中文版以人为本的基层卫生服务量表的信效度验证

Reliability and Validity of the Chinese Version of the People-Centered Primary Care Measure

背景以人为本的基层卫生服务量表(PCPCM)已在国外研究中被证明能够用于全面、准确、快速地评估基层卫生服务质量。目前,PCPCM已被我国香港学者汉化和修订成中文版,但中文版PCPCM在内地文化背景和医疗环境下的适用性与测量学性能尚未见文献报道。目的评价中文版PCPCM的信效度,旨在为内地基层卫生服务质量的评价提供可靠的工具。方法于2021年12月8—20日,采用方便抽样法,连续选取至河南省、上海市、黑龙江省6家社区卫生服务中心/乡镇卫生院、社区卫生服务站/村卫生室就诊的患者为研究对象,采用一般资料调查表、中文版PCPCM对其进行调查,并邀请10名相关领域(社会医学与卫生事业管理、全科医学、流行病与卫生统计学等)专家对量表进行评价。对中文版PCPCM进行项目分析、信度分析和效度分析。结果共回收有效问卷683份。项目分析结果显示:各条目得分与量表总分的相关系数为0.843~0.923(P均<0.001);各条目的临界值(CR值)为28.270~36.055(P均<0.001);逐一剔除各条目后,量表的Cronbach'sα系数为0.970~0.973(<0.974)。信度方面,中文版PCPCM的Cronbach'sα系数为0.974,Guttman折半系数为0.952,各条目分间的相关系数为0.667~0.913,重测信度系数为0.874,各条目的重测信度系数为0.724~0.886。条目水平的内容效度指数(I-CVI)为0.800~1.000,全体一致量表水平的内容效度指数(S-CVI)为0.818,平均S-CVI为0.973,K^(*)为0.79~1.00。探索性因子分析结果显示,KMO值为0.960,Bartlett's球形检验χ^(2)=4538.461、P<0.001,表明数据样本适合进行因子分析;按特征根>1.000的标准可提取1个公因子,公因子的累积方差贡献率为78.715%,各条目在公因子上的载荷值为0.826~0.925。验证性因子分析结果显示,初始模型拟合指标不理想。根据指标提示添加误差变量e1和e2、e3和e9、e10和e11间的相关关系后,修正模型各拟合指标处于可接受范围〔χ^(2)/df=2.983,拟合优度指数(GFI)为0.934,调整拟合优度指数(AGFI)为0.894,近似误差均方根(RMSEA)为0.081,规范拟合指数(NFI)为0.966,相对适配指数(RFI)为0.954,增值拟合指数(IFI)为0.977,非归准适配指数(TLI)为0.969,比较拟合指数(CFI)为0.977,标准化回归系数为0.790~0.900〕。已知族群效度分析结果显示,不同年龄、户口类型、家庭人均月收入、自评健康状况、慢性病患病情况、省份的患者PCPCM得分比较,差异有统计学意义(P<0.05)。结论由我国香港学者汉化的中文版PCPCM在内地文化背景和医疗环境下具有较好的测量学性能,但该工具的跨文化适用性,以及居民将量表内容概念化的能力仍有待进一步研究。

Background The People-Centered Primary Care Measure(PCPCM)scale has proven to be able to evaluate the quality of primary care measures in a accurate,all-round and rapid manner in foreign studies.And after being translated and revised,its Chinese version has been developed by Hong Kong scholars.However,there is not yet any report on the applicability and metrological performance concerning the Chinese version of the PCPCM(PCPCM-C)in the culture and healthcare settings in Chinese mainland.Objective To evaluate the reliability and validity of the PCPCM-C scale,providing evidence for its feasibility in evaluating primary care measures in Chinese mainland.Methods By use of convenience sampling,patients were consecutively selected from six primary care settings(community health center,community health station,township health center and village clinic)located in Henan Province,Shanghai,and Heilongjiang Province from December 8th to 20th,2021.They received a survey using the General Information Questionnaire developed by us,and the PCPCM-C. Then the PCPCM-C was revised in accordance with the survey results,and the assessment results by 10 Chinese professionals(public health managers,general practice experts and administrators of primary care settings)invited nationwide. After that,the PCPCM-C was analyzed using item,reliability and validity analyses. Results Altogether,683 cases who handed in responsive questionnaires were included for final analysis. The item-total correlation coefficients of the scale ranged from 0.843 to 0.923(P<0.001). The CR values of items ranged from 28.270 to 36.055(P<0.001). Leave-one-out analysis demonstrated that the Cronbach's α of the PCPCM-C ranged from 0.970 to 0.973 (<0.974). The reliability analysis of the Cronbach's α,Guttman Split-half coefficient and test-retest reliability of the PCPCM-C was 0.974,0.952,and 0.874,respectively. The inter-item correlation coefficients ranged from 0.667 to 0.913. The test-retest reliability of every item ranged from 0.724 to 0.886. The expert evaluation showed that content validity index(CVI)of every item was 0.800-1.000,the S-CVI/UA was 0.818,and the S-CVI/Ave was 0.973,the K^(*) was 0.79-1.00. The value of KMO(0.960)and result of Bartlett's test of sphericity(χ^(2)=4 538.461,P<0.001)derived from the exploratory factor analysis,suggested that the data sample was appropriate for factor analysis. One common factor with an eigenvalue >1.000 was extracted,and the cumulative variance explained by which was 78.715%,with the load value of each item of 0.826-0.925. The results of confirmatory factor analysis(CFA) showed that the fitting indicators of the initial model were undesirable. After the establishment of a covariance correlation between error variables e1 and e2,e3 and e9,and e10 and e11 according to the indication,the fitting indicators of the model were modified to be acceptable(modified CMIN:χ^(2)/df=2.983,GFI=0.934,AGFI=0.894,RMSEA=0.081,NFI=0.966,RFI=0.954, IFI=0.977,TLI=0.969,CFI=0.977,standardized regression coefficients ranged from 0.790 to 0.900). The results of known-group validity analysis showed that the total score of PCPCM-C among patients varied significantly by age,rural or urban hukou, household monthly income per person,self-rated health,and chronic disease prevalence,and region(P<0.05). Conclusion This PCPCM-C has proven to be with good psychometric quality in Chinese mainland. But further research is needed to test the cross-cultural applicability and residents' conceptualization of the scale.