A Simulation Study Using Inverse Probability Weighting to Adjust for Multiple Types of Bias in Observational Studies

• Activity Number: 372
• Type: Contributed
• Date/Time: Tuesday, August 11, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Diqiong Xie*
• Companies: FDA
• Keywords: bias ; observational study ; inverse probability weighting

Abstract

During the last five years, there has been an increase in the use of inverse probability weighting (IPW) methods to reduce bias in observational studies. IPW has proven to be an effective tool to reduce biases from selective treatment assignment, missing at random, and nonrandom censoring (for survival outcomes), one type of bias at a time. Many studies have employed IPW to adjust for multiple types of bias simultaneously, where separate weights were estimated for treatment assignment, loss of follow-up and censoring, and a total weight was estimated by the product of two or three weights. There is a need to better understand the use of multiple IPWs simultaneously. First, we aim to evaluate the efficiency of bias reduction using multiple IPWs through a simulation study, and secondly, investigate steps in the practical application of using multiple IPWs in observational studies.