Explainable Machine Learning for Predicting Major Adverse Renal and Cardiovascular Events in Acute Myocardial Infarction With Chronic Kidney Disease: Algorithm Development and Validation

Figures

• 

  Figure 1.. Formation of the discovery and validation cohorts and occurrence of major adverse renal and cardiovascular events in patients with acute myocardial infarction (AMI) and chronic kidney disease (CKD). MARCE: major adverse renal and cardiovascular events; SHAP: Shapley additive explanations; TCM: traditional Chinese medicine; XGBoost: extreme gradient boosting.

• 

  Figure 2.. Feature selection using the least absolute shrinkage and selection operator (LASSO) model. (A) Selection of the tuning parameter (λ) in the LASSO model via 10-fold cross-validation based on the 1-SE criterion. Binomial deviance from the LASSO cross-validation procedure was plotted as a function of log(λ). The y-axis indicates binomial deviance, and the x-axis indicates log(λ). Numbers along the upper x-axis represent the average number of predictors. Red dots indicate average deviance values for each model with a given λ, and vertical bars show upper and lower deviance values. The vertical black lines define the optimal values of λ. (B) LASSO coefficient profiles of clinical features. The dotted vertical line indicates the selected λ. Seven features with nonzero coefficients are shown.

• 

  Figure 3.. Performance of machine learning models in the test set. (A) Comparison of area under the receiver operating characteristic curves among extreme gradient boosting (XGBoost), random forest, logistic regression, and decision tree models in the test set. (B) Calibration plots of XGBoost, random forest, logistic regression, and decision tree models in the test set. The Brier score is reported in the lower right legend (the smaller the value, the better the calibration). AUC: area under the receiver operating characteristic curve.

• 

  Figure 4.. Shapley additive explanations (SHAP) summary plot of the extreme gradient boosting (XGBoost) model with selected features. (A) Average absolute impact of variables on the magnitude of the final model output, ordered by decreasing feature importance. (B) The plot depicts the dot estimates on the model output for the XGBoost model. Each dot represents an individual patient from the dataset. Purple represents higher SHAP values of specific features, whereas yellow represents lower SHAP values of specific features. Higher SHAP values indicate a greater risk of major adverse renal and cardiovascular events. ACEI: angiotensin-converting enzyme inhibitor; ARB: angiotensin receptor blocker; ARNI: angiotensin receptor-neprilysin inhibitor; eGFR: estimated glomerular filtration rate.

• 

  Figure 5.. Nomogram predicting major adverse renal and cardiovascular events in patients with acute myocardial infarction and chronic kidney disease. ACE: angiotensin-converting enzyme; ARB: angiotensin receptor blocker; ARNI: angiotensin receptor-neprilysin inhibitor; eGFR: estimated glomerular filtration rate; MARCE: major adverse renal and cardiovascular events.

• 

  Figure 6.. Performance comparison between the nomogram model and the Global Registry of Acute Coronary Events (GRACE) score in the test set. Receiver operating characteristic curve of the nomogram (A) and the GRACE score (B) in the test set; 400-resample bootstrapped calibration plots of the nomogram (C) and the GRACE score (D). AUC: area under the receiver operating characteristic curve; MARCE: major adverse renal and cardiovascular events.