Regression Calibration to Correct Correlated Errors in Outcome and Exposure

• Activity Number: 124
• Type: Contributed
• Date/Time: Monday, August 4, 2014 : 8:30 AM to 10:20 AM
• Sponsor: ENAR
• Author(s): Pamela Shaw*+ and Jiwei He and Bryan Shepherd
• Companies: University of Pennsylvania and University of Pennsylvania and Vanderbilt University
• Keywords: Differential error ; HIV/AIDS ; Linear regression ; Measurement error ; Moment correction ; Regression calibration

Abstract

Measurement error arises through a variety of mechanisms. A rich literature exists on the bias introduced by covariate measurement error and on methods of analysis to address this bias. By comparison, less attention has been given to errors in outcome assessment or differential error. We consider an extension of the regression calibration method to settings with errors in a continuous outcome that may be correlated with prognostic covariates or with covariate measurement error. This method adjusts for the observed measurement error in the data and can be applied with either a validation subset, on which the true data are also observed (e.g., a study audit), or a reliability subset, where repeat measures of the error prone measurements are available. We show for certain measurement error structures that the proposed method leads to unbiased estimates of the regression parameters. We also draw connections with an existing moment correction method developed for the setting of audited studies. We examine the performance of the method with simulations, and we demonstrate the method's application using data from a cohort of HIV-infected persons starting antiretroviral therapy.