An Environmental Vulnerability Index framework supporting targeted public health interventions at the census tracts level

BACKGROUND: Analyzing and visualizing disparities in environmental risks can help in assessing place-based vulnerabilities and provide civic leaders and community members with essential data about promoting health equity and inform public health strategies. However, there is a lack of effective and integrative tools for evaluating census tract vulnerabilities.

We investigated the adaption of a previously developed environmental vulnerability index to evaluate cumulative impacts of diverse stressors in Louisville Metro-Jefferson County, KY, with the goal of supporting multi-faceted targeted public health interventions at the census tract-level.

We assessed countywide variability in vulnerability using Toxicological Prioritization Index interface across five domains with 32 indicators and modeled the effects of theoretical public health interventions.

Our findings suggest similarly vulnerable areas are not always geographically clustered. Higher vulnerability scores are observed along the western and central areas of the county with lower vulnerability scores in the central urban core and eastern regions. The index enabled the selection of the most at-risk census tracts for modeling targeted public health interventions to reduce cumulative environmental vulnerability. SIGNIFICANCE: Environmental vulnerabilities are not invariant features of urban environments, rather the knowledge of these risks can guide the development and implementation of targeted solutions. IMPACT STATEMENT: Targeted interventions to modify environmental conditions that are supportive of health can be developed and implemented locally with greater precision at the census tract level, yielding impactful outcomes.

census tract; environmental justice; intervention; public health; ToxPi; vulnerability . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 3

Given the importance of the role that environmental factors play in health outcomes, assessing and mapping population health and environmental hazards together could better estimate place- based vulnerabilities and furnish civic leaders and community members with vital data on health equity and environmental risks. This knowledge can facilitate informed decisions regarding the implementation of public health interventions tailored to address specific vulnerabilities within communities. However, addressing the unequal distribution of environmental hazards and vulnerabilities across geographic areas presents multifaceted challenges. 1–4 Health risk and environmental hazards are well described at the state and county levels, but the characteristics that comprise vulnerability are often localized at the census tract or neighborhood level. 5 For instance, infectious disease tracking by the National Notifiable Diseases Surveillance System6 happens at the state level despite the fact that individual disease occurrence is shaped by geographical proximity at a city or a neighborhood scale. 7 Similarly, some determinants of health, such as access to healthy foods, or proximity to only fast food and convenience outlets, are well established to be most impactful at a neighborhood scale. 8 Further, risk and vulnerability do not recognize political boundaries; pollution can cross census tract, neighborhood, and even state boundaries. 9 However, there is paucity of integrative, census tract, vulnerability screening tools to inform targeted public health interventions. Existing vulnerability indices at smaller scales typically focus on a relatively narrow set of issues such as environmental pollution, extreme heat, flooding, disease, or lead exposure and depend on specific data sources which may limit appropriate use and effectiveness. 10–14 The World Health Organization (WHO) provides a framework for global application to evaluate the cost- . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 4 effectiveness of environmental health interventions for air pollution, water supply, sanitation, climate change, food safety, water management, and vector control;15 while lacking place-based intervention framework customization. Moreover, literature regarding public health interventions to address climate-related environmental vulnerabilities and extreme weather events is lacking. 16 Many of these evaluations focus on modifying individual stressors or exposures to individuals, rather than using a place-based framework that integrates cumulative impacts, community exposure, and the natural and built environments. As health inequities continue to grow nationally and new vulnerabilities arise from a changing climate, frameworks that integrate environmental hazard and risk data to understand vulnerability will become increasingly important. There are some efforts to integrate national and local health data in a spatial context to address localized concerns across several environmental domains. 1 For instance, the Houston–Galveston–Brazoria (HGB) EnviroScreen’s Environmental Vulnerability Index (EVI)1 pinpoints which communities need the most support by analyzing health and environmental data geographically. This index preceded the U.S. Climate Vulnerability Index, which integrates indictors nationwide to inform a broad range of policy interventions ranging from health and environment to infrastructure and socio-economic factors. 2,17 A potential benefit of environmental vulnerability assessments includes the prioritization and implementation of layered interventions to reduce cumulative burdens across communities. By employing comprehensive screening tools that merge publicly accessible health data with location-specific environmental indicators, vulnerabilities can be pinpointed. This approach forms a strong foundation for implementing precise public health intervention strategies. The goal of this study was to demonstrate the development of an integrative screening . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 5 tool to inform targeted public health interventions at a metro-area scale. Specifically, we adapted the HGBEnviroScreen framework into a bespoke index for Louisville Metro-Jefferson County. We used the framework to model and evaluate interventions to reduce environmental risk and vulnerabilities, focusing on solutions for communities most burdened by multiple stressors. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 6

Study Area The Louisville Metro-Jefferson County area in Kentucky, USA, is a mid-sized, metropolitan area with a population of 780,000.18 Employment is largely in trade, transportation, and utilities.19 Manufacturing activities are dispersed throughout the county, including a chemical and rubber manufacturing corridor along the western edge of the city. The county features mixed-income housing in the north and west, high-income areas in the east, and middle-to-low-income zones in the south. Many census tracts in the northwestern part of the county are identified by the Climate and Economic Justice Screening Tool20 as facing significant burdens. The people, environment, and infrastructure covering 190 census tracts are additionally affected by the presence of federal and state Superfund sites, an international airport with a commercial air-transport hub, two large interstate highways, and the Ohio River abutting the northern and western boundaries of the county (Figure 1). . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 7 Figure 1. Study area, Louisville Metro-Jefferson County area in Kentucky, USA. The Climate and Economic Justice Screening Tool (CEJST) designation by census tract as well as federal and state Superfund sites, the international airport with a commercial air-transport hub, two large interstate highways, and the Ohio River on the north are represented. