Abstract
Background
Area‐based deprivation indices have been linked to hypertension and diabetes outcomes, but limited information exists about their relative performance. We compared 2 area‐based deprivation measures (Area Deprivation Index and Yost Index) as risk factors for hypertension and diabetes control.
Methods
This cohort study identified adults with hypertension or diabetes in a 19‐county region of Southeast Minnesota covered by the Rochester Epidemiology Project. The deprivation indices were operationalized as quintiles for state and national ranks. We assessed agreement using the weighted kappa statistic. We used modified Poisson regression to evaluate associations of the deprivation indices with blood pressure or diabetes control in 2022. We compared relative model fit using the area under the receiver operating characteristic curve.
Results
We identified 67 386 individuals with hypertension and 30 435 with diabetes. There was moderate to substantial agreement between indices, but the agreement was higher for the state rank (weighted kappa 0.68 for both cohorts) than the national rank (0.54 for both). Although the indices were not associated with hypertension control, greater deprivation was associated with worse diabetes control for the Yost Index (relative risk [RR], 1.21 [95% CI, 1.07–1.38]) and Area Deprivation Index (RR, 1.39 [95% CI, 1.12–1.73]) national ranks. The full‐model area under the receiver operating characteristic curve was similar across indices but differed across cohorts with worse fit for hypertension (0.53 for both) than diabetes (0.63 for both).
Conclusions
Two common area‐based deprivation indices had similar strength of association with diabetes control, suggesting that either index would be appropriate for researching and developing interventions for diabetes control.
Keywords: diabetes, hypertension, social determinants of health
Subject Categories: Hypertension, Epidemiology, Health Services
Nonstandard Abbreviations and Acronyms
- ADI
Area Deprivation Index
- REP
Rochester Epidemiology Project
- SDoH
social determinants of health
Research Perspective.
What Is New?
In this analysis, we compared the Area Deprivation Index and Yost Index for a sample in Southeast Minnesota to understand whether the strength of the associations with diabetes and hypertension control outcomes varied between the 2 area‐based measures.
We found that the strength of the association was similar for both indices for diabetes control but that neither index had a meaningful association with hypertension control outcomes.
What Question Should Be Addressed Next?
Do alternative area‐based deprivation indices or other measures have better performance in terms of their association with hypertension control outcomes?
The health care system and policymakers have increasingly recognized the importance of social determinants of health (SDoH) as contributors to health inequities. SDoH encompass the social and economic conditions that affect health and include both individual‐level factors such as income and food insecurity and area‐based factors such as neighborhood deprivation and environmental exposures. 1 Although greater efforts have been made in recent years to collect individual‐level SDoH characteristics from patients, the collection remains inconsistent and may be more likely to be missing for communities who are underserved. 2 In contrast, area‐based measures based on publicly available data are available for most patients as long as address information is available and can be geocoded. Area‐based measures for SDoH have value as independent predictors for outcomes even after accounting for individual SDoH factors. 3 Typically, the component measures for area‐based deprivation are combined into an index that measures deprivation in a defined geographical area. Area‐based deprivation indices have been linked to worse health outcomes across a variety of conditions, including cardiovascular disease 4 and diabetes. 5 Further, these measures have been linked to cardiovascular health disparities among racial and ethnic groups. 6 Therefore, area‐based measures have value for studying and mitigating health inequities.
Although area‐based measures have shown promise for tracking health inequities, there are many alternative indices available for area‐based deprivation 7 with little evidence on the relative performance of these indices to each other for different health conditions and in different subpopulations. These indices differ in terms of how they are calculated and which indicators are included in the overall index. 7 There is also evidence that the indices may perform worse in particular geographic areas due to local characteristics. 8 Identification of the best performing index in a particular context would help guide local efforts to address health inequities. For example, patients living in areas with high deprivation and clinics serving these areas could be targeted for additional resources to help address health inequities.
In this analysis, we aimed to provide evidence on the relative performance of 2 frequently used area‐based indices in terms of the strength of their association with blood pressure control and diabetes management in Southeast Minnesota. We also evaluated whether the relative performance differed for racial and ethnic minority groups. The 2 area‐based measures for deprivation we compared were the Area Deprivation Index (ADI) 4 and the Yost Index. 9 We selected these indices because they are both commonly used indices that are publicly available for researchers. 7 The ADI is a composite measure of 17 census indicators capturing education, employment, income, poverty, and housing characteristics. 4 , 10 , 11 The Yost index is a composite measure of 7 census socioeconomic status indicators 9 that are a subset of the ones used to construct the ADI. 4 The Yost Index is considered by some to be less sensitive to outliers and more transparent in its construction than the ADI. 12 The ADI has also been criticized for overemphasizing housing value, especially in urban areas, 8 which does not appear to be a problem for the Yost Index. 12 The results of this analysis have implications for selecting area‐based measures for applications relating to these cohorts in future work.
METHODS
Study Design
The primary data source for the blood pressure and diabetes outcomes was medical records obtained through the REP (Rochester Epidemiology Project). The results reported in this article use REP data, which includes identifiable patient information (protected health information) derived from electronic health records. As such, the full data sets cannot be shared publicly. However, limited or deidentified data sets may be shared in accordance with Mayo Clinic, Olmsted Medical Center, and Health Insurance Portability and Accountability Act (HIPAA) policies that require institutional review board approval and enactment of data use agreements for defined studies. Please send queries regarding data sharing to the corresponding author of the article. The REP is a medical‐records linkage system that covers individuals living in a 27‐county area of Southeast Minnesota and west‐central Wisconsin. 13 , 14 , 15 For this analysis, we were focused on understanding health inequities in Southeast Minnesota and so we limited the geographic scope to the 19 counties covered by the REP in Southeast Minnesota. We linked the medical records data to 2 publicly available area‐based measures for deprivation: the ADI 4 and the Yost Index. 9 The ADI is available for the entire United States at the block‐group level at https://www.neighborhoodatlas.medicine.wisc.edu/. The Yost Index is available for the entire United States at both the census tract and block group level at https://tinyurl.com/yost8years. This study was approved by the Mayo Clinic and Olmsted Medical Center Institutional Review Boards and the requirement for informed consent was waived.
