摘要
目的基于健康管理队列构建心血管事件风险预测模型。方法数据来源于山东多中心健康管理纵向观察队列,共72 843人纳入队列。随机抽取70%队列人群作为训练组,其余30%作为校验组,应用Cox比例风险回归模型对影响心血管事件发生的因素进行变量筛选,利用部分分布竞争风险模型建立心血管事件预测模型,并使用十折交叉验证法检验模型稳定性。结果队列随访期间共发生心血管事件2 463例,发病密度为88.79/1 000人年,死于非心血管事件164例。最终纳入模型的变量包括年龄、吸烟、体质量指数、高血压、糖尿病、血脂异常、ST-T改变、T波改变、异常Q波、心律失常及肾脏疾病。训练组ROC曲线下面积男性为0.837(95%CI:0.821~0.853),女性为0.897(95%CI:0.880~0.913);校验组ROC曲线下面积男性为0.838(95%CI:0.813~0.862),女性为0.893(95%CI:0.872~0.914)。结论构建的心脑血管事件预测模型在健康管理人群中有较好的预测能力。
Objective To establish a model to evaluate the risk of cardiovascular disease (CVD) among health man- agement population. Methods The cohort consisted of 72 843 individuals who had physical check-up at Shandong Multi-center Longitudinal Cohort for Health Management. They were all free of CVD events. We randomly divided the cohort into the training group (70%) and testing group (30%). Cox proportional hazards regression model was applied to choose risk factors of CVD, competing risk prediction model was used to establish a prediction model for CVD, and ten-fold cross validation was used to test the stability of the model. Discriminatory ability was determined by the area under the receiver operating characteristic curve (AUC). Results There were 2 463 CVD cases during the study period and the incidence was 88.79/1 000 person-year, and 164 people died of other causes. The risk factors included age,smoke, BMI, hypertension, diabetes, dyslipidemia, ST-T segment changes, T wave change, abnormal Q wave, arrhythmia and chronic kidney diseases. The estimated AUC of the model in the training group was 0.837(95% CI: 0. 821-0. 854 ) for males and 0. 897 ( 95 % CI:0. 880-0.913 ) for females. The estimated AUC of the model in the testing group was 0.838(95% CI:0. 813-0. 862) for males and 0.893 (95% CI:0. 872-0.914) for females. Conclusion The risk prediction model can be used to screen high-risk subjects of CVD in health management population.
出处
《山东大学学报(医学版)》
CAS
北大核心
2017年第6期56-60,65,共6页
Journal of Shandong University:Health Sciences
基金
国家国际科技合作专项项目(2014DFA32830)
山东省医药卫生科技发展计划项目(2013WS0230)
关键词
心血管事件
健康管理
风险预测模型
Cardiovascular event
Health management
Risk prediction model