To describe the evolution of the use and reporting of propensity score (PS) analysis in observational studies assessing a surgical procedure.
Assessing surgery in randomized controlled trials raises several challenges. Observational studies with PS analysis are a robust alternative for comparative effectiveness research.
In this methodological systematic review, we identified all PubMed reports of observational studies with PS analysis that evaluated a surgical procedure and described the evolution of their use over time. Then, we selected a sample of articles published from August 2013 to July 2014 and systematically appraised the quality of reporting and potential bias of the PS analysis used.
We selected 652 reports of observational studies with PS analysis. The publications increased over time, from 1 report in 1987 to 198 in 2013. Among the 129 reports assessed, 20% (n = 24) did not detail the covariates included in the PS and 77% (n = 100) did not report a justification for including these covariates in the PS. The rate of missing data for potential covariates was reported in 9% of articles. When a crossover by conversion was possible, only 14% of reports (n = 12) mentioned this issue. For matched analysis, 10% of articles reported all 4 key elements that allow for reproducibility of a PS-matched analysis (matching ratio, method to choose the nearest neighbors, replacement and method for statistical analysis).
Observational studies with PS analysis in surgery are increasing in frequency, but specific methodological issues and weaknesses in reporting exist.
*INSERM, UMR 1153, Centre of Research in Epidemiology and Statistics Sorbonne Paris Cité (CRESS), METHODS Team, Paris, France
†Orthopaedic Department, Assistance Publique-Hôpitaux de Paris, European Hospital Georges Pompidou, Paris, France
‡Medical school, Paris Descartes university, Sorbonne Paris cité, Paris, France
§Centre of research in epidemiology and statistics Sorbonne Paris Cité , Hopital Hotel dieu, place du Parvis notre dame, Paris, France
¶Department of Surgery, McGill University Health Centre, Montreal, Canada.
Reprints: Guillaume Lonjon, MD, INSERM, UMR 1153, Centre of Research in Epidemiology and Statistics Sorbonne Paris Cité (CRESS), METHODS team, Paris, France. E-mail: email@example.com.
Disclosure: The authors report no conflicts of interest.
Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal's Web site (www.annalsofsurgery.com).