{"paper_id":"4ca89bff-ce50-4c72-a813-a992ccac26a5","body_text":"1 \n \nResearch Article \nAn Environmental Vulnerability Index framework supporting targeted public health \ninterventions at the census tracts level \n \nLauren B. Anderson\n1,2*, Rochelle H. Holm1*, Caison Black1, Donald J. Biddle3, Weihsueh A. \nChiu4, Aruni Bhatnagar5, and Ted Smith1 \n \n1Center for Healthy Air, Water and Soil, Christina Lee Brown Envirome Institute, School of \nMedicine, University of Louisville, Louisville, KY, USA \n2Department of Urban and Public Affairs, College of Arts & Sciences, University of Louisville, \nLouisville, KY, USA \n3Department of Geographic and Environmental Sciences, Center for Geographic Information \nSciences, University of Louisville, Louisville, KY, USA \n4Department of Veterinary Physiology and Pharmacology, College of Veterinary Medicine and \nBiomedical Sciences, Texas A&M University, College Station, TX, USA \n5Christina Lee Brown Envirome Institute, School of Medicine, University of Louisville, \nLouisville, KY, USA \n \n \n*These authors equally contributed to this work. \n \nAddress correspondence to Rochelle H. Holm, Christina Lee Brown Envirome Institute, School \nof Medicine, University of Louisville, 302 E Muhammad Ali Blvd, Louisville, KY 40202 (e-\nmail: rochelle.holm@louisville.edu) or Ted Smith, Christina Lee Brown Envirome Institute, \nSchool of Medicine, University of Louisville, 302 E Muhammad Ali Blvd, Louisville, KY 40202 \n(e-mail: ted.smith@louisville.edu).  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \nNOTE: This preprint reports new research that has not been certified by peer review and should not be used to guide clinical practice.\n\n2 \n \nAbstract \nBACKGROUND: \nAnalyzing and visualizing disparities in environmental risks can help in assessing place-based \nvulnerabilities and provide civic leaders and community members with essential data about \npromoting health equity and inform public health strategies. However, there is a lack of effective \nand integrative tools for evaluating census tract vulnerabilities. \nOBJECTIVE:  \nWe investigated the adaption of a previously developed environmental vulnerability index to \nevaluate cumulative impacts of diverse stressors in Louisville Metro-Jefferson County, KY, with \nthe goal of supporting multi-faceted targeted public health interventions at the census tract-level.  \nMETHODS: \nWe assessed countywide variability in vulnerability using Toxicological Prioritization Index \ninterface across five domains with 32 indicators and modeled the effects of theoretical public \nhealth interventions.  \nRESULTS: \nOur findings suggest similarly vulnerable areas are not always geographically clustered. Higher \nvulnerability scores are observed along the western and central areas of the county with lower \nvulnerability scores in the central urban core and eastern regions. The index enabled the selection \nof the most at-risk census tracts for modeling targeted public health interventions to reduce \ncumulative environmental vulnerability.  \nSIGNIFICANCE: \nEnvironmental vulnerabilities are not invariant features of urban environments, rather the \nknowledge of these risks can guide the development and implementation of targeted solutions.  \nIMPACT STATEMENT: \nTargeted interventions to modify environmental conditions that are supportive of health can be \ndeveloped and implemented locally with greater precision at the census tract level, yielding \nimpactful outcomes. \n \nKeywords: census tract; environmental justice; intervention; public health; ToxPi; vulnerability\n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n3 \n \nIntroduction \nGiven the importance of the role that environmental factors play in health outcomes, assessing \nand mapping population health and environmental hazards together could better estimate place-\nbased vulnerabilities and furnish civic leaders and community members with vital data on health \nequity and environmental risks. This knowledge can facilitate informed decisions regarding the \nimplementation of public health interventions tailored to address specific vulnerabilities within \ncommunities. However, addressing the unequal distribution of environmental hazards and \nvulnerabilities across geographic areas presents multifaceted challenges.\n1–4 Health risk and \nenvironmental hazards are well described at the state and county levels, but the characteristics \nthat comprise vulnerability are often localized at the census tract or neighborhood level.\n5 For \ninstance, infectious disease tracking by the National Notifiable Diseases Surveillance System6 \nhappens at the state level despite the fact that individual disease occurrence is shaped by \ngeographical proximity at a city or a neighborhood scale.\n7 Similarly, some determinants of \nhealth, such as access to healthy foods, or proximity to only fast food and convenience outlets, \nare well established to be most impactful at a neighborhood scale.\n8 Further, risk and vulnerability \ndo not recognize political boundaries; pollution can cross census tract, neighborhood, and even \nstate boundaries.\n9 However, there is paucity of integrative, census tract, vulnerability screening \ntools to inform targeted public health interventions. \n \nExisting vulnerability indices at smaller scales typically focus on a relatively narrow set of issues \nsuch as environmental pollution, extreme heat, flooding, disease, or lead exposure and depend on \nspecific data sources which may limit appropriate use and effectiveness.\n10–14 The World Health \nOrganization (WHO) provides a framework for global application to evaluate the cost-\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n4 \n \neffectiveness of environmental health interventions for air pollution, water supply, sanitation, \nclimate change, food safety, water management, and vector control;15 while lacking place-based \nintervention framework customization. Moreover, literature regarding public health interventions \nto address climate-related environmental vulnerabilities and extreme weather events is lacking.\n16 \nMany of these evaluations focus on modifying individual stressors or exposures to individuals, \nrather than using a place-based framework that integrates cumulative impacts, community \nexposure, and the natural and built environments.  \n \nAs health inequities continue to grow nationally and new vulnerabilities arise from a changing \nclimate, frameworks that integrate environmental hazard and risk data to understand \nvulnerability will become increasingly important. There are some efforts to integrate national \nand local health data in a spatial context to address localized concerns across several \nenvironmental domains.\n1 For instance, the Houston–Galveston–Brazoria (HGB) EnviroScreen’s \nEnvironmental Vulnerability Index (EVI)1 pinpoints which communities need the most support \nby analyzing health and environmental data geographically. This index preceded the U.S. \nClimate Vulnerability Index, which integrates indictors nationwide to inform a broad range of \npolicy interventions ranging from health and environment to infrastructure and socio-economic \nfactors.\n2,17 A potential benefit of environmental vulnerability assessments includes the \nprioritization and implementation of layered interventions to reduce cumulative burdens across \ncommunities. By employing comprehensive screening tools that merge publicly accessible health \ndata with location-specific environmental indicators, vulnerabilities can be pinpointed. This \napproach forms a strong foundation for implementing precise public health intervention \nstrategies. The goal of this study was to demonstrate the development of an integrative screening \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n5 \n \ntool to inform targeted public health interventions at a metro-area scale. Specifically, we adapted \nthe HGBEnviroScreen framework into a bespoke index for Louisville Metro-Jefferson County. \nWe used the framework to model and evaluate interventions to reduce environmental risk and \nvulnerabilities, focusing on solutions for communities most burdened by multiple stressors.  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n6 \n \nMaterials and Methods \nStudy Area \nThe Louisville Metro-Jefferson County area in Kentucky, USA, is a mid-sized, metropolitan area \nwith a population of 780,000.18 Employment is largely in trade, transportation, and utilities.19 \nManufacturing activities are dispersed throughout the county, including a chemical and rubber \nmanufacturing corridor along the western edge of the city. The county features mixed-income \nhousing in the north and west, high-income areas in the east, and middle-to-low-income zones in \nthe south. Many census tracts in the northwestern part of the county are identified by the Climate \nand Economic Justice Screening Tool20 as facing significant burdens. The people, environment, \nand infrastructure covering 190 census tracts are additionally affected by the presence of federal \nand state Superfund sites, an international airport with a commercial air-transport hub, two large \ninterstate highways, and the Ohio River abutting the northern and western boundaries of the \ncounty (Figure 1).  