Exposure measurement error in air pollution health effect studies: a pooled analysis of personal exposure validation studies in 17 communities across the United States.

Abstract

Despite demonstrated adverse health effects of air pollution, the impact of exposure measurement error on these associations remains unexplored, especially for NO₂ and PM_2.5 components. We compiled daily personal measurements of PM_2.5, NO₂, and PM_2.5 components - including Al, Cd, Fe, K, Ni, Pb, S, and Si - from previous studies as true exposure indicators. These were compared against ambient concentrations from the nearest monitors. We used Spearman correlation to examine relationships between monthly averages of personal exposures and ambient concentrations. Calibration coefficients were derived using linear mixed models to quantify measurement errors. Results showed strong correlations between monthly personal exposures and ambient concentrations for PM_2.5, NO₂, Cd, Ni, S, and Si across the US. Calibration coefficients for personal PM_2.5 and NO₂ were 0.46 (95% CI: 0.13, 0.78) and 0.97 (0.35, 1.59), respectively. Significant coefficients were also found for S (0.48; 95% CI: 0.27, 0.68), Cd (0.47; 0.17, 0.76), and Ni (0.17; 0.02, 0.32). Point estimates for calibration coefficients were all below one, indicating that using the nearest monitors as exposure surrogates would attenuate associations with health risks. The measurement error in component-wise analysis highlights the need for incorporating these calibration coefficients into future studies to adjust for such errors adequately.