2398

Background: Cohorts based on administrative data often lack information on socio-economic status or lifestyle, which could be potential confounders in the association between air pollution and health outcomes. Indirect adjustment via partitioned regression adopts an ancillary dataset to control for unmeasured confounders. We aimed to evaluate the bias correction of the indirect adjustment for individual-level variables in PM₁₀ mortality estimates in South Korea. Methods: The Korea National Health and Nutrition Examination Survey (2007-2015)-mortality linked cohort was used as a primary and ancillary dataset simultaneously to produce an adjusted mortality risk estimate of PM10. A total of 26,752 subjects (146,905 person-years), aged over 30 and resided in 122 districts of South Korea at baseline were included in the analyses. The district-level long-term PM₁₀ exposures (2007-2016 average concentrations) were assigned to individuals based on the residential address. A cox-proportional hazard regression model was used to estimate the effects of PM₁₀ on all-cause mortality. The effect estimate was indirectly and directly adjusted for smoking, physical activity, education, and occupation. Results: During the follow up by 2016, a total of 890 deaths were identified. PM₁₀ concentrations were higher in those who are smokers, less educated, do not participate in physical activity, and who are non-manual workers. The unadjusted hazard ratio for an increase of 10 μg/m³ in PM₁₀ was 1.11 (95% CI: 1.00, 1.24). The indirectly adjusted hazard ratio was 1.03 (95% CI: 0.93, 1.16), and direct adjustment yielded a hazard ratio of 1.07 (95% CI: 0.95, 1.19). Conclusions: Individual socioeconomic and lifestyle factors might confound the association between long-term PM₁₀ exposure and mortality, and the effects of PM₁₀ can be over-estimated if no correction is made for these variables. Part of the difference in estimates between direct and indirect adjustment may be due to a high censoring rate.