Education Differences in the Adverse Impact of PM_2.5 on Incident Cognitive Impairment Among U.S. Older Adults

Air pollution exposure is an important risk factor for both cognitive decline and dementia [1]. Fine particulate matter air pollution (PM_2.5) is ubiquitous in the air we breathe and is particularly harmful to the aging brain [2]. Older adults living in areas with higher concentrations of PM_2.5, for example, perform worse on assessments of cognitive function [3–7] and experience more rapid cognitive decline [8, 9] than older adults living in areas with lower concentrations of PM_2.5. This may be due to changes in brain structure, including white matter loss, which neuroimaging data has linked to PM_2.5 exposure [10, 11]. These findings are consistent with evidence that air pollution increases risk of both brain inflammation and accumulation of amyloid-β, which are implicated in the pathogenesis of Alzheimer’s disease [12, 13]. Older adults experiencing air pollutant-induced inflammation and neurodegeneration may therefore be more likely to develop cognitive deficits [14]. Research to-date, however, has not found strong evidence for an association between PM_2.5 and incident cognitive impairment [15].

Individuals with low levels of education may be particularly vulnerable to the negative impacts of air pollution on brain health. Prior research shows that less educated older adults report lower cognitive function [16–18] and are at increased risk for cognitive decline [19] and the onset of dementia and Alzheimer’s disease [20]. Less educated individuals are also more likely to live in more polluted neighborhoods [21, 22]; however, even when similarly exposed to air pollution, individuals with less education often do not have the cognitive, economic, and health resources that more educated people have to mitigate the potential damage of air pollution to the brain.

Highly educated individuals may have a greater reserve against, or ability to compensate for, damage to the structure and functions of the brain that can arise from exposure to toxins such as air pollution. The cognitive decline that would typically accompany damage to brain function caused by air pollution exposure may be mitigated either through a reserve of pre-existing brain networks that have higher capacity (i.e., are more efficient) and are, therefore, less susceptible to damage, or through an ability to compensate against damage by activating alternate networks [23, 24]. Research has found, for instance, that among individuals whose brains had pathology consistent with Alzheimer’s disease in postmortem examination, the more educated individuals were less likely to have exhibited clinical symptoms of dementia during their lifetimes compared to those with less education [25].

Education also confers socioeconomic and health advantages that may reduce susceptibility to the adverse cognitive effects of air pollution [26]. Highly educated individuals earn more money and accumulate more wealth, and these economic resources increase one’s ability to engage in healthier behaviors, including eating more nutritious food and smoking less, which have been shown to mitigate the effects of air pollution on health [27–29]. Having access to greater economic resources may also decrease co-exposure to other pollutants (e.g., indoor pollutants and second-hand smoke) that could exacerbate the adverse effects of ambient air pollution. Finally, less educated individuals are more likely to have pre-existing medical conditions, such as diabetes and cardiorespiratory disease, which are known to increase susceptibility to air pollution [30, 31].

Thus, education may play a key role in buffering the effect of air pollution on cognitive impairment. The current study tests the hypothesis that the association between PM_2.5 and incident cognitive impairment is stronger among those with less educational attainment. Two major societal changes related to education and air pollution could, however, alter the relationship between air pollution, education, and cognitive impairment among successive cohorts of older adults. First, educational attainment and access to quality education increased markedly in the first half of the 20th century [32, 33]. As a result, individuals who completed less than a high school education comprise a smaller proportion of the older adult population today than they did merely a decade ago and also represent a much more disadvantaged group. Second, in the past 20 years the EPA began regulating PM_2.5, which resulted in a 43% decline in the national average PM_2.5 concentration [34]. Therefore, this study examines the role of education in the association between PM_2.5 and cognitive impairment in two cohorts of older adults assessed 10 years apart.

Characteristics of the 2004 and 2014 samples are shown in Table 1. In 2004, 13.9% of respondents had incident cognitive impairment. Most respondents had completed at least 12 years of education, with 6.4% completing 8 or fewer years of school and 11.8% completing 9–11 years of education. In 2014, incident cognitive impairment was found in only 10.5% of respondents. Over 90% had completed at least 12 years of education, with 3.5% completing 8 or fewer years of school, and 6.8% completing 9–11 years of education. In both 2004 and 2014, the sample was about 70 years old on average, just over half were women, and the majority were white.

We plot the distribution of PM_2.5 for both 2004 and 2014 in Fig. 1. The mean 2004 concentration of PM_2.5 was 12.4, with a standard deviation of 2.8. Thus, just over half of the sample lived in areas with PM_2.5 concentrations above the EPA regulatory threshold of 12μg/m³. By 2014, the distribution of PM_2.5 had changed considerably. The mean concentration was 9.2, with a standard deviation of 1.7. In 2014, most respondents lived in areas with PM_2.5 concentrations below the threshold at which the EPA considers harmful to human health.

