Ambient Air Pollution Exposure and Adverse Outcomes Among Medicare Beneficiaries With Heart Failure

Abstract

Background

Exposure to fine particulate matter (<2.5 um, particulate matter with an aerodynamic diameter <2.5 microns [PM_2.5]) has been implicated in atherogenesis. Limited data in animal studies suggest that PM_2.5 exposure leads to myocardial fibrosis and increased incidence of heart failure (HF). Whether PM_2.5 is associated with adverse outcomes in patients with preexisting HF has not been widely studied.

Methods and Results

In this retrospective cohort study, Medicare patients hospitalized with first HF between 2013 and 2020 were identified from the Medicare Provider Analysis and Review Part A 100% files. Patients were linked with integrated estimates of ambient PM_2.5 obtained at 1×1 km using the zip code of participants' residence. The study outcomes were all‐cause death, HF, and all‐cause readmissions burden. A total of 2 599 525 patients were included in this study, with 6 321 731 person‐years of follow‐up. Mean PM_2.5 was 7.3±1.7 μg/m³. Each interquartile range of PM_2.5 was associated with 0.9% increased hazard of all‐cause death, 4.5% increased hazard of first HF readmission, 3.1% increased risk of HF hospitalization burden, and 5.2% increase in all‐cause readmission burden, after adjusting for 11 sociodemographic and medical factors. Subgroup analyses showed that the effects were more pronounced at levels <7 μg/m³ and in patients aged <75 years, Asians, and those residing in rural areas.

Conclusions

Ambient air pollution is associated with higher risk of adverse events in Medicare beneficiaries with established HF. These associations persist below the National Air Quality Standards (12 μg/m³), supporting that no threshold effect exists for health effects of air pollution exposure.

Nonstandard Abbreviations and Acronyms

PM_2.5

particulate matter with an aerodynamic diameter <2.5 microns

PURI‐HF

Air Purifiers on Heart Failure

SDI

social deprivation index

Clinical Perspective.

What Is New?

• Ambient particulate matter (particulate matter with an aerodynamic diameter <2.5 microns) exposure is associated with higher risk of death and heart failure readmissions in Medicare beneficiaries with heart failure, independent of social deprivation.

• The effects were more pronounced below air quality standard threshold of 12 μg/m³.

What Are the Clinical Implications?

• Studies should examine whether reducing personal exposure to particulate matter with an aerodynamic diameter <2.5 microns using practical interventions improves outcomes.

Heart failure (HF) remains a major cause of morbidity and death in the United States and globally, despite advancement in therapeutics, with >200 000 patients dying of HF annually in the United States. ¹ While preventive efforts have focused on controlling traditional risk factors, such as diabetes, obesity, hypertension, and renal disease, a significant portion of HF morbidity and death is attributed to less defined social and environmental factors. Despite their importance, socioenvironmental determinants of HF have received little attention.

Air pollution is the most important environmental risk factor, estimated to result in >9 million deaths globally, most of which are due to cardiovascular causes. ² Patients with preexisting chronic illness and older individuals are at potentiated risk of poor outcomes in response to air pollution exposure. An extensive body of literature has linked air pollution with cardiovascular disease (CVD), especially myocardial infarction and cardiovascular death. ³ , ⁴ Prior smaller studies, mostly outside the United States, have shown that air pollution is associated with HF incidence. ⁵ Small studies have also linked air pollution with surrogate markers of HF, such as lower peak oxygen consumption during exercise testing, ⁶ brain natriuretic peptides, ⁷ and diastolic function. ⁸ The biological mechanisms underlying the cardiovascular toxicity of particulate matter with an aerodynamic diameter <2.5 microns (PM_2.5) are multifactorial, including oxidative stress, inflammation, autonomic dysfunction, and thrombosis, and evidence from animal studies suggest PM_2.5 exposure leads to cardiac remodeling, myosin heavy chain isoform switch, myocardial fibrosis, and diastolic dysfunction. ⁹ , ¹⁰ While the epidemiological and experimental evidence clearly implicates PM_2.5 in cardiovascular morbidity, critical knowledge gaps remain. In particular, limited data exist on the effects of PM_2.5 on clinical outcomes among patients with established CVD, who may be at heightened susceptibility. Additionally, it remains unknown whether PM_2.5 air pollution is associated with poor outcomes in patients with preexisting HF, especially at low exposures, such as those seen in the United States, remains unknown. ¹⁰ Further, the intersection between area‐based social deprivation and PM_2.5 exposure in HF is not studied at the national level. Our study aims to address this gap by investigating the associations between PM_2.5 exposure and adverse events in a national cohort of older adults with preexisting HF.

