Comparison of the Social Vulnerability Index, Area Disadvantage Index, and Social Deprivation Index for Adults With Out-of-Hospital Emergencies

Abstract

Objectives:

Multiple measures are used to assess neighborhood disadvantage. Although each was created with a unique purpose, they share conceptual and methodological overlap. We examined the correlation among 3 measures of neighborhood disadvantage (Social Vulnerability Index [SVI], Area Disadvantage Index [ADI], and Social Deprivation Index [SDI]) and their association with quality indicators for adults with prehospital emergencies.

Methods:

We performed a retrospective analysis using a national multiagency emergency medical services (EMS) database, including emergency scene encounters for adults (aged ≥18 y) with available census-tract SDI and SVI data and census-block ADI data from January 1 through December 31, 2023. We compared the SVI, ADI, and SDI using overall and pairwise intraclass correlation coefficients (ICCs). We evaluated the association of each index with 7 quality indicators for prehospital care developed by the National EMS Quality Alliance.

Results:

We included 9 259 983 encounters (median [IQR] age, 63 [44-77] y). The overall ICC between indices was 0.65, indicating moderate agreement. We found higher agreement between SVI and SDI (ICC = 0.84) than between SVI and ADI (ICC = 0.54) or ADI and SDI (ICC = 0.59). We found overlap among the indices for most outcomes, although we found differences in ADI associations compared with SVI and SDI for some outcomes. These included bronchodilator use in asthma (SDI/SVI positively associated, ADI not associated), treatment of hypoglycemia (ADI negatively associated, SDI/SVI not associated), and screening of suspected stroke (SDI/SVI negatively associated, ADI not associated).

Conclusions:

We found moderate agreement among 3 commonly used indices of neighborhood disadvantage. Research is needed to refine the application of these indices to prehospital care and explore their utility in reducing health disparities across health care settings.

Social drivers of health play an important role in emergency medical services (EMS) care. Research has demonstrated associations between social drivers of health and EMS care provision,^1,2 access,^3-5 and overall health disparities.^6,7 Patient-level measures, such as income or health insurance type, are often used as a proxy for socioeconomic status. However, broader structural factors in a community also play an important role in health outcomes. ⁸ Community-level measures aggregate indicators of wealth, income, education, employment, and other social factors across various populations or geographies. These measures are used to examine the influence of community-level factors on health outcomes and are often derived from publicly available data sources, facilitating opportunities for their investigation. ⁸

Multiple community-level indices of social drivers of health have been described and are widely used in disparities research.^9-15 The intended role of these indices differs, but variables used in their development overlap. The various indices of community-level disparity or vulnerability incorporate variables such as educational attainment, poverty level, homeownership, income, and employment. While these measures are frequently used in research on disparities in the provision of prehospital care,^1,5,7,16 a rationale for selecting a specific index is frequently lacking. Given reports of disparities in care based on neighborhood disadvantage in other contexts,^2,5,17,18 using standard evidence-based measures of prehospital care may be a starting point to evaluate differences in care between commonly used indices of neighborhood disadvantage.

Community-level measures of disadvantage influence EMS quality indicators for various reasons, including access to care, resource allocation, clinician perceptions, and differences in health-seeking behavior.^19-22 Because many of these indices overlap with respect to economic, educational, housing, and environmental factors; household structure; transportation; health insurance coverage; and racial and ethnic minority status and language status, ⁸ it may be postulated that these variables are correlated with each other. However, while similarities exist, the exact composition of commonly used measures of community disparity demonstrates distinct differences. Direct comparisons of these indices are limited in the literature.^22-24 In this study, we sought to compare the overlap among 3 commonly used indices of neighborhood disadvantage among adults with prehospital emergencies. Second, we sought to evaluate and compare their association with National EMS Quality Alliance (NEMSQA) prehospital quality indicators.

Methods

Data Source

We conducted a retrospective analysis of EMS encounters using patient medical record data from the ESO Data Collaborative. ESO is a large provider of electronic health record (EHR) software in the United States. This EHR software complies with the standards of the National Emergency Medical Services Information System and encompasses details about service dispatch, patient demographic characteristics, clinical presentations, assessments, and medical interventions administered by EMS personnel. ESO compiles a standard dataset annually for academic research, which is accessible at no cost upon submission of a study proposal, data use agreement, and institutional review board (IRB) approval. For the present study, we analyzed data from the 2023 research dataset, which includes records from more than 3000 EMS agencies. This study was approved by the Ann & Robert H. Lurie Children’s Hospital of Chicago Institutional Review Board (IRB) with a waiver of the requirement for informed consent (IRB no. 2024-6685).

