健康管理人群高脂血症风险预测模型

A prediction model of hyperlipidemia risk based on the health management population

目的建立20岁以上健康管理人群高脂血症风险预测模型并对其预测效果进行评价。方法依托山东多中心健康管理纵向观察队列共纳入30 056人,采用Cox比例风险回归建立高脂血症预测模型,利用ROC曲线下面积(AUC)进行模型评价,十折交叉验证法检验模型的预测效果和判别能力。结果随访期间共新发高脂血症5 063例,发病密度为47.78‰。预测模型纳入的变量为年龄、性别、吸烟、饮酒、总胆固醇、甘油三酯、总胆红素、高密度脂蛋白、糖尿病和高血压10个变量。预测模型的ROC曲线下面积AUC为0.741(95%CI:0.731~0.752),经十折交叉验证平均AUC为0.741。结论构建的高脂血症风险预测模型在健康管理人群中具有较好预测能力。

Objective To construct a risk prediction model of hyperlipidemia for people aged 20 years and over. Methods A total of 30 056 people without hyperlipidemia at baseline were included based on the Shandong Multi-cen- ter Longitudinal Cohort for Health Management. The prediction model was built on Cox proportional hazards regression model. The predictability was evaluated with the area under the receiver operating characteristic （ROC） curve （AUC）. The predictive effect and distinguishing ability were verified with ten-fold cross-validation. Results During the follow- up of 3.53 ± 2.65 years, there were 5 063 new hyperlipidemia cases, and the incidence was 47.78%0. The risk factors of hyperlipidemia included age, sex, drinking, smoking, total cholesterol, triglyceride, total bilirubin, high-density lip- oprotein cholesterol, diabetes and hypertension. The AUC was 0.741 （95 % CI： 0. 731-0. 752 ）. Ten-fold cross-valida- tion verified that the AUC was 0. 741. Conclusion The prediction model of hyperlipidemia has good prediction ability in the health management population.