The Role of Neighborhood Air Pollution in Disparate Racial and Ethnic Asthma Acute Care Use

Abstract

Rationale

The share of Black or Latinx residents in a census tract remains associated with asthma-related emergency department (ED) visit rates after controlling for socioeconomic factors. The extent to which evident disparities relate to the within-city heterogeneity of long-term air pollution exposure remains unclear.

Objectives

To investigate the role of intraurban spatial variability of air pollution in asthma acute care use disparity.

Methods

An administrative database was used to define census tract population-based incidence rates of asthma-related ED visits. We estimate the associations between census tract incidence rates and 1) average fine and coarse particulate matter, nitrogen dioxide (NO₂), and sulfur dioxide (SO₂), and 2) racial and ethnic composition using generalized linear models controlling for socioeconomic and housing covariates. We also examine for the attenuation of incidence risk ratios (IRRs) associated with race/ethnicity when controlling for air pollution exposure.

Measurements and Main Results

Fine and coarse particulate matter and SO₂ are all associated with census tract–level incidence rates of asthma-related ED visits, and multipollutant models show evidence of independent risk associated with coarse particulate matter and SO₂. The association between census tract incidence rate and Black resident share (IRR, 1.51 [credible interval (CI), 1.48–1.54]) is attenuated by 24% when accounting for air pollution (IRR, 1.39 [CI, 1.35–1.42]), and the association with Latinx resident share (IRR, 1.11 [CI, 1.09–1.13]) is attenuated by 32% (IRR, 1.08 [CI, 1.06–1.10]).

Conclusions

Neighborhood-level rates of asthma acute care use are associated with local air pollution. Controlling for air pollution attenuates associations with census tract racial/ethnic composition, suggesting that intracity variability in air pollution could contribute to neighborhood-to-neighborhood asthma morbidity disparities.

At a Glance Commentary

Scientific Knowledge on the Subject

Although it is well recognized that air pollution exposure increases the risk of asthma-related acute care use, it is not well understood how within-city air pollution gradients contribute to neighborhood-to-neighborhood differences in asthma-related emergency department (ED) visit rates. Racial and ethnic disparity in asthma morbidity has been documented within cities, but the role of air pollution has not been established.

What This Study Adds to the Field

This study reveals that within-city variation of multiple pollutants is associated with census tract–level incidence rates of asthma-related ED visits, even at low pollution concentrations. Because associations between census tract racial/ethnic composition and asthma-related ED visit rates are attenuated when controlling for within-city pollution variability, these findings suggest that air pollution variability could contribute to within-city racial and ethnic disparity in asthma morbidity.

Although the relationship between short-term variation in air pollution concentration at the ZIP code and county levels and asthma acute care use is well characterized (1–5), the relationship between fine-scale spatial heterogeneity in air pollution concentrations and within-city variation in neighborhood-level rates of asthma acute care use remains unexplored. This holds particular relevance for race- and ethnicity-linked asthma disparities, which are apparent at small spatial scales, as is the disproportionate siting of pollution sources near communities of color in the United States (6). The latter has resulted in systematically higher exposure to ambient (outdoor) air pollution for people of color compared with White residents of the same city (7, 8). Higher rates of asthma-related emergency department (ED) visits among Black Americans and other people of color relative to their White counterparts have been attributed in part to higher acute exposure to particulate matter with an aerodynamic diameter ⩽2.5 μm (PM_2.5) among a range of environmental exposures and social determinants of health (9–13). However, a paucity of asthma morbidity data at fine geographic scales has been a barrier to the investigation of the relationship between persistent localized elevations in air pollution and race- and ethnicity-related disparate neighborhood-level asthma-related ED visit rates.

