Analyzing Racial Disparities in Pediatric Atopic Comorbidity Emergency Department Visitation using Electronic Health Records

Abstract

Background:

Although atopic diseases and associated co-morbidities are prevalent in children, little is known about racial differences in emergency department (ED) visitation.

Objective:

We sought to examine racial differences in ED visitation among children with allergic comorbidities.

Methods:

We conducted a retrospective study of patients (<21 years) who visited the ED at a large pediatric hospital for atopic dermatitis (AD), food allergy (FA), asthma, allergic rhinitis (AR), and eosinophilic esophagitis (EoE) from 2015 to 2019. We determined the probability of ED encounter-free using hazard ratios (HR) and time to recurrence (TTR) of ED encounter for patients identified as Black/African American (AA) and White/European American (EA). We assessed potentially underlying allergic, demographic, and place-based factors, and potential interactions between factors.

Results:

A total of 30,894 patients (38% AA, 62% EA) had 83,078 ED encounters (38,378 first ED encounters, and 44,700 recurrent ED encounters) during the study period. Asthma and AR showed the highest rate of comorbidity in ED encounters in both AA and EA children. AA children exhibited higher HR for encounter following index AD and asthma encounters. We found an interaction between the type of insurance and race in ED encounters for AD, FA, AR, and EoE. In AA children, those insured by Medicaid demonstrated a higher HR for any encounter compared to those with commercial insurance. Conversely, in EA children, those with Medicaid insurance showed a lower HR compared to their commercially insured peers. Regardless of race, allergic comorbidity increased the HR of ED encounter (1.12–1.62) for all allergic diseases. At 5-years follow up, mean differences in TTR were shorter in AA children compared to EA children in AD, FA, and asthma.

Conclusion:

Identification of disease-specific racial disparities in ED visitation related to atopic diseases is a necessary first step toward the design and implementation of interventions capable of equitably reducing emergency care in atopic comorbid children.

Graphical Abstract

Introduction

In the atopic march, atopic comorbidity refers to the co-occurrence of early-life allergic manifestations of atopic dermatitis (AD), with subsequent development of other atopic diseases, including food allergy (FA), asthma, allergic rhinitis (AR) and eosinophilic esophagitis (EoE) in sequential patterns.(1, 2) Although the epidemiology of allergic diseases, and the comorbidities they cause, are well established in White/European American (EA) children, less is known for Black/African American (AA) children. While the underlying mechanisms in pathogenesis of allergic diseases are not fully understood, it is generally accepted that the interplay of individual genetic predisposition and environmental factors contribute to the development.(3–5)

Racial disparities in prevalence and morbidity of allergic diseases (e.g., AD, FA, asthma, AR and EoE) are well documented. A 2003 survey conducted in the United States found Black race was significantly associated with a higher prevalence of AD after controlling for possible confounders.(6) Several studies have shown that AA children are at increased risk of food sensitization.(7) However, other studies were inconclusive on differences in FA prevalence across racial groups.(8) EA children have a higher prevalence of AR compared to AA children.(9) Asthma morbidity and mortality are both higher in children who identify as AA compared to those who identify as EA.(10) Finally, Weiler et al., identified significant differences between AA and EA patients in the type and timing of symptoms of EoE.(11) Such variable findings may be explained by the reality that ancestry and race are related but distinct concepts – the latter is a social construct and not one rooted in genetics or biology.

Few studies have looked into racial differences in atopic comorbidities. To our knowledge, the Mechanisms of Progression of Atopic Dermatitis to Asthma in Children (MPAACH) cohort was the first to investigate disparities between AA and EA children for AD, FA, and asthma diagnoses.(12) More recently, researchers have used genome-wide association studies (GWAS) to understand genetics of allergic diseases.(13, 14) Despite the high prevalence of AD in children and its role in predisposing them to, or being an antecedent to, other allergic conditions, there is a limited understanding of racial differences in atopic comorbidity, and links to emergency care utilization.(15, 16) Through such an understanding, we may be able to develop interventions aimed at both prevention and treatment. In this study, we examined racial differences in atopic comorbidity-related emergency department (ED) recurrent encounters, using individual-level factors and place-based exposures.

