Cluster analysis of MELD and EuroSCORE for predicting operative mortality after concomitant tricuspid valve surgery

Objective Patients requiring concomitant tricuspid surgery represent a heterogeneous cohort with significant comorbidities and varying degrees of organ and right ventricular dysfunction. However, surgeons can rely on little beyond intuition and experience when discussing operative risks. The objective of the study is to assess how Model for End-stage Liver Disease (MELD) score complements EuroSCORE II (ESII) in risk assessment. Methods Single-centre retrospective cohort study of 369 consecutive patients who underwent concomitant tricuspid valve surgery from 2011 to 2020. Multivariate analysis of factors affecting operative mortality was performed, producing two multivariate risk prediction models – one consisting of ESII components and the other consisting of both ESII components and MELD. The models were compared by measuring c-statistic, using the Hanley-McNeil method. This was further evaluated with category-free net reclassification improvement index (NRI). K-means clustering was performed using MELD and ESII values after scalar transformation as independent variables and operative mortality as the dependent variable. Results MELD is an independent predictor of operative mortality, with an adjusted odds ratio of 1.286 per point. Inclusion of MELD improves the discriminatory power of ESII for operative mortality, with a difference in AUC of 0.128 [0.0341 – 0.222] (p = 0.0076). The NRI of incorporating MELD with ESII was 0.959 [0.515 – 1.392], indicating significant improvement in risk reclassification. Cluster analysis identified a unique cohort of patients with intermediate-to-high MELD, not previously identified with ESII alone, who experienced high operative mortality. Conclusions MELD score as a quantifier of hepatorenal dysfunction complements EuroSCORE II in predicting operative mortality after tricuspid valve surgery.