Using surgical risk scores in nonsurgically treated infective endocarditis patients

The accuracy of surgical scores in predicting in-hospital mortality for nonsurgically treated patients with infective endocarditis (IE) has not yet been explored.

Patients with definite IE who did not undergo valve surgery were selected from the database of seven French administrative areas (Association pour l'Étude et la Prévention de l'Endocardite Infectieuse [AEPEI] Registry, 2008). The patients were scored using (a) six systems specifically devised to predict in-hospital mortality after surgery for IE, (b) three commonly used risk scores for heart surgery, and (c) a risk score for predicting six-month mortality in IE after either surgery or medical therapy. Calibration (Hosmer-Lemeshow test) and discriminatory power (receiver operating characteristic [ROC] analysis) were assessed for each score. Areas under ROC curves were compared one-to-one (Hanley-McNeil method).

A total of 192 patients (mean age, 65.2±15.2 years) were considered for analysis. There were 38 (19.8%) in-hospital deaths. Age >70 years (p=0.001), Staphylococcus aureus as causal agent (p=0.05), and severe sepsis (p=0.027) were independent predictors of in-hospital mortality. Despite many differences in the number and type of variables, all but two of the investigated scores showed good calibration (p>0.66). However, discriminatory power was satisfactory (area under ROC curve >0.70) only for three of the scores specific for IE and two of the scores used to predict mortality after cardiac surgery.

Among the 10 surgical scores evaluated in this study, five could be adopted to predict in-hospital mortality even for IE patients receiving medical treatment only.