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 8 Data Sources Our methodological approach was based on the HGB region tool1 (Table S1) with customization to reflect Louisville Metro-Jefferson County-relevant indicators. The index includes 32 indicators organized into five domains (Table 1; Table S2). Transportation noise exposure (vehicle, railways, aviation), heat wave exposure, and tornado indicators were added. Some indicators included in the original HGB index were removed for the Jefferson County index as being either not applicable or with no local source data (Table S3). Indicators were derived from source datasets that were publicly available, with source data from 2015 through 2023. For proximity analysis including number of hospitals, Superfund sites, and point sources of pollution, a sum of sites within a 5 km radius buffer from the census tract centroid was used. In most cases, low indicator values reflect low vulnerability. This relationship was inverted for three indicators (life expectancy, number of hospitals, and tree canopy). The baseline health domain contains six indicators pertaining to disease, longevity, and healthcare access to show where residents themselves are most vulnerable to adverse environmental impacts. The environmental exposures and risks domain encompasses nine indicators related to ambient pollution and general environmental hazards. The environmental sources domain features seven indicators, six reflect point sources of environmental exposures and the seventh is tree canopy. The social vulnerability domain includes nine indicators related to the resilience capacity of communities to recover from natural disasters and other crises. The extreme weather domain includes three indicators: flood, heat exposure, and tornados. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 9 Table 1. Domains and vulnerability Indicators in the Environmental Vulnerability Index model for Louisville Metro- Jefferson County, Kentucky (USA), data variance, source dataset publication year, and assessment of if modification is possible by public health intervention or policy. Domain Indicator Original data variance Source data year Modifiable? (yes/no)

Baseline Health Adult asthma (>18 years old) 8.7 – 18.4 2020 No 21 Chronic obstructive pulmonary disease 4.2 – 17.3 2020 No 21 Coronary heart disease 2.2 – 14.0 2020 No 21 Hospitals within 5km radius 0.0 – 6.0 2020 No 22 Life expectancy 63.7 – 86.8 2015 No 23 Stroke 1.3 – 8.7 2020 No 21 Environment al Exposures and Risks Environmental Hazard Index 1 – 62 2023 No 24 National Air Toxics Assessment cancer risk 20 – 80 2019 Yes 25 National Air Toxics Assessment reproductive risk 0.02 – 0.07 2019 Yes 25 National Air Toxics Assessment respiratory risk 0.3 – 0.6 2019 Yes 25 Noise exposure 46.8189 – 60.1937 2020 Yes 26 PM2.5 Community Multiscale Air Quality 5.3892 – 12.949 2016 Yes 27 Risk-screening environmental indicators (RSEI) 0 – 11,311,551 2021 Yes 28 Environment al Sources Cement batch plants within 5km radius 0 – 7 2023 No 29 Metal recyclers within 5km radius 0 – 3 2023 No 29 Petrochemical and oil refineries within 5km radius 0 – 1 2023 No 29 Power plants within 5km radius 0 – 1 2023 No 29 Superfund sites within 5km radius 0 – 1 2023 No 30 Traffic proximity and volume 16.4009 – 8243.5887 2020 No 31 Tree canopy coverage 1.6807 – 68.0316 2021 Yes 32 Social Vulnerability % Income for housing 0 – 0.5498 2023 Yes 33 % Renters 2.6892 – 96.711 2023 Yes 33 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 10 Domain Indicator Original data variance Source data year Modifiable? (yes/no)

% Without health insurance 0.2 – 28.3 2020 Yes 34 Food desert low access 0 – 2 2019 Yes 35 Household composition and disability 0.0072 – 0.9892 2020 No 34 Housing and transportation 0.0018 – 1 2020 No 34 Median renter income vs. median area income 0.3108 – 0.5789 2023 Yes 33 Minority status and language 0.1182 – 1 2020 No 34 Socioeconomic status 0.001 – 1 2020 No 34 Weather 1% Annual probability flood hazarda 0 – 40.3884 2020 Yes 36 Warm season average maximum temperature 86.13766667- 88.39227273 2018 Yes 37 Tornado 0.000912796 – 0.019012543 2023 Yes 38 a24 census tracts have missing data; the Louisville Metro-Jefferson County average was used in these cases. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 11 Data Analysis The Toxicological Prioritization Index (ToxPi) version 1.2.1 (Durham, North Carolina) was used to produce EVI scores by census tract (n=190).3,4,39 Indicators were converted to percentiles to put them on the same scale before being combined within domains. Equal weights were applied to each indicator within a domain, and each domain was weighted equally as one-fifth. Scores are relative rankings; a composite score of 0 indicates that an area has no vulnerabilities while higher scores indicate more vulnerability. The Louisville International Airport comprises an entire census tract and was excluded. Maps, geocoding, and Local Moran’s I analysis were performed using ArcGIS Pro version 2.9.5 (Redlands, CA). Environmental Vulnerability Intervention Simulation To evaluate potential public health interventions for reducing environmental vulnerabilities, modifiable indicators were identified, and three intervention scenarios were developed (Figure 2). Indicators were determined to be modifiable if they were changeable by policy or other intervention activity within five years, for instance – tree planting efforts can increase the percentage of tree canopy and air toxics can be reduced through policy or power plant retirement, retrofit, and conversion to natural gas. 7,40 To define modifiable factors, particular emphasis was placed on their potential for theoretically feasible implementation in real-world settings,41 within the context of census tract-level influence. Non-modifiable indicators included factors such as interstate highways, industrial corridors, rivers, and other natural landscape features. The five most vulnerable census tracts, those with the highest ToxPi composite scores, were selected as intervention sites with three illustrative case studies for each (A, B, C). To model . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 12 interventions, we substituted selected indicator score(s) with the lowest (least vulnerable) score observed in Louisville Metro-Jefferson County. This change represents a theoretical intervention aimed at improving the tract’s overall vulnerability. Intervention A modeled the potential impacts improving singular indicators such as tree canopy, noise pollution exposure, or respiratory risk from air toxics. Intervention B modeled the potential impacts of improving an entire domain such as environmental exposures and risks, environmental sources, or weather; for example, an intervention that included removing a group of point sources of pollution, decreasing traffic and increasing tree canopy coverage through joint economic and policy investment. Intervention C modeled the potential impacts of improving a thematic cluster of indicators, such as air pollution, environmental infrastructure, or point sources of pollution (Table S4). . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 13 Figure 2. Framework for Environmental Vulnerability Index supporting simulation of targeted public health interventions at the census tracts level. Using the Toxicological Prioritization Index composite scores, the five most vulnerable census tracts were identified and modifiable indicators were selected under three intervention scenarios as theoretical action for solutions. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 14 Ethics Data used in the analysis are available in online public records, sources are provided in Table S2. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 15