Methods

This retrospective population-based study included ED encounters occurring between January 1, 2015 and December 31, 2019 for patients aged from 0 to 21 years old on January 1, 2015, living in Butler, Clermont, Hamilton, and Warren counties in Ohio, and Boone, Campbell, and Kenton counties in Kentucky. Details of data extraction and processing were previously described.(17, 18) Briefly, we obtained data from the electronic health record (EHR) at Cincinnati Children’s Hospital Medical Center’s (CCHMC). We included patients with at least one ED encounter for one of the following atopic-related diseases during the study period, using International Classification of Diseases, 10^th Revision (ICD-10) diagnosis codes: AD (ICD-10=L20), FA (ICD-10=Z91), asthma (ICD-10=J45), AR (ICD-10=J30), and EoE (ICD-10=K20) (Figure 1). For the processing of ED encounters, if two recorded visits for the same disease had a difference in date that was less than seven days, then they were counted as one visit. If two recorded visits for different allergic diseases happened within seven days, they were considered as having the same first visit date. Only patients self- or parent-reported as AA (Black) or EA (White) that visited the ED for an allergic diagnosis within the study period were included. Exclusion criteria and the analytic framework are shown in Figure 1. The study protocol was approved by the CCHMC Institutional Review Board (IRB). Additional information about the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) checklist is included in the Supplementary Materials (Table E1).

Figure 1.

Flowchart of emergency department (ED) encounters exclusion criteria and analysis. The study includes a total of 30,894 children with 83,078 valid ED encounters for atopic dermatitis (AD), food allergy (FA), asthma, allergic rhinitis (AR), and eosinophilic esophagitis (EoE).

Health outcome

Two outcomes of interest were estimated for each individual disease (AD, FA, asthma, AR, EoE): patient probability of ED encounter-free survival time as measured by HR, and time to recurrence (TTR) of ED encounter for each disease (disease-specific). Recurrent visits (patients with two or more repeat ED visits of the same disease type) during the study period were included in the analysis.

Covariates

The primary independent variable was race, defined according to self-report in the EHR (AA and EA). To account for allergic comorbidities in disease-specific ED models, we developed a time-varying covariate to track historical atopic diagnoses prior to each ED event during the study period (“Yes” if the patient had one or more previous comorbid atopic diseases and “No” otherwise). For instance, in the asthma ED encounter model, if a specific patient received a ED diagnosis of AD, FA, AR, or EoE at any given time point during or prior to the study period, the time-dependent comorbid status was set to “Yes” for any asthma ED encounters occurring after that diagnosis. We also extracted additional individual-level data from the EHR, including age, sex, insurance type, and street address at each visit.(14, 19, 20) Residential addresses at the time of each encounter were geocoded and linked to place-based risk factors using DeGAUSS. DeGAUSS is a decentralized and reproducible system designed to characterize community and environmental exposures (e.g., greenspace, pollution, among others) under a specific geographic scale using geospatial data sources, all while preserving the confidentiality of sensitive health data.(21, 22) Greenspace was computed as the average normalized difference vegetation index (NDVI) at 500 meters, 1500 meters, and 2500 meters within a buffer area around the patient’s address. Weekly averages of daily particulate matter less than 2.5 μm (PM_2.5) were derived from satellite imagery for the timing and location of each ED encounter. Greenspace and PM_2.5 were treated independently due to near zero correlation in our dataset. Time-varying covariates were included for each subject longitudinally.

Statistical analysis

Descriptive statistics (medians with interquartile range [IQR] for continuous variables and frequencies with percentages for categorical variables) of patient and ED encounter characteristics were generated and stratified by self-reported race. To compare AA and EA children, we used Wilcoxon rank-sum tests or Chi-square (χ²) tests for independent observations of patient and ED encounter characteristics, and univariate linear mixed models for longitudinal measurements of patient and ED encounter characteristics. For atopic comorbidities, we estimated a race-specific Jaccard index to measure similarity between atopic diseases at ED encounters. The Jaccard index can be a value between 0 and 1 (representing 0 to 100%), with 0 indicating no overlap and 1 complete overlap between the two sets (co-occurrent diseases).(23) We assumed the patient’s age at the initial disease-specific ED encounter during the study period as age of onset for constructing incidence curves. Finally, patient probability of ED encounter-free survival time, and TTR of ED encounter were calculated under a Bayesian time-to-event framework with an explicit causal thinking.(24)

Time-to-event analysis

For time-to-event analyses, we modeled each disease (AD, FA, asthma, AR, and EoE) separately, during the study period, from January 1, 2015 to December 31, 2019 encompassing 60 total months (5 years). For each patient, the starting date was the date of his or her first ED encounter after the starting date of the study period (January 1, 2015). Then time to each subsequent event was calculated relative to that common patient’s starting date; right censoring was applied at the end of the follow-up period on December 31, 2019.