A recent study of census tract–level population-based incidence rates (PBIRs) of asthma-related ED visits in central Texas revealed substantial intraurban variation, spatial clustering, and race- and ethnicity-linked patterns of asthma disparity, with variation in the racial and ethnic composition of census tracts explaining 33% of spatial variation in ED visit rates after controlling for census tract–level socioeconomic status (SES) and housing conditions (14). These population health observations provide a novel opportunity to link ambient air pollution variation and asthma-related ED visit disparity at equally high spatial resolution. Associations between neighborhood-to-neighborhood variability in air pollution exposure and neighborhood-level variability in asthma-related acute health care use would suggest that finer spatial scale considerations of air pollution concentrations are important for understanding and addressing the excess burden of asthma ED visits among some neighborhoods. Furthermore, associations between neighborhood-to-neighborhood variability in air pollution concentrations and neighborhood-level racial and ethnic disparities in asthma-related acute health care use would be among the first data to tie within-city environmental injustice to measured health disparities.

Using a retrospective, cross-sectional ecologic study design, we examine for associations of census tract–level rates of asthma-related ED visits with average ambient concentrations of four air pollutants: PM_2.5, particulate matter with an aerodynamic diameter ⩽10 μm (PM₁₀), nitrogen dioxide (NO₂), and sulfur dioxide (SO₂). We then quantify excess exposure to one or more pollutants for Latinx or Black populations within this metropolitan area and examine the attenuation of race- and ethnicity-based incidence risk ratios in models that control for pollution exposure, indicating a relationship between ambient air pollution and apparent racial disparities in asthma-related ED visits.

Methods

Study Population and Health Outcomes

This ecologic analysis considers populations at the census tract level within the five-county Austin metropolitan statistical area (MSA) for 2016 and 2017. Tract-level PBIRs of asthma-related ED visits were determined from the Texas Health Care Information Collection administrative dataset (see the online supplement). This study was approved by the Institutional Review Board of the University of Texas at Austin and the Texas Department of State Health Services. Asthma-related ED visits were identified as having a primary International Classification of Diseases, 10th Revision, diagnostic code of J45.x. The PBIR for each census tract, expressed as rate per 10,000 person-years, was calculated as the sum of tract asthma-related ED visits for 2016 and 2017 divided by the sum of the tract populations in both years (15). PBIRs include repeat visits by the same patient as separate events (i.e., people are not removed from the risk set), and the population at risk is regarded as constant across the two years of the study period. We examine all-age PBIR and stratify by age category. From 2016 to 2017, Texas Health Care Information Collection records for the Austin MSA include 8,066 asthma-related ED visits by children (patients aged <18 yr) and 8,360 by adults (those aged ⩾18 yr). As our focus is on disparities between Black and Latinx patients and their White counterparts, we analyze the PBIR of White, Black, and Latinx records grouped together, excluding other racial and ethnic groups. Among a total of 16,426 ED visits, this results in the exclusion of 742 records (4.5%).

Census Tract Environmental and Demographic Covariates

Census tract average concentrations of NO₂, PM_2.5, PM₁₀, and SO₂ were obtained from the Center for Air, Climate, and Energy Solutions (16). The racial and ethnic compositions of the population of each census tract were obtained from the American Community Survey (ACS) as averages spanning 2013–2017. Categories of racial and ethnic identity provided by ACS were grouped as Latino or Hispanic members of any race (hereafter referred to as Latinx); non-Hispanic Black (hereafter referred to as Black); non-Hispanic White (hereafter referred to as White); and non-Hispanic Asian American, Pacific Islander, Native American, or other racial identity.

Because low SES and poor housing conditions are known to be associated with asthma morbidity burden, we include 10 additional ACS adjustment covariates (14). These covariates include the proportion of census tract residents 1) with household income below federal poverty line, 2) in households without access to a private vehicle, 3) in households without a computer, and 4) without health insurance; 5) adults with less than a high school education; the proportion of occupied housing units 6) using natural gas for heating, 7) with more than one resident per room, 8) with 10 or more attached units, and 9) with two or fewer bedrooms; and 10) the share of total unoccupied housing units.

