External and Temporal Validation 10 Years after the Development of the First Latin-American Risk Stratification Score for Cardiac Surgery (ArgenSCORE)

Background During the last decades, several risk assessment modelshave been applied to predict the risk of mortality after cardiacsurgery; however, none of them have been developed in Latin American populations. These models have inferiorperformance when applied to patient groups other than theones on whom they were developed.ObjectivesTo perform external and temporal validation of a local riskscore for cardiac surgery [Argentinean System for CardiacOperative Risk Evaluation (ArgenSCORE)] and compare itto the EuroSCORE.Material and MethodsA total of 5268 consecutive adult patients undergoing cardiacsurgery were included from June 1994 to December 2009. Therisk model was developed through logistic regression on thedata of 2903 patients who underwent cardiac surgery betweenJune 1994 and December 1999 at a center. Prospective internalvalidation was performed on 708 patients between January2000 and June 2001. External and temporal validation of therecalibrated model were performed between February 2000and December 2009, evaluating model discrimination andcalibration in patients operated on at four centers differentfrom the one where the score had been originally developed.The method was also compared to the EuroSCORE.ResultsThe external validation was performed on 1657 patients,mean age was 62.8±13.3 years and global mortality was4.58%. The ArgenSCORE showed both good discriminatorypower with an area under the ROC curve of 0.80 and predictivecapacity for risk assessment in all patients (observedmortality 4.58% vs. expected mortality 4.54%; p=0.842). TheEuroSCORE showed good discriminatory power (area underthe ROC curve of 0.79) but overestimated the risk (observedmortality 4.58% vs. expected mortality 5.23%; p <0.0001).ConclusionsThe ArgenSCORE showed an adequate capacity to predictin-hospital mortality in cardiac surgery 10 years after beingdeveloped. The score can be applied to populations with similargeographic characteristics, showing a better performance comparedto an established international risk stratification model.