Using an M-Spline function with 10 degrees of freedom (δ = 10), we performed disease-specific Bayesian mixed-effects survival analysis for AA and EA children with and without atopic comorbidity for AD, FA, asthma, AR, and EoE.(25) Following previous work,(26) a directed acyclic graph was built to infer causal effects to the observational data.(27, 28) We applied the backdoor criterion to remove open paths and checked for colliders to avoid overcontrol in the final adjustment set (Figure 2). Regression models were adjusted for interactions between age at visit and sex, and between race and other individual (insurance, comorbidities) and place-based (greenspace, PM_2.5) variables. Continuous predictors (age at visit, greenspace and PM_2.5) were scaled (using their grand means and standard deviations) for modeling and descaled for interpretation accordingly.(29) We accounted for subject-specific differences including a frailty term for patients with recurrent ED encounters. We also calculated the disease-specific differences in TTR at five years of follow-up time (τ) among AA and EA children with and without allergic comorbidities. Disease-specific TTR was defined as the average event-free time from the date the first ED encounter occurs to the date of a recurrent ED visit among a population during a fixed follow-up time (τ). Disease-specific TTR differences provide estimates of the duration of event-free time gained (less frequent ED encounters) or lost (more frequent ED encounters) between the two racial groups in each disease.(30) Finally, we report Bayesian credible intervals (CI) using a 95% probability. R code for this article has been annotated and deposited as open-source code in GitHub at https://github.com/maurosc3ner/edreencounters_ttr_survival_2023. See article’s Online Repository at www.jacionline.org for further details.

Figure 2.

Causal diagram for the independent variable (race, comorbidities, individual- and place-based variables) and the dependent variable (disease-specific emergency department [ED] encounters). Causal and biased pathways, covariates, confounders, and unobserved (latent) variables are shown.

Model assessment

We compared the M-Spline function against exponential models (null and full set of covariates unscaled, scaled with frailty term) using the expected log predictive density (ELPD).(31) ELPD quantifies the theoretical expected log pointwise predictive density for new observations, where a higher ELPD score indicates better model fit. Because our models account for recurrent ED encounters, we used a leave-one-group-out (LOGO) scheme in the cross-validation computation (one patient, multiple ED encounters). Finally, we provide disease-specific censoring and Pareto Smoothed Importance Sampling (PSIS) plots for diagnostics and outliers (See article’s Online Repository at www.jacionline.org, Table E3, Figures E1–E4).(31)

Results

Demographic and racial difference in atopic comorbidity

The study cohort included 30,894 unique children (38.2% AA, 61.8% EA) aged 0 to 21 years on January 1, 2015 who used the ED for an atopic disease at least once between January 1, 2015 and December 31, 2019. These patients contributed a total of 83,078 ED encounters (38,378 first ED encounters and 44,700 recurrent ED encounters). All patients were right censored when they turned 21 years old or on December 31, 2019, whichever came first.

Racial differences and disease comorbidity

On average, AA and EA patients differed in the median age at their first ED encounter in all diseases except AD: FA (5.0 years vs 4.0 years; P<0.001), asthma (5.2 years vs 6.6 years; P<0.001), AR (7.0 years vs 7.9 years; P<0.001), and EoE (3.3 years vs 13.1 years; P<0.001) (Figure 3A and Table 1). AA children with ED encounters for AD, FA, asthma, AR, and EoE were more likely to live in areas with less greenspace than EA children (all P<0.001). EA children’s addresses had slightly higher exposure to PM_2.5 in their asthma-related ED encounters compared to AA’s addresses (P<0.001). Figure 3B and Table E2 show race-specific progression proportions for the first two ED encounters of different disease types. Asthma and AR co-occurrence exhibited the highest rate of comorbidity in AA and EA children, 0.35 and 0.21, respectively.

Figure 3.

Race-specific incidence curves for atopic dermatitis (AD), food allergy (FA), asthma, allergic rhinitis (AR), and eosinophilic esophagitis (EoE) using age at the initial incidence of emergency department (ED) care during the study period (A), and Jaccard Index Similarity Matrix between diseases (B).

Table 1.

Patient and emergency department (ED) encounter characteristics by self-reported race including age at first disease-specific ED encounter as disease-specific age of onset and disease-specific individual- and place-based risk factors.

