Early Mortality in Reoperative Cardiac Surgery: Determinants, Prediction, and Calibration.

Background/Objectives: Reoperative (redo) cardiac surgery carries elevated risk, yet determinants and mechanisms of early death in valve- and endocarditis-predominant cohorts are poorly described. We aimed to identify preoperative determinants of early mortality, derive a clinically usable risk model, examine the mechanisms of death, and evaluate EuroSCORE II. Methods: We analysed all redo cardiac operations through repeat sternotomy at a single tertiary centre (December 2010-December 2025); the cohort comprised 821 patients with verified in-hospital mortality status. Independent preoperative predictors were identified by multivariable logistic regression. Discrimination and calibration of the model and of EuroSCORE II were compared, and procedural complexity, graded low-cardiac-output severity, postoperative morbidity, and failure-to-rescue were examined. Results: The early mortality rate was 22.8% (187/821). The independent preoperative predictors were chronic kidney disease (adjusted odds ratio 2.25), age ≥70 years (2.32), active endocarditis (1.89), advanced functional class (1.78), non-elective status (1.69), and systolic pulmonary artery pressure (1.15 per 10 mmHg); the model discriminated and calibrated well (concordance statistic 0.755). EuroSCORE II discriminated similarly but systematically under-predicted mortality (observed-to-expected ratio 1.46; worst 2.06 at intermediate risk). Mortality rose with procedural complexity (17.3% to 37.3%) and graded haemodynamic support (3.4% to 79.1%). Failure-to-rescue was 38.5%, compared with 5.8% mortality in patients without a major complication. A standard-risk subgroup (n = 209) had 12.4% mortality. Conclusions: Early mortality after redo cardiac surgery is driven by identifiable preoperative comorbidity, procedural complexity, and low-cardiac-output physiology, and is concentrated in failure-to-rescue. A parsimonious preoperative model predicted mortality well; EuroSCORE II, although discriminating, systematically under-predicted risk, supporting redo-specific assessment and recalibration.