Association of Neighborhood Social Determinants of Health with Acute Kidney Injury during Hospitalization

Visual Abstract

Abstract

Key Points

• AKI is common among hospitalized patients. However, the contribution of neighborhood social determinants of health to AKI risk is not known.

• We found that among 26,769 hospitalized patients, 26% developed AKI. Patients who lived in the most disadvantaged areas (highest tertile of Area Deprivation Index) had a 10% greater odds of developing AKI than counterparts in the lowest Area Deprivation Index tertile. Patients who lived in rural areas had 25% greater odds of not recovering from AKI by hospital discharge.

• This study demonstrates an association between neighborhood disadvantage and rurality on the development of AKI and lack of recovery from AKI. Further work is needed to understand the mechanisms of these associations and to develop community-level interventions to mitigate the health care burden of AKI for disadvantaged populations.

Background

AKI is common among hospitalized patients. However, the contribution of social determinants of health (SDOH) to AKI risk remains unclear. This study evaluated the association between neighborhood measures of SDOH and AKI development and recovery during hospitalization.

Methods

This is a retrospective cohort study of adults without ESKD admitted to a large Southern US health care system from October 2014 to September 2017. Neighborhood SDOH measures included (1) socioeconomic status: Area Deprivation Index (ADI) scores, (2) food access: Low-Income, Low-Access scores, (3) rurality: Rural–Urban Commuting Area scores, and (4) residential segregation: dissimilarity and isolation scores. The primary study outcome was AKI on the basis of serum creatinine Kidney Disease Improving Global Outcomes criteria. Our secondary outcome was lack of AKI recovery (requiring dialysis or elevated serum creatinine at discharge). The association of SDOH measures with AKI was evaluated using generalized estimating equation models adjusted for demographics and clinical characteristics.

Results

Among 26,769 patients, 26% developed AKI during hospitalization. Compared with those who did not develop AKI, those who developed AKI were older (median 60 versus 57 years), more commonly men (55% versus 50%), and more commonly self-identified as Black (38% versus 33%). Patients residing in most disadvantaged neighborhoods (highest ADI tertile) had 10% (95% confidence interval, 1.02 to 1.19) greater adjusted odds of developing AKI during hospitalization than counterparts in least disadvantaged areas (lowest ADI tertile). Patients living in rural areas had 25% higher adjusted odds of lack of AKI recovery by hospital discharge (95% confidence interval, 1.07 to 1.46). Food access and residential segregation were not associated with AKI development or recovery.

Conclusions

Hospitalized patients from the most socioeconomically disadvantaged neighborhoods and from rural areas had higher odds of developing AKI and not recovering from AKI by hospital discharge, respectively. A better understanding of the mechanisms underlying these associations is needed to inform interventions to reduce AKI risk during hospitalization among disadvantaged populations.

Introduction

AKI, characterized by abrupt deterioration in kidney function, represents a complex systemic syndrome induced by numerous insults.^1,2 Furthermore, AKI confers an increased likelihood of developing incident or progressive CKD, ESKD, and multiple other adverse clinical outcomes and is, therefore, an important health care burden.^1–4

Social determinants of health (SDOH) broadly encompass the economic, structural, environmental, sociocultural, and health care context and have been linked with a heightened risk of CKD prevalence and faster progression of kidney disease across the United States.^5–11 AKI disproportionately affects underserved populations, including racial and ethnic minorities, older adults, and those with limited health care access.^7,12–18 Moreover, AKI and CKD rates are elevated in Southern states, which are characterized by greater rurality, high levels of neighborhood poverty, and inadequate social and health care infrastructure, alongside high levels of cardiometabolic disease burden compared with the rest of the United States.^19–23 Individuals residing in disadvantaged neighborhoods frequently confront challenges that may increase AKI risk, including exposure to environmental toxins, crowded living conditions, limited availability of healthy foods, and constrained health care access.^5,13,20 General associations between SDOH and health outcomes offer limited insights into opportunities to intervene. More granular understanding of specific factors (e.g., socioeconomic disadvantage, food access, rurality, and segregation) may offer insights into the mechanisms underlying the link between SDOH and AKI and inform efforts to improve equity in AKI prevention and management.^20,24–33 The neighborhood-level SDOH measures in this study, namely socioeconomic disadvantage, rurality, food access, and residential segregation, were selected on the basis of the existing literature noting their significant role for health disparities in cardiovascular-kidney-metabolic health.^28–32 We hypothesize that individuals living in low SDOH areas will have higher rates of AKI during hospitalization. If proven, potential interventions at the community level can help improve kidney health equity and mitigate AKI risk.^20,27–33

Electronic health records (EHRs) allow for the investigation of health care outcomes using real-world data from large patient populations.^34,35 Linkage of EHR with geographic information system tools enables investigation of the relationship between neighborhood-based SDOH indicators and AKI.^20,36,37 We leveraged data of over 26,000 hospitalized patients from one large health care system in the Southern United States to examine the associations of neighborhood-level SDOH measures with incident hospitalized AKI and AKI recovery. Our study focused on several key SDOH domains, including socioeconomic deprivation, food access, rural–urban location, and residential segregation.

