Assessment of Population‐Level Disadvantage Indices to Inform Equitable Health Policy

Abstract

Policy Points.

• The rapid uptake of disadvantage indices during the pandemic highlights investment in implementing tools that address health equity to inform policy.

• Existing indices differ in their design, including data elements, social determinants of health domains, and geographic unit of analysis. These differences can lead to stark discrepancies in place‐based social risk scores depending on the index utilized.

• Disadvantage indices are useful tools for identifying geographic patterns of social risk; however, indiscriminate use of indices can have varied policy implications and unintentionally worsen equity. Implementers should consider which indices are suitable for specific communities, objectives, potential interventions, and outcomes of interest.

Context

There has been unprecedented uptake of disadvantage indices such as the Centers for Disease Control and Prevention Social Vulnerability Index (SVI) to identify place‐based patterns of social risk and guide equitable health policy during the COVID‐19 pandemic. However, limited evidence around data elements, interoperability, and implementation leaves unanswered questions regarding the utility of indices to prioritize health equity.

Methods

We identified disadvantage indices that were (a) used three or more times from 2018 to 2021, (b) designed using national‐level data, and (c) available at the census‐tract or block‐group level. We used a network visualization to compare social determinants of health (SDOH) domains across indices. We then used geospatial analyses to compare disadvantage profiles across indices and geographic areas.

Findings

We identified 14 indices. All incorporated data from public sources, with half using only American Community Survey data (n = 7) and the other half combining multiple sources (n = 7). Indices differed in geographic granularity, with county level (n = 5) and census‐tract level (n = 5) being the most common. Most states used the SVI during the pandemic. The SVI, the Area Deprivation Index (ADI), the COVID‐19 Community Vulnerability Index (CCVI), and the Child Opportunity Index (COI) met criteria for further analysis. Selected indices shared five indicators (income, poverty, English proficiency, no high school diploma, unemployment) but varied in other metrics and construction method. While mapping of social risk scores in Durham County, North Carolina; Cook County, Illinois; and Orleans Parish, Louisiana, showed differing patterns within the same locations depending on choice of disadvantage index, risk scores across indices showed moderate to high correlation (r_s 0.7‐1). However, spatial autocorrelation analyses revealed clustering, with discrepant distributions of social risk scores between different indices.

Conclusions

Existing disadvantage indices use varied metrics to represent place‐based social risk. Within the same geographic area, different indices can provide differences in social risk values and interpretations, potentially leading to varied public health or policy responses.

In the united states, the covid‐19 pandemic highlighted the consequences of underinvesting in public health infrastructure and the urgency of addressing intersectional health needs to advance health equity and justice. ¹ , ² Social disadvantage and structural racism have markedly influenced the evolution of the pandemic among historically marginalized populations, including Black, Latino, and American Indian communities. ³ , ⁴ States have collected data stratified by race, ethnicity, and social factors ⁵ , ⁶ , ⁷ to identify communities experiencing disproportionate risk of COVID‐19 exposure and adverse health outcomes to rapidly guide emergency response efforts that quantify structural disadvantage and social risk.

Disadvantage indices use publicly available population‐level data to facilitate identification of place‐based patterns of social risk that can influence health outcomes. ⁸ , ⁹ There has been unprecedented uptake of disadvantage indices during the pandemic to inform equitable COVID‐19 testing and vaccine allocation efforts. ¹⁰ Motivated in part by the National Academies of Science, Engineering, and Medicine (NASEM) framework for equitable vaccine allocation and Centers for Disease Control and Prevention (CDC) recommendations to use indices such as the CDC Social Vulnerability Index (SVI) in vaccine distribution planning, ¹¹ by March 2021 at least 34 states were using disadvantage indices to inform data‐driven COVID response strategies. ¹⁰ , ¹² States have used these tools to guide interventions, including allocating vaccines to communities with highest social vulnerability (e.g., CA, CT, KS, MA, MI, NH, NM) or historically marginalized populations (e.g., NC); increasing appointments in areas with high social risk (e.g., CA, CO, DC, MT, RI, UT, VA); placing vaccine sites in disproportionately affected locations (e.g., AK, CO, IN, MD, MI, PA, UT); and increasing outreach through mobile clinics or community partnerships (e.g., AK, LA, MI, MN, NV, TN, WA). ¹³ , ¹⁴ , ¹⁵

Proactive planning informed by disadvantage indices may impact the successful and equitable implementation of public health interventions. However, many indices were not developed with sensitivity to multifaceted determinants of health or their interaction during a prolonged public health crisis such as the COVID‐19 pandemic. ¹⁶ Multiple disadvantage indices exist, and it is not clear how policymakers or researchers should choose between various indices. There is limited standardization around data elements in existing disadvantage indices, and it is unknown how indices perform in unique community contexts. Therefore, despite the uptake of disadvantage indices during COVID‐19 and recommendations to use indices more broadly, ¹⁰ , ¹¹ an implementation gap has emerged due to the lack of evidence around application of these tools to promote health equity. ¹⁷ Here, we describe and compare disadvantage indices and use spatial analyses to analyze social risk profiles generated when using various indices across different geographic areas. We aim to describe evidence gaps, synthesize methodological challenges, and provide recommendations to improve the use of disadvantage indices moving forward.

Methods

The study spanned two stages, beginning with a review of the literature, followed by geospatial mapping.

Literature Review

We reviewed academic and gray literature for disadvantage indices to estimate place‐based social risk or resilience. Table 1 presents a summary of all indices and their parameters. We searched PubMed, Google Scholar and Google using the following search terms social vulnerability index, social risk index, social deprivation index, neighborhood deprivation index, COVID‐19 social vulnerability index, child vulnerability index. We categorized indicators in the following social determinants of health (SDOH) domains described by Healthy People 2030: economic stability; education access and quality; health care access and quality; neighborhood and built environment; and social and community context. ¹⁸

Table 1.

