Objective To construct and validate machine learning-based models to predict the risk of acute kidney injury(AKI)following percutaneous coronary intervention(PCI)in patients with acute ST-segment elevation myocardial infarction(STEMI).Methods A total of 2 315 STEMI patients who underwent PCI between January 2020 and June 2023 were included;306(13.2％)of them developed AKI.Baseline variables were screened using LASSO regression,with the optimal λ value selected via 10-fold cross-validation to identify AKI-associated features.Subsequently,eight distinct machine learning models were constructed and evaluated for their predictive performance.SHAP value analysis was employed to assess the impact of key variables on model predictions.Results LASSO regression identified seven variables significantly associated with AKI,including age,multivessel disease,preoperative creatinine,heart failure,white blood cell count,hemoglobin,and albumin levels.Among all the models,the light gradient boosting machine(LGBM)and extreme gradient boosting(XGB)demonstrated the best predictive performance,with training set AUCs being 0.899(95％CI:0.877-0.921)and 0.893(95％CI:0.868-0.918),and validation set AUCs being 0.809(95％CI:0.763-0.856)and 0.871(95％CI:0.833-0.909),respectively.SHAP analysis revealed that albumin,age,preoperative creatinine,and white blood cell count were the primary contributors to AKI risk.Conclusion This study successfully developed and validated machine learning-based predictive models capable of effectively identifying the risk of AKI following PCI in STEMI patients,thus providing valuable support for clinical decision-making.

Objective To construct and validate machine learning-based models to predict the risk of acute kidney injury(AKI)following percutaneous coronary intervention(PCI)in patients with acute ST-segment elevation myocardial infarction(STEMI).Methods A total of 2 315 STEMI patients who underwent PCI between January 2020 and June 2023 were included;306(13.2％)of them developed AKI.Baseline variables were screened using LASSO regression,with the optimal λ value selected via 10-fold cross-validation to identify AKI-associated features.Subsequently,eight distinct machine learning models were constructed and evaluated for their predictive performance.SHAP value analysis was employed to assess the impact of key variables on model predictions.Results LASSO regression identified seven variables significantly associated with AKI,including age,multivessel disease,preoperative creatinine,heart failure,white blood cell count,hemoglobin,and albumin levels.Among all the models,the light gradient boosting machine(LGBM)and extreme gradient boosting(XGB)demonstrated the best predictive performance,with training set AUCs being 0.899(95％CI:0.877-0.921)and 0.893(95％CI:0.868-0.918),and validation set AUCs being 0.809(95％CI:0.763-0.856)and 0.871(95％CI:0.833-0.909),respectively.SHAP analysis revealed that albumin,age,preoperative creatinine,and white blood cell count were the primary contributors to AKI risk.Conclusion This study successfully developed and validated machine learning-based predictive models capable of effectively identifying the risk of AKI following PCI in STEMI patients,thus providing valuable support for clinical decision-making.