A data-driven artificial intelligence model for remote triage in the prehospital environment.

Area of Science:

• Medical Technology
• Artificial Intelligence in Healthcare
• Trauma Care

Background:

• Triage time is critical for survival in mass casualty incidents, but limited medical personnel restrict efficiency.
• Current triage methods are manpower-dependent, limiting scalability and speed.
• Developing automated, rapid triage solutions is essential for improving patient outcomes.

Purpose of the Study:

• To develop a machine learning classification model for survival prediction to enable rapid, personnel-independent triage.
• To design a consciousness index, monitored via wearable devices, as a surrogate for manpower in triage.
• To enhance classification accuracy for survival prediction using advanced machine learning algorithms.

Main Methods:

• A consciousness index was designed and validated using logistic regression with vital signs from wearable devices.
• Machine learning algorithms (logistic regression, random forest, deep neural network) were employed to build survival prediction models.
• A large dataset of 460,865 trauma cases (blunt and penetrating injuries) from the national trauma databank was utilized.

Main Results:

• The consciousness index demonstrated high efficiency in remote monitoring via wearable devices.
• Machine learning models achieved high accuracy in survival prediction, comparable to existing injury severity scoring systems.
• The deep neural network model achieved the highest Area Under the Curve (AUC) of 0.89 (95% CI = 0.882 to 0.890).

Conclusions:

• Remote triage using wearable devices and a machine learning-based survival prediction model is feasible.
• The developed model can significantly reduce the time required for triage in mass casualty incidents.
• This technology has the potential to improve patient survival rates by enabling faster and more precise initial assessments.