Summary and Comparison of Disadvantage Indices

				SDOH Domains Included a
Index (Owner)	Purpose	Data Source	Granularity	Economic	Education	Health Care	Neighborhood	Community
ADI: Area Deprivation Index (Univ of Wisconsin) b	Rank neighborhoods by socioeconomic disadvantage in a region of interest	American Community Survey (ACS)	Census block group	✓	✓			✓
17 indicators: % aged ≥ 25 years with < 9 years of education, % aged ≥ 25 years with greater than or equal to a high school diploma, % of employed persons ≥ 16 years of age in white‐collar occupations, median family income, income disparity, median home value, median gross rent, median monthly mortgage, % owner‐occupied housing units (homeownership rate), % civilian labor force population ≥ 16 years of age unemployed (unemployment rate), % of families below the poverty level, % of population below 150% of the poverty threshold, % of single‐parent households with children < 18 years of age, % of occupied housing units without a motor vehicle, % of occupied housing units without a telephone, % of occupied housing units without complete plumbing (log), % of occupied housing units with more than one person per room (crowding).
BRIC: Baseline Risk Indicators for Communities (Univ of South Carolina)	Provides an overall baseline assessment for monitoring existing attributes of resilience to natural hazards	30 sources including ACS 3‐Year and 5‐Year Estimates, Decennial Census, and Census of Agriculture	County	✓	✓	✓	✓	✓
49 indicators in 6 summary themes: Social resilience (educational attainment equality, preretirement age, transportation, communication capacity, English‐language competency, non–special needs, health insurance, mental health support, food provisioning capacity, physician access) Economic resilience (homeownership, employment rate, race/ethnicity income equality, nondependence on primary/tourism sectors, gender income equality, business size, large retail–regional/national geographic distribution, ederal employment) Community capital (place attachment–not recent immigrants, place attachment–native born residents, political engagement, social capital–religious organizations, social capital–civic organizations, social capital–disaster volunteerism, citizen disaster preparedness, and response skills) Institutional resilience (mitigation spending, flood insurance coverage, jurisdictional coordination, disaster aid experience, local disaster training, performance regimes–state capital, performance regimes–nearest metro area, population stability, nuclear plant accident planning, crop insurance coverage) Housing/infrastructural resilience (sturdier housing types, temporary housing availability, medical care capacity, evacuation routes, housing stock construction equality, temporary shelter availability, school restoration potential, industrial resupply potential, high‐speed internet infrastructure) Environmental resilience (local food suppliers, natural flood buffers, efficient energy use, pervious surfaces, efficient water use)
CCVI: COVID‐19 Community Vulnerability Index (Surgo Ventures) b	Assesses which US communities may be less resilient to the impacts of the COVID pandemic	12 sources including ACS, Centers for Medicare and Medicaid Services (CMS), CDC SVI	Census tract	✓	✓	✓	✓	✓
40 indicators in 7 summary themes. The first 3 themes—socioeconomic status, minority status, and language—are modified from SVI themes. Two variables (uninsured population, access to indoor plumbing) were added to these themes. Theme 4 is epidemiological factors [estimated % of adults diagnosed with high cholesterol; estimated % of adults diagnosed with a stroke; estimated % of adults ever diagnosed with heart disease; estimated % of adults diagnosed with chronic obstructive pulmonary disease, emphysema, or chronic bronchitis; estimated % of adults reporting to smoke cigarettes; annual cancer incidence per 100,000 persons; rate of persons living with an HIV diagnosis per 100,000 people; estimated % of adults reporting to be obese (a body mass index of 30 or greater); estimated % of adults ever diagnosed with diabetes; persons aged 65 and older estimate]. Theme 5 is health care system factors [Intensive care unit beds per 100,000; hospital beds per 100,000; epidemiologists per 100,000, AHRQ—Prevention Quality Indicator Overall Composite (PQI): admission rates for preventable conditions (via good outpatient care) adjusted per population, health spending per capita, aggregate cost of medical care, % of population with a primary care physician, total Public Health Emergency Preparedness funding per capita, health labs per 100,000, emergency services per 100,000]. Theme 6 is high‐risk environments [long‐term care (nursing homes, assisted living, and care homes) residents per 100,000, prisons population per 100,000, % of population employed in high‐risk industry]. Theme 7 is population density [estimated total number of people per unit area (sq. miles)].
Chicago CCVI: Chicago COVID‐19 Community Vulnerability Index [Chicago Department of Public Health (CDPH)]	Identifies communities that have been disproportionately impacted by COVID‐19 and are uniquely vulnerable to barriers to COVID‐19 vaccine uptake	ACS, Healthy Chicago Survey, BlueDot, CDPH Communicable Disease Surveillance	Community area	✓	✓	✓	✓	✓
25 indicators in 4 summary themes: Sociodemographic factors (% individual poverty, % 16+ unemployment, % per capita income, % 25+ no high school diploma, % uninsured, % population 17 and under, % living with disability, % crowded housing, % single‐parent households, % with no primary care provider) Epidemiological factors (% 65+ years, % adult obesity, % adult current smoking, % adult diabetes) Occupational factors [education (teachers, support staff), manufacturing/food production, material moving (grocery store, laborers, freight), personal service (Child care, recreation and entertainment workers), public safety (first responders/corrections), transportation (public transit, airport truck, taxi), food service (cooks, servers, food prep, kitchen), Ratio of mobility in 2020 compared to 2019] Cumulative COVID‐19 burden [Diagnosed COVID‐19 cases (rate per 100,000), COVID‐19 hospital admissions (rate per 100,000), COVID‐19 mortality rate (rate per 100,000)]
COI: Child Opportunity Index (Heller School for Social Policy and Management at Brandeis University) b	Captures neighborhood resources and conditions that matter for children's healthy development in a single metric	Numerous data sources including: National Center for Health Statistics, Dept. of Education, Environmental Protection Agency (EPA); proprietary data licensed from Great Schools	Census tracts	✓	✓		✓
29 indicators in 3 summary themes: Education [number of early childhood education centers within a 5‐mile radius; number of NAEYC‐accredited centers within a 5‐mile radius; % 3‐ and 4‐year‐olds enrolled in nursery school, preschool, or kindergarten; % third graders scoring proficient on standardized math tests, converted to NAEP scale score points; % ninth graders graduating from high school on time; ratio of students enrolled in ≥ 1 AP courses to the number of 11th and 12th graders; % 18‐24‐year‐olds enrolled in college within 25‐mile radius, % students in elementary schools eligible for free or reduced‐price lunches, reversed; % teachers in their first and second year, reversed; % adults ages 25 and over with a college degree or higher] Health and Environment [% households without a car located farther than a half mile from the nearest supermarket, reversed; % impenetrable surface areas such as rooftops, roads, or parking lots, reversed; EPA Walkability Index; % housing units that are vacant, reversed; average number of Superfund sites within a 2‐mile radius, reversed; index of toxic chemicals released by industrial facilities, reversed; mean estimated microparticle (PM2.5) concentration, reversed; mean estimated 8‐hour average ozone concentration, reversed; summer days with maximum temperature above 90 degrees F, reversed; % individuals ages 0–64 with health insurance coverage] Social and Economic (% adults ages 25–54 who are employed; % workers commuting more than one hour one way, reversed; % individuals living in households with incomes below 100% of the federal poverty threshold, reversed; % households receiving cash public assistance or food stamps/Supplemental Nutrition Assistance Program, reversed; % owner‐occupied housing units; % individuals aged 16 and over employed in management, business, financial, computer, engineering, science, education, legal, community service, health care practitioner, health technology, arts and media occupations; median income of all households; % family households that are single‐parent headed, reversed) Abbreviations: NAEYC, National Association for the Education of Young Children; NAEP, National Assessment of Educational Progress.
CRIA: Community Resilience Indicator Analysis (Federal Emergency Management Agency)	Provides relative assessment of community's potential resilience	ACS	County	✓	✓	✓	✓	✓
20 indicators in 2 summary themes: Population focused (% population without health insurance, % population unemployed, % population without a high school education, % population with a disability, % population without access to a vehicle, % population with home ownership, % population over 65, % population single‐parent households, % population with limited English proficiency, median household income, Gini index: income inequality) Community focused (connection to civic/social organizations, hospital capacity, medical professional capacity, affiliation with a religion, presence of mobile homes, public school capacity, population change, hotel/motel capacity, rental property capacity)
