Ambient Air Pollution and Mortality among Older Patients Initiating Maintenance Dialysis

Elevated levels of fine particulate matter (particulate matter with diameter<2.5 µm [PM_2.5]) are a well-recognized risk factor for cardiovascular mortality [1] and are associated more broadly with an increased risk of death and a reduced life expectancy [2‒5]. Older adults are one of the most vulnerable groups to the harmful effect of air pollution [1, 6, 7]. It was estimated that more than 2 million deaths among older adults were attributed to ambient particulate matter globally in 2015 [8].

Furthermore, there is a high global burden of kidney disease that is attributable to ambient air pollution [9, 10]. Experimental studies showed that pollutants in the air such as diesel exhaust and heavy metal could promote oxidative stress, inflammation, and DNA damages in renal tissues [11, 12]. In the USA, elevated levels of fine particulate matter are associated with incident CKD, decline in estimated glomerular filtration rate, and progression to kidney failure [13, 14]. In fact, proximity to major roads and increased levels of PM_2.5 lead to reduced estimated glomerular filtration rate [15].

Previous studies also suggested that patients with diabetes or pre-existing coronary disease were susceptible to the harmful effect of PM_2.5 [1, 16, 17]. Therefore, older patients undergoing dialysis were extremely vulnerable to air pollution, given their advanced age and high burden of comorbidity [18]. However, it remains unclear whether older dialysis patients who live in areas with elevated levels of fine particulate matter are at increased risk of mortality and whether the current US Environmental Protection Agency (EPA) National Ambient Air Quality Standard for annual PM_2.5 (≤12 μg/m³) [19] is sufficient to protect this vulnerable population.

Our a priori hypothesis was that older dialysis patients were at elevated risk of mortality when living in a ZIP code with high ambient air pollution exposure. Understanding the role of air pollution in mortality among older patients undergoing dialysis will help inform policy discussions and clinical practice in the USA. Therefore, we estimated the risk of mortality by level of fine particulate matter among adults aged ≥65 years who initiated dialysis in the USA.

Older kidney failure patients aged ≥65 years at dialysis initiation who started maintenance dialysis (either hemodialysis or peritoneal dialysis) between 1/1/2010 and 12/31/2016 without a previous solid organ transplant were identified from the US Renal Data System (USRDS). The USRDS is a national data system that collects, analyzes, and distributes information about CKD and kidney failure in the USA. The database includes kidney failure patient demographic and diagnosis data, biochemical values, dialysis claims and information on treatment history, hospitalization events, and physician/supplier services. The data in the USRDS originated from the Centers for Medicare & Medicaid Service, the Organ Procurement and Transplantation Network, the Centers for Disease Control, the ESRD Networks, the US Census, and select data from past and current USRDS Special Studies [20]. This study was reviewed by the Johns Hopkins School of Medicine Institutional Review Board and was determined to be exempt.

Only older patients who were diagnosed with renal failure and received their first chronic dialysis treatment were included in our analysis. Among 396,259 older patients who initiated dialysis between 2010 and 2016, we excluded those who had missing or unidentified ZIP codes or lived in ZIP codes without available air pollution data (N = 9,183). We also excluded older patients who lived in ZIP code areas without available socioeconomic status (SES) data (N = 1,333) and older patients with missing age, sex, race/ethnicity, BMI, cause of kidney failure, smoking status, nephrology care, and comorbidities (N = 1,467). Our analytic sample included 384,276 (97% of the original population) older patients.

The primary exposure of our analysis was annual PM_2.5 concentration at the ZIP code level. Ambient air pollution data gridded at 0.01° between 2010 and 2016 were obtained from NASA’s Socioeconomic Data and Application Center (SEDAC) Global Annual PM_2.5 Grids from NASA Moderate Resolution Imaging Spectroradiometer, Multi-Angle Imaging Spectroradiometer, and the Sea-Viewing Wide Field-of-View Sensor Aerosol Optical Depth with Geographically Weighted Regression [21, 22]. PM_2.5 data were then aggregated to ZIP code level by year. We linked older kidney failure patients and the PM_2.5 data through their ZIP code of residence and the year of first dialysis. In this analysis, we defined a concentration ≤12 μg/m³ as lower level and >12 μg/m³ as higher level of PM_2.5, as is suggested by the US EPA National Ambient Air Quality Standard for annual PM_2.5.