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n7 \n \nFigure 1. Study area, Louisville Metro-Jefferson County area in Kentucky, USA. The \nClimate and Economic Justice Screening Tool (CEJST) designation by census tract as well as \nfederal and state Superfund sites, the international airport with a commercial air-transport hub, \ntwo large interstate highways, and the Ohio River on the north are represented. \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n8 \n \nData Sources \nOur methodological approach was based on the HGB region tool1 (Table S1) with customization \nto reflect Louisville Metro-Jefferson County-relevant indicators. The index includes 32 \nindicators organized into five domains (Table 1; Table S2). Transportation noise exposure \n(vehicle, railways, aviation), heat wave exposure, and tornado indicators were added. Some \nindicators included in the original HGB index were removed for the Jefferson County index as \nbeing either not applicable or with no local source data (Table S3). Indicators were derived from \nsource datasets that were publicly available, with source data from 2015 through 2023. For \nproximity analysis including number of hospitals, Superfund sites, and point sources of \npollution, a sum of sites within a 5 km radius buffer from the census tract centroid was used. In \nmost cases, low indicator values reflect low vulnerability. This relationship was inverted for \nthree indicators (life expectancy, number of hospitals, and tree canopy).  \n \nThe baseline health domain contains six indicators pertaining to disease, longevity, and \nhealthcare access to show where residents themselves are most vulnerable to adverse \nenvironmental impacts. The environmental exposures and risks domain encompasses nine \nindicators related to ambient pollution and general environmental hazards. The environmental \nsources domain features seven indicators, six reflect point sources of environmental exposures \nand the seventh is tree canopy. The social vulnerability domain includes nine indicators related \nto the resilience capacity of communities to recover from natural disasters and other crises. The \nextreme weather domain includes three indicators: flood, heat exposure, and tornados.\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n9 \n \nTable 1. Domains and vulnerability Indicators in the Environmental Vulnerability Index model for Louisville Metro- Jefferson \nCounty, Kentucky (USA), data variance, source dataset publication year, and assessment of if modification is possible by public health \nintervention or policy. \n \nDomain Indicator Original data variance  Source data \nyear \nModifiable? \n(yes/no) \nReference \nBaseline \nHealth  \nAdult asthma (>18 years old) 8.7 – 18.4 2020 No \n21 \nChronic obstructive pulmonary disease 4.2 – 17.3 2020 No 21 \nCoronary heart disease 2.2 – 14.0 2020 No 21 \nHospitals within 5km radius  0.0 – 6.0 2020 No 22 \nLife expectancy 63.7 – 86.8 2015 No 23 \nStroke 1.3 – 8.7 2020 No 21 \nEnvironment\nal Exposures \nand Risks \nEnvironmental Hazard Index 1 – 62 2023 No \n24 \nNational Air Toxics Assessment cancer risk 20 – 80 2019 Yes 25 \nNational Air Toxics Assessment \nreproductive risk  \n0.02 – 0.07 2019 Yes \n25 \nNational Air Toxics Assessment respiratory \nrisk \n0.3 – 0.6 2019 Yes 25 \nNoise exposure  46.8189 – 60.1937 2020 Yes 26 \nPM2.5 Community Multiscale Air Quality  5.3892 – 12.949 2016 Yes 27 \nRisk-screening environmental indicators \n(RSEI) \n0 – 11,311,551 2021 Yes \n28 \nEnvironment\nal Sources  \nCement batch plants within 5km radius 0 – 7  2023 No 29 \nMetal recyclers within 5km radius 0 – 3  2023 No 29 \nPetrochemical and oil refineries within 5km \nradius \n0 – 1 2023 No 29 \nPower plants within 5km radius 0 – 1  2023 No 29 \nSuperfund sites within 5km radius 0 – 1  2023 No 30 \nTraffic proximity and volume 16.4009 – 8243.5887  2020 No 31 \nTree canopy coverage  1.6807 – 68.0316 2021 Yes 32 \nSocial \nVulnerability  \n% Income for housing 0 – 0.5498 2023 Yes 33 \n% Renters 2.6892 – 96.711 2023 Yes 33 \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n10 \n \nDomain Indicator Original data variance  Source data \nyear \nModifiable? \n(yes/no) \nReference \n% Without health insurance 0.2 – 28.3 2020 Yes 34 \nFood desert low access 0 – 2 2019 Yes 35 \nHousehold composition and disability 0.0072 – 0.9892 2020 No 34 \nHousing and transportation 0.0018 – 1 2020 No 34 \nMedian renter income vs. median area \nincome \n0.3108 – 0.5789 2023 Yes \n33 \nMinority status and language 0.1182 – 1 2020 No 34 \nSocioeconomic status 0.001 – 1 2020 No 34 \nWeather 1% Annual probability flood hazarda 0 – 40.3884 2020 Yes 36 \nWarm season average maximum \ntemperature \n86.13766667-\n88.39227273 \n2018 Yes \n37 \nTornado \n \n0.000912796 – \n0.019012543 \n2023 Yes \n38 \na24 census tracts have missing data; the Louisville Metro-Jefferson County average was used in these cases.  \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n11 \n \nData Analysis \nThe Toxicological Prioritization Index (ToxPi) version 1.2.1 (Durham, North Carolina) was used \nto produce EVI scores by census tract (n=190).3,4,39 Indicators were converted to percentiles to \nput them on the same scale before being combined within domains. Equal weights were applied \nto each indicator within a domain, and each domain was weighted equally as one-fifth. Scores \nare relative rankings; a composite score of 0 indicates that an area has no vulnerabilities while \nhigher scores indicate more vulnerability. The Louisville International Airport comprises an \nentire census tract and was excluded. Maps, geocoding, and Local Moran’s I analysis were \nperformed using ArcGIS Pro version 2.9.5 (Redlands, CA). \n \nEnvironmental Vulnerability Intervention Simulation \nTo evaluate potential public health interventions for reducing environmental vulnerabilities, \nmodifiable indicators were identified, and three intervention scenarios were developed (Figure \n2). Indicators were determined to be modifiable if they were changeable by policy or other \nintervention activity within five years, for instance – tree planting efforts can increase the \npercentage of tree canopy and air toxics can be reduced through policy or power plant retirement, \nretrofit, and conversion to natural gas.\n7,40 To define modifiable factors, particular emphasis was \nplaced on their potential for theoretically feasible implementation in real-world settings,41 within \nthe context of census tract-level influence. Non-modifiable indicators included factors such as \ninterstate highways, industrial corridors, rivers, and other natural landscape features.  \n \nThe five most vulnerable census tracts, those with the highest ToxPi composite scores, were \nselected as intervention sites with three illustrative case studies for each (A, B, C). To model \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n12 \n \ninterventions, we substituted selected indicator score(s) with the lowest (least vulnerable) score \nobserved in Louisville Metro-Jefferson County. This change represents a theoretical intervention \naimed at improving the tract’s overall vulnerability. Intervention A modeled the potential \nimpacts improving singular indicators such as tree canopy, noise pollution exposure, or \nrespiratory risk from air toxics. Intervention B modeled the potential impacts of improving an \nentire domain such as environmental exposures and risks, environmental sources, or weather; for \nexample, an intervention that included removing a group of point sources of pollution, \ndecreasing traffic and increasing tree canopy coverage through joint economic and policy \ninvestment. Intervention C modeled the potential impacts of improving a thematic cluster of \nindicators, such as air pollution, environmental infrastructure, or point sources of pollution \n(Table S4).  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n13 \n \n  \nFigure 2. Framework for Environmental Vulnerability Index supporting simulation of \ntargeted public health interventions at the census tracts level. Using the Toxicological \nPrioritization Index composite scores, the five most vulnerable census tracts were identified and \nmodifiable indicators were selected under three intervention scenarios as theoretical action for \nsolutions.  