Tree cover, health care access, Sociome Data Commons, and pediatric asthma: Chicago, 2010-2019

Tree cover, health care access, Sociome Data Commons, and pediatric asthma: Chicago, 2010-2019 | Authorea try { document.documentElement.classList.add('js'); } catch (e) { } var _gaq = _gaq || []; _gaq.push(['_setAccount', 'G-8VDV14Y67G']); _gaq.push(['_trackPageview']); (function() { var ga = document.createElement('script'); ga.type = 'text/javascript'; ga.async = true; ga.src = ('https:' == document.location.protocol ? 'https://ssl' : 'http://www') + '.google-analytics.com/ga.js'; var s = document.getElementsByTagName('script')[0]; s.parentNode.insertBefore(ga, s); })(); Skip to main content Preprints Collections Wiley Open Research IET Open Research Ecological Society of Japan All Collections About About Authorea FAQs Contact Us Quick Search anywhere Search for preprint articles, keywords, etc. Search Search ADVANCED SEARCH SCROLL Pediatric Allergy and Immunology This is a preprint and has not been peer reviewed. Data may be preliminary. 20 January 2025 V1 Latest version Share on Tree cover, health care access, Sociome Data Commons, and pediatric asthma: Chicago, 2010-2019 Authors : Sandra Tilmon 0000-0002-1990-1197 [email protected] , Shashi Bellam , Kathy Bobay , Ellen Cohen , Emily Dillon , Brian Furner , Sarah E. Gray , … Show All … , Julie Johnson , David Meltzer , Doriane Miller , Sharmilee Nyenhuis , Jonathan Ozik , Carlos Santos , Anthony Solomonides , Julian Solway , Elizabeth Zampino , Sanjaya Krishnan , and Samuel L. Volchenboum Show Fewer Authors Info & Affiliations https://doi.org/10.22541/au.173738104.40873058/v1 416 views 230 downloads Contents Abstract Authors Credit statement Key Messages Abbreviations Introduction Materials and methods Results Variable selection Modeling Discussion Appendices Supplement 2: Highly correlated variables Supplement 4: Correlation matrix References Information & Authors Metrics & Citations View Options References Figures Tables Media Share Abstract Background Pediatric asthma exacerbations remain a critical public health concern, particularly in historically underserved urban settings. In Chicago, non-Hispanic Black children 5-19 years old had 2.5 times the emergency visits for asthma as non-Hispanic White children. Objective This study investigates sociome factors – the social context of disease – associated with asthma exacerbations among children living in Chicago’s South Side, leveraging clinical and census tract-level datasets. The aim is to uncover novel influences for potential new interventions. Methods A generalized linear model assessed associations while accounting for clustering at the patient level. Results Predictors of decreased risk included patient age (+4.8 years, -22%), tree crown density (+6% coverage, -17%), parks per acre (+0.41, -8%), and labor market engagement (+0.8 points, -9%). Conversely, predictors or increased risk included increased distance to the nearest pharmacy (+0.28 miles, +12%), limited English skills (+2.3%, +10%), higher inequality (+0.08 points, +8%), and visits in the Spring (+11%) and Fall (+20%). Conclusion The results suggest that tree crown density, a novel finding in the context of asthma exacerbations, may play a protective role. Limited access to health care facilities such as pharmacies continues to complicate care. However, the complexity of neighborhood-level influences require broader geographic sampling; limitations include the study’s restricted geographic and demographic scope. Integrating data from multiple hospitals will be essential for replicating these findings and translating them into actionable strategies for improving pediatric asthma care. Authors Sandra Tilmon MS MPH 1 , Shashi Bellam MD 3 , Kathy Bobay MSN PhD, Ellen Cohen, MPP 1 , Emily Dillon, PhD 4 , Brian Furner MS 1 , Sarah E. Gray MD ,6 , Julie Johnson PhD MPH 5 , David Meltzer MD PhD 6 , Doriane Miller MD 6 , Sharmilee Nyenhuis MD 1,6 , Jonathan Ozik PhD 7 , Carlos Santos MD MPHS 8 , Anthony Solomonides PhD 3 , Julian Solway MD 6 , Elizabeth Zampino MS 1 , Sanjaya Krishnan PhD 9 , Samuel L. Volchenboum MD PhD 1 Affiliations : 1 University of Chicago, Pediatrics; 2 Loyola University Parkinson School of Health Sciences and Public Health; 3 Endeavor Health Research Institute; 4 Carroll University, Life Sciences; 5 University of Chicago, Clinical Research Informatics; 6 University of Chicago Medicine; 7 Argonne National Laboratory, Decision and Infrastructure Sciences Division; 8 Rush University Medical Center, Internal Medicine; 9 University of Chicago, Computer Science Corresponding author Sandra Tilmon, MS MPH Healthcare Data Scientist Data for the Common Good Department of Pediatrics University of Chicago Medicine 5841 S. Maryland Ave. | Chicago, IL 60637 [email protected] Cell: 773-255-6497 Funding: NIH UL1TR002389-07 Word count: 3,416 Tables: 2 Figures: 4 Credit statement Conceptualization: Cohen, Furner, Krishnan, Miller, Nyenhuis, Ozik, Santos, Solomonides, Solway, Tilmon, Volchenboum. Methodology: Krishnan, Tilmon. Software: Furner, Krishnan, Tilmon. Validation: Tilmon. Formal analysis: Tilmon. Investigation: Bellam, Gray, Nyenhuis, Santos, Solomonides, Solway, Tilmon. Data curation: Krishnan, Tilmon. Writing - original draft: Tilmon. Writing - review and editing: Bobay, Gray, Dillon, Furner, Krishnan, Meltzer, Miller, Nyenhuis, Ozik, Solomonides, Tilmon. Visualization: Tilmon. Supervision: Cohen, Furner, Krishnan. Project administration: Zampino. Funding acquisition: Cohen, Krishnan, Ozik, Solway, Volchenboum. ORC-ID Last name ORC-ID Bellam 0000-0003-3974-3613 Bobay 0000-0003-274-6602 Cohen 0000-0002-8596-0769 Dillon 0000-0002-9362-7085 Furner 0000-0001-7074-7247 Gray 0000-0003-3599-0203 Johnson 0000-0001-6563-5101 Krishnan 0000-0001-6968-4090 Meltzer 0000-0003-2790-7393 Miller 0009-0002-7231-1353 Nyenhuis 0000-0002-7381-9733 Ozik 0000-0002-3495-6735 Santos 0000-0002-6224-0371 Solomonides 0000-0003-2117-2461 Solway 0000-0002-0898-8530 Tilmon 0000-0002-1990-1197 Volchenboum 0000-0001-9863-851X Zampino 0000-0002-8233-5132 Conflicts of interest S.B. served on an advisory board for Sanofi. E.F.D. receives consulting fees from Argus Cognitive for software medical devices. S.E.G. served on an advisory board for Sanofi. S.N. served on an advisory board for Avillion/Astra Zeneca, receives royalties from Wolters-Kluwer and Springer, and research funding from NIH and Asthma Allergy Foundation of America. A.S. Holds voluntary positions in the American Medical Informatics Association and is an equity investor in healthcare companies and other industries. C.S. served on advisory boards for Gilead and Merck, receives royalties from Wolters-Kluwer, and research funding from CDC. J.S. reports a potential financial interest in PulmOne Advanced Medical Devices, Ltd, Israel, and research grant funding from NIH, NSF, and Respiratory Health Association of Metropolitan Chicago. S.T. receives consulting fees from Aetna. S.L.V. is co-founder and Chief Medical Officer for Litmus Health, Inc., and receives consulting royalties from CVS Accordant. Abstract Background Pediatric asthma exacerbations remain a critical public health concern, particularly in historically underserved urban settings. In Chicago, non-Hispanic Black