ORAHS2024
Abstract Submission

73. A simulation-based analysis of machine learning algorithms for length of stay classification in the intensive care unit

Contributed abstract in session HB-2: Analytics, stream Regular talks.

Thursday, 11:00-12:30
Room: Room S1

Authors (first author is the speaker)

1. Sara Garber
Statistics and Data Science, University of Augsburg

Abstract

A recent report by the German Council of Experts on the assessment of developments in the healthcare sector shows that the German healthcare system is not crisis-resistant and not sufficiently prepared for the consequences of climate change or pandemics. In the event of crisis-related challenges, the intensive care unit is one of the areas particularly affected. In order to enable efficient capacity control in high-load situations, an accurate classification of the length of stay is essential. In the literature, binary classification is usually applied, however, a more precise distinction between more than two classes offers advantages in terms of capacity management. Therefore, we apply multiple machine learning algorithms for multiclass classification differing in terms of transparency and explainability on a real-world intensive care dataset. In addition to state-of-the-art performance indicators, we use a Monte Carlo simulation for further analysis and evaluation of the classification algorithms. This approach provides in-depth insights into the effects of an actual application of the machine learning algorithms in the intensive care unit.

Keywords

Status: accepted


Back to the list of papers