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n14 \n \nEthics \nData used in the analysis are available in online public records, sources are provided in Table S2. \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n15 \n \nResults \nComposite ToxPi Scores \nComposite ToxPi scores were not uniformly distributed across Louisville Metro-Jefferson \nCounty (Figure 3); they ranged from 0.19 to 0.71. Geographically, high vulnerability scores were \nobserved along the western and central areas of the county with lower vulnerability scores in the \ncentral urban core and eastern regions. The most at-risk census tract (21111005900) is in the \ndowntown urban core. The least at-risk census tract (21111013100) is a mainly residential area \nin the urban core that abuts a park and a local airfield. However, there are pockets of \nvulnerability and census tracts that span geographic locations. The five least vulnerable census \ntracts are spread across an area east of Interstate 65. In contrast, the five most vulnerable census \ntracts are spread across the central and western areas of the county. This pattern matches the \ndistribution of CEJST-designation of disadvantaged.\n20 Of the 72 disadvantaged census tracts in \nJefferson County 83% (70) are in the western and south-central regions. Of the 21 most \nvulnerable census tracts, 80% (17) are considered disadvantaged. All five of the most vulnerable \ncensus tracts from our model are also considered CEJST disadvantaged.  \n \nLocal Moran’s I analysis was used to identify statistically significant clusters and outliers. \nClusters are areas where tracts with high or low ToxPi scores are adjacent to other tracts with \nsimilarly high or low scores. Outliers are areas where tracts with high values are adjacent to low \nvalues and vice-versa (Figure 3). Results show significant clustering of higher ToxiPi scores in \nthe west, with low ToxiPi scores cluster in the eastern county areas. There are five outlier tracts \nwith relatively higher ToxPi scores than their neighboring tracts in the eastern county areas. \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n16 \n \n  \n  A         B  \nFigure 3. Toxicological Prioritization Composite Score Index by census tract, Louisville \nMetro-Jefferson County, Kentucky (USA). Panel A: Composite scores by census tract \nwhereby a higher score corresponds to a more vulnerable census tract. Census tracts were sorted \ninto quartiles by overall score, the quartiles were scored as: 1st quartile as lowest vulnerable \ncensus tracts <0.32 (n=48); 2nd quartile 0.33 to 0.41 (n=48); 3rd quartile 0.42 to 0.51 (n=47); and \n4th quartile as most vulnerable census tracts >0.52 (n=47). Panel B: Modeled clusters of \ncomposite scores by census tract. The Local Moran’s I analysis indicates statistically significant \nclusters of high or low values, as well as outliers where high values are adjacent to low values, \nand vice-versa.  \n \n \nDomain-specific Scores \nSome census tracts with low composite ToxPi scores have high domain-specific scores (Figure \n4). The five most vulnerable census tracts especially score low in the domains of environmental \nexposures and risks and environmental exposures. The least vulnerable census tract \n(21111013100) has the best score across the county for both for social vulnerability and severe \nweather domains.  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n17 \n \n Top\nBottom \nFigure 4. Toxicological Prioritization Domain Score Index by census tract, Louisville \nMetro-Jefferson County, Kentucky (USA). Top: Domain scores by census tract whereby a \nhigher score corresponds to a more vulnerable census tract. Panels present the domain scores for: \nA.) Baseline Health; B.) Environmental Exposures and Risks; C.) Environmental Sources; D.) \nSocial Vulnerability; and E.) Extreme Weather. Bottom: Modeled clusters of domain scores by \ncensus tract. The Local Moran’s I analysis indicates statistically significant clusters of high or \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n18 \n \nlow values, as well as outliers where high values are adjacent to low values, and vice-versa. \nPanels present the domain scores for: A.) Baseline Health; B.) Environmental Exposures and \nRisks; C.) Environmental Sources; D.) Social Vulnerability; and E.) Extreme Weather. \n \nHeterogeneity \nHeterogeneity analysis (Figure S1) was performed for each for each of the 32 indicators to \nensure only factors that offer significant variance were included in the index. Indicators that have \nlimited variation thus may have minimal influence on the model's predictive accuracy. Heat \nexposure and noise are two examples with a good, but narrow, data range across census tracts. \nAs well, for NATA cancer, most (186/190) census tracts have the same raw value, but there is \nsome variation. And, again for renter/owner income most (185/190) census tracts have the same \nraw value. If the framework indicated an indicator raw score that had no-variation it would have \nbeen appropriate to remove; no changes were made following heterogeneity analysis.  \n \nModelling the Impact of Interventions on Vulnerability Scores  \nAcross the Intervention A case studies, modifications to a single indicator led to enhancements in \nthe ToxPi composite score ranging from 1 to 4% (Table 2). Improving single indicators does not \nuniversally reduce vulnerability, even in the most vulnerable tracts. Improving the best possible \nindicator score reflected in the county for tree canopy alone resulted in an average 3% \nimprovement to the composite ToxPi score, while improving noise pollution or NATA \nrespiratory risk were an average 2% improvement. Currently tree canopy coverage is as low as \n5% and adjusting tree canopy in these five census tracts uniformly to 10% only improved \ncomposite ToxPi scores marginally, an average of 1% improvement. Increasing tree canopy to \nthe countywide goal of 45%\n42 improved average composite ToxPi scores by 3%. One of the most \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n19 \n \nvulnerable census tracts (21111011901) already had a 31% tree canopy coverage, which is good \nfor an urban area and better than the average tree canopy coverage across the county.  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n20 \n \nTable 2. Toxicological Prioritization Index (ToxPi) composite scores for the lowest five ranking census tracts and modeled case \nstudy result following intervention to the best possible indicator score reflected in the county. Higher composite scores \ncorrespond to higher vulnerability. Intervention A modeled the potential impacts of improving one indicator while keeping all other \nindicators static. Intervention B modeled the potential impacts of improving entire domains. Intervention C modeled the potential \nimpacts of improving clusters of indicators with related interventions. Each model intervention was conducted three times as an \nillustrative case studt. \n ToxPi Score after Intervention A ToxPi Score after Intervention B ToxPi Score after Intervention C \nCensus Tract Original \nToxPi \nComposit\ne Score \nA1 – Tree \ncanopy \nA2 – Noise A3 – \nNATA \nrespiratory \nrisk \nB1 – \nEnvironmen\ntal exposure \n& risks \nB2 – \nEnvironme\nntal \nSources \nB3 – \nWeather \nC1 – Air \nquality \nC2 – \nEnvironme\nntal \ninfrastruct\nure \nC3 – \nPotential \nhealth \nhazards \n21111005900 0.7077 0.68 0.6862 0.6941 0.6199 0.5695 0.6029 0.6397 0.6533 0.6254 \n21111009103 0.6831 0.6559 0.6618 0.6695 0.6025 0.5699 0.5194 0.6272 0.6509 0.5986 \n21111011002 0.6497 0.6233 0.6416 0.6361 0.574 0.5641 0.5217 0.5928 0.6058 0.5973 \n21111011901 0.6594 0.6548 0.6325 0.6458 0.5955 0.5948 0.4958 0.6127 0.6532 0.6105 \n21111012701 0.7 0.676 0.6929 0.6864 0.6498 0.546 0.5734 0.6657 0.6477 0.5894 \nLogistically \nfeasible public \nhealth \nintervention \n Tree \nplanting \nefforts \ninclusive of \nsite selection \nand \npreparation, \nlong term \nmaintenance \nand \ncommunity \nengagement, \nand \nprotective \npolicy. \nNoise \npollution \nreduction \nachieved \nthrough \nnoise \nreduction \nfeatures \n(sound \nwalls, \nvegetative \nbuffers, etc.) \nand policy \n(elevation \nstandards \nfor aircraft \nand \nresidential \nnoise \nregulations.) \nAir quality \nimproveme\nnts through \nlocal clean \nair \nstandards, \npollution \ncontrol \nrequirement\ns, best \navailable \ntechnologie\ns, and \ncompliance \nmonitoring \nand \nreporting \n \nA \ncombination \nof air \nmonitoring \nand noise \nexposure \ninterventions\n. \n \nEconomic \nand policy \ninvestment \nto remove or \nreduce the \nnumber of \nenvironment\nal sources of \npollution \nfrom \nresidential \nareas; and \nincreasing \ntree \ncoverage. \n \nWetland \nrestoration \nand green \ninfrastructur\ne; upgrading \ndrainage \nsystems and \ncritical \ninfrastructur\ne; integration \nof flood \nmanagement \ninto broader \nenvironment\nal planning. \n \nA \ncombinatio\nn of \nsuggested \nintervention\ns from each \nof the other \ncategories. \n \nA \ncombination \nof suggested \nintervention\ns from each \nof the other \ncategories. \n \nA \ncombination \nof suggested \ninterventions \nfrom each of \nthe other \ncategories. \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n21 \n \nFor Intervention B, entire domains with several indicators were adjusted: environmental \nexposures and risks, environmental sources, or weather. Across the case studies, domain \nmodifications led to enhancements in the ToxPi composite scores ranging from 7% to 25%. \nWhen indicator scores within a domain were adjusted to the best scores observed in the county, \nthe census tract’s composite ToxPi scores were reduced by an average of 11% for environmental \nexposures and risks, 16% for environmental sources, and 20% for extreme weather.  \n \nFor Intervention C, thematically related intervention clusters of indicators were adjusted: air \npollution, environmental infrastructure, or potential health hazards. Across the case studies, \nmodifications to these clusters led to enhancements in the ToxPi composite scores ranging from \n1 to 16%. When indicator scores for each cluster were reduced to the best scores observed in the \ncounty, the census tract’s composite ToxPi scores were reduced on average for air quality at 8%, \nfor environmental infrastructure at 6%, and for potential health hazards at 11%. \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n22 \n \nDiscussion \nIn this study, we used a modified version of community based EVI framework1 to assess \nvulnerabilities in a new geographic region. We extended the tool to model interventions and to \nassess potential impact on hyperlocal risks. Our approach provides a data-driven guide for \nLouisville Metro-Jefferson County as examples of census tract scale solutions as opposed to \nmore general public health programming. The index is customized to the particular concerns of \nour area, including the addition of indicators such as noise pollution exposure, heatwave \nexposure, and tornados. While Messer et al.\n43 developed neighborhood socioeconomic context \nfor 19 cities, it excluded flood, growing urban heat island effect, and tornados which are locally \nimportant vulnerabilities in the Louisville Metro-Jefferson County area. The presence of a \ncommercial air-hub led to the addition of the noise pollution indicator. These risks are real for \nmany cities and the expanded evaluation demonstrates the benefits of adaptation to local \nconditions, risks, and vulnerabilities. The Louisville Metro-Jefferson County has unique \nvulnerabilities when compared with most other counties across of the nation, although the area \ndoes resemble several towns in the United States Midwest. For example, all 190 census tracts in \nLouisville Metro-Jefferson County have higher average PM 2.5 concentrations than the \nnationwide average of 3.79 µg/m3. Identification and incorporation of such local vulnerabilities \nmay be key to developing local and well-targeted interventions. \n \nPrevious work suggests that single-indicator interventions to improve health are theoretically \nfeasible. Wang et al.\n40 reported on the effectiveness of an urban green and blue space \nintervention in improving wellbeing. Brown et al.44 and Hammer et al.45 discovered that direct \nregulation on lowering noise at its source, expanding and increasing access to noise maps, and \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n23 \n \naltering the built environment can be effective. Further, Casey et al.7 documented improved \nhealth outcomes resulting from the removal of an environmental point source of pollution when \nthe retirement of a coal-fired power plant led to improved asthma outcomes. \n \nOur index integrates several related indicators into a single domain and it incorporates social, \nclinical, and environmental data to provide a more comprehensive evaluation of area \nvulnerability. In previous work, other environmental health vulnerability studies conducted at \ncensus tract level have focused on only clinical data or only a few public data sources,\n40,46 \ndespite the probability that a wide range of structural environmental variables can impact health. \nFor instance, social vulnerability combines data on socioeconomic status, housing and \ntransportation, and health insurance coverage. When assessing access to healthcare, this \ncombined view of indicators may be more explanatory of lack of healthcare access than a simple \nhospital proximity indicator. \n \nOur findings suggest that environmental vulnerability could be feasibly estimated at census tract \nscale, rather than county level, consistent with the Glassman et al.\n5 report. In addition, our work \ndemonstrates that county-level data could be too broad to support intervention evaluation. \nCensus tract level data offers balanced spatial granularity useful to discern hyperlocal variations \nwithin urban areas whereby environmental risks can be viewed within the context of the entire \nspectrum of risks.  \n \nThe index enabled the selection of the most at-risk census tracts for modeling targeted public \nhealth interventions to reduce cumulative environmental vulnerability within specific \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n24 \n \ncommunities. Our work also suggests that while some census tracts may require plans tailored to \nthose specific areas, it may be possible to develop interventions effective in addressing common \nproblems that span boundaries across multiple tracts. Moreover, our work across the three case \nstudies also suggests that improving any single indicator may not be highly beneficial in \naddressing the overall area vulnerability and that it may be necessary to address multiple \nindicators collectively. Interesting, a high composite vulnerability scores can mask low indicator-\nlevel vulnerabilities (e.g., good tree coverage). Overall, our index and our method for computing \naggregate vulnerabilities maybe helpful in identifying focal points for resource allocation. \nHowever, additional work will be required to assess the efficacy of any intervention. For \ninstance, Louisville’s 14 hospitals are clustered in one geographic area thus limiting access, as \nmeasured by geographical proximity, for many census tracts. Although this proximity indicator \nreflects an equal impact geographically, it may result in a stronger impact on individuals at the \nlow end of the socioeconomic spectrum compared to those at the highest; affluent individuals \nhave significant resources, personal transportation, and insurance to access healthcare, regardless \nof distance to the nearest hospital. Thus, a healthcare access intervention may be warranted for \nlow-income census tracts to offset this inequity. Moreover, in our area of interest seven census \ntracts are within a 5-kilometer radius of Superfund sites, and all of these are west of a structural \nenvironment variable, Interstate 65, mirroring larger vulnerability patterns in the county. This \nsuggests purposively modifying some indicators may have a disproportionately positive impact \non the county’s most vulnerable census tracts. Therefore, as indicated by these two examples, \nseveral indicators are highly correlated and therefore changes in one could have far reaching \neffect on the other or may benefit only related aspects of those vulnerabilities. \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n25 \n \nThreshold values across both the composite score and indicators can be used to manage \nexpectations related to interventions. More affluent census tracts in Jefferson County have ToxPi \ncomposite scores of less than 0.3. These areas are often characterized by high life expectancy, \nlow social vulnerability, and limited pollution exposure. As well per a single indicator, while 40-\n45% is a goal set by Louisville’s Urban Tree Canopy Assessment\n42 the lowest tree canopy score \nin our study area is 2%, and any short-term aim to increase the canopy indicator to 40% may be a \nhighly unrealistic goal. When we modeled the 45% canopy coverage, it only improved the ToxPi \ncomposite score to be 3% better. These results put in perspective the expected indicator gains \nthat could be accrued by specific interventions and should help in prioritizing interventions. \nMoreover, some indicator scores may not change without a modification in direct policy and/or \nfinancial intervention, such as the presence of industrial facilities. Finally, when selecting areas \nfor an intervention, it may be important to consider ethical questions around withholding \nenvironmental health interventions known to be effective.\n47 \n \nThe main purport of the framework developed by our work may be to guide actions to mitigate \nvulnerability and risk. For example, local governance could use this framework to create action \nplans for increasing neighborhood resilience and preparedness following major events. Local \npublic health agencies could use also digital platforms and targeted advertising campaigns to \nraise awareness and promote health actions in specific places.