2型糖尿病危险评分表筛查效果及其协变量因素的ROC分析

目的 构建受试者工作特征（ROC）曲线模型，评价2型糖尿病危险评分表（Score）筛查无症状糖尿病人群的效果，并评价年龄和性别等协变量因素对Score筛查效果的影响。方法 使用2006年青岛糖尿病预防计划中的4070名受试者的横断面调查资料，所有受试者均填写Score，并行口服糖耐量实验（OGTT）以确诊无症状2型糖尿病患者。研究使用Score累计得分筛查无症状糖尿病人群的灵敏度、特异度及ROC曲线下的面积（AUC）。然后基于协变量ROC 模型，将Score转化为有序变量，分别以其为因变量，以是否为糖尿病、年龄和性别及其交互作用项为协变量拟合有序Logistic回归模型，根据AUC，评价年龄和性别对Score筛查糖尿病效果的影响。 结果 Score筛查无症状糖尿病人群的灵敏度为0.62，特异度为0.62，AUC为0.67；年龄和性别等协变量因素对Score筛查效果的影响为低年龄组大于高年龄组，女性大于男性，即Score用于筛查无症状糖尿病人群的效果为低年龄组优于高年龄组，女性优于男性。 结论 Score是筛查无症状糖尿病人群简单有效的方法； Score制定和使用时应考虑年龄和性别等协变量因素，以期达到最佳筛查效能。

ROC analysis on the screening effect of the type 2 diabetes risk score and its covariate factors

Objective      To evaluate the effect of the type 2 diabetes risk score (Score) on screening for asymptomatic diabetes and its covariate factors, such as age and gender, by establishing a receiver operating characteristic (ROC) curve model.  Methods    The data from a cross-sectional survey of 4070 subjects who underwent oral glucose tolerance tests (OGTT) for type 2 diabetes screening in the Qingdao Diabetes Prevention Program in 2006 were analyzed. All the subjects filled out the Score designed for the program at the same time. Sensitivity, specificity, and the area under the ROC curve (AUC) of the Score for asymptomatic diabetes screening was calculated. Then the Score was transformed into an ordered variable, and it was used as the dependent variable, and developing diabetes or not, age and/or gender and their interaction as the independent variables to fit an ordered logistic regression model. The AUC was calculated through the model and the influence of the covariates, age or/and gender, on the screening effect of the Score was evaluated.  Results    Sensitivity, specificity and AUC of the Score for asymptomatic diabetes screening were 0.62, 0.62 and 0.67, respectively. The influence of age on the screening effect of the Score was greater in the young age group than that in the old age group, and the influence of gender was greater in women than in men. So the Score had a better effect on screening for asymptomatic diabetes in the young women group.  Conclusion     The Score is a simple and effective tool for screening for asymptomatic type 2 diabetes. In order to achieve a better screening effect,  covariate factors, age and gender should be considered, when developing and conducting the diabetes risk score for population screening.