children 5-19 years old had 2.5 times the emergency visits for asthma as non-Hispanic White children. Objective This study investigates sociome factors – the social context of disease – associated with asthma exacerbations among children living in Chicago’s South Side, leveraging clinical and census tract-level datasets. The aim is to uncover novel influences for potential new interventions. Methods A generalized linear model assessed associations while accounting for clustering at the patient level. Results Predictors of decreased risk included patient age (+4.8 years, -22%), tree crown density (+6% coverage, -17%), parks per acre (+0.41, -8%), and labor market engagement (+0.8 points, -9%). Conversely, predictors or increased risk included increased distance to the nearest pharmacy (+0.28 miles, +12%), limited English skills (+2.3%, +10%), higher inequality (+0.08 points, +8%), and visits in the Spring (+11%) and Fall (+20%). Conclusion The results suggest that tree crown density, a novel finding in the context of asthma exacerbations, may play a protective role. Limited access to health care facilities such as pharmacies continues to complicate care. However, the complexity of neighborhood-level influences require broader geographic sampling; limitations include the study’s restricted geographic and demographic scope. Integrating data from multiple hospitals will be essential for replicating these findings and translating them into actionable strategies for improving pediatric asthma care. Clinical implications These findings provide hypotheses for future interventions for long-standing asthma disparities. Key Messages Pediatric asthma in Chicago’s South Side has persistent health disparities. The Sociome Data Commons curates novel datasets along with social determinants of health to reflect the full social context of disease and intends to generate new hypotheses for action to reduce health disparities. Predicting asthma exacerbations, findings here reflect predictors of decreased risk of increased tree cover and parks and increased risk from lower health access, poverty-associated factors, and linguistic isolation. 7 keywords: pediatric asthma, Chicago, health disparities, trees, SDOH, sociome, social determinants of health Abbreviations ITM: Institute for Translational Medicine SDC: Sociome Data Commons ICC: Intraclass correlation coefficient GLM: General linear model EHR: Electronic health record IRB: Institutional Review Board ICD: International Classification of Diseases FAIR: Findable, Accessible, Interoperable, and Reusable PHI: Private health information HIPAA: Health Insurance Portability and Accountability Act of 1996 IQR: Interquartile range LASSO: Least absolute shrinkage and selection operator PCA: Principal component analysis LR: Likelihood ratio test QIC: Quasi information criterion AIC: Akaike information criterion SD: Standard deviation LOESS: Locally weighted scatterplot smoothing NOAA: National Oceanic and Atmospheric Association FEMA: Federal Emergency Management Association RAPT: Resilience Analysis and Planning Tool ACS: American Community Survey HUD: Housing and Urban Development DOT: Department of Transportation EPA: Environmental Protection Agency Introduction Asthma incidence, morbidity, and mortality are deeply influenced by social determinants, including access to health care and neighborhood-level factors such as pollution. 1 Such social vulnerabilities are rooted in long-standing structural inequities and systemic racism. 2 Health care in the U.S. has its own civil rights journey, evolving from the Freedmen’s Bureau for previously enslaved individuals into an era of segregated hospitals. 3 Although formal desegregation ended in 1963, 4 disparities in health care financing, access, and outcomes remain. 5 Access itself has been seen to modify asthma treatment effects. 6 The Institute for Translational Medicine (ITM), a consortium of Chicago-area research hospitals, launched an initiative aimed at addressing and reversing the negative health impacts of structural racism. By accounting for the social, environmental, behavioral, psychological, and economic factors that shape health – the full range of lived experience, or the “sociome” – ITM seeks to identify new strategies to reduce health disparities. Further, by intentionally framing health disparities within their historical and social contexts, inadvertent assumptions that such disparities are natural, genetic, or inevitable are avoided. This approach enhances medical and public health research, making it actionable and solution oriented rather than solely descriptive. 7 Focusing on the sociome is essential because: 1. sociome factors interact with human biology, intensifying or causing disease and injury, and 2. illness, regardless of origin, can amplify the adverse effects of these factors. 8 As part of its research platform, ITM has partnered with Argonne National Laboratory, the University of Chicago’s Data Science Institute, and Data for the Common Good to develop the Sociome Data Commons (SDC). This growing resource is designed to collect high-quality, generalizable public data on sociome factors, following best data practices as outlined in a prior publication. 9 The SDC brings together social determinants of health data while also integrating novel, carefully curated data, reducing barriers for researchers to incorporate sociome factors into their studies. This in turn supports the discovery of previously unrecognized sociome impacts on health and disease and helps identify potential avenues for novel interventions. Despite decades of community outreach and interventions by local consortia, Chicago continues to experience persistent disparities in asthma morbidity, 10–13 highlighting a need for novel interventions. This paper examines pediatric asthma exacerbations at an urban research hospital serving an under-resourced area of Chicago. By integrating clinical data with novel datasets from the SDC, we aim to identify factors contributing to acute asthma episodes, offering a foundation for developing new hypotheses and interventions. Materials and methods Clinical data Data from the University of Chicago Epic (Epic Systems, Verona, WI) electronic health record (EHR) for all pediatric (under 18) visits was collected; this study was approved by the University of Chicago Biological Sciences Division (BSD) IRB #21-1920, and a waiver of consent was granted. Clinical data (address history, demographics, diagnoses, encounters, labs, medications, and notes) were stored and analyzed on University of Chicago HIPAA-compliant infrastructure hosted by the Center for Research Informatics. Python 14 was used for data management. Encounters were limited to visits for asthma, defined as either an asthma diagnosis 15 or “asthma” as text in the encounter. Differential diagnoses were present 16 in 3% of visits and evaluated by a team of specialists (SB, SG, SN). The National Institute of Health’s Value Set Authority Center was our primary source of ICD codes 17 ( Supplement 1 ). An asthma visit for an exacerbation was defined by the specialist team as any of the following: 1. The asthma encounter contains a text description including “exacerbation”; 2. The visit was to the Emergency Department; 3. The