Propensity Score Matching: A Conceptual Review for Radiology Researchers

Abstract

The propensity score is defined as the probability of each individual study subject being assigned to a group of interest for comparison purposes. Propensity score adjustment is a method of ensuring an even distribution of confounders between groups, thereby increasing between group comparability. Propensity score analysis is therefore an increasingly applied statistical method in observational studies. The purpose of this article was to provide a step-by-step nonmathematical conceptual guide to propensity score analysis with particular emphasis on propensity score matching. A software program code used for propensity score matching was also presented.

Influence of neoadjuvant chemotherapy on resection of primary colorectal liver metastases: A propensity score analysis

“…It is a statistical tool to create a balanced distribution of the analyzed confounders. 32 Therefore, it is important to collect sufficient information on all relevant confounders. 32 Our study showed that the rates of complications, especially of non-surgical complications, were higher in patients receiving chemotherapy before surgical resection of CRLM (Table 1).…”

“…However, one has to keep in mind that PSM is not able to completely reduce the heterogeneity of data between two or more groups. It is a statistical tool to create a balanced distribution of the analyzed confounders . Therefore, it is important to collect sufficient information on all relevant confounders …”

“…31 It is one possible way to reduce the existing heterogeneity between several studied groups and to adjust for confounders in retrospective studies. 32 We performed nearest neighbor matching with a caliper width of 0.0 and without replacement, which created entirely matched pairs of patients between both groups (NC vs ØNC) resulting in a maximum of comparability. Moreover, the resulting patient cohort was statistically analyzed after PSM to assure comparability of the relevant characteristics of the patients (Table 1).…”

Effect of Retrievable Stent Size on Endovascular Treatment of Acute Ischemic Stroke: A Multicenter Study

“…We also used a statistical method, namely propensity score matching analysis, which could control bias effectively and minimize the imbalance among groups in observational studies. 34 Our research also has several limitations. The study was a retrospective, multicenter program, and the data were self-reported by the site investigator, which included selection bias.…”

Comparison of outcome between intrauterine balloon tamponade and uterine artery embolization in the management of persistent postpartum hemorrhage: A propensity score‐matched cohort study

“…The primary outcome was compared between women who were managed by intrauterine balloon tamponade and women who underwent uterine artery embolization using a logistic regression model, resulting in estimated odds ratios (OR) with 95% confidence intervals (CI) . Differences in secondary outcome measures were estimated by Mann‐Whitney U testing before propensity score‐matching, and by the Wilcoxon signed‐rank test after propensity score‐matching, where a two‐tailed P value <0.05 was considered statistically significant . To evaluate the robustness of our study findings with regard to propensity score‐matching, a sensitivity analysis was performed of the primary outcome measure by including the propensity score as a covariate in the logistic regression model to compare the primary outcome measure between both intervention groups, under the assumption that the propensity score has a linear functional relation with the log odds of the primary outcome …”