Statistical Methods

The association of census tract air pollution with asthma-related ED PBIR was estimated using Poisson generalized linear models. Because the population at risk is regarded as constant across each of the two years of the study period, the Poisson regression estimates incidence risk ratios (IRR). To investigate the interrelationship among air pollution, census tract racial and ethnic composition, and asthma-related ED PBIR, we first examined racial and ethnic inequity in air pollution exposure using summary statistics of pollution exposure stratified by racial and ethnic groups, calculated using population-weighted distributions of each pollutant (17), and Gaussian linear models used to estimate the association of the shares of Black and Latinx residents in the census tract population with average pollution concentrations. We then estimated IRR values associated with the share of Black or Latinx residents modeled with and without air pollution as a control covariate and checked for attenuation. All regressions adjust for the previously detailed socioeconomic and housing covariates; the association of racial and ethnic with air pollution is estimated without these control covariates as a sensitivity analysis. Additional details are provided in the online supplement.

Results

Study Population

The Austin MSA includes five counties (Bastrop, Caldwell, Hays, Travis, and Williamson), with a 2016–2017 population of 1.97 million residents, with the largest share in Travis County (59%). Table 1 describes characteristics of the study area. Tracts spanned a broad range of population density, from rural and sparsely populated (17 residents) to high-density urban populations (21,435 residents). The racial and ethnic composition of census tracts also ranged widely.

Table 1.

Study Population Characteristics and Census Tract Population Statistics

	MSA Total	Census Tract Mean	SD	IQR	Minimum	Maximum
Total population	1,971,603	5,649	2,982	3,580	17	21,435
Pediatric	434,705	1,246	938	1,130	0	6,183
Adult	1,388,052	3,977	1,943	2,361	15	14,291
Black	7.0%	6.8%	7.2%	7.2%	0.0%	46.3%
Latinx	32.1%	30.9%	19.5%	29.8%	2.4%	85.0%
Native American, Asian, Pacific Islander, and other	7.9%	7.5%	6.7%	7.1%	0.0%	39.2%
White	53.1%	54.8%	21.4%	33.6%	3.2%	96.7%

Definition of abbreviations: IQR = interquartile range; MSA = metropolitan statistical area.

We analyzed the records of 15,684 asthma-related ED visits from 2016 and 2017 (Table 2). The MSA-wide PBIRs per 10,000 person-years were 84.9 for children and 27.9 for adults. Tract-specific rates varied substantially, with mean (SD) PBIRs of 96.1 (102.4) for children and 30.1 (29.8) for adults and an interquartile range of 90.0 for children and 25.9 for adults.

Table 2.

Asthma-related Emergency Department Visits and Census Tract Population-based Incidence Rates (per 10,000 Person-Years)

	Total Cases (2016–2017)	PBIR Census Tract Mean	PBIR SD	PBIR IQR	PBIR Minimum	PBIR Maximum
Total	15,684	44.0	38.0	41.7	1.6	297.2
Pediatric	7,686	96.1	102.4	90.0	0.0	1228.1
Adult	7,998	30.1	29.8	25.9	0.0	213.4

Definition of abbreviations: IQR = interquartile range; PBIR = population-based incidence rate.

Totals and rates include only Black-, Latinx-, and White-identifying patients.

Exposure Assessment and Association of Asthma Acute Care Use with Ambient Pollution Concentrations

Pollution concentrations in the Austin MSA were low relative to major U.S. cities (see Table E1 in the online supplement), with population-weighted mean (SD) values of 8.2 (0.7) μg/m³ for PM_2.5, 19.1 (2.9) μg/m³ for PM₁₀, 4.4 (1.8) ppb for NO₂, and 0.7 (0.2) ppb for SO₂. Geographic patterns varied among pollutants, suggesting contributions from a range of local sources: PM_2.5 and NO₂ were highest within the urban core and along transportation corridors, and PM₁₀ and SO₂ were higher in the northern and eastern suburbs in areas with greater industrial activity (see Figure E1). Census tract concentrations of PM_2.5 were highly correlated with NO₂ because of the coemission of these two pollutants by many urban sources (Pearson’s R = 0.81; see Table E2); correlations among other pollutants are low to moderate.

