Nayan Madhur, Hamilton Robert J, Juurlink David N, Finelli Antonio, Kulkarni Girish S, Austin Peter C
Division of Urology, Departments of Surgery and Surgical Oncology, Princess Margaret Cancer Centre, University Health Network and the University of Toronto, Toronto, ON, Canada.
Institute for Clinical Evaluative Sciences, Toronto, ON, Canada.
BJU Int. 2017 Dec;120(6):873-880. doi: 10.1111/bju.13930. Epub 2017 Jul 7.
To determine whether studies that used propensity score (PS) methods in the urology literature provide sufficient detail to allow scientific reproducibility and whether appropriate statistical tests were used to obtain valid measures of effect.
We searched OVID Medline and the Science Citation Index from inception to November 2016 to identify studies that used PS methods in five general urology journals. From each included article, we extracted pertinent information related to the PS methodology, such as estimation of the PS, whether balance diagnostics were performed, and the statistical analysis performed.
We identified 114 articles for inclusion. Matching on the PS was the most common method used (62 studies, 54.4%). Of all studies, 103 (90.4%) described which covariates were used to estimate the PS; however, only 24 provided justification for the selected covariates. Although the majority of studies (70.2%) performed some sort of diagnostic evaluation to assess balance, few studies (24.6%) used appropriate methods for balance assessment. Only four (6.4%) studies that used PS matching provided sufficient detail to replicate the matching strategy. Finally, the majority (77.4%) of studies that used PS matching explicitly used inappropriate statistical methods to estimate the effect of an exposure on an outcome.
In the urology literature PS methods were poorly described and implemented. We provide recommendations for improvement to allow scientific reproducibility and obtain valid measures of effect from their use.
确定泌尿外科文献中使用倾向评分（PS）方法的研究是否提供了足够的细节以实现科学可重复性，以及是否使用了适当的统计检验来获得有效的效应测量值。
我们检索了从创刊至2016年11月的OVID Medline和科学引文索引，以识别在五种普通泌尿外科期刊中使用PS方法的研究。从每篇纳入的文章中，我们提取了与PS方法相关的相关信息，如PS的估计、是否进行了平衡诊断以及所进行的统计分析。
我们确定了114篇文章纳入研究。基于PS进行匹配是最常用的方法（62项研究，54.4%）。在所有研究中，103项（90.4%）描述了用于估计PS的协变量；然而，只有24项为所选协变量提供了理由。尽管大多数研究（70.2%）进行了某种诊断评估以评估平衡，但很少有研究（24.6%）使用适当的方法进行平衡评估。只有四项（6.4%）使用PS匹配的研究提供了足够的细节来复制匹配策略。最后，大多数（77.4%）使用PS匹配的研究明确使用了不适当的统计方法来估计暴露对结局的影响。
在泌尿外科文献中，PS方法的描述和实施较差。我们提供了改进建议，以实现科学可重复性，并从其使用中获得有效的效应测量值。
