Zakrison T L, Austin P C, McCredie V A
Department of Surgery, University of Miami Miller School of Medicine, 1800 NW 10th Ave., Miami, FL, 33136, USA.
Institute for Clinical Evaluative Sciences, Toronto, ON, Canada.
Eur J Trauma Emerg Surg. 2018 Jun;44(3):385-395. doi: 10.1007/s00068-017-0786-6. Epub 2017 Mar 24.
Propensity score methods are techniques commonly employed in observational research to account for confounding when estimating the effects of treatments and exposures. These methods have been increasingly employed in the acute care surgery literature in an attempt to infer causality; however, the adequacy of reporting and the appropriateness of statistical analyses when using propensity score matching remain unclear.
The goal of this systematic review is to assess the adequacy of reporting of propensity score methods, with an emphasis on propensity score matching (to assess balance and the use of appropriate statistical tests), in acute care surgery (ACS) studies and to provide suggestions for improvement for junior investigators.
We searched three databases, and other relevant literature (from January 2005 to June 2015) to identify observational studies within the ACS literature using propensity score methods (PROSPERO No: CRD42016036432). Two reviewers extracted data and assessed the quality of the studies retrieved by reviewing the adequacy of both overall reporting and of the propensity score matching methods used.
A total of 49/71 (69%) of studies adequately reported propensity score methods overall. Matching was the most common propensity score method used in 46/71 (65%) studies, with 36/46 (78%) studies reporting matching methods adequately. Only 19/46 (41%) of matching studies reported the balance of baseline characteristics between treated and untreated subjects while 6/46 (13%) used correct statistical methods to assess balance. There were 35/46 (76%) of matching studies that explicitly used statistical methods appropriate for the analysis of matched data when estimating the treatment effect and its statistical significance.
We have proposed reporting guidelines for the use of propensity score methods in the acute care surgery literature. This is to help investigators improve the adequacy of reporting and statistical analyses when using observational data to estimate effects of treatments and exposures.
倾向评分方法是观察性研究中常用的技术，用于在估计治疗和暴露的效果时考虑混杂因素。这些方法在急性护理手术文献中越来越多地被使用，试图推断因果关系；然而，使用倾向评分匹配时报告的充分性和统计分析的适当性仍不明确。
本系统评价的目的是评估急性护理手术（ACS）研究中倾向评分方法报告的充分性，重点是倾向评分匹配（以评估平衡性和使用适当的统计检验），并为初级研究人员提供改进建议。
我们检索了三个数据库以及其他相关文献（2005年1月至2015年6月），以识别ACS文献中使用倾向评分方法的观察性研究（国际前瞻性系统评价注册库编号：CRD42016036432）。两名评审员提取数据，并通过审查总体报告的充分性和所使用的倾向评分匹配方法来评估检索到的研究的质量。
共有49/71（69%）的研究总体上充分报告了倾向评分方法。匹配是46/71（65%）的研究中最常用的倾向评分方法，其中36/46（78%）的研究充分报告了匹配方法。只有19/46（41%）的匹配研究报告了治疗组和未治疗组受试者基线特征的平衡性，而6/46（13%）的研究使用了正确的统计方法来评估平衡性。在估计治疗效果及其统计学意义时，有35/46（76%）的匹配研究明确使用了适合分析匹配数据的统计方法。
我们提出了在急性护理手术文献中使用倾向评分方法的报告指南。这有助于研究人员在使用观察性数据估计治疗和暴露效果时提高报告和统计分析的充分性。
