Development and Validation of a Machine Learning–Based Model for Predicting All-Cause Mortality Risk in Patients With Type 2 Diabetes Mellitus Combined With Hypertension: National Cohort Study

Figures

• 

  Figure 1. Study flowchart. NHANES: National Health and Nutrition Examination Survey.

• 

  Figure 2. Presentation of the results of the least absolute shrinkage and selection operator (LASSO) regression analysis. (A) LASSO regression model screening variable trajectories. (B) LASSO regression model factor selection: the left dashed line represents the optimal λ value (“lambda.min”), whereas the right dashed line marks the λ value within 1 SE of the optimal (“lambda.1se”).

• 

  Figure 3. Receiver operating characteristic curve analysis of the machine learning models training set. DT: decision tree; LightGBM: light gradient boosting machine; LR: logistic regression; RF: random forest; ROCAUC: area under the receiver operating characteristic curve; XGBoost: extreme gradient boosting.

• 

  Figure 4. Receiver operating characteristic curve analysis of the machine learning models validation set. DT: decision tree; LightGBM: light gradient boosting machine; LR: logistic regression; RF: random forest; ROCAUC: area under the receiver operating characteristic curve; XGBoost: extreme gradient boosting.

• 

  Figure 5. Performance comparison of different machine learning models on the training sets across multiple evaluation metrics. The heat maps display the performance of decision tree (DT), extreme gradient boosting (XGBoost), logistic regression (LR), random forest (RF), and light gradient boosting machine (LightGBM). Each cell represents the value of a specific evaluation metric: accuracy, balanced accuracy (Bal_accuracy), F1-score (F_meas), J-index, κ, Matthew correlation coefficient (MCC), positive predictive value (PPV), negative predictive value (NPV), precision, recall, area under the receiver operating characteristic curve (ROC_AUC), sensitivity, and specificity. Higher values are indicated in red, whereas lower values are indicated in green, showing the model’s effectiveness in both the training and test sets. PR_AUC: area under the precision-recall curve.

• 

  Figure 6. Performance comparison of different machine learning models on the test sets across multiple evaluation metrics. The heat maps display the performance of decision tree (DT), extreme gradient boosting (XGBoost), logistic regression (LR), random forest (RF), and light gradient boosting machine (LightGBM). Each cell represents the value of a specific evaluation metric: accuracy, balanced accuracy (Bal_accuracy), F1-score (F_meas), J-index, κ, Matthew correlation coefficient (MCC), positive predictive value (PPV), negative predictive value (NPV), precision, recall, area under the receiver operating characteristic curve (ROC_AUC), sensitivity, and specificity. Higher values are indicated in red, whereas lower values are indicated in green, showing the model’s effectiveness in both the training and test sets. PR_AUC: area under the precision-recall curve.

• 

  Figure 7. Feature-ranking plots of random forest for predicting frailty risk. ASCVD: atherosclerotic cardiovascular disease; BUN: blood urea nitrogen; CKD: chronic kidney disease; COPD: chronic obstructive pulmonary disease; PIR: poverty-to-income ratio; SHAP: Shapley additive explanations.

• 

  Figure 8. Summary plots of random forest for predicting frailty risk. ASCVD: atherosclerotic cardiovascular disease; BUN: blood urea nitrogen; CKD: chronic kidney disease; COPD: chronic obstructive pulmonary disease; PIR: poverty-to-income ratio; SHAP: Shapley additive explanations.

• 

  Figure 9. Shapley additive explanations (SHAP) dependency plots for continuous and categorical variables in the random forest model: partial dependence plot for some categorical variables. ASCVD: atherosclerotic cardiovascular disease; BUN: blood urea nitrogen; CKD: chronic kidney disease; COPD: chronic obstructive pulmonary disease; PIR: poverty-to-income ratio.

• 

  Figure 10. Shapley additive explanations (SHAP) dependency plots for continuous and categorical variables in the random forest model: partial dependence plot for some categorical variables. ASCVD: atherosclerotic cardiovascular disease; BUN: blood urea nitrogen; CKD: chronic kidney disease; COPD: chronic obstructive pulmonary disease; PIR: poverty-to-income ratio.

• 

  Figure 11. Shapley additive explanations (SHAP) dependency plots for continuous and categorical variables in the random forest model: partial dependence plot for some continuous variables. ASCVD: atherosclerotic cardiovascular disease; BUN: blood urea nitrogen; CKD: chronic kidney disease; COPD: chronic obstructive pulmonary disease; PIR: poverty-to-income ratio.

• 

  Figure 12. Shapley additive explanations (SHAP) dependency plots for continuous and categorical variables in the random forest model: partial dependence plot for some continuous variables. ASCVD: atherosclerotic cardiovascular disease; BUN: blood urea nitrogen; CKD: chronic kidney disease; COPD: chronic obstructive pulmonary disease; PIR: poverty-to-income ratio.