Assessing the magnitude and distribution of health impacts from nitrogen dioxide exposure in Mexico: a comparative analysis of global exposure datasets

Nitrogen dioxide (NO _2 ) affects a range of population health outcomes. However, comprehensive assessment of exposure patterns and associated health burdens has often been constrained by limited geographic coverage of air monitoring networks, particularly in low- and middle-income countries. To calculate the health impacts of NO _2 exposure in Mexico using multiple newly available models of ground-level concentrations, assessing how the magnitude, geographic distribution, and sociodemographic distribution of health risks differ across alternative datasets. We analyzed exposures and health risks using two globally modeled ground-level NO _2 datasets, and additionally used TROPOMI tropospheric NO _2 column densities to compare spatial patterns across datasets. For both ground-level data sets, we investigated differences in population-weighted exposures among various sociodemographic and geographic subgroups, examined spatial variations of NO _2 concentrations across Mexico, and calculated health impacts attributable to NO _2 . Across Mexico, we estimated between 24 000 (95% CI: 12 000–36 000) and 33 000 (95% CI: 17 000–50 000) annual premature deaths attributable to NO _2 exposure among adults aged 25 years and older, depending on the exposure dataset. In both models, most of the health burden was concentrated in Mexico City, though the relative contribution of each state varied by dataset. When mortality was modeled at the state level, the proportion of NO _2 -attributable premature deaths occurring in Mexico City and the State of Mexico—together comprising most of the metropolitan area—ranged from 52% to 81%, depending on the exposure dataset. Our analysis demonstrates substantial NO _2 -related health burdens across Mexico, including both urban and rural areas. We highlight the variation in results associated with exposure modeling, particularly for subnational and local analyses. These results indicate the need to select exposure datasets that are fit for the purpose of the research questions and test the sensitivity of health impact estimates to dataset choice.