Where You Live Matters: Area Deprivation Predicts Poor Survival and Liver Transplant Waitlisting

Abstract

Social determinants of health (SDOH) are an important predictor of poor clinical outcomes in chronic diseases, but its associations among the general cirrhosis population and liver transplantation (LT) are limited. We conducted a retrospective, multi-institutional analysis of adult (≥18-years-old) patients with cirrhosis in metropolitan Chicago to determine the associations of poor neighborhood-level SDOH on decompensation complications, mortality, and LT waitlisting. Area Deprivation Index (ADI) and covariates extracted from the American Census Survey were aspects of SDOH that were investigated. Among 15,101 patients with cirrhosis, the mean age was 57.2 years; 6,414 (42.5%) were women, 6,589 (43.6%) were Non-Hispanic White, 3,652 (24.2%) were Non-Hispanic Black, and 2,662 (17.6%) were Hispanic. Each quintile increase in area deprivation was associated with poor outcomes in decompensation (sHR 1.07; 95%CI 1.051.10; P<0.001), waitlisting (sHR 0.72; 95%CI 0.67–0.76; P<0.001), and all-cause mortality (sHR 1.09; 95%CI 1.06–1.12; P<0.001). Domains of SDOH associated with lower likelihood of waitlisting and survival included low income, low education, poor household conditions, and social support (P<0.001). Overall, patients with cirrhosis residing in poor neighborhood-level SDOH had higher decompensation, mortality, and were less likely to be waitlisted for LT. Further exploration of structural barriers towards LT or optimizing health outcomes are warranted.

1.0. Introduction

Cirrhosis prevalence and mortality have increased by 2-fold over the past two decades in the United States (US).^1–4 Liver transplantation (LT) is the only curative treatment for cirrhosis but is highly limited by the organ shortage and access to specialty care. Although LT waitlisting is largely guided by severity of liver disease, there are still disparities in access to LT waitlisting, particularly among historically disadvantaged populations, and its drivers are poorly understood.^5–10

Social determinants of health (SDOH) have emerged as an important risk factor in chronic disease, accounting for as much as 40% of the total variance in clinical outcomes.¹¹ SDOH is multidimensional and encompasses economic, education, health care, neighborhood/built environments, and social contexts. Various tools, such as the area deprivation index (ADI), has been used to capture the multifactorial nature of SDOH to approximate the associations of neighborhood-level socioeconomic status (SES) by ZIP code with clinical outcomes such as: rehospitalization rates, cancer mortality, surgical morbidity, and poor access to care.^12–17

Despite the growing literature of SDOH in chronic diseases^12–14,16, their associations are still poorly understood among patients with cirrhosis. SDOH may influence the entire continuum of cirrhosis care and single institution investigations in the United States (US) involving select cohorts have suggested that SDOH may influence LT access and mortality.^18–20 Although disparities among race/ethnicity, gender, age, and insurance status have been documented, these characteristics can be imprecise measurements of SDOH.^21–23 In addition to the use of validated tool such as the ADI, it is important to identify which specific SDOH are modifiable to advocate for targeted public health policy. Thus, we used a large longitudinal, population cohort of patients in the Chicago metropolitan area to examine the associations between ADI, its specific SDOH domains, and cirrhosis-related outcomes. We hypothesized that patients residing in more SES deprived areas have higher risks for hepatic decompensation, mortality, and are less likely to be waitlisted for LT.

2.0. Materials and Methods

2.1. Study Design

This was a retrospective, longitudinal cohort study using the HealthLNK data repository from January 1^st, 2006 through December 31^st, 2012. The findings from our study were reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines.²⁴ The Northwestern University Institutional Review Board deemed this study exempt from review and waived the need for patient informed consent.