	Overall n (%), Median [IQR] or Estimate (SE)	AA n (%), Median [IQR] or Estimate (SE)	EA n (%), Median [IQR] or Estimate (SE)	P-value
Patients	30,894	11,794 (38.2)	19,100 (61.8)
Patients with AD	5,662	3,442 (60.8)	2,220 (39.2)
Patients with FA	7,607	1,554 (20.4)	6,053 (79.6)
Patients with Asthma	13,014	4,540 (34.9)	8,474 (65.1)
Patients with AR	11,875	5,633 (47.4)	6,242 (52.6)
Patients with EoE	481	41 (8.5)	440 (91.5)
Age at initial ED encounter (yr)
AD	1.9 [0.7, 5.1]	2.0 [0.5, 5.5]	1.8 [0.8, 4.4]	0.413
FA	4.2 [1.4, 11.7]	5.0 [1.8, 12.7]	4.0 [1.3, 11.4]	<0.001*
Asthma	6.1 [2.7, 12.9]	5.2 [2.4, 11.4]	6.6 [2.9, 13.4]	<0.001*
AR	7.5 [4.0, 12.7]	7.0 [3.6, 12.3]	7.9 [4.3, 13.1]	<0.001*
EoE	12.7 [6.4, 15.9]	3.3 [2.4, 9.3]	13.1 [7.2, 16.1]	<0.001*
ED encounters	83,078	33,928 (40.8)	49,150 (59.2)
Allergic diseases
AD	10,782	6,743 (62.5)	4,039 (37.5)
Sex
Female	4865 (45.1)	3292 (48.8)	1573 (38.9)	<0.001*
Male	5917 (54.9)	3451 (51.2)	2466 (61.1)
Insurance type
Commercial	3,620 (33.6)	958 (14.2)	2,662 (65.9)	<0.001*
Medicaid	7,149 (66.3)	5,779 (85.7)	1,370 (33.9)
Time varying covariates
Greenspace	0.43 (0.0)	0.42 (0.0)	0.45 (0.0)	<0.001*
PM2.5 (μg/m3)	8.83 (0.02)	8.85 (0.03)	8.81 (0.08)	0.379
FA	15,820	3165 (20.0)	12655 (80.0)
Sex
Female	7097 (44.9)	1513 (47.8)	5584 (44.1)	<0.001*
Male	8723 (55.1)	1652 (52.2)	7071 (55.9)
Insurance type
Commercial	11214 (70.9)	1057 (33.4)	10157 (80.3)	<0.001*
Medicaid	4578 (28.9)	2104 (66.5)	2474 (19.5)
Time varying covariates
Greenspace	0.45 (0.0)	0.42 (0.0)	0.45 (0.0)	<0.001*
PM2.5 (μg/m3)	8.90 (0.02)	8.95 (0.05)	8.89 (0.10)	0.227
Asthma	31,956	13773 (43.1)	18183 (56.9)
Sex
Female	13869 (43.4)	5990 (43.5)	7879 (43.3)	0.785
Male	18087 (56.6)	7783 (56.5)	10304 (56.7)
Insurance type
Commercial	13,784 (43.1)	2,452 (17.8)	11,332 (62.3)	<0.001*
Medicaid	18,108 (56.7)	11,288 (82.0)	6,820 (37.5)
Time varying covariates
Greenspace	0.44 (0.0)	0.42 (0.0)	0.45 (0.0)	<0.001*
PM2.5 (μg/m3)	8.88 (0.01)	8.85 (0.02)	8.91 (0.05)	0.046
AR	22,240	9956 (44.8)	12284 (55.2)
Sex
Female	10885 (48.9)	5107 (51.3)	5778 (47.0)	<0.001*
Male	11355 (51.1)	4849 (48.7)	6506 (53.0)
Insurance type
Commercial	10119 (45.5)	1950 (19.6)	8169 (66.5)	<0.001*
Medicaid	12084 (54.3)	7993 (80.3)	4091 (33.3)
Time varying covariates
Greenspace	0.44 (0.0)	0.42 (0.0)	0.45 (0.0)	<0.001*
PM2.5 (μg/m3)	8.76 (0.02)	8.73 (0.02)	8.78 (0.5)	0.140
EoE	2,202	226 (10.3)	1976 (89.7)
Sex
Female	675 (30.7)	43 (19.0)	632 (32.0)	<0.001*
Male	1527 (69.3)	183 (81.0)	1344 (68.0)
Insurance type
Commercial	1,721 (78.2)	78 (34.5)	1,643 (83.1)	<0.001*
Medicaid	480 (21.8)	148 (65.5)	332 (16.8)
Time varying covariates
Greenspace	0.46 (0.0)	0.43 (0.0)	0.46 (0.02)	<0.001*
PM2.5 (μg/m3)	8.88 (0.06)	9.09 (0.18)	8.88 (0.37)	0.263

AD: Atopic dermatitis; FA: Food allergy; AR: Allergic rhinitis; ED: Emergency department; EoE: Eosinophilic esophagitis; AA: African American; EA: European American; IQR: Interquartile range; SE: Standard error; PM2.5: Fine particulate matter;

^*: P-value<0.05

Association of risk factors to atopic comorbidity ED encounters

Figure 4 and Table 2 summarize the results for the time-to-event analysis. Examining AD-related ED encounters, AA children had an 18% increased hazard ratio (HR=1.18, 95% credible interval [CI]=1.06 to 1.33) of AD-related ED encounters compared EA children. Among children with Medicaid, only AA (HR=1.13, CI=1.02 to 1.24) children had a higher likelihood for subsequent AD-related ED encounters when compared to AA children with commercial insurance. Compared to children with a single allergic diagnosis (non-comorbid children), EA children with allergic comorbidities other than AD had a higher HR for ED return (HR 1.12, CI=1.02 to 1.23). Finally, there was a negative association between exposure to PM_2.5 and the risk of AD-related ED encounters. For every one-unit increase in PM_2.5, the HR for AD-related ED encounters decreased by 1% in EA children after descaling.