Methods

Study Population

This retrospective cohort study used EHR data from adult patients 18 years and older who were hospitalized at the University of Alabama at Birmingham Hospital, a major academic medical center providing tertiary care to Alabama and neighboring states. We included patients who were discharged between October 1, 2014, and September 30, 2017. Patients were excluded if they had fewer than three inpatient serum creatinine (SCr) measurements; were prisoners were transferred from other hospitals; were classified as outpatients; had preexisting ESKD, including chronic dialysis or kidney transplantation (identified using International Classification of Diseases [ICD]-9/10 codes)³⁸; or lacked address information needed to link with neighborhood SDOH indicators or had a post office box address (Figure 1). The data were obtained from the institution's Oracle Cerner EHR data warehouse containing detailed information on demographics, vital signs, comorbidities (on the basis of ICD-9/10 codes—Supplemental Table 1), laboratory tests, procedures, medications, diagnoses, and hospital outcomes.^38,39 Demographic data were self-reported by the patients at the time of their first encounter with the health care system. Only the first hospitalization during the study period for each patient was evaluated. Individuals with extremes of body mass index (BMI) were excluded to avoid potential confounding of effect estimates by population outliers. The study was approved by the University of Alabama at Birmingham Institutional Review Board (No: 300009692).

Figure 1.

Study population selection flow chart. PO, post office.

SDOH Measures

The SDOH measures selected in this study were conceptualized as per the World Health Organization framework.⁴⁰ These neighborhood-level measures reflect the material circumstances, as part of the intermediary determinants in the World Health Organization framework. The patients' residential addresses at the time of admission were geocoded using Esri ArcGIS software and assigned a Census tract and block group identifier in the form of Federal Information Processing Standards (FIPS) code. FIPS codes were then linked to four neighborhood SDOH indicators.

1. Area Deprivation Index (ADI): ADI is a composite index of describing neighborhood socioeconomic status on the basis of 17 variables from the domains of income, education, employment, housing and living conditions, and family structure collected by the American Community Survey and aggregated to Census block groups.⁴¹ We used 2015 state ADI (scale 1–10, with higher scores denoting greater deprivation). Scores were categorized into tertiles on the basis of the variable's distribution in the primary analysis.

2. Low-Income, Low-Access (LILA) score: LILA is a dichotomous measure of food access on a Census tract level (scores of 1 indicate a food desert). Low income is defined as tracts with ≥20% poverty rate or median income <80% of the metropolitan area median income. Low access is defined as tracts where at least 500 people or 33% of population reside >1 mile (urban areas) or >10 miles (rural areas) from the nearest supermarket, supercenter, or large grocery store; vehicle access is included in the index (LILA version 2015).⁴²

3. Rural–Urban Commuting Area (RUCA) codes: RUCA represents a classification of census tracts along an urban–rural continuum on the basis of population density, urbanization, and daily commuting patterns (1–10 scale, RUCA version 2010) with values <4 indicating an urban area and values ≥4 indicating a rural area.⁴³ RUCA was considered as a dichotomous variable: urban (1–3) versus rural (4–10).

4. Residential segregation: This is a measure of spatial separation of racial or ethnic groups within neighborhoods, with higher values indicating greater segregation (US Census 2010). Two common measures of residential segregation were used: (1) evenness (dissimilarity index) and (b) exposure/isolation (isolation index).^20,44 These measures were categorized into tertiles on the basis of the variable distribution in the primary analysis.

Study Outcomes

The primary study outcome was the development of AKI during hospitalization on the basis of Kidney Disease Improving Global Outcomes SCr criteria: increase in SCr ≥0.3 mg/dl within 48 hours or ≥1.5 times baseline within 7 days.^1,2,45 Baseline SCr was the lowest value among the first three inpatient measures. AKI severity was categorized as stage 1 (increase in SCr ≥0.3 mg/dl within 48 hours or ≥1.5–1.9 times baseline within 7 days), 2 (increase in SCr ≥2.0–2.9 times baseline within 7 days), or 3 (increase in SCr ≥3 times baseline or to >4.0 mg/dl within 7 days or initiation of KRT).

Among patients who developed AKI and survived the hospitalization (9.6% of patients who died during the hospitalization were excluded to avoid competing risk of death), the secondary outcome that was evaluated was lack of AKI recovery by the time of hospital discharge. This was defined as (1) requiring dialysis within the last 72 hours before discharge or (2) last SCr before discharge ≥25% or ≥0.3 mg/dl above baseline SCr.

