Determinants of personal exposure to fine particulate matter in the retired adults – Results of a panel study in two megacities, China

This study aimed to investigate the relationship between outdoor, indoor, and personal PM₂.₅ exposure in the retired adults and explore the effects of potential determinants in two Chinese megacities. A longitudinal panel study was conducted in Nanjing (NJ) and Beijing (BJ), China, and thirty-three retired non-smoking adults aged 43–86 years were recruited in each city. Repeated measurements of outdoor-indoor-personal PM₂.₅ concentrations were measured for five consecutive 24-h periods during both heating and non-heating seasons using real-time and gravimetric methods. Time-activity and household characteristics were recorded. Mixed-effects models were applied to analyze the determinants of personal PM₂.₅ exposure. In total, 558 complete sets of collocated 24-h outdoor-indoor-personal PM₂.₅ concentrations were collected. The median 24-h personal PM₂.₅ exposure concentrations ranged from 43 to 79 μg/m³ across cities and seasons, which were significantly greater than their corresponding indoor levels (ranging from 36 to 68 μg/m³, p < 0.001), but significantly lower than outdoor levels (ranging from 43 to 95 μg/m³, p < 0.001). Indoor and outdoor PM₂.₅ concentrations were the strongest determinants of personal exposures in both cities and seasons, with RM² ranging from 0.814 to 0.915 for indoor and from 0.698 to 0.844 for outdoor PM₂.₅ concentrations, respectively. The personal-outdoor regression slopes varied widely among seasons, with a pronounced effect in BJ (NHS: 0.618 ± 0.042; HS: 0.834 ± 0.023). Ventilation status, indoor PM₂.₅ sources, personal characteristics, and meteorological factors, were also found to influence personal exposure levels. The city and season-specific models developed here are able to account for 89%–93% of the variance in personal PM₂.₅ exposure. A LOOCV analysis showed an R² (RMSE) of 0.80–0.90 (0.21–0.36), while a 10-fold CV analysis demonstrated a R² (RMSE) of 0.83–0.90 (0.20–0.35). By incorporating potentially significant determinants of personal exposure, this modeling approach can improve the accuracy of personal PM₂.₅ exposure assessment in epidemiologic studies.