In this national analysis, we sought to investigate the association between air pollution and adverse outcomes in Medicare beneficiaries with HF and its intersection with social deprivation.

Methods

Because of the sensitive nature of the data collected for this study, requests to access the data set from qualified researchers trained in human subject confidentiality protocols may be sent to Medicare Research Data Assistance Center at resdac@umn.edu. Strengthening the Reporting of Observational Studies in Epidemiology cohort reporting guidelines were followed. ¹¹ The Cleveland Clinic Institutional Review Board approved the use of the Medicare data set.

Study Cohort

We used Medicare Provider Analysis and Review 100% inpatient files to identify Medicare beneficiaries aged ≥65 years, who were admitted with a new diagnosis of HF in the primary position from January 2013 through September 2020 using International Classification of Diseases (ICD) codes (I50, 428, I110, I130’, ‘I132, 40201, 40211, 40291 40401, 40411, 40491, 40403, 40413, 40493). Patients with <1 year Medicare Fee‐for‐Service coverage before HF admission date, patients who were admitted and discharged on the same date, received palliative care within 30 days, left against medical advice, or were discharged to hospice were excluded. We also excluded patients who resided in a zip code with <10 Medicare beneficiaries or zip codes with missing PM_2.5 concentration.

Study Variables

Demographic information including age, sex, race and ethnicity, and residential zip code were extracted from Medicare Beneficiaries Summary files. The main exposure of interest in our study was the annual PM_2.5 particulate air pollution reported at the zip code level. We used validated PM_2.5 exposure estimates developed by the Atmospheric Composition Analysis Group. ¹² These estimates combine information from satellite remote sensing, chemical transport modeling, and calibration to ground‐based observations to generate concentration estimates. Data are provided in 1×1 km grids, which we imported to an open‐source geographical information system software QGIS version 3.16 and mapped to the 2018 zip code boundaries from the US Census Bureau. We then calculated mean zip PM_2.5 exposures using zonal statistics in QGIS. Each patient was assigned the 2016 annual exposure (the midpoint of our study duration). The year‐to‐year zip‐level PM_2.5 exposures are highly correlated (R>0.99), and thus a single‐year exposure was felt to be sufficient.

We used a look‐back period of 1 year and used all ICD codes submitted in that period, including the index HF admission, to ascertain patient comorbidities using Elixhauser's method. ¹³ We also adjusted for the patient's zip code census region (West, Southwest, Southeast, Midwest, and Northeast), socioeconomic distress, and rurality. A patient was considered to live in a rural area if the zip code is designated rural by the Federal Office of Rural Health Policy. ¹⁴ Socioeconomic distress was determined using the social deprivation index (SDI). The SDI is a composite measure of area‐level deprivation based on 7 demographic characteristics collected in the American Community Survey, including income less than federal poverty level, education (high school dropout rate), employment, housing (percentage living in rented or crowded housing units), household characteristics (single‐parent households), and transportation (percentage of population not owning a car). ¹⁵

Study Outcomes

The primary study outcome was all‐cause death. Secondary outcomes included first readmission with a primary diagnosis of HF, HF readmissions burden, and burden of all‐cause readmissions (defined as number of admissions per person‐time). Data on mortality rate were available through August 2021, and data on readmissions were available through December 2020. The Institutional Review Board at the Cleveland Clinic approved the study with waiver of informed consent.

Statistical Analysis

Continuous variables are presented as mean and SD if normally distributed and compared using ANOVA, or median and interquartile range (IQR) if not normally distributed and compared using the Kruskal–Wallis test. Categorical variables are presented as percentages and compared using the χ² test.