DCI: Distressed Communities Index (Economic Innovation Group)	Examines economic well‐being to provide a detailed view of the landscape of American prosperity	US Census Bureau's Business Patterns and ACS	Zip Codes	✓	✓		✓
7 indicators: no high school diploma (population ≥ 25 who lack a high school diploma or equivalent), poverty rate, adults not working, housing vacancy, median household, change in employment (change from 2014 to 2018 in the number of employees working), change in establishments (change from 2014 to 2018 in the number of establishments located in the geography)
HPI: California Healthy Places Index (Public Health Alliance of Southern California)	Summarizes the conditions and the levels of key resources in a community that foster a healthy population and health equity	14 data sources including ACS, Comprehensive Housing Assessment System, CA EPA	California census tract	✓	✓		✓
25 indicators in 8 summary themes: Economic (% of population with income exceeding 200% of federal poverty level, % of population aged 25‐64 who are employed, median household income) Education (% of population over age 25 with a bachelor's education or higher, % of 15‐17‐year‐olds enrolled in school, % of 3‐ and 4‐year‐olds enrolled in preschool) Health care access (% of adults aged 18 to 64 years currently insured) Housing (% of occupied housing units occupied by property owners, % of households with complete kitchen facilities and plumbing, % of low‐income homeowners paying more than 50% of income on housing costs, % of low‐income renter households paying more than 50% of income on housing costs, % of households with ≤1 occupant per room) Neighborhood [% of the population living within a half mile of a park, beach, or open space greater than 1 acre; population‐weighted % of the census‐tract area with tree canopy; % of urban and small‐town population residing less than a half mile from a supermarket/large grocery store; % of rural population living less than 1 mile from a supermarket/large grocery store; % of the population residing within a quarter mile of an off‐site sales alcohol outlet; combined employment density for retail, entertainment, and educational uses (jobs/acre)] Clean environment [Spatial distribution of gridded diesel PM emissions from on‐road and nonroad sources for a 2012 summer day in July (kg/day), Cal EnviroScreen 3.0 drinking water contaminant index for selected contaminants, mean of summer months (May‐October) of the daily maximum 8‐hour ozone concentration (ppm), averaged over 3 years (2012 to 2014); nnual mean concentration of PM2.5 (average of quarterly means, μg/m3) over 3 years (2012 to 2014)] Social (% of registered voters voting in the 2012 general election, % of family households with children under 18 with 2 parents) Transportation [% of households with access to an automobile, % of workers (16 years and older) commuting by walking, cycling, or transit (excluding working from home)] Abbreviations: ACS, American Community Survey; CA EPA, California Environmental Protection Agency; PM, particulate matter; ppm, parts per million.
Minority Health SVI: The Minority Health Social Vulnerability Index (Office of Minority Health)	Helps identify communities that may need support before, during, and after disasters, with a focus on minority groups as well as medical vulnerability	ACS	County	✓	✓	✓		✓
34 variables in 6 summary themes: Socioeconomic status [persons below poverty estimate, civilian (age 16 or older) unemployed estimate, per capita income estimate, persons with no high school diploma (age 25+) estimate] Household composition and disability (persons aged 65 or older estimate, persons aged 17 or younger estimate, civilian noninstitutionalized population with a disability estimate, single‐parent households with children under 18 estimate) Minority status and language (American Indian/Alaska Native estimate, Asian estimate, African American or Black estimate, Native Hawaiian/Pacific Islander estimate, Hispanic or Latino/a estimate, Some Other Race Alone estimate, Spanish speakers who speak English less than “very well,” Chinese speakers who speak English less than “very well,” Vietnamese speakers who speak English less than “very well,” Korean speakers who speak English less than “very well,” Russian speakers who speak English less than “very well”) Housing type and transportation [housing in structures with 10 units or more estimate, mobile homes estimate, at household level (occupied housing units): more people than rooms estimate, households with no vehicle available estimate, persons in institutionalized group quarters estimate] Health care infrastructure and access (hospitals per 100,000, urgent care clinics per 100,000, pharmacies per 100,000, primary care physicians working in patient care, nonfederal, persons without health insurance estimate) Medical vulnerability (total cardiovascular disease death rate per 100,000, chronic respiratory diseases, diagnosed diabetes in adults aged 20+ years, obesity in adults aged 20+ years, no internet access estimate)
MDI: Multidimensional Deprivation Index (US Census Bureau)	Estimates deprivation across multiple dimensions, intended to supplement measures of poverty	ACS, ADI	County	✓	✓	✓	✓
Variables across 6 summary themes: Standard of living (in poverty according to the official poverty measures) Education (aged 19 or older and without a high school diploma or GED) Health (for people under age 65: lacked health insurance; for people age 65 and over: lacked health insurance or reported at least 2 disabilities) Economic security [for people under age 65: aged 18 and older and unemployed at the time of the survey OR lived in a household in which average household hours worked OR average household weeks worked for working‐age adults (aged 18–64, not currently enrolled in school) was less than 20 hours a week or less than 26 weeks a year, respectively. For people age 65 and over: ynemployed at the time of the survey OR orked less than 20 hours a week OR less than 26 weeks a year AND had minimal retirement income] Housing quality (lived in a housing unit with more than 2 people per bedroom or lived in a shelter) Neighborhood quality (lived in a deprived block group as measured by the Area Deprivation Index: all block groups with an ADI score greater than 90).
NRI: National Risk Index (Federal Emergency Management Agency)	Helps illustrate the nation's communities most at risk of natural hazards	SoVI, BRIC, and annual hazard frequency data sources	Census tract	✓	✓	✓	✓	✓
Risk is calculated by expected annual loss (a likelihood and consequence component of risk that measures the expected loss of building value, population, and agricultural value each year due to natural hazards) times Social Vulnerability (outlined by the variables included in the SoVI), divided by community resilience (outlined by the variables included in the BRIC). Abbreviations: BRIC, Baseline Risk Indicators for Communities (University of South Carolina); SoVI, Social Vulnerability Index (University of South Carolina).
SDI: Social Deprivation Index (Robert Graham Center)	Quantifies level of disadvantage across small areas, evaluates association with health outcomes, and addresses health inequities	ACS	Primary Care Service Area	✓	✓		✓
7 indicators in 6 summary themes: income (% living in poverty), education (% with less than 12 years of education), household characteristics (% single‐parent household), housing (% living in rented housing unit, % living in overcrowded housing unit), transportation (% of households without a car), employment (% nonemployed adults under 65 years of age)
SoVI: Social Vulnerability Index (Univ of South Carolina)	Measures the social vulnerability of US counties to environmental hazards	ACS	County	✓	✓	✓	✓
29 indicators: % Asian, % Black, % Hispanic, % Native American, % population under 5 years or 65 and over, % children living in 2‐parent families, median age, % households receiving Social Security benefits, % poverty, % households earning over $200,000 annually, per capita income, % speaking English as a second language with limited English proficiency, % female, % female‐headed households, nursing home residents per capita, hospital per capita (county level only), % of population without health insurance (county level only), % with less than 12th‐grade education, % civilian unemployment, people per unit, % renters, median housing value, median gross rent, % mobile homes, % employment in extractive industries, % employment in service industry, % female participation in labor force, % of housing units with no car, % unoccupied housing units, % of all households spending more than 40% of their income on housing expenses (tract level only)
SVI: Social Vulnerability Index (CDC ASTDR) b	Identifies communities that may need support before, during, and after a hazardous event	ACS	Census tract	✓	✓			✓
15 indicators in 4 summary themes: Socioeconomic status [persons below poverty estimate, civilian (age 16+) unemployed estimate, per capita income estimate, persons with no high school diploma (age 25+) estimate] Household composition and disability (persons aged 65 or older estimate, persons aged 17 or younger estimate, civilian noninstitutionalized population with a disability estimate, single‐parent households with children under 18 estimate) Minority status and language [minority (all persons except white, non‐Hispanic) estimate, persons (age 5+) who speak English “less than well” estimate] Housing type and transportation (housing in structures with 10 units or more estimate, mobile homes estimate, at household level (occupied housing units): more people than rooms estimate, households with no vehicle available estimate, persons in institutionalized group quarters estimate)