patient was prescribed systemic corticosteroids 18 ( Supplement 1 ). This definition ensures that exacerbations are identified comprehensively, capturing both direct documentation of exacerbations and clinical actions likely to be associated with them. Asthma phenotypes were also defined with the specialist team. Atopic was defined via elevated immunoglobulin E (IgE) or diagnosis of any of rhinitis, eczema, or atopic dermatitis. Other phenotypes included inflammation- (T2-High, elevated eosinophils) or obesity-related (obesity diagnosis or BMI percentile >= 95%). IgE and eosinophil values came from laboratory results. Asthma control test scores were extracted from the EHR via regular expression and categorized by age but were highly missing. Missing data were handled in the following ways. All columns with greater than 30% missing were inspected for relevance and imputation feasibility; most were found unnecessary and removed. Phenotypes and asthma control were deemed infeasible for imputation due to a lack of relevant information. Missing data were also examined for patterns: missing insurance (2%) and race/ethnicity (<1%) were not associated with one another by correlation or dendrogram. Filling in missing information backwards and forwards by patient reduced missing insurance by half but had no effect on missing race/ethnicity. Multiple imputation 19 was conducted for insurance and race/ethnicity, resolving the remaining missing values. In 2016, visit definitions in the EHR were changed, altering record counts; sensitivity testing was conducted to determine if records before 2017 would be included. Values were tested for statistical significance via chi-square test for categorical variables and t-tests for numeric variables. Chicago lies along Lake Michigan to its East and is commonly split into three areas: the South, West, and North sides. Because 97% of asthma patients were from the South Side, analysis was limited to this geography. Many patients visited the hospital solely for asthma exacerbations; we assumed they received routine care elsewhere and only sought hospital treatment for acute episodes. To avoid bias in our patient sample, we restricted the dataset to only those patients who had at least one routine asthma visit. Before this restriction, asthma exacerbations were 62% of all asthma visits; this fell to 38% after excluding exacerbation-only patients. Sociome Data Commons data In our previous work, we described the development of the SDC, designed according to FAIR (Findable, Accessible, Interoperable, and Reusable) principles, and which included a poverty index derived from the American Community Survey. 9,20,21 Since then, we have enhanced the commons by curating census-tract-level datasets into thematic domains – social, environmental, behavioral, psychological, and economic – to address key challenges such as: 1. Finding relevant datasets; 2. Evaluating data quality and generalizability; and 3. Managing, processing, and integrating complex datasets. These enhancements aim to streamline the integration of sociome factors into research by reducing barriers for researchers. Additionally, select pollution and housing variables were analyzed for spatial clustering based on adjacency of census tracts. 22 Clustering was assessed using Moran’s I , with emphasis placed on the metric’s magnitude and scatterplots of the clusters instead of significant p-values. Only clustering exhibiting strong and consistent patterns in the scatterplots were retained. Raw data and Python code are included in the SDC to ensure transparency to users. In this paper, we further refined SDC data. Threshold effects identified through partial dependence plots led to the conversion of facility counts, such as community centers and landfills, into binary indicators of presence within census tracts. Analytic dataset Geocoding to latitude, longitude, and census tract was conducted with GoogleV3 and GeoPy 23 for SDC public dataset addresses; this method was chosen for ease of use given no need to protect protected health information (PHI). To maintain security and private of PHI, the HIPAA-compliant DeGAUSS platform was used; patient address history was available for visit-specific addresses and geocoded to latitude, longitude, census block, and census tract. 24 For interpretability, distance in miles between patient location and points of interest was calculated via the Haversine formula. 25 Computation time was reduced by using spatial indices and parallel processing. Points of interest included the closest health care site (primary care, hospital, and pharmacy) and various pollution sites. 26,27 To prevent extreme values from influencing the analysis, continuous variables were identified as outliers by Tukey’s 1.5 * the interquartile range (IQR) rule.28 The IQR, defined as the 25th to 75th percentiles, represents the middle 50% of the data. With this method, outliers are values outside the lower limit (quartile 1 − 1.5 * IQR) or upper limit (quartile 3 + 1.5 * IQR). Rather than excluding these outliers, any data points beyond these thresholds were capped at the boundary values. This approach minimizes the impact of extreme values while retaining the full dataset for analysis. Given seasonal effects with exacerbations, visits were categorized by quarter of the year. Summer (quarter 3) was used as the referent season and excluded from the analytic dataset. Clinical and SDC data were joined at the patient’s census tract. Weather metrics (temperature, humidity) with changes from the previous day were joined at the visit date. 29 After making binary indicators for categorical variables, the final dataset had 199 variables, 35 of which were patient-specific and the remainder describing the census tract. Excluding binary variables, continuous variables were rescaled to mean 0 and standard deviation (SD) 1 with standard scaler. 30 Variable selection Variable selection was performed using R 31 by combining variables chosen through three disparate methods: stepwise selection, least absolute shrinkage and selection operator (LASSO), and principal component analysis (PCA). Stepwise selection was applied in two ways: forward (starting with an empty model and adding variables) and backward (starting with a full model and removing variables). LASSO is a regularization method that penalizes variables by shrinking less important variables’ coefficients to 0. Unlike stepwise selection or LASSO, PCA does not predict an outcome. Instead, it combines variables into a smaller set of components, with each variable’s weight reflecting its importance, thus ranking variables. For PCA, variables in the first component with loadings above the median were retained. Variables from all methods were combined, and highly correlated variable pairs (|0.8| or greater) were reduced to a single representative variable ( Supplement 2) . Asthma visits were modeled with an exacerbation as the outcome. Patient race/ethnicity and insurance variables were excluded to prevent overfitting on race and because public health insurance is a known proxy for poverty. Modeling A multi-level generalized linear model (GLM) was conducted using STATA 18. 32 GLMs are semi-parametric and account for clustering at the patient level while also allowing for variability in both the number and timing of repeat visits. 