Field Foresight-Predicting the Need for Massive Transfusion, ICU Utilization, and Mechanical Ventilation for...

Area of Science:

• Trauma care and emergency medicine
• Artificial intelligence in healthcare
• Clinical decision support systems

Background:

• Accurate and rapid patient triage is crucial for effective resource allocation in trauma care.
• Existing triage algorithms often rely solely on vital signs, potentially missing critical information from external injuries.
• The Field AI Triage (FAIT) tool was developed to enhance triage accuracy by incorporating injury details.

Purpose of the Study:

• To develop and validate an AI tool (FAIT) for predicting resource needs in trauma patients.
• To evaluate if incorporating external injury information improves triage performance compared to vital signs alone.
• To assess FAIT's effectiveness in both civilian and military trauma populations.

Main Methods:

• Leveraged large-scale civilian (TQIP) and military (DoDTR) trauma datasets.
• Developed a logistic regression model incorporating vital signs and injury characteristics (mechanism, location).
• Validated the model using 10-fold cross-validation, with ROC AUC as the primary performance metric.

Main Results:

• FAIT demonstrated significantly improved prediction accuracy across multiple outcomes (massive transfusion, mechanical ventilation, ICU admission) for both civilian and military patients.
• For civilian patients, ROC AUC improved from 0.70 to 0.89 for massive transfusion and 0.65 to 0.90 for mechanical ventilation.
• For military patients, ROC AUC improved from 0.67 to 0.88 for massive transfusion and 0.76 to 0.90 for mechanical ventilation.

Conclusions:

• The FAIT tool effectively supports early assessment at the point of injury.
• FAIT shows significant capability in forecasting resource allocation for diverse trauma mechanisms in both civilian and military settings.
• Integrating external injury data with vital signs enhances AI-driven triage performance.