Composite ToxPi Scores Composite ToxPi scores were not uniformly distributed across Louisville Metro-Jefferson County (Figure 3); they ranged from 0.19 to 0.71. Geographically, high vulnerability scores were observed along the western and central areas of the county with lower vulnerability scores in the central urban core and eastern regions. The most at-risk census tract (21111005900) is in the downtown urban core. The least at-risk census tract (21111013100) is a mainly residential area in the urban core that abuts a park and a local airfield. However, there are pockets of vulnerability and census tracts that span geographic locations. The five least vulnerable census tracts are spread across an area east of Interstate 65. In contrast, the five most vulnerable census tracts are spread across the central and western areas of the county. This pattern matches the distribution of CEJST-designation of disadvantaged. 20 Of the 72 disadvantaged census tracts in Jefferson County 83% (70) are in the western and south-central regions. Of the 21 most vulnerable census tracts, 80% (17) are considered disadvantaged. All five of the most vulnerable census tracts from our model are also considered CEJST disadvantaged. Local Moran’s I analysis was used to identify statistically significant clusters and outliers. Clusters are areas where tracts with high or low ToxPi scores are adjacent to other tracts with similarly high or low scores. Outliers are areas where tracts with high values are adjacent to low values and vice-versa (Figure 3). Results show significant clustering of higher ToxiPi scores in the west, with low ToxiPi scores cluster in the eastern county areas. There are five outlier tracts with relatively higher ToxPi scores than their neighboring tracts in the eastern county areas. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 16 A B Figure 3. Toxicological Prioritization Composite Score Index by census tract, Louisville Metro-Jefferson County, Kentucky (USA). Panel A: Composite scores by census tract whereby a higher score corresponds to a more vulnerable census tract. Census tracts were sorted into quartiles by overall score, the quartiles were scored as: 1st quartile as lowest vulnerable census tracts 0.52 (n=47). Panel B: Modeled clusters of composite scores by census tract. The Local Moran’s I analysis indicates statistically significant clusters of high or low values, as well as outliers where high values are adjacent to low values, and vice-versa. Domain-specific Scores Some census tracts with low composite ToxPi scores have high domain-specific scores (Figure 4). The five most vulnerable census tracts especially score low in the domains of environmental exposures and risks and environmental exposures. The least vulnerable census tract (21111013100) has the best score across the county for both for social vulnerability and severe weather domains. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 17 Top Bottom Figure 4. Toxicological Prioritization Domain Score Index by census tract, Louisville Metro-Jefferson County, Kentucky (USA). Top: Domain scores by census tract whereby a higher score corresponds to a more vulnerable census tract. Panels present the domain scores for: A.) Baseline Health; B.) Environmental Exposures and Risks; C.) Environmental Sources; D.) Social Vulnerability; and E.) Extreme Weather. Bottom: Modeled clusters of domain scores by census tract. The Local Moran’s I analysis indicates statistically significant clusters of high or . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 18 low values, as well as outliers where high values are adjacent to low values, and vice-versa. Panels present the domain scores for: A.) Baseline Health; B.) Environmental Exposures and Risks; C.) Environmental Sources; D.) Social Vulnerability; and E.) Extreme Weather. Heterogeneity Heterogeneity analysis (Figure S1) was performed for each for each of the 32 indicators to ensure only factors that offer significant variance were included in the index. Indicators that have limited variation thus may have minimal influence on the model's predictive accuracy. Heat exposure and noise are two examples with a good, but narrow, data range across census tracts. As well, for NATA cancer, most (186/190) census tracts have the same raw value, but there is some variation. And, again for renter/owner income most (185/190) census tracts have the same raw value. If the framework indicated an indicator raw score that had no-variation it would have been appropriate to remove; no changes were made following heterogeneity analysis. Modelling the Impact of Interventions on Vulnerability Scores Across the Intervention A case studies, modifications to a single indicator led to enhancements in the ToxPi composite score ranging from 1 to 4% (Table 2). Improving single indicators does not universally reduce vulnerability, even in the most vulnerable tracts. Improving the best possible indicator score reflected in the county for tree canopy alone resulted in an average 3% improvement to the composite ToxPi score, while improving noise pollution or NATA respiratory risk were an average 2% improvement. Currently tree canopy coverage is as low as 5% and adjusting tree canopy in these five census tracts uniformly to 10% only improved composite ToxPi scores marginally, an average of 1% improvement. Increasing tree canopy to the countywide goal of 45% 42 improved average composite ToxPi scores by 3%. One of the most . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 19 vulnerable census tracts (21111011901) already had a 31% tree canopy coverage, which is good for an urban area and better than the average tree canopy coverage across the county. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 20 Table 2. Toxicological Prioritization Index (ToxPi) composite scores for the lowest five ranking census tracts and modeled case study result following intervention to the best possible indicator score reflected in the county. Higher composite scores correspond to higher vulnerability. Intervention A modeled the potential impacts of improving one indicator while keeping all other indicators static. Intervention B modeled the potential impacts of improving entire domains. Intervention C modeled the potential impacts of improving clusters of indicators with related interventions. Each model intervention was conducted three times as an illustrative case studt. ToxPi Score after Intervention A ToxPi Score after Intervention B ToxPi Score after Intervention C Census Tract Original ToxPi Composit e Score A1 – Tree canopy A2 – Noise A3 – NATA respiratory risk B1 – Environmen tal exposure & risks B2 – Environme ntal Sources B3 – Weather C1 – Air quality C2 – Environme ntal infrastruct ure C3 – Potential health hazards 21111005900 0.7077 0.68 0.6862 0.6941 0.6199 0.5695 0.6029 0.6397 0.6533 0.6254 21111009103 0.6831 0.6559 0.6618 0.6695 0.6025 0.5699 0.5194 0.6272 0.6509 0.5986 21111011002 0.6497 0.6233 0.6416 0.6361 0.574 0.5641 0.5217 0.5928 0.6058 0.5973 21111011901 0.6594 0.6548 0.6325 0.6458 0.5955 0.5948 0.4958 0.6127 0.6532 0.6105 21111012701 0.7 0.676 0.6929 0.6864 0.6498 0.546 0.5734 0.6657 0.6477 0.5894 Logistically feasible public health intervention Tree planting efforts inclusive of site selection and preparation, long term maintenance and community engagement, and protective policy. Noise pollution reduction achieved through noise reduction features (sound walls, vegetative buffers, etc.) and policy (elevation standards for aircraft and residential noise regulations.) Air quality improveme nts through local clean air standards, pollution control requirement s, best available technologie s, and compliance monitoring and reporting A combination of air monitoring and noise exposure interventions . Economic and policy investment to remove or reduce the number of environment al sources of pollution from residential areas; and increasing tree coverage. Wetland restoration and green infrastructur e; upgrading drainage systems and critical infrastructur e; integration of flood management into broader environment al planning. A combinatio n of suggested intervention s from each of the other categories. A combination of suggested intervention s from each of the other categories. A combination of suggested interventions from each of the other categories. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 21 For Intervention B, entire domains with several indicators were adjusted: environmental exposures and risks, environmental sources, or weather. Across the case studies, domain modifications led to enhancements in the ToxPi composite scores ranging from 7% to 25%. When indicator scores within a domain were adjusted to the best scores observed in the county, the census tract’s composite ToxPi scores were reduced by an average of 11% for environmental exposures and risks, 16% for environmental sources, and 20% for extreme weather. For Intervention C, thematically related intervention clusters of indicators were adjusted: air pollution, environmental infrastructure, or potential health hazards. Across the case studies, modifications to these clusters led to enhancements in the ToxPi composite scores ranging from 1 to 16%. When indicator scores for each cluster were reduced to the best scores observed in the county, the census tract’s composite ToxPi scores were reduced on average for air quality at 8%, for environmental infrastructure at 6%, and for potential health hazards at 11%. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 22