The HealthLNK dataset is a de-duplicated database from six large health care systems in the greater metropolitan Chicago area and includes: Northwestern Medicine, University of Chicago Hospitals and Clinic, Rush University Medical Center, University of Illinois at Chicago Medical Center, Loyola University Medical Center, and Cook County Health and Hospitals System. The United Network for Organ Sharing (UNOS) dataset is a national registry of all patients waitlisted for organ transplantation in the US and was merged with the HealthLNK database. Death data were extracted from the Social Security Death Master File for the state of Illinois and also merged as previously described.^25,26

2.2. Setting and Subjects

Adult (≥18-years-old) patients with cirrhosis (N=20,020) were identified by having at least one of the validated International Classification of Diseases, 9th Revision (ICD-9) codes.^27–31 The index date of cirrhosis diagnosis, the first time a patient was diagnosed with cirrhosis, was defined as the first appearance of a cirrhosis diagnosis code. Patients with an index date prior to June 1^st, 2006, were excluded (N=1,662, 8.3%) to limit selection bias of patients who may have had advanced cirrhosis prior to the beginning of the observational period. Patients were also excluded from analyses if they received a LT prior to cirrhosis diagnosis code, had missing ZIP Code (N=1,514, 7.6%), or if they resided outside of Illinois (N=1,743, 8.7%) since it is likely these patients would receive routine cirrhosis care in their home state (Figure 1). Follow-up was calculated from the time of first recorded diagnosis of cirrhosis to the date of: last known alive, death, LT receipt, or the end of observation period (12/31/2012). LT receipt and death were identified as the competing events of interest.

Figure 1.

Flow diagram of inclusion and exclusion criteria.

2.3. Outcomes and Covariates

Complete list of codes used in this study were outlined in Table S1 of the Supplement. The primary outcome of our study was mortality while secondary outcomes include decompensation complications and LT waitlisting. Demographic information was captured at the index date and included age, sex, race and ethnicity, and insurance type. Clinical comorbidity burden was defined by the Charlson Comorbidity Index.²⁷ Etiologies of cirrhosis were defined by ICD-9 codes and includes: alcohol-related, metabolic dysfunction associated steatohepatitis (MASH, formally termed non-alcoholic steatohepatitis [NASH]), hepatitis C virus (HCV), hepatitis B virus (HBV), cholestasis (e.g., primary sclerosing cholangitis, primary biliary cirrhosis), autoimmune, hemochromatosis, and Wilson’s disease.^25,28–31 Hepatic decompensation complications were defined as the occurrence of any of the following diagnoses identified by medications, ICD-9 or CPT codes: hepatic encephalopathy (HE), ascites, spontaneous bacterial peritonitis (SBP), variceal bleeding, hepatorenal syndrome (HRS) and/or hepatopulmonary syndrome (HPS). The Model for End-Stage Liver Disease with Sodium (MELD-Na) score was calculated with the four components of lab data (serum creatinine, bilirubin, international normalized ratio [INR], and sodium) when they were present within 90 days of each other and inclusion code. Patients receiving dialysis were assigned a creatinine cap of 4 mg/dL per UNOS guidelines.^32,33 If multiple scores were available, the highest value was reported since the LT organ allocation policy prioritizes the sickest patients first based on their maximum MELD-Na score.²⁵ Missing laboratory results were encountered in 6,940 (45.9%) patients. Multiple imputation was not conducted in order to minimize bias and regression analyses were limited to cohorts with calculated MELD-Na scores. Sensitivity analyses comparing patients with and without MELD-Na were further described in the Supplement (Table S2).

2.4. ADI and other SDOH covariates

A conceptual framework for the potential relationship between SDOH covariates, clinicodemographic characteristics, and outcomes was provided in Figure 2 and adapted from Kardashian et al.³⁴ and Mohamed et al.¹⁹ The ADI is a publicly available dataset maintained by the University of Wisconsin School of Medicine and Public Health.^13,14 It is a 17-point composite score, from 0 to 100, that aggregates SDOH variables from the American Community Survey (ACS) and provides a standardized ranking to identify disadvantaged census tracts. The ADI and individual SDOH from the ACS were available at the level of the census tract and linked with our database at the ZIP code level with a crosswalk provided by the U.S. Department of Housing and Urban Development.³⁵ Because multiple census tracts can exist in a single ZIP code, we calculated the population-weighted mean ADI for each ZIP code from each census-tract as previously published.^13,14 Higher ADI scores signify areas with lower SES. Patients were categorized into the following groups based on the ZIP code’s ADI where they reside: Least Deprived (1^st ADI quintile; 0–20), Less Deprived (2^nd ADI quintile; 21–40), Moderately Deprived (3^rd ADI quintile; 41–60), More Deprived (4^th ADI quintile; 61–80), and Most Deprived (5^th ADI quintile; 81–100) Areas (Figure 3). Specific SES covariates that were incorporated into the ADI composite score were publicly available at the level of the census tract from the 2012 ACS. In order to link these variables with our dataset, the population-weighted means of each ACS covariate for each ZIP code was extracted from each census tract in the same manner ADI was calculated. The following neighborhood-level SDOH domains were included for our analysis:

Figure 2.

Conceptual framework for the role of social determinants of health and how they may contribute to disparities in cirrhosis outcomes.^1,2

Figure 3.

Map of Illinois and the distribution of ZIP codes and Area Deprivation Index (ADI) quintiles.

• Income: per capita income, median family income, proportion (%) of families below the federal poverty level, income disparity.

• Education: % with <9 years of education, % with ≥high school diploma, % unemployed.

• Housing conditions/Transportation: median home value, median gross rent, median monthly mortgage, % owner-occupied housing units, % households with >1 person per room, % households without a motor vehicle, % occupied housing units without complete plumbing.

• Social Support: % single-parent households with children.

2.5. Statistical Analyses

Demographic variables were compared using analysis of variance (ANOVA) and chi-square tests for categorical and continuous variables, respectively. Two-sided P-values were used and those with an alpha level <0.05 were considered statistically significant. The Kaplan-Meier method and Restricted Mean Survival Time (RMST) were used to conduct time-to-event analyses. The Kaplan-Meier method estimates the probability of mortality and time to first decompensation event stratified by ADI category. Differences in survival estimates among each ADI category were compared using log-rank tests. Differences in Restricted Mean Survival Time (RMST) between the Least (ADI 1^st Quintile) and Most (ADI 5^th Quintile) Deprived Areas were tested using bootstrap resampling.^36,37 RMST analysis was conducted to measure the average time difference until decompensation, LT waitlist, or death between the Least and Most Deprived Areas over 3-, 6-, 12-, 24-, and 60-months of follow-up.³⁶ Fine-Gray multivariable, competing-risk survival analyses were conducted to identify specific clinicodemographic covariates associated with decompensation, LT waitlisting, and mortality. LT and mortality were the competing events of interest. Individual SDOH from the ACS that were accounted by the ADI were incorporated into the Fine-Gray multivariable regression analyses to estimate their unique associations with all-cause mortality, LT waitlisting, and decompensation. The subdistribution hazard ratios for all other variables incorporated into the model were reported in Table S3. To assess modification effects between race and insurance with ADI, we added several sensitivity analyses reported in Table S4. Pearson correlation coefficients of SDOH variables incorporated within the ADI were conducted to investigate collinearity and reported in Table S5. To avoid collinearity, each SDOH risk factor was included singularly into the regression model, separately from the ADI and all other SDOH risk factor. Thus, a new regression model was constructed for each ACS covariate (15) and the table reported is the aggregate of multiple models. All models predicting all-cause mortality and LT waitlisting were adjusted for known predictors of morbidity and mortality such as age³⁸, race and ethnicity²¹, sex^8,22, insurance type³⁹, cirrhosis etiology⁴⁰, decompensating events⁴¹, HCC⁴², MELD-Na^25,32, and Charlson Comorbidity Index.³⁸ Models predicting decompensation adjusted for the same covariates but did not include decompensation events since they were used as the outcome of the model. Variables in the model were expressed in subdistribution hazard ratios (sHR) with respective 95% confidence intervals (95%CI). Data processing and analysis were performed using R studio (version 4.1.0).