Participants
We identified 2 separate cohorts: a diabetes cohort and a hypertension cohort. Patients could be in both cohorts if they met the inclusion and exclusion criteria for both (Figure S1). For the diabetes cohort, patients had to have ≥2 International Classification of Diseases, Tenth Revision (ICD‐10) diagnosis codes (see Table 1) for diabetes at least 30 days apart between January 1, 2019 and June 30, 2022. We included diagnosis codes for both Type 1 and Type 2 diabetes. For the cohort with hypertension, patients had to have ≥2 ICD‐10 codes for hypertension in the same period. The diagnosis code sets come from quality measures that are used by the Centers for Medicare & Medicaid Services and private payers to promote better patient outcomes and measure the quality of care. 16 We selected this approach because the quality measures were designed to assess quality of care for diabetes and hypertension as measured by hemoglobin A1c and blood pressure control, which are the same outcomes used in our study. Although we are not aware of direct validations of these specific code sets, similar algorithms relying exclusively on ICD‐10 codes have found positive predictive value of 77% for diabetes 17 and 81% for hypertension. 18 In both cohorts, patients had to be aged ≥18 and be a resident of the 19‐county area of Minnesota covered by the REP as of July 1, 2022. Patients also had to have ≥1 diagnosis, laboratory, or procedure code during 2022 to ensure that they were still receiving care in the area. Lastly, patients had to have complete demographic information and an address that could be linked to the ADI and Yost Index data.
Table 1.
Patient Characteristics
| Cohort with hypertension | Cohort with diabetes | ||||
|---|---|---|---|---|---|
| Included (N=67 386) | Excluded (N=9770) | Included (N=30 435) | Excluded (N=4195) | ||
| Systolic BP | Mean±SD | 133.9±16.4 | |||
| Median (IQR) | 133 (123–143) | ||||
| Diastolic BP | Mean±SD | 77.8±10.2 | |||
| Median (IQR) | 78 (71.6–84) | ||||
| BP out of control | N (%) | 20 853 (32.8) | |||
| BP missing | N (%) | 3813 (5.7) | |||
| HbA1c | Mean±SD | 7.2±1.4 | |||
| Median (IQR) | 6.9 (6.3–7.8) | ||||
| HbA1c out of control | N (%) | 2566 (10.0) | |||
| HbA1c missing | N (%) | 4697 (15.4) | |||
| Age, y | Mean±SD | 65.7±14.1 | 66.3±14.0 | 64.4±14.7 | 65.1±14.5 |
| Median (IQR) | 67.0 (57.0, 76.0) | 67.0 (58.0, 76.0) | 66.0 (56.0, 75.0) | 67.0 (57.0, 75.0) | |
| Sex | Female | 34 395 (51.0%) | 4848 (49.6%) | 14 294 (47.0%) | 1874 (44.7%) |
| Male | 32 991 (49.0%) | 4922 (50.4%) | 16 141 (53.0%) | 2321 (55.3%) | |
| Race | Black | 1738 (2.6%) | 190 (1.9%) | 1225 (4.0%) | 134 (3.2%) |
| Asian | 1276 (1.9%) | 111 (1.1%) | 894 (2.9%) | 72 (1.7%) | |
| Hawaiian/Pacific Islander | 99 (0.1%) | 9 (0.1%) | 73 (0.2%) | 10 (0.2%) | |
| American Indian/Alaska Native | 211 (0.3%) | 32 (0.3%) | 167 (0.5%) | 27 (0.6%) | |
| Other/Mixed* | 1051 (1.6%) | 147 (1.5%) | 763 (2.5%) | 102 (2.4%) | |
| White | 62 896 (93.5%) | 9263 (95.0%) | 27 251 (89.7%) | 3832 (91.7%) | |
| Missing | 115 | 18 | 62 | 18 | |
| Hispanic ethnicity | 2648 (3.9%) | 413 (4.2%) | 1976 (6.5%) | 258 (6.2%) | |
| Education level | Eighth grade or less | 905 (1.6%) | 152 (1.9%) | 642 (2.5%) | 89 (2.5%) |
| Some high school | 1727 (3.0%) | 283 (3.4%) | 1061 (4.1%) | 160 (4.5%) | |
| High school/general educational development | 16 209 (28.2%) | 2568 (31.3%) | 7726 (30.1%) | 1174 (33.3%) | |
| Some college or 2‐y degree | 20 843 (36.2%) | 3045 (37.1%) | 9179 (35.8%) | 1278 (36.2%) | |
| 4‐y college degree | 8691 (15.1%) | 1080 (13.2%) | 3494 (13.6%) | 392 (11.1%) | |
| Postgraduate studies | 9154 (15.9%) | 1075 (13.1%) | 3569 (13.9%) | 433 (12.3%) | |
| Missing | 9857 | 1567 | 4764 | 669 | |
| Body mass index | Underweight | 342 (0.5%) | 45 (0.5%) | 77 (0.3%) | 7 (0.2%) |
| Healthy weight | 8649 (12.9%) | 1188 (12.3%) | 2563 (8.5%) | 299 (7.2%) | |
| Overweight | 19 348 (28.9%) | 2809 (29.0%) | 7103 (23.6%) | 954 (23.1%) | |
| Obese | 38 536 (57.6%) | 5650 (58.3%) | 20 357 (67.6%) | 2872 (69.5%) | |
| Missing | 511 | 78 | 335 | 63 | |
| Tobacco use | Never | 31 547 (46.9) | 4485 (46.0) | 13 850 (45.6) | 1806 (43.2) |
| Current | 8177 (12.1) | 1277 (13.1) | 3634 (12.0) | 528 (12.6) | |
| Former | 27 618 (41.0) | 3989 (40.9) | 12 899 (42.5) | 1843 (44.1) | |
| Missing | 44 | 19 | 52 | 18 | |
| Charlson Comorbidity Index | Mean±SD | 1.5±2.0 | 1.5±2.0 | 1.5±2.0 | 1.5±2.0 |
| Median (IQR) | 1.0 (0.0, 2.0) | 1.0 (0.0, 2.0) | 1.0 (0.0, 2.0) | 1.0 (0.0, 2.0) | |
| Rural–urban commuting area | Urban | 39 535 (58.7) | 17 550 (57.7) | ||
| Large rural | 19 032 (28.2) | 8640 (28.4) | |||
| Small rural | 4787 (7.1) | 2420 (8.0) | |||
| Isolated | 4032 (6.0) | 1825 (6.0) | |||
| ADI national rank | Q1 (least deprived) | 3177 (4.7%) | 1045 (3.4%) | ||
| Q2 | 16 862 (25.0%) | 6886 (22.6%) | |||
| Q3 | 24 920 (37.0%) | 11 107 (36.5%) | |||
| Q4 | 16 480 (24.5%) | 8172 (26.9%) | |||
| Q5 (most deprived) | 5947 (8.8%) | 3225 (10.6%) | |||
| Yost national rank | Q1 (least deprived) | 12 291 (18.2%) | 4842 (15.9%) | ||
| Q2 | 21 731 (32.3%) | 9106 (29.9%) | |||
| Q3 | 18 129 (26.9%) | 8420 (27.7%) | |||
| Q4 | 11 255 (16.7%) | 5751 (18.9%) | |||
| Q5 (most deprived) | 3980 (6.9%) | 2316 (7.6%) | |||
| ADI state rank | Q1 (least deprived) | 6779 (10.1%) | 2399 (7.9%) | ||
| Q2 | 13 260 (19.7%) | 5532 (18.2%) | |||
| Q3 | 13 553 (20.1%) | 6072 (20.0%) | |||
| Q4 | 17 083 (25.4%) | 7819 (25.7%) | |||
| Q5 (most deprived) | 16 711 (24.8%) | 8613 (28.3%) | |||
| Yost state rank | Q1 (least deprived) | 10 801 (16.0%) | 4214 (13.9%) | ||
| Q2 | 13 303 (19.7%) | 5491 (18.0%) | |||
| Q3 | 15 325 (22.7%) | 6755 (22.2%) | |||
| Q4 | 14 881 (22.1%) | 6910 (22.7%) | |||
| Q5 (most deprived) | 13 076 (19.4%) | 7065 (23.2%) | |||
| ADI national rank (continuous) | Mean±SD | 52.3±19.2 | 54.4±19.2 | ||
| Median (IQR) | 52 (37–66) | 54 (40–67) | |||
| Yost national rank (continuous) | Mean±SD | 43.0±22.6 | 45.6±23.0 | ||
| Median (IQR) | 40 (25–59) | 45 (27–62) | |||
ADI indicates Area Deprivation Index; BP, blood pressure; and IQR, interquartile range. *Other/Mixed is a separate option that patients can choose if they feel that the other categories do not apply.
