An introduction to propensity score analysis in acute care surgery research: Methodology and pitfalls.

Abstract

Observational studies assessing causal effects of interventions are subject to indication (selection) bias, which may be difficult to eliminate using traditional multivariable techniques. When properly specified, propensity score-adjusted analysis may offer an advantage over traditional regression by ensuring that investigators explicitly assess comparability of baseline prognostic factors between the treated and untreated. However, it is important to note that the effectiveness of a propensity score-adjusted analysis depends on the variables selected for the model and the analytic approach. Noninclusion of important prognostic factors and model misspecification among other errors may in fact increase bias; thus, in performing propensity score analysis, these errors must be minimized as much as possible or assessed using sensitivity analysis to ensure validity.