Table 2 shows the main effects and interactions for PM_2.5 and education from mixed-effects models of incident cognitive impairment. Results showing all model covariates are provided in Supplementary Tables 1 and 2. Log odds and standard errors for 2004 are shown in Table 2A. In Model 1, a 5μg/m³ increase in PM_2.5 concentration was associated with a 12% increase in the odds of incident cognitive impairment (b = 0.12, p < 0.05). This relationship was no longer statistically significant with adjustment for individual sociodemographic characteristics in Model 2, in particular race/ethnicity. In addition, compared to older adults with 8 or fewer years of education, those with 12 years of education had lower log odds of cognitive impairment (b = –0.900, p < 0.001), as did those with 13 or more years of education (b = –1.558, p < 0.001). Interactions between PM_2.5 and education included in Model 3 indicate that a 5μg/m³ increase in PM_2.5 concentration was associated with an increase in the log odds of incident cognitive impairment among those with 8 or fewer years of education (b = 0.314, p < 0.05). There was also a statistically significant interaction between PM_2.5 and 13 or more years of education (b = –0.436, p < 0.05), indicating no association between PM_2.5 and incident cognitive impairment among the most educated. These associations were largely unchanged with further adjustments for socioeconomic factors, health status, and health behaviors in Model 4. We found nearly identical results in models using a 1-year average for PM_2.5 in 2003.

We plot the predicted probability of cognitive impairment in 2004 in Fig. 2 based on estimates from Model 4 across PM_2.5 concentrations for each education group. Predicted probabilities and associated 95% confidence intervals are provided in Supplementary Table 3. To avoid out-of-sample predictions, we plotted data over values of PM_2.5 with cell sizes of 100 or greater. The flat line found for those with 13 years or more of education indicates no association between PM_2.5 and cognitive impairment for this group. We only find a statistically positive association between PM_2.5 and incident cognitive impairment among those with 8 or fewer years of education, indicating that it is the least educated that are at increased risk of cognitive impairment from living in more polluted environments.

Next, we examined these associations in 2014 and present those results in Table 2B. Model 1 shows that a 5μg/m³ increase in PM_2.5 concentration was associated with a 32% increase in the odds of incident cognitive impairment (b = 0.323, p < 0.05). This relationship was no longer statistically significant with adjustment for individual sociodemographic characteristics in Model 2. Similar to the results for 2004, the association was reduced primarily due to the inclusion of race/ethnicity. Compared to older adults with 8 or fewer years of education, we found lower log odds of cognitive impairment among those with 9–11 years of education (b = –0.539, p < 0.05), 12 years of education (b = –1.206, p < 0.001), and 13 or more years of education (b = –2.074, p < 0.001). We did not find any evidence for an interaction between education and PM_2.5 in 2014.

As shown in Fig. 1, the distribution of PM_2.5 substantially shifted from 2004 to 2014 resulting in fewer older adults living in high pollution environments in 2014. Furthermore, this shift in PM_2.5 occurred throughout the distribution of education (Supplementary Table 4), including among those with cognitive impairment (Supplementary Table 5). This may explain the lack of an association between education and PM_2.5 on incident cognitive impairment in 2014. To determine whether the difference in findings between the 2004 and 2014 samples was due to differences in the pollution environment, we conducted additional analysis of the 2004 sample living in areas with PM_2.5 less than 12μg/m³. There was no association between education and PM_2.5 on cognitive impairment among older adults in less polluted environments in 2004.

Lower educational attainment and greater exposure to air pollution are both associated with poor cognitive function among older adults, and theory and evidence suggests education may moderate the relationship between air pollution and brain health. But no studies to our knowledge have examined whether the association between air pollution and cognitive impairment varies according to level of education. Our study aimed to address this gap by determining if the relationship between PM_2.5 and incident cognitive impairment differed according to older adults’ educational attainment. We found some evidence that education serves as a buffer against the adverse impact of air pollution on cognitive impairment, but that this may only be true in populations exposed to high pollution environments.

