研究背景：心血管疾病（Cardiovascular disease，CVD）已成为全球重大公共卫生问题，准确评估人群心血管疾病风险，早期识别并有效控制CVD危险因素，对预防和控制心血管疾病的发生与发展至关重要。

研究内容：1. 构建中老年人群心血管疾病基线风险评估模型，识别CVD高风险人群。2. 构建中老年人群心血管疾病风险轨迹评估模型，反映CVD风险的动态变化趋势，提高心血管疾病风险评估准确性。3. 设计心血管疾病风险评估系统，有助于实现CVD风险评估和风险分层管理。

研究方法：首先，采用LASSO回归和Boruta算法选择特征，通过逻辑回归、决策树、随机森林、k近邻、支持向量机、极端梯度提升（Extreme Gradient Boosting，XGB）、梯度提升机和神经网络8种机器学习方法构建中老年人群心血管疾病基线风险评估模型，使用SHAP解释最优模型，采用Kaplan-Meier（K-M）曲线进行心血管疾病风险分层分析。其次，使用群组轨迹模型拟合动态发展轨迹曲线，采用Cox比例风险回归模型筛选CVD发生风险较高的指标轨迹组，构建中老年人群心血管疾病风险轨迹评估模型，采用K-M曲线进行心血管疾病风险分层分析。

研究结果：1. 构建中老年人群心血管疾病基线风险评估模型，其中，XGB模型综合性能表现较为突出，AUC为0.708（95%CI：0.680 - 0.737），准确率为0.755，灵敏度0.455，精确度0.461，F1-score为0.459。SHAP结果显示，年龄大、腰围大、TYG水平高、SBP水平高、高血压、抑郁状态、女性、血脂异常等与CVD风险增加有关。K-M曲线也表明本研究模型的CVD风险评估能力良好（P＜0.001）。2. 构建中老年人群心血管疾病风险轨迹评估模型。在调整所有潜在的混杂因素后，SBP-中等水平稳定组、DBP-高水平稳定组、腰围-较低水平稳定组、腰围-中等水平稳定组、腰围-高水平增加组、BRI-高水平增加组、抑郁状态-高水平波动组与CVD发病风险增加有关，HR（95%CI）分别为1.25（1.01-1.54）、1.33（1.03-1.72）、1.35（1.03-1.77）、1.53（1.09-2.16）、2.24（1.52-3.30）、1.52（1.03-2.23）、1.92（1.30-2.84）。简化版心血管疾病风险评估模型显示，年龄大、女性、高血压、糖尿病、血脂异常、抑郁状态会增加CVD风险。心血管疾病风险轨迹评估模型显示，年龄大、女性、高血压、糖尿病、血脂异常、腰围-高水平增加组、抑郁状态-高水平波动组会增加CVD风险。与简化版心血管疾病风险评估模型相比，心血管疾病风险轨迹评估模型的重分类比例提高了17.95%（95% CI：2.61% ~ 33.30%）。K-M曲线表明模型的CVD风险评估能力良好（P＜0.001）。3. 设计心血管疾病风险评估系统。首先明确心血管疾病风险评估系统的设计目标，依据实用性、科学性、安全性、灵活性、高效性的原则，分别从基层医生、居民和管理人员角度进行需求调研和分析。在此基础上进行了系统总体架构设计，采用分层架构模式，从上向下依次为用户层、表现层、功能层、逻辑层、数据层。最后进行系统功能设计，基层医生端集成健康档案、风险评估、健康干预、随访管理、医学知识库、统计分析等功能模块；居民端提供健康知识、健康监测、健康指导、服务预约、健康咨询等功能模块；管理人员端主要包括疾病流行情况监测、疾病危险因素监测、干预效果评估等功能模块。

研究结论：本研究构建了中老年人群心血管疾病基线风险评估模型和中老年人群心血管疾病风险轨迹评估模型，为心血管疾病风险评估和分层管理提供科学依据。同时，设计了具备风险评估、健康监测、健康干预和效果评估等功能的心血管疾病风险评估系统，通过提供个性化的健康建议和干预措施，帮助中高风险人群调整生活方式、控制心血管危险因素，降低心血管疾病的发生风险。

Background: Cardiovascular disease (CVD) has become a major global public health issue. Accurately assessing the risk of cardiovascular disease in the population, early identification and effective control of CVD risk factors are crucial for preventing and controlling the occurrence and development of cardiovascular disease.