33 The intraclass correlation coefficient (ICC) estimates how much variance is accounted for by the multilevel factor 33 and was calculated separately for the patient, the patient’s census tract, and a combination of both. All parameters were tested. A logit or probit link was chosen via a likelihood ratio (LR) test. Different correlation structures (independent, exchangeable, and auto-regressive(1)) were tested via quasi information criterion (QIC).34 A “saturated” model was run, including all 129 variables from the variable selection as a preliminary model. To balance model complexity with predictive performance, this model was compared to a reduced version limited to predictors with an upper p-value threshold of 0.10. This reduced model is nested within the saturated model, enabling the use of a LR test to evaluate the fit between them. The null hypothesis is that the reduced model performs as well as the saturated (e.g., p ≥ .05), while the alternative hypothesis is that they perform differently (e.g., p < 0.05). Akaike information criterion (AIC) values, which assess model fit while accounting for the number of variables, were also compared,34 with lower AIC values indicating better fit. To examine overall explanatory power, McFadden’s Pseudo R-Squared was calculated as 1 - (log likelihood of final model / log likelihood of null model).35 A null model is a baseline model that contains the outcome, but no predictors. Results EHR artifact at the end of 2016: Sensitivity check The sensitivity check for the EHR artifact showed significant and meaningful differences in insurance types. However, the plan was to exclude insurance, as the majority of patients have public insurance, mostly Medicare with some Medicaid, and this is a known proxy for poverty. Statistically significant differences were also found with asthma phenotypes. However, these variables were too sparse to include in the model ( Supplement 3 ). Descriptives 157 The pediatric asthma patient population mainly consisted of non-Hispanic Black children (84%), covered by public insurance (65%), an average age of 8 years old, and a slight male predominance (58%). Residential mobility was small, with most residing in a single census tract during the study period. On average, patients had 4 visits, though visit frequency varied widely with a SD of 5 visits ( Table 1 ). Most patients had not experienced exacerbations. Patients without exacerbations tended to be slightly younger (7.8 years vs. 8.7 years), had fewer visits (5.1 vs. 6.7), and resided in fewer home census tracts (0.6 vs. 1.5). Race- and insurance-based differences were present: non-Hispanic Black patients were slightly more likely to experience exacerbations (87% to 82%) compared to non-Hispanic White patients (4% vs. 7%). Similarly, patients with public insurance were slightly more likely to have had exacerbation visits (69% vs. 62%) compared to those with private insurance (29% to 37%) ( Table 1 ). Overall Ever had an exacerbation Never had an exacerbation Totals 3,421 1,585 1,836 Male n (%) 1,970 (58%) 907 (57%) 1,063 (58%) Age mean (SD) 8.3 (5.1) 7.8 (4.8) 8.7 (5.3) Race/ethnicity, n (%) Hispanic 241 (7%) 92 (6%) 149 (8%) Non-Hispanic: Asian or Middle Eastern 50 (2%) 19 (1%) 31 (2%) Black/African-American 2,887 (84%) 1,385 (87%) 1502 (82%) White 198 (6%) 66 (4%) 132 (7%) Multiple and all others 45 (1%) 23 (1%) 22 (1%) Insurance n (%) Private 1,139 (33%) 465 (29%) 674 (37%) Private with Medicaid 40 (1%) 20 (1%) 20 (1%) Public: Medicare, Medicaid 2,219 (65%) 1,089 (69%) 1,130 (62%) Self pay and other 23 (<1%) 11 (<1%) 12 (<1%) Number of visits, mean (SD) 4.1 (5.1) 5.1 (6.7) 6.7 (6.4) Number of census tracts, mean (SD) 1.3 (0.6) 0.6 (1.5) 1.5 (0.8) Table 1 : Descriptive characteristics of pediatric patients with at least one routine visit for asthma by ever having a visit for an asthma exacerbation. SD: Standard deviation. Variable selection When comparing methods, 37 variables were selected by all 3 methods. 33 variables were selected by both stepwise selection and LASSO, 27 by LASSO and PCA, and only 1 by stepwise and PCA. For variables not detected by either other method, stepwise selected 2, PCA 5, and LASSO 37. From the original 199 candidate variables, 88 were selected from stepwise selection, 152 from LASSO, and 84 from PCA. After removing 23 highly-correlated variables, the final count was 129 ( Supplement 2 ). Modeling GLM ICC values were as follows: for patients, 27% of variance; census tract, 6%; both, 27%. Given the negligible amount of increase in variance explained from census tracts, the model included patients only as the multiple level variable. After filtering the saturated model to variables with p-values <= 0.10, the variable count was reduced from 129 to 25 ( Table 2 and Figure 1 ). The reduced model did not differ in explanatory power from the saturated one and so was selected (χ 2 (101) 79.05, p = .9480). The AIC for the reduced model (16,750.82) was also lower than for the saturated one (16,873.78), showing a better fit. The final model, however, had a notably low McFadden’s pseudo r 2 of 0.02. For variable selection, the most successful selection came from LASSO, with 19 of the 20 significant variables in the final model. In contrast, stepwise selection and PCA resulted in 14 and 10 variables, respectively. For patient-specific factors, higher age predicted decreased risk, with a one-SD increase reducing exacerbation odds (4.8 years, -22%). In contrast, odds increased with visits during Spring (quarter 2, +11%) and Fall (quarter 4, +20%). Among continuous variables, increased odds occurred with 1 SD increases in distance to the closest pharmacy (0.28 miles, +12%) and average sea level pressure (0.21 millibars, indicating weather changes, +6%). For census tract-level factors, several predictors of decreased risk were observed, described as the reduction in exacerbation odds per 1 SD increase: poverty index (3.1 points, -21%), tree crown density (6.4%, -17%), labor market engagement index (16.2 points, -13%), recent movers (7.8%, -12%), log-transformed traffic (0.8, -9%), and parks per acre (0.41, -8%). Conversely, predictors of increased risk were linked to other census tract factors: the presence of a community center (+6%), hazardous material business licenses (+8%), and designation as a HUD renewal community (+9%). Among continuous factors, a 1 SD increase raised odds as follows: GINI coefficient of inequality (0.08, +8%), low transportation cost index (6.4 points, +13%), NDVI (6 points, +11%), non-family household percentage (13.6%, +12%), limited English proficiency percentage (2.3%, +10%), and foreign-born population percentage (7.7%, +12%) ( Table 2 ). Category Variable Data source OR (95% CI) p-value Selection method(s) Patient-specific variables Age Electronic health record (EHR) 0.78 (0.74, 0.83) < 0.001 LASSO, Stepwise Closest pharmacy in miles (closest_pharma_miles) EHR, Chicago Data Portal 36 1.12 (1.06, 1.19) < 0.001 All Patient-specific visit date variables Average sea level pressure (avgsealvlpress) National Oceanic and Atmospheric Association (NOAA) 29 1.06 (1.01, 1.10) 0.008 Stepwise Visit in quarter 2 (visitq2) EHR 1.11 (1.00, 1.24) 0.042 