In single-pollutant models, average concentrations of PM_2.5, PM₁₀, and SO₂ were associated with an increase in asthma-related ED visit rate (Table 3) when controlling for SES and housing covariates. For particulate pollutants, a 1-SD increment of PM₁₀ (3 μg/m³) was associated with a 30% (28–33%) higher PBIR of asthma-related ED visits, and a 1-SD increment of PM_2.5 (0.7 μg/m³) was associated with a 22% (19–25%) higher PBIR. For gaseous pollutants, a 1-SD increment of SO₂ (0.2 ppb) was associated with a 32% (30–35%) higher PBIR. NO₂ did not show a detectable association with PBIR in all-age models. Age-stratified single-pollutant models showed overlapping 95% credible intervals for child- and adult-specific IRRs except for NO₂, which was associated with a statistically significant increase in asthma-related ED visit rate for adults only.

Table 3.

IRR (CI) for Asthma-related Emergency Department Visits Associated with Census Tract Annual Average Pollution Concentration

	IRR (CI)
	All Ages	Pediatric	Adult
Single-pollutant models
NO2	1.01 (0.98–1.04)	0.99 (0.95–1.04)	1.09 (1.05–1.14)
PM2.5	1.22 (1.19–1.25)	1.14 (1.1–1.18)	1.34 (1.29–1.38)
PM10	1.3 (1.28–1.33)	1.26 (1.23–1.29)	1.31 (1.28–1.34)
SO2	1.32 (1.3–1.35)	1.29 (1.26–1.33)	1.33 (1.29–1.36)
Multipollutant model
NO2	1.02 (0.98–1.06)	1.09 (1.03–1.15)	0.99 (0.93–1.04)
PM2.5	1 (0.96–1.03)	0.88 (0.84–0.93)	1.15 (1.1–1.21)
PM10	1.22 (1.19–1.25)	1.24 (1.2–1.28)	1.15 (1.12–1.19)
SO2	1.24 (1.21–1.27)	1.24 (1.2–1.28)	1.23 (1.19–1.27)

Definition of abbreviations: CI = credible interval; IRR = incidence risk ratio; NO₂ = nitrogen dioxide; PM_2.5 = particulate matter with an aerodynamic diameter ⩽2.5 μm; PM₁₀ = particulate matter with an aerodynamic diameter ⩽10 μm; SO₂ = sulfur dioxide.

Models account for census tract socioeconomic and housing characteristics.

In a multipollutant model simultaneously investigating the four pollutants, the IRRs of asthma-related ED visits associated with PM₁₀ and SO₂ were attenuated but still significant in all-age and age-stratified models, while the IRR credible interval (CI) bounds for PM_2.5 included the null effect in the all-age model and show conflicting patterns across pediatric and adult models. First-order interactions between PM_2.5 and SO₂ were examined on the basis of a previous finding that the presence of particulate sulfur, expected from sources coemitting these pollutants, may increase the risk of respiratory effects (4). We found PM_2.5 to be associated with a greater increase in PBIR in areas with higher SO₂ concentrations, a pattern consistent across children and adults (Figure 1).

Figure 1.

IRR for asthma-related ED visits associated with a 1-SD change in PM_2.5 modeled with interaction effects between PM_2.5 and SO₂. The model adjusts for census tract socioeconomic and housing covariates. CI = credible interval; ED = emergency department; IRR = incidence risk ratio; PM_2.5 = particulate matter with an aerodynamic diameter ⩽2.5 μm.

Racial and Ethnic Inequity in Air Pollution Exposure

Population-weighted summary statistics showed that White residents on average live in neighborhoods with lower concentrations of all four pollutants compared with all other racial and ethnic groups (see Figure E2). The median concentration of each pollutant was highest for Black residents, followed by Latinx residents. Differences in median concentrations between Black or Latinx residents and White residents were small in absolute terms (NO₂, 1 ppb; PM_2.5, 0.5 μg/m³; PM₁₀, 1 μg/m³; and SO₂, 0.1 ppb) and of comparable magnitude with the SD for the population within the Austin MSA. The upper quartiles of NO₂, PM₁₀, and SO₂ for Black and Latinx residents appeared at higher concentrations than for White residents, indicating disproportionate representation of Black and Latinx residents in high-pollution neighborhoods. In addition, differences in lower quartiles among racial/ethnic groups exceeded differences in medians for PM_2.5, PM₁₀, and SO₂, with White residents disproportionately represented in the cleanest areas of the MSA.