Content: 1 Build a baseline risk assessment model for cardiovascular disease in middle-aged and elderly populations, and identify high-risk populations for CVD. 2. Construct a cardiovascular disease risk trajectory assessment model for middle-aged and elderly populations, reflecting the dynamic trend of CVD risk and improving the accuracy of cardiovascular disease risk assessment. 3. Designing a cardiovascular disease risk assessment system can help achieve CVD risk assessment and risk stratification management.

Method: Firstly, LASSO regression and Boruta algorithm were used to select features. Eight machine learning methods, including Logistic Regression, Decision Tree, Random Forest, K-Nearest Neighbor, Support Vector Machine, Extreme Gradient Boosting (XGB), Gradient Boosting Machine, and Neural Network, were used to construct a baseline risk assessment model for cardiovascular disease in middle-aged and elderly populations. SHAP was used to explain the optimal model, and Kaplan Meier (K-M) curve was used for cardiovascular disease risk stratification analysis. Secondly, group-based trajectory models were used to fit the dynamic development trajectory curve, and Cox risk regression models were used to screen for indicator trajectory groups with higher CVD risk. A cardiovascular disease risk trajectory assessment model was constructed for middle-aged and elderly populations, and K-M curve was used for cardiovascular disease risk stratification analysis.

Result: 1 Constructing a baseline risk assessment model for cardiovascular disease in middle-aged and elderly populations, among which the XGB model showed outstanding comprehensive performance, with AUC of 0.708 (95% CI: 0.680-0.737), accuracy of 0.755, sensitivity of 0.455, accuracy of 0.461, and F1-score of 0.459. The SHAP results showed that older age, larger waist circumference, higher TYG levels, higher SBP levels, hypertension, depression status, female gender, and dyslipidemia were associated with increased risk of CVD. The K-M curve also indicated that the CVD risk assessment ability of this research model was good (P<0.001). 2. Build a cardiovascular disease risk trajectory assessment model for middle-aged and elderly populations. After adjusting for all potential confounding factors, the SBP moderate stable group, DBP high stable group, waist circumference low stable group, waist circumference moderate stable group, waist circumference high level increased group, BRI high level increased group, and depression state high level fluctuation group were associated with an increased risk of CVD, with HR (95% CI) values of 1.25 (1.01-1.54), 1.33 (1.03-1.72), 1.35 (1.03-1.77), 1.53 (1.09-2.16), 2.24 (1.52-3.30), 1.52 (1.03-2.23), and 1.92 (1.30-2.84), respectively. The simplified CVD risk assessment model showed that older, female, hypertension, diabetes, dyslipidemia, and depression will increase the risk of CVD. CVD risk trajectory assessment model showed that older, female, hypertension, diabetes, dyslipidemia, waist circumference high level increased group, depression high level fluctuation group will increase CVD risk. Compared with the simplified version of the cardiovascular disease risk assessment model, the reclassification rate of the cardiovascular disease risk trajectory assessment model had increased by 17.95% (95% CI: 2.61% ~ 33.30%). The K-M curve indicated that the model had good CVD risk assessment ability (P<0.001). 3. Design a cardiovascular disease risk assessment system. Firstly, clarify the design objectives of the cardiovascular disease risk assessment system, and conduct requirement research and analysis from the perspectives of grassroots doctors, residents, and management personnel based on the principles of practicality, scientificity, safety, flexibility, and efficiency. On this basis, the overall architecture design of the system was carried out, adopting a layered architecture pattern, from top to bottom, consisting of user layer, presentation layer, functional layer, logical layer, and data layer. Finally, the system function design was carried out, integrating functional modules such as health records, risk assessment, health intervention, follow-up management, medical knowledge base, and statistical analysis on the grassroots doctor side. The residential end provided functional modules such as health knowledge, health monitoring, health guidance, service appointment, and health consultation. The management team mainly included functional modules such as disease epidemic monitoring, disease risk factor monitoring, and intervention effect evaluation.

Conclusion: This study constructed a baseline risk assessment model and a cardiovascular disease risk trajectory assessment model for middle-aged and elderly populations, providing scientific basis for cardiovascular disease risk assessment and hierarchical management. At the same time, a cardiovascular disease risk assessment system has been designed with functions such as risk assessment, health monitoring, health intervention, and effectiveness evaluation. By providing personalized health advice and intervention measures, it helps middle and high-risk populations adjust their lifestyle, control cardiovascular risk factors, and reduce the risk of cardiovascular disease.