Automated remote decision-making algorithm as a primary triage system using machine learning techniques.

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Emergency Medicine

Background:

• Mass casualty incidents (MCIs) necessitate efficient patient triage due to limited resources and personnel.
• Current triage methods can be subjective and challenging in remote or resource-scarce environments.

Purpose of the Study:

• To develop and evaluate an automated, remote decision-making algorithm for patient triage in MCIs.
• To categorize patient emergency levels using clinical parameters measurable by wearable devices.

Main Methods:

• Utilized the National Trauma Data Bank dataset for algorithm development.
• Applied machine learning models: logistic regression, random forest, and deep neural network (DNN).
• Evaluated algorithm performance using macro-averaged f1 score, mean absolute error (MMAE), and area under the receiver operating characteristic curve (AUC).

Main Results:

• Deep neural network (DNN) and random forest models achieved high performance, with f1 scores of 0.784 and 0.783, respectively.
• Area under the curve (AUC) values for DNN and random forest were 0.883 and 0.882, indicating strong discriminative ability.
• Logistic regression showed a lower f1 score of 0.673 and AUC of 0.844.

Conclusions:

• The developed algorithm presents a viable, automated approach for remote patient triage in MCI settings.
• The algorithm can aid in prioritizing patient transfer and optimizing resource redistribution during mass casualty events.