A Cox regression model was used to estimate hazard ratios (HRs) for the relationship between PM_2.5 exposure and the study outcomes, using a robust sandwich covariance matrix estimate and robust standard error estimates to account for the clustering of patients within zip codes. ¹⁶ To test the association of PM_2.5 concentration and the outcomes (heart failure rehospitalization, all‐cause death), we first modeled PM_2.5 as a continuous variable, then repeated the analysis using tertiles of PM_2.5 concentration as a categorical variable. For each exposure variable (PM_2.5 as continuous or categorical), we fit 2 models: the first model adjusted for age, sex, and race; rurality of zip code; census region; and SDI. The second model included model 1+diabetes, hypertension, coronary artery disease, and obesity.

To visualize the association between continuous PM_2.5 and composite outcome, we built a Cox proportional hazard model with penalized smoothed spline of PM_2.5, adjusting for factors in model 2. To visualize the additive effects of PM_2.5 and SDI, we constructed a Cox proportional hazard model with penalized smoothed splines of PM_2.5 and its interaction with penalized smoothed spline of the SDI. We plotted the continuous natural log (HR) over the entire spectrum of SDI and PM_2.5 in the study cohort with all‐cause death.

For the outcome of time to first readmission with HF, subdistribution hazards Fine–Gray model accounting for the competing risk of death was used. ¹⁷ For the outcome of HF readmission burden in follow‐up, a generalized linear mixed model was used, with number of admissions as the dependent variable and log of time of follow‐up as offset term, with log link and Poisson distribution and random intercept for the zip code. For the primary study outcome, interaction was tested between the PM_2.5 concentration and age (<75 years or >75 years), sex, race, diabetes, coronary artery disease status, and quartiles of SDI.

The analysis was performed using SAS version 9.4 (SAS Institute, Inc, Cary, NC), R version 4.0.2 (R Foundation for Statistical Computing, Vienna, Austria), and Prism version 8 (GraphPad Software, La Jolla, CA). A 2‐sided P value <0.05 was considered statistically significant. QGIS was used for geospatial visualization and map creation.

Results

The final study cohort included 2 599 525 patients from 22 992 zip codes. Figure S1 shows the study cohort flowchart. The distribution of PM_2.5 and SDI are shown in Figures S2 and S3. The geospatial distribution of PM_2.5 and cohort patients are shown in Figure 1. Table 1 shows baseline characteristics for the study cohort divided into 3 groups based on tertiles of PM_2.5 exposure. Patient with higher PM_2.5 exposures were younger, more likely to be women, and less likely to be of White race. Patients with higher PM_2.5 exposure also had higher prevalence of diabetes, hypertension, and obesity and were more likely to be dual Medicaid enrolled. Patients with higher exposure were more likely to reside in urban locations, with higher socioeconomic distress as evidenced by higher SDI.

Figure 1. Distribution of annual PM_2.5 concentration and study cohort in zip code level in the United States.

Maps demonstrating annual PM_2.5 exposure and geospatial distribution of study patients (A) annual PM_2.5 and (B) zip code–level number of study patients. PM_2.5 indicates particulate matter with an aerodynamic diameter <2.5 microns.

Table 1.

Characteristics of Patients by Tertile of PM_2.5

	PM2.5 (μg/m3)			P value
	Tertile 1	Tertile 2	Tertile 3
PM2.5 (μg/m3)	5.6±1.1	7.3±0.3	9.0±1.3
Number of patients	866 497	866 511	866 517
Number of zip codes	11 086	6955	4951
Age, y	80.6 (8.7)	80.0 (8.7)	79.9 (8.9)	<0.001
Male sex, %	48.1	47.0	45.8	<0.001
Race and ethnicity, %				<0.001
White	90.8	79.5	71.0
Black	4.2	12.4	17.6
Asian	0.6	1.0	2.0
Hispanic	2.6	5.7	8.2
Native American	0.8	0.4	0.1
Other/unknown	1.0	1.0	1.1
Diabetes, %	44.2	47.9	48.4	<0.001
Hypertension, %	91.9	93.7	94.1	<0.001
Obesity, %	25.2	25.1	25.6	<0.001
Coronary artery disease, %	62.7	63.8	62.4	<0.001
Prior coronary artery bypass grafting, %	18.5	19.2	18.0	<0.001
Ischemic stroke, %	6.6	6.8	6.9	<0.001
Prior gastrointestinal bleed, %	7.1	7.5	7.6	<0.001
Depression, %	18.2	17.0	16.3	<0.001
Connective tissue disease, %	5.3	4.7	4.8	<0.001
Tumor without metastasis, %	3.8	3.8	3.8	0.2
Lymphoma, %	1.6	1.5	1.5	0.001
Liver disease, %	3.7	3.8	4.2	<0.001
Chronic kidney disease, %	49.0	49.3	51.0	<0.001
Medicaid eligible, %	20.1	24.0	24.8	<0.001
Rural residence, %	34.2	25.1	6.2	<0.001
Census region, %				<0.001
Midwest	25.0	20.7	25.9
Northeast	31.8	19.3	21.6
Southeast	16.1	43.8	25.0
Southwest	9.9	10.9	9.4
West	17.2	5.2	18.1
Social deprivation index	39 (20–61)	53 (28–74)	59 (31–83)	<0.001