Outcomes
The outcomes were diabetes control for the cohort with diabetes and hypertension control for the cohort with hypertension. Diabetes was considered “in control” if the most recent hemoglobin A1c measure was <9% and hypertension was considered “in control” if the most recent blood pressure measure was <140 mm Hg systolic and < 90 mm Hg diastolic. The outcome thresholds come from the quality measures mentioned previously. 16 We considered a single observation for each individual based on the most recent hemoglobin A1c or blood pressure value between January 1, 2022 and December 31, 2022. For the main analysis, individuals without a hemoglobin A1c or blood pressure measurement were considered to be “not in control.”
Key Exposures
The 2 key exposures are the ADI and Yost Index, which are both measures of neighborhood socioeconomic status. Both the ADI and Yost Index are constructed using American Community Survey 5‐year data. We used the 2020 v4 version of the ADI (2016–2020) and the 2018 version of the Yost Index (2014–2018). We operationalized the ADI and Yost Index as quintiles with respect to both the state distributions and the national distributions, with higher quintiles indicating greater neighborhood deprivation. We chose quintiles to facilitate comparisons across the indices and because quintiles are the most common grouping reported in the literature. 19 Both indices were linked to health outcome data at the census‐block group level using geocoded patient addresses. We used the patient address on July 1, 2022 or the address that was active closest to July 1, 2022.
Other Variables
We identified other factors that may act as confounders between the area‐based deprivation measures and diabetes or hypertension control (see Figure S2 for directed acyclic graph). We controlled for demographics including age, sex, race and ethnicity, and education. Lastly, we controlled for rurality using Rural–Urban Commuting Area (RUCA) codes. We operationalized rurality using the secondary Rural–Urban Commuting Area codes as being urban, large rural, small rural, and isolated rural using the Rural Health Research Center definitions. 20 We identified other factors that may affect health status including body mass index, overall comorbidity burden measured by the Charlson comorbidity index, and tobacco use as potential mediators. We did not control for these variables because they are potential mediators, but we did include them in Table 1 to provide further characteristics of the sample. Demographics and risk factor data were obtained from the REP, and Rural–Urban Commuting Area codes are publicly available.
Statistical Analysis
We report the descriptive statistics using frequencies and percentages for categorical variables and means±SD for continuous variables. We conducted modified Poisson regression 21 with cluster‐robust error variances with clustering at the census block‐group level for both outcomes. For each cohort, we completed 4 regressions: (1) ADI using state rank+other variables; (2) Yost Index using state rank+other variables; (3) ADI using national rank+other variables; and (4) Yost Index using national rank+other variables. The regression results are reported as relative risk (RR) of the highest deprivation quintile compared with the lowest quintile, with 95% CI. We conducted a test for trend from lower to higher quintiles of the distribution using an auxiliary regression in which we respecified the quintiles as a continuous variable ranging from 1 to 5. We compared the relative performance of the indices using the Vuong closeness test, 22 which is a likelihood ratio test for nonnested models. We also conducted the same comparisons in terms of area under the receiver operating characteristic curve) for predicting the outcomes for both the univariate and full models. Finally, we assessed whether including interaction terms between the indices and race and ethnicity led to statistically significant improvements in model fit. We excluded observations with missing data for the main analysis.
Sensitivity Analyses
To assess the robustness of our findings, we conducted 5 sensitivity analyses. First, we studied more stringent thresholds for hypertension control (<130 mm Hg systolic and <80 mm Hg diastolic) and diabetes control (hemoglobin A1c <8%). Second, we classified the outcome for individuals missing hemoglobin A1c or blood pressure measurements as “in control” rather than “not in control” as used in the primary analysis. Third, we excluded individuals who did not have a hemoglobin A1c or blood pressure measurement in the period. Fourth, we included observations with missing data for categorical variables by creating a “missing” category. Fifth, we specified the index as a continuous variable from 1 to 100 based on the percentiles. We limited this analysis to the national rankings because percentiles are not available for the state rankings.
RESULTS
Sample Characteristics
Table 1 shows the sample characteristics for the cohorts with hypertension and diabetes as well as the characteristics for patients who were excluded due to not having address information that could be linked to the ADI or Yost Index. Overall, 13% of the patients with hypertension and 12% of the patients with diabetes were excluded due to missing address data or address data that could not be geocoded (eg, post office boxes). Characteristics of excluded patients were similar to those included in the analyses (Table 1).
The cohorts with hypertension and diabetes were similar in terms of mean age (65.7 versus 64.4 years), current tobacco use (12.1% versus 12.0%), Charlson comorbidity index mean (1.5 for both), and residing in an urban area (58.7% versus 57.7%). In contrast, patients in the cohort with hypertension were less likely to have obesity (57.6% versus 67.6%) and more likely to live in the least deprived quintile (Q1) for the Yost Index national rank (18.2% versus 15.9%). There was also a higher proportion of patients “out of control” in the cohort with hypertension (32.8%) than the cohort with diabetes (10.0%).
