Population-level SDOH and Pediatric Asthma Health Care Utilization: A Systematic Review

The research team, including an expert medical librarian, created the search strategy listed in Appendix 1. The initial literature search included 4 major databases from database inception to present: PubMed, Web of Science, Scopus, and Cumulative Index to Nursing and Allied Health Literature. Databases were last checked on May 25, 2021. This review was not registered.

We created a review protocol that identified studies evaluating the association between population-level SDOH and children’s asthma-related health care utilization. Empirical, observational studies were included in this review if: (1) outcome(s) included health care utilization, which refers to asthma-related ED visits, ED revisits, hospitalizations and/or readmissions in this review; (2) exposure(s) included ≥1 SDOH described by the HP SDOH framework; (3) outcome(s) and exposure(s) were both analyzed at the population-level using a geographic entity (eg, census-tract, zip-code); (4) results were explicitly reported for children ≤18 years; (5) the setting was in the United States; and (6) the full-text was available.

Studies were excluded if: (1) the results reported a mixed study population of children and adults; (2) the setting was outside of the United States; or (3) outcome(s) examined asthma prevalence; asthma-related clinic or urgent care encounters; or asthma-related costs and not our outcomes of interest. To optimize the specificity of any identified associations between SDOH and asthma outcomes, we excluded studies that used diagnosis codes for wheezing to identify their study population and/or outcome.

ED visits and hospitalizations were selected as outcomes to align with HP’s 2030 objectives that aim to reduce pediatric asthma-related ED visits and hospitalizations.⁷  In addition, we intentionally restricted study location to the United States given the variability in geographic entities used to characterize neighborhoods between countries and the authors’ objective to use the study’s findings to inform US-based population-level solutions.

Two independent reviewers (J.T. and A.G.) completed the title and abstract screening and full-text review utilizing Covidence’s online program (https://www.covidence.org/). For any differences, the 2 reviewers met to discuss and reach a consensus. Inter-rater reliability was assessed with Cohen’s κ (ĸ) coefficient, created through Covidence. Aligning with prior literature, ĸ coefficient values reflected the following amount of agreement: <0 none, 0 to 0.2 slight, 0.2 to 0.4 fair, 0.4 to 0.6 moderate, 0.6 to 0.8 substantial, and 0.8 to 1.0 almost perfect.^17,18  In addition, 1 reviewer (J.T.) examined reference lists of included full-text articles for additional studies that met inclusion criteria.

An Excel-based data form was created and piloted by 2 reviewers (J.T. and A.G.). One reviewer (J.T.) completed data extraction. Data extraction was cross-checked and confirmed for accuracy by a second reviewer (A.G.) for all articles. Extracted data elements included: study characteristics (authors, publication year, study design, study period); study population (number of participants or encounters, age range, location); exposure details (specific SDOH evaluated, overarching HP SDOH domain, source); outcome details (specific outcome studied, study period, source); spatial details (geographic entity of analysis); statistical analysis method used; and study findings.

This study was executed and reported according to the Preferred Reporting Items for Systematic Reviews and Meta-analyses checklist.¹⁹  Since most studies were observational and heterogeneous, only a qualitative synthesis was performed to identify patterns within and between studies. We classified studies and reported results according to the specific SDOH they examined as exposure variables, and more broadly, organized according to the social determinants’ overarching HP SDOH domain(s). These 5 domains include: Economic Stability (Economy), Education Access and Quality (Education), Health care Access and Quality (Health care), Social and Community Context (Community), and Neighborhood and the Built Environment (Environment).⁷  Studies that evaluated only 1 SDOH are only represented under that determinant section. For studies that evaluated >1 SDOH, findings are broken down and reported under each determinant evaluated. If a study evaluated a SDOH measure that combined >1 determinant, these were included under “Composite Measures.” For this review, if a study evaluated population-level proportions of individuals of a specific race and/or ethnicity as an exposure variable or covariate, we considered this an evaluation of structural racism belonging to the “Discrimination” determinant of the Community domain.²⁰  Study characteristics were summarized using counts and proportions.

Study quality was assessed for all studies included in the systematic review by 2 reviewers’ (J.T. and A.G.) consensus using the Joanna Briggs Institute checklists for cross-sectional and cohort studies.²¹  These tools assess (1) study population; (2) confounding variables; and (3) outcome and exposure variable ascertainment and analysis. Study quality was assessed after the full-text review.