Figure 4.

Model-based simulation of emergency department (ED) survival curves for atopic dermatitis (AD) (A), food allergy (FA) (B), asthma (C), allergic rhinitis (AR) (D), and eosinophilic esophagitis (EoE) ED encounters (E). We conditioned hazard ratios (HRs) on race (African American [AA] and European American [EA]), comorbid status (with and without a comorbid allergic condition), Medicaid-insured male, and marginal means for continuous risk factors.

Table 2.

Disease-specific hazard ratios (HRs) showing interactions between race, individual- and place-based risk factors and emergency department (ED) encounters for atopic dermatitis (AD), food allergy (FA), asthma, allergic rhinitis (AR), and eosinophilic esophagitis (EoE).

	AD HR (95% CI)	FA HR (95% CI)	Asthma HR (95% CI)	AR HR (95% CI)	EoE HR (95% CI)
Age at visit for females	0.88 (0.85 to 0.92)	0.86 (0.82 to 0.89)	0.84 (0.81 to 0.86)	0.97 (0.95 to 1.0)	1.0 (0.86 to 1.17)
Age at visit for males	0.80 (0.77 to 0.84)	0.85 (0.81 to 0.88)	0.75 (0.73 to 0.77)	0.93 (0.90 to 0.96)	0.79 (0.70 to 0.89)
Race
EA	Ref	Ref	Ref	Ref	Ref
AA	1.18 (1.06 to 1.33)	0.89 (0.79 to 1.0)	1.16 (1.07 to 1.25)	0.93 (0.86 to 1.0)	1.02 (0.76 to 1.39)
Race*Insurance type
Medicaid-insured AA vs Commercially-insured AA	1.13 (1.02 to 1.24)	1.19 (1.05 to 1.34)	1.21 (1.12 to 1.30)	1.06 (0.99 to 1.13)	0.98 (0.72 to 1.35)
Medicaid-insured EA vs Commercially-insured EA	1.08 (0.98 to 1.19)	0.94 (0.87 to 1.01)	1.06 (1.01 to 1.12)	0.95 (0.90 to 1.01)	0.76 (0.61 to 0.94)
Race*Previous allergic comorbidity
Comorbid AA vs Non-comorbid AA	0.95 (0.88 to 1.02)	1.35 (1.21 to 1.50)	1.42 (1.35 to 1.51)	1.62 (1.52 to 1.72)	1.07 (0.8 to 1.43)
Comorbid EA vs Non-comorbid EA	1.12 (1.02 to 1.23)	1.48 (1.40 to 1.58)	1.56 (1.48 to 1.64)	1.55 (1.46 to 1.64)	1.21 (1.03 to 1.41)
Race*Place-based exposures
Greenspace for AA	0.99 (0.95 to 1.02)	1.02 (0.96 to 1.07)	0.98 (0.95 to 1.0)	1.02 (0.99 to 1.05)	0.86 (0.74 to 1.01)
Greenspace for EA	0.97 (0.93 to 1.02)	1.02 (0.99 to 1.05)	0.99 (0.97 to 1.02)	0.99 (0.96 to 1.02)	1.0 (0.91 to 1.10)
PM2.5 (μg/m3) for AA	1.0 (0.97 to 1.03)	0.99 (0.95 to 1.04)	1.0 (0.98 to 1.02)	0.96 (0.94 to 0.99)	1.07 (0.93 to 1.24)
PM2.5 (μg/m3) for EA	0.95 (0.92 to 0.99)	1.0 (0.97 to 1.02)	1.01 (0.99 to 1.03)	0.97 (0.95 to 0.99)	1.04 (0.98 to 1.09)

AD: Atopic dermatitis; FA: Food allergy; AR: Allergic rhinitis; EoE: Eosinophilic esophagitis; AA: African American; EA: European American; HR: Hazard ratio; CI: Credible intervals; PM2.5: Fine particulate matter; Ref: Reference group

Among children seen in the ED for FA, AA children had an 11% decreased HR of FA ED encounters, and the 95% credible interval ranged from 0.79 to 1.0, suggesting that this finding is close to statistical significance. Our survival analysis revealed a significant interaction between race and insurance type. Specifically, we found that AA children with Medicaid were significantly more likely to return to the ED than those with commercial insurance (HR=1.19, CI= 1.05 to 1.34). Conversely, there was no significant association between insurance type and the risk of subsequent FA-related ED encounters among EA children. We also found that children with allergic comorbidities other than FA were more likely to experience subsequent FA-related ED encounters, an increase of 35% in AA children (CI=1.21 to 1.50) and 48% in EA children (CI=1.40 to 1.58) compared to those without comorbidity. There were no significant interactions between race, greenspace, PM_2.5, and the risk of subsequent FA-related ED encounters.