Cross‐Comparison of the ADI and Yost Index
The concordance between the ADI and Yost Index was moderate for both the cohorts with hypertension and diabetes. For the cohort with hypertension, the weighted kappa was lower for the national rank (0.538 [95% CI, 0.535–0.542]) than for the state rank (0.677 [95% CI, 0.674–0.680]) (Table 2). For the cohort with diabetes, the weighted kappa was also lower for the national rank (0.541 [95% CI, 0.536–0.546]) than the state rank (0.678 [95% CI, 0.673–0.683]; Table 3). When there was disagreement for the quintiles in the cohort with hypertension or diabetes, the ADI tended to rank individuals in a more deprived quintile than the Yost Index (Figure; Figure S2). This pattern of the ADI ranking individuals as having higher area‐level deprivation was consistent across the study region for the national rank. However, for the state rank some of the areas in the northern part of the study region were rated as having higher deprivation for the Yost Index compared with the ADI.
Table 2.
ADI and Yost Index Cross‐Comparison for Cohort With Hypertension
| Yost national rank | ADI national rank | |||||
|---|---|---|---|---|---|---|
| Q1 | Q2 | Q3 | Q4 | Q5 | Total | |
| Q1 (least deprived) | 3100 | 7496 | 1679 | 16 | 0 | 12 291 |
| Q2 | 66 | 8270 | 12 757 | 623 | 15 | 21 731 |
| Q3 | 11 | 1071 | 9543 | 6976 | 528 | 18 129 |
| Q4 | 0 | 24 | 925 | 7035 | 3271 | 11 255 |
| Q5 (most deprived) | 0 | 1 | 16 | 1830 | 2133 | 3980 |
| Total | 3177 | 16 862 | 24 920 | 16 480 | 5947 | 67 386 |
| Weighted kappa | Estimate | Std. Error | 95% CI | |||
| 0.5384 | 0.0019 | 0.5347 | 0.5420 | |||
| Yost state rank | ADI state rank | |||||
|---|---|---|---|---|---|---|
| Q1 | Q2 | Q3 | Q4 | Q5 | Total | |
| Q1 (least deprived) | 5182 | 4249 | 1185 | 177 | 8 | 10 801 |
| Q2 | 1452 | 6825 | 4209 | 764 | 53 | 13 303 |
| Q3 | 128 | 2072 | 6911 | 5826 | 388 | 15 325 |
| Q4 | 15 | 94 | 944 | 8214 | 5614 | 14 881 |
| Q5 (most deprived) | 2 | 20 | 304 | 2102 | 10 648 | 13 076 |
| Total | 6779 | 13 260 | 13 553 | 17 083 | 16 711 | 67 386 |
| Weighted kappa | Estimate | Std. error | 95% CI | |||
| 0.6769 | 0.0016 | 0.6737 | 0.6801 | |||
ADI indicates Area Deprivation Index.
Table 3.
ADI and Yost Index Cross‐Comparison for Cohort With Diabetes
| Yost national rank | ADI national rank | |||||
|---|---|---|---|---|---|---|
| Q1 | Q2 | Q3 | Q4 | Q5 | Total | |
| Q1 (Least deprived) | 1019 | 3086 | 732 | 5 | 0 | 4842 |
| Q2 | 22 | 3269 | 5521 | 287 | 7 | 3009 |
| Q3 | 4 | 522 | 4364 | 3293 | 237 | 2501 |
| Q4 | 0 | 8 | 480 | 3558 | 1705 | 2990 |
| Q5 (Most deprived) | 0 | 1 | 10 | 1029 | 1276 | 2935 |
| Total | 1045 | 6886 | 11 107 | 8172 | 3225 | 30 435 |
| Weighted kappa | Estimate | SE | 95% CI | |||
| 0.5410 | 0.0027 | 0.5357 | 0.5464 | |||
| Yost state rank | ADI state rank | |||||
|---|---|---|---|---|---|---|
| Q1 | Q2 | Q3 | Q4 | Q5 | Total | |
| Q1 (Least deprived) | 1804 | 1798 | 532 | 77 | 3 | 4214 |
| Q2 | 534 | 2774 | 1818 | 347 | 18 | 5491 |
| Q3 | 57 | 902 | 3098 | 2534 | 164 | 6755 |
| Q4 | 3 | 52 | 453 | 3804 | 2598 | 6910 |
| Q5 (Most deprived) | 1 | 6 | 171 | 1057 | 5830 | 7065 |
| Total | 2399 | 5532 | 6072 | 7819 | 8613 | 30 435 |
| Weighted kappa | Estimate | SE | 95% CI | |||
| 0.6778 | 0.0024 | 0.6730 | 0.6826 | |||
ADI indicates Area Deprivation Index.
Figure 1. Heat map of ADI and Yost Index using the national level distribution.
ADI indicates Area Deprivation Index.
Model Results
In the cohort with hypertension, the ADI and Yost Index were not significantly associated with blood pressure control (Table 4). Further, the performance did not vary for the ADI or the Yost Index when using the national or state ranks and the test for trend was not statistically significant. The Vuong closeness test was not significant for either the national or state rank, which suggests that there was not better performance in the models using the ADI or Yost Index. Additionally, the area under the receiver operating characteristic curve was low for all models (0.54 for full models and 0.51–0.52 for the univariate models [Table S6]). The interaction terms of the indices with race and ethnicity were not statistically significant.
Table 4.
Regression Results
| National ank | State rank | |||
|---|---|---|---|---|
| Yost Index | ADI | Yost Index | ADI | |
| Cohort with hypertension | ||||
| RR | RR | RR | RR | |
| (95% CI) | (95% CI) | (95% CI) | (95% CI) | |
| Q1 | Reference | |||
| Q2 | 1.00 | 1.01 | 1.00 | 1.04 |
| (0.96–1.04) | (0.95–1.07) | (0.96–1.05) | (0.99–1.09) | |
| Q3 | 1.02 | 1.02 | 1.03 | 1.03 |
| (0.99–1.06) | (0.96–1.08) | (0.99–1.07) | (0.97–1.08) | |
| Q4 | 1.04* | 1.03 | 1.04 | 1.06* |
| (1.00–1.08) | (0.97–1.09) | (0.99–1.08) | (1.00–1.11) | |
| Q5 | 0.97 | 1.01 | 1.02 | 1.04 |
| (0.91–1.03) | (0.95–1.09) | (0.98–1.07) | (0.99–1.09) | |
| Test for trend | P=0.25 | P=0.40 | ||
| Vuong closeness test | P=0.11 | P=0.75 | ||
| Area under the curve | 0.531 | 0.530 | 0.530 | 0.530 |
| LR test for model including interactions with race and ethnicity | P>0.99 | N/C | P=0.91 | P=0.73 |
| Cohort with diabetes | ||||
|---|---|---|---|---|
| RR | RR | RR | RR | |
| (95% CI) | (95% CI) | (95% CI) | (95% CI) | |
| Q1 | Reference | |||
| Q2 | 1.04 | 1.16 | 1.02 | 1.11 |
| (0.94–1.15) | (0.95–1.42) | (0.91–1.14) | (0.97–1.27) | |
| Q3 | 1.06 | 1.17 | 1.07 | 1.09 |
| (0.96–1.17) | (0.95–1.43) | (0.95–1.20) | (0.95–1.25) | |
| Q4 | 1.24‡ | 1.29* | 1.07 | 1.16* |
| (1.11–1.37) | (1.05–1.59) | (0.96–1.19) | (1.01–1.32) | |
| Q5 | 1.21† | 1.39† | 1.23‡ | 1.25‡ |
| (1.07–1.38) | (1.12–1.73) | (1.10–1.37) | (1.09–1.42) | |
| Test for trend | P<0.001 | P<0.001 | ||
| Vuong closeness test | P=0.39 | P=0.25 | ||
| Area under the curve | 0.630 | 0.630 | 0.630 | 0.629 |
| LR test for model including interactions with race and ethnicity | P=0.96 | N/C | P=0.81 | P=0.91 |
Models control for race and ethnicity, age, sex, education, and rural–urban commuting area. ADI indicates Area Deprivation Index; LR, likelihood ratio; N/C, model did not converge; and RR, relative risk.
