The links of fine airborne particulate matter exposure to occurrence of cardiovascular and metabolic diseases in Michigan, USA

Air pollutants, particularly airborne particulate matter with aerodynamic diameter < 2.5μm (PM2.5), have been linked to the increase in mortality and morbidity associated with cardiovascular and metabolic diseases. In this study, we investigated the dose-risk relationships between PM2.5 concentrations and occurrences of cardiovascular and metabolic diseases as well as the confounding socioeconomic factors in Michigan, USA, where PM2.5 levels are generally considered acceptable. Multivariate linear regression analyses were performed to investigate the relationship between health outcome and annual ground-level PM2.5 concentrations of 82 counties in Michigan. The analyses revelated significant linear dose-response associations between PM2.5 concentrations and cardiovascular disease (CVD) hospitalization. A 10 μg/m3 increase in PM2.5 exposure was found to be associated with a 3.0% increase in total CVD, 0.45% increase in Stroke, and a 0.3% increase in Hypertension hospitalization rates in Medicare beneficiaries. While the hospitalization rates of Total Stroke, Hemorrhagic Stroke, and Hypertension in urbanized counties were significantly higher than those of rural counties, the death rates of coronary heart disease and ischemic stroke in urbanized counties were significantly lower than those of rural counties. These results were correlated with the facts that PM2.5 levels in urbanized counties were significantly higher than that in rural counties and that the percentage of the population with health insurance and the median household income in rural counties were significantly lower. While obesity prevalence showed evidence of a weak positive correlation (ρ = 0.20, p-value = 0.078) with PM2.5 levels, there was no significant dose-response association between county diabetes prevalence rates and PM2.5 exposure in Michigan. In summary, this study revealed strong dose-response associations between PM2.5 concentrations and CVD incidence in Michigan, USA. The socioeconomic factors, such as access to healthcare resources and median household income, represent important confounding factors that could override the impact of PM2.5 exposure on CVD mortality.

Introduction Air pollution is a sustained world-wide public health concern for the general population. Accumulating evidence has linked exposure to high-levels of airborne particulate matter in fine and ultrafine ranges (aerodynamic diameter < 2.5 μm, PM 2.5 ) to increased mortality and morbidity associated with cardiovascular and metabolic diseases [1][2][3][4][5]. There is a linear doserisk relationship between PM 2.5 concentrations and occurrence of cardiovascular or metabolic disease, even in the countries within guidelines for EPA PM 2.5 exposure limits [6]. Long-term PM 2.5 exposure increases the risk of cardiovascular, metabolic, and neurodegenerative diseases to an even greater degree and effects may be even more pronounced in susceptible sub-populations such as older adults, those with lower socioeconomic status, and individuals with preexisting conditions. Most recently, epidemiological studies indicated that short-term exposure to polluted air, even at levels generally considered "acceptable," may impair mental ability in elderly people in US [7]. Long-term exposure to PM 2.5 is associated with an increased risk of anosmia, the inability to smell [8]. In developing countries, such as China, India, and Latin America, where daily and annual PM 2.5 levels range from 100 to 200 μg/m 3 , detrimental effects of PM 2.5 exposure on public health have been grossly underestimated. Traffic-related PM 2.5 is a complex mixture of particles and gases from gasoline and diesel engines, together with dust from wear of road surfaces, tires, and brakes [9,10]. Airborne PM 2.5 exhibits an incremental capacity to penetrate to the most distal airway units and potentially the systemic circulation [11,12]. Studies have demonstrated that these PM 2.5 particles cause cytotoxic effects, especially upon forming complexes of PM 2.5 particles rather than as a single or few particles [11]. Recent studies by our group and others in the field have addressed that traffic-related PM 2.5 may promote cardiovascular and metabolic diseases, possibly by: exaggerating systemic inflammation [1,4,5], causing oxidative and endoplasmic reticulum (ER) stress damages [13][14][15], and disrupting energy homeostasis [4,16]. While the link between PM 2.5 pollution and occurrence of cardiovascular and metabolic diseases has been established, dose-response associations between PM 2.5 concentrations and the subtypes of CVD or metabolic diseases, especially in the areas where PM 2.5 levels are generally considered "acceptable", have not been precisely defined.
