Abstract
BACKGROUND: Patients with coronary artery disease (CAD) and type 2 diabetes mellitus (T2DM) are at markedly increased risk of developing heart failure (HF), yet early identification of high-risk individuals remains challenging. The remnant cholesterol inflammatory index (RCII) has been proposed as a predictor of adverse cardiovascular outcomes, but its role in patients with CAD and T2DM has not been fully elucidated. METHODS: We retrospectively analyzed clinical data from patients treated at our center. Demographic characteristics, comorbidities, medication use, and laboratory parameters were collected. Key features were selected using the Boruta algorithm, and five machine learning models-logistic regression (Logistic), decision tree (DT), elastic net regression (ENet), LASSO regression, and naïve Bayes (NB)-were constructed. Discrimination was assessed by receiver operating characteristic (ROC) curves and area under the curve (AUC), calibration by calibration plots and Brier scores, and interpretability by SHAP analysis. RESULTS: Among 1181 enrolled patients, 73 developed HF. Median RCII levels were significantly higher in the HF group. Boruta feature selection identified 13 key predictors for model development. Logistic regression demonstrated the best performance, achieving AUCs of 0.88 in the training set and 0.85 in the testing set, with overall accuracy of 0.87 and F1-score of 0.79 in the testing cohort. SHAP analysis revealed that elevated RCII, poor nutritional status, and smoking were major contributors to HF occurrence, with RCII showing a positive association with HF risk. CONCLUSION: RCII is a valuable predictor of HF in patients with CAD and T2DM. Higher RCII levels are closely linked to an increased risk of HF.