Reducing bias in cancer research: application of propensity score matching.

Abstract

In cancer observational studies, differences between groups on confounding variables may have a significant effect on results when examining health outcomes. This study demonstrates the utility of propensity score matching to balance a non-cancer and cancer cohort of older adults on multiple relevant covariates. This approach matches cases to controls on a single indicator, the propensity score, rather than multiple variables. Results indicated that propensity score matching is an efficient and useful way to create a matched case-control study out of a large cohort study, and allows confidence in the strength of the observed outcomes of the study.