Reassessing surgical risk models in infective endocarditis: are they still valid beyond surgery?

Background Infective endocarditis (IE) remains associated with high in-hospital mortality, despite advances in diagnostic and therapeutic strategies. Several risk scores originally developed for cardiac surgery populations, such as EuroSCORE I, EuroSCORE II, AEPEI II and PALSUSE, are frequently used to estimate early postoperative risk. However, their validity in predicting overall in-hospital mortality, including patients managed medically, remains uncertain. Purpose Assess the discriminative performance of these scores in predicting in-hospital mortality among all IE patients, irrespective of surgical indication or treatment strategy. Methods Retrospective observational study including all patients hospitalized for definitive IE between 2017 and 2022 at a secondary cardiac center according to the modified Duke criteria. The dependent variable was in-hospital mortality (death during index admission). Four pre-established risk scores—EuroSCORE I, EuroSCORE II, PALSUSE, and AEPEI modified—were evaluated individually as predictors in univariate binary logistic regression models. For each model, we calculated the odds ratio (OR) with 95% confidence interval (CI). Predictive discrimination was assessed using receiver operating characteristic (ROC) curve analysis, with computation of area under the curve (AUC), sensitivity, specificity, and overall accuracy. Results From a total of 89 patients, 69% were male and the mean age was 69.9 ±12.6 years. From the latter, 63 had surgical indication.In-hospital mortality occurred in 30,3% (n=27) of patients. Among the four tested models, only the AEPEI modified score demonstrated significant association with mortality and acceptable discriminative capacity. Model comparison confirmed a statistically significant improvement in model fit for AEPEI (Δχ² = 7.807, p = 0.008), while no significant improvement was observed for EuroSCORE I, EuroSCORE II, or PALSUSE (all p > 0.10). Discrimination indices confirmed that only AEPEI achieved an AUC ≥ 0.70, a commonly accepted threshold for acceptable predictive performance. Conclusions In this single-center cohort of 89 patients with IE—managed either conservatively or surgically—only the AEPEI modified score demonstrated significant predictive value for in-hospital mortality, with an AUC of 0.70 (p=0.014) and balanced sensitivity (0.69) and specificity (0.84). Traditional surgical risk models (EuroSCORE I, EuroSCORE II, and PALSUSE) showed poor discriminative capacity in this mixed population regarding general all cause in-hospital mortality and were not significantly associated with mortality outcomes. These results may suggest that the AEPEI modified score can retain prognostic validity beyond surgical candidates and could serve as a valuable, easily applicable tool for early mortality risk stratification in the overall population of patients with IE.