\n48 Qualitative researchers could use \nthe vulnerability index to gather community input to validate proposed solutions. Finally, \ngrassroots community organizations could use an assessment of environmental vulnerability to \nidentify problems, construct potential solutions, and advocate for policy change. Establishing a \ndata-driven index to inform interventions tailored to specific needs, rather than generic \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n26 \n \napproaches would also ensure accountability by verifying that the interventions are directed \ntowards areas where they are most needed and address the area’s most significant vulnerabilities.  \n \nOverall, we believe our study advances the integrated vulnerability index approach for selecting \ncensus tracts for health interventions in several key ways: \n1. Consideration of multiple domains and indicators . Consideration of health parameters \nbeyond traditional clinical and environmental data can inform more effective, data-driven, public \nhealth communications, and interventions. \n2. Data anchored to census tract boundaries . Focus on census tract level vulnerability \nrather than countywide scale.  \n3. Prioritization of intervention and investment by data-driven vulnerability . Our \napproach shows which census tracts are in most need of vulnerability mitigation and \nintervention. In some cases, the same intervention may benefit multiple census tracts even if they \nare not geographically adjacent.  \n4. Shift focus away from the single indicator interventions . Modeled results show \nimproving a group of related indicators is more impactful than adjusting a single indicator.  \n \nDespite its many strengths, the study has some limitations. We relied solely on publicly available \nnational, state, and local government data sources, potentially overlooking data requiring Data \nTransfer Agreements or open records requests. Application may be limited for rural census tracts \ndue to a lack of local data available. We restricted our analysis to Louisville Metro-Jefferson \nCounty, neglecting the bordering counties of the metro area. All domains are equally weighted as \nwe could not justify indicator-specific weighting. Qualitative research and community input can \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n27 \n \nbe used in tools to validate vulnerability frameworks; however, integration of qualitative and \nquantitative data sources can pose additional challenges.49 Our modeled interventions have \nlimitations; determining efficacy of intervention requires empirical testing to establish the degree \nto which environmental vulnerabilities are modifiable. One major limitation of our approach is \nthat scores are percentile-based which does not distinguish between indicators with low and high \nvariance. However, we addressed this limitation by selecting indicators for interventions that \ncould feasibly produce a material change, such as percent tree canopy, achievable in real-world \nconditions.   \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n28 \n \nConclusion \nIn this study we used census tract boundaries and integrated vulnerability data to develop a \ncomprehensive vulnerability index which could be used to assess the potential impact of targeted \npublic health interventions. While the original framework was designed to assign relative ranks \nbased on environmental vulnerability, the aggregation of vulnerability and intervention allows \nthe index to develop into a more comprehensive tool for public health, community planning, and \nenvironmental justice advocates. Chief among our insights is that places are not defined by their \nenvironmental vulnerabilities and that such vulnerabilities are not invariant features of living \ncommunities. Rather, knowledge of these risks can spur action for towards local and bespoke \nsolutions. Targeted census tract interventions may be more effective than broad-scale campaigns \nat state or national levels and may lead to health-supportive environmental conditions with \ngreater geographic precision.  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n29 \n \nContributors \nConceptualization: TS, WC, AB; Methodology: TS, AB; Formal analysis: LBA, CB, DJB, RHH; \nWriting-original draft preparation: LBA, RHH; Writing-review and editing: TS, LBA, CB, DJB, \nWC, RHH, AB; Supervision: TS, AB; Project administration: LBA. All the authors have read \nand agreed to the published version of the manuscript.  \n \nData sharing \nAll source data are publicly available, links are provided in Table S2. \n  \nFunding \nThis work was supported in part by the Owsley Brown II Family foundation, NIEHS (P42 \nES023716 to the University of Louisville; and the P30 ES029067 and P42 ES027704 to Texas \nA&M University) and the Robert Wood Johnson Foundation (Grant #: 80565).  \n \nAcknowledgments \nThe authors thank Ray Yeager, Rebecca Turney, and Angelina Rangel for their mapping \nexpertise, Dhiraj Kanneganti for ToxPi skills developed during this project, and Laney Taylor for \ngathering and organizing the initial data for this research.  \n \nDeclaration of interests \nThe authors declare they have nothing to disclose.  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n30 \n \nReferences \n1. Bhandari S, Lewis PGT, Craft E, Marvel SW, Reif DM, Chiu WA. 2020. \nHGBEnviroScreen: enabling community action through data integration in the Houston–\nGalveston–Brazoria region. Int J Env Res Pub He 17(4):1130. \nhttps://doi.org/10.3390/ijerph17041130 \n2. Lewis PG, Chiu WA, Nasser E, Proville J, Barone A, Danforth C, et al. 2023. \nCharacterizing vulnerabilities to climate change across the United States. Environ Int \n172:107772. https://doi.org/10.1016/j.envint.2023.107772 \n3. Marvel SW, To K, Grimm FA, Wright FA, Rusyn I, Reif DM. 2018. ToxPi graphical \nuser interface 2.0: Dynamic exploration, visualization, and sharing of integrated data \nmodels. BMC Bioinform 19:80. https://doi.org/10.1186/s12859-018-2089-2  \n4. Reif DM, Sypa M, Lock EF, Wright FA, Wilson A, Cathey T, et al. 2013. ToxPi GUI: an \ninteractive visualization tool for transparent integration of data from diverse sources of \nevidence. Bioinformatics 29(3):402–3. https://doi.org/10.1093/bioinformatics/bts686  \n5. Glassman B. 2020. The multidimensional deprivation index using different neighborhood \nquality definitions. United States Census Bureau Social, Economic, and Housing \nStatistics Division. https://www.census.gov/content/dam/Census/library/working-\npapers/2020/demo/SEHSD-WP2020-08.pdf [accessed 24 April 2024]. \n6. Centers for Disease Control and Prevention. 2024. National Notifiable Diseases \nSurveillance System (NNDSS). https://www.cdc.gov/nndss/index.html [accessed 24 April \n2024]. \n7. Casey JA, Su JG, Henneman LR, Zigler C, Neophytou AM, Catalano R, et al. 2020. \nImproved asthma outcomes observed in the vicinity of coal power plant retirement, \nretrofit and conversion to natural gas. Nature Energy 5(5):398–408. \nhttps://doi.org/10.1038/s41560-020-0600-2 \n8. Hilmers A, Hilmers DC, Dave J. 2012. Neighborhood disparities in access to healthy \nfoods and their effects on environmental justice. Am J Public Health 102(9):1644–54. \nhttps://doi.org/10.2105/AJPH.2012.300865  \n9. Gochfeld M, Burger J. 2011. Disproportionate exposures in environmental justice and \nother populations: the importance of outliers. Am J Public Health 101(S1):S53–63. \nhttps://doi.org/10.2105/AJPH.2011.300121  \n10. Johnson DP, Stanforth A, Lulla V, Luber G. 2012. Developing an applied extreme heat \nvulnerability index utilizing socioeconomic and environmental data. Appl Geogr 35(1–\n2):23–31. https://doi.org/10.1016/j.apgeog.2012.04.006  \n11. Marvel SW, House JS, Wheeler M, Song K, Zhou YH, Wright FA, et al. 2021. The \nCOVID-19 Pandemic Vulnerability Index (PVI) Dashboard: monitoring county-level \nvulnerability using visualization, statistical modeling, and machine learning. Environ \nHealth Perspect 29(1):017701. https://doi.org/10.1289/EHP8690  \n12. Nasiri H, Yusof MJM, Ali TAM, Hussein MKB. 2019. District flood vulnerability index: \nurban decision-making tool. Int J Environ Sci Te 16(5):2249–58. \nhttps://doi.org/10.1007/s13762-018-1797-5  \n13. Newman G, Malecha M, Atoba K. 2023. Integrating ToxPi outputs with ArcGIS \nDashboards to identify neighborhood threat levels of contaminant transferal during flood \nevents. J Spat 68(1):57–69. https://doi.org/10.1080/14498596.2021.1891149  \n14. Xue J, Zartarian V, Tornero-Velez R, Stanek LW, Poulakos A, Walts A, et al. 2022. A \ngeneralizable evaluated approach, applying advanced geospatial statistical methods, to \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n31 \n \nidentify high lead exposure locations at census tract scale: Michigan case study. Environ \nHealth Persp 130(7):077004. https://doi.org/10.1289/EHP9705 \n15. World Health Organization. 2000. Considerations in evaluating the cost-effectiveness of \nenvironmental health interventions. https://www.who.int/publications/i/item/WHO-SDE-\nWSH-00.10 [accessed 9 April 2024]. \n16. Bouzid M, Hooper L, Hunter PR. 2013. The effectiveness of public health interventions \nto reduce the health impact of climate change: a systematic review of systematic reviews. \nPLoS One 8(4):e62041. https://doi.org/10.1371/journal.pone.0062041 \n17. Environmental Defense Fund & Texas A&M University. 