3.0. Results

3.1. Cohort Characteristics

Among 15,101 patients with cirrhosis, the mean (±SD) age was 57.2 (±11.7) years; 6,414 (42.5%) were women, 6,589 (43.6%) were Non-Hispanic White, 3,652 (24.2%) were Non-Hispanic Black, 2,662 (17.6%) were Hispanic, and 435 (2.9%) were Asian. There were 7,698 (51.0%) patients enrolled in Medicare/Medicaid, 4,858 (32.2%) in private insurance, and 2,545 (16.9%) in other insurance. Patients had a mean follow-up time of 26.5 (±23.7) months. The most frequent cirrhosis etiologies were alcohol-related (N=5,974, 39.6%), MASH (N=3,284, 21.8%) and HCV (N=6,310, 41.8%). Mean Charlson score was 4.8 (±3.13) and MELD-Na was 17.9 (±8.8) (Table 1).

Table 1.

Socioeconomic demographics of cirrhotic patients stratified by quintiles of Area Deprivation Index (ADI).

Characteristic	Full Cohort (N=15,101)	Least Deprived (ADI 1st Quintile, N=1533)	Less Deprived (ADI 2nd Quintile, N=4153)	Moderately Deprived (ADI 3rd Quintile, N=4342)	More Deprived (ADI 4th Quintile, (N=4267)	Most Deprived (ADI 5th Quintile, N=806)	P-value
Age, year, mean (±SD)	57.16 (±11.66)	58.94 (12.40)	57.39 (11.79)	57.01 (11.63)	56.6 (11.48)	56.34 (10.20)	<0.001
Female, No. (%)	6414 (42.5%)	600 (39.1%)	1709 (41.2%)	1853 (42.7%)	1880 (44.1%)	372 (46.2%)	<0.001
Race, No. (%)
NHW	6589 (43.6%)	978 (64.8%)	2279 (54.9%)	2108 (48.6%)	966 (22.6%)	258 (32.0%)	<0.001
NHB	3652 (24.2%)	159 (10.4%)	399 (9.6%)	778 (17.9%)	1916 (44.9%)	400 (49.6%)	<0.001
Hispanic	2662 (17.6%)	138 (9.0%)	659 (15.9%)	857 (19.7%)	959 (22.5%)	49 (6.1%)	<0.001
Asian	435 (2.9%)	60 (3.9%)	221 (5.3%)	113 (2.6%)	36 (0.8%)	5 (0.6%)	<0.001
Other	1763 (11.7%)	198 (12.9%)	595 (14.3%)	486 (11.2%)	390 (9.1%)	94 (11.7%)	<0.001
Insurance, No. (%)
Medicare/Medicaid	7698 (51.0%)	803 (52.4%)	1957 (47.1%)	2198 (50.6%)	2289 (53.6%)	451 (56.0%)	<0.001
Private	4858 (32.2%)	614 (40.1%)	1505 (36.2%)	1477 (34.0%)	1068 (25.0%)	194 (24.1%)	<0.001
Other	2545 (16.9%)	116 (7.6%)	691 (16.6%)	667 (15.4%)	910 (21.3%)	161 (20.0%)	<0.001
SDOH Variables
ADI, mean (±SD)	47.60 (±20.16)	12.7 (5.50)	30.35 (5.41)	48.3 (5.57)	67.11 (5.68)	85.93 (4.28)	<0.001
