Propensity Scores: uses and abuses

Big data are increasingly available describing the experience and outcomes of exposure to treatments in diabetes.  These data may be used to supplement and bridge from the results of regulatory randomised trials, which are necessarily restrictive in their inclusion of patients.  Approaches for analysis using propensity scores have rightly become the most commonly used approach to derive treatment estimates.  This presentation will describe how propensity scores are calculated and used, and their strengths and limitations, particularly focusing on how they may be used alongside randomised trials to inform our understanding of the effects of treatment in real world practice.