In models not adjusting for SES and housing covariates, concentrations of all four pollutants were positively associated with the shares of Black and Latinx census tract residents. With SES and housing covariates included, significant positive associations remained between concentrations of PM₁₀ and SO₂ and the share of Black census tract residents and between PM_2.5 and PM₁₀ concentrations and the share of Latinx residents (see Figure E4).

Association of Asthma Acute Care Use with Neighborhood Racial/Ethnic Composition

The shares of Black and Latinx residents were both significant predictors of census tract asthma-related ED visit rates (Table 4), and the magnitude of the association was not affected when adjusting for SES and housing covariates (see Tables E3 and E4), in agreement with results reported by Zárate and colleagues for Travis County alone (14). A 10% higher share of Black residents corresponded to an IRR of 1.51 (CI, 1.48–1.54), with age-stratified IRRs of 1.47 (CI, 1.43–1.52) for pediatric PBIR and 1.47 (CI, 1.42–1.51) for adult PBIR. The IRRs associated with the share of Latinx residents were lower: 1.12 (CI, 1.10–1.14) for the total population, 1.10 (CI, 1.08–1.13) for pediatric residents, and 1.11 (CI, 1.08–1.13) for adult residents.

Table 4.

IRR (CI) for Asthma-related Emergency Department Visits Associated with Census Tract Share of Residents Identifying as Black and Latinx

	IRR (CI)
	All Ages	Pediatric	Adult
Share Black	1.51 (1.48–1.54)	1.47 (1.43–1.52)	1.47 (1.42–1.51)
Share Latinx	1.12 (1.1–1.14)	1.1 (1.08–1.13)	1.11 (1.08–1.13)

Definition of abbreviations: CI = credible interval; IRR = incidence risk ratio.

Models account for census tract socioeconomic and housing characteristics.

Residual Disparities after Adjusting for Air Pollution

Figure 2 depicts the attenuation of the IRR of asthma-related ED visits associated with racial/ethnic composition from models adjusting for census tract SES and housing but not pollution (shown in blue), models accounting for each pollutant individually, and a model considering all four pollutants simultaneously (“All poll.”). As shown in Table 5, with the inclusion of all pollutants, the IRRs for the share of Black residents were attenuated by 24% (IRR, 1.39 [CI, 1.39–1.42]), 25% (IRR, 1.35 [CI, 1.31–1.40]), and 19% (IRR, 1.36 [CI, 1.32–1.41]) in all-age, pediatric, and adult models, respectively. The IRRs for the share of Latinx residents were attenuated by 32% (IRR, 1.08 [CI, 1.06–1.1]), 36% (IRR, 1.07 [CI, 1.04–1.09]), and 54% (IRR, 1.06 [CI, 1.04–1.09]) in all-age, pediatric, and adult models. When accounting for individual pollutants, SO₂ and PM₁₀ showed the greatest IRR attenuation for the share of Black residents, while PM_2.5 and PM₁₀ showed the greatest attenuation for the share of Latinx residents, in alignment with the patterns of race- and ethnicity-specific exposure disparity observed after controlling for census tract SES and housing covariates.

Figure 2.

Attenuation of race/ethnicity-related incidence risk ratio after the addition of one or all pollutant variables. The blue marker indicates the incidence risk ratio in models not adjusting for pollution. All models adjust for census tract socioeconomic and housing covariates. CI = credible interval; ED = emergency department; IRR = incidence risk ratio; PM_2.5 = particulate matter with an aerodynamic diameter ⩽2.5 μm; PM₁₀ = particulate matter with an aerodynamic diameter ⩽10 μm; poll. = pollutants.

Table 5.

