Arguelles Gabriel R, Shin Max, Lebrun Drake G, DeFrancesco Christopher J, Fabricant Peter D, Baldwin Keith D
Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA, USA.
Hospital for Special Surgery, New York, NY, USA.
HSS J. 2022 Nov;18(4):550-558. doi: 10.1177/15563316221082632. Epub 2022 Apr 4.
Propensity score matching (PSM) is a statistical technique used to reduce bias in observational studies by controlling for measured confounders. Given its complexity and popularity, it is imperative that researchers comprehensively report their methodologies to ensure accurate interpretation and reproducibility.
This systematic review sought to define how often PSM has been used in recent orthopedic research and to describe how such studies reported their methods. Secondary aims included analyzing study reproducibility, bibliometric factors associated with reproducibility, and associations between methodology and the reporting of statistically significant results.
PubMed and Embase databases were queried for studies containing "propensity score" and "match*" published in 20 orthopedic journals prior to 2020. All studies meeting inclusion criteria were used for trend analysis. Articles published between 2017 and 2019 were used for analysis of reporting quality and reproducibility.
In all, 261 studies were included for trend analysis, and 162 studies underwent full-text review. The proportion of orthopedic studies using PSM significantly increased over time. Seventy-one (41%) articles did not provide justification for covariate selection. The majority of studies illustrated covariate balance through values. We found that 19% of the studies were fully reproducible. Most studies failed to report the use of replacement (67.3%) or independent or paired statistical methods (34.0%). Studies reporting standardized mean differences to illustrate covariate balance were less likely to report statistically significant results.
Despite the increased use of PSM in orthopedic research, observational studies employing PSM have largely failed to adequately report their methodology.
倾向得分匹配（PSM）是一种统计技术，用于通过控制已测量的混杂因素来减少观察性研究中的偏差。鉴于其复杂性和受欢迎程度，研究人员必须全面报告其方法，以确保准确的解释和可重复性。
本系统评价旨在确定PSM在近期骨科研究中的使用频率，并描述此类研究如何报告其方法。次要目的包括分析研究的可重复性、与可重复性相关的文献计量因素，以及方法与统计学显著结果报告之间的关联。
在PubMed和Embase数据库中查询2020年前在20种骨科期刊上发表的包含“倾向得分”和“匹配*”的研究。所有符合纳入标准的研究均用于趋势分析。2017年至2019年发表的文章用于分析报告质量和可重复性。
总共纳入261项研究进行趋势分析，162项研究进行全文审查。随着时间的推移，使用PSM的骨科研究比例显著增加。71篇（41%）文章未提供协变量选择的理由。大多数研究通过值来说明协变量平衡。我们发现19%的研究完全可重复。大多数研究未报告使用替代方法（67.3%）或独立或配对统计方法（34.0%）。报告标准化均值差异以说明协变量平衡的研究不太可能报告统计学显著结果。
尽管PSM在骨科研究中的使用有所增加，但采用PSM的观察性研究在很大程度上未能充分报告其方法。