P<0.05.
P<0.01.
P<0.001.
In the cohort with diabetes, both measures of area deprivation were associated with diabetes control (Table 4). For the national rank, individuals in the most deprived quintile (Q5) were more likely to be out of control compared with those in the least deprived quintile (Q1) for both the Yost Index (RR, 1.21 [95% CI, 1.07–1.38]) and ADI (RR, 1.39 [95% CI, 1.12–1.73]). For the state rank, the RRs were also statistically significant for Q5 compared with Q1 for both the Yost Index (RR, 1.23 [95% CI, 1.10–1.37]) and ADI (RR, 1.25 [95% CI, 1.09–1.42 ]). The test for trend was also statistically significant in both models (P<0.001). The area under the receiver operating characteristic curvewas higher for these models (0.63 full models and 0.54–0.55 for the univariate models [Table S6]) than for the hypertension cohort models. However, similar to the cohort with hypertension models, the Vuong closeness test was not statistically significant and the interactions of the indices with race and ethnicity were not statistically significant for any comparison.
Sensitivity Analyses
The sensitivity analyses were consistent with the main findings (Tables S2 through S5). The RRs for Q5 versus Q1 for both the ADI and Yost Index were >1 and statistically significant for most comparisons in the cohort with diabetes. However, the RRs were generally nonsignificant for the cohort with hypertension. The Vuong closeness test was not statistically significant for any comparison and the interaction terms for race and ethnicity were not significant for any model. For the continuous specification, the RRs were significant only for the cohort with diabetes (RR, 1.03 [95% CI, 1.02–1.05] Yost; RR, 1.04 [95% CI, 1.02–1.05] ADI) and were consistent in direction with the quintile specification.
DISCUSSION
In this analysis, we found that the ADI and Yost Index, 2 commonly used area‐based deprivation indices, had similar performance in terms of their association with blood pressure and diabetes control outcomes in Southeast Minnesota. For both indices, higher deprivation was associated with worse diabetes control outcomes and the performance was similar for the national and state ranks. However, higher deprivation was not associated with blood pressure control outcomes for either index, and the predictive performance was poor with respect to the area under the receiver operating characteristic curve for both indices with the national and state ranks. Additionally, the interaction terms between the indices and race or ethnicity were not statistically significant for any comparison. Overall, these findings suggest that the 2 deprivation indices perform similarly in analyses of blood pressure and diabetes control.
Our results differ from prior research that has found conflicting results when comparing alternative area‐based deprivation measures in different contexts. A prior study comparing the ADI and the Yost Index found substantial differences in their construction and that the ADI may be sensitive to outliers and areas with missing census data. 12 It is possible that these differences are less applicable to the region of Southeast Minnesota we focused on for this analysis. In a study among pediatric surgical patients, the concordance between the Social Vulnerability Index, 23 ADI, and Child Opportunity Index 24 was poor to fair (kappa statistic ranged from 0.15 to 0.30) within different hospital sites. 25 Another analysis comparing the ADI and SVI across all census tracks found modest agreement (44.1% were within 1 decile in ranking) and that disagreement was greater for urban areas with characteristics like higher median monthly mortgages. 26 The greater dissimilarity between indices in these studies may have been driven by the indices being compared having fewer measures in common than the ADI and Yost Index. 4 , 7 , 9 We note that both indices use measures from the American Community Survey. As such, our findings may be driven by the similarities in the measures used to construct the ADI and Yost Index. Additionally, a key limitation of the ADI is that it appears to use unstandardized housing variables, which can lead to high rates of misclassification in urban areas. 8 , 27 Given our focus on a largely rural area of Southeast Minnesota, this limitation may not have affected the results as much in our analysis since there is less variation in housing costs. Another analysis compared the Social Vulnerability Index, the Graham Social Deprivation Index (SDI), 28 the unstandardized version of the ADI that is commonly used, and a standardized version of the ADI in terms of predictive performance for mortality among a Medicare fee‐for‐service population. The results suggested similar performance for the Social Vulnerability Index, Social Deprivation Index, and the standardized version of the ADI but worse performance for the unstandardized version of the ADI. 29 Taken together, these findings suggest that even though our analysis did not find meaningful differences between the ADI and Yost Index in terms of performance, it is important to consider local context and the construction of the indices when choosing a particular index for a research application.
The poor performance of the indices in predicting blood pressure control suggests alternative measures are needed to better understand and improve blood pressure control outcomes. In contrast, prior research has found higher prevalence of hypertension in areas with greater area‐level deprivation. 30 , 31 , 32 , 33 Although one such study focusing on adults in Cuyahoga County, Ohio, found a much higher prevalence comparing ADI quintile 5 (50.7%) to ADI quintile 1 (25.5%), being prescribed an antihypertensive medication was only slightly lower in ADI quintile 5 (61.3%) compared with quintile 1 (64.5%). 31 In another analysis that included many of the same areas of Southeast Minnesota that were the focus of our study, blood pressure control was not significantly different between the highest and lowest quintiles of the ADI. 34 Similarly, another article found comparable hypertension control rates for patients in high and low deprivation areas in states that expanded Medicaid. 35 As such, it is possible that area‐level deprivation may matter more for incidence of hypertension than treatment and control of prevalent hypertension. Additionally, the findings may be due in part to the local context in Minnesota, which has expanded Medicaid, has generous Medicaid eligibility criteria, and has a 5% uninsurance rate (whereas the rate is 10% nationally). 36 Following Medicaid expansion, the largest gains in hypertension control were found to be in areas with the highest area‐based deprivation. 35 Thus, the findings of minimal differences in blood pressure control may reflect the success of policies put in place to address health inequities. 34
We did find a consistent association between both the ADI and Yost Index and diabetes control. Associations between area deprivation and diabetes control have been reported elsewhere. In the same study that used a similar geographic scope and found no association for blood pressure control, there was a significant association for ADI Q5 (most deprived) compared with Q1 (least deprived) for glycemic control. 34 Similarly, a national study found that counties with higher deprivation for the ADI have more health care use related to uncontrolled diabetes 5 and another national study found a higher prevalence of diabetes among non‐Hispanic White patients living in higher deprivation areas. 33 Additionally, a study in Maryland found that higher levels of the ADI were associated with a greater number of diabetic ketoacidosis readmissions, especially among children. 37 In contrast, a study focused on adults in Pennsylvania found that ADI was not associated with glycemic control. However, high ADI was associated with lower adherence to diabetes medications, being less likely to have a hemoglobin A1c test in the year, and higher health care use. 38 A study focusing on a population in Northern California found that the highest quartile of a neighborhood deprivation index was associated with higher hemoglobin A1c levels but not blood pressure control levels. 39 Altogether, these findings suggest that unlike blood pressure control, there is a mostly consistent relationship between higher levels of area‐level deprivation and worse diabetes control.
