医学观察性研究论文统计报告质量评价指标体系的构建

发布时间：2018-05-18 13:02

本文选题：观察性研究 + 统计报告　；参考：《郑州大学》2015年硕士论文

【摘要】：目的构建医学观察性研究论文统计报告质量评价指标体系,计算各级指标权重,为定量评价观察性研究论文统计报告质量提供依据。运用构建的评价指标体系对随机抽取的医学期刊刊载的观察性研究论文进行统计报告质量评价,客观评价论文的统计报告质量。方法运用文献分析法构建医学观察性研究论文统计报告质量评价指标体系的初步框架;运用改进的专家咨询法对初步框架进行修订;运用层次分析法,依据专家对指标重要性的打分计算各级指标的权重。采用分层随机等额抽样的原则,随机抽取8本SCI收录和中文核心医学类期刊,下载270篇期刊刊载的观察性研究论文,评价并比较不同级别期刊刊载的观察性研究论文的统计报告质量。结果1三轮专家咨询的应答率分别是95%,100%,100%。第一轮专家咨询后删除5个指标,分别是二级指标“研究对象”下的“纳入标准”、“排除标准”、“对照设置”,和“研究真实性”下的“偏倚类型”和“偏倚影响”;合并两个三级指标,即二级指标“统计推断”下的“统计量”和“P值”合并为“统计量与P值”;增加两个三级指标,即“研究真实性”下的“内部真实性”和“外部真实性”。第二轮专家咨询后指标重要性和可行性的中位数均大于7,变异系数均小于0.25,专家意见达成一致。评价指标体系构建完成,包括4个一级指标,11个二级指标,27个三级指标。2 4个一级指标“题目和摘要”、“材料与方法”、“结果”、“讨论”的权重分别是15.9%、33.7%、39.0%和11.4%。三级指标中综合权重最大的是“材料与方法”下的“统计方法”,其次是“结果”下的“统计量与P值”,“模型假设”,“变量赋值”,“集中/离散趋势”,“统计图表”。3 27个三级指标中,中文核心期刊与SCI收录期刊观察性研究论文统计报告得分均较低的指标为“题目和摘要”部分的“统计方法”,“材料与方法”下的“偏倚类型”和“偏倚处理”,“结果”下的“变量赋值”;前者得满分的论文数低于总论文数50%的指标共19个,后者为6个。SCI收录期刊刊载的观察性研究论文统计报告质量高于中文核心期刊(Z=13.807,P0.001)。两个级别期刊内部横断面研究、病例对照研究、队列研究论文的统计报告得分差异无统计学意义(2c=0.578,P=0.749;c2=1.224,P=0.542)。结论1专家积极性好,可靠程度高,评价指标体系构建过程严格,各指标的相对重要性符合统计理论,观察性研究论文统计报告质量评价指标体系的构建是成功的。2与SCI医学期刊载的观察性研究论文统计报告质量相比,中文核心医学类期刊刊载的观察性研究论文统计报告质量较低。
[Abstract]:Objective to construct the evaluation index system of statistical report quality of medical observational research papers and calculate the weight of indexes at all levels so as to provide the basis for quantitative evaluation of statistical report quality of observational research papers. The evaluation index system is used to evaluate the statistical report quality of the observational research papers published in randomly selected medical journals, and objectively evaluate the statistical report quality of the papers. Methods A preliminary framework of evaluation index system for statistical report quality of medical observational research papers was constructed by using literature analysis method, the preliminary framework was revised by using improved expert consultation method, and the analytic hierarchy process was used. The weight of the index at all levels is calculated according to the expert's score on the importance of the index. Based on the principle of stratified random equal sampling, 8 SCI and Chinese core medical journals were randomly selected, and 270 observational research papers were downloaded. To evaluate and compare the quality of statistical reports of observational research papers published in different levels of journals. Results the response rates of 3 rounds of expert consultation were 95% and 100%, respectively. After the first round of expert consultation, five indicators were deleted, namely, "inclusion criteria", "exclusion criteria", "control settings" and "types of bias" and "influence of bias" under "research objects", respectively; Combining two third-level indicators, that is, "statistics" and "P values" under "statistical inference", to form "statistics and P values", and adding two third-level indicators, That is, "internal authenticity" and "external authenticity" under "Research authenticity". After the second round of expert consultation, the median of the importance and feasibility of the index is more than 7, and the coefficient of variation is less than 0.25. The evaluation index system was constructed, including 4 primary indexes, 11 secondary indexes, 27 third grade indexes, 2.24 first-level indexes, "title and summary", "materials and methods", "results" and "discussion", the weights of which were 15.93.733.79.0% and 11.4%, respectively. Among the three levels of indicators, the most comprehensive weight is "statistical method" under "material and method", followed by "statistics and P value" under "result", "model hypothesis", "variable assignment", "concentration / dispersion trend". Of the 27 three levels of statistical charts, the statistical methods of the part of "title and Abstract" were the indicators with lower scores of statistical reports in Chinese core journals and SCI journals. "bias type" and "bias processing" under "materials and methods" and "variable assignment" under "result"; The quality of statistical reports published in 6 sci journals was higher than that in Chinese core journals (ZHZ 13.807 / P0.001). There was no significant difference in the scores of internal cross-sectional studies, case-control studies and cohort study papers between the two levels of journals. Conclusion 1 the experts have good enthusiasm, high degree of reliability, strict construction process of evaluation index system, and the relative importance of each index accords with the statistical theory. The construction of statistical report quality evaluation index system of observational research papers is compared with the quality of observational research papers published in SCI medical journals. The quality of statistical reports published in Chinese core medical journals is low.
【学位授予单位】：郑州大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：G353.1;R195.1

【参考文献】