In this study, we investigated the dose-risk relationships between PM 2.5 concentrations and occurrence of cardiovascular and metabolic diseases in Michigan, USA, where the levels of PM 2.5 are generally considered "acceptable" compared to the developing countries. Michigan is a representative state of the Midwestern region of the USA with a complex environmental and socioeconomic background, representing an ideal state for studying complex relationships between PM 2.5 exposure and health outcome. The state of Michigan possesses a typical urban/ rural county status and racial diversity as well as counties of specific health interest, such as Wayne County. Particularly, the great Detroit area possesses a number of industrial facilities and transportation routes. Residents in this area form a racially diverse community, including approximately 40% Hispanic, 27% African American, 30% White, Arabic, Native American, and others within a highly urbanized, industrial environment dominated by auto factories, steel mills, metal finishing industries, and waste processing. Our studies reveal that the levels of PM 2.5 exposure were significantly associated with increased hospitalization rates of cardiovascular diseases (CVD), Stroke, and Hypertension in Michigan. Interestingly, while the hospitalization rate of CVD in urbanized counties was higher than that in rural counties, the death rates of CVD in rural counties were significantly higher than that in urbanized counties, implicating the impact of socioeconomic factors, such as access to healthcare resources and median household income, in conjunction with air pollution-associated mortality and morbidity.

Data collection
Annual data was collected for county Diabetes prevalence rate from 2010 to 2016 from the CDC National Health Interview Survey (NIHS) (https://gis.cdc.gov/grasp/diabetes/ DiabetesAtlas.html#). Prevalence was deemed as a more appropriate measure for rates of diabetes, as diabetes is a chronic disease treated on an outpatient basis, whereas the other diseases studied tend to be more acute events, thus hospitalization rates would not be an accurate representation of diabetes rates within Michigan. Diabetes death rate data was obtained using Michigan Department of Health and Human Services (MDHHS) yearly county mortality datasets (https://www.mdch.state.mi.us/osr/deaths/DiabetesUS.asp). All other health outcome data (hospitalizations and deaths) was retrieved from the CDC's Division for Heart Disease and Stroke Prevention (DHDSP) (https://www.cdc.gov/dhdsp/index.htm). Based on the US Census Bureau database, 2016 Michigan population estimate is 9,950,571. The number of 2016 Michigan county Original Medicare Beneficiaries is 1,270,636. Biannual data was collected for Total Cardiovascular Disease (CVD), Coronary Heart Disease (CHD), Heart Disease (HD), Total Stroke, Ischemic Stroke, and Hemorrhagic Stroke hospitalizations and deaths as well as Hypertension hospitalizations. Death rate data was not available for both hypertension and obesity, likely because they are generally considered secondary causes of death rather than the primary cause of death. All hospitalization data was collected for Medicare Beneficiaries ages 65+ for all races and genders and was reported as Hospitalizations per 1,000 Medicare Beneficiaries; this data will be referred to as "hospitalizations" throughout the study. Death data was collected for all ages, races and genders, and was reported as Deaths per 100,000 population (all ages, races, and genders); this data will be referred to as "deaths" throughout the study. Hospitalization rate and death rate are calculated as follows: Death Rate % ð Þ ¼ Deaths 100; 000 people in county x 100% ð2Þ International Classification of Diseases, 9 th Revision, Clinical Modification (ICD-9-CM, for hospitalizations before 2015) and ICD-10-CM (for hospitalizations after 2015) codes for each health outcome are listed below in S1 Table. All data for the confounding variables mentioned below were also obtained from the CDC's DHDSP. Data for 82 counties in the state of Michigan was collected. This included all Michigan counties with the exception of Keweenaw County, Michigan, which was excluded due to lack of sufficient data. Annual concentrations of ground-level fine particulate matter (PM 2.5 , measured in μg/m 3 ) using high-resolution 0.01˚x0.01˚gridded data retrieved from multiple satellite, surface-level monitor, and simulated data sources (S1 Fig) [17]. These gridded datasets were processed into county-wise measurements across the state of Michigan by averaging all grid values whose coordinates are contained within a respective county polygon. Coordinates for county polygons were acquired from the Homeland Infrastructure Foundation-Level Data (HIFLD) public domain (https://hifld-geoplatform.hub.arcgis.com/).

