HEROES V-A: HEmoRrhagic cOmplications in veno-arterial Extracorporeal life Support: Development and internal validation of a multivariable prediction model in adult patients

Abstract

Background: Risk factors for bleeding complications during extracorporeal life support (ECLS) indicated for cardiac support remain poorly investigated. The aim is to develop and internally validate a prediction model to calculate the risk for bleeding complications in adult patients receiving veno-arterial (V-A) ECLS.

Methods: Data of the Extracorporeal Life Support Organization registry of adult patients undergoing V-A ECLS between 2010 and 2020 were analyzed. The primary outcome was bleeding complications recorded during V-A ECLS. Multivariable logistic regression with backward stepwise elimination was used to develop the prediction model. Performance of the model was tested by discriminative ability and calibration with receiver operator characteristic, area under the curve, and visual inspection of the calibration plot. Internal validation was performed to detect overfitting of the model.

Results: In total 28 767 adult patients were included, of which 29.0% developed bleeding complications. Sex, body mass index, surgical cannulation, pre-ECLS respiratory and hemodynamic variables, pre-ECLS support and interventions, and different type of diagnosis were included in the prediction model. This prediction model showed a predictive capability with an AUC of 0.66.

Conclusion: The model is based on the largest cohort of V-A ECLS patients and is the best available predictive model for bleeding events given the predictors that are available in V-A ECLS compared to current literature. The model can help in identifying patients at high risk for bleeding complications and will help in developing further research and decision-making in terms of anticoagulation management. External validation is warranted to extrapolate this model in the clinical setting.

Keywords: anticoagulation; bleeding complications; hemorrhagic complications; prediction model; registry data; veno-arterial extracorporeal life support.