For disease-specific asthma ED encounters, AA children had a 16% increased HR of asthma ED encounters (HR=1.16, CI=1.07 to 1.25). AA children with Medicaid insurance were 21% more likely to return to the ED for an asthma-related encounter compared to those with commercial insurance (CI=1.12 to 1.30). EA children with Medicaid insurance were similarly more likely to return to the ED compared to their commercially-insured counterparts (HR=1.06, CI=1.01 to 1.12). Presence of allergic comorbidities other than asthma increased the HR for subsequent asthma ED encounters by 42% in AA (HR=1.42, CI=1.35 to 1.51) and 56% in EA children (HR=1.56, CI=1.48 to 1.64), compared to children with no comorbid allergic conditions. There were not significant interactions between race and place-based exposures pertaining to subsequent asthma-related ED encounters.

For AR-related ED encounters, AA had a 7% decreased HR of FA ED visits. It approached significance with a 95% confidence interval of 0.86 to 1.0. Children with allergic comorbidities other than AR had a higher HR of ED encounters than those without comorbidities (for AA children – HR=1.62, CI=1.52 to 1.72; and for EA children – HR=1.55, CI=1.46 to 1.64). There was a negative association between exposure to PM_2.5 and the risk of AR ED encounters. For every 1 μg/m³ increase in PM_2.5, the HR for AR-related ED encounters decreased by 4% for AA and 3% for EA children after descaling. There were no significant interactions between race, insurance type, greenspace, and the risk of subsequent AR-related ED encounters.

With disease-specific EoE ED encounters, EA children with Medicaid insurance were less likely to return to the ED for EoE-related encounters compared to their commercially-insured counterparts (HR=0.76, CI=0.61 to 0.94). We found no such association for AA children. EA children with allergic comorbidities other than EoE had 21% higher HR of ED encounters than those without comorbidities (HR=1.21, CI=1.03 to 1.41).

Time to recurrence of ED encounters

Table 3 summarizes disease-specific TTR differences during 5 years of follow-up. Overall, TTR estimates were shorter when comparing comorbid versus non-comorbid children. Specifically, among non-comorbid children, TTR differences between AA and EA children showed that, on average, AA had significantly shorter time free before recurrence in AD (TTR_Diff=−2.4, CI=−3.5 to −1.3), FA (TTR_Diff=−1.4, CI=−2.6 to −0.4), and asthma (TTR_Diff=−3.5, CI=−5.0 to −1.1) compared to EA children. Among children with allergic comorbidities, TTR differences revealed that EA children had significantly longer time free before recurrence in asthma (TTR_Diff=−1.8, CI=−3.6 to −0.4) compared to AA comorbid children. There were no TTR differences in non-comorbid AR and EoE, as well as comorbid AD, FA, AR, and EoE.

Table 3.

Disease-specific differences in time to recurrence (TTR) of emergency department (ED) encounters for atopic dermatitis (AD), food allergy (FA), asthma, allergic rhinitis (AR), and eosinophilic esophagitis (EoE), conditioned on a Medicaid insured male with and without comorbid allergic condition, and marginal means.

tau=60 months	AA RMST (95% CI)	EA RMST (95% CI)	Difference in RMST (95% CI)
AD
Non-comorbid	14.9 (6.7 to 27.8)	17.4 (8.1 to 30.8)	−2.4 (−3.5 to −1.3)
Comorbid	15.6 (7 to 28.6)	16.0 (7.3 to 29.2)	−0.4 (−1.4 to 0.5)
FA
Non-comorbid	17.5 (6.5 to 34.2)	19.2 (7.4 to 36.1)	−1.4 (−2.6 to −0.4)
Comorbid	10.3 (9.8 to 10.7)	10.4 (10.1 to 10.8)	−0.2 (−0.7 to 0.3)
Asthma
Non-comorbid	14.8 (3.3 to 37.3)	18.5 (4.4 to 41.3)	−3.5 (−5.0 to −1.1)
Comorbid	10.8 (2.3 to 31.8)	12.7 (2.8 to 34.9)	−1.8 (−3.6 to −0.4)
AR
Non-comorbid	22.2 (7.0 to 42.5)	22.7 (7.3 to 42.9)	−0.5 (−1.4 to 0.3)
Comorbid	15.4 (4.4 to 35.8)	16.5 (4.8 to 37.1)	−0.9 (−2.3 to 0.0)
EoE
Non-comorbid	6.8 (1.1 to 33.3)	9.4 (1.5 to 37.9)	−2.0 (−9.4 to 1.3)
Comorbid	6.2 (1.1 to 32.4)	7.6 (1.2 to 35)	−0.9 (−7.3 to 2.6)

AD: Atopic dermatitis; FA: Food allergy; AR: Allergic rhinitis; EoE: Eosinophilic esophagitis; AA: African American; EA: European American; RMST: Restricted mean survival time; CI: Credible interval.