In this study, we used a modified version of community based EVI framework1 to assess vulnerabilities in a new geographic region. We extended the tool to model interventions and to assess potential impact on hyperlocal risks. Our approach provides a data-driven guide for Louisville Metro-Jefferson County as examples of census tract scale solutions as opposed to more general public health programming. The index is customized to the particular concerns of our area, including the addition of indicators such as noise pollution exposure, heatwave exposure, and tornados. While Messer et al. 43 developed neighborhood socioeconomic context for 19 cities, it excluded flood, growing urban heat island effect, and tornados which are locally important vulnerabilities in the Louisville Metro-Jefferson County area. The presence of a commercial air-hub led to the addition of the noise pollution indicator. These risks are real for many cities and the expanded evaluation demonstrates the benefits of adaptation to local conditions, risks, and vulnerabilities. The Louisville Metro-Jefferson County has unique vulnerabilities when compared with most other counties across of the nation, although the area does resemble several towns in the United States Midwest. For example, all 190 census tracts in Louisville Metro-Jefferson County have higher average PM 2.5 concentrations than the nationwide average of 3.79 µg/m3. Identification and incorporation of such local vulnerabilities may be key to developing local and well-targeted interventions. Previous work suggests that single-indicator interventions to improve health are theoretically feasible. Wang et al. 40 reported on the effectiveness of an urban green and blue space intervention in improving wellbeing. Brown et al.44 and Hammer et al.45 discovered that direct regulation on lowering noise at its source, expanding and increasing access to noise maps, and . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 23 altering the built environment can be effective. Further, Casey et al.7 documented improved health outcomes resulting from the removal of an environmental point source of pollution when the retirement of a coal-fired power plant led to improved asthma outcomes. Our index integrates several related indicators into a single domain and it incorporates social, clinical, and environmental data to provide a more comprehensive evaluation of area vulnerability. In previous work, other environmental health vulnerability studies conducted at census tract level have focused on only clinical data or only a few public data sources, 40,46 despite the probability that a wide range of structural environmental variables can impact health. For instance, social vulnerability combines data on socioeconomic status, housing and transportation, and health insurance coverage. When assessing access to healthcare, this combined view of indicators may be more explanatory of lack of healthcare access than a simple hospital proximity indicator. Our findings suggest that environmental vulnerability could be feasibly estimated at census tract scale, rather than county level, consistent with the Glassman et al. 5 report. In addition, our work demonstrates that county-level data could be too broad to support intervention evaluation. Census tract level data offers balanced spatial granularity useful to discern hyperlocal variations within urban areas whereby environmental risks can be viewed within the context of the entire spectrum of risks. The index enabled the selection of the most at-risk census tracts for modeling targeted public health interventions to reduce cumulative environmental vulnerability within specific . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 24 communities. Our work also suggests that while some census tracts may require plans tailored to those specific areas, it may be possible to develop interventions effective in addressing common problems that span boundaries across multiple tracts. Moreover, our work across the three case studies also suggests that improving any single indicator may not be highly beneficial in addressing the overall area vulnerability and that it may be necessary to address multiple indicators collectively. Interesting, a high composite vulnerability scores can mask low indicator- level vulnerabilities (e.g., good tree coverage). Overall, our index and our method for computing aggregate vulnerabilities maybe helpful in identifying focal points for resource allocation. However, additional work will be required to assess the efficacy of any intervention. For instance, Louisville’s 14 hospitals are clustered in one geographic area thus limiting access, as measured by geographical proximity, for many census tracts. Although this proximity indicator reflects an equal impact geographically, it may result in a stronger impact on individuals at the low end of the socioeconomic spectrum compared to those at the highest; affluent individuals have significant resources, personal transportation, and insurance to access healthcare, regardless of distance to the nearest hospital. Thus, a healthcare access intervention may be warranted for low-income census tracts to offset this inequity. Moreover, in our area of interest seven census tracts are within a 5-kilometer radius of Superfund sites, and all of these are west of a structural environment variable, Interstate 65, mirroring larger vulnerability patterns in the county. This suggests purposively modifying some indicators may have a disproportionately positive impact on the county’s most vulnerable census tracts. Therefore, as indicated by these two examples, several indicators are highly correlated and therefore changes in one could have far reaching effect on the other or may benefit only related aspects of those vulnerabilities. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 25 Threshold values across both the composite score and indicators can be used to manage expectations related to interventions. More affluent census tracts in Jefferson County have ToxPi composite scores of less than 0.3. These areas are often characterized by high life expectancy, low social vulnerability, and limited pollution exposure. As well per a single indicator, while 40- 45% is a goal set by Louisville’s Urban Tree Canopy Assessment 42 the lowest tree canopy score in our study area is 2%, and any short-term aim to increase the canopy indicator to 40% may be a highly unrealistic goal. When we modeled the 45% canopy coverage, it only improved the ToxPi composite score to be 3% better. These results put in perspective the expected indicator gains that could be accrued by specific interventions and should help in prioritizing interventions. Moreover, some indicator scores may not change without a modification in direct policy and/or financial intervention, such as the presence of industrial facilities. Finally, when selecting areas for an intervention, it may be important to consider ethical questions around withholding environmental health interventions known to be effective. 47 The main purport of the framework developed by our work may be to guide actions to mitigate vulnerability and risk. For example, local governance could use this framework to create action plans for increasing neighborhood resilience and preparedness following major events. Local public health agencies could use also digital platforms and targeted advertising campaigns to raise awareness and promote health actions in specific places. 48 Qualitative researchers could use the vulnerability index to gather community input to validate proposed solutions. Finally, grassroots community organizations could use an assessment of environmental vulnerability to identify problems, construct potential solutions, and advocate for policy change. Establishing a data-driven index to inform interventions tailored to specific needs, rather than generic . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 26 approaches would also ensure accountability by verifying that the interventions are directed towards areas where they are most needed and address the area’s most significant vulnerabilities. Overall, we believe our study advances the integrated vulnerability index approach for selecting census tracts for health interventions in several key ways: 1. Consideration of multiple domains and indicators . Consideration of health parameters beyond traditional clinical and environmental data can inform more effective, data-driven, public health communications, and interventions. 