The outcome of interest in the study was all-cause mortality. The USRDS obtained death dates of the patients from multiple sources including CMS Medicare Enrollment Database, CMS forms 2746 and 2728 (1995–2005), Organ Procurement and Transplantation Network transplant follow-up form, CROWNWeb database, and Inpatient Claims [20]. Older patients were followed up from the date of their first dialysis to the date of death, date of transplant, or administrative censoring (12/31/2017), whichever came first. We also explored the association between PM_2.5 and cause-specific mortality. Given the large number of missingness in the data for cause of deaths, we only classified cause-specific mortality by 3 categories: CVD, other, and unknown.

Individual-level factors were obtained from USRDS database (patient profile and CMS-2728 Form), including demographics (age, sex, and race/ethnicity) and health-related factors (cause of kidney failure, smoking status, BMI, and nephrology care status). Considering comorbidities are potential mediators in the association between PM_2.5 and mortality, we did not include them in the models of our main analysis.

ZIP code-level SES factors included percent below 200% of the federal poverty line, mean years of education, median household income, median housing cost per month, percent non-Hispanic Black, percent Hispanics, population density, and urbanicity. Population density was obtained from SEDAC Gridded Population of the World 2010 [23]. Urbanicity of the area in the year of 2013 was retrieved from the Census Bureau TIGER/Line Shapefiles and was categorized into urban, urban clusters (incorporating places that contained at least 2,500 but less than 50,000 people within its boundaries), and rural [24]. Other ZIP code-level characteristics were abstracted from 2014 American Community Survey [25].

Baseline characteristics were summarized by levels of ZIP code-level PM_2.5 exposure (≤12 μg/m³, >12 μg/m³). Continuous variables were presented as median (IQR), while categorical variables were presented as frequency (%).

Cox proportional hazard models with robust variance accounting for the cluster effect of ZIP codes were used to investigate the unadjusted and adjusted continuous association between PM_2.5 and all-cause and cause-specific mortality among the study population. Considering the variation of air pollution by different time periods and that more than 50% of the older patients with PM_2.5 exposure level higher than 12 μg/m³ came from California, we accounted for different baseline hazard functions by the year of dialysis initiation and whether the older patients lived in California. We accounted for demographics, health-related factors, and ZIP code-level factors in our adjusted models. We fit linear spline models with a knot at 12 μg/m³ to allow for different dose-response associations below and above the current EPA National Ambient Air Quality Standard. To explore other potential nonlinear exposure-outcome associations, we additionally fit restricted cubic spline models. To investigate whether the association between ZIP code-level PM_2.5 concentration and all-cause mortality among the older dialysis population varied by age (<75 and ≥75 years), sex (male, female), race/ethnicity (non-Hispanic white, non-Hispanic Black, Hispanic/Latino, and other), history of smoking (current and noncurrent), and cause of kidney failure (diabetes, hypertension, glomerular nephritis, and other), we fit linear spline models within different subgroups.

Considering the levels and composition of PM_2.5 can vary greatly across the USA, we further investigated the association within different regions of the country. We grouped our study population by their states of residence into 8 regions (Fig. 1). Adjusted Cox proportional hazard models were used to obtain the linear association between continuous PM_2.5 and all-cause mortality in different regions, separately.