Covariates

Demographic covariates included age, sex, and self-reported race (Black versus other). Clinical covariates encompassed BMI, BP, smoking, baseline eGFR (derived using race-independent eGFR equation),⁴⁶ and comorbidities (defined per the Elixhauser Comorbidity Index on the basis of the ICD-9 or 10 codes; Supplemental Table 2).^35,38

Statistical Analyses

We compared the characteristics of patients who did and did not develop AKI during hospitalization using chi-square test for proportions of categorical variables and t tests or Wilcoxon rank-sum test for continuous variables as appropriate. We also compared the characteristics of patients who developed AKI and survived versus those who developed AKI and did not survive the hospitalization. We then compared the characteristics of patients with and without AKI by ADI tertiles, LILA (food desert, yes/no), RUCA (urban versus rural), index of dissimilarity tertiles, and social isolation tertiles. We assessed the correlation between each of the individual neighborhood SDOH measures using the Pearson correlation coefficient and ANOVA test. We used generalized estimating equation regression with an exchangeable working correlation matrix with robust standard errors to account for correlated observations between patients living in the same census tract (FIPS code).⁴⁷

We studied the association between each of the SDOH measures and outcomes using the following modeling: (1) model 1: unadjusted; (2) model 2: adjusted for age, sex, baseline eGFR, systolic BP, and Elixhauser score; and (3) model 3: model 2+race. We did not adjust for obesity in the model because BMI was strongly associated with the obesity measure included in the Elixhauser score. We also assessed whether the association between each of the SDOH measures and relevant outcomes differed by race (Black versus other), sex (male versus female), and obesity (obese: BMI ≥30 kg/m² versus BMI <30 kg/m²) because these have been independently associated with AKI recovery and SDOH.^17,48–50 When considering continuous SDOH measures (ADI and residential segregation measures), we used second-order terms to test for nonlinearities in the association with the outcomes. We did not find any evidence of nonlinearities (second-order terms P value > 0.05) and, therefore, present the models using linear terms. We defined statistical significance at P < 0.05.

In post hoc analyses, we studied the association of SDOH measures with developing stage 2 or 3 AKI compared with stage 1 AKI and lack of AKI recovery among those who survived the hospitalization and had AKI. Additional sensitivity analyses were conducted wherein the subset of patients with SCr measured 7 to 365 days before the index hospitalization. The closest SCr value before the index hospitalization was considered the baseline SCr and was used for estimating AKI. All analyses were performed using R (version 4.00, Vienna, Austria).

Results

Cohort Characteristics

The study cohort included 26,769 adults (Figure 1), of whom 26% (n=6940) developed AKI during their hospital stay. Most of the patients developed stage 1 AKI (93%), and only 7% developed stage 2 or 3 AKI (Table 1). Compared with those without AKI, patients who developed AKI were older (60 [47–71] versus 57 [42–68] years), more frequently men (55% versus 50%), and more likely to self-identify as Black (38% versus 33%). Patients who developed AKI had lower baseline eGFR (82 [51–107] versus 96 [75–115] ml/min per 1.73 m²), longer hospital length of stay (7 [4–13] versus 4 [3–7] days), and higher in-hospital mortality (3.2% versus 0.4%) compared with those who did not develop AKI. The correlation between all SDOH measures was statistically significant (P < 0.001) (Supplemental Table 2). The characteristics of patients by SDOH measure categories are shown in Supplemental Tables 3–7. Stage 1 AKI was more common among patients who lived in the highest ADI tertile (Supplemental Table 3), in LILA tracts (Supplemental Table 4), and in the highest social isolation tertile (Supplemental Table 7) compared with other categories. AKI was more common among patients living in urban areas than in rural areas (Supplemental Table 5).

Table 1.

