Measurement Error in HIV Research

• Activity Number: 436
• Type: Invited
• Date/Time: Wednesday, August 6, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Health Policy Statistics Section
• Author(s): Erica Moodie and Nema Dean and Nabila Parveen*+
• Companies: McGill University and University of Glasgow and McGill University
• Keywords: Measurement error ; Causal inference ; Marginal structural models ; Simulation-Extrapolation ; HIV

Abstract

Sound health policy decisions require unbiased estimators of intervention effects. However it is frequently the case that confounding variables are measured with error; in some cases, the error may not be centred at zero, or may depend on the value of other observed covariates. For example, while marginal structural models have become a standard tool in epidemiological analyses -- developed for, and frequently used in, HIV research -- to date, little attention has been paid to the potential for measurement error in confounding variables such as CD4 cell counts. In this presentation, I will demonstrate the impact of error in confounding variables, and the improvements in estimation that are gained using a straightforward measurement error correction.