Per capita income, mean (±SD)	$28,908 (±14,980)	$58175 ($20157)	$34694 ($9718)	$25508 ($6014)	$18300 ($5176)	$17901 ($5260)	<0.001
Median family income in US dollars, mean (±SD)	56948 ($23070)	$90728 ($27199)	$67162 ($18756)	$56462 ($15018)	$39491 ($10841)	$35107($10791)	<0.001
% Families below federal poverty level, mean (±SD)	16.8% (±10.7%)	9.4% (6.0%)	11.4% (7.8%)	14.3% (8.3%)	25.1% (9.3%)	28.3% (12.5%)	<0.001
Income disparity, mean (±SD)	0.44 (±0.06)	0.49 (0.07)	0.44 (0.06)	0.42 (0.06)	0.44 (0.05)	0.46 (0.05)	<0.001
% Population aged 25 years or older with less than 9 years of education, mean (±SD)	16.7% (±11.0%)	4.7% (3.5%)	11.7% (6.8%)	17.5% (9.8%)	24.5 (11.5%)	19.5% (6.4%)	<0.001
% Population aged 25 years or older with at least a high school diploma, mean (±SD)	25.9% (8.7%)	11.6% (7.9%)	21.9% (6.6%)	28.1% (6.2%)	31.2% (4.8%)	34.3 (4.4%)	<0.001
% Civilian labor force population aged 16 years and older who are unemployed, mean (±Sd)	3.2% (±2.5%)	1.3% (2.1%)	2.3% (2.2%)	3.1% (2.6%)	4.3% (1.9%)	5.2 (3.1%)	<0.001
% Single-parent households with children younger than 18, mean (±SD)	35.2% (±21.0%)	20.8% (15.3%)	23.5% (13.1%)	31.2% (14.6%)	51.0% (20.2%)	60.7 (20.5%)	<0.001
Home value in US dollars, mean (±SD)	$243,728 ($106,460)	$438804 ($146337)	$304006 ($54378)	$220272 ($32117)	$165060 ($28084)	$104546 ($21053)	<0.001
Gross rent in US dollars, mean (±SD)	$974 (±$226)	$1231 ($280)	$1070 ($230)	$952 ($161)	$848 ($119)	773 ($134)	<0.001
Monthly mortgage in US dollars, mean (±SD)	$1,932 (±$465)	$2716 ($569)	$2193 ($261)	$1887 ($175)	$1574 ($192)	$1232 ($184)	<0.001
% Owner-occupied housing units, mean (±SD)	60.6% (±20.0%)	64.6% (20.9%)	64.3% (20.4%)	65.3% (19.8%)	51.8% (16.8%)	55.1% (16.4%)	<0.001
% Households with more than 1 person per room, mean (±SD)	4.0% (±3.4%)	1.2% (1.1%)	2.8% (2.1%)	4.1% (2.9%)	6.2% (4.3%)	3.0% (1.7%)	<0.001
% Households without a motor vehicle, mean (±SD)	8.8% (±9.3%)	12.0% (13.4%)	7.3% (8.4%)	6.3% (7.9%)	11.2% (8.5%)	11.1% (9.9%)	<0.001
% Occupied housing units without complete plumbing, mean (±SD)	2.2% (±2.4%)	0.8% (0.7%)	1.1 (1.1%)	1.6% (1.2%)	3.8% (2.4%)	6.1% (4.8%)	<0.001