Characteristics of study participants stratified by AKI status

Characteristic	No AKI, n=19,793 (74%)	AKI, n=6976 (26%)	P Value
Age, yr	57 (42–68)	60 (47–71)	<0.001
Male sex	9949 (50.3)	3856 (55.3)	<0.001
Race/ethnicity			<0.001
American Indian or Alaska Native	28 (0.1)	11 (0.2)
Asian	269 (1.4)	92 (1.3)
Black	6441 (32.5)	2625 (37.6)
Declined	166 (0.8)	57 (0.8)
Hispanic	303 (1.5)	109 (1.6)
Multiple	163 (0.8)	79 (1.1)
Other	9 (0)	3 (0)
Unknown	4 (0)	5 (0.1)
White	12,410 (62.7)	3995 (57.3)
Baseline SCr, mg/dl	0.8 (0.6–1.0)	1.0 (0.7–1.3)	<0.001
eGFR, ml/min per 1.73 m2	96.5 (74.6–114.8)	82.1 (50.8–10.6.8)	<0.001
Systolic BP	135 (119–153)	135 (116–155)	0.006
Diastolic BP	79 (69–88)	77 (67–88)	<0.001
BMI, kg/m2	28.1 (24.0–33.3)	28.3 (34.4–33.7)	<0.001
BMI ≥30 kg/m2	7704 (39.7)	2829 (41.1)	<0.001
AKI stage
1	—	6465 (92.7)	—
2	—	244 (3.5)	—
3	—	267 (3.8)	—
Required dialysis during hospitalization	11 (0.10)	184 (2.6)	<0.001
Hospital LOS, d	4 (3–7)	7 (4–13)	<0.001
Hospital mortality	76 (0.4)	220 (3.2)	<0.001
Elixhauser comorbidities
Congestive heart failure	2.862 (14.5)	1751 (25.1)	<0.001
Cardiac arrhythmia	1797 (9.1)	1002 (14.4)	<0.001
Diabetes	3764 (19.0)	1643 (23.6)	<0.001
Valvular heart disease	1412 (7.1)	743 (10.7)	<0.001
Pulmonary circulation disease	955 (4.8)	633 (9.1)	<0.001
Peripheral vascular disease	715 (3.6)	376 (5.4)	<0.001
Hypertension w/o complications	9922 (50.1)	3425 (49.1)	0.15
Hypertension with complications	2381 (12)	1676 (24)	<0.001
Paralysis	602 (3)	22 (3.2)	0.58
Other neurologic disorder	1152 (5.8)	404 (5.8)	0.96
Chronic pulmonary disease	3755 (19)	1419 (20.3)	0.01
Hypothyroidism	1925 (9.7)	752 (10.8)	0.01
Kidney failure	2076 (10.5)	1662 (23.8)	<0.001
Liver disease	1098 (5.5)	627 (9)	<0.001
Acute immunodeficiency syndrome	220 (1.1)	86 (1.2)	0.450
Lymphoma	244 (1.2)	89 (1.3)	0.828
Metastatic cancer	1402 (7.1)	449 (6.4)	0.072
Solid tumor without metastasis	2652 (13.4)	905 (13)	0.381
Rheumatoid arthritis	650 (3.3)	217 (3.1)	0.508
Coagulopathy	650 (3.3)	684 (9.8)	<0.001
Weight loss	1725 (8.7)	1304 (18.7)	<0.001
Fluid and electrolyte disorder	10,069 (50.9)	4938 (70.8)	<0.001
Blood loss anemia	122 (0.6)	49 (0.7)	0.49
Deficiency anemias	330 (1.7)	178 (2.6)	<0.001
Alcohol abuse	870 (4.4)	307 (4.4)	0.99
Drug abuse	946 (4.8)	319 (4.6)	0.51
Psychoses	336 (1.7)	112 (1.6)	0.65
Depression	2888 (14.6)	1002 (14.4)	0.66
Elixhauser score	3 (2–4)	4 (2–5)	<0.001
ADI—state scale (0–10, higher score reflects more disadvantaged areas)	5 (2–8)	5 (3–8)	<0.001
Low-income and low-access tract measured at 1 mile for urban and 10 miles for rural areas (yes versus no)	4493 (22.7)	1743 (25)	<0.001
RUCA (4–10)—rural	3302 (16.7)	1127 (16.2)	0.32
Segregation (all values range between 0 and 1, higher values indicate greater vulnerability)
Index of dissimilarity	0.62 (0.47–0.72)	0.62 (0.47–0.72)	0.99
Social isolation	0.49 (0.24–0.78)	0.53 (0.26–0.83)	<0.001

All continuous data are presented as median (interquartile range, 25th–75th centile). All categorical parameters are presented as N (%). ADI, Area Deprivation Index; BMI, body mass index; LOS, length of stay; RUCA, Rural–Urban Commuting Area; SCr, serum creatinine.

SDOH and Incident AKI

Patients who lived in the most disadvantaged neighborhoods (highest ADI tertile), when compared with those living in the least disadvantaged (lowest ADI tertile), had 10% higher odds of developing AKI during hospitalization (odds ratio [OR], 1.10 [1.02 to 1.19]) even after adjustment for demographics, clinical characteristics, and self-identified race (Table 2; model 3). The nested models demonstrated that the association between ADI and AKI was attenuated by race (OR for ADI in model 2: 1.22 and OR for ADI in model 3: 1.10). In the fully adjusted model, patients living in the most disadvantaged areas had 11% greater odds of developing AKI stage 1 compared with patients in the least disadvantaged areas (OR, 1.11 [1.02 to 1.20]) (Table 3). We did not observe an association of food access, rurality, and residential segregation with AKI development, overall or by stage after adjustment (Table 3). Moreover, there was no interaction between race, sex, or obesity with any of the SDOH measures on the odds of the primary outcome (P > 0.05 for all). We observed positive monotonous associations between incident AKI and ADI and residential segregation (dissimilarity and isolation) when used as continuous variables (Supplemental Figure 1, A–C).

Table 2.

Association of social determinants of health measures with AKI development in hospitalized patients

Neighborhood Social Determinants of Health Measures	Model 1, OR (95% CI)	Model 2 OR, (95% CI)	Model 3, OR (95% CI)
Area deprivation (ADI 1–10 state scale, in tertiles)
Low (1–3)	Reference	Reference	Reference
Medium (4–6)	1.08 (1.01 to 1.16)a	1.08 (1.00 to 1.16)	1.06 (0.98 to 1.14)
High (7–10)	1.21 (1.13 to 1.30)a	1.22 (1.13 to 1.31)a	1.10 (1.02 to 1.19)a
Food desert (LILA at 1 mile for urban, 10 miles for rural areas)	1.13 (1.06 to 1.21)a	1.13 (1,06 to 1.21)a	1.02 (0.95 to 1.10)
Rural (RUCA 4–10)	0.96 (0.89 to 1.04)	0.94 (0.87 to 1.02)	1.00 (0.92 to 1.08)
Residential segregation
Dissimilarity index
Low (0.005–0.52)	Reference	Reference	Reference
Medium (0.52–0.69)	1.02 (0.95 to 1.09)	0.98 (0.92 to 1.06)	0.97 (0.90 to 1.04)
High (0.69–1)	0.98 (0.92 to 1.05)	0.95 (0.88 to 1.02)	0.98 (0.91 to 1.05)
Isolation index
Low (0.01–0.32)	Reference	Reference	Reference
Medium (0.32–0.70)	1.07 (1.00 to 1.14)	1.06 (0.99 to 1.14)	1.02 (0.95 to 1.10)
High (0.70–1)	1.19 (1.11 to 1.27)a	1.18 (1.10 to 1.27)a	0.98 (0.90 to 1.07)