Inclusion

We included EMS encounters for emergency scene responses involving adults (aged ≥18 y), excluding those with missing age information and non-911 scene responses (eg, interfacility transports, standby, canceled, scheduled transports, scene assists [defined as support from additional personnel or resources at the encounter site]). From this sample, we further restricted the analysis to encounters with available Social Deprivation Index (SDI), Social Vulnerability Index (SVI), and Area Deprivation Index (ADI) data for the scene location of the EMS encounter.^9,10,14,15

Exposure

We assessed 2 census tract–level indices of community disadvantage (the SDI and the SVI, both published in 2018) and the census-block index of ADI (published in 2022; Table 1). Although multiple other criteria for area-level disadvantage have been reported, ⁸ we selected these indices because they are broadly applicable (ie, not designed for a specific population) and because they are among those most frequently used in the scientific literature. The SDI is intended to measure area-level deprivation to quantify socioeconomic variation in health outcomes. ¹⁵ The SVI supports emergency response planners and public health officials by pinpointing areas likely to need assistance in preparing for, responding to, and recovering from emergencies. ¹⁴ The ADI focuses on block-level socioeconomic deprivation to guide program planning, health care delivery, and policy making.^9,10 These indices were provided in the ESO Data Collaborative research datasets for the scene of each event, using an honest broker process to ensure data deidentification. We used the z-scored SDI, the National Rank of the ADI, and the overall ranked percentile of the SVI. To facilitate comparisons, we rescaled each index among the included encounters to have a mean of 0 and an SD of 1. With all scores, a higher z score indicated greater disadvantage.

Table 1.

Summary of variables used to construct neighborhood disadvantage indices in the United States, 2023 ^a

Indicator	Area Deprivation Index	Social Deprivation Index	Social Vulnerability Index
No. of indicators	17	7	16
Education
Percentage of the block group’s population aged ≥25 years with <9 years of education	X
Percentage of population aged ≥25 years with <12 years of education		X	X
Percentage of population aged ≥25 years with ≥high school diploma	X
Employment and income
Percentage of employed people aged ≥16 years in white-collar occupations	X
Percentage of nonemployed for population aged 16 to 64 years		X
Median family income	X
Housing cost burden			X
Income disparity	X
Percentage of civilian labor force population aged ≥16 years unemployed (unemployment rate)	X		X
Percentage of families below the federal poverty level	X	X
Percentage of population below 150% of the federal poverty threshold	X		X
Housing and resources
Median home value	X
Median gross rent	X
Median monthly mortgage	X
Percentage of owner-occupied housing units (home ownership rate)	X
Percentage of households living in renter-occupied housing units		X
Percentage of occupied housing units without a motor vehicle	X	X	X
Percentage of occupied housing units without a telephone	X
Percentage of occupied housing units without complete plumbing	X
Percentage of occupied housing units with >1 person per room (crowding)	X	X	X
Percentage multiunit structures			X
Percentage mobile homes			X
Percentage of people in group quarters			X
Household characteristics
Percentage of single-parent households with children aged <18 years	X	X	X
Percentage of people aged ≥65 years			X
Percentage of people aged ≤17 years			X
Percentage of people aged ≥5 years with a disability			X
Percentage of people with racial and ethnic minority status			X
Percentage of people aged ≥5 years who speak English less than “well” (English language proficiency)			X
No health insurance			X

^aEach index incorporates a unique set of indicators reflecting varied sociodemographic characteristics, primarily derived from the US Census and American Community Survey data.^9,10,14,15

Variables

From the EMS record, we used the following variables: demographic characteristics, including age, sex, race and ethnicity (categorized as Hispanic or Latino, non-Hispanic Black, non-Hispanic White, other or multiracial, or unknown), and scene location. We evaluated whether the encounter involved a basic life support or advanced life support/critical care unit. We classified encounter urbanicity by using definitions from the Centers for Medicare & Medicaid Services, categorizing locations as urban, rural, or super rural. ²⁶ We extracted data on transport status, assessments, medication administered, and procedures.

