Integrating Geospatial Data With Birth Cohorts to Explore Social Determinants of Health and Asthma

Abstract

Place-based measures of environmental exposures, climate, neighborhood characteristics, housing, and other social determinants of health (SDoH) are powerful predictors of health outcomes, including asthma. In addition, SDoH are likely causes of the persistent racial and ethnic disparities in asthma prevalence and morbidity. The objectives of this commentary are to 1) provide an overview of geospatial data and resources available to researchers to incorporate into studies of asthma-related outcomes, 2) provide a general approach to consider geospatial data in birth cohorts, 3) demonstrate the use of geospatial data in asthma-related research, and 4) highlight challenges and future opportunities for the use of geospatial data in asthma-related research and birth cohort studies. By integrating place-based data into longitudinal studies, researchers may identify critical drivers of asthma and asthma-related disparities and develop strategies to mitigate their impact.

Social Determinants of Health and Asthma

More than 4.6 million children suffer from asthma in the United States, and it remains the most common chronic disease of childhood.¹ While the overall prevalence of asthma has declined among children, racial and ethnic disparities in both prevalence and morbidity persist. Non-Hispanic Black children have more than double the prevalence (11.6%) of asthma compared to Non-Hispanic White children (5.5%).¹ Non-Hispanic Black children and Hispanic children of Afro-Caribbean descent also have higher rates of emergency department visits, hospitalizations, and death due to asthma compared to Non-Hispanic White children.

The primary cause of these disparities is unlikely to be genetic. Race is a social construct, and previous studies have identified social, economic and environmental factors as key contributors to asthma prevalence, severity, and control inequalities.^2–7 More likely, these arise from social determinants of health (SDoH), defined as the “conditions in which people are born, grow, work, live, and age, and the wider set of forces and systems shaping the conditions of daily life”.⁸ SDoH are powerful predictors of health, well-being, and quality of life across multiple diseases, including asthma, and contribute to more than 50% of the variation in health outcomes. The scope of SDoH encompasses economic stability, educational access and quality, health care access and quality, availability of affordable healthy, well-maintained housing, characteristics of the neighborhood and built environment, and social and community context. These operate at the individual, household, community, and systemic level; specific examples identified to play a role in asthma include individual socioeconomic status (e.g. household income, educational attainment, employment), housing quality (e.g. home disrepair, mold, water damage), stress and exposure to stressful events (e.g. crime, adverse childhood experiences), access to quality health care, and racism and discrimination.⁴

Neighborhood characteristics such as air quality, greenspace, access to healthy food, safety, and other aspects of the physical environment are also critical aspects of SDoH that contribute to the risk of asthma incidence and morbidity. Unlike some SDoH that require an individuals’ direct assessment (i.e. via questionnaires), many neighborhood and built environment SDoH are ‘place-based’ risks. Thus, knowing a person’s geographic location can yield valuable insights through geospatial approaches and data linkages. Birth cohorts offer exceptional platforms to study the role of SDoH in asthma incidence, morbidity, and disparities because their longitudinal design allows for specific periods of susceptibility (e.g. prenatally, first years of life) to be evaluated. Here, our objective is to provide an overview of the methods, strengths, and limitations associated with integrating place-based data into new and existing studies to examine the role of environmental exposures, community characteristics, and other SDoH in asthma incidence and morbidity to enhance our understanding of these critical drivers of health disparities and identify strategies to mitigate their impact.