PM_2.5 indicates particulate matter with an aerodynamic diameter <2.5 microgram/m³.

Primary Outcome

The continuous relationship between PM_2.5 and all‐cause death is shown in Figure 2. Every IQR of PM_2.5 was associated with a 1.1% higher risk of death (HR, 1.011 [95% CI, 1.008–1.013]), which was unchanged after adjustment in model 2 (HR, 1.009 [95% CI, 1.006–1.011]). This relationship was more pronounced in the cohort that is confined to those exposed to <7 μg/m³ of PM_2.5 (model 1: HR, 1.028 [95% CI, 1.021–1.134]; model 2: 1.025 [95% CI, 1.020–1.030]). Compared with tertile 1 of exposure, patients residing in tertile 2 and tertile 3 had 1.6% and 2.0% increased hazards of all‐cause death after adjustments in model 2 (HR, 1.016 [95% CI, 1.011–1.022]; and HR, 1.020 [95% CI, 1.014–1.025], respectively; Table 2). The associations were more pronounced in patients aged <75 years, Asian individuals, without diabetes, and without coronary artery disease. There were no differences in the association with the SDI (Figure 3). SDI and PM_2.5 exposures were additive to the hazard of all‐cause death (Figure S4). Notably, there were no significant differences in the association between PM_2.5 exposure and death on the basis of sex (Figure 3).

Figure 2. Association between PM_2.5 and study outcomes.

Association between PM_2.5 and HF admission (left) and all‐cause death (right) using penalized smoothed spline. HF indicates heart failure; HR, hazard ratio; and PM_2.5 particulate matter with an aerodynamic diameter <2.5 microns.

Table 2.

Association of PM_2.5 With Outcomes as a Continuous Variable and as a Categorical Variable (3 Groups) in Patients With New Diagnosis of HF

	HR	95% CI	P value
All‐cause death
Model 1
Per IQR PM2.5	1.011	1.008–1.013	<0.001
Per IQR PM2.5 (in those <7 μg/m3)	1.028	1.021–1.134	<0.001
Tertile 2 vs tertile 1	1.020	1.015–1.026	<0.001
Tertile 3 vs tertile 1	1.024	1.018–1.029	<0.001
≥90th percentile vs ≤10th percentile of cohort limited to those <7 μg/m3 (PM2.5 1.23–4.15 vs 6.87–7.00)	1.080	1.060–1.110	<0.001
Model 2
Per IQR PM2.5	1.009	1.006–1.011	<0.001
Per IQR PM2.5 (in those <7 μg/m3)	1.025	1.020–1.030	<0.001
Tertile 2 vs tertile 1	1.016	1.011–1.022	<0.001
Tertile 3 vs tertile 1	1.020	1.014–1.025	<0.001
≥90 percentile vs ≤10 percentile* of cohort limited to those <7 μg/m3	1.070	1.060–1.100	<0.001
First HF admission
Model 1
Per IQR PM2.5 first segment below median 7.31	1.051	1.043–1.06	<0.001
Per IQR PM2.5 second segment >7.31	1.005	0.990–1.010	0.1
Tertile 2 vs tertile 1	1.028	1.021–1.035	<0.001
Tertile 3 vs tertile 1	1.023	1.015–1.030	<0.001
≥90th percentile vs ≤10th percentile* of cohort limited to those <7 μg/m3	1.180	1.140–1.220	<0.001
Model 2
Per IQR PM2.5 first segment below median 7.31	1.045	1.037–1.053	<0.001
Per IQR PM2.5 second segment >7.31	1.004	0.998–1.011	0.2
Tertile 2 vs tertile 1	1.023	1.016–1.030	<0.001
Tertile 3 vs tertile 1	1.018	1.010–1.025	<0.001
≥90th percentile vs ≤10th percentile* of cohort limited to those <7 μg/m3	1.160	1.130–1.190	<0.001

