Prediction model of postoperative mortality for congenital heart disease: evidence from two large-scale cohorts

Fig. 1. Forest plot of independent predictors of postoperative mortality. Multivariable logistic regression identified six independent predictors. Odds ratio (OR) with 95% confidence intervals (CI) are shown.

Fig. 2. Clinical nomogram for postoperative mortality risk estimation. The tool integrates six routinely available predictors (preoperative creatinine, total protein, glucose, triglycerides, and postoperative lactate, cystatin C). The summed points from the top scale correspond to an individualized risk probability on the bottom scale, facilitating bedside assessment.

Fig. 3. Receiver operating characteristics (ROC) curves for the prediction model. (A) the ROC curve of the model in the training cohort, showing an area under the curve (AUC) of 0.952 with its 95% confidence intervals (95% CI: 0.932–0.972). (B) The ROC curve in the external validation cohort, showing an AUC of 0.761 with its 95% confidence intervals (95% CI: 0.621–0.901). The solid diagonal line represents the reference line of no discrimination (AUC = 0.5).

Fig. 4. Calibration curves assessing the agreement between predicted and observed mortality risk. (A) Calibration plot for the training cohort. (B) Calibration plot for the external validation cohort. The diagonal dashed line represents perfect calibration (ideal fit).

Fig. 5. Decision curve analysis (DCA) for the clinical utility of the nomogram. The curve assesses the net of using the nomogram to predict postoperative mortality across a range of threshold probabilities. The solid red line represents the net benefit of the prediction model. The solid grey line represents the strategy of assuming all patients die (“treat all”), and the solid black line represents the strategy of assuming no patients die (“treat none”).