2023. The U.S. Climate \nVulnerability Index. https://climatevulnerabilityindex.org/ [accessed 21 May 2024]. \n18. United States Census. Jefferson County, Kentucky. Population, Census, April 1, 2020. \nhttps://www.census.gov/quickfacts/fact/table/jeffersoncountykentucky/POP010220 \n[accessed 24 April 2024].  \n19. University of Kentucky. 2023. Kentucky Annual Economic Report 2023. \nhttps://cber.uky.edu/sites/cber/files/publications/UK%20CBER%20Kentucky%20Annual\n%20Economic%20Report%202023_Web.pdf [accessed 3 May 2024]. \n20. United States Climate Resilience Toolkit. The Climate and Economic Justice Screening \nTool (CEJST) https://toolkit.climate.gov/tool/climate-and-economic-justice-screening-\ntool \n[accessed 9 April 2024]. \n21. Centers for Disease Control and Prevention. 2021. Places: Census Tract Data – GIS \nFriendly Format https://chronicdata.cdc.gov/500-Cities-Places/PLACES-Census-Tract-\nData-GIS-Friendly-Format-2021-/yjkw-uj5s/data [accessed 1 Nov 2023]. \n22. Kentucky Cabinet for Health and Family Services. Division of Health Care Facilities. \nhttps://chfs.ky.gov/agencies/os/oig/dhc/Pages/hcf.aspx [accessed 2 November 2023]. \n23. National Center for Health Statistics. 2018. U.S. Small-Area Life Expectancy Estimates \nProject (USALEEP): Life Expectancy Estimates File for Kentucky, 2010-2015. \nhttps://www.cdc.gov/nchs/nvss/usaleep/usaleep.html [accessed 1 November 2023]. \n24. United States Department of Housing and Urban Development. 2023. Environmental \nHealth Hazard Index. https://data.lojic.org/datasets/HUD::environmental-health-hazard-\nindex/about [accessed 9 November 2023]. \n25. United States Environmental Protection Agency. 2020. Chemical concentrations, \nexposures, health risks by census tract from National Scale Air Toxics Assessment \n(NATA). https://catalog.data.gov/dataset/chemical-concentrations-exposures-health-\nrisks-by-census-tract-from-national-scale-air-toxics [accessed 9 November 2023]. \n26. United States Department of Transportation. 2020. National Transportation Noise Map. \nttps://www.bts.gov/geospatial/national-transportation-noise-map [accessed 2 November \n2023]. \n27. United States Environmental Protection Agency. 2021. Daily Census Tract-Level PM2.5 \nConcentrations, 2016 - 2020 https://data.cdc.gov/Environmental-Health-\nToxicology/Daily-Census-Tract-Level-PM2-5-Concentrations-2016/96sd-\nhxdt/about_data [accessed 8 November 2023]. \n28. United States Environmental Protection Agency. 2021. Risk-Screening Environmental \nIndicators (RSEI) Model RSEI Version 2.3.11 (RY 2021). \nhttps://edap.epa.gov/public/extensions/EasyRSEI/EasyRSEI.html [accessed 2 November \n2023]. \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n32 \n \n29. United States Environmental Protection Agency. 2023. Facility Registry Service \nFacilities State Single File CSV Download. https://www.epa.gov/frs/epa-frs-facilities-\nstate-single-file-csv-download [accessed 16 November 2023]. \n30. Kentucky Energy and Environmental Cabinet. Superfund Site List. \nhttps://dep.gateway.ky.gov/eSearch/AvailableReports/Details?id=47 [accessed 14 \nNovember 2023]. \n31. United States Environmental Protection Agency. 2021a. EJScreen: Environmental Justice \nScreening and Mapping Tool. Traffic proximity and volume dataset. \nhttps://gaftp.epa.gov/EJScreen/ [accessed 16 November 2023]. \n32. United States Geological Survey, U.S. Department of the Interior, and U.S. Forest \nService. 2021. Tree Canopy Cover of the Conterminous United States, 2021 NLCD 2021 \nUSFS Tree Canopy Cover (CONUS). Multi-Resolution Land Characteristics (MRLC) \nConsortium. https://www.mrlc.gov/data/nlcd-2021-usfs-tree-canopy-cover-conus \n[accessed 11 November 2023]. \n33. United States Department of Housing and Urban Development. 2023. Location \nAffordability Index (Version 3). https://data.lojic.org/datasets/HUD::location-\naffordability-index-v-3/about [accessed 17 November 2023].  \n34. Centers for Disease Control and Prevention. 2020. Agency for Toxic Substances and \nDisease Registry/ Geospatial Research, Analysis, and Services Program. CDC/ATSDR \nSocial Vulnerability Index, Database Kentucky. \nhttps://www.atsdr.cdc.gov/placeandhealth/svi/data_documentation_download.html. \nAccessed on 20 November 2023. \n35. United States Department of Agriculture, Economic Research Service. 2019. Food \nAccess Research Atlas. https://www.ers.usda.gov/data-products/food-access-research-\natlas/download-the-data/ [accessed 17 November 2023]. \n36. Esri. 2020. USA_Flood_Hazard_Reduced_Set_gdb. ArcGIS Online Map Viewer. \nhttps://www.arcgis.com/apps/mapviewer/index.html?layers=bf9585afc2934b648f391869\n3285b7c8 [accessed 20 November 2023]. \n37. Jefferson County Information Consortium. 2022. Jefferson County Kentucky Urban Heat \nManagement Study (LOJIC). https://hub.arcgis.com/datasets/LOJIC::jefferson-county-\nky-urban-heat-management-study [accessed 8 May 2024]. \n38. Federal Emergency Management Agency. 2023. National Risk Index (NRI) Data & \nResources. https://hazards.fema.gov/nri/data-resources [accessed 23 September 2023]. \n39. Toxicological Prioritization Index (ToxPi) version 1.2.1 (https://CRAN.R-\nproject.org/package=toxpiR ) \n40. Wang R, Browning MH, Kee F, Hunter RF. 2023. Exploring mechanistic pathways \nlinking urban green and blue space to mental wellbeing before and after urban \nregeneration of a greenway: Evidence from the Connswater Community Greenway, \nBelfast, UK. Landscape Urban Plan 235:104739. \nhttps://doi.org/10.1016/j.landurbplan.2023.104739 \n41. O'Cathain A, Croot L, Duncan E, Rousseau N, Sworn K, Turner KM, et al. 2019. \nGuidance on how to develop complex interventions to improve health and healthcare. \nBMJ Open 9(8):e029954. http://dx.doi.org/10.1136/bmjopen-2019-02995 \n42. Louisville Metro-Jefferson County Government. Louisville Urban Tree Canopy \nAssessment, 2015. https://louisvilleky.gov/urban-forestry/document/louisville-urban-tree-\ncanopy-assessment-2015 [accessed 9 April 2024].  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n33 \n \n43. Messer LC, Laraia BA, Kaufman JS, Eyster J, Holzman C, Culhane J, et al. 2006. The \ndevelopment of a standardized neighborhood deprivation index. J Urban Health \n83(6):1041–62. https://doi.org/10.1007/s11524-006-9094-x  \n44. Brown AL, Van Kamp I. 2017. WHO environmental noise guidelines for the European \nRegion: A systematic review of transport noise interventions and their impacts on health. \nInt J Env Res Pub He 14(8):873. https://doi.org/10.3390/ijerph14080873 \n45. Hammer MS, Swinburn TK, Neitzel RL. 2014. Environmental noise pollution in the \nUnited States: developing an effective public health response. Environ Health Perspect \n122(2):115–9. http://dx.doi.org/10.1289/ehp.1307272. \n46. Uong SP, Zhou J, Lovinsky-Desir S, Albrecht SS, Azan A, Chambers EC, et al. 2023. \nThe creation of a multidomain neighborhood environmental vulnerability index across \nNew York City. J Urban Health 100(5):1007–23. https://doi.org/10.1007/s11524-023-\n00766-3  \n47. Resnik DB, Zeldin DC, Sharp RR. 2005. Research on environmental health interventions: \nethical problems and solutions. Accountability Res 12(2):69–101. \nhttps://doi.org/10.1080/08989620590957157 \n48. Anderson LB, Ness HD, Holm RH, Smith T. 2024. Wastewater-informed digital \nadvertising as a Covid-19 geotargeted neighborhood intervention: Jefferson County, \nKentucky, 2021–2022. Am J Public Health 114(1):34–7. \nhttps://doi.org/10.2105/AJPH.2023.307439  \n49. Ho HC, Wong MS, Man HY, Shi Y, Abbas S. 2019. Neighborhood-based subjective \nenvironmental vulnerability index for community health assessment: development, \nvalidation and evaluation. Sci Total Environ 654:1082–90. \nhttps://doi.org/10.1016/j.scitotenv.2018.11.136   \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n34 \n \nFigure Legends \n \nFigure 1. Study area, Louisville Metro-Jefferson County area in Kentucky, USA. The \nClimate and Economic Justice Screening Tool (CEJST) designation by census tract as well as \nfederal and state Superfund sites, the international airport with a commercial air-transport hub, \ntwo large interstate highways, and the Ohio River on the north are represented. \n \nFigure 2. Framework for Environmental Vulnerability Index supporting simulation of \ntargeted public health interventions at the census tracts level. Using the Toxicological \nPrioritization Index composite scores, the five most vulnerable census tracts were identified and \nmodifiable indicators were selected under three intervention scenarios as theoretical action for \nsolutions.  \n \nFigure 3. Toxicological Prioritization Composite Score Index by census tract, Louisville \nMetro-Jefferson County, Kentucky (USA). Panel A: Composite scores by census tract \nwhereby a higher score corresponds to a more vulnerable census tract. Census tracts were sorted \ninto quartiles by overall score, the quartiles were scored as: 1\nst quartile as lowest vulnerable \ncensus tracts <0.32 (n=48); 2nd quartile 0.33 to 0.41 (n=48); 3rd quartile 0.42 to 0.51 (n=47); and \n4th quartile as most vulnerable census tracts >0.52 (n=47). Panel B: Modeled clusters of \ncomposite scores by census tract. The Local Moran’s I analysis indicates statistically significant \nclusters of high or low values, as well as outliers where high values are adjacent to low values, \nand vice-versa.  \n \nFigure 4. Toxicological Prioritization Domain Score Index by census tract, Louisville \nMetro-Jefferson County, Kentucky (USA). Top: Domain scores by census tract whereby a \nhigher score corresponds to a more vulnerable census tract. Panels present the domain scores for: \nA.) Baseline Health; B.) Environmental Exposures and Risks; C.) Environmental Sources; D.) \nSocial Vulnerability; and E.) Extreme Weather. Bottom: Modeled clusters of domain scores by \ncensus tract. The Local Moran’s I analysis indicates statistically significant clusters of high or \nlow values, as well as outliers where high values are adjacent to low values, and vice-versa. \nPanels present the domain scores for: A.) Baseline Health; B.) Environmental Exposures and \nRisks; C.) Environmental Sources; D.) Social Vulnerability; and E.) Extreme Weather. \n \n \n \n  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n35 \n \nSupplementary Information \nAn Environmental Vulnerability Index framework supporting targeted public health \ninterventions at the census tracts level \n \nLauren B. Anderson\n1,2*, Rochelle H. Holm1*, Caison Black1, Donald J. Biddle3, Weihsueh A. \nChiu4, Aruni Bhatnagar5, and Ted Smith1 \n \n1Center for Healthy Air, Water and Soil, Christina Lee Brown Envirome Institute, School of \nMedicine, University of Louisville, Louisville, KY, USA \n2Department of Urban and Public Affairs, College of Arts & Sciences, University of Louisville, \nLouisville, KY, USA \n3Department of Geographic and Environmental Sciences, Center for Geographic Information \nSciences, University of Louisville, Louisville, KY, USA \n4Department of Veterinary Physiology and Pharmacology, College of Veterinary Medicine and \nBiomedical Sciences, Texas A&M University, College Station, TX, USA \n5Christina Lee Brown Envirome Institute, School of Medicine, University of Louisville, \nLouisville, KY, USA \n  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n36 \n \nTable of Contents \nTable S1. Vulnerabilities of Houston–Galveston–Brazoria region, Texas, and Louisville Metro- \nJefferson County, Kentucky. ................................................................................................... .... 37 \nTable S2. Data sources of indicators incorporated into Environmental Vulnerability Index for \nLouisville Metro- Jefferson County, Kentucky (USA). ............................................................... 38 \nTable S3. Indicators included in the Houston–Galveston–Brazoria (HGB) EnviroScreen’s \nEnvironmental Vulnerability Index (EVI) (Bhandari et al., 2020) which were removed for the \nLouisville Metro-Jefferson County index. ................................................................................... 42 \nTable S4. Indicators subgrouped based upon related impacts across domains for Intervention C \ncase studies. ................................................................................................................. ............... 43 \nFigure S1. Heterogeneity analysis for the 32 indicators. Performed for each indicators to ensure \nonly factors that offer significant variance were included in the index (32 panels). .................... 44 \n \n  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n37 \n \nTable S1. Vulnerabilities of Houston–Galveston–Brazoria region, Texas, and Louisville Metro- \nJefferson County, Kentucky.  \nVulnerability Houston–\nGalveston–\nBrazoria region, \nTexas \nLouisville Metro- \nJefferson County, \nKentucky \nData Source \nCOVID-19 impact 1,547,659 cases \n13586 deaths \n289,020 cases \n2263 deaths \nhttps://www.nytimes.com/\n \nPopulation 5,453,858 773,399 https://data.census.gov/ \nArea 3,446.2 square \nmiles \n380.6 square miles https://data.census.gov/ \nLarge industrial \ncomplexes \nYes Yes  \nPoverty rate 7.6%-16.4% 14.0%  https://data.census.gov/ \nNatural environment Along the Gulf of \nMexico \nOhio River  \nResearch focused \nhigher-education \ninstitution  \nRice University \nBaylor College of \nMedicine \nUniversity of \nLouisville \n \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n38 \n \nTable S2. Data sources of indicators incorporated into Environmental Vulnerability Index for Louisville Metro- Jefferson County, \nKentucky (USA). \n \nDomain Indicator Reference Data Description (File names and filter applied as \napplicable) \nAccess \nDate \n(MM-DD-\nYEAR) \nBaseline Healtha Adult asthma (>18 years \nold) \n21 PLACES: Census Tract Data: Column C, \"Current asthma \namong adults aged >=18 years\" \n11-01-2023 \n \n Chronic obstructive \npulmonary disease \n21 PLACES: Census Tract Data: Column C, \"Chronic \nobstructive pulmonary disease among adults aged >=18 \nyears\" \n11-01-2023 \n Coronary heart disease 21 PLACES: Census Tract Data: Column C, \"Coronary heart \ndisease among adults aged >=18 years\" \n11-01-2023 \n Hospitals within 5km \nradius  \n22 Hospital Directory By County: Columns C/D/E entered \ninto ArcGIS Pro to calculate 5km radius \n11-02-2023 \n Life expectancy 23 Kentucky Life Expectancy: Column E, \"Life Expectancy in \nyears\" \n11-01-2023\n \n Stroke 21 PLACES: Census Tract Data: Column C, \"Stroke among \nadults aged >=18 years\" \n11-01-2023 \nEnvironmental \nExposures and \nRisksb \nEnvironmental Hazard \nIndex \n24 Hazard Index: Column D, “Haz_IDX” 11-09-2023 \n National Air Toxics \nAssessment cancer risk \n25 NATA 2019 National: Sheet 2, Column G, “Total Cancer \nRisk (Per million)” \n11-09-2023 \n National Air Toxics \nAssessment \nreproductive risk  \n25 NATA 2019 National: Column G, “Reproductive Health” 11-09-2023  \n National Air Toxics \nAssessment respiratory \nrisk \n25 NATA 2019 National: Column C, “Respiratory Health” 11-09-2023  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n39 \n \nDomain Indicator Reference Data Description (File names and filter applied as \napplicable) \nAccess \nDate \n(MM-DD-\nYEAR) \n Noise exposure  26 NoiseToCensus_JeffCo: Column B, “MEAN_Noise_dB” 11-02-2023 \n PM2.5 Community \nMultiscale Air Quality  \n27 Daily Census Tract Level PM2.5 Concentration: Column \nH, “DS_PM_pred” \n11-08-2023  \n Risk-screening \nenvironmental \nindicators (RSEI) \n28 EPA Easy RSEI Dashboard: Column F, “RSEI Score” 11-02-2023  \nEnvironmental \nSourcesc  \nCement batch plants \nwithin 5km radius \n29 EPA FRS Facilities State Single File: SIC codes \n3273,3272,3271,2951. Entered into ArcGIS Pro to locate 5 \nkm radius of census tract. \n11-16-2023\n \n Metal recyclers within \n5km radius \n29 EPA FRS Facilities State Single File: SIC code 5093. \nEntered into ArcGIS Pro to locate 5 km radius of census \ntract. \n11-16-2023  \n Petrochemical and oil \nrefineries within 5km \nradius \n29 EPA FRS Facilities State Single File: SIC code 2911. \nEntered into ArcGIS Pro to locate 5 km radius of census \ntract. \n11-16-2023 \n Power plants within \n5km radius \n29 EPA FRS Facilities State Single File: SIC code 4911,3612. \nEntered into ArcGIS Pro to locate 5 km radius of census \ntract. \n11-16-2023  \n Superfund sites within \n5km radius \n30 EPA Superfund Sites: Column B,C,D,E. Entered into \nArcGIS Pro to locate 5 km radius of census tract. \n11-14-2023  \n Traffic proximity and \nvolume \n31 EPA EJScreen: PTRAF calculated by Annual Average \nDaily Traffic divided by distance in meters \n11-16-2023 \n \n Tree canopy coverage  32 NLCD 2021 USFS Tree Canopy: Entered into ArcGIS Pro \nto locate 5 km radius of census tract. \n11-11-2023 \nSocial \nVulnerabilityd \n% Income for housing 33 Location Affordability Index v.3: column \nR\"avg_h_cost\"X12/column AC \"median_hh_income\" \n11-17-2023\n \n % Renters 33 Location Affordability Index v.3: column K, “pct_renters” 11-17-2023  \n % Without health \ninsurance \n34 CDC/ATSDR SVI Data: Column DR, “EP_UNINSUR” 11-17-2023  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n40 \n \nDomain Indicator Reference Data Description (File names and filter applied as \napplicable) \nAccess \nDate \n(MM-DD-\nYEAR) \n Food desert low access 35 FoodDesert, FoodAccessResearchAtlasData:A sum of \nColumn O “LowIncomeTracts” and Column U \n“LATracts_half” \n11-17-2023 \n \n Household composition \nand disability \n34 CDC/ATSDR SVI Data: Column CH, “RPL_THEME2” 11-17-2023  \n Housing and \ntransportation \n34 CDC/ATSDR SVI Data: Column CS, “RPL_THEME4” 11-17-2023  \n Median renter income \nvs. median area income \n33 Location Affordability