Model 1 adjusted for age, sex, race, and Medicaid eligibility status, rurality of zip code, Census region, and social deprivation score. Model 2 adjusted for same variables as model 1 + diabetes, hypertension, coronary artery disease, and obesity. HF indicates heart failure; IQR, interquartile range; and PM_2.5, particulate matter with an aerodynamic diameter <2.5 microns.

^*10th percentile PM_2.5=1.23–4.15 μg/m³ and 90th percentile PM_2.5=6.87–7.00 μg/m³.

Figure 3. Association between PM_2.5 (per IQR) with all‐cause death in subgroups.

Association between PM_2.5 (per IQR) with all‐cause death in subgroups. Cox model 2 was fit for each subgroup, and interactions were tested between the subgroups and PM_2.5 with respect to composite outcome. Mean effect (HR, 1.011) and No effect (HR, 1) lines are shown. HR indicates hazard ratio; IQR, Interquartile range; and PM2.5, particulate matter with an aerodynamic diameter <2.5 microns.

Secondary Outcomes

The association between PM_2.5 and HF readmission was nonlinear, so analysis was done with piecewise regression below and above the median PM_2.5. The piecewise regression showed that each IQR of PM_2.5 <7.3 μg/m³ was associated with a 5.1% increase in first HF hospitalization (HR, 1.051 [95% CI, 1.043–1.060]) in model 1, which did not change after adjusting in model 2 (HR, 1.045 [95% CI, 1.037–1.053]). There was no association between PM_2.5 and first HF readmission >7.3 μg/m³. The continuous relationship between PM_2.5 and first HF rehospitalization is shown in Figure 2. Compared with tertile 1, those residing in tertile 2 and tertile 3 had 2.3% and 1.8% increase in hazard of PM_2.5 (model 2: HR, 1.023 [95% CI, 1.016–1.030]; and HR, 1.018 [95% CI, 1.010–1.025]), respectively. Those residing at ≥90th percentile of exposure had 16% increased hazard of first HF hospitalization compared with those residing at ≤10th percentile of exposure (HR, 1.160 [95% CI, 1.130–1.190]; P<0.001). The additive effect of PM_2.5 and SDI are shown in Figure S4. Each IQR of PM_2.5 was associated with 3.1% increase in overall HF readmission burden (relative risk [RR] 1.031 [1.027–1.036] in model 2) and 5.2% increase in all‐cause readmission burden (RR 1.052 [1.048–1.056] in model 2). Additional adjustment for chronic obstructive pulmonary disease for did not alter the association (HR, 1.032 [95% CI, 1.028–1.036]; and HR, 1.049 [95% CI, 1.046–1.053], respectively). These relationships were more pronounced in patients exposed to <7 μg/m³ (8.6% and 10.6% in model 2, respectively, Table 3).

Table 3.

Association of PM_2.5 With Total HF Hospitalization Burden and All‐Cause Readmissions in Patients With New Diagnosis of HF

HF admissions burden
	Relative risk	95% CI	P value
Model 1
Per IQR PM2.5	1.035	1.031–1.04	<0.001
Per IQR PM2.5 (in those <7 μg/m3)	1.094	1.081–1.107	<0.001
Tertile 2 vs tertile 1	1.055	1.045–1.065	<0.001
Tertile 3 vs tertile 1	1.044	1.034–1.055	<0.001
≥90th percentile vs ≤10th percentile* of cohort limited to those <7 μg/m3	1.266	1.212–1.323	<0.001
Model 2
Per IQR PM2.5	1.031	1.027–1.036	<0.001
Per IQR PM2.5 (in those <7 μg/m3)	1.086	1.073–1.099	<0.001
Tertile 2 vs tertile 1	1.048	1.039–1.058	<0.001
Tertile 3 vs tertile 1	1.038	1.028–1.049	<0.001
≥90th percentile vs ≤10th percentile* of cohort limited to those <7 μg/m3	1.244	1.191–1.298	<0.001

