¿Podrían ayudarnos los algoritmos de machine learning en la predicción de hemorragia masiva a nivel prehospitalario? / Could machine learning algorithms help us in predicting massive hemorrhage at the prehospital level?

Objetivo: Comparación de la capacidad predictiva de diferentes algoritmos de machine learning (AML) respecto a escalas tradicionales de predicción de hemorragia masiva en pacientes con enfermedad traumática grave (ETG). Diseño: Sobre una base de datos de una cohorte retrospectiva con variables clínicas prehospitalarias y de resultado de hemorragia masiva se realizó un tratamiento de la base de datos para poder aplicar los AML, obteniéndose un conjunto total de 473 pacientes (80% entrenamiento, 20% validación). Para la modelización se realizó imputación proporcional y validación cruzada. El poder predictivo se evaluó con la métrica ROC y la importancia de las variables mediante los valores Shapley. Ámbito: Atención extrahospitalaria del paciente con ETG. Pacientes: Pacientes con ETG atendidos en el medio extrahospitalario por un servicio médico extrahospitalario desde enero de 2010 hasta diciembre de 2015 y trasladados a un centro de trauma en Madrid. Intervenciones: Ninguna. Variables de interés principales: Obtención y comparación de la métrica ROC de 4 AML: random forest, support vector machine, gradient boosting machine y neural network con los resultados obtenidos con escalas tradicionales de predicción. Resultados: Los diferentes AML alcanzaron valores ROC superiores al 0,85, teniendo medianas cercanas a 0,98. No encontramos diferencias significativas entre los AML. Cada AML ofrece un conjunto de variables diferentes, pero con predominancia de las variables hemodinámicas, de reanimación y de deterioro neurológico. Conclusiones: Los AML podrían superar a las escalas tradicionales de predicción en la predicción de hemorragia masiva. (AU)

Objective: Comparison of the predictive ability of various machine learning algorithms (MLA) versus traditional prediction scales for massive hemorrhage in patients with severe traumatic injury (ETG). Design: On a database of a retrospective cohort with prehospital clinical variables and massive hemorrhage outcome, a treatment of the database was performed to be able to apply the different MLA, obtaining a total set of 473 patients (80% training and 20% validation). For modeling, proportional imputation and cross validation were performed. The predictive power was evaluated with the ROC metric and the importance of the variables using the Shapley values. Setting: Out-of-hospital care of patients with ETG. Participants: Patients with ETG treated out-of-hospital by a prehospital medical service from January 2010 to December 2015 and transferred to a trauma center in Madrid. Interventions: None. Main variables of interest: Obtaining and comparing the ROC curve metric of 4 MLAs: random forest, support vector machine, gradient boosting machine and neural network with the results obtained with traditional prediction scales. Results: The different MLA reached ROC values higher than 0.85, having medians close to 0.98. We found no significant differences between MLAs. Each MLA offers a different set of more important variables with a predominance of hemodynamic, resuscitation variables and neurological impairment. Conclusions: MLA may be helpful in patients with massive hemorrhage by outperforming traditional prediction scales. (AU)

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

OBJECTIVE: Comparison of the predictive ability of various machine learning algorithms (MLA) versus traditional prediction scales (TPS) for massive hemorrhage (MH) in patients with severe traumatic injury (STI). DESIGN: On a database of a retrospective cohort with prehospital clinical variables and MH outcome, a treatment of the database was performed to be able to apply the different AML, obtaining a total set of 473 patients (80% training, 20% validation). For modeling, proportional imputation and cross validation were performed. The predictive power was evaluated with the ROC metric and the importance of the variables using the Shapley values. SETTING: Out-of-hospital care of patients with STI. PARTICIPANTS: Patients with STI treated out-of-hospital by a out-of-hospital medical service from January 2010 to December 2015 and transferred to a trauma center in Madrid. INTERVENTIONS: None. MAIN VARIABLES OF INTEREST: Obtaining and comparing the "Receiver Operating Characteristic curve" (ROC curve) metric of four MLAs: "random forest" (RF), "vector support machine" (SVM), "gradient boosting machine" (GBM) and "neural network" (NN) with the results obtained with TPS. RESULTS: The different AML reached ROC values higher than 0.85, having medians close to 0.98. We found no significant differences between AMLs. Each AML offers a different set of more important variables with a predominance of hemodynamic, resuscitation variables and neurological impairment. CONCLUSIONS: MLA may be helpful in patients with HM by outperforming TPS.