^a Social Determinants of Health (SDOH) domains. Variables were categorized using the five SDOH domains (Economic Stability, Education Access and Quality, Health Care Access and Quality, Neighborhood and Built Environment, Social and Community Context) identified in Healthy People 2030.

^b Indicates index met inclusion criteria for further analysis.

We selected indices for additional analyses based on the following inclusion criteria: (1) it was designed using nationally available data (i.e., not specific to a particular city or state); (2) it was used by health policymakers or in health‐outcomes research at least three times from 2018 to 2021; and (3) it is available at the census‐tract or block‐group level given well‐established variability with less granular geographic levels. ¹⁹ Indices developed for the purpose of a single study and indices that did not include publicly accessible data were excluded. We used Gephi, an open‐source software, to create a network visualization among the indices that met inclusion criteria (Figure 1).

Figure 1.

Social determinants of health represented across illustrative disadvantage indices. [Colour figure can be viewed at wileyonlinelibrary.com]

Geospatial Mapping

We used publicly available polygon shapefiles of census tracts and block groups of Durham County, North Carolina; Cook County, Illinois; and Orleans Parish, Louisiana, to construct maps of disadvantage risk profiles for each county. In addition to demographic and socioeconomic diversity, the three localities face varied environmental conditions associated with disparate health outcomes. Orleans Parish (which encompasses the city of New Orleans) has a coastal location and low elevation that leave it vulnerable to extreme storm surges and flooding; Cook County (which encompasses Chicago) faces extreme winter temperatures, tornado threats, and erratic weather impacting Lake Michigan levels, which have detrimental impacts on the city; ²⁰ and Durham County (which encompasses Durham) faces extreme heat and high risk of inland flooding. ²¹

We computed Spearman's nonparametric rank correlation to quantify pairwise correlations between indices across counties (Figure A1 in the online appendix). Moran's I was used to quantify autocorrelation and identify clusters of high and low disadvantage. Moran's I values range from −1 to +1 (where −1 indicates dissimilar values clustered together, +1 indicates similar values clustered together, and a value of 0 indicates no autocorrelation, or perfect randomness). ²² To make geospatial mapping comparable between the University of Wisconsin Area Deprivation Index (ADI) and the other indices, block‐group‐level data were used to generate Figure 2. To calculate the local Moran's I statistic in Figure 3, we aggregated ADI to the census‐tract level by calculating population‐weighted mean ADI values for each census tract from block‐group data using the total population count from the five year American Community Survey (ACS) estimate for the years 2015–2019 as weights. Figures were created in R. Code and data can be found at (https://github.com/nb1014/MilibankIndices).

Figure 2.

Maps comparing the Area Deprivation Index (ADI), Social Vulnerability Index (SVI), COVID‐19 Community Vulnerability Index (CCVI), and the Child Opportunity Index (COI) in Durham County, NC; Cook County, IL; and Orleans Parish, LA. [Colour figure can be viewed at wileyonlinelibrary.com]

Increasing levels of disadvantage are indicated in red (higher scores for ADI, SVI, CCVI and lower scores for COI). Because indices include diverse factors to summarize social risk at differing geographic levels in some cases, different indices can show varied patterns of social vulnerability within the same city. The ADI produces the most granular results as it is measured at the block group level, compared to other indices at the census tract level. The CCVI differs from the SVI in including data about health conditions and health care infrastructure. While the SVI and CCVI show similar results in Durham, the patterns of risk in Cook County and New Orleans Parish differ between the two indices. The COI, which aims to map childhood opportunity, shows substantially different patterns than other indices. For example, in Cook County, the COI identifies much larger regions of disadvantage (indicated by lower child opportunity) than all other indices. Gray areas indicate missing data.

Figure 3.

Spatial autocorrelation comparing the ADI, SVI, CCVI, and COI in Durham County, NC; Cook County, IL; and Orleans Parish, LA. [Colour figure can be viewed at wileyonlinelibrary.com]

Global Moran's I test was significant (p<0.05) across all indices. Disadvantage and social risk were not randomly distributed. Positive values for local Moran's I statistic indicate areas of clustering, where adjacent neighborhoods demonstrate similar levels of disadvantage.

Results

We identified 14 indices in the academic and gray literature (Table 1). These indices broadly aimed to summarize how contextual factors shape a community's susceptibility or resilience to risk, with multiple indices initially developed to quantify vulnerability to disasters and hazards (n = 4). ²³ , ²⁴ Indices included different SDOH domains, with all indices including at least three domains, and five indices including all five domains. Each index used different variables to represent SDOH domains, even when using the same data source. All indices incorporate data from publicly available sources to measure population‐level social risk or resilience, with half of the indices using only the ACS data (n = 7) and the remaining half combining data from multiple sources (n = 7). Indices differed in geographic unit granularity: county level (n = 5), census‐tract level (n = 5), zip code level (n = 1), community‐area level (n = 1), block‐group level (n = 1), and primary care service area level (n = 1). Various methods and weighting strategies were used to construct indices, with some indices built using a variable reduction approach ²⁵ (e.g., ADI ²⁶ , University of South Carolina Social Vulnerability Index ²⁷ , ²⁸ ) and others constructed via variable addition (e.g., SVI).