Model 1: unadjusted, model 2: adjusted for age, sex, baseline eGFR, systolic BP, and Elixhauser Comorbidity Index, model 3: model 2 additionally adjusted for race. We assessed interactions by race, sex, and obesity with each of social determinants of health measure, and none were statistically significant. ADI, Area Deprivation Index, CI, confidence interval; LILA, Low-Income, Low-Access; OR, odds ratio; RUCA, Rural–Urban Commuting Area.

^aP < 0.05.

Table 3.

Association of social determinants of health measures with AKI stratified by stage in hospitalized patients (no AKI [n=19,793] versus AKI stage 1 [n=6465] and no AKI [n=19,793] versus AKI stage 2/3 [n=511])

Neighborhood Social Determinants of Health Measures	Model 1, OR (95% CI)	Model 2, OR (95% CI)	Model 3, OR (95% CI)
Area deprivation (ADI 1–10 state scale, in tertiles)
AKI stage 1 versus no AKI
Low (1–3)	Reference	Reference	Reference
Medium (4–6)	1.08 (1.00 to 1.16)a	1.08 (1.00 to 1.16)a	1.05 (0.98 to 1.14)
High (7–10)	1.22 (1.13 to 1.31)a	1.23 (1.14 to 1.32)a	1.11 (1.02 to 1.20)a
AKI stage 2/3 versus no AKI
Low (1–3)	Reference	Reference	Reference
Medium (4–6)	1.17 (0.93 to 1.47)	1.08 (0.86 to 1.37)	1.07 (0.85 to 1.36)
High (7–10)	1.17 (0.93 to 1.48)	1.03 (0.82 to 1.31)	1.00 (0.78 to 1.28)
Food desert (LILA at 1 mile for urban, 10 miles for rural areas)
AKI stage 1 versus no AKI	1.14 (1.07 to 1.22)a	1.14 (1.07 to 1.22)a	1.03 (0.96 to 1.11)
AKI stage 2/3 versus no AKI	1.01 (0.82 to 1.25)	0.97 (0.78 to 1.20)	0.93 (0.74 to 1.17)
Rural (RUCA 4–10)
AKI stage 1 versus no AKI	0.96 (0.89 to 1.04)	0.94 (0.87 to 1.01)	1.00 (0.92 to 1.08)
AKI stage 2/3 versus no AKI	1.00 (0.79 to 1.26)	0.95 (0.75 to 1.22)	0.97 (0.75 to 1.25)
Measures of segregation
Dissimilarity index (highest tertile reflects the highest vulnerability tertile)
AKI stage 1 versus no AKI
Low (0.005–0.52)	Reference	Reference	Reference
Medium (0.52–0.69)	1.03 (0.96 to 1.10)	0.99 (0.92 to 1.07)	0.97 (0.91 to 1.05)
High (0.69–1)	0.98 (0.92 to 1.05)	0.95 (0.88 to 1.02)	0.98 (0.91 to 1.05)
AKI stage 2/3 versus no AKI
Low (0.005–0.52)	Reference	Reference	Reference
Medium (0.52–0.69)	0.90 (0.73 to 1.12)	0.84 (0.67 to 1.05)	0.84 (0.67 to 1.04)
High (0.69–1)	0.95 (0.77 to 1.18)	0.94 (0.76 to 1.16)	0.95 (0.76 to 1.18)
Isolation index (highest tertile reflects the highest vulnerability tertile)
AKI stage 1 versus no AKI
Low (0.01–0.32)	Reference	Reference	Reference
Medium (0.32–0.70)	1.06 (0.99 to 1.14)	1.06 (0.99 to 1.14)	1.02 (0.95 to 1.10)
High (0.70–1)	1.19 (1.11 to 1.27)a	1.19 (1.11 to 1.28)a	0.98 (0.90 to 1.07)
AKI stage 2/3 versus no AKI
Low (0.01–0.32)	Reference	Reference	Reference
Medium (0.32–0.70)	1.15 (0.92 to 1.42)	1.11 (0.89 to 1.38)	1.09 (0.87 to 1.36)
High (0.70–1)	1.16 (0.93 to 1.45)	1.06 (0.85 to 1.33)	0.99 (0.76 to 1.30)

Model 1: unadjusted, model 2: adjusted for age, sex, baseline eGFR, systolic BP, and Elixhauser Comorbidity Index, model 3: model 2 additionally adjusted for race. We assessed interactions by race, sex, and obesity with each of social determinants of health measure, and none were statistically significant. ADI, Area Deprivation Index; CI, confidence interval; LILA, Low-Income, Low-Access; OR, odds ratio; RUCA, Rural–Urban Commuting Area.