Eight hundred and fifteen unique studies were identified after removing duplicates (Fig 1). Ninety-four studies were included in the full-text review. An additional 6 studies were identified from the reference lists of included full-text studies. Forty studies were ultimately included in the systematic review. Studies were frequently excluded for having a mixed population of children and adults or for the wrong study design. Inter-rater reliability was moderate for the title and abstract (ĸ = 0.44) and full text (ĸ = 0.52) screens.

Studies were published across 4 decades (1992–2021), and 85% used a cross-sectional study design. Study locations included 21 states in the United States, most frequently New York (n = 8), Texas (n = 7), and California (n = 6) (Table 1). Included studies evaluated SDOH from all HP framework’s domains. Most frequently, the determinants “Environmental Conditions” (Environment) followed by “Poverty” (Economy) were evaluated. Most studies evaluated multiple SDOH (n = 25) rather than a single SDOH (n = 15). Studies often used zip-code or zip-code tabulation areas (n = 16) as their geographic entity of analysis, followed by census-tract (n = 9), and census-block group (n = 6). Table 2 provides descriptions of geographic entities evaluated by included studies.

We identified 10 determinants across all HP SDOH domains evaluated as exposure variables: poverty, unemployment, high school graduation, enrollment in higher education, access to health services, access to primary care, discrimination, environmental conditions, housing quality, and crime and violence. Notably, there were SDOH across all HP domains that have yet to be evaluated in relation to pediatric asthma (listed in gray in Fig 2). These gaps in current literature include food insecurity and housing instability (Economic Stability domain); incarceration, civic participation, and social cohesion (Social and Community Context); language and literacy, as well as early childhood development and education (Education Access and Quality); health literacy (Health care Access and Quality; and foods that support healthy eating patterns (Neighborhood and the Built Environment).

For 8 of the evaluated SDOH, significant associations between adverse measures of these determinants and an asthma outcome were consistent among most studies (Fig 2). Conversely, only 33% (n = 1) of studies evaluating unemployment or educational attainment less than high school graduation found significant associations between these and pediatric asthma outcomes (Fig 2). Detailed findings are reported below, organized by HP SDOH domain and further separated by specific SDOH (Table 3). Supplemental Table 4 provides descriptions of the composite measures evaluated in included studies.

Sixteen studies evaluated poverty measures as an exposure variable, often using median household income (MHI) or percentage of the population living in poverty. Eleven studies found that decreased MHI or increased percentages of individuals living in poverty were associated with increased asthma-related ED visits and hospitalizations.^16,22–31  All studies (n = 8) evaluating MHI identified significant associations with asthma health care utilization.^16,24–30  Only 5 of 9 studies evaluating poverty found significant associations with asthma health care utilization. Another 5 studies found no significant association between percentages of the population living in poverty and asthma health care utilization.^32–36  Two studies analyzed data using analysis of variance (ANOVA) or Fisher’s exact test.^16,23  The remaining 14 studies used regression. Geographic entities varied by study: zip-code or zip-code tabulation area (ZCTA) (n = 9), census-tract (n = 3), county (n = 2), census block group (n = 1), and ozone region (n = 1).

Three studies evaluated population-level unemployment as an exposure variable measured by the percentage of unemployed adults. One study that used ANOVA found that unemployment rates significantly correlated with increased asthma-related hospitalizations.¹⁶  In the 2 other studies using regression, there was no association between unemployment rates and asthma-related ED visits or hospitalizations.^25,32  Two studies used census-tract as their geographic entity, and 1 used ZCTA.

Three studies, all using regression, evaluated the population-level percentage of adults whose educational attainment was less than high school graduation. Garcia et al found that this measure significantly correlated with increased asthma-related ED visits and hospitalizations in Georgia but only with increased ED visits in North Carolina.³⁰  Two studies identified no significant association between this measure and asthma-related ED visits²⁵  or hospitalizations.³⁶  Each study used a different geographic entity: ZCTA, census-block group, and county.

