Deep Learning Improves the MAGGIC Risk Score in Predicting Contrast-Induced Nephropathy in ST Elevation Myocardial Infraction Patients

Abstract

Contrast-induced nephropathy (CIN) is a serious complication in ST-elevation myocardial infarction (STEMI) patients undergoing primary percutaneous coronary intervention (pPCI). Early identification of high-risk patients is essential to improve outcomes and reduce mortality. The Meta-Analysis Global Group in Chronic Heart Failure (MAGGIC) risk score was originally designed to predict mortality in heart failure patients, but its role in predicting CIN has not been fully explored. In the present retrospective study, 1403 STEMI patients treated with pPCI were analyzed. Those who developed CIN had higher mortality, longer hospital stays, and more comorbidities. The MAGGIC score and 21 clinical parameters were incorporated into deep learning (DL) models, including multilayer perceptrons, TabNet, TabTransformer, and Kolmogorov-Arnold Networks (KAN) and one machine learning algorithm such as logistic regression. The best-performing model, KAN, significantly improved CIN prediction with an area under curve (AUC) of 0.92. SHapley Additive exPlanations (SHAP) analysis revealed key predictors such as pain-to-balloon time, contrast volume, baseline creatinine, and MAGGIC score. Our findings demonstrate that combining MAGGIC risk scoring with DL substantially enhances CIN prediction in STEMI patients. This approach enables identification of at-risk individuals and supports implementation of nephroprotective strategies at an early stage. The web-based calculator may assist clinical decision making.