LASSO, Stepwise Visit in quarter 4 (visitq4) EHR 1.20 (1.09, 1.32) < 0.001 LASSO, Stepwise Neighborhood (census tract) variables Community center, present (commctrcount_yn) Chicago Data Portal 1.06 (1.00, 1.12) 0.039 LASSO, Stepwise Gini index of income inequality (gini): 0 as perfect equality, 1 as perfect inequality Federal Emergency Management Association (FEMA), Resilience Analysis and Planning Tool (RAPT) 37 1.08 (1.02, 1.15) 0.014 LASSO, PCA Hazardous material business licenses, present (hazardcount_yn) Chicago Data Portal 36 1.08 (1.02, 1.15) 0.009 LASSO, Stepwise Homicide rate per 1,000 persons (homiciderate1000) Chicago Data Portal, 36 American Community Survey (ACS) 21 1.07 (0.99, 1.16) 0.093 LASSO, Stepwise HUD empowerment zone (hudempowerzone) Housing and Urban Development (HUD) 38 0.95 (0.89, 1.02) 0.143 LASSO, PCA HUD renewal community (hudrenewalcomm) HUD 38 1.09 (1.02, 1.16) 0.010 LASSO, Stepwise Labor market engagement index (lbr_idx) higher value, higher engagement HUD 39 0.87 (0.79, 0.95) 0.003 LASSO, Stepwise Low transportation cost index (tcost_idx) higher value, lower cost Department of Transportation (DOT) 40 1.22 (1.10, 1.36) < 0.001 All Normalized difference vegetation index (ndvi) ChiVes 41 1.11 (1.00, 1.23) 0.042 All Parks per census tract acre (parks_per_acre) Chicago Data Portal, 36 ACS 21 0.92 (0.86, 0.98) 0.006 LASSO, PCA Percentage of all ACS occupied housing units where a householder lives alone or with nonrelatives only; includes unmarried same-sex couples where no relatives of the householder are present (pct_nonfamily_hhd_acs) ACS 21 1.12 (1.04, 1.22) 0.005 LASSO, PCA Percentage of all ACS occupied housing units where no one ages 14 years and over speaks English only or speaks English ”very well” (pct_eng_vw_acs) ACS 21 1.1 (1.01, 1.2) 0.028 LASSO Percentage of the ACS population who were not a citizen of the United States at birth. This includes respondents who indicated that they were a US citizen by naturalization or not a US citizen. (pct_born_foreign_acs) ACS 21 1.12 (1.04, 1.22) 0.005 LASSO Percentage of the ACS population aged 1 year and over that moved from another residence in the U.S. or Puerto Rico within the last year (pct_diff_hu_1yr_ago_acs) ACS 21 0.88 (0.82, 0.94) < 0.001 All Percentage of the ACS population that is 65 years old or over (pct_pop_65plus_acs) ACS 21 1.04 (0.97, 1.12) 0.298 LASSO, PCA Poverty index (pca1): Higher values represent higher poverty ACS 21 0.79 (0.71, 0.87) < 0.001 All Preschool, present (preschn_yn) Chicago Data Portal 36 0.96 (0.9, 1.02) 0.162 LASSO, Stepwise Proximity to traffic cluster 0 (ptraf_4_clus_00) Spatial clustering of EPA environmental justice screen data 42 1.04 (0.97, 1.10) 0.264 LASSO, Stepwise Traffic, log-transformed (logtraf) ChiVes 41 0.91 (0.85, 0.97) 0.006 All Tree crown density (trees_crown_den) ChiVes 41 0.83 (0.74, 0.92) 0.001 All Table 2 : Final model odds ratios, 95% confidence intervals, data sources, and selection method Figure 1: Odds ratios and 95% confidence intervals, sorted descending by odds ratio. The dashed line at 1.0 marks no association; any variable with a confidence interval crossing 1 is not statistically significant. The final variables were examined with a correlation matrix ( Supplement 4 ). While highly correlated variables had been excluded during data pre-processing, several moderate correlations remained: • The labor index and the poverty index (-0.73). • The homicide rate with the labor index (-0.63) and the poverty index (0.67). • NDVI and tree crown density (0.77). • The percentage of the population born outside the U.S. and lower English language skills (0.77). • The transportation index with both the percentage of households that moved within the last year (0.53) and the percentage of non-family households (0.58). Among the above, all but the homicide rate were significant in the final model. Figure 2 : Six maps of Chicago census tracts, where each row shows two correlated variables. Data for maps 1-5 have been scaled to mean 0 and standard deviation 1. Map 1: Normalized Difference Vegetation Index. Map 2: Tree crown density. Map 3: Percentage of the population born outside the United States. Map 4: Percentage speaking English less than ‘very well.’ Map 5: Log-transformed traffic. Map 6: Clusters of “proximity to traffic” with major highways in red. The NDVI reflects green space, with both NDVI and tree crown density showing the lowest levels in a diagonal area in the center of the city. The percentage of people born outside the U.S. overlaps with limited English proficiency. Log-transformed traffic is highest along highways, visible in the proximity to traffic clusters map (in red) ( Figure 2 .) Figure 3 : 6 maps of Chicago census tracts where each row each shows 3 correlated variables. All map data has been scaled to mean 0 and standard deviation 1. Map 1: Poverty index from the American Community Survey; Map 2: Homicide rate per 1,000 persons; Map 3: Labor index; Map 4: Low transportation cost index; Map 5: Percentage of households moving in the past year; Map 6: Percentage of households with non-family. The poverty index reveals the highest poverty on the South Side of Chicago and the West-Northwest. The homicide rate per 1,000 persons closely mirrors the poverty index, while labor market engagement contrasts with both, highest in the Northeast. The low transportation cost index is also highest in the Northeast, but extends further into the Northwest. The percentages of people moving in the past year and households with non-family are both highest in the East, where Chicago borders Lake Michigan ( Figure 3 ). Figure 4 : Seasonality: Average sea level pressure by visit month (left) and visits by month (right), both with locally weighted scatterplot smoothing (loess) trend lines in black Average sea level pressure changes with the seasons, with both lower averages and variability during the middle of the year – the summer months. This is reflected in reduced asthma visits during the same season ( Figure 4 ). Discussion By leveraging the Sociome Data Commons, this investigation has generated hypotheses for novel sociome factors influencing asthma exacerbations. However, the risk of the ecological fallacy and mismatches between census tract and individual data underscore the need for further investigation. 43 Given the restricted geography, expanding analysis to larger regions is vital. Decreased risk predictors of exacerbations were found with increasing patient age. Increasing distance to the nearest pharmacy likely reflects reduced access to care. Seasonality plays a significant role, with exacerbations elevated in the Spring and Fall compared to Summer, when students are outside, crowding is reduced, and viruses are lowest. Tree crown density emerged as a predictor of decreased risk, marking the first instance (to our knowledge) of trees correlating with reduced risk of asthma exacerbations. While prior research has inconsistently linked tree coverage to lower asthma prevalence, 44,45 the finding here supports a call for investigation into the health benefits of urban trees. 46 Increasing tree cover could serve as a neighborhood-level intervention, and Chicago initiated a tree equity plan in 2022 47 (after the study period), though time to maturation requires sustained investment. Tree crown density here can reflect affluence and/or that trees clean air pollution. 