A novel model for predicting a composite outcome of major complications after valve surgery

Figures

• 

  Figure 1. Predictor selection using least absolute shrinkage and selection operator (LASSO) regression. (A) Optimization of the penalty tuning parameter λ using 10-fold cross-validation with the minimum criterion. The number of predictors is indicated at the top of the plot. Dotted vertical lines indicate the minimum criterion and one standard error away from it. An optimal value of log (λ) = −4.82 was selected. (B) Coefficient profile plot of the predictors. The number of predictors is indicated at the top of the plot. The ordinate is the coefficient value. Each curve in the figure represents the trajectory of the change of the coefficient of each independent variable, and the coefficient of variable is eventually compressed to 0 as the λ parameter increases. Dotted vertical lines were plotted at ideal values, based on the same criteria as in (A). Six predictors with non-zero coefficients were selected.

• 

  Figure 2. Development and performance assessment of a nomogram to predict composite major complications after valve surgery. (A) The nomogram. NYHA, New York Heart Association; BUN, blood urea nitrogen; RBC, red blood cell. (B) Receiver operating characteristic curves for the nomogram against the development cohort (red) or validation cohort (blue). The area under the curve (AUC) together with the 95% CI: 0.735 (95% CI, 0.686–0.784) in the development cohort and 0.761 (95% CI, 0.694–0.828) in the validation cohort.

• 

  Figure 3. Calibration curve of development and validation cohorts. Calibration curves showing agreement between the nomogram's predicted probability of a composite of major complications and actual probability of the composite in the (A) development or (B) validation cohort. The dashed diagonal line in the figure indicates when predicted and observed probabilities match perfectly. The red and green fitted curves were apparent and bias-corrected calibration curves respectively. The corresponding Brier scores are shown at the upper left of each plot.

• 

  Figure 4. Decision curves of development and validation cohorts. Decision curve analysis assessing the clinical utility of the nomogram for the (A) development or (B) validation cohort as a function of probability thresholds. Dotted lines represent the states when no patient (black) or all patients (gray) experience the composite outcome of major complications. Clinical benefit is indicated when the orange line lies to the right of the black dotted line and above the gray dotted line.