Some indices, such as the Surgo Ventures COVID‐19 Community Vulnerability Index (CCVI), ²⁹ were specifically developed during COVID‐19. Other indices, such as the Public Health Alliance of Southern California Healthy Places Index (HPI) and Chicago Department of Public Health COVID Community Vulnerability Index (Chicago CCVI), were developed to tailor strategies and answer research questions specific to local contexts and communities. The US Office of Minority Health Social Vulnerability Index (MH SVI) was the most recently released of included indices and has not yet been used by policymakers or researchers. ³⁰

Analysis of the Area Deprivation, Child Opportunity, COVID‐19 Community Vulnerability, and Social Vulnerability Indices

Four indices (ADI, COI, CCVI, and SVI) met inclusion criteria for further analysis (see Figures 1 and 2). All indices share at least three SDOH domains (economic, education, and environment) and five indicators (income, poverty, English proficiency, no high school diploma, and unemployment), yet each index also includes a unique set of indicators (see Figure 1). For example, both the SVI and the ADI incorporate economic, education, and community domains, yet neither accounts for neighborhood factors or health care access. The CCVI includes indicators in all domains but, like the ADI and the SVI, only one indicator within education. The Child Opportunity Index (COI) incorporates many unique education and neighborhood factors but just one indicator related to community.

The SVI uses 15 ACS variables across four themes of socioeconomic status; household composition; race, ethnicity, and language; and housing and transportation. The SVI is available at the census‐tract level and was created to allow policymakers to consider social vulnerability in emergency responses (e.g., hurricanes, climate change, or chemical spills). ³¹ During COVID‐19, policymakers have most frequently used the SVI to inform equitable allocation of vaccines and other COVID‐19 mitigation resources; ¹⁰ assess correlations between COVID‐19 incidence and social vulnerability; ³² , ³³ and identify locations for testing and vaccination sites. ¹³ , ³⁴

The ADI uses 17 ACS variables to measure socioeconomic disadvantage at the census‐block‐group level. ³⁵ Researchers have frequently used the ADI to examine associations between social risk and health outcomes. ²³ The SVI and the ADI incorporate different variables that represent similar key SDOH domains, with major differences in geographic unit and exclusion of race and ethnicity in the ADI. ³⁶ During the pandemic, the ADI has been used to understand the relationship between area deprivation and COVID‐19 risk in different localities, ³⁷ assess how COVID‐19 policies impacted communities differentially, ³⁸ and identify communities for focused public health information outreach. ¹⁰ The ADI produces the most granular results, as it measures deprivation at the block‐group level, compared to the other indices, which are reported at the census‐tract level.

The CCVI ²⁹ was developed during the pandemic to identify communities most vulnerable to the effects of COVID‐19 and help policymakers determine where to focus resources. CCVI builds on existing SVI domains with additional data on health conditions and health care infrastructure. This index is the only included index that incorporates indicators in all five SDOH domains. Policymakers have used the CCVI to allocate personal protective equipment to health care facilities and frontline workers, guide allocation of financial resources (e.g., grants) to businesses and organizations, identify vaccine and testing clinic locations, and track COVID‐19 trends. ²⁹

Finally, the COI captures neighborhood resources and conditions that facilitate healthy child development into one composite measure at the census‐tract level. ³⁹ The COI includes 29 indicators across three summary themes: education, health and environment, and social and economic. While all indices include educational factors related to disadvantage (e.g., “no high school diploma”), the COI includes data on the quality of education and academic opportunities (e.g., AP course enrollment, college enrollment). The COI also includes the most detailed information on environmental conditions that impact quality of life and safety (e.g., waste dump sites, ozone concentration). Policymakers have used the COI to connect families to housing resources, allocate funding for safe recreational sites for children, and produce community‐level health needs assessments. ⁴⁰ The COI has also been used to understand how COVID‐19 is exacerbating inequities among children. ³⁹

Spatial Comparisons of Social Risk Indices

Geospatial maps comparing these four indices across Durham County, Cook County, and Orleans Parish illustrate differences across social risk indices (Figure 2). Spearman's rank correlation indicated indices have high pairwise correlation in census‐tract rankings of disadvantage (Figure A1 in the online appendix), with the SVI and the CCVI having the highest positive correlation (between 0.9‐1.0) across counties; the correlation between the CCVI and the ADI (0.7) was slightly lower. The COI showed strong negative correlation with all indices because areas with high child opportunity scores correspond to areas of low disadvantage. The Global Moran's I test was significant (p < 0.05) across all indices, indicating that disadvantage and social risk were not randomly distributed (Figure 3). In addition, positive values for local Moran's I statistic indicate areas of clustering, where adjacent neighborhoods demonstrate similar ADI, COI, SVI, and CCVI values or similar levels of disadvantage. Negative local Moran's values indicate neighborhoods with dissimilar index score values (e.g., neighborhoods with high ADI next to neighborhoods with low ADI).

Durham County's risk profile is similar across the four indices, with concentrated vulnerability in more urban areas and neighborhoods heavily impacted by redlining (Figure 2). ⁴¹ However, the ADI shows more variability in Durham County, identifying fewer block groups in the city center with severe levels of disadvantage compared to other indices. Geographic clusters are evident across indices in Durham County, but local spatial autocorrelation shows differential distributions of clusters, with more census tracts identified proximal to dissimilar census tracts using the SVI, the CCVI, and the COI as compared with the ADI (Figure 3). The SVI and the CCVI show similar local distributions of disadvantage, while the COI shows lower child opportunity (and therefore higher risk) in central and northern Durham County.

Cook County's spatial risk profile differs substantially across the illustrated indices (Figure 2). The ADI highlights greatest disadvantage in the southern part of the county. While the SVI and the CCVI show the most similar patterns of disadvantage, one of Cook County's largest census tracts (representing the South Deering neighborhood on Chicago's far South Side) shows greatest disadvantage using the SVI but moderate disadvantage using the CCVI. Compared to Durham County and Orleans Parish, the patterns of disadvantage in Cook County are more diffuse where neighboring census tracts can display opposite index rankings. Additionally, in Cook County the COI identifies much larger regions of disadvantage (indicated by lower child opportunity) in contrast to the other indices. Local spatial autocorrelation reveals similar clusters of disadvantages across indices for Cook County, albeit with varying degrees of intensity (Figure 3).

Finally, in Orleans Parish relatively similar spatial patterns of disadvantage are observed across the ADI, the SVI, and the COI (Figure 2). In contrast, the CCVI shows high levels of disadvantage in the northeast quadrant and southernmost tip of Orleans Parish, whereas the other indices display relatively low disadvantage in these areas. These areas correspond to areas with fewer health care clinics and hospitals than other parts of the parish, areas with high proportions of individuals with diagnosed comorbidities, and/or proximity to high‐risk workplace environments, including nursing homes, meatpacking plants, or chemical factories. ⁴² Local spatial autocorrelation analysis reveals neighborhoods in the northeast quadrant are more likely to have dissimilar disadvantage rankings when utilizing the SVI or CCVI as compared to the COI.

