Related Experiment Video
Updated: Aug 30, 2025
09:09
Description of a Swine Infant Model of Volume-Controlled Hemorrhagic Shock
Published on: November 3, 2023
1.1K
K-Means Clustering for Shock Classification in Pediatric Intensive Care Units
María Rollán-Martínez-Herrera1,2,3, Jon Kerexeta-Sarriegi3,4,5, Javier Gil-Antón1,2
1Cruces University Hospital, 48903 Barakaldo, Spain.
Diagnostics (Basel, Switzerland)
|August 26, 2022
Summary
Unsupervised learning, specifically k-means clustering, effectively stratified pediatric intensive care unit (PICU) patients with shock. This data-driven approach demonstrated superior correlation with mortality compared to traditional clinical methods.
Area of Science:
- Pediatric critical care medicine
- Data science in healthcare
- Machine learning applications
Background:
- Shock, characterized by inadequate tissue oxygenation, requires precise classification for effective treatment.
- Current clinical methods for shock stratification in pediatric intensive care units (PICUs) are often outdated.
- There is a need for advanced methods to improve patient stratification and treatment outcomes.
Purpose of the Study:
- To investigate the utility of unsupervised classification methods for stratifying pediatric shock patients.
- To compare the efficacy of unsupervised clustering with traditional clinical classification in predicting patient outcomes.
- To enhance patient management through data-driven stratification in the PICU.
Main Methods:
- Application of the k-means clustering algorithm to a cohort of 90 pediatric patients admitted to the PICU.
- Utilized physiological, analytical variables, and device requirements within the first 24 hours of admission for classification.
- Compared clustering results with discharge diagnoses and patient outcomes.
Main Results:
- The k-means algorithm identified three distinct patient groups based on initial data.
- Significant differences were observed in both used and unused variables, discharge diagnoses, and outcomes across the identified groups.
- Clustering classification showed a stronger association with mortality (p < 0.04) than traditional methods (p = 0.16) and equal association with length of stay (LOS) (p = 0.01).
Conclusions:
- Unsupervised learning algorithms, such as k-means, offer a more outcome-correlated method for classifying pediatric shock patients compared to traditional approaches.
- This data-driven stratification can lead to more appropriate and effective patient treatment in the PICU.
- The study supports the integration of machine learning for improved patient classification and management in critical care settings.