In our sensitivity analysis, we further adjusted for comorbidities (cardiovascular disease, chronic obstructive pulmonary disease, and cancer) in our linear spline models as additional potential confounders. The first 6 months after dialysis initiation was a complicated period with high mortality; we did additional analysis among those who survived the first 3 and 6 months of maintenance dialysis. Considering a large proportion of the older patients residing in areas with a higher level of PM_2.5 concentration were from California, we also conducted analyses separately for older patients who came from California and those who were from other states to see if the association was consistent between the 2 groups. Statistical analyses were conducted using Stata SE, version 15 (Stata Corp, College Station, TX, USA) and R 3.6.1 [26, 27], and a p value of less than 0.05 was considered statistically significant.

There were 384,276 older patients (across 25,100 ZIP codes) who initiated dialysis between 2010 and 2016 and were followed up for a median of 1.84 (IQR: 0.77–3.25) years. Overall, 14% were female, 63% were non-Hispanic white, 20% were non-Hispanic Black, and 11% were Hispanic/Latino. The median age at dialysis initiation was 74 (IQR: 69–80) years.

The median ZIP code-level PM_2.5 concentration for older patients was 9.17 μg/m³ (IQR: 7.63–10.78, range 0.70–23.62 μg/m³). Among our study population, 48,885 (13%) had a ZIP code-level PM_2.5 concentration >12 μg/m³. Furthermore, the distribution of ZIP code-level PM_2.5 varied across the USA, with the highest ZIP code-level air pollution occurring in the Far West region.

When compared to those residing in areas with lower PM_2.5 concentrations (≤12 μg/m³), older patients in areas with a higher level of concentration (PM_2.5 > 12 μg/m³) were less likely to be non-Hispanic white (49 vs. 65%) and more likely to be Hispanic/Latino (20 vs. 10%). Only 23% of the older patients living in areas with a higher level of PM_2.5 had been on nephrology care for more than 12 months at the time of dialysis initiation, compared to 32% in the older patients with lower exposure (Table 1).

The older patients in ZIP codes with higher levels of PM_2.5 concentrations were more likely to live in ZIP codes with a higher proportion of Hispanic residents (median 19.0 vs. 6.9%), population density (median 1,974/km² vs. 393/km²), and housing cost (median $1,145 vs. $928 USD/month). Furthermore, they were more likely to be in urban areas; 93.0% of the older patients in high PM_2.5 concentration ZIP codes lived in urban areas compared to 73.3% in the lower group (Table 1).

Overall, 117,091 patients residing in areas with a low PM_2.5 level and 29,995 patients in areas with high PM_2.5 died during the follow-up period, among which 84,109 (34%) died from CVD, 94,923 (38%) died from other causes, and 69,093 (28%) had unknown cause of death. A restricted cubic spline model suggested that there was nonlinear association between PM_2.5 and mortality risk and shown that the dose-response association changed when the PM_2.5 level reached ∼12 μg/m³ (Fig. 2).

In the unadjusted linear spline model, a 10 μg/m³ increase in PM_2.5 was associated with a 1.03-fold (95% CI: 1.01–1.06) increased risk of mortality when PM_2.5 was at lower level; a 10 μg/m³ increase in PM_2.5 was associated with a 1.21-fold (95% CI: 1.11–1.33) increased risk of mortality when PM_2.5 was at higher level. After accounting for demographics, health-related factors, and ZIP code-level characteristics, no association was found between PM_2.5 and mortality risk when the PM_2.5 level was below 12 μg/m³ (HR = 1.01 per 10 μg/m³ increase, 95% CI: 0.99–1.03). However, when PM_2.5 was above 12 μg/m³, 10 μg/m³ increase was associated with a 1.16-fold (95% CI: 1.08–1.25) increase in mortality risk among older dialysis patients (Table 2). At higher levels, increased PM_2.5 was associated with an increased risk of CVD mortality (HR = 1.38; 95% CI: 1.21–1.58) and unknown causes of mortality (HR = 1.33; 95% CI: 0.1.14–1.55), but not death from other cause (online suppl. Table 1; see www.karger.com/doi/10.1159/000514233 for all online suppl. material).

