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

作　　者：张光[1] 王广银[2] 吴红彦[1] 张红玉[1] 王停停[3,4] 李吉庆[3,4] 李敏[3,4] 康凤玲 刘言训[3,4] 薛付忠[3,4] 

机构地区：[1]山东大学附属千佛山医院健康管理中心,山东济南250014 [2]胜利石油管理局胜利医院健康管理科,山东东营257055 [3]山东大学公共卫生学院生物统计学系,山东济南250012 [4]山东大学齐鲁生物医学大数据研究中心,山东济南250012

出　　处：《山东大学学报（医学版）》2017年第6期72-76,共5页Journal of Shandong University:Health Sciences

基　　金：国家国际科技合作专项项目(2014DFA32830)

摘　　要：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.

关 键 词：高脂血症 风险预测模型 队列 健康管理人群 

分 类 号：R589.2[医药卫生—内分泌]