Figures & data
Table 1. Baseline data table for heart failure endpoints.
Figure 1. The ROC curve for the random forest model for predicting heart failure in the test set. AUC: area under the curve.
Figure 2. Variable importance analysis. Results indicate the decrease in accuracy of the final model on exclusion of each specific variable, quantified on a scale of 0 to 0.14, 250, respectively. While 0 represents the minimum importance (lowest decrease in the accuracy when excluded), 0.14 and 250 represent the maximum importance (highest decrease in the accuracy when excluded). (A) Variable importance analysis of heart failure. (B) Variable importance analysis of 1-year heart failure. (C) Variable importance analysis of 5-year heart failure; CCI1: myocardial infarction; CCI2: congestive heart failure; CCI5: dementia; CCI7: connective tissue disease/rheumatic disease; CCI9: mild liver disease; CCI13: renal disease. ESRD1: primary glomerulonephritis; ESRD2: diabetes; SBP: systolic blood pressure; BMI: body mass index.
Table 2. Performance each index of the optimal model in the test set of AUC, accuracy, sensitivity, specificity, PPV, NPV, and F1-score.
Figure 3. The ROC curve for the random forest model for predicting heart failure at year 1 in the test set. AUC: area under the curve.
Figure 4. The ROC curve for the XGBoost model for predicting heart failure at year 5 in the test set. AUC: area under the curve.
Figure 5. Nomograms based on heart failure endpoints. The nomogram was quantified according to the weights of the variables selected by the model. BMI: body mass index; ECG: electrocardiograph; SBP: systolic blood pressure.
Figure 6. External validation of the risk score system for HF incidence. Kaplan–Meier’s survival curves for PD patients with higher and lower risks of HF. The risk score is divided into three levels. The survival prognosis of PD patients in the high-risk group was significantly worse than that in the low-risk group follow-up time (m): the time from study enrollment to the end of HF.
Supplemental Material
Download PDF (1.4 MB)Data availability statement
Data are not publicly available due to ethical reasons. Further enquiries can be directed to the corresponding author.