We did not observe statistically significant interactions between the indices and race and ethnicity. One potential explanation is that the interaction terms were underpowered despite the large sample size. The power to detect interaction effects is much smaller than main effects. 40 In addition, the percentages of non‐Hispanic White patients were 93.5% in the cohort with hypertension and 89.7% in the cohort with diabetes. With such small percentages of other races and ethnicities, this would further compound the decreased power. Another explanation is that residential segregation is one of the main mechanisms through which race and ethnicity can affect health outcomes. 41 As such, race and ethnicity may not have much explanatory power beyond what is accounted for by the neighborhood deprivation indices. Lastly, it may be possible that the results are specific to the geographic area covered in this study, which focused on a region of Southeast Minnesota with a high proportion of rural residents, no large cities, and a relatively high proportion of non‐Hispanic White residents. There may be stronger interaction effects for other geographic areas such as those including residents in large cities. 42 , 43
Our study findings have implications for how these measures can be used to identify and address inequities. First, the similar performance of the 2 indices and the national versus state ranks suggests that the choice between the ADI or Yost Index and the choice between the national and state ranks do not matter much in the context of Southeast Minnesota. Second, the indices appear more useful for identifying inequities in diabetes control outcomes than blood pressure control outcomes. Third, in addition to the potential for identifying health inequities, either index could be used as part of efforts to identify and intervene in patients at high risk for poor diabetes control. Many health systems have interventions to refer patients with health‐related social needs to community resources. One such example are community health workers who are individuals in the community that assist patients with navigating community resources to address health‐related social needs. 44 One of the health systems that serves many of the patients in this region of Southeast Minnesota is the Mayo Clinic and it has had a community health worker program in place since 2013. 45 The program relies heavily on health‐related social needs identified in SDoH questionnaires. However, these questionnaires have a low completion rate, especially among groups that may be more likely to have health‐related social needs. 2 As such, area‐based measures can be used alongside screening questions to help identify patients who may be at higher risk for poor health outcomes and may benefit from referrals to interventions like the community health workers program.
The strengths of this analysis include the large sample size and the comprehensive medical record information available through the REP. There are also important limitations. First, the study focused on blood pressure and diabetes control outcomes in Southeast Minnesota, and results may not generalize to other geographic locations in which neighborhood characteristics differ dramatically. For example, Southeast Minnesota has a high rural population and does not include a large urban city. As such, the findings may not generalize to areas that are more urban. The results may also not generalize to other health conditions we did not consider that may be differentially affected by neighborhood characteristics. Second, the findings may not generalize to other area‐based deprivations measures that use different methodologies than the ADI or Yost Index. For example, the RTI Local Social Inequity score is trained on a much larger number of variables and is developed using machine learning models trained for specific outcomes; the Local Social Inequity appears to outperform other commonly used indices such as the ADI. 46 , 47 Although the Local Social Inequity may have better performance than other area‐based deprivation indices, it is not freely available, which may be a constraint for researchers with limited budgets. Third, we had limited information on individual's SDoH other than education; thus, it is unclear how much the area‐based deprivation indices add in addition to individual‐level information in this context. Prior work has found that the area‐level and individual‐level measures capture distinct information. 48 Additionally, the use of neighborhood‐level measures rather than individual‐level measures can lead to Berkson error in which coefficients are not biased but the estimates are less precise. 49 However, as many large database analyses lack information on individual‐level SDoH such as housing stability and food insecurity, it is important to understand the performance of these measures in absence of individual‐level measures.
CONCLUSIONS
This analysis found that 2 commonly used area‐based measures of deprivation had similar performance for diabetes control but were not predictors of blood pressure control. These findings suggest that either index would be equally appropriate for researching and developing interventions for diabetes control. Additional research is needed to determine if an alternative area‐based deprivation index has better performance for predicting hypertension control.
Sources of Funding
Research reported in this publication was supported by the National Institute on Minority Health And Health Disparities of the National Institutes of Health under Award Number P50MD017342. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Disclosures
Dr St. Sauver reports grant funding from Moderna for studies of cytomegalovirus and infectious mononucleosis that is unrelated to this work. The remaining authors have no disclosures to report.
Supporting information
Tables S1–S6
Figures S1–S2
Acknowledgments
This study used the resources of the Rochester Epidemiology Project medical records‐linkage system, which is supported by the National Institute on Aging (AG 058738), by the Mayo Clinic Research Committee, and by fees paid annually by REP users. The content of this article is solely the responsibility of the authors and does not represent the official views of the National Institutes of Health or the Mayo Clinic.
Part of this work was presented at the C2DREAM Conference, June 3, 2025, in St. Paul, MN.
This article was sent to Mahasin S. Mujahid, PhD, MS, FAHA, Associate Editor, for review by expert referees, editorial decision, and final disposition.
Supplemental Material is available at https://www.ahajournals.org/doi/suppl/10.1161/JAHA.125.045324
For Sources of Funding and Disclosures, see page 11.
References
- 1. World Health Organization . Social determinants of health. 2008.