Validation

Table E4 shows disease-specific validation results for included ED encounters. M-spline had better ELPD in all five diseases compared to the classic exponential function (ELPD diff>4). PSIS plots showed non-over-optimistic estimates in all diseases (k<0.7).

Discussion

We used a large EHR database, inclusive of all of our institution’s ED visits and a recurrent time-to-event approach, to identify the effect of race on atopic-related ED encounters. We used a Bayesian framework to model recurrent ED encounters and time-dependent covariates at each ED encounter time point. Overall, asthma and AR co-occurrence shared the highest rate of comorbidity in AA and EA children. In our multivariable time-to-event analysis, AA children had a higher HR of AD and asthma ED encounters. Results suggest that having any allergic comorbidity increased the HR of subsequent ED encounters for patients in all allergic diseases by at least 12%. Our disease-specific survival analyses revealed significant interactions between race and insurance type, indicating that the effect of Medicaid insurance on the HR for subsequent ED encounters in allergic diseases varies by race. Finally, our study found differences in ED utilization with shorter time free of recurrence in non-comorbid and comorbid AA children compared to EA children in AD, FA, and asthma.

Our research investigated racial disparities in atopic ED encounters. Although our dataset is based on age at initial incidence of ED care, our incidence data share similar patterns to others reported in prior literature.(1, 11) However, it is likely that for the majority of school-age children and young adults, the first diagnosis would have been established well before the study period in a non-ED encounter. Also, it may be unusual that a child would need to present to the ED with AR as the primary complaint, suggesting perhaps the child had presented with another acute complaint and the rhinitis was part of recognized comorbidities. Alternatively, presentation to the ED for such a complaint may indicate diminished access to primary, preventive care. Future research including data from across diagnosis and encounter types, emergent and non-emergent, might focus on investigating the natural course of the atopic march, as well as how interventions might explicitly halt/prevent allergic manifestations experienced during the atopic march sequence.(4, 32)

Patterns of disease comorbidity differed by race and specific allergic condition. Also, after adjusting for individual- and place-based factors, we found that children with any allergic comorbidity experienced an increased HR of subsequent ED encounters following the index encounter, ranging from 15% to 28% across atopic diseases. Moreover, ED utilization, and time free from ED utilization, varied by race and disease. This was particularly true for AA children with AD, FA, asthma, and EoE. Biagini et al. showed that different developmental trajectories of allergic diseases probably exist based on race.(20) However, more ED utilization by AA children is likely reflective of a multitude of factors, including differential social, economic, and environmental exposures and differences in access. Indeed, race is a social construct and factors associated with race, and racism, may influence disease progression and severity. Such exposures may also influence likelihood of comorbidities and may relate to our finding of higher HRs for comorbid atopic patients.(33–38) Although, we explored disease-specific comorbidity effects by race, comorbidity variance remains poorly understood and requires further investigation to explore other relationships (e.g., age, place-based risk factors) to improve atopic comorbid outcomes.

Our time-to-event analysis revealed AA children had a higher HR of AD and asthma ED visits, whereas close to statistical significance credible intervals in FA and AR suggest a higher HR for EA children. We found an interaction between race and insurance in AD, FA, AR, and EoE. Generally, AA children insured by Medicaid exhibited a higher HR, while EA children with Medicaid insurance showed a lower HR, as well as shorter TTR for AA children and longer TTR for EA children. We hypothesize that these findings might be partly explained by social (e.g., family behaviors, education attainment), economic, and environmental factors that disproportionately burden marginalized population groups. Our team and others have extensively documented racial disparities in asthma outcomes, which are often mediated by socioeconomic factors and associated hardships, likely emerging from structural racism and not anything inherent to race itself.(17, 26, 39) Such disparities shown here extend to other atopic diseases, following consistent patterns that characterize the experience of multiple health outcomes. We see depiction of such disparities as a vital step in the development of approaches aimed at their elimination.

There are certain implications of our findings, in the context of consistent findings illustrative of noteworthy disparities. We suggest that the mechanisms through which individuals experience disease (and health) are important for research studies to consider. We structured our analyses to document the patterns of morbidity over time, and disparities in the patterns of atopic diseases among children. There are also clinical implications that warrant investigation. For example, healthcare providers, from clinicians to community health workers, may see an index atopic ED visit as a sentinel event, a time to identify those health-promoting or health-detracting exposures which may prove relevant for a child and child’s family. Indeed, are there interventions that could be put into place during or after an ED visit to prevent the atopic comorbidity from occurring or, at least, to prevent it from resulting in further morbidity?

