Development and preliminary evaluation of a nomogram for predicting 30-day mortality after surgery for infective endocarditis: a single-center retrospective prediction-model study

Infective endocarditis (IE) remains associated with substantial early postoperative mortality. This study aimed to develop an exploratory, parsimonious nomogram for estimating 30-day all-cause mortality after IE surgery and describe its apparent performance, including exploratory comparisons with EuroSCORE II and SYSUPMIE, in the development cohort. Consecutive adults who underwent surgery for modified Duke criteria-defined and surgically confirmed IE at a tertiary hospital in China between June 2018 and April 2026 were retrospectively screened. The outcome was all-cause death within 30 days after surgery. Patients lacking essential preoperative test results or ascertainable 30-day outcomes were excluded. Variables with more than 30% missingness were omitted; remaining missing values were handled using single imputation by chained equations with five iterations, resulting in one completed dataset for subsequent analyses. Least absolute shrinkage and selection operator logistic regression reduced candidate predictors, followed by backward stepwise multivariable logistic regression. Apparent performance was assessed using discrimination, calibration, decision curve analysis, and reclassification metrics. Among 187 patients, 16 died within 30 days. The final model included sex, stroke/transient ischemic attack, ventricular tachycardia/ventricular fibrillation, estimated glomerular filtration rate, C-reactive protein-to-albumin ratio, and hemoglobin. The apparent area under the curve (AUC) was 0.927 (95% confidence interval, 0.857–0.996). This estimate and the apparent calibration findings were derived from the development cohort and may have been optimistic. Within the post hoc exploratory threshold-probability range of 5%–20%, the nomogram’s net-benefit curve was higher than the treat-all and treat-none curves in the development cohort; this finding was descriptive and potentially affected by overfitting. Reclassification comparisons with EuroSCORE II and SYSUPMIE were exploratory and did not establish superiority. This exploratory nomogram was developed in a single-center cohort with only 16 deaths. Its predictor structure and apparent performance remain uncertain and may have been optimistic. The model is hypothesis-generating and should not be used for clinical decisions before adequate internal evaluation and independent external validation.