Abstract: Background　There are no effective treatments for dementia，a disease with a heavy burden，which is becoming increasingly prevalent among the rapidly growing aging population. As a pre-stage of dementia，mild cognitive impairment（MCI）has attracted wide attention，but the screening for it using scales has not been popularized due to various scale limitations. So it is very important to find a simple tool suitable for screening for MCI in primary care. Objective　To explore the value of the Chinese version of the General Practitioner Assessment of Cognition（GPCOG-C） in screening for MCI in primary care. Methods　By use of simple random sampling and systematic sampling，860 cases were selected from 4 962 physical examinees（≥ 60 years old）in Shanghai to attend a questionnaire survey conducted from May to July 2020 by unified trained college student volunteers using a demographic questionnaire developed by our research group，four MCI assessment scales〔Montreal Cognitive Assessment Basic（MoCA-B），Activities of Daily Living（ADL） Scale，Dementia Rating Scale（CDR）and GPCOG-C〕. MCI was diagnosed using the criteria proposed by Petersen et al. Participants with MCI were matched 1:1 to those with normal cognitive function randomly in terms of age，sex，and years of education using SPSS software to run a comparative analysis. ROC curve of the GPCOG-C in screening for MCI was plotted to examine its cut-off threshold，sensitivity，specificity，positive and negative predictive values. Spearman's rank correlation analysis was used to measure the reliability of the GPCOG-C. The internal consistency of the scale was measured by Cronbach's α. Results　Altogether，812 cases completed the survey，and 192 of them（24.27%） were assessed to have MCI. 187 MCI cases（MCI group） and 187 matched cases of normal cognitive function（control group）were finally included. Compared with the criteria for MCI defined by Petersen et al，the gold standard，the GPCOG-C was found by ROC analysis to have a maximal Youden index of 0.43. When the cut-off threshold was determined as 12 points，the sensitivity，specificity，positive predictive value and negative predictive value of the GPCOG-C were 70.05%，69.52%，69.68% and 69.89%，respectively. The control group had higher mean scores of MoCA-B and GPCOG（P<0.05）. The mean time used for completing the MOCA-B differed significantly between the groups（P<0.05），but that used for completing the GPCOG-C did not（P>0.05）. MCI group had statistically significantly lower mean scores of time orientation，clock drawing，message，memory（first name，last name，city，road，house number），patient part，informed part and total score（P<0.05）. The inter-item correlation analysis of the GPCOG-C showed that the correlation between clock drawing and memory（last name） was negative and the correlation between clock drawing and informant section was negative. Other items were positively correlated to a certain extent，and the correlation coefficient r ranged 0.021 to 0.836. The Cronbach's α of GPCOG-C was 0.78. The mean time used for completing the MoCA-B was longer than that used for completing the GPCOG-C in all participants（P<0.001）. And the mean total GPCOG-C score was positively correlated with the mean MoCA-B score（rs=0.484，P<0.001）. Conclusion　The reliability and validity of the GPCOG-C have proved to be reasonable，suggesting that it may used for the screening for MCI in primary care.

Key words: Mild cognitive impairment, General Practitioner Assessment of Cognition, Montreal Cognitive Assessment Basic

摘要： 背景　痴呆患病率随着人口老龄化加剧而上升，痴呆负担重却无有效治疗方法，轻度认知功能障碍（MCI）作为痴呆前阶段引起关注，但用于筛查MCI的各种量表因存在不足未广泛开展，因此，寻找适合基层医疗卫生机构筛查MCI的简易工具非常重要。目的　探讨全科医生认知功能评估量表（GPCOG）在基层医疗卫生机构MCI筛查中的应用价值。方法　采用简单随机抽样加系统抽样的方法从2020年5—7月在嘉定区嘉定镇参加60岁以上老年人健康体检的人群4 962人中抽出860人进行问卷调查，由经过统一培训的大学生志愿者通过采用自制调查问卷和认知功能相关量表〔蒙特利尔认知评估基础量表（MoCA-B）、日常生活能力量表（ADL）、临床痴呆量表（CDR）、GPCOG〕进行调查，根据Petersen标准诊断MCI。根据MCI确诊者年龄、性别、受教育年限特征用SPSS软件对认知正常者按照1∶1随机抽样进行匹配，之后通过对比分析，绘制受试者工作特征曲线（ROC曲线），得出划界分、灵敏度、特异度、阳性预测值、阴性预测值。GPCOG量表信度采用Spearman秩相关分析法，量表内部一致性以Cronbach's α进行评价。结果　参加此次调查并配合完成所有测试的人数为812人，符合MCI诊断者192人（24.27%）。进行抽样匹配后MCI组和与认知正常组（对照组）各187例。与MCI诊断“金标准”进行对比，GPCOG用于筛查MCI的约登指数最大为0.43，其曲线下面积（AUC）为0.78，GPCOG总分以12分为界，灵敏度为70.05%，特异度为69.52%，阳性预测值为69.68%，阴性预测值为69.89%。对照组MoCA-B评分、GPCOG分高于MCI组，差异有统计学意义（P<0.05）。两组MoCA-B用时比较，差异有统计学意义（P<0.05）；而GPCOG用时比较，差异无统计学意义（P>0.05）。MCI组时间定位、绘制时钟、消息、记忆（姓、名、市、路、门牌号）、患者部分、知情者部分及总分均低于对照组，差异有统计学意义（P<0.05）。GPCOG量表各项目间的相关性结果显示，画钟与记忆（名）、知情者部分呈负相关，其他各项目间部分呈正相关，相关系数为0.097~0.836。GPCOG量表的Cronbach's α系数为0.78。MCI组与对照组合并后，374例研究对象MoCA-B用时大于GPCOG用时，差异有统计学意义（P<0.001）；GPCOG总评分与MoCA-B总评分呈正相关（rs=0.484，P<0.001）。结论　GPCOG在MCI筛查中的信度效度尚可，可用于基层医疗卫生机构进行MCI的初步筛查。

关键词: 轻度认知障碍, 全科医生认知功能评估量表, 蒙特利尔认知评估基础量表