Inverse Probability Weighted Methods for Estimating Treatment Effects in Observational Studies
*Kevin J Anstrom, Duke University Medical Center 

Keywords: observational study, propensity score

Clinical research studies using observational data are often conducted to estimate the effectiveness of therapies. Whereas properly conducted randomized trials are viewed as the gold-standard for evaluating treatments, observational studies are typically considered to provide a lower quality of evidence due to the possibility of selection bias. Several statistical methods have been developed to reduce selection bias due to imbalances of the observed covariates. Among the most frequently used methods are those based on the so-called propensity score. Using the estimated propensity scores, inverse probability weighted estimators will be applied to estimate the survival benefit of beta-blocker therapy among heart failure patients. These methods will be applied in two cohorts of heart failure patients.