Attenuation of Association of Racial/Ethnic Composition with Asthma-related Emergency Department Visits in Models Accounting for Concentrations of One or More Pollutants

	IRR (CI)
	All Ages	Pediatric	Adult
NO2
Share Black	1.52 (1.49–1.55)	1.48 (1.44–1.53)	1.49 (1.44–1.53)
Share Latinx	1.12 (1.1–1.14)	1.10 (1.07–1.13)	1.10 (1.07–1.13)
PM2.5
Share Black	1.51 (1.48–1.54)	1.48 (1.44–1.53)	1.47 (1.43–1.51)
Share Latinx	1.10 (1.08–1.12)	1.09 (1.06–1.12)	1.07 (1.04–1.10)
PM10
Share Black	1.42 (1.38–1.45)	1.40 (1.36–1.44)	1.36 (1.32–1.41)
Share Latinx	1.08 (1.06–1.10)	1.07 (1.04–1.1)	1.06 (1.03–1.08)
SO2
Share Black	1.40 (1.37–1.43)	1.38 (1.34–1.43)	1.36 (1.31–1.4)
Share Latinx	1.11 (1.09–1.13)	1.09 (1.07–1.12)	1.1 (1.07–1.12)
Multipollutant
Share Black	1.39 (1.35–1.42)	1.35 (1.31–1.4)	1.36 (1.32–1.41)
Share Latinx	1.08 (1.06–1.10)	1.07 (1.04–1.09)	1.06 (1.04–1.09)

Definition of abbreviations: CI = credible interval; IRR = incidence risk ratio; NO₂ = nitrogen dioxide; PM_2.5 = particulate matter with an aerodynamic diameter ⩽2.5 μm; PM₁₀ = particulate matter with an aerodynamic diameter ⩽10 μm; SO₂ = sulfur dioxide.

All models account for census tract socioeconomic and housing characteristics.

Discussion

This study demonstrates that geographic heterogeneity in long-term air pollution concentrations within the Austin metropolitan area is associated with census tract–level incidence rates of asthma-related ED visits, even though overall pollution concentrations are low and neighborhood-to-neighborhood variation is of small magnitude. Furthermore, the association between asthma-related ED visit rates and neighborhood racial and ethnic composition that persists after controlling for neighborhood SES and housing indicators is attenuated when accounting for neighborhood air pollution. Given that the localized influence of emission sources and urban infrastructure produces substantial within-city air pollution heterogeneity, with differences in patterns among pollutants arising from the locations of their dominant emission sources (18–21), these findings implicate the siting of air pollution sources in neighborhood-level variability in asthma-related ED visits and disparity. Collectively, these findings of significant associations between asthma-related ED visit rates and SO₂ and PM₁₀ in multipollutant models, together with differences in the set of pollutants that attenuate disparities for the Black versus Latinx population shares, motivate increased examination of the role of specific urban pollution sources in asthma health disparities related to intraurban segregation (18, 22–26).

Within the Austin MSA, Black and Latinx populations are exposed to higher concentrations of all four pollutants considered (PM_2.5, PM₁₀, NO₂, and SO₂), as shown by higher median degrees of exposure and disproportionate representation in the highest exposure categories. Similar exposure inequity has been documented at the national level, although studies have focused primarily on PM_2.5 (17, 27–31). A complementary ecologic view shows that neighborhoods with predominantly Black or Latinx populations tend to be more polluted, and the types of pollutants remaining associated with each race/ethnicity differ when controlling for SES and housing characteristics. It is worth noting that confounding by SES does not discount the role of race and ethnicity in exposure inequity for those pollutants but does influence results pertaining to modeling of asthma ED visits that controls for those covariates. Accounting for PM_2.5 and PM₁₀ reduced the IRR of asthma-related ED visits associated with Latinx residents by 19% and 38%, respectively, and accounting for SO₂ and PM₁₀ reduced the IRR associated with Black residents by 22% and 19%, respectively. Previous investigations of multilevel risk factors contributing to asthma ED disparities have relied on PM_2.5 as the single marker of outdoor air pollution (11, 13, 24). In light of these findings, a broader view of exposures is warranted in future investigations, especially when comparing exposure-based risks among three or more racial and ethnic groups.