Geospatial Data For Asthma Birth Cohorts

A large set of environmental, socioeconomic, and health-related data are collected or generated by local, regional, state, and federal agencies, as well as other organizations, and made publicly available. Repositories provided by the Centers for Disease Control and Prevention (CDC), the US Environmental Protection Agency (US EPA), the US Census Bureau, the Agency for Healthcare Research and Quality (AHRQ), and other agencies include data collected from outdoor air monitoring, the American Community Survey (ACS), the Food Environment Atlas, and other sources.^9–11 The spatial resolution of each dataset varies according to their collection and source but can be at the point level (e.g. location of air monitor) or within a boundary (e.g. county, city, zip code, or census block group or tract). Table 1 provides examples of frequently used geospatial data relevant to asthma grouped by NIH Health People 2030 domain. The datasets included in this table are not intended to be exhaustive of the extensive geospatial data available and all geospatial data, includes those listed, should be considered with respect to their source data, spatial and temporal resolution, and other strengths and limitations. . In addition, geospatial data at higher spatial or temporal resolution, including city neighborhoods, streets, parcel, and address level is available from many states and local governments. These datasets often include information on crime, drug overdoses, housing code violations, and other data that can be linked to local health and birth cohort addresses to study these characteristics and asthma outcomes. One example of the use of this type of local data is the derivation and application of the HOUSES index based on property data available from local assessor’s office in two South Dakota counties and linked to asthma control.¹² Using geospatial approaches and geographic information systems (GIS), these datasets can be used to derive additional place-based information including distances to nearby pollutant sources, drive times to local hospitals, and other relevant metrics.

Table 1.

Examples of Available Geospatial Data Relevant for Asthma Cohorts

NIH Healthy People 2030 Domain	Exposure / Characteristic	Example Source Data	Temporal and/or Spatial Resolution	Link to Asthma
Neighborhood and Built Environment	Ambient air pollution (PM2.5, NO2, O3)	Model derived estimates47 papers U.S. EPA	Daily concentrations at 1 km x 1 km grid, US (2000 – 2016) Monitoring locations	Exposure to air pollutants is associated with respiratory morbidity and asthma incidence24,48
	Temperature, Relative Humidity	PRISM Climate Data49	800 m x 800 m resolution	Changes in temperature associated with asthma morbidity50
	Major highways	U.S. Census Bureau, TIGER/Line Files51	1992, 2000, 2006-2024	Residential proximity to major roadways is associated with childhood wheeze and asthma morbidity52,53
	Greenspace	Landsat Normalized Difference Vegetation Index (NDVI) / MODIS54	16-day composite, 250 m – 1 km resolution	Residential proximity to greenspace linked to reduced risk of asthma55
Social and Community Context	Food Security	Food Environment Atlas56	County	Food insecurity increases risk of asthma and asthma exacerbations57
	Crime	U.S. Federal Bureau of Investigation, Crime Data Explorer58	State, County, City	Children residing in areas with increased crime at higher risk for childhood asthma59
	Demographic Characteristics (e.g. Race / Ethnicity, Language, Household Composition)	American Community Survey / US Census Data60	U.S. Census tracts, updated annually	Characteristics of the neighborhood including income, poverty, urbanicity, employment, educational attainment, housing, and health insurance associated with asthma morbidity and incidence3,4,61–65
	Housing (e.g. Type, Costs, Internet availability)	American Community Survey / US Census Data60	U.S. Census tracts, updated annually
Health Care Access and Quality	Health Insurance	American Community Survey / US Census Data60	U.S. Census tracts, updated annually
Economic Stability	Neighborhood poverty	American Community Survey / US Census Data60	U.S. Census tracts, updated annually
	Household income	American Community Survey / US Census Data60	U.S. Census tracts, updated annually
	Employment	American Community Survey / US Census Data60	U.S. Census tracts, updated annually
Education Access and Quality	Educational Attainment (e.g. percentage of population with high school education)	American Community Survey / US Census Data	U.S. Census tracts, updated annually

In addition to specific variables from sources like the ACS, composite area-level indices of community level SDoH including the Environmental Justice Index (EJI), the Child Opportunity Index (COI), the Social Vulnerability Index (SVI), the Area Deprivation Index (ADI), Material Community Deprivation Index, and others have been created for research, disaster planning, resource allocation, and other uses. While these indices are developed using similar source data, often the ACS and US Census, they differ in their development methods, specific variable composition, geographic extant, and intended purpose. As such, these factors should be carefully considered when determining their application in research.^13,14