Two studies evaluated population-level percentages of adults enrolled in higher education. Beck et al specifically found that the percentage of adults older than 25 years with college education or greater was significantly and negatively associated with asthma-related hospitalizations using ANOVA and linear regression.¹⁶  Largent et al used negative binomial regression to identify that the proportion of the population with at least an associate degree was negatively and significantly associated with asthma-related ED visits and hospitalizations.²⁸  The first study used census-tract as its geographic entity, the latter used zip-code.

Six studies evaluated access to health services as an exposure variable using 4 different measures. One used ANOVA and linear regression to identify a significant association between the population-level percentage of households without access to a car and increased asthma-related hospitalizations by census-tract.¹⁶  Two studies evaluated populations of uninsured individuals at the county-level and demonstrated a significant correlation with increased asthma-related ED visits²⁵  and hospitalizations.³⁵  Neither of the 2 studies that evaluated provider density identified a significant association with asthma-related ED visits or hospitalizations.^35,37  Alcala et al found that the distance from a child’s home to the ED correlated to increased asthma-related ED visits and hospitalizations.²²  Garcia et al identified a significant county-level association between increased distance between a child’s home and an asthma specialist with asthma-related ED visits and hospitalizations in 1 of the 2 state populations (North Carolina).³⁰  This association was not significant in the second state (Georgia).³⁰  All studies except for the first, as noted, used regression.

Two of the above studies also specifically evaluated access to primary care as an exposure variable using regression. Alcala et al found that primary care shortage areas significantly correlated to increased asthma-related hospitalizations but not ED visits at the county-level.²²  Garcia et al found that increased distances to a primary care provider significantly correlated to increased asthma-related ED visits in 1 of 2 state populations.³⁰  However, the same study paradoxically found increased distances to a primary care provider significantly correlated to decreased hospitalizations in both states studied.³⁰ 

Nine studies evaluated population-level percentages of race or ethnicity composition as exposure variables. Seven studies used regression, 1 used cox proportional hazards, and 1 used Fisher’s exact test to analyze these associations. All studies that evaluated percentages of Black and/or African American individuals (n = 6) or Native American individuals (n = 1) living in a specific geographic entity found significant correlations with increased asthma-related ED visits and hospitalizations.^{24,27,34,36,38,39}  Beck-Sague et al found zip-codes with greater than 50% of African American individuals had an increased risk of asthma-related ED visits and hospitalizations.²³  One study found that the percentage of non-white residents significantly and positively correlated with asthma-related rehospitalizations.³³  Findings among studies (n = 6) that evaluated percentages of Hispanic individuals living in a specific geographic entity were mixed. Three studies found no significant association between this measure and asthma-related hospitalizations.^24,34,38  Only 1 study found a significant association with increased percentages of Hispanic individuals with decreased hospitalizations.³⁹  One study evaluated percentages of native- and foreign-born noncitizen Hispanic individuals: the former correlated to increased hospitalizations but the latter to decreased hospitalizations.²⁷  Six studies used zip-code or ZCTA as their geographic entity, 2 studies used census-block group, and 1 used census-tract.

Twenty-three studies evaluated measures of environmental conditions as their exposure variables with varied geographic entities: zip-code (n = 11), census-tract or census-block group (n = 6), other (n = 5), and county (n =1). Measures were varied and included air pollution markers, traffic and roadway proximity, waste and drilling sites, and indoor school hazards. Fourteen studies evaluated air pollution: 8 studies (57%) found a significant positive association between at least 1 studied pollutant and asthma-related health care utilization.^{31,38,40–45}  These studies varied greatly in terms of studied pollutants, measures of pollutants, and the geographic entity used. One study found a significant association between traffic-related air pollution and increased ED visits and hospitalizations.³²  Of 3 studies that evaluated roadway proximity using various distances, only 2 studies found significant associations between decreased proximity and increased hospitalizations.^36,39,46  One study evaluated traffic counts and road density and found significant associations between those 2 measures and hospitalizations.³⁶  All studies (n = 4) that evaluated the impact of waste, pollution, and drilling sites found a significant association with increased asthma-related ED visits or hospitalizations at the zip-code level.^29,47–49  One study (geographic entity = school district) found that some indoor school hazards (eg, windows, floor finishes) were associated with increased asthma-related hospitalizations.⁵⁰ 