^aP < 0.05.

In post hoc analyses, we found no association between SDOH measures and stage 2 or 3 AKI development compared with stage 1 AKI (Supplemental Table 8). The results from the sensitivity analyses in the subset of patients with prehospitalization SCr available were similar to those of the primary analyses (Supplemental Tables 9 and 10).

SDOH and Lack of AKI Recovery

Among patients who developed AKI, 9.6% died before discharge and were additionally excluded from the cohort evaluated for the risk of AKI nonrecovery (Supplemental Table 12). Those who died were older, had lower baseline eGFR, and were more likely to have developed stage 3 AKI. SDOH measures were similar among those who developed AKI and survived compared with those who developed AKI and died during hospitalization (Supplemental Table 12). Of the 6305 who developed AKI and survived until discharge, 1662 did not recover kidney function (29 of whom were receiving dialysis on discharge) while 4643 recovered kidney function by the time of hospital discharge. Rurality was associated with lack of AKI, recovery with patients living in rural areas having 25% greater adjusted odds of AKI nonrecovery than those who were living in urban areas (95% confidence interval, 1.07 to 1.46). There was no association between the remaining SDOH measures and AKI recovery status by hospital discharge (Table 4). The association of food access (LILA) and AKI recovery differed by obesity status (P value for interaction=0.02). Patients who were not obese and lived in food deserts were 24% more likely to recover from AKI by hospital discharge (OR, 0.76 [0.62 to 0.93]). This association was not observed among obese patients. In addition, the association between rurality (RUCA) and AKI recovery differed by race (P value for interaction=0.02). Black adults who lived in rural areas were 25% more likely to recover from AKI by hospital discharge (OR, 0.75 [0.56 to 0.99]) when compared with their rural non-Black counterparts (OR, 1.05 [0.91 to 1.22]) (Supplemental Table 13). The results from the sensitivity analyses in the subset of patients with prehospitalization SCr are presented in Supplemental Table 11. We observed negative monotonous associations between lack of AKI recovery and both ADI and dissimilarity when used as continuous variables (Supplemental Figure 2, A and B). We also noted a positive monotonous association between lack of AKI recovery and isolation when used as a continuous variable (Supplemental Figure 2C).

Table 4.

Association of social determinants of health measures with lack of AKI recovery in hospitalized patients who survived to discharge

Neighborhood Social Determinants of Health Measures	Model 1, OR (95% CI)	Model 2, OR (95% CI)	Model 3, OR (95% CI)
Area deprivation (ADI 1–10 state scale, in tertiles)
Low (1–3)	Reference	Reference	Reference
Medium (4–6)	0.93 (0.80 to 1.07)	0.93 (0.80 to 1.08)	0.94 (0.81 to 1.10)
High (7–10)	0.87 (0.75 to 1.01)	0.82 (0.70 to 0.95)a	0.88 (0.75 to 1.03)
Food desert (LILA at 1 mile for urban, 10 miles for rural areas)	0.85 (0.75 to 0.97)a	0.82 (0.71 to 0.94)a	0.87 (0.76 to 1.01)
Rural (RUCA 4–10)	1.22 (1.05 to 1.41)a	1.30 (1.12 to 1.52)a	1.25 (1.07 to 1.46)a
Residential segregation
Dissimilarity index
Low (0.005–0.52)	Reference	Reference	Reference
Medium (0.52–0.69)	0.97 (0.85 to 1.11)	0.98 (0.85 to 1.13)	0.99 (0.86 to 1.14)
High (0.69–1)	0.92 (0.81 to 1.06)	0.95 (0.83 to 1.10)	0.93 (0.80 to 1.07)
Isolation index
Low (0.01–0.32)	Reference	Reference	Reference
Medium (0.32–0.70)	1.05 (0.92 to 1.21)	0.97 (0.84 to 1.12)	1.01 (0.87 to 1.17)
High (0.70–1)	0.97 (0.85 to 1.12)	0.88 (0.76 to 1.02)	1.03 (0.86 to 1.22)

Model 1: unadjusted, model 2: adjusted for age, sex, baseline eGFR, systolic BP, and Elixhauser Comorbidity Index, model 3: model 2 additionally adjusted for race: model 2, race. We checked interaction by race, sex, and obesity with each of social determinants of health measures; none were significant for all measures except for Low-Income, Low-Access scores and obesity (P for interaction=0.02) and Rural–Urban Commuting Area and Black race (P for interaction=0.0. ADI, Area Deprivation Index; CI, confidence interval; LILA, Low-Income, Low-Access; OR, odds ratio; RUCA, Rural–Urban Commuting Area.

^aP < 0.05.