Statistical analyses
Linear regression analysis was used to evaluate for strength of association between each health outcome and PM 2.5 concentration. Health outcome and PM 2.5 data were averaged over the study period, from 2010 to 2016. Two multivariate models were calculated: one with disease hospitalizations as the dependent variables and the other with disease deaths as the outcome variables. One multiple linear regression model was calculated using diabetes prevalence rate as the dependent variable. All three models were constructed using PM 2.5 concentration as the explanatory variable of interest. Additionally, each model was adjusted for the following 15 county-level confounding factors: obesity prevalence (ages 20+), leisure time physical inactivity (ages 20+), % black, % Hispanic, % above age 65, number of hospitals, % without health insurance, % with less than high school education (ages 25+), % poverty, median home income, county urban/rural status, % blood pressure medication nonadherence (Medicare Part D beneficiaries), % cholesterol-lowering medication nonadherence (Medicare Part D beneficiaries), % high cholesterol (among adults ages 18+ screened in 5 years), and % current smoker status (ages 18+) ( Table 2). The inclusion of these confounding variables controls for lifestyle, race, age, access to healthcare, socioeconomic factors, and cardiovascular health predictors-each of which has been established to be associated with cardiovascular and metabolic disease incidence and mortality.
Given the ecological nature of the study, we calculated Moran indices (Moran I) to test for spatial autocorrelation in each measured outcome with respect to county distance. County weights used in calculating each Moran I were obtained by using an inverse distance matrix, where counties were represented as a single [longitude, latitude] point within each county polygon and gridded Euclidean distances were computed between each point [18]. The Moran I significance test was then performed to test the null hypothesis of spatial independence, and all outcome variables except diabetes prevalence showed statistically significant differences from expected values (p < 0.05), indicating significant spatial autocorrelation in these outcomes. To accommodate for this spatial autocorrelation, spatial correlation was accounted for in the parameter estimates of the two multivariate regression models, as detailed below.
Eq (3) shows the multivariate linear regression model for disease hospitalization rates: Where Y 1 , . . ., Y 7 represent outcome variables: total CVD, CHD, HD, total stroke, ischemic stroke, hemorrhagic stroke, and hypertension hospitalization rates. Eq (4) shows the multivariate linear regression model for disease death rates: Where Y 1 , . . ., Y 6 represent outcome variables: total CVD, CHD, HD, total stroke, ischemic stroke, and hemorrhagic stroke death rates. Eq (5) shows the multiple linear regression model for diabetes prevalence rate: Where Y represents the outcome variable diabetes prevalence rate. In each of the equations above, α i represents the intercept, X i, PM2.5 represents the explanatory variable average county PM 2.5 concentration and β i,PM2.5 is its respective regression coefficient, X i,1 , . . ., X i,12 represents each of the 12 confounding variables and β i,1 , . . ., β i,12 are their respective regression coefficients. The errors ε i in Eqs (3) and (4) are taken to be spatially correlated with respect to county proximity (determined by using a single point within each county polygon, as described above). This correlation was computed using an exponential correlation structure, in which the correlation degrades with an exponential rate of decay with respect to distance. Regression coefficient estimates were then obtained by maximum likelihood (ML) estimation [19]. The error vector ε in Eq (5) was taken to consist of independently normally distributed errors, as there was no significant spatial autocorrelation in diabetes prevalence data, as described above. The normality assumption is validated with the adjustment of the random errors for spatial correlation, and the multivariate linear regression models detailed above are deemed appropriate for our data analysis. A summary of the results of the regression analyses is shown in Table 1. Table 2 shows descriptive statistics for each of the variables included in the study for all 82 counties (Total), as well as for urban and rural counties. Non-parametric Wilcoxon Rank Sum tests were used in comparing means of urban vs. rural counties. P-values < 0.05 are considered statistically significant in this study. Lastly, Spearman's correlation coefficients were calculated between all variables and are shown as a correlation matrix in S2 Table. All statistical analyses in this study were performed using R (version 3.6.2).

