Five Steps to Successfully Implement and Evaluate Propensity Score Matching in Clinical Research Studies

In clinical research, the gold standard level of evidence is the randomized controlled trial (RCT). The availability of nonrandomized retrospective data is growing; however, a primary concern of analyzing such data is comparability of the treatment groups with respect to confounding variables. Propensity score matching (PSM) aims to equate treatment groups with respect to measured baseline covariates to achieve a comparison with reduced selection bias. It is a valuable statistical methodology that mimics the RCT, and it may create an "apples to apples" comparison while reducing bias due to confounding. PSM can improve the quality of anesthesia research and broaden the range of research opportunities. PSM is not necessarily a magic bullet for poor-quality data, but rather may allow the researcher to achieve balanced treatment groups similar to a RCT when high-quality observational data are available. PSM may be more appealing than the common approach of including confounders in a regression model because it allows for a more intuitive analysis of a treatment effect between 2 comparable groups. We present 5 steps that anesthesiologists can use to successfully implement PSM in their research with an example from the 2015 Pediatric National Surgical Quality Improvement Program: a validated, annually updated surgery and anesthesia pediatric database. The first step of PSM is to identify its feasibility with regard to the data at hand and ensure availability of data on any potential confounders. The second step is to obtain the set of propensity scores from a logistic regression model with treatment group as the outcome and the balancing factors as predictors. The third step is to match patients in the 2 treatment groups with similar propensity scores, balancing all factors. The fourth step is to assess the success of the matching with balance diagnostics, graphically or analytically. The fifth step is to apply appropriate statistical methodology using the propensity-matched data to compare outcomes among treatment groups. PSM is becoming an increasingly more popular statistical methodology in medical research. It often allows for improved evaluation of a treatment effect that may otherwise be invalid due to a lack of balance between the 2 treatment groups with regard to confounding variables. PSM may increase the level of evidence of a study and in turn increases the strength and generalizability of its results. Our step-by-step approach provides a useful strategy for anesthesiologists to implement PSM in their future research. Accepted for publication November 21, 2017. Funding: None. The authors declare 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 website (http://ift.tt/KegmMq). American College of Surgeons National Surgical Quality Improvement Program (ACS NSQIP) and the hospitals participating in the ACS NSQIP are the source of the data used herein; they have not verified and are not responsible for the statistical validity of the data analysis or the conclusions derived by the authors. Reprints will not be available from the authors. Address correspondence to Steven J. Staffa, MS, Department of Anesthesiology, Perioperative and Pain Medicine, Boston Children's Hospital, 300 Longwood Ave, Boston, MA 02115. Address e-mail to Steven.Staffa@childrens.harvard.edu. © 2018 International Anesthesia Research Society

from Anaesthesiology via xlomafota13 on Inoreader http://ift.tt/2ml7i2C
via IFTTT