Is memory in older adulthood influenced by changes in air pollution over the previous decade?: Examining concurrent...

The temporal pattern of air pollution exposure may affect cognitive aging. We characterized 10-year trajectories of particulate matter (PM_2.5) and nitrogen dioxide (NO₂) exposure at the United States census tract level using sequence and cluster analysis and evaluated their association with memory among 6750 National Health and Aging Trends Study (NHATS) participants aged 65 or older. For each year between 2000 and 2010, we trichotomized PM_2.5 and NO₂ concentrations in each census tract using the national 2010 25th and 75th percentiles of PM_2.5 (7.9 and 10.8 µg/m³) and NO₂ (4.8 and 10.8 ppb), creating nine mutually exclusive air pollution categories for each census tract each year (e.g., high PM_2.5 and high NO₂, high PM_2.5 and medium NO₂, etc.). We quantified the dissimilarity between each census tract's 10-year PM_2.5 and NO₂ trajectory and all other trajectories in NHATS using sequence analysis, and combined similar air pollution trajectories using hierarchical clustering. We assigned a 10-year air pollution trajectory cluster to each participant based on their residential census tract location in 2011. Episodic memory score in 2011 was the mean of immediate and delayed word recall scores. We estimated the association of air pollution trajectory with this episodic memory score using linear models adjusted for age, gender, education level, race/ethnicity, neighborhood socioeconomic status, census region, and urbanicity. We clustered 1080 unique air pollution trajectories from 2000 to 2010 into nine distinct groups. In general, participants belonging to clusters with higher pollutant concentrations in 2000 had lower memory scores in 2011, even when pollutant concentrations in the cluster later declined. For example, those exposed to high PM_2.5 and high NO₂ concentrations in 2000, with decreasing NO₂ concentration (and consistently high PM_2.5) starting in 2006, had an adjusted mean memory score that was 0.20 units lower (95% confidence interval: -0.35, -0.046) than those in clusters with consistently low concentrations of both pollutants. This application of sequence analysis offers a new approach for characterizing temporal patterns of PM_2.5 and NO₂ exposure, simultaneously incorporating pollutant type, timing, duration, and concentration.