External validation of the SHARPEN score for predicting in-hospital mortality in infective endocarditis

Risk stratification of infective endocarditis (IE) using IE-specific variables remains limited. In this study, we evaluated SHARPEN in an Italian cohort and performed an exploratory paired comparison with baseline Sequential Organ Failure Assessment (SOFA) score.

We performed a retrospective study of consecutive adults with definite IE hospitalized at a tertiary center from 2000 to 2021.Primary endpoint was all-cause in-hospital mortality. Discrimination was assessed by ROC curves, calibration by Brier score, calibration intercept and slope, Hosmer-Lemeshow test, and decile-based plots. The optimal threshold was identified using the Youden index. A prespecified multivariable model including SHARPEN, hemoglobin, INR, and pre-existing heart failure was evaluated. Baseline SOFA was compared with SHARPEN in paired complete-case subset.

245 patients were included. In-hospital mortality occurred in 44 (18%). Non-survivors had higher SHARPEN scores (median 11 vs. 9,p < 0.001). The score showed good discrimination (AUC 0.748;95%CI0.660-0.834) and acceptable calibration (Brier 0.1269; Hosmer-Lemeshow p = 0.458), with an optimal threshold of 11, yielding sensitivity 0.703, specificity 0.737, and NPV 0.918. A simple multivariable model combining SHARPEN with hemoglobin, INR, and chronic heart failure significantly improved discrimination (AUC 0.859; 95%CI0.796-0.921). Baseline SOFA was available in 75 patients (20 deaths); in this selected paired subset, AUCs were 0.759 (95%CI0.625-0.893) for SHARPEN and 0.878 (95%CI0.788-0.968) for SOFA.

The SHARPEN score demonstrated reproducible discrimination for predicting in-hospital mortality in IE in our cohort. The exploratory SOFA comparison did not establish superiority. SHARPEN could serve as an adjunct to multidisciplinary assessment rather than a stand-alone decision tool.