Nine studies evaluated various population-level measures of housing quality as exposure variables. Regression was often used to analyze these associations, though 1 study used ANOVA and another used Fisher’s exact test. Two studies that evaluated vacancy rates demonstrated significant associations with increased ED revisits, rehospitalizations, and hospitalizations.^16,51  Two studies evaluated the proportion of renter-occupied homes, which were not significantly associated with ED revisits or rehospitalization in 1 study that used regression, but significantly associated with increased asthma-related hospitalizations in the second study that used ANOVA.^16,51  Housing code violation density was significantly associated with asthma-related ED revisits and rehospitalizations in adjusted analysis in the first study, but not significant in the second study that importantly adjusted for crime.^32,51  Four studies evaluated housing age as an exposure variable: 2 found that older homes were significantly associated with increased ED visits but only 1 of 3 studies found a significant association with increased hospitalizations.^23,27,34,52  Two studies that evaluated median home value found this significantly and negatively correlated to ED visits and hospitalizations.^16,36  Indoor hazards were evaluated in 2 studies with mixed findings of significance.^34,52  Studies typically used census-tract (n = 4) or zip-code (n = 4) as their geographic entity. One study used census-block group.

One study used regression to evaluate types of population-level crime as an exposure variable by census-tract. The authors found that both violent crime and all crime were significantly associated with increased ED visits and hospitalizations.³² 

Eight studies evaluated measures that combined a variety of SDOH. Studies that evaluated the Childhood Opportunity Index (n = 1), Neighborhood Risk Index (n = 1), and other smaller composite socioeconomic measures (n = 3) found that decreased opportunity, increased neighborhood risk, and decreased socioeconomic status were associated with increased asthma-related ED visits, hospitalizations, and reutilization.^27,52–55  Two studies evaluated the Area Deprivation Index but had mixed findings: increased deprivation correlated to increased readmissions in the first study⁵⁶  but not with reutilization or hospitalizations in the second study.⁵⁷  The 1 study that evaluated the Social Vulnerability Index found no association with asthma-related hospitalizations.⁵³  Four studies used census-tract as the geographic entity, 3 used census-block group, and 1 used zip-code.⁵⁸ 

Twenty-seven of 40 studies met all Joanna Briggs Institute checklist criteria. All studies clearly defined inclusion criteria for their study population. Most studies (n = 39) clearly described objective, standard methods for measuring their exposure, and outcome variables. Studies most frequently lost points for identification of confounding variables^{16,23,28,29,46,52}  (n = 6) and strategies for accounting for confounding variables^16,28,29,52  (n = 4). Supplemental Tables 4 and 5 provide comprehensive quality assessment results for all included cross-sectional and cohort studies, respectively.

Research that examines population-level health care and SDOH data by geographic entities is evolving, yet often restricted to single locations (eg, cities or states), limiting the generalizability of findings. This systematic review qualitatively synthesizes spatial associations between population-level SDOH and asthma health care utilization among children across the United States. This review’s results highlight consistent and significant spatial associations between 8 out of 10 studied place-based SDOH and pediatric asthma health care utilization throughout the United States. These SDOH represent all HP domains and include poverty, enrollment in higher education, access to health care, access to primary care, discrimination, environmental conditions, housing quality, and crime. Of these, we found that environmental conditions and poverty were most frequently evaluated, lending additional strength to the associations identified in this review between poorer environmental conditions and increased poverty with pediatric asthma-related health care utilization. Importantly, most studies included in this review met all quality criteria and, in turn, provide reliable, foundational data to inform which determinants should be the focus of solutions to improve asthma-related health among children.