2. Data anchored to census tract boundaries . Focus on census tract level vulnerability rather than countywide scale. 3. Prioritization of intervention and investment by data-driven vulnerability . Our approach shows which census tracts are in most need of vulnerability mitigation and intervention. In some cases, the same intervention may benefit multiple census tracts even if they are not geographically adjacent. 4. Shift focus away from the single indicator interventions . Modeled results show improving a group of related indicators is more impactful than adjusting a single indicator. Despite its many strengths, the study has some limitations. We relied solely on publicly available national, state, and local government data sources, potentially overlooking data requiring Data Transfer Agreements or open records requests. Application may be limited for rural census tracts due to a lack of local data available. We restricted our analysis to Louisville Metro-Jefferson County, neglecting the bordering counties of the metro area. All domains are equally weighted as we could not justify indicator-specific weighting. Qualitative research and community input can . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 27 be used in tools to validate vulnerability frameworks; however, integration of qualitative and quantitative data sources can pose additional challenges.49 Our modeled interventions have limitations; determining efficacy of intervention requires empirical testing to establish the degree to which environmental vulnerabilities are modifiable. One major limitation of our approach is that scores are percentile-based which does not distinguish between indicators with low and high variance. However, we addressed this limitation by selecting indicators for interventions that could feasibly produce a material change, such as percent tree canopy, achievable in real-world conditions. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 28

In this study we used census tract boundaries and integrated vulnerability data to develop a comprehensive vulnerability index which could be used to assess the potential impact of targeted public health interventions. While the original framework was designed to assign relative ranks based on environmental vulnerability, the aggregation of vulnerability and intervention allows the index to develop into a more comprehensive tool for public health, community planning, and environmental justice advocates. Chief among our insights is that places are not defined by their environmental vulnerabilities and that such vulnerabilities are not invariant features of living communities. Rather, knowledge of these risks can spur action for towards local and bespoke solutions. Targeted census tract interventions may be more effective than broad-scale campaigns at state or national levels and may lead to health-supportive environmental conditions with greater geographic precision. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 29 Contributors Conceptualization: TS, WC, AB; Methodology: TS, AB; Formal analysis: LBA, CB, DJB, RHH; Writing-original draft preparation: LBA, RHH; Writing-review and editing: TS, LBA, CB, DJB, WC, RHH, AB; Supervision: TS, AB; Project administration: LBA. All the authors have read and agreed to the published version of the manuscript. Data sharing All source data are publicly available, links are provided in Table S2. Funding This work was supported in part by the Owsley Brown II Family foundation, NIEHS (P42 ES023716 to the University of Louisville; and the P30 ES029067 and P42 ES027704 to Texas A&M University) and the Robert Wood Johnson Foundation (Grant #: 80565). Acknowledgments The authors thank Ray Yeager, Rebecca Turney, and Angelina Rangel for their mapping expertise, Dhiraj Kanneganti for ToxPi skills developed during this project, and Laney Taylor for gathering and organizing the initial data for this research. Declaration of interests The authors declare they have nothing to disclose. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 30

1. Bhandari S, Lewis PGT, Craft E, Marvel SW, Reif DM, Chiu WA. 2020. HGBEnviroScreen: enabling community action through data integration in the Houston– Galveston–Brazoria region. Int J Env Res Pub He 17(4):1130. https://doi.org/10.3390/ijerph17041130 2. Lewis PG, Chiu WA, Nasser E, Proville J, Barone A, Danforth C, et al. 2023. Characterizing vulnerabilities to climate change across the United States. Environ Int 172:107772. https://doi.org/10.1016/j.envint.2023.107772 3. Marvel SW, To K, Grimm FA, Wright FA, Rusyn I, Reif DM. 2018. ToxPi graphical user interface 2.0: Dynamic exploration, visualization, and sharing of integrated data models. BMC Bioinform 19:80. https://doi.org/10.1186/s12859-018-2089-2 4. Reif DM, Sypa M, Lock EF, Wright FA, Wilson A, Cathey T, et al. 2013. ToxPi GUI: an interactive visualization tool for transparent integration of data from diverse sources of evidence. Bioinformatics 29(3):402–3. https://doi.org/10.1093/bioinformatics/bts686 5. Glassman B. 2020. The multidimensional deprivation index using different neighborhood quality definitions. United States Census Bureau Social, Economic, and Housing Statistics Division. https://www.census.gov/content/dam/Census/library/working- papers/2020/demo/SEHSD-WP2020-08.pdf [accessed 24 April 2024]. 6. Centers for Disease Control and Prevention. 2024. National Notifiable Diseases Surveillance System (NNDSS). https://www.cdc.gov/nndss/index.html [accessed 24 April 2024]. 7. Casey JA, Su JG, Henneman LR, Zigler C, Neophytou AM, Catalano R, et al. 2020. Improved asthma outcomes observed in the vicinity of coal power plant retirement, retrofit and conversion to natural gas. Nature Energy 5(5):398–408. https://doi.org/10.1038/s41560-020-0600-2 8. Hilmers A, Hilmers DC, Dave J. 2012. Neighborhood disparities in access to healthy foods and their effects on environmental justice. Am J Public Health 102(9):1644–54. https://doi.org/10.2105/AJPH.2012.300865 9. Gochfeld M, Burger J. 2011. Disproportionate exposures in environmental justice and other populations: the importance of outliers. Am J Public Health 101(S1):S53–63. https://doi.org/10.2105/AJPH.2011.300121 10. Johnson DP, Stanforth A, Lulla V, Luber G. 2012. Developing an applied extreme heat vulnerability index utilizing socioeconomic and environmental data. Appl Geogr 35(1– 2):23–31. https://doi.org/10.1016/j.apgeog.2012.04.006 11. Marvel SW, House JS, Wheeler M, Song K, Zhou YH, Wright FA, et al. 2021. The COVID-19 Pandemic Vulnerability Index (PVI) Dashboard: monitoring county-level vulnerability using visualization, statistical modeling, and machine learning. Environ Health Perspect 29(1):017701. https://doi.org/10.1289/EHP8690 12. Nasiri H, Yusof MJM, Ali TAM, Hussein MKB. 2019. District flood vulnerability index: urban decision-making tool. Int J Environ Sci Te 16(5):2249–58. https://doi.org/10.1007/s13762-018-1797-5 13. Newman G, Malecha M, Atoba K. 2023. Integrating ToxPi outputs with ArcGIS Dashboards to identify neighborhood threat levels of contaminant transferal during flood events. J Spat 68(1):57–69. https://doi.org/10.1080/14498596.2021.1891149 14. Xue J, Zartarian V, Tornero-Velez R, Stanek LW, Poulakos A, Walts A, et al. 2022. A generalizable evaluated approach, applying advanced geospatial statistical methods, to . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 31 identify high lead exposure locations at census tract scale: Michigan case study. Environ Health Persp 130(7):077004. https://doi.org/10.1289/EHP9705 15. World Health Organization. 2000. Considerations in evaluating the cost-effectiveness of environmental health interventions. https://www.who.int/publications/i/item/WHO-SDE- WSH-00.10 [accessed 9 April 2024]. 16. Bouzid M, Hooper L, Hunter PR. 2013. The effectiveness of public health interventions to reduce the health impact of climate change: a systematic review of systematic reviews. PLoS One 8(4):e62041. https://doi.org/10.1371/journal.pone.0062041 17. Environmental Defense Fund & Texas A&M University. 2023. The U.S. Climate Vulnerability Index. https://climatevulnerabilityindex.org/ [accessed 21 May 2024]. 18. United States Census. Jefferson County, Kentucky. Population, Census, April 1, 2020. https://www.census.gov/quickfacts/fact/table/jeffersoncountykentucky/POP010220 [accessed 24 April 2024]. 