Personalized risk prediction of mortality and rehospitalization for heart failure in patients undergoing mitral valve repair surgery

Figures

• 

  Figure 1. Analysis overview for identifying best-performing risk prediction model.

• 

  Figure 2. Nomogram of mortality. The nomogram was constructed based on the regression coefficient in the model.

• 

  Figure 3. Nomogram of rehospitalization for heart failure. The nomogram was constructed based on the regression coefficient in the model.

• 

  Figure 4. Performance evaluation in terms of mortality prediction model in the derivation and validation cohorts. (A) The area under the receiver operating characteristic curve (AUC) of the mortality prediction model in the derivation cohort was 0.825 [95% confidence interval (CI): 0.764–0.886], which was better than that of Euro score II (p < 0.001). (B) The AUC of the mortality prediction model in the external validation cohort was 0.808 (95% CI: 0.699–0.917), which was also better than that of Euro score II (p < 0.001). (C,D) Calibration plot of the mortality prediction model in the derivation and validation cohort. The dashed diagonal line represents perfect calibration. The x-axis is the predicted probability estimated by the model and the y-axis is the actual probability.

• 

  Figure 5. Performance evaluation in terms of rehospitalization for heart failure prediction model in the derivation and validation cohorts. (A) The area under the receiver operating characteristic curve (AUC) of the rehospitalization prediction model was 0.794 [95% confidence interval (CI): 0.756–0.832] in the derivation cohort. (B) The AUC of the rehospitalization prediction model was 0.812 (95% CI: 0.758–0.866) in the validation cohort. (C,D) Calibration plot of the mortality prediction model in the derivation and validation cohort. The dashed diagonal line represents perfect calibration. The x-axis is the predicted probability estimated by the model and the y-axis is the actual probability.

• 

  Figure 6. Kaplan–Meier estimates of mortality in the derivation and validation cohorts. The survival rate was observed using the Kaplan–Meier curves and compared using the log-rank test. (A) Kaplan–Meier estimates of the survival rate in the derivation cohort using the mortality prediction model (p < 0.0001). (B) Kaplan–Meier estimates of the survival rate in the external validation cohort using the mortality prediction model.As shown by the Kaplan–Meier curves, the two cohorts showed significant differences in mortality among the three risk-score groups (p < 0.0001).

• 

  Figure 7. Kaplan–Meier estimates of rehospitalization for heart failure in the derivation and validation cohorts. The survival rate was observed using the Kaplan–Meier curves and compared using the log-rank test. (A) Kaplan–Meier estimates of rehospitalization for heart failure in the derivation cohort using the rehospitalization prediction model (p < 0.0001). (B) Kaplan–Meier estimates of rehospitalization for heart failure in the validation cohort using the rehospitalization prediction model (p < 0.0001). As shown by the Kaplan–Meier curves, the two cohorts showed significant differences in rehospitalization for heart failure among the three risk-score groups (p < 0.0001).