A dose-response association between PM 2.5 exposure and hospitalization of cardiovascular disease, stroke, and hypertension, but not diabetes, exists in the state of Michigan
Health outcome data for each of the 82 Michigan counties studied was plotted against average county PM 2.5 concentration, and Spearman's correlation coefficient (ρ) was calculated for each Table 1. A summary of the results of the regression analyses. Regression coefficient (β) for PM 2.5 is shown for each regression model, as well as the p-value for each variable. P-value < 0.05 is considered a statistically significant association. Column a shows results of multivariate multiple regression for hospitalization/death rates as well as multiple regression results for diabetes prevalence. Coefficients of confounding variables in the models are not shown. Colum b shows results of univariate simple regression with respect to each disease outcome variable, with PM 2.5 concentration as the explanatory variable.  Unlike the hospitalizations, deaths of cardiovascular disease did not appear to be significantly correlated with PM 2.5 levels in Michigan (Fig 3). Of note, however, total  [20][21][22][23]. However, diabetes prevalence rates showed no correlation (ρ = 0.04) with county PM 2.5 concentrations in Michigan, while obesity prevalence rates did show a weak positive correlation (ρ = 0.20, p-value = 0.078) with county PM 2.5 concentrations, though it was not statistically significant (Fig 2). We performed multivariate multiple linear regression analyses to assess for associations between ground-level PM 2.5 concentrations and county health outcome data for the years 2010 to 2016. These models were adjusted for major health determinants that are well-known to be linked to cardiovascular, cerebrovascular, and metabolic diseases, such as obesity and sedentary lifestyle, race, socioeconomic status, access to healthcare, hyperlipidemia, and medication nonadherence. Adjusting for these factors allows for correction of the concurrent impact of confounding factors, as well as comparison with simple linear regression results to analyze the degree of distortion caused by these factors. As shown in Table 1, all cardiovascular disease and cerebrovascular disease hospitalizations were found to be significantly positively associated with PM 2.5 when accounting for the confounding factors. For each 1 μg/m 3 increase in PM 2.5 concentration, it was found that total CVD hospitalizations increased by 3.006 hospitalizations per 1,000 Medicare Beneficiaries (3.01% increase in hospitalization rate per 10 μg/ m 3 , p-value = 0.011), while cardiovascular disease subsets HD hospitalizations and CHD hospitalizations increase by 1.039 (1.04% increase in hospitalization rate per 10 μg/m 3 , pvalue = 0.010) and 2.604 (2.64% increase in hospitalization rate per 10 μg/m 3 , p-value = 0.011), respectively. Furthermore, a 1 unit increase in PM 2.5 concentration was also found to be associated with an increase in the hospitalizations of total stroke and its subsets ischemic stroke and hemorrhagic stroke by 0.448 (0.45% increase per 10 μg/m 3 , p-value < 0.001), 0.371 (0.37% increase per 10 μg/m 3 , p-value = 0.002), and 0.039 (0.04% increase per 10 μg/m 3 , pvalue = 0.004) hospitalizations per 1,000 Medicare Beneficiaries, respectively. Hypertension hospitalizations were also found to increase by 0.290 for each 1 μg/m 3 increase in PM 2.5 concentration (0.29% increase per 10 μg/m 3 , p-value = 0.025). Diabetes prevalence rate had no significant association with PM 2.5 concentration.

Multivariate Multiple Regression
Of the diseases studied, only Total CVD death rates were found to be significantly associated with PM 2.5 concentration, while mortality rates of CVD subsets CHD and HD and all stroke subsets showed no significant association with PM 2.5 when adjusting for confounding factors (Table 1, Fig 3). A 1 μg/m 3 increase in county PM 2.5 concentration was associated with an increase in Total CVD deaths by 4.537 per 100,000 people (0.05% increase in death rate per 10 μg/m 3 , p-value = 0.020) ( Table 1). Heart disease mortality rates showed some association with average county PM 2.5 (β = 3.382, p-value = 0.084), but this association was not statistically significant. Though Ischemic Stroke deaths were found to have a significant negative association by simple linear regression (p-value = 0.002), this association disappeared once the model was adjusted for confounding factors (p-value = 0.659). This indicates that the strong negative correlation between Ischemic Stroke deaths and PM 2.5 concentration (Fig 3) noted above is likely due to interactions between PM 2.5 concentration and confounding factors resulting in a potentially misleading apparent association between PM 2.5 and Ischemic stroke mortality.  This was also seen with diabetes deaths, where simple linear regression showed a significant negative association (p-value = 6.64 x 10 −6 ) which disappeared in the multiple linear regression model which adjusted for confounding factors (p-value = 0.807) ( Table 1).

