Machine learning-based risk assessment for cardiovascular diseases in patients with chronic lung diseases.

The association between chronic lung diseases (CLDs) and the risk of cardiovascular diseases (CVDs) has been extensively recognized. Nevertheless, conventional approaches for CVD risk evaluation cannot fully capture the risk factors (RFs) related to CLDs. This research sought to construct a CLD-specific CVD risk prediction model based on machine learning models and evaluate the prediction performance. The cross-sectional study design was adopted with data retrieved from Waves 1 and 3 of the China Health and Retirement Longitudinal Study, including 1357 participants. Multiple RFs were integrated into the models, including conventional RFs for CVDs, pulmonary function indicators, physical features, and measures of quality of life and psychological state. Four machine learning algorithms, including extreme gradient boosting (XGBoost), logistic regression, random forest, and support vector machine, were evaluated for prediction performance. The XGBoost model displayed superior performance to machine learning algorithms for predictive accuracy (area under the receiver operating characteristic curve [AUC]: 0.788, accuracy: 0.716, sensitivity: 0.615, specificity: 0.803). This model pinpointed the top 5 RFs for CLD-specific CVD RFs: body mass index, age, C-reactive protein, uric acid, and grip strength. Moreover, the prediction performance of the random forest model (AUC: 0.709, accuracy: 0.633) was higher relative to the logistic regression (AUC: 0.619, accuracy: 0.584) and support vector machine (AUC: 0.584, accuracy: 0.548) models. Nonetheless, these models performed less favorably compared to the XGBoost model. The XGBoost model presented the most accurate predictions for CLD-specific CVD risk. This multidimensional risk assessment approach offers a promising avenue for the establishment of personalized prevention strategies targeting CVD in patients with CLDs.