中国台湾省35～74岁健检人群慢性肾病5年发病风险预测模型

目的构建中国台湾省35～74岁健检人群慢性肾病(CKD)5年发病风险(个体化)预测模型。方法选择1996～2006年初次参加台湾省美兆健检的35～74岁人群19 987人,将其中基线未患CKD者18 275人分为建模队列(用于建立5年发病预测模型)和验证队列(用于评估模型外部效度)。采用logistic回归构建预测模型,以ROC曲线下面积(AUC)评价拟合优度,并将人群的预测风险概率进行风险等级划分。结果去除基线患者后CKD5年发病率为5.24%(958/18275)。纳入5年发病风险预测模型的变量包括年龄、教育程度、糖尿病、高血压、低密度脂蛋白胆固醇、甘油三酯和血肌酐水平7个指标。建模队列建立的预测模型的ROC曲线下面积(AUC)约为0.734(95%CI:0.714～0.754),验证队列外部效度验证结果AUC=0.768(95%CI:0.734～0.801)。将建模队列划分为4个风险等级后,显示中危(占14.1%)和高危(占5.4%)的个体5年内发生CKD的危险分别比一般人群高2.0倍和4.8倍。结论利用中国台湾省美兆健检纵向数据建立的CKD 5年个体风险预测模型效应与信度均较高,且简单实用,对于今后CKD风险个体评价及群体监测均有较大的应用价值。

Estimated risks on developing chronic kidney disease within 5 years for adults aged 35-74 based on Taiwan MJ health-checkup Database

Objective The purpose of this study was to provide the epidemiological model evaluating the risks of developing chronic kidney disease（CKD）within 5 years among adults in Taiwan.Methods After excluding 1 712 individuals with CKD at baseline,a cohort of 18 275 subjects aged 35-74 years without CKD in the checkup point were enrolled during the year of 1996 to 2006.They were divided into a modeling cohort（for risk factors）and a verifying cohort（for evaluation）.Logistic regression models was used for the estimation of risk factors,and the area under the receiver-operating characteristic（ROC）curve（AUC）was applied for the evaluation of model predictability as well as the verification of its diagnostic property for the verifying cohort.A total of 4 different risk levels were classified based on cut points of these probabilities after transforming into normal distribution by log-transformation.Results After excluding 1 712 CKD individuals at baseline,the incidence of CKD was found to be 5.24%（958/18 275）in 5 years.Age,education,diabetes,hypertension,triglyceride,low density lipoprotein cholesterol and serum creatinine were 7 variables included in Logistic regression model for the estimation of risk factors.The results showed that the AUC was 0.734（95%CI： 0.714-0.754）for modeling cohort and 0.768（95%CI：0.734-0.801）for verifying cohort.Also individuals who were in the intermediate（14.1%）and high（5.4）risk levels were 2 times and 4.8 times more likely to develop CKD in 5 years.Conclusions The CKD Risk Score model has higher predictability and reliability and is practicable,which can be used for CKD risk evaluation for both individual and community.