Estimating elemental constituents of personal PM2.5: A modeling approach of older individuals of the China BAPE study

Abstract

Most previous studies on air pollution have used mass concentrations and elemental constituents of fine particulate matter (PM 2.5 ) from monitoring stations as dose metrics when determining health risks associated with PM 2.5 exposure, which may result in error in their estimations. Herein we conducted an exploratory panel study in Jinan, China, to investigate potential determinants and estimation of personal PM 2.5 elemental constituents using a modeling approach. For 76 old individuals aged 60–69 years, 5 repeated measurements of personal PM 2.5 were taken from September 2018 to January 2019. The average concentrations of personal PM 2.5 was 57.1 ± 46 . 3 μ g/m 3 , which was lower than ambient levels. The concentrations of Zn, Fe, and Ti in personal PM 2.5 were lower than those in ambient PM 2.5 . Personal PM 2.5 , Zn, Fe, Ti, and Ca were significantly correlated with their corresponding ambient concentrations. Ambient measurements, the demographic characteristics and lifestyles of subjects, and meteorological factors were crucial factors influencing personal exposure. The final predictive models based on ambient measurements and accessible variables explained 70% of the variance of personal PM 2.5 , and 52%–64% of the variance of personal Zn, Fe, Ti, and Ca. The average ratios of ambient to personal measurements and of modeling predictions to personal measurements of PM 2.5 , Zn, Fe, Ti, and Ca were 1.23 ± 0.47 and 1.08 ± 0.41, 1.85 ± 0.57 and 1.09 ± 0.30, 1.25 ± 0.42 and 1.04 ± 0.26, 2.20 ± 0.97 and 1.07 ± 0.41, and 1.08 ± 0.39 and 1.09 ± 0.34, respectively. The results suggest that models integrating ambient measurements and related variables can be a reliable candidate for estimating the levels of personal exposure to elemental constituents of PM 2.5 . • A panel of 76 old healthy individuals aged 60–69 years were enrolled in Jinan, China. • Using ambient elements in PM 2.5 as the proxy of personal exposure may increase exposure misclassification. • Ambient measurement and accessible variables were critical influencing factors for personal elements. • The predictive models increased the accuracy of personal exposure assessment.