University of Twente Student Theses

Login

From Laboratory Test Results to Emergency Department Admission Status: Forecasting with Machine Learning and Predictive Process Mining

Naguine, P.V. (2024) From Laboratory Test Results to Emergency Department Admission Status: Forecasting with Machine Learning and Predictive Process Mining.

[img] PDF
1MB
Abstract:Emergency department overcrowding is a significant issue impacting healthcare systems globally, influencing patient care and resource allocation. This study investigates whether predictive process mining offers an improvement over traditional machine learning methods for classifying emergency department admissions using sequential medical data. By leveraging the MIMIC-IV dataset, which includes laboratory tests conducted during hospital admissions and captures dynamic changes in test results over time, the research compares the performance of predictive process mining and machine learning models. Results show that the standalone machine learning model, which differs from predictive process mining models primarily in the data itself and its format, has a performance comparable to the predictive process mining model with event-level features. However, it outperforms the predictive process mining model with case-level features in terms of accuracy, precision and recall. The study also identifies limitations, such as the exclusion of general practitioner visits and pre-hospitalisation tests from the dataset and challenges related to class imbalance, which impact model training and generalisability.
Item Type:Essay (Master)
Clients:
Medisch Spectrum Twente, Enschede, Netherlands
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:54 computer science
Programme:Computer Science MSc (60300)
Link to this item:https://purl.utwente.nl/essays/102778
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page