19. University of Kentucky. 2023. Kentucky Annual Economic Report 2023. https://cber.uky.edu/sites/cber/files/publications/UK%20CBER%20Kentucky%20Annual %20Economic%20Report%202023_Web.pdf [accessed 3 May 2024]. 20. United States Climate Resilience Toolkit. The Climate and Economic Justice Screening Tool (CEJST) https://toolkit.climate.gov/tool/climate-and-economic-justice-screening- tool [accessed 9 April 2024]. 21. Centers for Disease Control and Prevention. 2021. Places: Census Tract Data – GIS Friendly Format https://chronicdata.cdc.gov/500-Cities-Places/PLACES-Census-Tract- Data-GIS-Friendly-Format-2021-/yjkw-uj5s/data [accessed 1 Nov 2023]. 22. Kentucky Cabinet for Health and Family Services. Division of Health Care Facilities. https://chfs.ky.gov/agencies/os/oig/dhc/Pages/hcf.aspx [accessed 2 November 2023]. 23. National Center for Health Statistics. 2018. U.S. Small-Area Life Expectancy Estimates Project (USALEEP): Life Expectancy Estimates File for Kentucky, 2010-2015. https://www.cdc.gov/nchs/nvss/usaleep/usaleep.html [accessed 1 November 2023]. 24. United States Department of Housing and Urban Development. 2023. Environmental Health Hazard Index. https://data.lojic.org/datasets/HUD::environmental-health-hazard- index/about [accessed 9 November 2023]. 25. United States Environmental Protection Agency. 2020. Chemical concentrations, exposures, health risks by census tract from National Scale Air Toxics Assessment (NATA). https://catalog.data.gov/dataset/chemical-concentrations-exposures-health- risks-by-census-tract-from-national-scale-air-toxics [accessed 9 November 2023]. 26. United States Department of Transportation. 2020. National Transportation Noise Map. ttps://www.bts.gov/geospatial/national-transportation-noise-map [accessed 2 November 2023]. 27. United States Environmental Protection Agency. 2021. Daily Census Tract-Level PM2.5 Concentrations, 2016 - 2020 https://data.cdc.gov/Environmental-Health- Toxicology/Daily-Census-Tract-Level-PM2-5-Concentrations-2016/96sd- hxdt/about_data [accessed 8 November 2023]. 28. United States Environmental Protection Agency. 2021. Risk-Screening Environmental Indicators (RSEI) Model RSEI Version 2.3.11 (RY 2021). https://edap.epa.gov/public/extensions/EasyRSEI/EasyRSEI.html [accessed 2 November 2023]. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 32 29. United States Environmental Protection Agency. 2023. Facility Registry Service Facilities State Single File CSV Download. https://www.epa.gov/frs/epa-frs-facilities- state-single-file-csv-download [accessed 16 November 2023]. 30. Kentucky Energy and Environmental Cabinet. Superfund Site List. https://dep.gateway.ky.gov/eSearch/AvailableReports/Details?id=47 [accessed 14 November 2023]. 31. United States Environmental Protection Agency. 2021a. EJScreen: Environmental Justice Screening and Mapping Tool. Traffic proximity and volume dataset. https://gaftp.epa.gov/EJScreen/ [accessed 16 November 2023]. 32. United States Geological Survey, U.S. Department of the Interior, and U.S. Forest Service. 2021. Tree Canopy Cover of the Conterminous United States, 2021 NLCD 2021 USFS Tree Canopy Cover (CONUS). Multi-Resolution Land Characteristics (MRLC) Consortium. https://www.mrlc.gov/data/nlcd-2021-usfs-tree-canopy-cover-conus [accessed 11 November 2023]. 33. United States Department of Housing and Urban Development. 2023. Location Affordability Index (Version 3). https://data.lojic.org/datasets/HUD::location- affordability-index-v-3/about [accessed 17 November 2023]. 34. Centers for Disease Control and Prevention. 2020. Agency for Toxic Substances and Disease Registry/ Geospatial Research, Analysis, and Services Program. CDC/ATSDR Social Vulnerability Index, Database Kentucky. https://www.atsdr.cdc.gov/placeandhealth/svi/data_documentation_download.html. Accessed on 20 November 2023. 35. United States Department of Agriculture, Economic Research Service. 2019. Food Access Research Atlas. https://www.ers.usda.gov/data-products/food-access-research- atlas/download-the-data/ [accessed 17 November 2023]. 36. Esri. 2020. USA_Flood_Hazard_Reduced_Set_gdb. ArcGIS Online Map Viewer. https://www.arcgis.com/apps/mapviewer/index.html?layers=bf9585afc2934b648f391869 3285b7c8 [accessed 20 November 2023]. 37. Jefferson County Information Consortium. 2022. Jefferson County Kentucky Urban Heat Management Study (LOJIC). https://hub.arcgis.com/datasets/LOJIC::jefferson-county- ky-urban-heat-management-study [accessed 8 May 2024]. 38. Federal Emergency Management Agency. 2023. National Risk Index (NRI) Data & Resources. https://hazards.fema.gov/nri/data-resources [accessed 23 September 2023]. 39. Toxicological Prioritization Index (ToxPi) version 1.2.1 (https://CRAN.R- project.org/package=toxpiR ) 40. Wang R, Browning MH, Kee F, Hunter RF. 2023. Exploring mechanistic pathways linking urban green and blue space to mental wellbeing before and after urban regeneration of a greenway: Evidence from the Connswater Community Greenway, Belfast, UK. Landscape Urban Plan 235:104739. https://doi.org/10.1016/j.landurbplan.2023.104739 41. O'Cathain A, Croot L, Duncan E, Rousseau N, Sworn K, Turner KM, et al. 2019. Guidance on how to develop complex interventions to improve health and healthcare. BMJ Open 9(8):e029954. http://dx.doi.org/10.1136/bmjopen-2019-02995 42. Louisville Metro-Jefferson County Government. Louisville Urban Tree Canopy Assessment, 2015. https://louisvilleky.gov/urban-forestry/document/louisville-urban-tree- canopy-assessment-2015 [accessed 9 April 2024]. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 33 43. Messer LC, Laraia BA, Kaufman JS, Eyster J, Holzman C, Culhane J, et al. 2006. The development of a standardized neighborhood deprivation index. J Urban Health 83(6):1041–62. https://doi.org/10.1007/s11524-006-9094-x 44. Brown AL, Van Kamp I. 2017. WHO environmental noise guidelines for the European Region: A systematic review of transport noise interventions and their impacts on health. Int J Env Res Pub He 14(8):873. https://doi.org/10.3390/ijerph14080873 45. Hammer MS, Swinburn TK, Neitzel RL. 2014. Environmental noise pollution in the United States: developing an effective public health response. Environ Health Perspect 122(2):115–9. http://dx.doi.org/10.1289/ehp.1307272. 46. Uong SP, Zhou J, Lovinsky-Desir S, Albrecht SS, Azan A, Chambers EC, et al. 2023. The creation of a multidomain neighborhood environmental vulnerability index across New York City. J Urban Health 100(5):1007–23. https://doi.org/10.1007/s11524-023- 00766-3 47. Resnik DB, Zeldin DC, Sharp RR. 2005. Research on environmental health interventions: ethical problems and solutions. Accountability Res 12(2):69–101. https://doi.org/10.1080/08989620590957157 48. Anderson LB, Ness HD, Holm RH, Smith T. 2024. Wastewater-informed digital advertising as a Covid-19 geotargeted neighborhood intervention: Jefferson County, Kentucky, 2021–2022. Am J Public Health 114(1):34–7. https://doi.org/10.2105/AJPH.2023.307439 49. Ho HC, Wong MS, Man HY, Shi Y, Abbas S. 2019. Neighborhood-based subjective environmental vulnerability index for community health assessment: development, validation and evaluation. Sci Total Environ 654:1082–90. https://doi.org/10.1016/j.scitotenv.2018.11.136 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 34 Figure Legends Figure 1. Study area, Louisville Metro-Jefferson County area in Kentucky, USA. The Climate and Economic Justice Screening Tool (CEJST) designation by census tract as well as federal and state Superfund sites, the international airport with a commercial air-transport hub, two large interstate highways, and the Ohio River on the north are represented. Figure 2. Framework for Environmental Vulnerability Index supporting simulation of targeted public health interventions at the census tracts level. Using the Toxicological Prioritization Index composite scores, the five most vulnerable census tracts were identified and modifiable indicators were selected under three intervention scenarios as theoretical action for solutions. Figure 3. Toxicological Prioritization Composite Score Index by census tract, Louisville Metro-Jefferson County, Kentucky (USA). Panel A: Composite scores by census tract whereby a higher score corresponds to a more vulnerable census tract. Census tracts were sorted into quartiles by overall score, the quartiles were scored as: 1 st quartile as lowest vulnerable census tracts 0.52 (n=47). Panel B: Modeled clusters of composite scores by census tract. The Local Moran’s I analysis indicates statistically significant clusters of high or low values, as well as outliers where high values are adjacent to low values, and vice-versa. Figure 4. Toxicological Prioritization Domain Score Index by census tract, Louisville Metro-Jefferson County, Kentucky (USA). Top: Domain scores by census tract whereby a higher score corresponds to a more vulnerable census tract. Panels present the domain scores for: A.) Baseline Health; B.) Environmental Exposures and Risks; C.) Environmental Sources; D.) Social Vulnerability; and E.) Extreme Weather. Bottom: Modeled clusters of domain scores by census tract. The Local Moran’s I analysis indicates statistically significant clusters of high or low values, as well as outliers where high values are adjacent to low values, and vice-versa. Panels present the domain scores for: A.) Baseline Health; B.) Environmental Exposures and Risks; C.) Environmental Sources; D.) Social Vulnerability; and E.) Extreme Weather. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 35 Supplementary Information An Environmental Vulnerability Index framework supporting targeted public health interventions at the census tracts level Lauren B. Anderson 1,2*, Rochelle H. Holm1*, Caison Black1, Donald J. Biddle3, Weihsueh A. Chiu4, Aruni Bhatnagar5, and Ted Smith1 1Center for Healthy Air, Water and Soil, Christina Lee Brown Envirome Institute, School of Medicine, University of Louisville, Louisville, KY, USA 2Department of Urban and Public Affairs, College of Arts & Sciences, University of Louisville, Louisville, KY, USA 3Department of Geographic and Environmental Sciences, Center for Geographic Information Sciences, University of Louisville, Louisville, KY, USA 4Department of Veterinary Physiology and Pharmacology, College of Veterinary Medicine and Biomedical Sciences, Texas A&M University, College Station, TX, USA 5Christina Lee Brown Envirome Institute, School of Medicine, University of Louisville, Louisville, KY, USA . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 36 Table of Contents Table S1. Vulnerabilities of Houston–Galveston–Brazoria region, Texas, and Louisville Metro- Jefferson County, Kentucky. ................................................................................................... .... 37 Table S2. Data sources of indicators incorporated into Environmental Vulnerability Index for Louisville Metro- Jefferson County, Kentucky (USA). ............................................................... 38 Table S3. Indicators included in the Houston–Galveston–Brazoria (HGB) EnviroScreen’s Environmental Vulnerability Index (EVI) (Bhandari et al., 2020) which were removed for the Louisville Metro-Jefferson County index. ................................................................................... 42 Table S4. Indicators subgrouped based upon related impacts across domains for Intervention C case studies. ................................................................................................................. ............... 43 Figure S1. Heterogeneity analysis for the 32 indicators. Performed for each indicators to ensure only factors that offer significant variance were included in the index (32 panels). .................... 44 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 37 Table S1. Vulnerabilities of Houston–Galveston–Brazoria region, Texas, and Louisville Metro- Jefferson County, Kentucky. Vulnerability Houston– Galveston– Brazoria region, Texas Louisville Metro- Jefferson County, Kentucky Data Source COVID-19 impact 1,547,659 cases 13586 deaths 289,020 cases 2263 deaths https://www.nytimes.com/ Population 5,453,858 773,399 https://data.census.gov/ Area 3,446.2 square miles 380.6 square miles https://data.census.gov/ Large industrial complexes Yes Yes Poverty rate 7.6%-16.4% 14.0% https://data.census.gov/ Natural environment Along the Gulf of Mexico Ohio River Research focused higher-education institution Rice University Baylor College of Medicine University of Louisville . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 38 Table S2. Data sources of indicators incorporated into Environmental Vulnerability Index for Louisville Metro- Jefferson County, Kentucky (USA). Domain Indicator Reference Data Description (File names and filter applied as applicable) Access Date (MM-DD- YEAR) Baseline Healtha Adult asthma (>18 years old) 21 PLACES: Census Tract Data: Column C, "Current asthma among adults aged >=18 years" 11-01-2023 Chronic obstructive pulmonary disease 21 PLACES: Census Tract Data: Column C, "Chronic obstructive pulmonary disease among adults aged >=18 years" 11-01-2023 Coronary heart disease 21 PLACES: Census Tract Data: Column C, "Coronary heart disease among adults aged >=18 years" 11-01-2023 Hospitals within 5km radius 22 Hospital Directory By County: Columns C/D/E entered into ArcGIS Pro to calculate 5km radius 11-02-2023 Life expectancy 23 Kentucky Life Expectancy: Column E, "Life Expectancy in years" 11-01-2023 Stroke 21 PLACES: Census Tract Data: Column C, "Stroke among adults aged >=18 years" 11-01-2023 Environmental Exposures and Risksb Environmental Hazard Index 24 Hazard Index: Column D, “Haz_IDX” 11-09-2023 National Air Toxics Assessment cancer risk 25 NATA 2019 National: Sheet 2, Column G, “Total Cancer Risk (Per million)” 11-09-2023 National Air Toxics Assessment reproductive risk 25 NATA 2019 National: Column G, “Reproductive Health” 11-09-2023 National Air Toxics Assessment respiratory risk 25 NATA 2019 National: Column C, “Respiratory Health” 11-09-2023 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 39 Domain Indicator Reference Data Description (File names and filter applied as applicable) Access Date (MM-DD- YEAR) Noise exposure 26 NoiseToCensus_JeffCo: Column B, “MEAN_Noise_dB” 11-02-2023 PM2.5 Community Multiscale Air Quality 27 Daily Census Tract Level PM2.5 Concentration: Column H, “DS_PM_pred” 11-08-2023 Risk-screening environmental indicators (RSEI) 28 EPA Easy RSEI Dashboard: Column F, “RSEI Score” 11-02-2023 Environmental Sourcesc Cement batch plants within 5km radius 29 EPA FRS Facilities State Single File: SIC codes 3273,3272,3271,2951. Entered into ArcGIS Pro to locate 5 km radius of census tract. 11-16-2023 Metal recyclers within 5km radius 29 EPA FRS Facilities State Single File: SIC code 5093. Entered into ArcGIS Pro to locate 5 km radius of census tract. 11-16-2023 Petrochemical and oil refineries within 5km radius 29 EPA FRS Facilities State Single File: SIC code 2911. Entered into ArcGIS Pro to locate 5 km radius of census tract. 11-16-2023 Power plants within 5km radius 29 EPA FRS Facilities State Single File: SIC code 4911,3612. Entered into ArcGIS Pro to locate 5 km radius of census tract. 11-16-2023 Superfund sites within 5km radius 30 EPA Superfund Sites: Column B,C,D,E. Entered into ArcGIS Pro to locate 5 km radius of census tract. 11-14-2023 Traffic proximity and volume 31 EPA EJScreen: PTRAF calculated by Annual Average Daily Traffic divided by distance in meters 11-16-2023 Tree canopy coverage 32 NLCD 2021 USFS Tree Canopy: Entered into ArcGIS Pro to locate 5 km radius of census tract. 