Outcomes

We analyzed several outcomes to characterize EMS care delivery, both overall and for common prehospital conditions. The NEMSQA measures help EMS agencies assess the quality and safety of care using standardized definitions. ²⁵ We selected 7 NEMSQA measures for 6 conditions: asthma, trauma, nontransport (patient refusals, in which the patient declines transport to the hospital), seizure, hypoglycemia, and stroke. We selected these measures because of their common occurrence in prehospital settings and because of prior work suggesting an association between these measures and the social determinants of health.^2,7,27-32 Specifically, the studied NEMSQA measures included the provision of bronchodilators for patients with asthma (NEMSQA measure Asthma-01), documentation of a pain score in injured patients (Trauma-01), improvement in pain for injured patients (Trauma-03), documentation of vital signs in patients who were not transported by EMS (TTR-01), provision of benzodiazepines for patients with suspected status epilepticus (Seizure-02), provision of glucose for patients with hypoglycemia (Hypoglycemia-01), and performance of a stroke assessment in patients with a suspected stroke (Stroke-01) (Table 2).

Table 2.

EMS quality metrics used as outcome measures in EMS agencies in the 2023 ESO Data Collaborative ^a

Case scenario	Measure description	Outcome
Asthma-01	Percentage of EMS responses originating from a 911 request for patients with a diagnosis of asthma who had an aerosolized beta-agonist administered	Provision of levalbuterol, albuterol, alupent, atrovent, combivent, duoneb, ipratropium, xopenex
Trauma-01	Percentage of EMS responses originating from a 911 request for patients with injury who were assessed for pain	At least 1 pain score
Trauma-03	Percentage of EMS transports originating from a 911 request for patients whose pain score was lowered during the EMS encounter	Last pain score is less than first pain score
TTR-01	Percentage of EMS responses originating from a 911 request for patients not transported by EMS during which a basic set of vital signs is documented	Documented systolic blood pressure, diastolic blood pressure, heart rate, respiratory rate
Seizure-02	Percentage of EMS responses originating from a 911 request for patients with status epilepticus who received benzodiazepine during the EMS response	Provision of diazepam, midazolam, midazolam infusion, lorazepam, lorazepam infusion, valium, versed, versed drip, or ativan
Hypoglycemia-01	Percentage of EMS responses originating from a 911 request for patients with symptomatic hypoglycemia who received treatment to correct their hypoglycemia	Glucagon, glucose tablets, oral glucose, dextrose 10%, dextrose 25%, dextrose 5%, dextrose 50%
Stroke-01	Percentage of EMS responses originating from a 911 request for patients who had a suspected stroke who had a stroke assessment performed during the EMS response	Completion of any stroke assessment form

Abbreviations: EMS, emergency medical services; ID, identification; TTR, treated, not transported record.

^aTable includes quality metrics from the National EMS Quality Alliance (NEMSQA) used as outcome measures to assess prehospital care in EMS responses initiated through 911 calls. The “measure description” refers to the goal of each quality metric, and the “outcome” describes the measured action or finding used to determine compliance. Quality measures reflect treatment and documentation practices across clinical scenarios such as asthma, trauma, hypoglycemia, seizure, stroke, and nontransport encounters. Data source: NEMSQA Clinical and Operational Performance Measures. ²⁵

Analysis

We described encounter-level demographic characteristics. To assess the agreement between the continuous measures (both overall and pairwise), we calculated a fixed-effects intraclass correlation coefficient (ICC), interpreting the agreement as follows: <0.5 as poor, 0.5 to 0.75 as moderate, >0.75 to 0.9 as good, and >0.9 as excellent. ³³ We evaluated the correlation of the nontransformed indices with each other by using the Spearman rank-correlation test. For each NEMSQA measure, we built logistic regression models using each community index and a linear tail-restricted cubic spline function. We chose this approach, as opposed to categorizing each index into predefined units, to preserve as much information as possible and to detect nonlinear associations between each index of neighborhood disadvantage and outcomes. Models were fitted using maximum likelihood estimation with 5 knots, with the first and last knots positioned at the 5th and 95th percentiles. We accounted for clustering by applying robust variance estimation using a sandwich estimator. We visualized the relationships between neighborhood indices and NEMSQA measures using spline plots and reported representative odds ratios (ORs) with 95% CIs, comparing the third quartile with the first quartile of each outcome. We identified the best-fitting index for each performance measure as the one having the lowest Akaike information criterion (AIC). ³⁴ As a sensitivity analysis, we repeated this approach by restricting our sample to EMS encounters that occurred in a private residence, given that the studied indices were determined based on the incident location. We conducted analyses using the rms (version 6.7-0) and psych (version 2.3.6) packages in R version 4.3.2 (R Foundation for Statistical Computing).