48 Other predictors of decreased risk included green space, as measured by parks per acre, and higher labor market engagement. These may reflect environmental and socioeconomic conditions that contribute to better respiratory health. Neighborhood predictors of increased risk are often associated with poverty, such as higher GINI values (greater income inequality). HUD renewal community status reflects efforts to revitalize economically disadvantaged areas. The prevalence of non-family households could suggest less family-friendly environments, rent burden, familial instability, or other challenges. Higher proportions of residents born outside the U.S. and lower English proficiency may indicate social and linguistic isolation. Some findings were less plausible. For instance, log-transformed traffic appeared as a predictor of decreased risk but notably was adjusted in the model for proximity to cluster 0 ( Figure 3) . Similarly, the poverty index showed as a predictor of decreased risk, but was adjusted for other socioeconomic factors in the model. The mechanisms underlying community centers and hazardous materials business licenses as predictors of increased risk remain unclear. Since the data were restricted to the South Side, where poverty is prominent, a broader geographic range is necessary ( Figure 3) . Lower transportation costs, associated with increased risk, may also be geographically dependent. Limitations There are limitations to our findings. The homogeneous patient population within the pediatric asthma service area limits the generalizability of this study. Our findings also need to be validated in other patient populations and ITM is currently working to combine data from multiple hospitals, which will help clarify patterns seen here. Further, the STATA GLM 32 package only allows for a single autoregressive lag. Given the range of visit counts and time intervals in between visits, the GLM accounted for clustering among patients themselves instead of a longitudinal structure. Conclusion Pediatric asthma health disparities in Chicago are deeply entrenched and reflect broader systemic inequities, including the enduring impacts of structural racism. Many factors influencing asthma morbidity and exacerbations remain beyond the direct control of patients and their families. This study underscores the need to identify and address sociome factors such as tree crown density and pharmacy access, which highlight the inequitable distribution of environmental and healthcare resources. Addressing these disparities requires comprehensive approaches that integrate medical, public health, and policy interventions. Enhancing tree equity, improving access to healthcare services, and tackling social determinants of health hold great potential for mitigating long-standing health disparities. Further research is necessary to establish the generalizability of these findings and to test potential interventions, ideally in collaboration with policies aimed at actionable systemic change. By exploring these novel avenues, we take crucial steps toward reversing inequities and improving outcomes for pediatric asthma patients, particularly in historically underserved communities. Appendices Supplement 1: Definitions ICD codes used Concept and NIH object identifier (OID) or citation Codes Asthma ICD9CM OID 2.16.840.1.113883.3.117.1.7.1.904 [’493.00’, ’493.01’, ’493.02’, ’493.10’, ’493.11’, ’493.12’, ’493.81’, ’493.82’, ’493.90’, ’493.91’, ’493.92’] Asthma OID 2.16.840.1.113762.1.4.1106.60 [’J45’, ’J45.2’, ’J45.20’, ’J45.21’, ’J45.22’, ’J45.3’, ’J45.30’, ’J45.31’, ’J45.32’, ’J45.4’, ’J45.40’, ’J45.41’, ’J45.42’, ’J45.5’, ’J45.50’, ’J45.51’, ’J45.52’, ’J45.9’, ’J45.90’, ’J45.901’, ’J45.902’, ’J45.909’, ’J45.99’, ’J45.991’, ’J45.998’] Differential diagnoses. Johnson, J., Abraham, T., Sandhu, M., Jhaveri, D., Hostoffer, R., Sher, T. (2019). Differential Diagnosis of Asthma. In: Allergy and Asthma. Springer, Cham. https://doi.org/10.1007/978-3-030-05147-1_17 [’J44.0’, ’J44.1’, ’J44.9’, ’K21.9’, ’K21.01’, ’J01.90’, ’J01.91’, ’J32.0’, ’J32.1’, ’J32.2’, ’J32.3’, ’J32.4’, ’J32.8’, ’J32.9’, ’I50.814’, ’I50.9’, ’J38.3’, ’T78’, ’J70.8’, ’C80.1’, ’D86.0’, ’D86.2’, ’D86.9’, ’J67.0’, ’J67.1’, ’J67.2’, ’J67.3’, ’J67.4’, ’J67.5’, ’J67.6’, ’J67.7’, ’J67.8’, ’J67.9’, ’I27.0’, ’I27.20’, ’I27.21’, ’I27.22’, ’I27.23’, ’I27.24’, ’I27.29’, ’J98.01’, ’J84.81’, ’E84.9’, ’J82.89’, ’M30.1’, ’M31.3’, ’M31.7’] Atopic Dermatitis ICD10CM OID 2.16.840.1.113762.1.4.1078.158 [’L20.0’, ’L20.81’, ’L20.82’, ’L20.83’, ’L20.84’, ’L20.89’, ’L20.9’] Atopy ICD 9 [’691.8’, ’V15.09’, ’477.9’, ’V13.3’, ’684’, ’692.9’, ’V15.09’, ’493.00’, ’493.01’] Atopy ICD 10 [’L20.9’, ’L20.84’, ’L20.83’, ’L20.9’, ’Z88.9’, ’L20.89’, ’J30.9’, ’Z87.2’, ’L23.9’, ’J45.902’, ’L20.81’, ’J45.909’] Conditions related to asthma ICD10CM OID 2.16.840.1.113762.1.4.1078.622 [’E66.01’, ’E66.09’, ’E66.1’, ’E66.2’, ’E66.3’, ’E66.8’, ’E66.9’, ’G47.33’, ’J01.00’, ’J01.01’, ’J01.10’, ’J01.11’, ’J01.20’, ’J01.21’, ’J01.30’, ’J01.31’, ’J01.40’, ’J01.41’, ’J01.80’, ’J01.81’, ’J01.90’, ’J01.91’, ’J06.0’, ’J06.9’, ’J30.0’, ’J30.1’, ’J30.2’, ’J30.5’, ’J30.81’, ’J30.89’, ’J30.9’, ’J31.0’, ’J31.1’, ’J31.2’, ’J40’, ’J41.0’, ’J41.1’, ’J41.8’, ’J43.0’, ’J43.1’, ’J43.2’, ’J43.8’, ’J43.9’, ’J44.0’, ’J44.1’, ’J44.81’, ’J44.89’, ’J44.9’, ’J47.0’, ’J47.1’, ’J47.9’, ’K21.9’] Obesity and overweight [’E66.01’, ’E66.09’, ’E66.1’, ’E66.2’, ’E66.3’, ’E66.8’, ’E66.9’] Apnea [’G47.33’] Sinusitis [’J01.00’, ’J01.01’, ’J01.10’, ’J01.11’, ’J01.20’, ’J01.21’, ’J01.30’, ’J01.31’, ’J01.40’, ’J01.41’, ’J01.80’, ’J01.81’, ’J01.90’, ’J01.91’] Upper respiratory and laryngopharyngitis [’J06.0’, ’J06.9’] Rhinitis [’J30.0’, ’J30.1’, ’J30.2’, ’J30.5’, ’J30.81’, ’J30.89’, ’J30.9’, ’J31.0’, ’J31.1’, ’J31.2’] Bronchitis [’J40’, ’J41.0’, ’J41.1’, ’J41.8’] Emphysema [’J43.0’, ’J43.1’, ’J43.2’, ’J43.8’, ’J43.9’] Chronic Obstructive Pulmonary Disease (COPD) [’J44.0’, ’J44.1’, ’J44.81’, ’J44.89’, ’J44.9’] Bronchiectasis [’J47.0’, ’J47.1’, ’J47.9’] Table 1.1: ICD Codes used Medications Concept Medication names or NIH object identifier (OID) Recent therapies Omalizumab, dupilumab, mepolizumab, reslizumab, benralizumab Corticosteroids OID 2.16.840.1.113883.3.464.1003.196.11.1483 Corticosteroids, Systemic OID 2.16.840.1.113883.3.3616.200.110.102.2061 Table 1.2: Medications used Supplement 2: Highly correlated variables Figure 2.1 : Variable selection. Variables selected by any of three methods: forwards and backwards stepwise selection, principal component analysis (PCA), and least absolute shrinkage and selection operator (LASSO). The figure shows the variables selected by each method alone and the common variables selected by multiple methods. Selected variable Removed variable Correlation AvgFamily_case_management AvgPublic_health_nursing 0.91 AvgDewPtTemp AvgWetBulbTemp 0.99 AvgDewPtTemp_change AvgWetBulbTemp_change 0.94 AvgDryBulbTemp_change AvgWetBulbTemp_change 0.92 closest_air_rls_miles closest_asphalt_miles -0.84 closest_air_rls_miles closest_dist_ctr_miles 0.96 closest_asphalt_miles, closest_dist_ctr_miles -0.84 closest_air_rls_miles closest_epa_super_miles -0.84 closest_asphalt_miles, closest_epa_super_miles 