All‐cause readmissions burden
	Relative risk	95% CI	P value
Model 1
Per IQR PM2.5	1.049	1.045–1.052	<0.001
Per IQR PM2.5 (in those <7 μg/m3)	1.106	1.096–1.116	<0.001
Tertile 2 vs tertile 1	1.076	1.067–1.084	<0.001
Tertile 3 vs tertile 1	1.068	1.059–1.077	<0.001
≥90th percentile vs ≤10th percentile* of cohort limited to those <7 μg/m3	1.286	1.242	<0.001
Model 2
Per IQR PM2.5	1.052	1.048–1.056	<0.001
Per IQR PM2.5 (in those <7 μg/m3)	1.10	1.09–1.11	<0.001
Tertile 2 vs tertile 1	1.071	1.063–1.079	<0.001
Tertile 3 vs tertile 1	1.063	1.054–1.072	<0.001
≥90th percentile vs ≤10th percentile* of cohort limited to those <7 μg/m3	1.227	1.186–1.27	<0.001

Model 1 adjusted for age, sex, race, and Medicaid eligibility status, rurality of zip code, Census region, and social deprivation score. Model 2 adjusted for same variables as model 1+diabetes, hypertension, coronary artery disease, and obesity. HF indicates heart failure; IQR, interquartile range; and PM_2.5, particulate matter with an aerodynamic diameter <2.5 microns.

^*10th percentile PM_2.5 = 1.23–4.15 μg/m³ and 90th percentile PM_2.5 = 6.87–7.00 μg/m³.

Discussion

To our knowledge, this is the largest study to date investigating the association between PM_2.5 exposure and adverse outcomes in patients with preexisting HF. We used fine‐scale PM_2.5 exposure model throughout the United States and linked it with a contemporary cohort of Medicare enrollees with preexisting HF throughout the continental United States. We show that PM_2.5 is associated with a substantial increase in HF readmission and death, and it is additive to social deprivation.

Significant literature has linked chronic exposure to PM_2.5 with incident CVD (mainly ischemic events) and death. The association between chronic exposure to PM_2.5 and HF risk has been less well studied. Prior studies linking PM_2.5 and HF have been limited by small sample size, focus on short‐term exposure (in case‐crossover or time‐series design), or incident HF in patients without baseline CVD. For example, a systematic review of 35 studies showed that short‐term exposure (days to weeks) to multiple pollutants (CO, SO₂, NO₂, and PM_2.5, but not ozone) was associated with increased risk for HF.⁵ In a longitudinal analysis of 432 530 participants free of HF, atrial fibrillation, or coronary heart disease in the UK Biobank study, over 10‐year follow‐up, each 10 μg/m³ increase in chronic PM_2.5 was associated with 85% risk increase in incident HF (HR, 1.85 95% CI, 1.34–2.55) over a follow‐up of 10 years. ¹⁸ To our knowledge, our study is the first large study to examine the impact of PM_2.5 in racially and geospatially diverse patients with preexisting HF.

The mechanism linking PM_2.5 and CVD is an active area of investigation. In animal models, PM_2.5 induces inflammation, thrombosis, autonomic dysfunction, oxidative stress, blood pressure elevations, and robust epigenetic modifications. ³ , ¹⁹ Chronic exposure to PM_2.5 in mice leads to myocardial fibrosis, myosin heavy chain isoform switching, and myocardial dysfunction. ¹⁰ PM_2.5 may lead to cardiotoxicity through hypermethylation in regulatory genes that promote apoptosis. ²⁰ , ²¹ , ²² In humans, a randomized trial of air filtration showed that diesel exhaust exposure led to brain natriuretic peptide increase, endothelial dysfunction, and reduction in exercise oxygen consumption in patients with HF, which was abrogated by air filtration. ⁶ , ⁷

