POS-257 Machine learning in evaluation on the role of microinflammation in cardiovascular events in patients with chronic kidney disease

Abstract

Lipid metabolism disorder is one of the main complications in patients with chronic kidney disease (CKD), which increases the risk of cardiovascular disease (CVD). Traditional lipid-lowering statins has been found limited benefit on the final CVD outcome of CKD patients. Therefore, this study aimed to investigate the effect of microinflammation on CVD in statin-treated CKD patients. We retrospectively analyzed statin-treated CKD patients at the Department of Nephrology of Zhongda Hospital from January 2013 to September 2020. Machine learning methods were employed to develop models of low-density lipoprotein (LDL) level and CVD indexes. Five-fold cross validation method was used to avoid the overfitting of predicting models. The accuracy and area under the receiver operating characteristic (ROC) curve (AUC) were used to evaluate the models. Gini impurity index of the predictors in the Random Forest (RF) model was ranked for the analysis of importance. The RF algorithm performed best in models of both LDL and CVD events, with an accuracy of 82.27% and 74.15%, respectively, making it the most suitable method in clinical data processing. The Gini impurity ranking of LDL model revealed that hypersensitive C-reactive protein (hs-CRP) was high relevant, while the use of satin plus gender played the least important role on the outcome both in the CVD model and in the LDL model. Meanwhile, hs-CRP was the strongest predictors of CVD events in this model. Microinflammation is closely associated with potential CVD events in CKD patients, suggesting that therapeutic strategies against microinflammation should be taken when lipid-lowering statins are used to prevent CVD events in CKD patients.