- 2. Savitz ST, Nyman MA, Kaduk A, Loftus C, Phelan S, Barry BA. Association of patient and system‐level factors with social determinants of health screening. Med Care. 2022;60:700–708. doi: 10.1097/MLR.0000000000001754 [DOI] [PubMed] [Google Scholar]
- 3. Shohaimi S, Luben R, Wareham N, Day N, Bingham S, Welch A, Oakes S, Khaw K‐T. Residential area deprivation predicts smoking habit independently of individual educational level and occupational social class. A cross sectional study in the Norfolk cohort of the European Investigation into Cancer (EPIC‐Norfolk). J Epidemiol Community Health. 2003;57:270–276. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Kind AJ, Jencks S, Brock J, Yu M, Bartels C, Ehlenbach W, Greenberg C, Smith M. Neighborhood socioeconomic disadvantage and 30‐day rehospitalization: a retrospective cohort study. Ann Intern Med. 2014;161:765–774. doi: 10.7326/m13-2946 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Kurani SS, Heien HC, Sangaralingham LR, Inselman JW, Shah ND, Golden SH, McCoy RG. Association of area‐level socioeconomic deprivation with hypoglycemic and hyperglycemic crises in US adults with diabetes. JAMA Netw Open. 2022;5:e2143597. doi: 10.1001/jamanetworkopen.2021.43597 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Akinyelure OP, Jaeger BC, Oparil S, Carson AP, Safford MM, Howard G, Muntner P, Hardy ST. Social determinants of health and uncontrolled blood pressure in a National Cohort of Black and white US adults: the REGARDS study. Hypertension. 2023;80:1403–1413. doi: 10.1161/HYPERTENSIONAHA.122.20219 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Trinidad S, Brokamp C, Mor Huertas A, Beck AF, Riley CL, Rasnick E, Falcone R, Kotagal M. Use of area‐based socioeconomic deprivation indices: a scoping review and qualitative analysis: study examines socioeconomic deprivation indices. Health Aff. 2022;41:1804–1811. [DOI] [PubMed] [Google Scholar]
- 8. Hannan EL, Wu Y, Cozzens K, Anderson B. The neighborhood atlas area deprivation index for measuring socioeconomic status: an overemphasis on home value. Health Aff. 2023;42:702–709. doi: 10.1377/hlthaff.2022.01406 [DOI] [PubMed] [Google Scholar]
- 9. Yost K, Perkins C, Cohen R, Morris C, Wright W. Socioeconomic status and breast cancer incidence in California for different race/ethnic groups. Cancer Causes Control. 2001;12:703–711. doi: 10.1023/a:1011240019516 [DOI] [PubMed] [Google Scholar]
- 10. Knighton AJ, Savitz L, Belnap T, Stephenson B, VanDerslice J. Introduction of an area deprivation index measuring patient socioeconomic status in an integrated health system: implications for population health. EGEMS (Wash DC). 2016;4:1238. doi: 10.13063/2327-9214.1238 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Singh GK. Area deprivation and widening inequalities in US mortality, 1969–1998. Am J Public Health. 2003;93:1137–1143. doi: 10.2105/ajph.93.7.1137 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Boscoe FP, Liu B, Lee F. A comparison of two neighborhood‐level socioeconomic indexes in the United States. Spat Spatio‐Temporal Epidemiol. 2021;37:100412. doi: 10.1016/j.sste.2021.100412 [DOI] [PubMed] [Google Scholar]
- 13. Rocca WA, Yawn BP, Sauver JLS, Grossardt BR, Melton LJ. History of the Rochester Epidemiology Project: half a century of medical records linkage in a US population. Mayo Clin Proc. 2012;87:1202–1213. doi: 10.1016/j.mayocp.2012.08.012 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. St Sauver JL, Grossardt BR, Leibson CL, Yawn BP, Melton LJ III, Rocca WA. Generalizability of epidemiological findings and public health decisions: an illustration from the Rochester Epidemiology Project. Mayo Clin Proc. 2012;87:151–160. doi: 10.1016/j.mayocp.2011.11.009 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Rocca WA, Grossardt BR, Brue SM, Bock‐Goodner CM, Chamberlain AM, Wilson PM, Finney Rutten LJ, St Sauver JL. Data resource profile: expansion of the Rochester Epidemiology Project medical records‐linkage system (E‐REP). Int J Epidemiol. 2018;47:368–368j. doi: 10.1093/ije/dyx268 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Partnership for Quality Measurement . Core Quality Measure Collaborative (CQMC). 2025. Accessed July 28. https://p4qm.org/CQMC.
- 17. Chen G, Khan N, Walker R, Quan H. Validating ICD coding algorithms for diabetes mellitus from administrative data. Diabetes Res Clin Pract. 2010;89:189–195. doi: 10.1016/j.diabres.2010.03.007 [DOI] [PubMed] [Google Scholar]
- 18. Quan H, Khan N, Hemmelgarn BR, Tu K, Chen G, Campbell N, Hill MD, Ghali WA, McAlister FA. Validation of a case definition to define hypertension using administrative data. Hypertension. 2009;54:1423–1428. doi: 10.1161/HYPERTENSIONAHA.109.139279 [DOI] [PubMed] [Google Scholar]
- 19. Markey C, Bello O, Hanley M, Loehrer AP. The use of area‐level socioeconomic indices in evaluating cancer care delivery: a scoping review. Ann Surg Oncol. 2023;30:2620–2628. doi: 10.1245/s10434-023-13099-x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Rural Health Research Center . RUCA data: using RUCA data. 2022. Accessed July 14. https://depts.washington.edu/uwruca/ruca‐uses.php.
