摘要
目的构建基于山东省健康管理队列的冠心病风险预测模型。方法构建山东省健康管理队列,基于国际上较为通用的冠心病风险预测模型变量,应用Cox比例风险回归模型进行单因素分析,利用竞争风险模型建立心脑血管事件预测模型,使用十折交叉验证法检验模型稳定性。结果共纳入队列73 386人,其中男41 968人,女31 418人。队列中位随访时间3.10年。经随访共有1 545人发生冠心病,其中男958人,发病密度为5.95/1 000人年;女587人,发病密度为4.90/1 000人年。建立的男性模型AUC为0.809(95CI:0.804~0.815),O/E值为0.98;女性模型AUC为0.869(95%CI:0.863~0.874),O/E值为1.02。经十折交叉内部验证,男性模型AUC为0.806(95%CI:0.801~0.812),女性为0.866(95%CI:0.860~0.872)。结论构建的冠心病预测模型在健康管理队列中有较好的预测能力。
Objective To establish a risk predicting model for coronary heart disease in health management cohort in Shandong province. Methods The cohort consisted of the health management cohort in Shandong province. Cox pro- portional hazards regression model was applied to screen the variables based on other predictive models. We used com- peting risk prediction model to establish the prediction model, and ten-fold cross validation to test the stability of the model. Results There were 73 386 subjects in the cohort, including 41 968 males and 31 418 females. The median fol- low-up time was 3.10 years. Coronary heart disease occurred in 1 545 sbujects, including 958 males and 587 females. The incidence density was 5.95 per 1 000 person-years for males and 4.90 per 1 000 person-years for females. The AUC of the male model was 0.809（95CI： 0.8044）. 815 ）, and the O/E value was 0.98. The AUC of the female model was 0. 869 （95 % CI： 0. 863-0.874）, and the O/E value was 1.02. In the ten-fold cross validation model, the AUC of the male mod- el was 0. 806（95% CI： 0. 8014）. 812）, and the AUC of the female model was 0.866（95% CI： 0. 860-0. 872） . Conclusion The predictive model for coronary heart disease has good predictive ability in the health management cohort.
出处
《山东大学学报（医学版）》
CAS
北大核心
2017年第6期66-71,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
冠心病
健康管理队列
风险预测模型
Coronary heart disease
Health management cohort
Risk prediction model
