Background: Coronary heart disease (CHD) and cerebral ischemic stroke (CIS) are two major types of cardiovascular disease (CVD) that are increasingly exerting pressure on the healthcare system worldwide.Machine learning holds great promise for improving the accuracy of disease prediction and risk stratification in CVD. However, there is currently no clinically applicable risk stratification model for the Asian population. This study developed a machine learning-based CHD and CIS model to address this issue.Methods: A case-control study was conducted based on 8,624 electronic medical records from 2008 to 2019 at the Tongji Hospital in Wuhan, China. Two machine learning methods (the random down-sampling method and the random forest method) were integrated into 2 ensemble models (the CHD model and the CIS model). The trained models were then interpreted using Shapley Additive exPlanations (SHAP). Results:The CHD and CIS models achieved good performance with the areas under the receiver operating characteristic curve (AUC) of 0.895 and 0.884 in random testing, and 0.905 and 0.889 in sequential testing, respectively. We identified 4 common factors between CHD and CIS: age, brachial-ankle pulse wave velocity, hypertension, and low-density lipoprotein cholesterol (LDL-C). Moreover, carcinoembryonic antigen (CEA) was identified as an independent indicator for CHD.Conclusions: Our ensemble models can provide risk stratification for CHD and CIS with clinically applicable performance. By interpreting the trained models, we provided insights into the common and unique indicators in CHD and CIS. These findings may contribute to a better understanding and management of risk factors associated with CVD.

BackgroundDeath due to cardiovascular diseases (CVD) increased significantly in China. One possible way to reduce CVD is to identify people at risk and provide targeted intervention. We aim to develop and validate a CVD risk prediction model for Chinese males (CVDMCM) to help clinicians identify those males at risk of CVD and provide targeted intervention.MethodsWe conducted a retrospective cohort study of 2,331 Chinese males without CVD at baseline to develop and internally validate the CVDMCM. These participants had a baseline physical examination record (2008–2016) and at least one revisit record by September 2019. With the full cohort, we conducted three models: A model with Framingham CVD risk model predictors; a model with predictors selected by univariate cox proportional hazard model adjusted for age; and a model with predictors selected by LASSO algorithm. Among them, the optimal model, CVDMCM, was obtained based on the Akaike information criterion, the Brier's score, and Harrell's C statistic. Then, CVDMCM, the Framingham CVD risk model, and the Wu's simplified model were all validated and compared. All the validation was carried out by bootstrap resampling strategy (TRIPOD statement type 1b) with the full cohort with 1,000 repetitions.ResultsCVDMCM's Harrell's C statistic was 0.769 (95% CI: 0.738–0.799), and D statistic was 4.738 (95% CI: 3.270–6.864). The results of Harrell's C statistic, D statistic and calibration plot demonstrated that CVDMCM outperformed the Framingham CVD model and Wu's simplified model for 4-year CVD risk prediction.ConclusionsWe developed and internally validated CVDMCM, which predicted 4-year CVD risk for Chinese males with a better performance than Framingham CVD model and Wu's simplified model. In addition, we developed a web calculator–calCVDrisk for physicians to conveniently generate CVD risk scores and identify those males with a higher risk of CVD.

Aims: Death due to cardiovascular diseases (CVD) increased significantly in China. One way to reduce CVD is to identify people at risk and provide targeted intervention. We aim to develop and validate a CVD risk prediction model for Chinese males (CVDMCM) to help clinicians identify those at risk of CVD and provide targeted intervention. Methods and Results: We conducted a retrospective cohort study of 2331 Chinese males without prior CVD to develop and internally validate the CVDMCM. These participants had a baseline physical examination record (2008-2016) and one revisits record by September 2019. With the full cohort, we used single factor cox regression to examine each candidate predictor adjusted for age. 16 sequential prediction models were built on significant predictors. CVDMCM was selected based on the Akaike information criterion, the area under the ROC curve, and the percentage of variation in outcome values explained by the model (R2). This model, the Framingham CVD risk model, and Wu simplified model were all validated by bootstrapping with 1000 repetitions. CVDMCM C statistics (0.779, 95% CI: 0.733-0.825), D statistic (4.738, 95% CI: 3.270-6.864), and calibration plot demonstrated that CVDMCM outperformed the other two models. Conclusions: We developed and validated CVDMCM, which predicted 4-year CVD risk for Chinese males with better performance than the Framingham CVD model and Wu simplified model. In addition, we developed a web calculator for physicians to conveniently generate CVD risk scores and identify those with a higher risk of CVD. We believe CVDMCM had great potential for clinical usage.