These results describe how estimates of racial and ethnic disparities differ depending on whether differences in urban air pollution are taken into account. Importantly, the assumptions underlying causal mediation analysis are not claimed or evaluated, so our results should not be interpreted as such. In service of better intervention design, previous studies have apportioned distinct shares of asthma-related ED burden and racial disparity to poverty, urban residence, environmental exposures, and other social determinants of health (9, 11, 13, 24, 32). Complicating the interpretation of such an apportionment, even within the bounds of asthma-related ED visit investigations, is that the interrelationship between race and ethnicity, SES, and pollution exposure has been variously framed as moderation (33–35), as confounding (9, 11, 36), and as mediation (13, 37). Given the observational and ecologic nature of this study and the inability to rule out unmeasured confounding or moderation, this cross-sectional view of environmental injustice does not elucidate the social, legal, and economic processes leading to exposure inequity among racial and ethnic groups (6, 38). Nonetheless, these results suggest that even small persistent differences in air pollution concentrations among neighborhoods within a metropolitan area may relate to evident racial and ethnic asthma disparities and indicate that substantial disparities remain after accounting for air pollution.

We find that long-term air pollution concentrations show significant association with census tract asthma-related ED PBIR, despite the temporal mismatch of predicting acute health outcomes on the basis of chronic exposure. Studies examining the effects of air pollution exposure on acute asthma outcomes commonly consider daily or short-term lagged exposure (1, 5, 39–41). However, for the purposes of ecologic comparison, we find that it is advantageous to consider exposure to multiple pollutants at high geographic granularity, even if those exposure data do not provide daily time resolution. Annual averages do not account for daily and seasonal pollutant dynamics that affect acute risk but may indicate areas with a greater occurrence of high acute exposures. Because of the strong seasonal pattern in ozone concentrations, with clear peaks in the summer and comparatively low concentrations throughout the rest of the year, that pollutant was not investigated within our study design. It would be useful to extend this approach to other, more polluted cities, which may present different pollution dynamics.

Our approach to characterizing the effects of a multipollutant mix, although not definitive, uncovered a result that is consistent with the recent literature on source-linked particulate matter toxicity. Oxidative potential (OP) has been identified as a key factor influencing respiratory response to air pollution exposure (2, 4, 42) and provides a mechanism to explain observations that traffic proximity and wildfire contributions alter the per-unit asthma risks associated with PM_2.5 and NO₂ (43, 44). The interaction effect between PM_2.5 and SO₂ identified in this study, although not directly linked with OP, aligns with observations that particulate sulfur and sulfate increase the OP of trace metals and result in higher per-unit risks from PM_2.5 coemitted with SO₂ (2, 4). Emerging methods to determine how mixtures of exposures affect human health offer opportunities to further pursue this line of inquiry (45).

There are several other important limitations to consider. The health outcome data were collected for administrative purposes and could have been affected by miscoding. In addition, the population-based incidence rate does not differentiate between areas with a few individuals who visit the ED for asthma-related reasons many times per year and areas with many individuals visiting at a lower frequency. A related limitation of the study design is the possibility of ecologic fallacy: the risks associated with the share of Black or Latinx residents in an area are not equivalent to the risks associated with Black or Latinx identity. We could not control for individual risk factors, including exposure to secondhand smoke, although we do not expect a strong systematic relationship between secondhand smoke exposure and air pollution that is not already accounted for by other SES variables, which would introduce confounding bias. In addition, residuals of some models (in particular models of pediatric PBIR) showed significant spatial clustering, indicating the influence of unobserved environmental or social factors that exhibit similar spatial patterns. Finally, the focus on ED visits presents a limited view of asthma disparity.

Conclusions

In identifying a relationship between within-city spatial heterogeneity of air pollution and both census tract–level asthma-related ED visit rates and race- and ethnicity-linked asthma disparity, this study has several significant implications for public health and disparity research. First, moving beyond regional regulatory measurements to higher density monitoring of within-city air pollution is important for public health decision making, even in cities meeting federal standards. Second, mitigating pollution from sources located in predominantly Black and Latinx neighborhoods should be considered among interventions targeting environmental exposures and social determinants of health. To more closely target that mitigation, future investigations of air pollution–linked health disparities may benefit from a source-focused approach that considers both toxicity and degree of disproportionate exposure for communities of color.