健康管理人群代谢综合征发病风险预测模型

发布时间：2018-02-12 02:54

本文关键词： 代谢综合征纵向队列风险预测模型健康管理人群 Cox比例风险回归　出处：《山东大学学报(医学版)》2017年06期 　论文类型：期刊论文

【摘要】：目的基于健康管理人群队列,构建代谢综合征的5年发病风险预测模型。方法依托山东多中心健康管理纵向观察队列,选取20~80岁且基线未患代谢综合征者构建队列,采用Cox比例风险回归构建预测模型,并利用十折交叉验证法检验模型的稳定性,通过受试者工作特征曲线(ROC)下面积(AUC)和观测/期望(OE比)评价模型的预测效果。结果随访期间共发生代谢综合征1 591例(男1 273例,女318例),发病密度为38.57/1 000人年。男性代谢综合征预测模型纳入的变量包括年龄、BMI、空腹血糖、甘油三酯、高密度脂蛋白、血尿酸、总胆固醇和是否高血压,女性模型纳入变量包括年龄、BMI、空腹血糖、甘油三酯、血尿酸和是否高血压;模型ROC曲线下面积分别为0.751(95%CI:0.742~0.759)和0.745(95%CI:0.734~0.756);OE比分别为1.03和1.00;十折交叉验证ROC曲线下面积平均值分别为0.749和0.746。结论本研究利用健康管理纵向队列数据,建立了代谢综合征5年发病风险预测模型,经十折交叉验证结果表明,其在健康管理人群中有较好的预测效果,有助于识别高发病风险人群,进而减少和预防代谢综合征的发生。
[Abstract]:Objective to construct a 5-year risk prediction model of metabolic syndrome based on the health management population cohort. Methods based on the longitudinal observation cohort of multi-center health management in Shandong Province, the subjects aged 20 to 80 years without metabolic syndrome were selected to construct the cohort. The Cox proportional risk regression is used to construct the prediction model, and the 10% cross-validation method is used to verify the stability of the model. The predictive effect of the model was evaluated by using the area under the operating characteristic curve (ROC) and the observed / expected OE ratio. Results during the follow-up period, 1 591 patients (1 273 males) suffered from metabolic syndrome. There were 318 female patients with incidence density of 38.57 / 1 000 person-years. The male metabolic syndrome predictive model included BMIs, fasting blood glucose, triglyceride, high density lipoprotein, serum uric acid, total cholesterol and hypertension. The female model included age BMI, fasting blood glucose, triglyceride, serum uric acid and hypertension; The area under the model ROC curve is 0.751 / 95 CI: 0.7420.59) and the ratio of 0.74595 CI: 0.734 / 0.756 / OE is 1.03 and 1.00 respectively; the average area under ten fold cross-validation of ROC curve is 0.749 and 0.7466.Conclusion the longitudinal cohort data of health management are used in this study. A 5-year risk prediction model for metabolic syndrome was established. The results of 10% cross-validation showed that the model had a good predictive effect in the health management population and was helpful to identify the high risk population. And then reduce and prevent the occurrence of metabolic syndrome.
【作者单位】：山东大学公共卫生学院生物统计学系;山东大学齐鲁生物医学大数据研究中心;威海市立医院健康体检科;
【基金】：国家自然科学基金(81273177)
【分类号】：R589

【相似文献】