Predicting mortality in nonsurgical patients before cannulation for veno‐arterial extracorporeal life support: Development and validation of the LACT‐8 score

Authors:

Abstract
Objectives: We sought to derive and validate a model to predict inpatient mortality after veno‐arterial extracorporeal life support (VA‐ECLS) based on readily available, precannulation clinical data.Background: Refractory cardiogenic shock supported by VA‐ECLS is associated with high morbidity and mortality.Methods: VA‐ECLS cases at our institution from January 2014 through July 2019 were retrospectively reviewed. Exclusion criteria were cannulation: (1) at another institution; (2) for primary surgical indication; or (3) for extracorporeal cardiopulmonary resuscitation. Multivariable logistic regression compared those with and without inpatient mortality. Multiple imputation was performed and optimism‐adjusted area under the curve (oAUC) values were computed.Results: VA‐ECLS cases from August 2019 through November 2020 were identified as a validation cohort. In the derivation cohort (n = 135), the final model included Lactate (mmol/L), hemoglobin (g/dl; Anemia), Coma (Glasgow Coma Scale [GCS] < 8) and resusciTATEd cardiac arrest (LACTATE score; oAUC = 0.760). In the validation cohort (n = 30, LACTATE showed similar predictability [AUC = 0.710]). A simplified (LACT‐8) score was derived by dichotomizing lactate (>8) and hemoglobin (<8) and summing together the number of components for each patient. LACT‐8 performed similarly (derivation, oAUC = 0.724; validation, AUC = 0.725). In the derivation cohort, both scores outperformed SAVE (oAUC = 0.568) and SOFA (oAUC = 0.699) scores. A LACT‐8 ≥ 3 had a specificity for mortality of 97.9% and 92.9%, in the derivation and validation cohorts, respectively.Conclusions: The LACT‐8 score can predict inpatient mortality prior to before cannulation for VA‐ECLS. LACT‐8 can be implemented utilizing clinical data without the need for an online calculator.
Funding Information
  • National Center for Advancing Translational Sciences (UL1TR001998)

This publication has 34 references indexed in Scilit: