内科学硕士学位论文中统计学方法应用及误用调查

USE AND MISUSE OF STATISTICAL METHODS IN MASTER’S THESIS OF INTERNAL MEDICINE

目的了解某大学医学院内科学硕士研究生学位论文中统计学方法应用和误用情况及其变化趋势。方法检索万方数据库中某大学医学院2014年和2019年内科学硕士研究生学位论文177篇作为研究对象,逐篇评阅,收集其中统计学方法的应用和误用情况。结果在是否应用统计推断(包括两两比较)、统计推断方法的种类及数量方面2019年与2014年差别无统计学意义。2019年统计方法应用率的前5位为:t检验和方差分析(并列)、列联表分析(包括卡方检验和Fisher确切概率法)、线性相关分析、非参数检验和Logistic回归等;2014年为:t检验、方差分析、列联表分析、线性相关分析和非参数检验等。2019年计算OR值并给出CI的使用率高于2014年(χ^(2)=7.251,P<0.05)。2019年与2014年统计表应用率均高于90%,统计图的应用率均不及60%,SPSS统计软件的应用率较高(94.3%,86.7%),但差别均无统计学意义(χ^(2)=0.003~2.932,P均>0.05)。2019年论文中有40.2%给出具体P值,高于2014年的24.4%(χ^(2)=5.049,P<0.05)。统计学表达有误的在2019年占36.8%(32/87),高于2014年的20.0%(18/90)(χ^(2)=6.147,P<0.05)。2019年统计推断方法的误用率是58.6%(51/87),与2014年的61.1%(55/90)差别无统计学意义(χ^(2)=0.114,P>0.05)。结论内科学硕士研究生学位论文中统计学方法应用率较高,但统计描述不规范、统计推断方法误用率高。

Objective To investigate the current status and changing trend of the use and misuse of statistical methods in master’s thesis of internal medicine in the medical school of a university.Methods Wanfang Data was searched to obtain 177 master’s theses of internal medicine published by the medical school of the university in 2014 and 2019,and the theses were evaluated one by one to investigate the use and misuse of statistical methods.Results There were no significant differences in the application of statistical inference(including pairwise comparison)and the number and type of statistical inference methods between the theses in 2014 and those in 2019.In 2019,the top five statistical methods in terms of rate of use were t-test and analysis of variance(both in the first place),contingency table analysis(including chi-square test and Fisher’s exact test),linear correlation analysis,non-parametric test,and logistic regression,and in 2014,the top five statistical methods were t-test,analysis of variance,contingency table analysis,linear correlation analysis,and non-parametric test.The rate of use of odds ratio and corresponding confidence interval in 2019 was significantly higher than that in 2014(χ^(2)=7.251,P<0.05).The rate of use of statistical table was above 90%and the rate of use of statistical graph was below 60%in both 2019 and 2014,and there was a relatively high rate of use of SPSS software(94.3%and 86.7%,respectively),with no statistical significance(χ^(2)=0.003-2.932,all P>0.05).The proportion of theses with specific P value in 2019 was significantly higher than that in 2014(40.2%vs 24.4%,χ^(2)=5.049,P<0.05).The proportion of theses with incorrect statistical expression in 2019 was significantly higher than that in 2014(36.8%(32/87)vs 20.0%(18/90),χ^(2)=6.147,P<0.05).There was no significant difference in the rate of misuse of statistical inference methods between 2019 and 2014(58.6%(51/87)vs 61.1%(55/90),χ^(2)=0.114,P>0.05).Conclusion There is a relatively high rate of use of statistical methods in master’s theses of internal medicine,with high rates of a lack of standard statistical description and misuse of statistical inference methods.