Optimizing cardiovascular disease diagnosis through machine learning models integrating oxidized low-density lipoprotein and routine clinical indicators.

Introduction: To evaluate the diagnostic value of oxidized low-density lipoprotein cholesterol (oxLDL-C) for cardiovascular disease (CVD) and to construct a machine learning model integrating routine clinical indicators, thereby providing an efficient and economical tool to aid clinical diagnosis. Methods: This retrospective analysis enrolled 3, 686 participants. The discriminatory performance of oxLDL-C was compared with traditional lipid markers using receiver operating characteristic curves, and its association with CVD risk was analyzed using multivariate logistic regression. Through Recursive Feature Elimination and multivariate logistic regression, six core variables (including oxLDL-C) were ultimately selected. Seven machine learning algorithms were employed to construct predictive models, and their performance was evaluated in an internal validation set. Results: The discriminatory efficacy of oxLDL-C (AUC = 0.642) was significantly superior to traditional indicators such as low-density lipoprotein cholesterol. Its level was independently and positively associated with CVD risk (OR for the highest quartile = 3.773). The diagnostic model based on XGBoost demonstrated excellent discriminative ability (AUC = 0.911) and good calibration in internal validation. Discussion: The machine learning model integrating oxLDL-C with routine clinical indicators performs well, offering a practical tool for preliminary CVD risk screening and patient triage in resource-limited settings.