The association between PM_2.5 concentration and mortality varied across different subgroups of older patients initiating dialysis (Fig. 3). At higher levels, we found that older patients who were female (HR = 1.26 per 10 μg/m³ increase, 95% CI: 1.13–1.42), Black (HR = 1.31 per 10 μg/m³ increase, 95% CI: 1.09–1.59), or had diabetes (HR = 1.25 per 10 μg/m³ increase, 95% CI: 1.13–1.38) as the primary cause of kidney failure had higher PM_2.5-associated mortality risk than the overall study population. Although the risk of mortality was stronger when PM_2.5 was above 12 μg/m³ (HR = 1.19 per 10 μg/m³ increase, 95% CI: 1.08–1.32), even at lower levels, PM_2.5 was significantly associated with increased mortality risk among older patients who were aged >75 years (HR = 1.04 per 10 μg/m³ increase, 95% CI: 1.00–1.07).

Across 8 geographic regions, the Great Lakes region had the highest median ZIP code-level PM_2.5 concentration (median = 11.17 μg/m³), while the Rocky Mountains had the lowest (median = 5.77 μg/m³) (Fig. 1). Significant associations between PM_2.5 and mortality risk were only found in the Far West region (median PM_2.5 = 10.22 μg/m³), where per 10 μg/m³ increase in PM_2.5 was associated with a 1.06-fold (95% CI: 1.01–1.11) increase in mortality risk. We also found marginally significant exposure-outcome association in the Great Lakes region (HR = 1.08 per 10 μg/m³ increase, 95% CI: 1.00–1.16). No association was observed in other regions with a lower PM_2.5 level (Fig. 1).

Further adjustment for comorbidities did not significantly change the inferences from the main analyses; for every 10 μg/m³ increase, PM_2.5 was associated with a 1.14-fold (95% CI: 1.07–1.22) increase in mortality risk for higher levels, and no association was observed at lower air pollution levels. After excluding patients who died within first 3 and 6 months of maintenance dialysis, the result remained consistent. The exposure-outcome association was similar between older patients from California and other states (online suppl. Table 1). Mortality rates among our study population between 2010 and 2016 are shown in online suppl. Table 2.

In this study of 384,276 older patients initiating dialysis between 2010 and 2016 in the USA, the median ZIP code-level PM_2.5 concentration was 9.17 μg/m³, and 13% lived in a ZIP code with a PM_2.5 level >12 μg/m³. A 10 μg/m³ increase in PM_2.5 was associated with a 1.23-fold (95% CI: 1.14–1.32) increase in mortality among older patients initiating dialysis when at a higher level of PM_2.5 and a 1.05-fold (95% CI: 1.02–1.07) increase when at a lower level after adjusting for demographics and ZIP code-level SES. When further adjusted for health-related factors, the association was attenuated such that PM_2.5 was associated with mortality only at levels above 12 μg/m³ (HR = 1.15, 95% CI: 1.07–1.23). We also found that PM_2.5 was associated with elevated mortality risk among older patients who were aged >75 years at dialysis initiation even at lower levels of PM_2.5. However, older patients who were female, Black, or had diabetes as a primary cause of kidney failure were most vulnerable when exposed to higher levels of PM_2.5 (>12 μg/m³).

To our knowledge, while many cohort studies have evaluated the association between long-term exposure of air pollution and mortality risk in the general population [3, 28‒30], this is the first study to examine this relationship with a focus on older patients with kidney failure in the USA. We observed a positive association between long-term exposure to PM_2.5 and all-cause and CVD mortality among older patients initiating dialysis, which is consistent with a recent study in Hong Kong (median PM_2.5 concentration = 38.9 μg/m³). Ran et al. [31] found an interquartile change (4 μg/m³) in PM_2.5 was associated with a 14% (95% CI: 1.01–1.30) increase in mortality, which could be translated into more than 35% increase in mortality every 10 μg/m³ increase in PM_2.5, among kidney failure patients. This larger effect size observed in Hong Kong could potentially be explained by the higher levels of air pollution persisting in the region than the USA. Furthermore, patients with kidney failure in Hong Kong were quite different from those in the USA in terms of primary renal diagnosis [32]. Another recent study investigating the short-term effect of wildfire smoke conducted among hemodialysis patients in the USA found that a 10 μg/m³ increase in wildfire PM_2.5 was associated with a 1.04-fold (95% CI: 1.01–1.17) increase in the same-day mortality [33]. All these findings support that air pollution is a potential threat to the health of patients with kidney failure.

