application of machine learning to leverage this data to support decision-making in the ICU. However, challenges related to the quality of ICU data, patient heterogeneity, and the integration of medical and computer science knowledge create a gap between machine learning research and its practical implementation in critical care. In this thesis, we introduce an innovative approach that combines a network representation of clinical data with non-parametric community detection in networks. With this method, we aim to complement and enhance existing models, addressing the aforementioned challenges related to ICU care. We apply our approach to two tasks in the ICU: the identification of multimorbidity profiles for patients and in-hospital mortality prediction. Through these two studies we concretely present the challenges to the application of machine learning, show the limitations of existing approaches, and present the methodological benefits of our approach. Our results highlight several advantages of our approach over existing methods. Firstly, the network representation of patients along with community detection, reveals statistically robust network structures that mitigate the impact of missing data. Secondly, our approach captures elements of heterogeneity, such as ethnicity and age, and automatically incorporates them into our clustering and prediction models. Third, the hierarchical structure of our approach offers flexibility to accommodate and adapt to varying levels of available data. Fourth, the non-parametric nature of our approach eliminates the need for complex parameter selection. Finally, the network representation of data offers a visual and more intuitive delivery of our models. In this way, our work represents a step forward in the use of networks and machine learning methods to enhance decision-making support in the ICU in consideration of its complexity. Future directions outlined in this thesis involve enriching clinical data representation, exploring network reconstruction for mortality prediction, and incorporating medical knowledge to augment model interpretability. Lay Summary In the intensive care unit (ICU), where patients receive critical life support, providing effective care is challenging due to the diverse needs, medical backgrounds, and d

FAQs

Why is predicting medical outcomes in the ICU challenging?

ICU patients vary widely in needs, medical background, and demographics, creating strong heterogeneity. Data quality issues further increase the complexity of assessment, treatment evaluation, and outcome prediction.

What is the core method proposed in this thesis?

The thesis combines a network representation of clinical data with non-parametric community detection in networks. This aims to complement existing models and address quality and heterogeneity challenges in ICU data.

Which two ICU tasks does the thesis focus on?

It applies the approach to identifying multimorbidity profiles and to predicting in-hospital mortality. These studies evaluate limitations of existing methods and demonstrate methodological benefits.

Aug 4, 2026

Upload by Hazel