Misclassification in the Partial Population Attributable Risk

Abstract

The population attributable risk (PAR) is often defined as the percent reduction in disease incidence that would be observed if the exposure were to be entirely removed from the population, given the exposure distribution in the population to which the results are intended to apply. The partial population attributable risk is used to quantify the population-level impact of preventive interventions in a multi-factorial disease setting. In this dissertation, we considered the effect of non-differential risk factor misclassification on the direction and magnitude of bias of partial population attributable risk (pPAR) estimands and related quantities in Chapter 1. We also developed methods for estimation and inference in Chapters 2 and 3, using both a likelihood-based approach and a Bayesian approach to correct this bias, under different study designs. We evaluated the performance of the methods through extensive simulation studies, and applied them to the Health Professionals Follow-Up Study for risk factors of colorectal cancer.