Discussion

Prompted by the rapid uptake of disadvantage indices by policymakers to promote equity in public health decision making during the COVID‐19 pandemic, we compared existing indices and used geospatial analyses to evaluate the use of indices across various geographic areas. We found 14 currently available disadvantage indices in the literature, all developed with varied objectives, methods, and indicators of social risk. While many disadvantage indices have been used interchangeably or for similar aims, to our knowledge researchers and policymakers have not selected indices in a strategic manner or compared the differential impact of choosing a particular index on policy decisions or health outcomes. Our analyses indicate that indices differ substantially in the variables and SDOH domains included, as well as the underlying data sources. Our spatial analyses suggest that index selection can impact how disadvantage is interpreted within the same geographic area, particularly in the local context. Although overall index scores are correlated, because Spearman's correlation ranks census tracts for each index, ranks may be similar between indices even if absolute values are different. Spatial autocorrelation analysis demonstrated clustering of disadvantage, with local analyses showing that adjacent neighborhoods can have similar levels of disadvantage with one index, but opposite findings when using another index. The implications of the margins of difference between index scores when considering health equity in policy decisions will depend on the policy objective. However, the differences shown here in spatial clustering of social risk depending on choice of index could lead to varied interventions, policy responses, and outcomes. These results illustrate key methodological challenges and evidence gaps that require further consideration by researchers, public health officials, and policymakers as indices are further incorporated into efforts to improve health equity and redress injustice.

Methodological Challenges: Underlying Data Source, Index Construction, and Ecological Fallacy

Underlying Data Source

Nearly all the indices identified here use ACS data, and Figure 1 shows the overlap of variables across indices. The ADI and the SVI rely fully on ACS data, while the CCVI and the COI incorporate additional data sources and indicators. Although the ACS is the most extensive national source of demographic and economic data currently available, there are critical challenges with ACS data and data collection. ⁴³ , ⁴⁴ The ACS uses imprecise small‐area estimates that often lead to large margins of error and uncertainty, rendering the data unreliable. ⁴⁵ Spielman and Folch have demonstrated key drivers of uncertainty in ACS data and proposed strategies for ameliorating some of these concerns, including development of an open‐source spatial optimization algorithm to improve the usability of ACS and other survey data. ⁴⁶ Although indices that use only ACS data might have the same relative uncertainty, indices with other data sources (such as the COI or the CCVI) are not directly comparable. As shown in our analyses, the risk profiles generated by the COI in particular can be substantially discrepant from those derived from ACS‐dependent indices, making interpretation of these values nearly impossible. Additionally, because indices are typically updated every two to five years, dynamic neighborhood changes (such as those caused by gentrification) are lagged in indices by several years. Continued testing and refinement of strategies such as those proposed by Spielman and Folch can generate needed evidence to improve the utility of existing data sources and improve the interpretation of disadvantage indices. ⁴⁶

Beyond improving the utility of underlying data sources, improved ingredients are needed to build better indices. One important challenge in the application of disadvantage indices to identify health inequities and inform strategies to generate measurable improvements in health outcomes is capturing the impact of structural racism and other forms of discrimination that drive social disadvantage. Race and ethnicity are captured in some indices (e.g., Baseline Risk Indicators for Communities, SVI, MH SVI) as proxies for the way disadvantage is stratified by race and ethnic background. Some critics raise ethical and legal concerns about including race and ethnicity indicators in disadvantage indices, especially as they are applied in resource allocation decision making. ⁴⁷ However, a simulation comparing the ADI versus the SVI under the NASEM guidelines for equitable COVID‐19 vaccine allocation showed that while the ADI was better at identifying specific geographic areas, it yielded lower vaccine shares to historically marginalized populations than the SVI. ³⁶ The inclusion or exclusion of variables such as race can differentially affect public health decision making and, in turn, health outcomes.

To address some of the gaps noted in the SVI during the pandemic, the US Office of Minority Health recently released the MH SVI. This adaptation of the original SVI expands the Minority Status and Language theme to include specific data for race, ethnicity, and languages as well as factors associated with COVID‐19 outcomes and other public health themes. By including factors that focus explicitly on minoritized populations and medical vulnerability, the MH SVI aims to fill key contextual gaps in the current SVI, but whether it fulfills this objective has not yet been assessed. ¹⁶ Despite the inclusion of indicators related to race and ethnicity, conflating race and structural racism or structural disadvantage can be, and often is, problematic. ⁴⁸ , ⁴⁹ , ⁵⁰ Distinguishing between race and racism is not yet well operationalized and tools that explicitly measure racism are not widely adopted.

Index Construction

Beyond the data source used, as shown in the network analysis here, existing disadvantage indices include differing indicators to measure disadvantage. However, not all metrics are sensitive predictors of health disparities, and the sensitivity of an index will depend on the intended objective of analysis. For example, the CCVI and the SVI share nearly all the same indicators, with the exception of “percentage uninsured” and all health care indicators. The CCVI was developed by adapting the SVI to include additional data on comorbidities and health care access to provide more robust information about community vulnerability to COVID‐19. ⁵¹ As the SVI was originally developed to measure social vulnerability in the contexts of hazards or natural disasters, it is not surprising that additional factors are needed to assess vulnerability to COVID‐19. While SVI and CCVI scores are often similar, nearly a quarter of US county scores differ, with the CCVI showing greater ability to identify communities exhibiting vulnerability to COVID‐19 transmission than the SVI. This is well depicted in our geospatial analyses, as the CCVI depicts higher levels of social disadvantage in some census tracts (e.g., Cook and Orleans counties, Figure 2) as compared to the SVI, reflecting the incorporation of health‐specific vulnerabilities. Similarly, the COI, which is designed to illustrate spatial patterns of child opportunity, incorporates a number of different variables and data sources compared to the other included indices, and it shows substantially different patterns of disadvantage and clustering of disadvantage compared to other indices. In addition to using differing indicators of social vulnerability from multiple sources, indices can also be constructed using different methods. For example, the ADI ³⁵ , ⁵² and the COI ⁵³ use weighting strategies that are based on factor analysis, while the SVI ⁵⁴ and CCVI ⁵⁵ use a variable additive approach with all variables weighted equally. A robust demography literature compares these different methods of index construction, input variables, and region of application, with repeated calls for nuanced interpretation and caution in applying and comparing disadvantage indices. ²⁵ , ²⁸ , ⁵⁶

Ecological Fallacy

Less granular measures are useful for understanding county‐level trends, yet these metrics can also mask health‐related disparities among heterogeneous populations, potentially leading to inappropriate or even harmful research or policy decisions. ⁵⁷ This bias, or ecological fallacy, results when incorrect inferences about individuals are made based on aggregate data derived from broad geographic levels. ⁵⁸ , ⁵⁹ A manifestation of ecological fallacy extensively described in the geographic literature is the modifiable areal unit problem (MAUP). ⁶⁰ , ⁶¹ , ⁶² , ⁶³ The MAUP describes the source of statistical bias that occurs through the differential spatial aggregation (by modifying either scale or zone) of point‐based measures. ⁶⁴ , ⁶⁵ , ⁶⁶ Disadvantage indices are clear case examples for exploring the bias created by the MAUP, as these indices aggregate spatially distributed social data represented at different geographic levels. In our analysis, the potential for different geographic units to lead to varied interpretations of the same data can be seen when comparing differences in social risk patterns between the ADI, which measures risk at the more granular block‐group level, and indices that measure risk at the census‐tract level.