Methods to Link Geospatial and Birth Cohort Data

Existing and new longitudinal cohorts can enrich their research by linking their participants’ location to geospatial databases (see examples in Table 1). The key piece of information to collect is the location of study participants – typically the street address of homes, schools, and other locations where children spend significant time and where SDoH may influence health outcomes. When possible, addresses should be collected prospectively from the time of enrollment, and retrospectively if prior locations are also relevant. Continued longitudinal collection thereafter reduces the potential for recall bias. Standard operating procedures for collecting and processing address histories are available and methods to reconstruct residential address histories through retrospective data collection or by utilizing commercial databases to ascertain participants’ prior addresses have also been described.^15,16 It is important to document the date(s) when participants began and ended living or spending time at each address to allow for longitudinal assessment. The careful compilation of address histories in birth cohorts prenatally throughout childhood allows researchers to evaluate the cumulative effects of SDoH and specific periods (e.g. childhood, adolescence) enabling lifecourse epidemiology.

Geocoding is the process of linking address text to geospatial points or boundaries, most often converting mailing addresses to latitude and longitude coordinates; though other partial address information (e.g. five-digit ZIP code, street intersection) can also be geocoded to spatial boundaries such as ZIP code tabulation (ZCTAs) provided by the US Census. Additionally, address-level matching or parcel-level geocoding can be used to link mailing addresses to physical buildings or parcels of land, facilitating linkage of housing code infractions or other address-level infractions.¹⁷ The positional accuracy, cost, computational resources, required software, and other advantages, disadvantages, and implications of the different methodological approaches to geocoding have been described.¹⁸ In addition, researchers must also consider participant confidentiality when selecting a geocoding method. Some methods and software, including those relying on sharing addresses with a third-party (e.g. Google Maps, ArcGIS Online, Census Geocoder), require a business associate agreement (BAA) which is a legal contract ensuring that both parties will comply with Health Insurance Portability and Accountability Act (HIPAA) regulations.¹⁹ One alternative approach is our Decentralized Geomarker Assessment for Multi-Site Studies (DeGAUSS) software, specifically developed and freely available to geocode and integrate addresses with geospatial data while maintaining protected health information by using a federated approach.^20,21

Linking geocoded addresses to place-based data involves the use of geospatial software to perform a ‘spatial join’ based on a defined spatial relationship between the two datasets. Common spatial relationships include containment (e.g. identifying the polygon or spatial boundary in which an address is located) and proximity (e.g. determining the distance to a geographic feature of interest). For example, identifying the census tract that an address is located within is a containment join, while measuring the distance to the nearest major road or the distance to a nearby industrial facility are proximity calculations. Multiple software options are available for mapping and analyzing geographic data including both proprietary GIS software (e.g. ArcGIS Pro, ESRI, Redlands CA) and open-source alternatives (e.g. QGIS). Additionally, open-source software and tools, including R and Python, are increasingly used for mapping and geospatial analysis and multiple packages specifically designed for spatial data are available.²²

Case Example:

The Children’s Respiratory and Environmental Workgroup (CREW) Consortium:

The CREW consortium, part of the NIH Environmental Influences on Child Health Outcomes (ECHO) program, consists of 12 individual birth cohorts studying the development of allergic disease and asthma in childhood. CREW, and the cohorts that comprise the consortium, have been described in detail.²³ Using our DeGAUSS software, each cohort site geocoded participant addresses and identified the census tract in which they were located at the time of birth, which ranged over more than 4 decades. After geocoding the addresses and identifying their corresponding census tracts, neighborhood-level information was derived for participants from the US Census and ACS. Census tract-level data included population density, median household income, percent of the population in poverty, percent of the population without a high school diploma, percent occupied housing, percent of single-parent households, and other place-based SDoH indicators.²¹ Further geospatial data linked to geocoded addresses included air pollution estimates and whether the child was born in a historically redlined neighborhood.²⁴