Discussion

This large study using data from over 26,000 patients hospitalized in a tertiary care academic medical center in the Southern United States suggests that residence in a socioeconomically disadvantaged neighborhood may be an independent risk factor of incident AKI during hospitalization. Individuals in the most socioeconomically disadvantaged neighborhoods had a 9% higher odds of developing AKI compared with those in the least disadvantaged areas, and patients living in rural areas had 25% greater odds of not recovering from AKI by hospital discharge than those living in urban areas even after adjusting for demographics, clinical characteristics, and self-reported race. Other area-level characteristics, such as food access and residential segregation, were not associated with AKI development or AKI recovery after adjustment for these factors. However, some SDOH factors (e.g., food access and rurality) were differentially associated with kidney recovery according to obesity status and racial group.

The finding that higher socioeconomic neighborhood deprivation and residence in rural areas are associated with adverse AKI outcomes aligns with the growing body of evidence that highlights the role of socioenvironmental disadvantage on kidney disease.^{14–16,18,51} Disadvantaged neighborhoods and rural communities often lack resources supporting healthy lifestyles while exposing residents to societal and environmental stressors that may affect kidney health.^24,52 These findings necessitate the development of specialized outreach programs and telemedicine services to mitigate kidney disease risk in these underserved areas. While the comorbidity index was similar across the SDOH categories, there were notable differences in the prevalence of individual comorbidities associated with kidney disease, such as diabetes, hypertension, and congestive heart failure, in some categories. Constrained access to primary and specialty health care in under-resourced and rural communities may also contribute to worsening risk factors of kidney disease.²⁴ Furthermore, while there is disproportionate clustering of kidney disease risk factors in the Southeastern United States,^19,21 there may also be heterogeneity in their prevalence across various strata of neighborhood-level SDOH measures. Therefore, it is plausible that a higher burden of comorbidities known to be associated with kidney disease and delayed presentation due to challenging access to health care may have contributed to the observed associations of ADI and rurality with adverse AKI outcomes. The elevated risk in rural regions calls for an urgent reassessment of health care delivery models to ensure equitable health care access, thereby mitigating the heightened vulnerability of rural populations to adverse kidney health outcomes.

Prior investigations from different international contexts and heterogeneous health care infrastructures have elucidated the association between neighborhood-level SDOH measures and incident AKI.^{14–16,18,51} These studies, drawing upon data derived from universally public health care frameworks and hybrid systems, provide a substantive foundation for interpreting our results within the framework of the US health care landscape. This study from the Southeastern US health care system investigates a patient population landscape that is particularly marked by a concentrated prevalence of risk factors of kidney disease and lower socioeconomic status compared with the rest of the United States. Furthermore, our analysis extends the discourse by probing the intricacies of racial disparities and their interplay with socioeconomic factors in the context of hospitalized AKI.

The current investigation reinforces the need for multifaceted approaches to advance acute care nephrology research. AKI disproportionately affects populations experiencing socioeconomic disadvantage, with prior work indicating that AKI may exacerbate financial instability,^51,53,54 and this may in turn make affording healthy housing, nutrition, and health care access more difficult.^7,13,25 Further investigations should work to disentangle the complex bidirectional relationships between SDOH factors and AKI susceptibility and post-AKI outcomes.³⁴ This work may inform policymakers in making significant public health investments to eradicate the disproportionate burden of kidney disease shouldered by underserved and rural populations.²⁶ The heterogeneous associations of discrete SDOH domains with AKI incidence and recovery noted in this study underline the need to examine individual socioeconomic and environmental factors separately because composite indices may occlude ascertainment of the nuanced effects of individual components. In addition, this research highlights the need to understand the complex and interconnected ways SDOH are related to each other. Special attention was given to race, a social construct shaped by sociopolitical, geographic, and cultural factors,⁵⁵ which is associated with neighborhood disadvantage and health.⁴⁸

In this study, the associations between neighborhood-SDOH and AKI were consistently attenuated by race, which may be due to structural racism that leads to worse neighborhood conditions, worse health care access (especially culturally sensitive health care), and fewer resources for specific minority racial groups.⁴⁸ Deficits in self-advocacy among patients from marginalized demographic subgroups with diminished educational attainment, income status, and health literacy may also propagate disparities in kidney health. The insights from this study also suggest that socioeconomic factors may be potentially modifiable targets to ameliorate the public health burden of AKI, warranting screening, monitoring, and early preventative interventions in specific disadvantaged groups.^56–58 This investigation also highlights the importance of analyzing the effect of neighborhood-level SDOH on specific patient subpopulations according to obesity status and racial group. Overall, this work fits within a broader paradigm shift to move beyond clinical or traditional risk factors into evaluating population-level kidney health risk factors to incorporate their neighborhood and social realities.