In another example of ecological fallacy, disadvantage index scores can prompt incorrect inferences from health care providers about individuals in communities. For example, while census tracts with high SVI scores may have higher rates of COVID‐19 transmission, health care providers cannot assume that individuals seeking care from these areas have high levels of social risk. As health systems consider integrating disadvantage indices into electronic health records, avoiding such assumptions and evaluating whether providers perpetuate systemic racism, discrimination, and stigma when using these data should be considered. Bias can also occur if index measures are poor proxies for underlying constructs. For example, neighborhoods surrounding large hospital systems may rank high on indices that measure access to care based on the number of providers in an area. Yet, this indicator captures availability of providers and may not translate into access to care for historically marginalized populations. Similarly, indices that include factors such as race, ethnicity, or language may perform differently in majority‐minority neighborhoods or areas with large immigrant or refugee populations.

Approaches for Improving Use of Disadvantage Indices

Given our analytic findings and the challenges outlined in this paper, applying disadvantage indices requires a thoughtful and nuanced approach, as indiscriminate application may unintentionally worsen equity. Here we discuss considerations for improving the use of disadvantage indices and synthesize recommendations and strategies to address aforementioned methodological challenges by different stakeholder groups (Table 2). When planning to utilize a disadvantage index, users must consider several factors, including factors that drive an outcome of interest and community‐specific demographic patterns. As indices synthesize information differently, users must critically assess the implications of inclusion or exclusion of variables, and how these can differentially affect decision‐making. Similar to our spatial analyses, policymakers might consider comparing more than one index before making key decisions. Developing long‐term strategies that promote longitudinal community feedback to supplement and critically assess the use of indices should also be a priority in planning for future public health emergencies.

Table 2.

Recommendations and Strategies to Address Methodological Challenges When Choosing and Using Disadvantage Indices

	Methodological Challenge
	Data Source	Index Construction	Ecological Fallacy
For All Disadvantage Index Users
When choosing index, consider factors driving outcome of interest and geographic levels most suitable for a given objective, intervention, or outcome	✓		✓
Assess implications of using indices that include different variables but apply caution when making comparisons across indices with varied data inputs	✓
Consider impacts of demographic changes (e.g., gentrification or gerrymandering) on interpretation of index results given typical rigid geographic boundary definitions			✓
Utilize adapted indices (e.g., Minority Health SVI, Healthy Places Index, Chicago CCVI) or consider local adaptations to incorporate contextualized variables or variables that most effectively represent root causes of disadvantage	✓	✓	✓
For Policymakers
Compare more than one index before making key public health decisions or implementing policy		✓
Ensure appropriate interpretation of neighborhood‐level disadvantage (e.g., through analyzing local clustering) before making allocation decisions or implementing interventions			✓
Develop processes that streamline inclusion of community feedback to supplement existing data and incorporate PGIS approaches during development of public health response strategies	✓		✓
Authorize capacity building and funding for the community health workforce and convene community members to support further development or refinement of disadvantage indices	✓	✓
Streamline the use of disadvantage indices based on application to guide allocation or intervention implementation in tandem with other public health initiatives (e.g., education and outreach programs)			✓
Create incentives for cross‐sector data sharing to improve interoperability of indices and support integration of public health, health care, and social services infrastructure			✓
For Researchers
Use methods such as spatial autocorrelation to examine the impact of longitudinal neighborhood‐level and structural changes not captured in underlying data sources (e.g., historical redlining, food and pharmacy deserts)	✓		✓
Develop evaluation framework to assess index performance and inform index adaptation. Key considerations include whether an index measures dimensions it aims to; produces consistent, replicable results across contexts; and generates results that align with similar indices		✓
Clarify key domains and indicators most salient for predicting risk and resilience to negative health outcomes and identify measurements of structural racism outside of solely race and ethnicity indicators		✓
Identify applications that might benefit from additional contextual data (e.g., policy responses for neighborhoods with large immigrant populations) and evaluate the impact of including such data	✓	✓
Address issues of imperfect “ingredients” through continued refinement of existing data sources by combining geographic units in novel ways (e.g., regionalization methods proposed by Spielman et al.) or linking individual‐level (e.g., census microdata) information to community‐level data	✓		✓

Some steps have been taken on a community‐by‐community basis to adapt available indices for specific contexts and address some of the aforementioned gaps. For example, while most US states and localities have used the SVI during the pandemic, presumably following NASEM and CDC recommendations, a few localities have used community‐derived data related to COVID‐19 vulnerability to provide a more comprehensive picture of health inequities. The Chicago CCVI and California's HPI are two indices that have been tailored to specific geographic contexts. The Chicago CCVI was adapted from the CCVI, and it includes the proportion of essential workers, proportion of adults with obesity or diabetes, and community data on COVID‐19 burden. ⁶⁷ Policymakers used the index to identify Chicago neighborhoods to allocate vaccine supply and resources, and to align outreach and mitigation strategies with identified barriers to vaccine uptake. ⁶⁸

Similarly, California updated the HPI, a California‐specific index developed to help policymakers and communities allocate resources to areas with the greatest disadvantage, during the pandemic. ⁶⁹ Policymakers added indicators like health insurance access, race and ethnicity, chronic conditions, and hospital bed availability and used the adapted tool to allocate a large portion of California's vaccine supply to areas experiencing disadvantage. ⁷⁰ In both Chicago and California, contextual data were used to develop locally relevant social risk scores. ⁶⁹ Further research is needed to assess the impact and effectiveness of incorporating additional data elements compared to utilizing existing indices, and identify situations in which including additional contextual data can promote public health measures and more equitable policy decisions.

It is incumbent upon index users to decide which geographic level is most suitable for a given objective, potential intervention, or outcome of interest. Returning to the MAUP and the issue of ecological fallacy, users should carefully consider how the level of geographic analysis will impact interpretation of results when using disadvantage indices. Resolving biases that result from the aggregation of spatial data requires critically appraising geographic scope and using the smallest groupings, where possible, to examine variables for mapping. ⁷¹ Before making allocation decisions or implementing interventions, policymakers should analyze local clustering and context to ensure appropriate interpretation of neighborhood‐level disadvantage.

Users should also consider that geographic boundary definitions might not be sensitive to demographic changes due to displacement or sociopolitical processes like gentrification or gerrymandering. Further, as noted in other research, many patterns of social risk overlap with historic redlining practices and racial housing segregation. ⁷² Researchers can adopt longitudinal mapping methods to examine the impact of including information around factors such as neighborhood‐level changes and structural circumstances (e.g., gentrification, historical redlining, food and pharmacy deserts) that are not captured in data sources like the ACS. For example, researchers can use spatial autocorrelation methods to track longitudinal changes in neighborhoods by identifying whether clusters of disadvantage shift or disappear, which may provide information on the impact of gentrification on the outcomes of index‐based measures of social disadvantage. This information may identify areas of high disadvantage or vulnerability and could increase sensitivity of existing indices, but it will not be necessary or helpful in all situations.