Paradoxically, PM_2.5 was found to be negatively associated with AD and AR related ED visits. A 1 μg/m³ increase in PM_2.5 was associated with decreased AD and AR-related ED encounters in both AA and EA children. Our group and others have previously reported the influence of air pollution on respiratory diseases.(20, 40, 41) It is highly probable that these associations are confounded by other unobserved clinical and social factors linked to PM_2.5. Therefore, further research is necessary to understand these complex relationships more clearly.

The primary strengths of our study were the longitudinal population-based design and the recurrent time-to-event approach. Thus, we were able to include covariates at the time of each ED encounter which allowed detailed specification of the outcome during the entire time course. Second, we captured aspects of the environment using two different place-based variables to explain the probability of event-free survival and the TTR of ED encounters for each allergic disease. To our knowledge, this is the first atopic comorbidity study evaluating racial disparities in atopic ED utilization using this breadth of available data. This set of strengths might inform the design of novel interventions for disease prevention, treatment, and prognosis of atopic comorbidity among patients with different environmental exposure burdens and experience of emergency care.

This study has some limitations. First, the clinical data were limited to information gathered from ED encounters only. Children might have received their first diagnosis of an atopic condition in different settings, such as primary or subspecialty care offices or during inpatient encounters. Diagnoses may also have been present before the commencement of our study period. Furthermore, children with restricted access to healthcare services may be underrepresented in diagnoses of atopic conditions at the ED. These limitations could lead to discrepancies between the age at initial ED care and the actual age of onset for each disease-specific incidence curve. Consequently, these constraints could also influence our estimates of time-to-event probability and time to recurrence. Second, our study relied on ICD-10 billing codes linked to primary diagnoses. This approach may have limited our ability to capture atopic conditions that were reported as secondary diagnoses. Therefore, any potential biases or inaccuracies in our data collection process could influence our findings. Third, even if CCHMC manages the large share of emergency pediatric care in Greater Cincinnati and our EHR ED encounters database only included patients from the Cincinnati metropolitan area, certain patients might be seen at other hospitals and would therefore be missed by our dataset. This could be particularly true in areas further from the main campus and those of older ages. Thus, our results may differ in, and may not generalize to, populations at other institutions. Fourth, our multivariable analysis is based on self-reported race in the EHR. Given prevailing demographics in our region, we focused our analyses on those identified as AA and EA. Future research might include genetic ancestry data, and additional population groups (e.g., Hispanic or Asian children).(42) It should be noted that we operationalized race in this study as the social construct it is, indicative of the complex milieu of exposures and experiences that constitute race in the US. Finally, due to the lack of more precise patient-level environmental data, the contribution of our place-based exposures estimated at area-level units may not accurately capture patient-level environmental exposures, leading to potential confounding in our multivariable analysis.

Conclusions

Using data from more than 30,000 patients and 83,000 ED encounters, this study identified racial differences in atopic related ED encounters. The study identified how ED encounters for atopic diseases differed by race. The racial differences in AD, FA, asthma, AR, and EoE could have been influenced by socioeconomic conditions, co-occurrence of atopic comorbidities, and environmental exposures. Future research in atopic diseases and health care disparities might focus on potential mechanisms, including relevant social and environmental risk factors, and enhanced prediction models to target enhanced interventions. Our time-to-event analysis may support the design of targeted interventions to equitably improve health outcomes in children with atopic comorbidities.

Highlight Box:

1. What is already known about this topic?

   Differences in prevalence and morbidity exist in allergic diseases (e.g., atopic dermatitis (AD), asthma, food allergy (FA), allergic rhinitis (AR), and eosinophilic esophagitis (EoE)) between Black/African American (AA) and White/European American (EA) children.

2. What does this article add to our knowledge?

   This study identified how racial differences in atopic comorbidity of emergency department (ED) visitation were mainly driven by socioeconomic factors and underlying allergic comorbidities. The study also found that AA children exhibited a higher hazard ratio (HR) for encounter in AD and asthma, and shorter time free of recurrence of ED encounter (i.e., more ED encounters) in AD, FA, and asthma compared to EA.

3. How does this study impact current management guidelines?

   This study does not directly impact current guidelines but knowing that there are racial differences in the morbidity (ED visits) related to allergic conditions should prompt identification of prevention strategies before and during emergency care.

Abbreviations

AA

African American

AD

Atopic Dermatitis

AR

Allergic Rhinitis

CCHMC

Cincinnati Children’s Hospital Medical Center

CI

Credible Interval

EA

European American

ED

Emergency department

EHR

Electronic health records

EoE

Eosinophilic Esophagitis

FA

Food Allergy

GWAS

Genome-wide Association Sequencing

HR

Hazard Ratio

ICD-10

International Classification of Diseases, 10^th Revision

IRB

Institutional Review Board

NDVI

Normalized Difference Vegetation Index

PM_2.5

Particulate matter less than 2.5 μm

TTR

Time to recurrence