Could machine learning algorithms help us predict massive bleeding at prehospital level?

Area of Science:

• Emergency Medicine
• Data Science
• Trauma Surgery

Background:

• Massive hemorrhage (MH) is a critical concern in severe traumatic injury (STI).
• Accurate prediction of MH is vital for timely and effective interventions.
• Traditional prediction scales (TPS) have limitations in predicting MH in STI.

Purpose of the Study:

• To compare the predictive performance of machine learning algorithms (MLAs) against TPS for MH in STI patients.
• To evaluate the utility of MLAs in out-of-hospital settings for trauma care.

Main Methods:

• Retrospective analysis of 473 STI patients with prehospital data.
• Development and validation of four MLAs (Random Forest, SVM, GBM, NN) using 80% training and 20% validation data.
• Evaluation of predictive power using Receiver Operating Characteristic (ROC) curves and variable importance via Shapley values.

Main Results:

• MLAs achieved high predictive accuracy with ROC values exceeding 0.85, with medians near 0.98.
• No significant differences were found between the performance of the tested MLAs.
• Key predictive variables identified by MLAs included hemodynamic status, resuscitation factors, and neurological impairment.

Conclusions:

• Machine learning algorithms demonstrate superior predictive ability for massive hemorrhage compared to traditional scales in severe traumatic injury.
• MLAs offer a promising advancement for improving prehospital care and patient outcomes in trauma.