Urbanized counties showed significantly higher PM 2.5 concentrations and hospitalization rates of stroke and hypertension than rural counties
We compared the hospitalization and death rates of cardiovascular and metabolic disease between urbanized and rural counties in Michigan. As shown in Fig 4 and Table 2, urban counties were found to have significantly higher PM 2.5 levels than rural counties (pvalue < 0.001). Disease hospitalization rates for Total CVD (p-value = 0.088), HD (pvalue = 0.088), Total Stroke (p-value = 0.002), Ischemic Stroke (p-value = 0.051), Hemorrhagic Stroke (p-value < 0.001), and Hypertension (p-value = 0.008) in urban counties were higher than those in rural counties, while CHD hospitalization rates and diabetes prevalence rates were similar in urban vs. rural counties. Opposite to hospitalization rates, CHD, Total Stroke, and Ischemic Stroke death rates were significantly lower in the urban counties, compared to those in the rural counties (p-values = 0.008, 0.062, and 0.005, respectively) (Fig 4, Table 2).
We compared values of various socioeconomic factors in urban versus rural counties to further unveil possible explanations for the above-mentioned trends in hospitalization and death rates in urban compared to rural counties. Percent Black and Hispanic populations (pvalues < 0.001) were significantly larger in urban compared to rural counties. Interestingly, though important population-level cardiovascular disease predictors, such as elderly age (above 65), high cholesterol, and current smoker percentages, were significantly higher in rural counties, cholesterol lowering medication and blood pressure medication nonadherence were significantly higher in urban counties (all p-values < 0.001). The number of hospitals on average (p-value < 0.001), percentage of the population with health insurance (pvalue < 0.001), and median household income (p-value < 0.001) were all significantly higher in urban counties compared to rural counties ( Table 2), indicating that residents of urban counties in Michigan had better access to healthcare as well as greater resources to seek medical aid with.

Discussion
In this retrospective study, we confirmed the link between PM 2.5 concentrations and CVD and cerebrovascular diseases and made significant expansions on past studies to show strong doseresponse links between PM 2.5 concentrations and CVD subtypes, Ischemic Stoke, Hemorrhagic Stroke, and Hypertension in Michigan, USA, where the levels of PM 2.5 are generally considered "acceptable." While the EPA's annual standard for PM 2.5 levels is 12 μg/m 3 [24], the counties we analyzed had, on average over the study period, a mean PM 2.5 level of 8.14 μg/ m 3 and a maximum of 12.08 μg/m 3 (Table 1). We demonstrated a strong and significant doseresponse association, in which the level of PM 2.5 exposure is positively associated with the occurrence of both CVD and cerebrovascular diseases, despite the fact that Michigan counties were on average within the EPA standard for PM 2.5 . Thus, though levels below this range may be less harmful, they are still significantly associated with disease. This is a significant finding, as it not only highlights the dramatic health consequences even small levels of PM 2.5 may carry. The strong dose-response links between PM 2.5 concentrations and Ischemic Stroke, Hemorrhagic Stroke, and Hypertension, among the CVD subtypes, are significant, as these associations have not been precisely defined in the past.
The state of Michigan has a complex urban/rural county status and diverse communities with variable healthcare resources and household income. Michigan represents an ideal place to study the complex impact of environmental and socioeconomic factors in public health. We found that percent Black and Hispanic populations are significantly higher in urban compared to rural counties in Michigan. Furthermore, we found that cholesterol and blood pressure controlling medication nonadherence rates were significantly higher in urban counties. Black and Hispanic races and medication nonadherence have been well documented to be associated with worsened cardiovascular health outcomes. This is further evidenced by our findings of their highly positive correlations with disease hospitalization rates in our study (S2 Table). Though we note that these urban risk factors are directly opposed by the significantly higher percentages of elderly aged residents, percentages of current smokers, and prevalence of high cholesterol in the rural counties, it is arguable that the urban risk factors were seen to have much stronger correlations with hospitalization rates (S2 Table). Thus, in addition to the strong positive associations we have shown between average county PM 2.5 levels and cardiovascular and cerebrovascular disease hospitalization rates, the factors of race and medication nonadherence may provide a partial explanation to our findings that the hospitalization rates of cardiovascular and cerebrovascular diseases were significantly higher in urban compared to rural counties.
