A Simulation Study To Quantify Attenuation And Bias Of Health Risk Estimates Due To Exposure Measurement Error In Bipollutant Epidemiologic Models

Introduction: Exposure measurement error in multipollutant epidemiologic models has the potential to introduce bias and attenuation of model coefficients. A simulation study was conducted using empirical data to quantify the impact of measurement error on resultant relative risks (RR). Methods: Empirical estimates of exposure for six pollutants (CO, NOx, EC, PM2.5, SO4, O3) from 1999-2002 in the Atlanta area for three tiers of exposure were used to calculate spatial, population, and total exposure measurement error. Empirically determined covariance of concentrations for pollutant pairs were used together with calculated spatial, population, and total exposure measurement error in a Monte Carlo simulation to simulate 'true' exposure (i.e., exposure estimates without measurement error) when the main pollutant has a RR=1.05, and the copollutant has a RR=1. 'True' and 'noisy' exposure estimates (i.e., exposure estimates with measurement error) were combined with data on emergency department visits for respir...