Abstract OBJECTIVES This observational study aimed to evaluate Association pour l’Etude et la Prévention de l’Endocardite Infectieuse (AEPEI) surgery score predictive performance in comparison to general (EuroSCORE I, II) and specific (De Feo, PALSUSE) surgical risk scores for infective endocarditis (IE). METHODS The study included patients who underwent surgery for IE during the acute phase at Bichat University Hospital (Paris, France) between 1 January 2006 and 31 December 2016 and at Nancy University Hospital (Nancy, France) between 1 January 2009 and 31 December 2019. Patients with IE complicating percutaneous aortic valve implantations or implantable intra-cardiac devices were excluded. Discrimination and calibration were assessed using receiver operating characteristic curve analysis, calibration curves and the Hosmer–Lemeshow test. RESULTS In-hospital mortality rates were 18% at Bichat and 16% at Nancy. Discrimination was high for all risk scores at Bichat (area under the receiver operating characteristic curve = 0.77 for EuroSCORE I, 0.78 for EuroSCORE II, 0.76 for De Feo score, 0.72 for PALSUSE and 0.73 for AEPEI with 95% confidence interval ranging from 0.67 to 0.83), but lower at Nancy (0.56 for EuroSCORE I, 0.65 for EuroSCORE II, 0.63 for De Feo score, 0.67 for PALSUSE and 0.66 for AEPEI score with 95% confidence interval ranging from 0.47 to 0.75). With Brier score, all scores were adequately calibrated in both populations between 0.129 (De Feo) and 0.135 (PALSUSE) for Bichat and between 0.128 (De Feo) and 0.135 (EuroSCORE I) for Nancy. With the Hosmer–Lemeshow test, the AEPEI score exhibited the best calibration (observed/predicted ratio 1.058 in Bichat, 1.087 in Nancy). CONCLUSIONS This surgical score external validation in 2 large independent populations demonstrated that the AEPEI surgical score had the best predictive performance compared to other prognosis scores. It could be helpful in clinical practice to assist the endocarditis team in decision-making.