Index v.3: column AB, \n“area_income_renter_frac” \n11-17-2023 \n \n Minority status and \nlanguage \n34 CDC/ATSDR SVI Data: Column CL, “RPL_THEME3” 11-17-2023  \n Socioeconomic status 34 CDC/ATSDR SVI Data: Column CB, “RPL_THEME1” 11-17-2023  \nWeathere 1% Annual probability \nflood hazard \n36 ESRI Living Atlas USA Flood Hazard Reduced Set: \nColumn B, “1% Flood Hazard Percent Area Coverage” \n11-20-2023 \n \n Warm season average \nmaximum temperature \n37 Jeff Co KY Urban Heat Management Study: Column I, \n“WS_B_APMAX ” \n05-08-2024 \n Tornado \n \n38 National Risk Index: Column MH, “Tornado Annualized \nFrequency” \n04-23-2024 \na The Baseline Health domain contains six indicators related to disease, longevity, and health care access to show where residents \nthemselves are more vulnerable to environmental impacts. CDC Places data for Jefferson County was isolated and downloaded by \ncensus tract. Disease prevalence estimates for asthma, chronic obstructive pulmonary disease (COPD), stroke, and coronary heart \ndisease (CHD) were assessed. CDC Small Area Life Expectancy Estimates for Jefferson County were isolated and downloaded by \ncensus tract. Acute and psychiatric hospital addresses were downloaded from Kentucky’s Cabinet for Health and Family Services, \ngeocoded, and summary statistics were calculated for distance between facilities and each census tract centroid.  \nb The Environmental Exposures and Risks domain contains nine indicators related to ambient pollution and general environmental \nhazards. Estimates for air pollution emissions impacting reproductive health, respiratory health, and total cancer risks were \ndownloaded from the National Air Toxics Assessment (NATA). PM2.5 data was downloaded from the CDC’s community \nmultiscale air quality (CMAQ) program. The U.S. Department of Transportation’s National Transportation Noise Map raster file \nwas downloaded and geocoded to census tracts to calculate mean 24-hour noise exposure. Environmental Hazard Index scores and \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n41 \n \nDomain Indicator Reference Data Description (File names and filter applied as \napplicable) \nAccess \nDate \n(MM-DD-\nYEAR) \nRisk-Screening Environmental Indicator (RSEI) scores were downloaded from the U.S. Department of Housing and Urban \nDevelopment (HUD) and the U.S. EPA, respectively. \nc The Environmental Sources domain contains seven indicators, six are related to sources of environmental exposures and the \nseventh is tree canopy. The U.S. EPA’s Facility Registry Service provided addresses for cement batch plants, metal recyclers, \npetrochemical and oil refineries, and power plants which were geocoded to census tracts to calculate distance between sites and \ncensus tract boundaries. The U.S. EPA’s Superfund program provided addresses for superfund sites which were geocoded to census \ntracts to calculate distance between sites and census tract boundaries. The U.S. Department of Transportation (DOT) provided \ntraffic proximity and volume which was calculated according to the U.S. EPA’s EJScreen methodology (U.S. EPA 2022). The \nMulti-Resolution Land Characteristics Consortium provided tree canopy cover data as a raster which was spatially joined with \ncensus tract polygons to calculate summary statistics. \nd The Social Vulnerability domain contains nine indicators related to the resilience capacity of communities to recover from natural \ndisasters and other crises. Three indicators related to housing affordability were provided by the U.S. Department of Housing and \nUrban Development (HUD) through their Location Affordability Index: percent income used for housing, percent renters, and \nmedian renter income vs. median area income. To calculate percent income used for housing the average housing cost was divided \nby median household income. Five indicators from the CDC and Agency for Toxic Substances and Disease Registry’s (ATSDR) \nSocial Vulnerability Index were included: housing and transportation, household composition, minority status, socioeconomic \nstatus, and health insurance coverage. The U.S. Department of Agriculture’s Food Access Research Atlas provided food access. To \ncalculate scores for food access, LowIncomeTracts and LATracts_half were summed. \ne The Weather domain contains three indicators related to severe weather. Flooding hazard was provided by ESRI’s Living Atlas \nUSA Flood Hazard geodatabase that categorizes FEMA Flood zones by census tract. Warm season average maximum temperature \nvalue represents the mean average, minimum and maximum temperatures over the 2012 warm season (May – September), the \nnumber of deaths attributable to urban heat over the 2012 warm season provided by the Jefferson County Urban Heat Management \nStudy. Tornado annualized frequency value represents the average number of recorded Tornado occurrences (event-days) per year \nprovided by FEMA’s National Risk Index. \n  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n42 \n \nTable S3. Indicators included in the Houston–Galveston–Brazoria (HGB) EnviroScreen’s \nEnvironmental Vulnerability Index (EVI) (Bhandari et al., 2020) which were removed for the \nLouisville Metro-Jefferson County index. \n \nDomain Indicator \nSocial Vulnerability  -Modified food retail environment index  \n-Low food security  \nBaseline Health  -Childhood asthma \nEnvironmental Exposures and \nRisks  \n-PM2.5 Satellite  \nEnvironmental Sources  -Leaking petroleum storage tanks  \n-Facilities with risk management plans  \n-Accident events reported in RMP  \n-Shelter-in-place events reported in RMP \nFlooding  -100-year flood plain  \n-500-year flood plain  \n-Harvey damage assessment “Affected”  \n-Harvey damage assessment “Minimal”  \n-Harvey damage assessment “Major”  \n-Harvey damage assessment “Destroyed”  \n-Families filing Harvey damage claims \n  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n43 \n \nTable S4. Indicators subgrouped based upon related impacts across domains for Intervention C \ncase studies. \n \nSubgroup Domain Indicator High Scores (Census tract) \nC1 - Air \nquality \nEnvironmental NATA Cancer 20 (21111011604) \nEnvironmental NATA 0.3 (21111007501) \nEnvironmental NATA 0.03 (21111010307) \nEnvironmental PM 2.5 5.38923766 (21111011403) \nC2 - \nEnvironmental \nInfrastructure \nEnvironmental Tree Canopy 68.031571 (21111012003) \nEnvironmental \nExposure and \nRisks \nEnvironmental \nHazard Index  62 (21111010307) \nEnvironmental RSEI 0 \nEnvironmental Traffic Exposure 16.40093089 (21111011604) \nC3 - Potential \nHealth \nHazards \nEnvironmental Cement Batch 0 \nEnvironmental Metal Recyclers 0 \nEnvironmental Petrochemical and 0 \nEnvironmental Power Plants 0 \nEnvironmental Superfund Sites 0 \n  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n44 \n \nFigure S1. Heterogeneity analysis for the 32 indicators. Analysis was performed for each \nindicator to ensure only factors that offer significant variance were included in the index (32 \npanels).   \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n45 \n \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n46 \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n47 \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n48 \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n49 \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n50 \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n51 \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n52 \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n53 \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n54 \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n55 \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n56 \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n57 \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n58 \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n59 \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n60 \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n61 \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n62 \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n63 \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n64 \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n65 \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n66 \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n67 \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n68 \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n69 \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n70 \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n71 \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n72 \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n73 \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n74 \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n75 \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint \n\n76 \n \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint","source_license":"CC-BY-4.0","license_restricted":false}