- 21. Zou G. A modified Poisson regression approach to prospective studies with binary data. Am J Epidemiol. 2004;159:702–706. doi: 10.1093/aje/kwh090 [DOI] [PubMed] [Google Scholar]
- 22. Vuong QH. Likelihood ratio tests for model selection and non‐nested hypotheses. 1989;57:307–333. doi: 10.2307/1912557 [DOI] [Google Scholar]
- 23. Flanagan BE, Gregory EW, Hallisey EJ, Heitgerd JL, Lewis B. A social vulnerability index for disaster management. J Homel Secur Emerg Manag. 2011;8:0000102202154773551792. doi: 10.2202/1547-7355.1792 [DOI] [Google Scholar]
- 24. Acevedo‐Garcia D, McArdle N, Hardy EF, Crisan UI, Romano B, Norris D, Baek M, Reece J. The child opportunity index: improving collaboration between community development and public health. Health Aff. 2014;33:1948–1957. doi: 10.1377/hlthaff.2014.0679 [DOI] [PubMed] [Google Scholar]
- 25. Stephens CQ, Yap A, Vu L, Saito JM, Barry D, Shui AM, Cockrell H, Cairo S, Wakeman D, Berman L, et al. Comparative analysis of indices for social determinants of health in pediatric surgical populations. JAMA Netw Open. 2024;7:e2449672. doi: 10.1001/jamanetworkopen.2024.49672 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Rollings KA, Noppert GA, Griggs JJ, Melendez RA, Clarke PJ. Comparison of two area‐level socioeconomic deprivation indices: implications for public health research, practice, and policy. PLoS One. 2023;18:e0292281. doi: 10.1371/journal.pone.0292281 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Rehkopf DH, Phillips RL. The neighborhood atlas area deprivation index and recommendations for area‐based deprivation measures. Health Aff. 2023;42:710–711. doi: 10.1377/hlthaff.2023.00282 [DOI] [PubMed] [Google Scholar]
- 28. Butler DC, Petterson S, Phillips RL, Bazemore AW. Measures of social deprivation that predict health care access and need within a rational area of primary care service delivery. Health Serv Res. 2013;48:539–559. doi: 10.1111/j.1475-6773.2012.01449.x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Franchi L, Barreto Parra PN, Chorniy A, Weston BW, Meurer J, Whittle J, Ackermann R, Black BS. A Comparative assessment of measures of area‐level socio‐economic status. Northwestern Law Econ Res Paper. 2024;23–43. [Google Scholar]
- 30. Xu J, Lawrence KG, O'Brien KM, Jackson CL, Sandler DP. Association between neighbourhood deprivation and hypertension in a US‐wide cohort. J Epidemiol Community Health. 2022;76:268–273. doi: 10.1136/jech-2021-216445 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Blazel MM, Perzynski AT, Gunsalus PR, Mourany L, Gunzler DD, Jones RW, Pfoh ER, Dalton JE. Neighborhood‐level disparities in hypertension prevalence and treatment among middle‐aged adults. JAMA Netw Open. 2024;7:e2429764. doi: 10.1001/jamanetworkopen.2024.29764 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Baker‐Smith CM, Yang W, McDuffie MJ, Nescott EP, Wolf BJ, Wu CH, Zhang Z, Akins RE. Association of area deprivation with primary hypertension diagnosis among youth Medicaid recipients in Delaware. JAMA Netw Open. 2023;6:e233012. doi: 10.1001/jamanetworkopen.2023.3012 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Vintimilla R, Seyedahmadi A, Hall J, Johnson L, O'Bryant S. Association of Area Deprivation Index and hypertension, diabetes, dyslipidemia, and obesity: a cross‐sectional study of the HABS‐HD cohort. Gerontol Geriatr Med. 2023;9:23337214231182240. doi: 10.1177/23337214231182240 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34. Kurani SS, Lampman MA, Funni SA, Giblon RE, Inselman JW, Shah ND, Allen S, Rushlow D, McCoy RG. Association between area‐level socioeconomic deprivation and diabetes care quality in US primary care practices. JAMA Netw Open. 2021;4:e2138438. doi: 10.1001/jamanetworkopen.2021.38438 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Angier H, Green B, Fankhauser K, Marino M, Huguet N, Larson A, DeVoe J. Role of health insurance and neighborhood‐level social deprivation on hypertension control following the affordable care act health insurance opportunities. Soc Sci Med. 2020;265:113439. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Kaiser Family Foundation (KFF) . Medicaid in Minnesota. 2024. Accessed March 6. https://files.kff.org/attachment/fact‐sheet‐medicaid‐state‐MN.
- 37. Everett E, Mathioudakis N. Association of area deprivation and diabetic ketoacidosis readmissions: comparative risk analysis of adults vs children with type 1 diabetes. J Clin Endocrinol Metab. 2019;104:3473–3480. doi: 10.1210/jc.2018-02232 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38. Laffey TN, Marr D, Modany A, McGraw M, Mounarath T, Bryk A, Christian N, Good C. Area deprivation index impact on type 2 diabetes outcomes in a regional health plan. J Manag Care Spec Pharm. 2024;30:1375–1384. doi: 10.18553/jmcp.2024.30.12.1375 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39. Laraia BA, Karter AJ, Warton EM, Schillinger D, Moffet HH, Adler N. Place matters: neighborhood deprivation and cardiometabolic risk factors in the Diabetes Study of Northern California (DISTANCE). Soc Sci Med. 2012;74:1082–1090. doi: 10.1016/j.socscimed.2011.11.036 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40. Sommet N, Weissman DL, Cheutin N, Elliot AJ. How many participants do I need to test an interaction? Conducting an appropriate power analysis and achieving sufficient power to detect an interaction. Adv Methods Pract Psychol Sci. 2023;6:25152459231178728. doi: 10.1177/25152459231178728 [DOI] [Google Scholar]
- 41. Williams DR, Collins C. Racial residential segregation: a fundamental cause of racial disparities in health. Public Health Rep. 2001;116:404–416. doi: 10.1093/phr/116.5.404 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42. Strauss AT, Moughames E, Jackson JW, Malinsky D, Segev DL, Hamilton JP, Garonzik‐Wang J, Gurakar A, Cameron A, Dean L, et al. Critical interactions between race and the highly granular area deprivation index in liver transplant evaluation. Clin Transpl. 2023;37:e14938. doi: 10.1111/ctr.14938 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43. Monsivais P, Amiri S, Denney JT, Amram O. Area deprivation amplifies racial inequities in premature mortality: analysis of 242,667 deaths in Washington state, USA 2011‐15. Health Place. 2020;61:102261. doi: 10.1016/j.healthplace.2019.102261 [DOI] [PubMed] [Google Scholar]
- 44. Rosenthal EL, Brownstein JN, Rush CH, Hirsch GR, Willaert AM, Scott JR, Holderby LR, Fox DJ. Community health workers: part of the solution. Health Aff (Millwood). 2010;29:1338–1342. doi: 10.1377/hlthaff.2010.0081 [DOI] [PubMed] [Google Scholar]
- 45. Njeru JW, Ridgeway JL, Asiedu GB, Radecki Breitkopf C, Gunderson JM, Quirindongo‐Cedeño O, O'Brien MW, Nelson TM, Buzard R, Wieland ML. Evaluating a community‐placed and clinically integrated community health worker program: a realist approach. J Ambul Care Manage. 2019;42:116–127. doi: 10.1097/jac.0000000000000268 [DOI] [PubMed] [Google Scholar]
- 46. Lines LM, Fowler CI, Kaganova Y, Peacock K. Development and validation of a community risk score for sexual and reproductive health in the United States. Health Affairs Scholar. 2024;2:2. doi: 10.1093/haschl/qxae048 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47. Lines LM, Long MC, Zangeneh S, DePriest K, Piontak J, Humphrey J, Subramanian S. Composite indices of social determinants of health: overview, measurement gaps, and research priorities for health equity. Popul Health Manag. 2023;26:332–340. doi: 10.1089/pop.2023.0106 [DOI] [PubMed] [Google Scholar]
- 48. Brown EM, Franklin SM, Ryan JL, Canterberry M, Bowe A, Pantell MS, Cottrell EK, Gottlieb LM. Assessing area‐level deprivation as a proxy for individual‐level social risks. Am J Prev Med. 2023;65:1163–1171. doi: 10.1016/j.amepre.2023.06.006 [DOI] [PubMed] [Google Scholar]
- 49. Armstrong BG. Effect of measurement error on epidemiological studies of environmental and occupational exposures. Occup Environ Med. 1998;55:651–656. doi: 10.1136/oem.55.10.651 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Tables S1–S6
Figures S1–S2