Importantly, our studies revealed that the associations between disease death rates and PM 2.5 exposure may not be consistent with the associations seen between hospitalization rates and PM 2.5 among these diverse populations. This was especially seen when observing the strong negative correlation between Ischemic Stroke deaths and PM 2.5 concentration. Though Ischemic Stroke death rate showed a significant negative correlation with PM 2.5 concentration, regression analysis adjusted for confounders showed no association between the two variables. Thus, possible explanation for the negative correlation seen is that counties with higher PM 2.5 tend to be more developed. We used urban status as a measure of the degree of development and found that PM 2.5 levels were significantly higher while Ischemic Stroke death rates were significantly lower in urbanized counties compared to rural counties. Furthermore, the number of hospitals on average, the percentage of the population with health insurance, and median household income were significantly higher in urban counties; thus, urban counties have more healthcare resources and greater access to healthcare than rural counties. As timely medical attention is critical to Ischemic Stroke survival, this is a likely explanation for the significantly lower death rates of Ischemic Stroke in urban counties, thus resulting in the observed negative correlation between Ischemic Stroke death rates and PM 2.5 concentrations. Simple linear regression did show a strong negative association between Ischemic Stroke deaths and average PM 2.5 concentration, however this association completely disappeared in the multivariate multiple linear regression model which corrected for confounding factorsthis is further evidence that the correlation between Ischemic Stroke and PM 2.5 levels is due to distortion by confounding factors, such as access to healthcare. Similar reasoning can also be applied to the other diseases studied. While the hospitalization rates for total CVD, HD, total stroke, hemorrhagic stroke, and hypertension were significantly higher in urban counties, the death rates of CHD and total stroke were all significantly lower in urban counties than in rural counties. This, again can be attributed to the decreased access to healthcare resources in rural counties. This observation suggests that the access to healthcare resources and household income could be important factors that mitigate the impact of environmental factors, such as PM 2.5 exposure, in CVD-associated mortality.
It has been reported that PM 2.5 exposure is associated with the prevalence of metabolic syndrome, particularly diabetes mellitus [20][21][22][23]. However, our studies showed that PM 2.5 may not be associated with the prevalence of diabetes in a state where the link between PM 2.5 and CVD and obesity was confirmed. The complex factors involving this discrepancy remains to be elucidated in future studies.
The conclusion that PM 2.5 is a causative risk factor for CVD is consistent with the previous studies with the populations in Asia [25,26], Europe [27,28], and USA [29,30]. The strengths of our study include a large sample size, investigation of CVD subtypes, and adjustment for and analysis of the effects of multiple confounding factors, including obesity, race, health care resources, household income, and variables directly linked to cardiovascular health, such as elderly age, smoking, high cholesterol prevalence, and blood pressure and cholesterol controlling medication nonadherence. A multivariate multiple linear regression model was used to assess for associations between health outcomes and PM 2.5 concentrations while adjusting for these confounders. Some limitations are present in our study. It is a retrospective, CDC databased study drawn from survey of public domain but not from the community. A potential bias in case-control studies is discerning the temporal relationship between risk factors and clinical outcomes, because of the complex relationships between PM 2.5 exposure and disease occurrence. Additionally, the assessment of some outcomes and covariates relies on selfreporting that may involve unknown misclassifications. Individual variations in long-term exposures to PM 2.5 are not trackable. Due to the nature of self-reporting and the dependency on the existing state records, we were limited in addressing the issues of unknown misclassification and individual variations. Nevertheless, to circumvent the weaknesses related to our study, we conducted the analysis with multiple CVD-related risk factors and demographics as well as socioeconomic confounding factors. Indeed, the positive correlation between risk factors and CVD indicated by our analyses were consistent with the established conclusions [31]. This validated the accuracy of our analyses, despite the use of retrospective data, in predicting the relationship between PM 2.5 concentrations and cardiovascular and metabolic diseases.

Conclusion
The hospitalization rates of Total CVD, Stroke and Hypertension are significantly associated with the levels of PM 2.5 exposure in Michigan, USA, where the air quality is generally considered as acceptable. While the hospitalization rates of CVD, Stroke, and Hypertension in rural counties were significantly lower than those in urbanized counties, the death rates of CVD and Stroke subtypes were significantly higher in rural counties, partially attributed to the decreased access to healthcare resources and lower median household income. These findings indicate that the socioeconomic factors, such as access to healthcare resources and median household income, are significantly associated with the incidence of air pollution-associated mortality and morbidity.
The findings of the current study indicate that PM 2.5 exposure in the USA is an important and major risk factor for both CVD and cerebrovascular disease incidence and mortality. Development of specific guidelines for the prevention and treatment of CVD and cerebrovascular disease needs to consider these findings. Both control of PM 2.5 levels and improvement in healthcare resources should be warranted. Such guidelines should outline appropriate and safe treatment to prevent CVD and cerebrovascular disease in patient populations that are at increased risk due to levels of PM 2.5 exposure, with the ultimate goal of preventing progressive disease development within the population.