Furthermore, approaches that construct geographic boundaries without community input may overgeneralize homogeneity among populations and may not reflect how communities define themselves. To address the need for contextualized information, policymakers and researchers can develop strategies to longitudinally engage communities in decision making through methods such as participatory geographical information systems (PGIS). Policymakers have used PGIS to prepare for environmental disasters and address structural and social determinants of health by mapping assets and barriers. ⁷³ Building capacity for the community health workforce and convening community members can further support the development or supplementation of disadvantage indices.

As noted earlier, although disadvantage indices include indicators that are strong proxies for factors that differentially disadvantage (or advantage) some communities over others, ⁶⁹ a nuanced approach is needed to understand which indicators measure structural barriers and pathways that influence key outcomes. Such an approach requires choosing the right index according to objectives (i.e., using the COI instead of the SVI to understand how structural racism impacts early childhood development), supplementing index data with context‐dependent information (e.g., including information from community stakeholders about neighborhood‐level factors), and accounting for real‐time changes (e.g., monitoring trends in communities). After careful selection and adaptation of an appropriate index, risk scores should be used not just for resource allocation but also in tandem with public health initiatives. For example, during the pandemic some states have implemented specific education and outreach programs, understanding that resource allocation alone does not always translate to resource uptake. ³⁴

In addition to strategies to augment strength of existing indicators in disadvantage indices, it is important for researchers to explore approaches for measuring structural racism outside of just race and ethnicity indicators, especially if the goal of applying an index is to guide equitable decision making. A growing body of work is helping to elucidate methodological and analytical approaches for measuring structural racism and its deleterious impacts on health. ⁷⁴ , ⁷⁵ , ⁷⁶ , ⁷⁷ A recent commentary scopes relevant challenges and provides recommendations for epidemiologists and other health researchers for measuring structural racism. ⁴⁹ Strategies proposed include linking interdisciplinary variables in data sets, using mixed‐methods approaches, and using life course theory in analytical approaches. ⁴⁹

Importantly, long‐term efforts to improve geographical information systems and the use of disadvantage indices may call for a paradigm shift in how we understand regional difference and group communities together. Addressing issues of imperfect ingredients, such as uncertainty with ACS data, will require continued testing of innovations such as combining geographic units in novel ways, as explored in other research. ⁴⁶ Balancing the need for contextually derived information at specific geographic units of analysis while also strategically aggregating data to address uncertainty is no simple feat but is an important aim for improved reliability and utility of disadvantage indices. One proposed recommendation for addressing this tension, and the related ecological fallacy, is adopting a “bottom‐up” approach that treats area‐level vulnerability as an emergent property of individual‐level vulnerability. ⁷⁸ For example, linking census microdata (individual‐level information) to community‐level data at the census tract can help identify typologies of individuals within communities and how distinct communities are made vulnerable to certain diseases, disasters, hazards, or sociopolitical disadvantages. ⁷⁸

Ultimately, an important future aim is developing an evaluation framework to assess disadvantage index performance in different contexts, as well as specific strategies for adaptation of indices when indicated. Such a framework should consider whether a given index measures the dimensions it aims to; produces consistent results longitudinally and across contexts; involves replicable and transparent methodology; produces interpretable results; is relevant to the objective of interest; and produces results that align with similar indices. ⁷⁹ To improve strategic use of the many existing indices, researchers should carefully elucidate key domains and indicators across indices that are most salient for predicting risk and resilience to negative health outcomes. Such efforts could lead to a standard index supplemented with additional community‐derived data depending on the objective, context, or health outcome of interest. Other benefits could include interoperability across indices and cross‐sector data systems to support needed integration of public health, health care, and social services infrastructure, and allow for better triage of finite resources.

Limitations

This study has several limitations. Although we reviewed a wide range of indices, we did not include many indices that were developed for one‐time usage, and we limited our geospatial analyses to the four indices that met additional criteria. These four indices were available at more granular geographic units such as the census‐tract and block‐group levels. While not as granular, other geographic levels of analysis (such as county or zip code levels) are important for different objectives beyond the scope of this work. We also considered these indices in the context of relatively urban areas. Comparisons of these indices may differ when considering more rural settings, as some constructs (e.g., multiunit housing and crowding) may be less applicable.

Conclusion

Existing health inequities exacerbated by the pandemic reemphasized the need for tools to quantify health disparities and systemic barriers to health care. The rapid and nearly ubiquitous uptake of disadvantage indices indicates that policymakers, health systems, and public health leaders are invested in implementing policies and tools to improve health equity. Evidence during the COVID‐19 pandemic suggests that use of these tools can inform more equitable policies that prioritize historically marginalized communities. However, as we demonstrate in these analyses, simply accepting the outputs from disadvantage indices as “truth” is problematic, as indiscriminate use of these tools could have unintended consequences.

As we move forward in further integrating indices for use beyond the pandemic, a one‐size‐fits‐all approach will not suffice. Although researchers and policymakers can use indices as one of several available tools, it is critical to understand how structural disadvantage manifests at local levels prior to implementing policies or interventions. Future analysis should investigate robust indicators for disadvantage and resilience, and develop rigorous strategies for using these tools in tandem with measuring the impact of structural racism and other underlying social disadvantage systems. The aim of our analysis was not to provide a clear “winner” in terms of index selection, because as we have explored here, index choice depends on a number of factors, including intended use, populations of interest, and geographic area of interest. However, through this examination of disadvantage indices and synthesis of important methodological challenges with their application, we hope that policymakers and researchers will be better equipped to employ disadvantage indices to promote equity and social justice moving forward.

Funding/Support: Dr. Cholera was supported by the Eunice Kennedy Shriver National Institute of Child Health and Human Development (K12‐HD105253).

Acknowledgements: We thank Dr. Eric Monson from Duke University Center for Data and Visualization Sciences for assisting us with data visualization, Hemi Tewarson from the National Academy for State Health Policy, and Mark McClellan, Gillian Sanders Schmidler, Charlene Wong, and Dave Anderson from the Duke‐Margolis Center for Health Policy for their review and guidance.

Potential Conflict of Interest Disclosures: None.

Supporting information

Figure A1. Spearman's non‐parametric rank correlation in Durham County, NC; Cook County, IL; and Orleans Parish, LA. Indices have high pairwise correlation in census track rankings of disadvantage. COI showed strong negative correlation with all indices because areas with high child opportunity scores correspond to areas of low disadvantage.

Abbreviations: Area Disadvantage Index (ADI); COVID Community Vulnerability Index (CCVI); Child Opportunity Index (COI); Social Vulnerability Index (SVI)