Additional Analyses

First, as an exploratory analysis, we compared the distribution of EMS encounters between the 2018 (for SVI and SDI) and 2022 (for ADI) ESO datasets and the 2023 dataset ³⁵ to assess potential changes in these measures between their publication year and the present analysis. We visualized differences between the studied measures using density plots. Second, we performed a falsification test to assess for differences among a variable not expected to be correlated. Given that many variables may show spurious associations with neighborhood advantage in large datasets, we selected several dichotomous variables that were expected to have minimal or no association with neighborhood advantage: documented patient pulse, respiratory rate, and Glasgow Coma Scale. ³⁶

Results

Inclusion

We identified 13 957 073 encounters in the 2023 ESO Collaborative dataset. After excluding encounters with missing age (n = 1 810 771), pediatric encounters (n = 722 549), and those not classified as scene responses (n = 1 393 975), we retained 10 029 778 encounters. Of these, 2.8% were missing SDI data, 3.1% were missing SVI data, and 7.4% were missing ADI data. After excluding records with any missing neighborhood disadvantage data, our analysis comprised a final sample of 9 259 983 encounters (93.6% of total). The median (IQR) age of included encounters was 63 (44-77) years, with 52.5% of encounters for female patients (Table 3).

Table 3.

Demographic and encounter characteristics of adult emergency medical services scene responses in the United States, 2023^a,b

Characteristic	No. (%) a (N = 9 259 983)
Age, median (IQR), y	63 (44-77)
Sex
Female	4 858 460 (52.5)
Male	4 374 966 (47.2)
Other	4629 (<0.1)
Missing	21 928 (0.2)
Race and ethnicity
Hispanic/Latino	630 485 (6.8)
Non-Hispanic Black	1 797 308 (19.4)
Non-Hispanic White	5 677 322 (61.3)
Other or >1 race	260 332 (2.8)
Missing	894 536 (9.7)
Level of care
Basic life support	1 638 179 (17.7)
Advanced life support/critical care	7 448 737 (80.4)
Missing	173 067 (1.9)
Disposition
Nontransport	1 267 413 (13.7)
Transport	7 992 570 (86.3)
Location
Home/residence	5 102 191 (55.1)
Street or highway	844 618 (9.1)
Place of business	582 256 (6.3)
Nursing home	593 869 (6.4)
Other	2 136 500 (23.1)

^aData source: ESO Data Collaborative. ³⁵ Variables include age, sex, race/ethnicity, level of care provided, patient disposition, and encounter location. Values are presented as number (percentage) unless otherwise indicated. Missing values are reported where applicable.

^bValues are number (percentage) unless otherwise indicated.

Index Agreement

Overall, the ICC among the 3 indices was 0.65, indicating moderate agreement. In the evaluation of pairwise agreement, agreement was higher between the SVI and SDI (ICC = 0.84) than between the ADI and SDI (ICC = 0.59) and the SVI and ADI (ICC = 0.54). Findings were consistent when we evaluated the correlation among the nontransformed indices, with the SVI and SDI having greater correlation than the other pairs. The Spearman rank correlation test showed the correlation between the SDI and SVI as 0.84, the correlation between the SDI and ADI as 0.50, and the correlation between the SVI and ADI as 0.46 (eFigure 1 in Supplemental Material).

Association of Community-Level Disadvantage Indices With NEMSQA Measures

Performance was highest in the Stroke-01 (84.0%) and Hypoglycemia-01 (83.4%) case scenarios and lowest in the Trauma-03 (17.9%) case scenario (Table 4). We identified an association between disadvantage indices and the NEMSQA quality outcomes for most measures (Figure). When we used the AIC to evaluate the best-fitting models, the ADI had better goodness of fit than the other indices did for scenarios of Asthma-01, Trauma-01, Seizure-02, Hypoglycemia-01, and Stroke-01. The SDI had better goodness of fit for Trauma-03 and TTR-01. When we evaluated indices as representative ORs (comparing the third quartile with the first quartile of the disadvantage variable), none of the indices were associated with the provision of benzodiazepines for status epilepticus or pain assessment in trauma. In addition, the ADI was not associated with the provision of bronchodilators with asthma, vital signs assessment of nontransports, or screening of stroke. The SVI was not associated with vital signs assessments in nontransports and management of hypoglycemia. The SDI was not associated with management of hypoglycemia. We found differences in alignment with respect to some study outcomes between the ADI and the SDI and SVI, including differences in the provision of bronchodilators for asthma and assessment of vital signs for nontransports. For provision of bronchodilators for asthma, the SDI and SVI were significantly associated with the study outcome, whereas the ADI was not. For assessment of vital signs for nontransports, the ADI suggested a nonsignificant positive association with this outcome, whereas the SDI and SVI were negatively associated with this outcome, although not significantly with the SVI. Findings were similar in a sensitivity analysis restricted to encounters occurring at home (eFigure 2 in Supplemental Material).