A prior large study of all >60 million Medicare enrollees (2000–2012) has shown that each 10 μg/m³ increase in PM_2.5 was associated with 7.3% increase in all‐cause death. ²³ Our study shows a more pronounced association between PM_2.5 and death, at lower exposure concentrations, and an even stronger relationship with first and total HF readmissions. This suggests that patients with preexisting HF may carry higher susceptibility for risk for harmful effects of PM_2.5 exposure. We have previously shown that each 10 μg/m³ increase in PM_2.5 is associated with a 26% increased risk of death in patients with HF undergoing heart transplantation. ⁴ Collectively, these studies highlight the importance of studying methods to reduce PM_2.5 exposure (eg, portable air cleaners ²⁴ ) in patients with HF to reduce morbidity and death. At the national and the global scale, PM_2.5 exposure affects a large number of patients, particularly those living in socioeconomically disadvantaged neighborhoods. As such, PM_2.5 exposure likely contributes to the disparities in HF outcomes noted in national studies and offers a target for interventions. An ongoing trial (PURI‐HF [Air Purifiers on Heart Failure], Clinicaltrials.gov, NCT05230784) will examine the impact of portable air filters in patients with HF exposed to high levels of air pollution in India.

Interestingly, we show that younger individuals, those without diabetes, those living in rural areas and those of Asian race to be at increased susceptibility for harmful effects of PM_2.5 exposure. The reasons remain speculative, and several hypotheses may explain these observations. First, there could be a survival bias, where more susceptible patients (eg, those with diabetes, older individuals) died earlier in life and did not survive to be included in our older cohort. Second, patients with diabetes may be more likely to be on protective cardiometabolic medications that may attenuate the cardiovascular effects of pollution exposure. Younger individuals may have fewer comorbidities, making the associations between PM_2.5 and outcomes more apparent compared with older individuals with multiple chronic conditions. These findings are in line with prior findings from other cohorts, such as veterans with coronary artery disease. ²⁵ The mechanisms underlying the heightened risks observed among Asian individuals remain unclear and warrant further investigation. Interestingly, we found that the associations between PM_2.5 and adverse outcomes were more pronounced among patients residing in rural areas. While the reasons for this remain speculative, it may relate to differences in air pollution sources, exposure patterns, or access to medical care in rural areas, which are different from urban environments. Future research should explore the rural–urban differences and how localized interventions may need to be tailored on the basis of the geographic context. Further research is also needed to clarify the interrelationships among diabetes, comorbidities, cardiovascular medications, and pollution susceptibility. Elucidating these complex interplays between comorbidities and pollutant exposures can help identify patient groups most vulnerable to air pollution health impacts, for whom targeted interventions can be investigated.

This study has many strengths, including large national and contemporary cohort, utilization of high‐resolution validated PM_2.5 estimates, and adjustment or zip code–level SDI. The findings of our study, however, have to be taken within the context of study limitations. First, the PM_2.5 exposure is a model‐based estimate, and while validated with ground monitors, it is prone to misclassification. Second, we lacked information on census tracts, which allows more granular geospatial analysis for clinical outcomes. Third, we lack data on ejection fraction, medications, and care processes and quality during the admission, specialized cardiology care after discharge, and the use of guideline‐directed therapy on follow‐up, which may differ by location and affect downstream adverse events. Fourth, we also lacked information on societal factors such as neighborhood violence, housing stability, access to transportation, and so on, which may influence the risk of adverse outcomes differently. Fifth, the risk adjustment was on the basis of claims data, which may be prone to coding error or missingness from lack of capture of events before Medicare enrollment. These factors, and other unmeasured factors, may be confounding the relationships observed. However, the use of claims‐based data for risk adjustment is standard for the Center for Medicare and Medicaid Services and most health policy and health services research, and the research findings should be interpreted within the context of biologic plausibility from decades of molecular and mechanistic studies discussed above. Additionally, while we tested for interactions by demographic/socioeconomic factors, we could not examine more granular neighborhood‐level characteristics or localized pollution sources/levels that may impact exposure heterogeneity within geographic areas. The heterogeneity in the associations between PM_2.5 and outcomes may present chance findings and should be taken within the context of the study limitations.

Conclusions

Air pollution exposure is associated with significant morbidity and death in Medicare patients with preexisting HF. The association is independent of social deprivation, suggesting an additive effect. Interventions to reduce PM_2.5 at the individual and community level should be investigated to reduce HF death and morbidity.

Sources of Funding

This work was partly funded by the National Institute on Minority Health and Health Disparities Award number P50MD017, and partly funded by philanthropic gifts by the Haslam Family, Bailey Family, and Khouri Family to the Cleveland Clinic (principal investigator, Milind Y. Desai).

Disclosures

None.

Supporting information

Figure S1–S4