Many of the studies found linear or supralinear (characterized by a decrease in slope with an increased exposure level) concentration-response associations between PM_2.5 and mortality, where increased air pollution-associated mortality risk occurred even at very low levels of PM_2.5 [3, 29, 34‒36]. Specifically, a study conducted among the Medicare population found that per 10 μg/m³ increase in PM_2.5 was linearly associated with a 7.3% (95% CI: 7.1–7.5%) increase in all-cause mortality without any thresholds [3], while another study conducted among a Canadian census cohort observed that PM_2.5 was associated with a steep increase in relative mortality risk at levels below 5 μg/m³, with associations attenuating at higher levels [29]. However, in this study, the reverse was observed; we found that there was increased mortality risk only when PM_2.5 was >12 μg/m³, and no association when PM_2.5 was at lower levels among the overall study population. Previous studies suggested that air pollution could affect health by inducing systematic inflammation [1]. When compared to the general population, older patients with kidney failure are more likely to have a high inflammatory state [37, 38]. It is possible that at low levels of PM_2.5, the pro-inflammatory effect from air pollution is small relative to the already high state of inflammation experienced by this vulnerable population, such that there is no observed increase in mortality risk. However, when PM_2.5 increased to a specific level, the effect was large enough to lead to adverse outcomes among older patients with kidney failure. This could also explain the comparable or even stronger association we observed at higher levels of air pollution than in other US studies [3, 28, 30]. Additionally, this potential mechanism was also supported by our findings of significant exposure-outcome associations in regions with highest levels of PM_2.5 (Great Lakes and Far West).

In this study, patients >75 years with kidney failure were more vulnerable to air pollution than their younger counterparts (65–75 years), such that elevated mortality risk was observed even at low levels of PM_2.5. This finding is in line with other studies demonstrating the susceptibility of the oldest older adults in general to harmful effects of air pollution [1, 16, 17], supporting current EPA recommendations for them to take extra precautions during high-pollution days [39]. A previous study had found that older adults (aged ≥65 years) who were frail were more vulnerable to the adverse effects of air pollution than their nonfrail counterparts [40]. The higher vulnerability to air pollution among older dialysis patients observed in our study may be attributable to frailty, a condition characterized by higher vulnerability to stressors [41], considering that the prevalence of frailty increases with age, and more than 25% of the kidney transplant candidates aged ≥65 years were frail [42‒44]. Meanwhile, the higher prevalence of comorbidities including cardiovascular disease and respiratory disease among older patients initiating dialysis could also contribute to the vulnerability of this population [1].

In this population of older dialysis patients, we found that the effect of air pollution on mortality differed by racial/ethnic groups, with older non-Hispanic Black patients being more vulnerable to the effects of PM_2.5 at levels above the current EPA standard, and non-Hispanic white patients at increased risk of mortality even at levels below the current EPA standard. One potential cause for this difference could be the geographical distribution of racial/ethnic groups as the composition of particulate matter varies greatly by region [45] and the particle compositions were found to be associated with mortality [46, 47]. Furthermore, Black patients are also more likely to be disproportionately affected by other risk factors such as low SES, which may increase their vulnerability to environmental stressors and thus contribute to this observed difference, yet the role of race/ethnicity on the association between air pollution and mortality remains inconclusive.