Table 4.

Performance on selected NEMSQA quality measures among adult EMS scene responses in the United States, 2023 ^a

Case scenario	Measure description	No. of eligible encounters	No. (%) of eligible encounters meeting performance criteria
Asthma-01	Percentage of EMS responses originating from a 911 request for patients with a diagnosis of asthma who had an aerosolized beta agonist administered	26 504	20 232 (76.3)
Trauma-01	Percentage of EMS responses originating from a 911 request for patients with injury who were assessed for pain	1 109 714	896 616 (80.8)
Trauma-03	Percentage of EMS transports originating from a 911 request for patients whose pain score was lowered during the EMS encounter	762 042	136 546 (17.9)
TTR-01	Percentage of EMS responses originating from a 911 request for patients not transported by EMS during which a basic set of vital signs is documented	40 516	25 327 (62.5)
Seizure-02	Percentage of EMS responses originating from a 911 request for patients with status epilepticus who received benzodiazepine during the EMS response	23 806	8974 (37.7)
Hypoglycemia-01	Percentage of EMS responses originating from a 911 request for patients with symptomatic hypoglycemia who received treatment to correct their hypoglycemia	39 923	33 309 (83.4)
Stroke-01	Percentage of EMS responses originating from a 911 request for patients who had a suspected stroke who had a stroke assessment performed during the EMS response	108 623	91 268 (84.0)

Abbreviations: EMS, emergency medical services; NEMSQA, National EMS Quality Alliance; TTR, treated, not transported record.

^aMeasures span a range of clinical scenarios and are derived from the ESO Data Collaborative. ³⁵ Performance was calculated as the number and percentage of encounters meeting each measure’s defined criteria out of all eligible cases.

Figure.

Splined plots demonstrating the association of each continuously measured community advantage index with each NEMSQA outcome measure for patients with out-of-hospital emergencies, United States, 2023. For each plot, the x-axis represents the scaled index (with higher numbers indicating greater disadvantage), and the y-axis represents the univariable probability of meeting the measure criteria. Shared areas represent 95% CIs. Note that y-axes differ between images to allow for visualization of differences. Some lines are not visible because of overlapping predictions. Expressed values are odds ratios (95% CI) comparing the third quartile with the first quartile of the stated predictor with the outcome. Labels with “a” indicate models with the best fit, defined by having the lowest Akaike information criterion. Abbreviations: ADI, Area Deprivation Index; NEMSQA, National EMS Quality Alliance; SDI, Social Deprivation Index; SVI, Social Vulnerability Index.

Comparison of 2023 Indices With Prior Year Distributions of SVI, SDI, and ADI

The distributions of the 2023 SVI and SDI, visualized with density polots, were similar to those of the 2018 SVI and SDI. The distribution of the 2023 ADI was similar to the distribution of the 2022 ADI (eFigure 3 in Supplemental Material).

Falsification Test

In the evaluated falsification tests, we found weak associations between the documentation of patient assessments and each of the studied indices (eFigure 4 in Supplemental Material).

Discussion

We used a large prehospital database to evaluate the agreement among 3 commonly used indices of neighborhood disparity. We found moderate agreement among the 3 indices; agreement was higher between the SVI and SDI than between the other pairings. When we evaluated the association of these indices with NEMSQA quality measures, the associations identified by using the ADI sometimes differed from associations identified by using the SVI and SDI.