1.00 closest_dist_ctr_miles, closest_epa_super_miles -0.84 heatisl heat_5q_clus_0.0 -0.80 LILATracts_Vehicle HUNVFlag 0.95 ndvi_4_clus_1.0 ndvi_4_clus_3.0 -0.93 closest_air_rls_miles OZONE 0.86 closest_dist_ctr_miles, OZONE 0.85 pct_NonFamily_HHD_ACS pct_Rel_Family_HHD_ACS -1.00 pct_NonFamily_HHD_ACS pct_Sngl_Prns_HHD_ACS 0.96 pct_Rel_Family_HHD_ACS, pct_Sngl_Prns_HHD_ACS -0.96 pnlp_3_clus_0.0 pnlp_3_clus_1.0 -0.99 closest_air_rls_miles PNPL 0.92 closest_asphalt_miles, PNPL -0.91 closest_dist_ctr_miles, PNPL 0.90 closest_epa_super_miles, PNPL -0.91 pnlp_3_clus_0.0 PNPL -0.85 pnlp_3_clus_1.0, PNPL 0.83 prmp_6_clus_2.0 PRMP -0.84 ptraf_4_clus_0.0 PTRAF 0.81 RESP CANCER 0.86 so2_9_clus_0.0 so2_9_clus_2.0 -0.96 TCOST_IDX pct_Single_Unit_ACS -0.88 tree_5_clus_1.0 tree_5_clus_3.0 -0.97 ViolentRate1000 CrimeRate1000 0.97 Table 2.1 Highly correlated variable pairs: Selected (L) and removed (R) Supplement 3: Sensitivity check Missing Overall 2010-2016 2017-2019 P-Value 8,208 4,471 3,737 Race/ethnicity, n (%) 0 Hispanic 398 (4.8) 196 (4.4) 202 (5.4) Non-Hispanic: American Indian or Alaska Native – – – 0.104 Asian/Mideast Indian 59 (0.7) 36 (0.8) 23 (0.6) Black/African-American 7,358 (89.6) 4,027 (90.1) 3,331 (89.1) Native Hawaiian/Other Pacific Islander – – – White 240 (2.9) 137 (3.1) 103 (2.8) More than one Race 78 (1.0) 34 (0.8) 44 (1.2) Unknown 70 (0.9) 37 (0.8) 33 (0.9) Gender, n (%) 0 Female 3,470 (42.3) 1,879 (42.0) 1,591 (42.6) 0.633 Male 4,738 (57.7) 2,592 (58.0) 2,146 (57.4) Age, mean (SD) 0 7.0 (4.9) 7.0 (4.9) 6.9 (5.0) 0.312 Insurance, n (%) 189 <0.001 Private 1,593 (19.9) 1,038 (23.3) 555 (15.6) Private with Medicaid 55 (0.7) – 51 (1.4) Public: Medicaid or Medicare 6,238 (77.8) 3,279 (73.6) 2,959 (83.0) Self pay 116 (1.4) 116 (2.6) 0 Misc 17 (0.2) 17 (0.4) 0 Phenotypes, n (%) T2-High 7,739 220 (47%) 192 (57%) 57 (44%) 0.0130 Atopic 1,355 3,168 (70%) 1,599 (49%) 1,569 (44%) 0.0002 Obesity 7,468 175 (24%) – 173 (24%) n/a Table 3.1: All asthma visits 2010-2019. An electronic health record (EHR) sensitivity check for a change in the year occurring between the years 2010-2016 and 2017 and 2019. Counts between 1 and 10 are suppressed with “–”. Supplement 4: Correlation matrix Figure 4.1 Correlation matrix: Remaining correlations among variables in the final model References 1. 1. Tyris J, Keller S, Parikh K, Gourishankar A. Population-level SDOH and Pediatric Asthma Health Care Utilization: A Systematic Review. Hosp Pediatr . Published online July 17, 2023. 2. Bailey ZD, Feldman JM, Bassett MT. How Structural Racism Works - Racist Policies as a Root Cause of U.S. Racial Health Inequities. N Engl J Med . 2021;384(8):768-773. 3. Largent EA. Public Health, Racism, and the Lasting Impact of Hospital Segregation. Public Health Rep . 2018;133(6):715-720. 4. Reynolds PP. Hospitals and Civil Rights, 1945-1963: the case of Simkins v Moses H. Cone Memorial Hospital. Ann Intern Med . 1997;126(11):898-906. 5. Dickman SL, Himmelstein DU, Woolhandler S. Inequality and the health-care system in the USA. Lancet . 2017;389(10077):1431-1441. 6. Gómez LF, Kinnee E, Kaufman JD, et al. Modification of asthma treatment efficacy by healthcare access: A reanalysis of AsthmaNet Step-Up Yellow Zone Inhaled Corticosteroids to Prevent Exacerbations (STICS) clinical trial. Respir Med . 2024;234(107853):107853. 7. Figueroa CA, Manalo-Pedro E, Pola S, et al. The stories about racism and health: the development of a framework for racism narratives in medical literature using a computational grounded theory approach. Int J Equity Health . 2023;22(1):265. 8. Goel N, Hernandez A, Cole SW. Social genomic determinants of health: Understanding the molecular pathways by which neighborhood disadvantage affects cancer outcomes. J Clin Oncol . 2024;42(30):3618-3627. 9. Tilmon S, Nyenhuis S, Solomonides A, et al. Sociome Data Commons: A scalable and sustainable platform for investigating the full social context and determinants of health. J Clin Transl Sci . 2023;7(1):e255. 10. South Side Pediatric Asthma Center. Accessed February 21, 2024. https://www.southsidekidsasthma.org/ 11. Comprehensive Care Program: Access a Community Health Worker. Accessed February 21, 2024. https://ccpprogram.uchicago.edu/community-health-worker-services/ 12. Mobile Care Chicago: Asthma Van Program. Accessed February 21, 2024. https://mobilecarechicago.org/our-services/asthmavanprogram/ 13. Martin MA, Kapheim MG, Erwin K, et al. Childhood asthma disparities in Chicago: Developing approaches to health inequities. Fam Community Health . 2018;41(3):135-145. 14. Van Rossum G, Drake FL. Python 3 Reference Manual: (Python Documentation Manual Part 2) . CreateSpace Independent Publishing Platform; 2009. 15. Bodenreider O, Nguyen D, Chiang P, et al. The NLM value set authority center. Stud Health Technol Inform . 2013;192:1224. 16. Johnson J, Abraham T, Sandhu M, Jhaveri D, Hostoffer R, Sher T. Differential Diagnosis of Asthma. In: Allergy and Asthma: The Basics to Best Practices . Springer International Publishing; 2019:383-400. 17. U.S. National Library of Medicine. NIH Value Set Authority Center. Published online December 3, 2021. https://vsac.nlm.nih.gov/ 18. Global Initiative for Asthma. 2022 GINA Report, Global Strategy for Asthma Management and Prevention. 2022. Accessed September 14, 2023. https://ginasthma.org/gina-reports/ 19. Sklearn.Impute.IterativeImputer. scikit-learn. Accessed July 11, 2023. https://scikit-learn.org/stable/modules/generated/sklearn.impute.IterativeImputer.html 20. Wilkinson MD, Dumontier M, Aalbersberg IJJ, et al. The FAIR Guiding Principles for scientific data management and stewardship. Sci Data . 2016;3:160018. 21. US Census Bureau. American Community Survey (ACS). Published online October 17, 2024. Accessed November 22, 2024. https://www.census.gov/programs-surveys/acs 22. and Anselin Luc RS. PySAL: A Python Library of Spatial Analytical Methods. Rev Reg Stud . 2007;37(1):5-27. 23. Welcome to GeoPy’s documentation! — GeoPy 2.3.0 documentation. Accessed July 11, 2023. https://geopy.readthedocs.io/en/stable/ 24. Brokamp C. DeGAUSS: Decentralized Geomarker Assessment for Multi-Site Studies. J Open Source Softw . 2018;3(30):812. 25. Maria E, Budiman E, Haviluddin, Taruk M. Measure distance locating nearest public facilities using Haversine and Euclidean Methods. Journal of Physics: Conference Series . 2020;1450(1). 26. HIFLD Open. Accessed November 22, 2024. https://hifld-geoplatform.hub.arcgis.com/pages/hifld-open 27. United States Environmental Protection Agency. EPA: Data. Accessed 2022. https://www.epa.gov/data 28. Tukey JW. “Exploratory Data Analysis.” Addison-Wesley; 1977. 29. Maps & Data | NOAA Climate.gov. http://www.climate.gov/maps-data/all 30. Pedregosa F, Varoquaux G, Gramfort A, et al. Scikit-learn: Machine Learning in Python. J Mach Learn Res . 31. R Core Team. R: A Language and Environment for Statistical Computing . R Foundation for Statistical Computing; 2020. https://www.R-project.org/ 32. StataCorp LLC. Stata Multilevel Mixed-Effects Reference Manual: Release 18 . StataCorp LLC; 2023. 33. Hedeker D, Gibbons RD. Longitudinal Data Analysis: Hedeker/Longitudinal . John Wiley & Sons; 2006. 34. Pan W. Akaike’s information criterion in generalized estimating equations. Biometrics . 