We found that a 5μg/m³ increase in PM_2.5 concentration was associated with a 15% increase in the odds of incident cognitive impairment in 2004, but that with adjustment for individual characteristics, particularly race/ethnicity, the association was no longer statistically significant. The lack of an association between PM_2.5 and incident impairment, and the confounding role of race, has also been reported in other studies of older U.S. adults [15]. We think race/ethnicity confounds the association between PM_2.5 and cognitive impairment because race/ethnic minorities have higher risk of cognitive impairment and also live in more polluted areas due to a history of racial residential segregation and environmental injustices in these communities. We did, however, find evidence of an association between PM_2.5 and incident impairment among the least educated. While the likelihood of incident cognitive impairment was unrelated to PM_2.5 among those with 13 or more years of education, it increased with increasing concentrations of PM_2.5 among those with 8 or fewer years of education. Similar interactions between PM_2.5 and education have been found in research on mental health [43]. Accounting for socioeconomic and health-related risk factors did not explain the greater impact of PM_2.5 on incident cognitive impairment among those with less education, suggesting education operates via other resources such as cognitive reserve. Air pollution-induced damage to the brain may have a weaker impact on cognitive function for those with higher levels of education because education itself is associated with greater cognitive processing efficiency and increased ability to mobilize alternative neural pathways necessary for maintaining cognitive function [23].

In 2014, a 5μg/m³ increase in PM_2.5 concentration was associated with a 37% increase in the odds of incident cognitive impairment. But the association was not statistically significant after adjusting for individual characteristics. Although there was an education gradient in incident cognitive impairment, we did not find any interactions between PM_2.5 and education. Research has shown that education is becoming a stronger predictor of health and mortality among older adults [44–46], due in part to compositional changes in who completes and does not complete higher education [33] as well as socio-historical changes occurring within the U.S. education system that changed the meaning of education across successive cohorts [32]. Dementia prevalence has been declining in the U.S. population, and globally, attributed primarily to increased educational attainment in younger cohorts of older adults, and the resulting improvements in health [47–51]. At the same time, PM_2.5 has been declining across the U.S. and may be a less important predictor of cognitive impairment at lower concentrations. Studies examining the association of PM_2.5 with mortality, for instance, have found that the human health burden from PM_2.5 declined from 2000 to 2010 [52, 53]. Thus, while education may protect against the adverse impact of PM_2.5 at the higher levels of pollution observed in 2004, it may not confer any additional protection under conditions of improved air quality.

Our findings provide preliminary evidence that environmental policy should be considered in approaches to reduce or delay cognitive impairment and dementia in the older adult population. Much of the focus in dementia research has been on reducing risk from individual-level factors such as poor health and behaviors [54, 55]. The current study adds to the growing literature showing that environmental factors are also key determinants of cognitive health and highlights the importance of considering the interaction between individual and environmental risks. In addition, we identified a potentially vulnerable group, those with the least education, who had the greatest risk of cognitive impairment with increasing PM_2.5 in 2004. However, though PM_2.5 has declined since 2000, data from the past few years indicate a return to higher concentrations, with potentially negative implications for human health [52]. A return to prior PM_2.5 levels could result in increasing rates of onset and progression of cognitive impairment and dementia in the population.

Strengths of this study include the use of a large, nationally representative and socioeconomically and geographically diverse cohort of U.S. older adults that allowed us to examine incident cognitive impairment across a wide distribution of educational attainment and PM_2.5. In addition, by examining two different time periods we were able to show that the impact of PM_2.5 on incident impairment, and the potentially modifying role of education, are highly dependent on cohort-specific levels of educational attainment and exposure to high pollution environments.

This study also has some limitations. The assessments on which we base our classification of cognitive impairment come from survey-based measures rather than physician’s diagnosis. However, classification of cognitive impairment in the HRS demonstrates very high agreement with that based on a consensus panel of experts in neuropsychiatric assessments of dementia [37]. In addition, our measure of outdoor PM_2.5 concentrations may not completely capture total individual exposure. Although outdoor pollution is an important contributor to total exposure, exposure can also occur in other contexts, such as in the workplace or during daily road travel. As the sample was aged 60 and older, most of the sample was not employed, so pollution concentrations at their place of residence may be a good approximation of their total exposure. Due to the relatively small spatial variability in PM_2.5 within urban areas, there is a high correlation between concentrations of outdoor and both indoor and personal PM_2.5 [56]. Moreover, we use a tract-specific measure of PM_2.5 concentrations because we do not have access to respondent addresses. However, because there is relatively little micro-scale spatial variability in PM_2.5 across small areas like a census tract, we expect the tract-level measures we used closely match concentrations at the respondent’s address. Importantly, we do not capture long-term, or lifetime, pollution exposure. Our measure may, however, reflect some degree of longer-term air pollution because pollution concentrations are highly correlated over time [56], and many respondents reported living in the same neighborhood for at least a decade.

This work was supported with funding from the National Institutes of Health National Institute on Aging [grant numbers: R00AG039528, P30 AG043073]. The pollution data were additionally financially supported through NIA grant number R21AG045625. No funding providers played a role in study design/conduct, analysis/interpretation of data, or manuscript preparation.

Authors’ disclosures available online (https://www.j-alz.com/manuscript-disclosures/20-0765r1).

The supplementary material is available in the electronic version of this article: https://doi.org/10.3233/JAD200765.