We found that the studied measures were correlated. The SVI and SDI had greater agreement with each other than with the ADI. Notably, the ADI is calculated at the census-block level, which is a smaller geographic area than used by the SVI and SDI (which use census tract). These differences may be due to the methodological construction of the indices, even though they all incorporate economic, education, housing and environment, and transportation-related variables. A lower correlation between the ADI and SVI was noted in another study that evaluated pediatric prehospital encounters, relative to comparisons made using the Child Opportunity Index. ²⁴ One study that compared individual census tracts with each other also noted a lower correlation between the ADI and SVI (ICC = 0.51), again suggesting these measures are not interchangeable measures of socioeconomic deprivation. ²² Another study that compared the SDI, SVI, ADI, and Neighborhood Deprivation Index similarly identified that the SDI and SVI had the highest pairwise correlation (0.88) and the ADI and SDI had the lowest pairwise correlation (0.51) among patients listed for transplants. ²³ Importantly, the SDI, SVI, and ADI were designed with distinct purposes, even though many of their features overlap. As such, our findings highlight the importance of selecting an appropriate index for analysis, including consideration of the potential causal pathways for study outcomes.

Many of the findings demonstrated associations with social health disparities, although the strength of association we observed varied based on the studied metric. Associations were stronger for some conditions, such as asthma, reduction of pain in trauma, and assessment in nontransports and stroke, compared with those pertaining to seizure management, pain assessment in trauma, and hypoglycemia. While evidence has suggested that many of these measures may be related to social determinants of health, the extent to which they are based on prior literature varies. Studies have used various metrics to assess this association, including patient-level factors such as race and ethnicity, employment, and income, as well as composite measures such as those used in this study. In addition, research has examined various patient populations, such as prehospital or emergency department patients,^2,7,28-32 which can limit direct comparisons. In addition, when comparing the indices for NEMSQA measures, we found all 3 indices were aligned for some outcome measures but not for others. Notably, in areas where the ADI differed in directionality (such as in Asthma-01, Trauma-01, and TTR-01), these differences in prediction predominantly occurred in the low extremes. These areas may be more influenced by local factors that are better described by the ADI than by the SVI and SDI, which tend to capture broader, regional-level socioeconomic and demographic factors. For example, the use of EMS for asthma may correlate with localized environmental hazards or housing conditions, leading to greater calls for EMS for patients who have these conditions but who may not have sufficient respiratory distress to require use of a bronchodilator. Trauma-related outcomes might be influenced by neighborhood-specific safety concerns or crime rates that vary within a census tract. Nontransport decisions may be driven by ease of access to alternative forms of transportation, which may be best described at the census-block level.

Examining overlapping indices of disadvantage offers insights for designing strategies and policies to improve emergency services, reducing inequities in health outcomes. Although these indices differ in purpose and composition, their application revealed similar patterns in relation to study outcomes. These results underscore the importance of further exploration into their connections with variables such as response times, transport durations, and encounter types (eg, trauma incidents, chronic condition flare-ups). For example, community EMS systems play a critical role in addressing these disparities by leveraging local resources to deliver tailored interventions and improve accessibility for medically vulnerable populations. A community paramedicine program by the University of Maryland Medical Center and Baltimore City Fire Department found that social determinants of health needs, particularly in health care coordination, durable medical equipment, utilities, and social services, were linked to increased predicted and actual 30-day hospital readmissions. ³⁷ Addressing these social vulnerabilities could drive initiatives such as base station placement and increased training and funding for EMS in medically underserved areas. Research focused on causal mechanisms underlying these associations will be important, as will evaluating how policies may be aimed at reducing differences in prehospital emergency care delivery.

Limitations

Our findings were subject to several limitations. First, this is a retrospective analysis of data gathered through routine prehospital care, which may have errors in medical records or data abstraction. Second, not all encounters included all 3 community advantage indices, representing a potential source of bias. Third, we used the NEMSQA measures, which reflect broad consensus but may be influenced by variations in EMS agency protocols, potentially affecting the interventions provided and assessments performed. Fourth, we were unable to evaluate the necessity or appropriateness of interventions or determine whether some were omitted because of protocol differences across agencies. Fifth, we used density plots to compare the distribution of the disadvantage indices between the 2018 and 2023 years; however, while the overall distributions may appear similar, this approach may mask changes at the local or area-specific level that are not captured in aggregate comparisons.

Conclusion

We evaluated agreement among 3 indices of neighborhood disadvantage and their associations with quality indicators and found moderate agreement between indices. The ADI, calculated at a more granular census-block level, showed some distinct associations with specific outcomes. Continued research is needed to refine the application of these indices and explore their utility in reducing health disparities across health care settings.