HEROES V-V-HEmorRhagic cOmplications in Veno-Venous Extracorporeal life Support-Development and internal validation of multivariable prediction model in adult patients.

ABSTRACT

BACKGROUND: During extracorporeal life support (ECLS), bleeding is one of the most frequent complications, associated with high morbidity and increased mortality, despite continuous improvements in devices and patient care. Risk factors for bleeding complications in veno-venous (V-V) ECLS applied for respiratory support have been poorly investigated. We aim to develop and internally validate a prediction model to calculate the risk for bleeding complications in adult patients receiving V-V ECLS support. METHODS: Data from adult patients reported to the extracorporeal life support organization (ELSO) registry between the years 2010 and 2020 were analyzed. The primary outcome was bleeding complications recorded during V-V ECLS. Multivariable logistic regression with backward stepwise elimination was used to develop the predictive model. The performance of the model was tested by discriminative ability and calibration with receiver operating characteristic curves and visual inspection of the calibration plot. RESULTS: In total, 18 658 adult patients were included, of which 3 933 (21.1%) developed bleeding complications. The prediction model showed a prediction of bleeding complications with an AUC of 0.63. Pre-ECLS arrest, surgical cannulation, lactate, pO2 , HCO3 , ventilation rate, mean airway pressure, pre-ECLS cardiopulmonary bypass or renal replacement therapy, pre-ECLS surgical interventions, and different types of diagnosis were included in the prediction model. CONCLUSIONS: The model is based on the largest cohort of V-V ECLS patients and reveals the most favorable predictive value addressing bleeding events given the predictors that are feasible and when compared to the current literature. This model will help identify patients at risk of bleeding complications, and decision making in terms of anticoagulation and hemostatic management.