Our systematic review applied the HP SDOH framework to classify studies spatially linking pediatric asthma health care utilization and population-level SDOH data. Prior asthma-specific literature reviews have organized SDOH using a biopsychosocial model as well as the World Health Organization’s conceptual framework for action on SDOH to emphasize the general relationship between SDOH and pediatric asthma.^59,60  Another review organized studies that have linked health outcomes data to SDOH data using geographic entities for all medical disorders across the entire population.⁶¹  Our current review builds on these prior works by incorporating both approaches – organizing studies using the HP SDOH framework and requiring included studies apply a geographic entity for analysis. In doing so, we can more specifically characterize gaps in current literature. We may also be able to better translate our findings to advocacy for population-level solutions that impact SDOH identified here as associated with asthma-related ED visits and hospitalizations. Further, as our review uniquely incorporated a place-based element to study associations between pediatric asthma health care utilization and SDOH, our findings complement existing work on patient-level social interventions. Prior asthma-specific work demonstrated that connecting families to resources to address unmet social needs is associated with reduced hospitalizations.¹¹  Additionally, social risk-based interventions were associated with reduced pediatric asthma-related ED visits and hospitalizations in a recent meta-analysis.⁶²  Our review’s findings provide a population-level perspective of SDOH that are important to consider, address, and perhaps incorporate into these effective patient-level interventions.

Our review highlights existing gaps and conflicting evidence in the current literature that can inform future research agendas. Nine SDOH, across all HP domains, have not been studied with respect to asthma-related ED visits and hospitals. It is difficult to identify reliable and accurate population-level measures to study these missing determinants (eg, social cohesion, health literacy). Thoughtfully constructing variables that measure these SDOH and examining their population-level impact on pediatric asthma outcomes using geographic entities may identify communities where interventions can be focused to optimize impact.

Our work identified conflicting results for some SDOH. We hypothesize that discordant findings are related to the variable methodology employed by these studies. For example, SDOH were evaluated by zip-codes or ZCTAs in one study versus counties or census-tracts in another. Zip-codes are larger geographic entities and contain more heterogenous populations compared with census-tracts or census-block groups, which may influence study findings.⁶³  Another possible explanation for disparate findings are the differing number and nature of confounding variables and covariates for which individual studies adjusted their analyses. For example, in 1 study,³²  associations between poverty, unemployment, and asthma-related health care utilization became nonsignificant after crime was included in the analysis. Finally, 5 studies included in this review evaluated composite measures of SDOH. Each of these includes various measures of SDOH that reflect multiple HP SDOH domains, and findings were not disentangled to isolate the effect of specific SDOH, which may have offered clarity for SDOH with mixed evidence.

SDOH are interrelated, have complex interactions with one another, and vary greatly by place. Altogether, it may be beneficial to advocate using a composite SDOH measure to streamline future work aiming to characterize the spatial relationship between SDOH and pediatric asthma outcomes. In addition, it would be helpful to determine which non-SDOH variables are essential to control for in this type of work and to analyze by smaller geographic entities (eg, census-tracts) that reflect homogenous populations when feasible. Doing so would strengthen the collective methodology of this research, possibly resulting in powerful results that may better drive the creation and integration of socioeconomic policies with health care to impact SDOH at the community level.⁶⁴ 

This review has limitations. First, there is notable heterogeneity in the included studies, specifically in the methods. As mentioned, exposure variables had differing measures and definitions; various geographic entities were used for analysis; and covariates and confounding variables were inconsistent among studies. Despite these challenges, we successfully synthesized our included studies’ findings qualitatively. Second, studies often used utilization event counts or population-based rates as outcome variables. Count outcomes are not normalized to a population, making it difficult to interpret whether geographic locations have high utilization counts because there is higher morbidity or because there are more people with asthma living there. Ideally, future research will use at-risk rates that divide utilization encounters by only the pediatric asthma population for a given area, which can be derived when city and state-level disease-specific registries are created.^65,66  Third, nearly all included studies used a cross-sectional study design analyzing retrospective data, making it not feasible to determine causality between SDOH and pediatric asthma health care utilization. Fourth, analyses that use population-level exposure variables may be biased by ecological fallacy, though research has demonstrated that individual and population-level SDOH typically align.^67,68  Finally, despite a rigorous literature search, we may have excluded conference proceedings or abstracts that could be particularly important for spatial science and related fields like computer science, informatics, and environmental science. The absence of gray literature within our review may indicate that our results are impacted by publication bias.

Population-level SDOH are associated with pediatric asthma health care utilization when evaluated by geographic entities and organized by the HP SDOH framework. However, these findings remain limited in their ability to inform population-level solutions given large heterogeneity in study methods. We recommend that future research prioritize using similar research methodology to measure SDOH and analyze these associations to strengthen this area of research and, in turn, its ability to impact pediatric asthma health outcomes.: