Aspects of the Use of Propensity Score Matching Methods in Neurosurgery

Abstract

Objective Observational studies and register data provide researchers with ample opportunities to obtain answers to questions that randomized controlled trials cannot answer for organizational or ethical reasons. One of the most common tools for solving this problem is the use of propensity score matching (PSM) methods. The purposes of our study were to compare various models and algorithms for selecting PSM parameters, using retrospective clinical data, and to compare the results obtained using the PSM method with those of prospective studies. Methods The results of two studies (randomized prospective and retrospective) conducted at the Novosibirsk Research Institute of Traumatology and Orthopedics were used for comparative analysis. The trials aimed to study the effectiveness and safety of surgical treatment of degenerative dystrophic lesions in the lumbar spine. We compared the results using the recommended PSM parameters (caliper=0.2 and 0.6) the propensity score is the probability of assignment to one treatment conditional on a subject’s measured baseline covariates. Propensity-score matching is increasingly being used to estimate the effects of exposures using observational data. In the most common implementation of propensity-score matching, pairs of treated and untreated subjects are formed whose propensity scores differ by at most a pre-specified amount (the caliper widthand the caliper values often used in real-life studies (0.05, 0.1, 0.25, 0.5, and 0.8) with the those obtained in a similar prospective study. Results After eliminating systematic selection bias, the results of the retrospective and randomized prospective studies were qualitatively comparable. Conclusion The results of this study provide recommendations for the use of PSM: when evaluating efficacy scores in neurosurgical studies (with a sample size < 150 patients), we recommend matching on the logit of the propensity score using calipers of width equal to 0.6 of the standard deviation of the logit of the propensity score. Level of evidence V; Type of study is expert opinion.