14种非损伤糖尿病筛查风险评分模型的流行病学特征及其验证

Epidemiology characteristic and verification of 14 non-invasive risk score models for screening patients with diabetes

目的总结公开发表的非损伤糖尿病筛查风险评分模型的特点,利用广州市人群进行验证,为建立广州市人群的风险评分模型提供参考。方法查阅国内外文献,选择出非损伤糖尿病筛查风险评分模型,总结文献与模型的特点及其诊断试验评价结果等流行病学特征,并用广州市人群数据来验证各模型在广州市人群中的可重复性。结果共有13篇文献的14个模型纳入研究。各模型使用的预测指标主要包括8类,分别为年龄、肥胖相关指标、性别、糖尿病家族史、高血压史、运动相关指标、吸烟和类固醇药使用史。约登指数最小为1.208,最大值为1.521,中位数为1.400。受试者工作特性曲线下面积(AUC)最小为0.673,最大为0.884,中位数为0.756。广州市人群数据中的AUC与原文相比,2个增加,12个减少;约登指数与原文比较,使用原文切点时,12个减少,1个增大,使用最佳切点时,11个减少,3个增大。结论 13个文献的14种非损伤糖尿病筛查风险评分模型绝大部分设计合理,结果可靠,在广州市人群中表现尚可,对于建立广州市人群的风险评分模型具有相当大的参考意义。

Objective To summarize the characteristics of the published non-invasive risk score models for screening the undiagnosed patients with diabetes to verify in residents of Guangzhou and provide the base for establishing the risk score models in residents of Guangzhou. Methods The literatures were searched for selecting the non-invasive risk score models and the epidemiological characteristics and the results of diagnostic tests of those models were summarized. The data of resident in Guangzhou were used to verify the repeatability of models in resident in Guangzhou. Results Thirteen articles and 14 models were included in the study. Eight predictive indexes in those models were age, obesity-related indicators, gender, family history of diabetes, history of hypertension, sports-related indicators, smoking and the history of using steroids. The minimum, maximum and median values of Youden index were 1.208, 1.521 and 1.400, respectively. The minimum, maximum and median values of area under the receiver operating characteristic curve（ AUC） were 0. 673, 0. 884 and 0. 756, respectively. As compared with the literaturess, 2 AUCs increased and 12 AUCs decreased in data of resident in Guangzhou. As compare with literatures, 12 Youden indexes reduced and one Youden index enhanced when using the original cut-off; 11 Youden indexes reduced and 3 Youden indexes enhanced when using the optimal cut-off. Conclusion The most designs of the 14 non-invasive risk score models in screening for undiagnosed diabetes from 13 articles are reasonable and whose results are reliable. Performance in resident in Guangzhou is acceptable. Those models are useful for establishing the risk score models in residents of Guangzhou.