In observational research, issues of bias and confounding relate to study design and analysis in the setting of non-random treatment assignment where compared subjects might differ substantially with respect to comorbidities. No control over the treatment assignment and the lack of balance in the covariates between the treatment and control groups can produce confounded estimates of treatment effect. Faculty will explain how propensity scores can be used to mitigate confounding through standard observational approaches (restriction, stratification, matching, regression, or weighting). The advantages and disadvantages of standard adjustment relative to propensity score-based methods will be discussed. Details of propensity score methodology (variable selection, use, and diagnostics) will also be discussed. The course will also elaborate briefly on risk adjustment models that collapse predictors of outcomes and their use relative to propensity scores. This course is designed for those with little experience with this methodology, but some knowledge of observational databases.

PREREQUISITE: Previous attendance at the short course “Introduction to the Design & Analysis of Observational Studies of Treatment Effects Using Retrospective Data Sources,” or equivalent knowledge, is recommended.

Click here for additional information and registration details.

Related Content

Evidence Matters 2022

Evidence Matters 2022

Evidence Matters is a virtual one-day summit that brings the literature review community together from different industries to learn, engage, and solve ever-pressing evidence-based research challenges.