This study has notable limitations. First, this analysis included a single Southern US health care system (with considerable representation of Black individuals), constraining generalizability to institutions serving different populations. Only individuals who interacted with the health care system were captured in this report. Second, individual-level socioeconomic specifications, including insurance, income, social support, and education, could not be captured from the EHR. Thus, residual confounding due to unmeasured individual-level parameters is plausible. Third, given the unavailability of prehospitalization SCr values in most of the study population, community-acquired AKI could not be differentiated from hospital-acquired AKI. However, sensitivity analyses, in which we limited our cohort to those with prehospitalization SCr data available, yielded consistent results with the primary analyses. Fourth, limited statistical power and predominantly mild forms of AKI (i.e., AKI stage 1) prevented more granular assessment of nuances across AKI severity stages. Fifth, our study findings should be interpreted within the context of the methods and definitions used to characterize AKI outcomes. Nevertheless, this study has several strengths including adjustment for confounders that affect AKI outcomes, use of real-world data from a substantial cohort of hospitalized patients, and use of geospatial techniques to link clinical data to multiple neighborhood-level SDOH measures, which have been previously evaluated in other fields of medicine but not in hospitalized AKI.^59–61 This large study from >26,000 hospitalized patients noted that patients from the most socioeconomically disadvantaged neighborhoods had greater likelihood of developing AKI during the hospitalization, and patients from rural areas had greater likelihood of not recovering from AKI by hospital discharge, even after accounting for demographics, clinical characteristics, and self-reported race. The association between neighborhood-level socioeconomic deprivation and AKI requires further investigation and offers a potential target for community-level interventions that could mitigate the health care burden of AKI. Specifically, understanding and ameliorating socioeconomic disparities may help advance health equity in AKI prevention, risk classification, and improving post-AKI outcomes.

Supplementary Material

cjasn-19-1371-s001.pdf ^{(1.4MB, pdf)}

cjasn-19-1371-s002.pdf ^{(571.3KB, pdf)}

Disclosures

Disclosure forms, as provided by each author, are available with the online version of the article at http://links.lww.com/CJN/C20.

Funding

O.M. Gutierrez: National Center for Advancing Translational Sciences (UL1TR003096). J.A. Neyra: National Institute of Diabetes and Digestive and Kidney Diseases (U01DK12998, R01DK128208, R01DK133539, and U54DK137307). This work was supported by National Institute of Diabetes and Digestive and Kidney Diseases (P30 DK079337).

Author Contributions

Conceptualization: Lama Ghazi, Orlando M Gutierrez, Javier A. Neyra.

Data curation: Elizabeth Baker, Orlando M. Gutierrez, Javier A. Neyra, Gabriela R. Oates, AKM Fazlur Rahman.

Formal analysis: Lama Ghazi.

Funding acquisition: Orlando M. Gutierrez, Javier A. Neyra.

Methodology: Xinyuan Chen, Lama Ghazi, Javier A. Neyra, Edward D. Siew.

Supervision: Javier A. Neyra.

Writing – original draft: Elizabeth Baker, Catherine R. Butler, Lama Ghazi, Orlando M. Gutierrez, Lucia Juarez, Ariann F. Nassel, Javier A. Neyra, Gabriela R. Oates, Vibhu Parcha, AKM Fazlur Rahman, Tomonori Takeuchi.

Writing – review & editing: Elizabeth Baker, Catherine R. Butler, Xinyuan Chen, Lama Ghazi, Orlando M. Gutierrez, Lucia Juarez, Ariann F. Nassel, Javier A. Neyra, Gabriela R. Oates, Vibhu Parcha, AKM Fazlur Rahman, Edward D. Siew, Tomonori Takeuchi.

Data Sharing Statement

All data are included in the manuscript and/or supporting information.

Supplemental Material

This article contains the following supplemental material online at http://links.lww.com/CJN/C21.

Supplemental Table 1. ICD-9/ICD-10 codes for Elixhauser comorbidity.

Supplemental Table 2. Correlation between social determinants of health (SDOH) measures evaluated in the study.

Supplemental Table 3. Characteristics of study participants by Area Deprivation Index at state-level tertiles during index hospitalization.

Supplemental Table 4. Characteristics of study participants by low-income and low-access tract measured at 1 mile for urban and 10 miles for rural areas during index hospitalization.

Supplemental Table 5. Characteristics of study participants by Rural–Urban Commuting Area during index hospitalization.

Supplemental Table 6. Characteristics of study participants by index of dissimilarity tertiles during index hospitalization.

Supplemental Table 7. Characteristics of study participants by social isolation tertiles during index hospitalization.

Supplemental Table 8. Association of social determinants of health (SDOH) measures among those with AKI by stage in hospitalized patients (overall).

Supplemental Table 9. Association of social determinants of health (SDOH) measures with AKI development in hospitalized patients (among those with prehospitalization serum creatinine data available).

Supplemental Table 10. Association of social determinants of health (SDOH) measures with AKI stratified by stage in hospitalized patients (among those with prehospitalization serum creatinine data available).

Supplemental Table 11. Association of social determinants of health (SDOH) measures with lack of AKI recovery in hospitalized patients who survived to discharge (among those with prehospitalization serum creatinine data available).

Supplemental Table 12. Characteristics of study participants with AKI status who died during hospitalization compared with survivors.

Supplemental Table 13. Association of low-income and low-access tract measured at 1 mile for urban and 10 miles for rural areas and Rural–Urban Commuting Area with no recovery of AKI in hospitalized patients who survived to discharge: stratified by obesity and race.

Supplemental Figure 1. (A–C) Association of each of Area Deprivation Index, index of dissimilarity, and social isolation with incident AKI during hospitalization.

Supplemental Figure 2. (A–C) Association of each of Area Deprivation Index, index of dissimilarity, and social isolation with lack of AKI recovery during hospitalization.