Development and Internal Validation of a Clinical Prediction Model and Nomogram for 30-Day Mortality After Post-Myocardial Infarction Ventricular Septal Rupture

<title>Abstract</title> <p>Background: Post-myocardial infarction ventricular septal rupture (VSR) is associated with early mortality frequently exceeding 40% in contemporary series. Although recent nomograms address 30-day mortality across medical and surgical pathways, existing tools remain limited by moderate sample size and incomplete resampling-based validation. We aimed to develop and internally validate a 30-day mortality prediction model and nomogram in the largest reported post-MI VSR cohort to date. Methods: We analyzed 469 consecutive patients screened at Fuwai Hospital (September 2001–December 2025); 384 met inclusion criteria after excluding congenital ventricular septal defect (n = 1) and myocardial infarction duration exceeding one month (n = 84). The primary outcome was 30-day mortality (147 events; 38.3%). From 42 pre-specified candidate predictors, least absolute shrinkage and selection operator (LASSO) regression with 10-fold cross-validation selected variables for multivariable logistic regression with multiple imputation (MICE; 20 datasets). Discrimination, calibration, decision curve analysis, bootstrap internal validation (1,000 resamples), subgroup analyses, and four sensitivity analyses were performed per TRIPOD Type 1b guidelines. Results: LASSO retained 32 predictors at λmin; 22 were stably selected (≥70% inclusion across 1,000 bootstrap resamples). Twelve predictors were independently associated with 30-day mortality (P &lt; 0.05), including cardiogenic shock, mechanical ventilation, white blood cell count, acute kidney injury, heart rate, and age. The model achieved excellent discrimination (C-statistic 0.965; 95% confidence interval [CI], 0.945–0.983; optimism-corrected 0.944) with acceptable calibration (slope 0.971; Hosmer–Lemeshow P = 0.171). At the Youden-optimal threshold (0.606), sensitivity was 85.7% and specificity 97.5%. Decision curve analysis confirmed positive net benefit from 3% to 98% threshold probabilities. A bedside nomogram was constructed. Conclusions: In the largest 30-day mortality prediction study for post-MI VSR reported to date (n = 384; 147 events), we developed a TRIPOD-aligned nomogram integrating hemodynamic, inflammatory, and organ-dysfunction markers with excellent discrimination and acceptable calibration. External validation and prospective recalibration are warranted before clinical deployment.</p>