首页|受试者工作特征曲线(ROC曲线)的应用分析

受试者工作特征曲线(ROC曲线)的应用分析

扫码查看
目的:通过分析《中国计划生育学杂志》中受试者工作特征曲线(简称ROC曲线)的应用情况,为规范应用ROC曲线提供建议.方法:在中国知网检索2018年1月1日-2022年12月31日《中国计划生育学杂志》发表的与ROC曲线相关的文献,根据是否使用多指标联合预测结局,将文献分为单独组和联合组,分析文献的发表数量变化趋势、研究领域、应用条件、统计软件和模型.结果:共获得323篇文献.根据纳入排除标准,排除11篇文献,最终纳入312篇,其中单独组125篇、联合组187篇.最常用的统计软件为SPSS,数量达303篇,占97.1%,286篇文章仅提及使用ROC分析诊断价值,而未指出具体的拟合方法.2018-2022年的文章发表数量呈逐年上升的趋势,所纳入学科领域文献数量的排名前3位的依次是围产医学(187篇,59.9%)、妇科肿瘤学(45篇,14.4%)和普通妇科学(40篇,12.8%).纳人文献样本量M(P25,P75)为180.5(127,242),其中样本量<100例的文章29篇,占9.3%.整体的AUC 值M(P25,P75)为 0.812(0.747,0.881),68.3%的 AUC 值位于 0.7~0.9,<0.5 的 AUC 值共 7 个,占 0.6%.在≥2个指标的诊断试验研究中,149篇文章中未进行AUC值统计学差异性检验,仅报告指标间诊断价值的高低;29篇文章仅报告统计学比较的P值,未报告具体使用的统计学比较方法.结论:在ROC曲线的应用中,需选择合适的模型,并建议在文章的方法部分详细报告所用的ROC模型和统计软件,若存在多个指标的比较,需进行统计学检验,并报告相应的Z值和P值.
Application analysis of receiver operator characteristic curve
Objective:To analyze the application situation of ROC curve in the articles from the Chinese Journal of Family Planning,and to provide suggestions for standardized application of ROC curve.Methods:The articles related to ROC curve from the Chinese Journal of Family Planning between January 1,2018 and December 31,2022 was re-trieved in CNKI.According to whether the multi-indicators used to predict the ending of the disease jointly or not in the articles,these articles were divided into Separate group and the Joint group.The variation tendency of the published number,the research fields,the application conditions and the statistical software and models of these articles were an-alyzed.Results:A total of 323 articles were retrieved.According to the inclusion and exclusion standards,11 articles were excluded and 312 articles were eventually included,including 125 articles in Separate groups and 187 articles in the Joint group.The most commonly used statistical software in these articles was SPSS,with a number of 303 arti-cles,accounting for 97.1%.286 articles only mentioned ROC curve was used to analyze the diagnostic value without pointing out the specific fitting method.The number of the articles with ROC curve analysis from 2018 to 2022 had in-creased year by year.The top three number of the articles were about perinatal medicine(187 articles,59.9%),gynecological oncology(45 articles,14.4%)and ordinary gynecological science(40 articles,12.8%).M(P25,P75)of the samples of the articles was 180.5(127,242).The number of articles with the sample less than 100 cases was 29 articles,accounting for 9.3%.M(P25,P75)of the overall AUC values was 0.812(0.747,0.881),68.3%of AUC value was between 0.7 to 0.9,and AUC values of 7 articles were less than 0.5,accounting for 0.6%.In the studies with two or more indicators about the diagnostic test,there were 149 articles without statistical difference test of AUC values,and which only reported the levels of the diagnostic values among the indicators.29 articles reported only P val-ue of the statistical comparison,without the specific statistical comparison methods.Conclusion:In the application of ROC curve of the articles,the appropriate model should be selected,and it is recommended to report the ROC model and statistical software used in the"methods part"of the articles detailly.If there are multiple indicators comparisons,statistical tests are required and the corresponding Z values and P values should be reported.

ROC CurveAUCApplication statusFamily Planning

邓猛聪、王昕、吴尚纯、茅群霞、王宏伟

展开 >

国家卫生健康委科学技术研究所(北京,100081)

北京协和医学院研究生院

中国劳动关系学院

ROC曲线 AUC 应用现状 计划生育

中国妇女发展金基会中央级公益性科研院所基本科研业务费专项

2022HX0062021WH001

2024

中国计划生育学杂志
国家人口计生委科学技术研究所

中国计划生育学杂志

CSTPCD
影响因子:1.759
ISSN:1004-8189
年,卷(期):2024.32(2)
  • 34