基于纵向健康体检数据的老年糖尿病发病风险预测模型分析  

Analysis of the risk prediction model of diabetes in the elderly based on longitudinal health examination data

作　　者：张娅惠 贺永梅[1] 刘芳铭 郝燕菲[1] Zhang Yahui;He Yongmei;Liu Fangming;Hao Yanfei(Department of Health Management,Aerospace Center Hospital,Beijing,100049,P.R.China)

机构地区：[1]航天中心医院健康管理部,北京100049

出　　处：《老年医学与保健》2022年第4期866-870,共5页Geriatrics & Health Care

摘　　要：目的 基于纵向健康体检数据,建立老年人群的糖尿病发病风险预测模型。方法 基于纵向健康体检数据,选取2019年1月—2021年12月于航天中心医院体检的老年人群,根据随访期间是否发生糖尿病分为糖尿病组与非糖尿病组,评估并比较2组间体检指标值。采用Cox比例风险回归模型预测疾病风险,受试者工作曲线(ROC)评估预测效果,十折交叉验证法行内部验证。结果 共纳入1 512例体检者,其中男性817例,女性695例,随访时间为0.25~2.75年;随访结果显示,新发糖尿病173例,其中男性96例,女性77例。糖尿病与非糖尿病组在性别、低密度脂蛋白胆固醇(LDL-C)和总胆固醇(TC)方面比较,无统计学意义(P>0.05);2组在年龄、体质量指数(BMI)、腰围、血压、吸烟饮酒史、糖尿病家族史、血糖、空腹胰岛素、胰岛素抵抗指数与高密度脂蛋白胆固醇(HDL-C)、三酰甘油(TG)水平方面比较,差异均有统计学意义(P<0.05)。Cox模型将年龄、BMI、腰围、收缩压、舒张压、吸烟史、饮酒史、糖尿病家族史、空腹血糖、餐后2 h血糖、空腹胰岛素、胰岛素抵抗指数、HDL-C、TG作为变量,结果显示,年龄大、BMI值高、有糖尿病家族史、空腹血糖、餐后2h血糖、空腹胰岛素、胰岛素抵抗指数、HDL-C、TG水平高为老年糖尿病发病的危险因素(P<0.05)。ROC曲线显示,模型的最佳截断值为88.82%、灵敏度为72.30%、特异度为89.50%。通过十折交叉验证显示,10次验证得到的曲线下面积(AUC)分别为0.84、0.87、0.88、0.86、0.85、0.87、0.85、0.87、0.88、0.86,平均AUC为0.86。结论 基于纵向健康体检数据建立的模型能很好预测老年体检人群中糖尿病发病风险,具有一定的临床参考价值。Objective To establishof a diabetes risk prediction model for the elderly based on longitudinal health examination data.Methods Based on the longitudinal health examination data,the elderly people who underwent physical examination in Aerospace Center Hospital from January 2019 to December 2021 were selected.According to whether diabetes occurred during the follow-up period,they were divided into diabetes group and non-diabetes group,and the physical examination index values between the two groups were evaluated and compared.The Cox proportional hazards regression model was used to predict the disease risk,the receiver operating curve(ROC) was used to evaluate the prediction effect,and the ten-fold cross-validation method was used for internal verification.Results A total of 1 512 physical examination subjects were included in this study,including 817 males and 695 females.The follow-up time was 0.25~ 2.75 years.The follow-up results showed that there were 173 new-onset diabetes cases,including 96 males and 77 females.There was no significant difference in gender,low density lipoprotein cholesterol(LDL-C) and total cholesterol(TC) between the diabetes group and non-diabetes group(P>0.05),but there were significant differences in age,body mass index(BMI),waist circumference,blood pressure,smoking and drinking history,family history of diabetes,blood glucose,fasting insulin,insulin resistance index,high-density lipoprotein cholesterol(HDL-C) and triglyceride(TG) levels between the two groups(P<0.05).Taking age,BMI,waist circumference,systolic blood pressure,diastolic blood pressure,smoking history,drinking history,family history of diabetes,fasting blood glucose,postprandial 2 h blood glucose,fasting insulin,insulin resistance index,HDL-C and TG as variables,the Cox model results showed thatolder age,high BMI value,family history of diabetes,fasting blood glucose,postprandial 2 h blood glucose,fasting insulin,insulin resistance index,high HDL-C,high TG levels were risk factors for the onset of diabetes in the elder

关 键 词：老年 糖尿病 风险预测模型 健康体检 队列研究 内部验证 

分 类 号：R587.1[医药卫生—内分泌]