Interpretable Machine Learning Models Using Peripheral Immune Cells to Predict 90-Day Readmission or Mortality in Acute Heart Failure Patients

Figures

• 

  Figure 1.. Flowchart of patient selection and cohort distribution for developing and validating predictive models in AHF patients.

• 

  Figure 2.. Comparative analysis of ML classifiers-namely Logit, SVM, RF, and XGBoost - on traditional clinical data, detailing their performance through (A) ROC curves, (B) calibration plots, and (C) DCA. They achieved ROC-AUCs of 0.804, 0.807, 0.675, and 0.752, respectively. ML, machine learning; ROC, receiver operating characteristic; AUC, area under the curve; DCA, decision curve analysis; Logit, logistic regression; SVM, support vector machine; RF, random forest; XGBoost, extreme gradient boosting.

• 

  Figure 3.. Comparative analysis of ML classifiers-namely Logit, SVM, RF, and XGBoost - using peripheral immune cells, as illustrated through (A) ROC curves, (B) calibration plots, and (C) DCA. They yielded AUCs of 0.823, 0.841, 0.789, and 0.822, respectively. ML, machine learning; ROC, receiver operating characteristic; AUC, area under the curve; DCA, decision curve analysis; Logit, logistic regression; SVM, support vector machine; RF, random forest; XGBoost, extreme gradient boosting.

• 

  Figure 4.. Comparative analysis of ML classifiers-namely Logit, SVM, RF, and XGBoost on combined traditional clinical data and peripheral immune cells, as demonstrated through (A) ROC curves, (B) calibration plots, and (C) DCA. They achieved AUCs of 0.946, 0.948, 0.912, and 0.728, respectively. ML, machine learning; ROC, receiver operating characteristic; AUC, area under the curve; DCA, decision curve analysis; Logit, logistic regression; SVM, support vector machine; RF, random forest; XGBoost, extreme gradient boosting.

• 

  Figure 5.. Assessment of optimal ML model's predictive performance using external verification cohort: (A) ROC curve (AUC = 0.888), (B) calibration curve, and (D) DCA. ML, machine learning; ROC, receiver operating characteristic; AUC, area under the curve; DCA, decision curve analysis.

• 

  Figure 6.. SHAP analysis of SVM model for predicting 90-day readmission or mortality: (A) feature importance ranking and (B) summary plot. SHAP, shapley additive explanations; SVM, support vector machine; BUN, blood urea nitrogen; SBP, systolic blood pressure; DBP, diastolic blood pressure; eGFR estimated glomerular filtration rate.