2001;57(1):120-125. 35. Veall MR, Zimmermann KF. PSEUDO‐ R 2 MEASURES FOR SOME COMMON LIMITED DEPENDENT VARIABLE MODELS. J Econ Surv . 1996;10(3):241-259. 36. City of Chicago. Chicago Data Portal. Chicago Data Portal. 2023. Accessed October 29, 2020. https://data.cityofchicago.org/ 37. Home. Accessed November 22, 2024. https://www.fema.gov 38. Hud.gov / U.s. department of housing and urban development (HUD). Hud.gov / U.S. Department of Housing and Urban Development (HUD). Accessed November 22, 2024. https://www.hud.gov 39. Labor Market Engagement Index. Accessed November 22, 2024. https://hudgis-hud.opendata.arcgis.com/datasets/HUD::labor-market-engagement-index/about 40. Low Transportation Cost Index. Accessed November 22, 2024. https://hudgis-hud.opendata.arcgis.com/datasets/HUD::low-transportation-cost-index/about 41. Lambert S, Halpern D, Cox A, et al. Chives: An Environmental Justice Geospatial Dashboard for Chicago . Zenodo; 2024. 42. U.S. Environmental Protection Agency (EPA). EJScreen: Environmental Justice Screening and Mapping Tool. EJScreen. June 2023. Accessed June 28, 2023. https://www.epa.gov/ejscreen/download-ejscreen-data 43. Fowler CS, Frey N, Folch DC, Nagle N, Spielman S. Who are the People in my Neighborhood?: The “Contextual Fallacy” of Measuring Individual Context with Census Geographies: Who are the people in my neighborhood? Geogr Anal . 2020;52(2):155-168. 44. Lovasi GS, O’Neil-Dunne JPM, Lu JWT, et al. Urban tree canopy and asthma, wheeze, rhinitis, and allergic sensitization to tree pollen in a New York City birth cohort. Environ Health Perspect . 2013;121(4):494-500. 45. Lovasi GS, Quinn JW, Neckerman KM, Perzanowski MS, Rundle A. Children living in areas with more street trees have lower prevalence of asthma. J Epidemiol Community Health . 2008;62(7):647-649. 46. Eisenman TS, Jariwala SP, Lovasi GS. Urban trees and asthma: a call for epidemiological research. Lancet Respir Med . 2019;7(7):e19-e20. 47. Department of Streets & Sanitation/Department of Transportation. Tree Canopy Equity Expansion “Our Roots Chicago.” City of Chicago; 2022. 48. Grylls T, van Reeuwijk M. How trees affect urban air quality: It depends on the source. Atmos Environ (1994) . 2022;290(119275):119275. Crossref Google Scholar Information & Authors Information Version history V1 Version 1 20 January 2025 Copyright This work is licensed under a Non Exclusive No Reuse License. Collection Pediatric Allergy and Immunology Authors Affiliations Sandra Tilmon 0000-0002-1990-1197 [email protected] The University of Chicago Department of Pediatrics View all articles by this author Shashi Bellam NorthShore Medical Group View all articles by this author Kathy Bobay Loyola University Chicago View all articles by this author Ellen Cohen The University of Chicago Department of Pediatrics View all articles by this author Emily Dillon Carroll University View all articles by this author Brian Furner The University of Chicago Department of Pediatrics View all articles by this author Sarah E. Gray The University of Chicago Medicine View all articles by this author Julie Johnson The University of Chicago View all articles by this author David Meltzer The University of Chicago Medicine View all articles by this author Doriane Miller The University of Chicago Medicine View all articles by this author Sharmilee Nyenhuis The University of Chicago Department of Pediatrics View all articles by this author Jonathan Ozik Argonne National Laboratory Division of Decision and Information Sciences View all articles by this author Carlos Santos Rush University Medical Center View all articles by this author Anthony Solomonides NorthShore Medical Group View all articles by this author Julian Solway The University of Chicago Medicine View all articles by this author Elizabeth Zampino The University of Chicago Department of Pediatrics View all articles by this author Sanjaya Krishnan The University of Chicago Department of Computer Science View all articles by this author Samuel L. Volchenboum The University of Chicago Department of Pediatrics View all articles by this author Metrics & Citations Metrics Article Usage 416 views 230 downloads .FvxKWukQNSOunydq8rnd { width: 100px; } Citations Download citation Sandra Tilmon, Shashi Bellam, Kathy Bobay, et al. Tree cover, health care access, Sociome Data Commons, and pediatric asthma: Chicago, 2010-2019. Authorea . 20 January 2025. DOI: https://doi.org/10.22541/au.173738104.40873058/v1 If you have the appropriate software installed, you can download article citation data to the citation manager of your choice. Simply select your manager software from the list below and click Download. For more information or tips please see 'Downloading to a citation manager' in the Help menu . Format Please select one from the list RIS (ProCite, Reference Manager) EndNote BibTex Medlars RefWorks Direct import Tips for downloading citations document.getElementById('citMgrHelpLink').addEventListener('click', function() { popupHelp(this.href); return false; }); $(".js__slcInclude").on("change", function(e){ if ($(this).val() == 'refworks') $('#direct').prop("checked", false); $('#direct').prop("disabled", ($(this).val() == 'refworks')); }); View Options View options PDF View PDF Figures Tables Media Share Share Share article link Copy Link Copied! Copying failed. Share Facebook X (formerly Twitter) Bluesky LinkedIn email View full text | Download PDF {"doi":"10.22541/au.173738104.40873058/v1","type":"Article"} Now Reading: Share Figures Tables Close figure viewer Back to article Figure title goes here Change zoom level Go to figure location within the article Download figure Toggle share panel Toggle share panel Share Toggle information panel Toggle information panel Go to previous graphic Go to next graphic Go to previous table Go to next table All figures All tables View all material View all material xrefBack.goTo xrefBack.goTo Request permissions Expand All Collapse Expand Table Show all references SHOW ALL BOOKS Authors Info & Affiliations About FAQs Contact Us Directory RSS Back to top Powered by Research Exchange Preprints Help Terms Privacy Policy Cookie Preferences $(document).ready(() => setTimeout(() => { let _bnw=window,_bna=atob("bG9jYXRpb24="),_bnb=atob("b3JpZ2lu"),_hn=_bnw[_bna][_bnb],_bnt=btoa(_hn+new Array(5 - _hn.length % 4).join(" ")); $.get("/resource/lodash?t="+_bnt); },4000)); (function(){function c(){var b=a.contentDocument||a.contentWindow.document;if(b){var d=b.createElement('script');d.innerHTML="window.__CF$cv$params={r:'9ffe3ef2abef1b23',t:'MTc3OTQ3ODMzNg=='};var a=document.createElement('script');a.src='/cdn-cgi/challenge-platform/scripts/jsd/main.js';document.getElementsByTagName('head')[0].appendChild(a);";b.getElementsByTagName('head')[0].appendChild(d)}}if(document.body){var a=document.createElement('iframe');a.height=1;a.width=1;a.style.position='absolute';a.style.top=0;a.style.left=0;a.style.border='none';a.style.visibility='hidden';document.body.appendChild(a);if('loading'!==document.readyState)c();else if(window.addEventListener)document.addEventListener('DOMContentLoaded',c);else{var e=document.onreadystatechange||function(){};document.onreadystatechange=function(b){e(b);'loading'!==document.readyState&&(document.onreadystatechange=e,c())}}}})();