Toward a hemorrhagic trauma severity score: fusing five physiological biomarkers

Area of Science:

• Biomedical Engineering
• Computational Biology
• Trauma Medicine

Background:

• Hemorrhagic trauma presents a critical challenge in patient management.
• Current assessment methods may not fully capture the dynamic physiological state of trauma patients.

Purpose of the Study:

• To introduce the Hemorrhage Intensive Severity and Survivability (HISS) score.
• To develop a patient-specific attribute for hemorrhagic trauma using multi-biomarker data fusion.
• To evaluate machine learning algorithms for predicting the HISS score.

Main Methods:

• Generated 100 synthetic patient datasets (Sensible Fictitious Rationalized Patient - SFRP).
• Assigned HISS scores by five clinical experts, stratifying trauma criticality (low to severe).
• Evaluated four classifier algorithms: SVM-L, EBDT, ANN:BR, and PRBF for HISS score prediction.

Main Results:

• All evaluated algorithms achieved high testing accuracies (0.91-0.93) in predicting HISS scores.
• No statistically significant difference was observed between the performance of the tested algorithms (p > 0.05).
• Target accuracies of 0.99 and 0.999 are achievable with increased data and expert scores.

Conclusions:

• Data-driven HISS score predictions can aid in patient stratification and triage.
• Integration with point-of-care multi-analyte biosensors can enhance continuous monitoring of trauma patients.
• The HISS score offers a valuable tool for objective assessment of trauma severity.