Related Experiment Video
Updated: Jun 26, 2025
03:40
Point-of-Care Ultrasound for Peripheral Veno-Arterial Extracorporeal Membrane Oxygenation Without Left Ventricular Venting
Published on: January 17, 2025
285
Machine Learning Differentiates Extracorporeal Membrane Oxygenation Mortality Risk Profiles Among Trauma Patients.
Bryan R Campbell1, Alexandra S Rooney1, Andrea Krzyzaniak1
1Division of Trauma and Acute Care Surgery, Department of Surgery, Scripps Mercy Hospital, San Diego, CA, USA.
The American Surgeon
|May 16, 2024
Summary
Machine learning identified three distinct risk profiles for trauma patients on extracorporeal membrane oxygenation (ECMO). These profiles, based on injury severity and patient characteristics, can refine ECMO treatment guidelines and decision-making.
Area of Science:
- Trauma Surgery
- Critical Care Medicine
- Machine Learning Applications
Background:
- Extracorporeal membrane oxygenation (ECMO) is a life-support technology with high resource demands and mortality rates.
- Current guidelines for ECMO use in trauma patients do not adequately identify individuals most likely to benefit.
- There is a need to refine patient selection criteria for ECMO in severe trauma.
Purpose of the Study:
- To identify unique patient risk profiles among trauma patients receiving ECMO.
- To leverage machine learning for a comprehensive analysis of a large trauma patient database.
- To enhance the understanding of factors influencing outcomes in ECMO-treated trauma patients.
Main Methods:
- Utilized the Trauma Quality Improvement Program database (2010-2019) for patients aged 16 years and older.
- Applied machine learning K-median clustering with 101 variables, including injury severity, demographics, comorbidities, and hospital stay.
- Evaluated mortality rates, patient characteristics, and center variations across identified risk profiles.
Main Results:
- Identified 3 distinct patient risk profiles from 1037 ECMO patients (33% mortality).
- Group 1: Young, severely injured patients with thoracic/head injuries and ARDS.
- Group 2: Moderately injured patients with exposure-related injuries.
- Group 3: Older patients with comorbidities and extremity injuries.
Conclusions:
- Machine learning effectively identifies unique mortality risk profiles in trauma patients undergoing ECMO.
- These findings can inform and improve existing ECMO treatment guidelines.
- The identified profiles support better clinical decision-making and institutional criteria for ECMO utilization.
