A substantial body of evidence links long-term exposure to air pollution to cardiovascular mortality and morbidity. However, research on the effects of air pollution below current regulatory limits remains scarce, and causal evidence is limited. This study aims to address this gap by examining the association between long-term exposure to low concentrations of PM2.5, NO2, and ozone (O3) and the risk of cardiovascular hospitalizations (specifically stroke, heart failure (HF), and atrial fibrillation (AF)) among a large cohort of US Medicare beneficiaries. Previous research has predominantly focused on PM2.5, neglecting the potential cardiovascular impacts of NO2 and O3, and often analyzing the entire exposure range rather than focusing on concentrations below regulatory standards. Understanding the health effects of air pollution at these lower levels is crucial for informing air pollution regulations and determining a truly “safe” exposure level. Therefore, this study employs a double negative control approach to strengthen causal inference by mitigating potential confounding bias from unmeasured variables, offering a more robust assessment of the relationships between low-level air pollution and cardiovascular diseases.

Existing epidemiological studies demonstrate a positive association between long-term air pollution exposure and cardiovascular events, although research on specific cardiovascular outcomes beyond overall mortality and morbidity is limited. While studies have linked PM2.5 to increased stroke risk, the cardiovascular effects of NO2 and O3 require further investigation. Inconsistencies and a scarcity of data exist regarding the association between air pollution and HF and AF. Most studies have examined the full range of pollution exposure, obscuring the health impacts of concentrations below regulatory standards. Previous research suggests that the exposure-response (E-R) relationship for PM2.5 and cardiovascular mortality is curvilinear with no apparent threshold. However, studies in the US and Europe indicate that cardiovascular risks associated with PM2.5 may persist or even increase at lower exposures below the EPA and EU recommended limits. The use of negative controls, especially a double negative control approach, offers a valuable methodology to improve causal inference in observational studies by accounting for potential confounding from unmeasured variables. Previous studies have utilized future air pollution levels as a negative exposure control or negative outcome controls independent of the primary exposure. The double negative control approach strengthens causal inference by using both negative exposure and outcome controls simultaneously.

This study utilized data from a national cohort of fee-for-service (FFS) Medicare beneficiaries (aged ≥65 years) across the contiguous US from 2000 to 2016. The study population was restricted to individuals consistently exposed to low-level air pollution (PM2.5 < 9 µg/m³, NO2 < 75.2 µg/m³ [40 ppb], warm-season O3 < 88.2 µg/m³ [45 ppb]) throughout the study period, creating separate datasets for each pollutant. Hospitalization data were obtained from the Medicare Provider Analysis and Review (MEDPAR) file, using ICD codes to identify stroke, HF, and AF hospitalizations. Daily concentrations of PM2.5, NO2, and O3 were obtained from ensemble prediction models with high spatial resolution (1 km × 1 km). Annual average exposures were linked to beneficiaries based on their residential ZIP codes. ZIP code-level covariates, including SES indicators, meteorological data (temperature and relative humidity), smoking rates, BMI, and access-to-care metrics, were obtained from various sources (US Census Bureau, CDC BRFSS, Dartmouth Atlas of Health Care). The primary analysis employed a double negative control approach to control for unmeasured confounding. The negative exposure control was the subsequent year's air pollution levels, and the negative outcome control was the preceding year's hospitalization counts for each outcome. Generalized linear models (GLMs) with quasi-Poisson regression were used, incorporating the predicted values of the negative outcome control to address confounding bias. Stratified analyses examined effect modification by demographic characteristics (age, sex, race, Medicaid eligibility). GLMs without negative controls were used for comparison, and natural spline functions were used to depict E-R curves.

The study found positive associations between long-term exposure to low-level PM2.5, NO2, and warm-season O3 and hospitalization rates for stroke, HF, and AF, even after controlling for co-pollutants and unmeasured confounders using the double negative control approach. The associations generally persisted in demographic subgroups. Specifically: * For PM2.5 (<9 µg/m³), each 1 µg/m³ increase was associated with a 1.82% increase in stroke hospitalizations, 2.83% in HF hospitalizations, and 0.13% in AF hospitalizations. * For NO2 (<75.2 µg/m³), each 1 µg/m³ increase was associated with a 0.01% increase in stroke hospitalizations, 0.18% in HF hospitalizations, and 0.09% in AF hospitalizations. * For warm-season O3 (<88.2 µg/m³), each 1 µg/m³ increase was associated with a 0.32% increase in stroke hospitalizations, 0.05% in HF hospitalizations, and 0.12% in AF hospitalizations. Stratified analyses revealed that Black individuals and those eligible for Medicaid were more vulnerable to the cardiovascular risks associated with PM2.5 and warm-season O3. The youngest-old (aged 65-74 years) were more susceptible to warm-season O3-related risks. However, NO2-related risks showed inconsistent effect modification patterns across demographic groups. E-R curves from GLMs showed a positive association between PM2.5 and stroke and HF even at the lowest concentrations. The association between PM2.5 and AF was more complex, showing a positive association above 5 µg/m³ but a negative association at lower levels. NO2 exhibited an almost linear positive association with HF, and a more complex relationship with stroke and AF. Warm-season O3 showed a linear positive association with stroke and non-linear relationships with HF and AF, with effects increasing at higher concentrations.

This study provides strong evidence that even low levels of air pollution, below current US EPA standards, are associated with increased risks of cardiovascular hospitalizations among older adults. The double negative control approach strengthened the causal interpretation of these findings by addressing unmeasured confounding. The findings underscore the potential health benefits of lowering air pollution levels even in areas currently considered to have relatively clean air. The observed disparities in vulnerability across demographic subgroups, particularly for Black individuals and Medicaid-eligible individuals, highlight environmental justice concerns and the need for targeted interventions. The inconsistent effect modification patterns observed for NO2 require further investigation. This research supports the need for stricter air quality guidelines and more robust pollution control policies.

This study, using a rigorous double negative control approach and a large national cohort of Medicare beneficiaries, demonstrates that long-term exposure to low levels of PM2.5, NO2, and warm-season O3 is associated with increased risks of stroke, HF, and AF hospitalizations. The findings strongly suggest that current US NAAQS may be insufficient to protect cardiovascular health. Future research should focus on the mechanisms underlying these associations, particularly the disparities observed across demographic groups, and should explore even stricter air quality standards to minimize the cardiovascular disease burden.

This study's findings may not be generalizable to younger populations or regions with significantly higher pollution levels. Residual confounding from unmeasured cardiovascular risk factors is possible despite the use of multiple covariates and negative controls. The use of ZIP code-level air pollution data may not perfectly capture individual exposure variations. Finally, relying on hospital discharge diagnoses might underestimate the true burden of cardiovascular disease.