Background: The study aimed to determine latent patterns of geographical distribution of health-related air pollutants across the USA, and to evaluate real-world cumulative effects of these patterns on public health metrics. Methods: It was an ecological study using county-level data on the concentrations of 12 air pollutants (i.e., ozone, CO, NO2 and SO2, PM2.5 mass, and speciation, PM10 mass and speciation, HAPs, VOCs, NONOxNOy and lead) over 20 years, and the rigid measurements of population health including life expectancy at birth, age-specific mortality risks and cause-specific mortality rates (21 mutually exclusive disease groups). Latent class analysis (LCA) was used to identify the common clusters of life expectancy-associated air pollutants based on their concentration characteristics in the final studied counties (n=699). Multivariate linear regression analyses were then applied to assess the relationship between the LCA-derived clusters and health measurements with confounding adjustment. Results: PM2.5 mass, PM10 speciation, and NONOxNOy were associated with life expectancy and thus were included in LCA. Five clusters were identified: the ‘all low’ cluster (n=115, 16.5%), ‘all medium’ cluster (n=285, 40.8%), ‘high particulates’ cluster (n=152, 21.8%), ‘all high’ cluster (n=136, 19.5%) and ‘mixed profile’ cluster (n=11, 1.6%). Cluster with a more severe pollutant profile was associated with a decreasing life expectancy, an increasing mortality risk among the middle-aged and elderly populations (≥45 years), and an increasing mortality rate caused by chronic respiratory conditions, cardiovascular diseases, and neoplasms. Conclusions: Our study brings new perspectives of real-world geographical patterns of air pollution to explain health disparities across the USA.