This geospatial data, coupled with individual-level information collected by the cohorts, enabled multiple analyses examining the impact of SDoH on asthma incidence. For example, we found that children residing in census tracts with higher proportions of low-income households, increased population density, and elevated poverty had higher risk for developing childhood asthma. In addition, we observed that SDoH indicators from census data did not modify the increased risk of developing asthma for Black and Hispanic children, who remained at higher risk compared to White children regardless of their census tract socioeconomic indicators potentially due to unmeasured discrimination.³

We further assessed early-life exposure to fine particulate matter (PM_2.5) and nitrogen dioxide (NO₂), estimated at the home addresses of CREW participants from birth through age three. Our results identified significant associations between NO₂ and PM_2.5 exposure and asthma incidence through ages 5 and 11. Moreover, by examining interactions between neighborhood and individual-level SDoH we identified that the association between air pollution and asthma was heightened among children whose mothers had less than a high school diploma and who lived in densely populated communities.²⁴ Using historical redlining maps, overlaid with census tracts and participants geocoded addresses, we also determined whether their home addresses were located in a historically redlined neighborhood. With this information, combined with present-day census tract data, we conducted a mediation analysis that demonstrated the relationship between historical redlining and present-day disparities in asthma incidence was due, in part, to changes in neighborhood socioeconomic status.²⁵

Challenges in the Use of Geospatial Data for Birth Cohorts

Data Privacy, Confidentiality, and Ethical Considerations:

Address information, including geocodes, are personally identifiable information and the use of these poses a risk to participant confidentiality. More precise spatiotemporal data could also inadvertently reveal identities of study participants, especially in less populated areas. When spatiotemporal data is combined with other personal information, the risk of re-identification always increases, often in ways that cannot be anticipated. Personal information linked to heath information, such as is done in research studies, is legally regulated by HIPAA as well as the Health Information Technology for Economic and Clinical Health Act (HITECH) and the Common Rule (Federal Policy for the Protection of Human Subjects). In addition to legal compliance, researchers should take a multidisciplinary approach that includes technical solutions, ethical considerations, and engagement with community stakeholders. Challenges exist within informed consent processes in that broad consent forms often do not adequately cover the use of location data and when engaged in a more detailed consent form, individuals may not fully understand how their location data will be used or the potential risks involved. Using location data for purposes beyond the original scope may violate participant trust. Lastly, identifying geographic areas with poor health outcomes may lead to stigmatization of those communities. Overall, by implementing robust data governance practices and engaging transparently with study participants, researchers can mitigate privacy risks while harnessing the full potential of geospatial data to improve health outcomes.

In addition to confidentiality and privacy concerns, researchers using geospatial data related to health and economic disparities, racial and ethnic segregation, historical redlining and housing policies, crime, and other potentially stigmatizing data must guard against the propagation of bias and discrimination. Prior studies have identified community residents’ concerns and experiences with geospatial visualizations of these types of data including its use to support gentrification and displacement of marginalized residents.²⁶ In addition to available ethical guidelines and resources, researchers may also consider community-based participatory research principles and soliciting input from affected community members when including geospatial data in the dissemination of study results and visualizations of sensitive geospatial data.²⁷

Summary Indices or Specific Variables of Neighborhood Socioeconomic Status:

When studying the role of neighborhood socioeconomic position on asthma, researchers frequently face the decision of whether use composite indices of neighborhood characteristics, like the SVI or COI, or to analyze the individual variables that comprise these indices. The SVI, for example, uses 16 U.S. Census variables from the 5-year American Community Survey (ACS) to identify communities that may need support before, during, or after disasters. These variables are grouped into four themes covering socioeconomic status, household characteristics, racial and ethnic minority status, and housing type and transportation, and are then summarized into a single measure of overall social vulnerability.²⁸ Similarly, the COI aggregates multiple sources to define space-based child opportunities in domains such as educational, health and environmental, and social and economic opportunities.²⁹