11-11-2023 Social Vulnerabilityd % Income for housing 33 Location Affordability Index v.3: column R"avg_h_cost"X12/column AC "median_hh_income" 11-17-2023 % Renters 33 Location Affordability Index v.3: column K, “pct_renters” 11-17-2023 % Without health insurance 34 CDC/ATSDR SVI Data: Column DR, “EP_UNINSUR” 11-17-2023 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 40 Domain Indicator Reference Data Description (File names and filter applied as applicable) Access Date (MM-DD- YEAR) Food desert low access 35 FoodDesert, FoodAccessResearchAtlasData:A sum of Column O “LowIncomeTracts” and Column U “LATracts_half” 11-17-2023 Household composition and disability 34 CDC/ATSDR SVI Data: Column CH, “RPL_THEME2” 11-17-2023 Housing and transportation 34 CDC/ATSDR SVI Data: Column CS, “RPL_THEME4” 11-17-2023 Median renter income vs. median area income 33 Location Affordability Index v.3: column AB, “area_income_renter_frac” 11-17-2023 Minority status and language 34 CDC/ATSDR SVI Data: Column CL, “RPL_THEME3” 11-17-2023 Socioeconomic status 34 CDC/ATSDR SVI Data: Column CB, “RPL_THEME1” 11-17-2023 Weathere 1% Annual probability flood hazard 36 ESRI Living Atlas USA Flood Hazard Reduced Set: Column B, “1% Flood Hazard Percent Area Coverage” 11-20-2023 Warm season average maximum temperature 37 Jeff Co KY Urban Heat Management Study: Column I, “WS_B_APMAX ” 05-08-2024 Tornado 38 National Risk Index: Column MH, “Tornado Annualized Frequency” 04-23-2024 a The Baseline Health domain contains six indicators related to disease, longevity, and health care access to show where residents themselves are more vulnerable to environmental impacts. CDC Places data for Jefferson County was isolated and downloaded by census tract. Disease prevalence estimates for asthma, chronic obstructive pulmonary disease (COPD), stroke, and coronary heart disease (CHD) were assessed. CDC Small Area Life Expectancy Estimates for Jefferson County were isolated and downloaded by census tract. Acute and psychiatric hospital addresses were downloaded from Kentucky’s Cabinet for Health and Family Services, geocoded, and summary statistics were calculated for distance between facilities and each census tract centroid. b The Environmental Exposures and Risks domain contains nine indicators related to ambient pollution and general environmental hazards. Estimates for air pollution emissions impacting reproductive health, respiratory health, and total cancer risks were downloaded from the National Air Toxics Assessment (NATA). PM2.5 data was downloaded from the CDC’s community multiscale air quality (CMAQ) program. The U.S. Department of Transportation’s National Transportation Noise Map raster file was downloaded and geocoded to census tracts to calculate mean 24-hour noise exposure. Environmental Hazard Index scores and . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 41 Domain Indicator Reference Data Description (File names and filter applied as applicable) Access Date (MM-DD- YEAR) Risk-Screening Environmental Indicator (RSEI) scores were downloaded from the U.S. Department of Housing and Urban Development (HUD) and the U.S. EPA, respectively. c The Environmental Sources domain contains seven indicators, six are related to sources of environmental exposures and the seventh is tree canopy. The U.S. EPA’s Facility Registry Service provided addresses for cement batch plants, metal recyclers, petrochemical and oil refineries, and power plants which were geocoded to census tracts to calculate distance between sites and census tract boundaries. The U.S. EPA’s Superfund program provided addresses for superfund sites which were geocoded to census tracts to calculate distance between sites and census tract boundaries. The U.S. Department of Transportation (DOT) provided traffic proximity and volume which was calculated according to the U.S. EPA’s EJScreen methodology (U.S. EPA 2022). The Multi-Resolution Land Characteristics Consortium provided tree canopy cover data as a raster which was spatially joined with census tract polygons to calculate summary statistics. d The Social Vulnerability domain contains nine indicators related to the resilience capacity of communities to recover from natural disasters and other crises. Three indicators related to housing affordability were provided by the U.S. Department of Housing and Urban Development (HUD) through their Location Affordability Index: percent income used for housing, percent renters, and median renter income vs. median area income. To calculate percent income used for housing the average housing cost was divided by median household income. Five indicators from the CDC and Agency for Toxic Substances and Disease Registry’s (ATSDR) Social Vulnerability Index were included: housing and transportation, household composition, minority status, socioeconomic status, and health insurance coverage. The U.S. Department of Agriculture’s Food Access Research Atlas provided food access. To calculate scores for food access, LowIncomeTracts and LATracts_half were summed. e The Weather domain contains three indicators related to severe weather. Flooding hazard was provided by ESRI’s Living Atlas USA Flood Hazard geodatabase that categorizes FEMA Flood zones by census tract. Warm season average maximum temperature value represents the mean average, minimum and maximum temperatures over the 2012 warm season (May – September), the number of deaths attributable to urban heat over the 2012 warm season provided by the Jefferson County Urban Heat Management Study. Tornado annualized frequency value represents the average number of recorded Tornado occurrences (event-days) per year provided by FEMA’s National Risk Index. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 42 Table S3. Indicators included in the Houston–Galveston–Brazoria (HGB) EnviroScreen’s Environmental Vulnerability Index (EVI) (Bhandari et al., 2020) which were removed for the Louisville Metro-Jefferson County index. Domain Indicator Social Vulnerability -Modified food retail environment index -Low food security Baseline Health -Childhood asthma Environmental Exposures and Risks -PM2.5 Satellite Environmental Sources -Leaking petroleum storage tanks -Facilities with risk management plans -Accident events reported in RMP -Shelter-in-place events reported in RMP Flooding -100-year flood plain -500-year flood plain -Harvey damage assessment “Affected” -Harvey damage assessment “Minimal” -Harvey damage assessment “Major” -Harvey damage assessment “Destroyed” -Families filing Harvey damage claims . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 43 Table S4. Indicators subgrouped based upon related impacts across domains for Intervention C case studies. Subgroup Domain Indicator High Scores (Census tract) C1 - Air quality Environmental NATA Cancer 20 (21111011604) Environmental NATA 0.3 (21111007501) Environmental NATA 0.03 (21111010307) Environmental PM 2.5 5.38923766 (21111011403) C2 - Environmental Infrastructure Environmental Tree Canopy 68.031571 (21111012003) Environmental Exposure and Risks Environmental Hazard Index 62 (21111010307) Environmental RSEI 0 Environmental Traffic Exposure 16.40093089 (21111011604) C3 - Potential Health Hazards Environmental Cement Batch 0 Environmental Metal Recyclers 0 Environmental Petrochemical and 0 Environmental Power Plants 0 Environmental Superfund Sites 0 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 44 Figure S1. Heterogeneity analysis for the 32 indicators. Analysis was performed for each indicator to ensure only factors that offer significant variance were included in the index (32 panels). . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 45 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 46 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 47 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 48 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 49 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 50 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 51 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 52 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 53 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 54 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 55 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 56 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 57 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 58 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 59 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 60 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 61 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 62 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 63 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 64 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 65 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 66 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 67 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 68 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 69 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 70 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 71 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 72 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 73 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 74 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 75 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint 76 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 17, 2024. ; https://doi.org/10.1101/2024.10.16.24315575doi: medRxiv preprint