Decision Support for Tactical Combat Casualty Care Using Machine Learning to Detect Shock

Area of Science:

• Medical Informatics
• Machine Learning in Healthcare
• Prehospital Care

Background:

• Complex Prolonged Field Care in austere settings requires enhanced support for inexperienced providers.
• Effective tools are needed to assist in treating patients during tactical combat casualty care.
• Accurate detection and differentiation of shock are critical in prehospital settings.

Purpose of the Study:

• To develop a phone-/tablet-based decision support system for prehospital tactical combat casualty care.
• To utilize machine learning for detecting and differentiating shock manifestations.
• To augment medic decision-making with salient patient data.

Main Methods:

• Software interface developed through literature review, rapid prototyping, and expert design reviews.
• Machine learning model trained on MIMIC and de-identified Mayo Clinic ICU data.
• Logistic regression model selected as the best performing algorithm.

Main Results:

• 17 military medical subject matter experts provided design requirements.
• The machine learning algorithm predicted shock onset 90 minutes prior to occurrence.
• The model achieved over 75% accuracy in predicting shock in the test dataset.

Conclusions:

• The Trauma Triage, Treatment, and Training Decision Support system will enhance medic decision-making.
• The system aids in diagnosing multiple types of shock.
• Remotely trained, field-deployed machine learning models support enhanced care.