Aggregated indices provide an overall measure of a neighborhood’s resources, simplifying analyses by condensing multiple dimensions of often correlated socioeconomic resources into a composite metric. These indices are often standardized across census tracts or other geographic regions, facilitating comparisons across communities. However, aggregating data into an index can obscure the effects of individual components and hinder efforts to identify and intervene on specific neighborhood characteristics. In addition, the accuracy of indices and their appropriate use in health studies depends, in part, on the type and quality of data used in their development leading to the potential for misuse when not properly understood. Alternatively, using individual data variables can provide specific insight into which aspects or components of a neighborhood drive health outcomes allowing for more targeted interventions. The use of individual data can also allow researchers to select the most appropriate variables for their studies and identify their specific contribution. However, due to the highly correlated nature of many neighborhood measures (e.g. household income, percentage of individual without health insurance) a composite index may still be beneficial in cases where statistical models cannot deal with highly correlated data (e.g., regression models), or to capture overall context.

Data Integration and Standardization:

Geospatial data is collected at different spatial and temporal resolutions and is provided in a variety of formats that can often be challenging to integrate and standardize. While US Census and ACS data is provided yearly at the block group and tract geographic units, these boundaries are re-mapped every 10 years and are subject to change based on shifting population distributions. For longitudinal studies, it is important to know which boundaries should be used at different points in time. ACS data is available yearly back to 2006, but prior to that census data is only available every decade requiring researchers to choose a dataset most appropriate for their research and population. Environmental data such as air pollution and climate are available at varying resolutions including raw data collected at the source point, and modelled surfaces produced at differing resolutions (e.g 800m, 1km, and 4km). The grid cell areas used for these modelled surfaces do not conform to zip codes, counties, census tracts, or other geographic units. Further, data formats among datasets may be in tabular, text, vector, raster, or other format. Often several pre-processing steps using modern GIS software applications or scripting languages are required to convert the various data formats into usable geospatial information. Downloading, pre-processing, and associating various geospatial datasets with cohort locations can often be a laborious, time-consuming process with specific, detailed workflows to follow.

In epidemiologic studies, after accounting for measured geospatially-defined risk factors or sources of opportunity, there frequently still remains a residual, and often strong association of self-described race or ethnicity with health outcomes. In the CREW multi-birth cohort investigation, census variables did not significantly modify the association between race and ethnicity and risk for asthma incidence; Black and Hispanic children remained at higher risk for asthma compared with White children across census tracts socioeconomic levels potentially due to unmeasured discrimination not captured well by the census variables.³

Opportunities in the Use of Geospatial Data for Birth Cohorts

Lifecourse Epidemiology:

Integrating geospatial data with longitudinal address data from birth cohorts can significantly enhance life course epidemiology, allowing researchers to study both cumulative exposures to SDoH and those experienced during specific periods of growth and development. The collection of location information beginning prenatally allows investigators to study the influence of geospatial data during pregnancy on later maternal and child health outcomes. Individuals may also experience changes in SDoH due to the temporality of exposures (e.g. air pollution, neighborhood gentrification) or residential mobility. Changes in participants’ locations can significantly influence lifetime exposure to greenspace, air pollution, and other place-based exposures and these may lead to differential exposure misclassification when the characteristics of participants who move are different from those who do not.^30,31 However, residential mobility also provides the opportunity to study changes in housing and neighborhood characteristics and the impact on asthma. For example, children residing in high poverty neighborhoods in Baltimore MD who participated in a housing mobility program were found to have improve asthma morbidity after moving to low-poverty neighborhoods.³² The collection of longitudinal addresses in birth cohorts also enables ‘natural experiments’ to examine associations between changes in SDoH and asthma. Examples of these analyses include data from the Upstate KIDS study in New York state demonstrating that residential mobility from neighborhoods with high to low COI was associated with increased risk of asthma diagnosis in early childhood and aa recent study using the UK Biobank and estimated concentrations of air pollutants across longitudinal address histories to identify a significantly increased risk for respiratory diseases among participants who moved from low exposure addresses to addresses with moderate and high air pollution.^{33 34}

