Development and validation of a risk model for intracardiac thrombosis in patients with dilated cardiomyopathy: a retrospective study

Figures

• 

  Figure 1. The flow chart of study population enrollment.

• 

  Figure 2. Use several methods for variable screening. (A1) The abscissa represents the numbers of variables inclusion, and the ordinate represents the value of BIC; When the number of variables is 4 or 5, the minimum BIC is − 48.4. (A2) Specific variables included in the model with minimum BIC. (B1) The abscissa represents the numbers of variables inclusion, and the ordinate represents the value of adjusted R-square; when the number of variables is 8, the maximum adjusted R-square is 0.143. (B2) Specific variables included in the model with maximum adjusted R-square. (C1) LASSO coefficient profiles of the features. A coefficient profile plot was produced against the log (lambda) sequence. (C2) Optimal parameter (lambda) selection in the LASSO model used fivefold cross-validation via minimum criteria40. The partial likelihood deviance (binomial deviance) curve was plotted versus log (lambda). Dotted vertical lines were drawn at the optimal values by using the minimum criteria and the 1 SE of the minimum criteria (the 1-SE criteria). BIC Bayesian Information Criterion, LASSO least absolute shrinkage and selection operator, SE standard error.

• 

  Figure 3. Baseline model selection (comparison between candidate models). (A1) Comparison of ROC curves between model BIC and model S/L. (A2) Comparison of NRI (Categorical, threshold = 0.163) between model BIC and model S/L. (A3) Comparison of NRI (Continuous) between model BIC and model S/L. (B1) Comparison of ROC curves between model S/L and model R2. (B2) Comparison of NRI (Categorical, threshold = 0.174) between model S/L and model R2. (B3) Comparison of NRI (Continuous) between model S/L and model R2. ROC receiver operating characteristic curve, NRI net reclassification improvement.

• 

  Figure 4. Comparison between the baseline model and the model S/L in different subgroups. In the left ventricular thrombosis subgroup: (A1) Comparison of ROC curves between the model S/L-LVT and the baseline model. (A2) Comparison of NRI (Categorical, threshold = 0.100) between the model S/L-LVT and the baseline model. (A3) Comparison of NRI (Continuous) between the model S/L-LVT and the baseline model. In the left atrial thrombosis subgroup: (B1) Comparison of ROC curves between the model S/L-LAT and the baseline model. (B2) Comparison of NRI (Categorical, threshold = 0.021) between the model S/L-LAT and the baseline model. (B3) Comparison of NRI (Continuous) between the model S/L-LAT and the baseline model. ROC receiver operating characteristic curve, NRI net reclassification improvement.

• 

  Figure 5. The final model is presented via nomogram. (a) Developed intracardiac thrombosis nomogram. (b) Example nomogram for risk of intracardiac thrombosis in a NIDCM patient.

• 

  Figure 6. The performance of the final model. (a) ROC curve for the final model. (b) Calibration curve for the final model. The x-axis represents the estimated intracardiac thrombosis risk. The y-axis represents the actually observed probability of intracardiac thrombosis. The diagonal dotted line represents a perfect estimation by an ideal model. The pink dotted line and the blue solid line represent the performance of the model before and after calibration, respectively. The closer to the diagonal, the better the calibration of the model. (c) Decision curve analysis for the final model. The y-axis measures the standardized net benefit. The blue line represents the intracardiac thrombosis risk model and its 95% CI. The thin solid line represents the assumption that all patients occur intracardiac thrombosis (All patients undergo intervention). The thick solid line represents the assumption that no patients occur intracardiac thrombosis (no patients undergo intervention). The decision curve shows that if the threshold probability is in the range of 2% to 78%, using this model to estimate intracardiac thrombosis risk adds more benefit than the intervention-none scheme or…

• 

  Figure 7. Internal validation of the final model. (A1) ROC curve for the model (30% of the study population). (A2) Calibration curve for the model (30% of the study population). (B1) ROC curve for the model (70% of the study population). (B2) Calibration curve for the model (70% of the study population).