Older patients initiating dialysis are a vulnerable group with high baseline mortality [18], such that even a small increase in relative risk would result in a relatively large increase in absolute mortality. Consequently, it is especially critical for this population to avoid unnecessary excess mortality risk from PM_2.5 exposure. Although air pollution has long been considered to be a threat to health, recent studies on air quality awareness among US adults found that only 29.2% of participants believed that air pollution was bad and 14.7% would change their behavior when there was low air quality; similar results were found among participants with heart disease [48, 49]. Under these circumstances, there is a need for healthcare providers, like geriatricians, to discuss the potential harms of air pollution with their vulnerable older patients and provide recommendations for behavioral changes. This study identified different subgroups of older patients initiating dialysis with varying vulnerability to air pollution, which could help geriatricians and nephrologists target air pollution exposure mitigation strategies for older patients most in need of them.

Our study has the strength of using the geographically diverse, registry-based data capturing the majority of older patients with kidney failure in the US. Data from a national registry database allowed us to adjusted for key health-related factors which were most relevant to older patients initiating dialysis. Additionally, the large sample size enabled us to do individual analysis in different subgroups and regions.

This study has its limitations in the measurements of exposure and confounders and ascertainment of outcomes. Air pollution concentration was assessed at ZIP code levels, and we were not able to account for indoor air pollution; this could result in potential exposure misclassification as variations in PM_2.5 can exist within a ZIP code and this level may not reflect all locations where older patients may spend time. Even within the same area, individual exposure of PM_2.5 could be affected by many factors such as proximation to major roadways and outdoor activity levels. Due to the limited availability of other data, we only investigated the effect of PM_2.5 in this study and were not able to control for other pollutants such as ozone and NO₂. Therefore, part of the observed effect of might be contributed to other types of pollutants. However, previous studies suggested that further controlling for other pollutants did not significantly change the estimated effects of PM_2.5 [3, 50]. We only have the ZIP code and health-related information for the patient at the time of listing, such that we were not able to account for time-varying exposure of PM_2.5 and other time-varying confounding. The ZIP code information was obtained solely from the USRDS patient profile, and we were not able to confirm whether the patients actually resided within the ZIP code area at the time of dialysis initiation and during the follow-up. However, the ZIP code data had been used to link ZIP code-level exposure in other studies [51, 52], and these are the only data we have regarding the patients’ area of residence. For SES confounders, we were only able to measure at the ZIP code level, which could lead to residual confounding. In terms of outcome ascertainment, the large number of missingness in cause-of-death data limited our ability to investigate and interpret the association between PM_2.5 and cause-specific mortality.

In conclusion, ambient air pollution, measured by PM_2.5, was associated with all-cause mortality among older patients initiating dialysis. More notably, increased mortality risk was observed among patients aged >75 years, even at levels below the current EPA National Ambient Air Quality Standard for PM_2.5. Our results suggested that the current EPA standard for PM_2.5 might not be enough to protect the most vulnerable populations. Geriatricians and nephrologists should consider ambient air pollution as an important environmental exposure and implement ambient air pollution exposure mitigation strategies for their vulnerable older patients and actively engage in advocacy for population-level solutions.

The data reported here have been supplied by the United States Renal Data System (USRDS). The interpretation and reporting of these data are the responsibility of the author(s) and in no way should be seen as an official policy or interpretation of the US government.

No authors report a conflict of interest.

Funding for this study was provided in part by the National Institute of Diabetes and Digestive and Kidney Disease (NIDDK) and the National Institute on Aging (NIA) (Grant No. R01DK120518 [PI: Mara McAdams-DeMarco], R01AG055781 [PI: Mara McAdams-DeMarco], K01AG064040 [PI: Chu], and K24AI144954 [PI: Dorry Segev]).

M.M.D., Y.F., M.J., and D.L.S. designed the study; all authors drafted and revised the manuscript; all authors approved the final version of the manuscript.

The data underlying this article were provided by the USRDS under license/by permission. Data will be shared on request to the corresponding author with permission of the USRDS.