Precision Public Health:

The promise of precision health is to provide tailored interventions to the appropriate population at the relevant time, but this necessitates understanding which geospatial factors, combined with individual data, are most influential. Considering geospatial data in birth cohorts can advance precision public health by identifying communities and individuals at heighted risk for asthma and asthma morbidity. Studies with large sample sizes are needed to consider the joint effects of multiple SDoH to identify specific characteristics of communities and individuals that increase asthma, enabling the delivery of tailored and effective interventions. For example, Sharma et al. examined environmental exposure to air pollution and temperature, as well as neighborhood crime and socioeconomic deprivation, and found that neighborhood characteristics modified the effects of environmental exposures.³⁵

Geocoding patients with specific health outcomes can also help identify place-based risk factors. One well-known example of this approach is the use of geocoded locations of children with elevated blood lead levels to identify neighborhoods and housing factors contributing to pediatric lead exposure in Flint, MI.³⁶ Similarly, Samuels et al. geocoded the addresses of children and adults who had an emergency department visit related to asthma. By linking these addresses to city parcel-level housing inspection data, the investigators showed that increased asthma ED visits were identified approximately one year prior to failed housing inspections.³⁷ In Massachusetts, Lemire and colleagues used geospatially data to demonstrate disparities in housing code enforcement in response to tenants complaints of housing-related exposures known to be potential asthma triggers.³⁸ Investigators have also created geospatial datasets of housing laws, including landlord-tenant laws, fair housing laws, and nuisance property ordinances that can be used to assess effects on health.³⁹ In addition to housing violation records to identify neighborhoods with poor living conditions that may exacerbate asthma symptoms, voter and driver registration data can provide insights into community engagement and access to resources.⁴⁰ When combined with health information, these highly resolved geospatially and temporally data can inform efforts to increase civic engagement, improve housing quality, enforce existing housing codes, and help to determine the success of interventions aimed at improving health, including respiratory health.

Summary:

There is a wealth of geospatial data available for researchers to integrate with new and ongoing birth cohort studies to better understand how place-based measures of the physical and social environment contribute to asthma incidence, morbidity, and disparities. Community characteristics, environmental exposures, housing information, and other geospatial data have been used to identify drivers of asthma related outcomes including asthma diagnoses, hospitalizations, emergency department visits, and medication access – as well as disparities in these ^41,42 Results of studies using geospatial data can be used to inform interventions and policies to reduce the influence of these external and modifiable contributors to asthma in childhood. Examples of interventions based on geospatially identified risk factors include housing remediation (e.g. smoke-free policies, pest-control, improved ventilation and other ‘green housing’ solutions), targeted building and zoning restrictions or greenspace additions, deployment of community health workers, or increasing food security.^43–46 In addition, historical and current data on temperatures, humidity, and other climate-related factors are available to better understand the existential threat posed by climate change on asthma and asthma disparities. The integration of geospatial data holds significant promise to mitigate the effects of SDoH on asthma and promote equitable outcomes for all children.

Abbreviations

SDoH

Social Determinants of Health

CDC

Centers for Disease Control and Prevention

US EPA

United States Environmental Protection Agency

AHRQ

Agency for Healthcare Research and Quality

ACS

American Community Survey

NIH

National Institutes of Health

EJI

Environmental Justice Index

COI

Child Opportunity Index

SVI

Social Vulnerabilities Index

ADI

Area Deprivation Index

PM2.5

Particulate matter < 2.5 microns

NO2

Nitrogen dioxide

O3

Ozone

NDVI

Normalized Difference Vegetation Index

ZCTA

Zip Code Tabulation Area

GIS

Geographic Information System

HIPAA

Health Insurance Portability and Accountability Act

DeGUASS

Decentralized Geomarker Assessment for Multi-Site Studies

CREW

Children’s Respiratory and Environmental Workgroup

ECHO

Environmental Influences on Child Health Outcomes

HITECH

Health Information Technology for Economic and Clinical Health Act