ADI = Area Deprivation Index, HCC = hepatocellular carcinoma, HE = hepatic encephalopathy, NHB = Non-Hispanic Black, NHW = Non-Hispanic White, SBP = spontaneous bacterial peritonitis, SD = standard deviation, SDOH = social determinants of health

There were 1,533 (10.2%) patients that resided in Least Deprived Areas, 4,153 (27.5%) in Less Deprived Areas, 4,342 (28.8%) in Moderately Deprived Areas, 4,267 (28.3%) in More Deprived Areas, and 806 (5.3%) in Most Deprived Areas. For the total cirrhosis cohort, the mean ADI score was 47.6 (±20.2) (Table 1). Patients in increasingly deprived areas had increasing Charlson and MELD-Na scores despite being younger (P<0.001). Patients residing in increasingly deprived areas experienced higher proportions of decompensating events (P<0.001), while all-cause mortality were similar (P=0.058) (Table 2).

Table 2.

Clinical outcomes of cirrhotic patients stratified by quintiles of Area Deprivation Index (ADI).

Characteristic	Full Cohort (N=15,101)	Least Deprived (ADI 1st Quintile, N=1533)	Less Deprived (ADI 2nd Quintile, N=4153)	Moderately Deprived (ADI 3rd Quintile, N=4342)	More Deprived (ADI 4th Quintile, (N=4267)	Most Deprived (ADI 5th Quintile, N=806)	P-value
Mean days of follow-up time (±SD)	26.49 (23.73)	27.2 (24.06)	27 (22.74)	26.74 (22.11)	27.17 (21.97)	24.63 (20.8)	0.051
Charlson Comorbidity Index, mean (±SD)	4.76 (3.13)	4.66 (3.17)	4.66 (3.13)	4.64 (3.04)	5.01 (3.20)	4.78 (3.11)	<0.001
Etiologies of Cirrhosis, No. (%)
HBV	1324 (8.8%)	137 (8.9%)	407 (9.8%)	359 (8.3%)	368 (8.6%)	53 (6.6%)	0.018
HCV	6310 (41.8%)	544 (35.5%)	1602 (38.6%)	1742 (40.1%)	2018 (47.3%)	404 (50.1%)	<0.001
Alcohol-related	5974 (39.6%)	553 (36.1%)	1595 (38.4%)	1751 (40.3%)	1759 (41.2%)	316 (39.2%)	0.003
MASH	3284 (21.8%)	349 (22.8%)	924 (22.3%)	983 (22.6%)	871 (20.4%)	157 (19.5%)	0.033
Cholestasis	1552 (10.3%)	208 (13.6%)	458 (11.0%)	464 (10.7%)	355 (8.3%)	67 (8.3%)	0.002
Autoimmune	574 (3.8%)	44 (2.9%)	148 (3.6%)	186 (4.3%)	164 (3.8%)	32 (4.0%)	0.130
Hemochromatosis	132 (0.9%)	16 (1.0%)	46 (1.1%)	36 (0.8%)	31 (0.7%)	3 (0.4%)	0.156
Wilson’s Disease	49 (0.3%)	6 (0.4%)	14 (0.3%)	18 (0.4%)	11 (0.3%)	0 (0.0%)	0.644
MELD-Na	17.85 (8.80)	16.04 (7.44)	17.19 (8.12)	17.79 (8.68)	18.83 (9.06)	19.04 (8.86)	<0.001
Listed Patients, No. (%)	1109 (7.3%)	202 (13.2%)	346 (8.3%)	340 (7.8%)	181 (4.2%)	40 (5.0%)	<0.001
Median days on Waitlist (IQR)	19.35 (24.98)	19.63 (26.15)	19.23 (25.21)	19.36 (25.01)	18.57 (23.04)	22.35 (26.30)	<0.001
Transplant, No. (%)	567 (3.8%)	106 (6.9%)	181 (4.4%)	178 (4.1%)	82 (1.9%)	20 (2.5%)	<0.001
Decompensation, No. (%)	7507 (49.7%)	706 (46.1%)	2028 (48.8%)	2195 (50.6%)	2138 (50.1%)	440 (54.6%)	<0.001
Ascites	5330 (35.3%)	512 (33.4%)	1450 (34.9%)	1600 (36.9%)	1473 (34.5%)	295 (36.6%)	0.067
SBP	1012 (6.7%)	98 (6.4%)	290 (7.0%)	296 (6.8%)	283 (6.6%)	45 (5.6%)	0.645
HE	5421 (35.3%)	513 (33.4%)	1473 (34.9%)	1598 (36.9%)	1509 (34.5%)	328 (36.6%)	<0.001
Variceal Bleeding	4443 (29.4%)	415 (27.1%)	1318 (31.7%)	1311 (30.2%)	1225 (28.7%)	174 (21.6%)	<0.001
HRS	960 (6.4%)	99 (6.5%)	277 (6.7%)	311 (7.2%)	237 (5.6%)	36 (4.5%)	0.005
HPS	40 (0.3%)	6 (0.4%)	14 (0.3%)	10 (0.2%)	8 (0.2%)	2 (0.3%)	0.571
HCC, No. (%)	2089 (13.8%)	248 (16.2%)	613 (14.8%)	572 (13.2%)	542 (12.7%)	114 (14.1%)	0.003
Mortality, No. (%)	4837 (32.0%)	481 (31.4%)	1316 (31.7%)	1355 (31.2%)	1393 (32.7%)	292 (36.2%)	0.058
Liver-related	3045 (20.2%)	300 (19.6%)	857 (20.6%)	852 (19.6%)	851 (19.9%)	185 (23.0%)	0.221
Non-Liver	1247 (8.3%)	143 (9.3%)	328 (7.9%)	329 (7.6%)	374 (8.8%)	73 (9.1%)	0.024
Non-descript	545 (3.6%)	38 (2.5%)	131 (3.2%)	174 (4.0%)	168 (3.9%)	34 (4.2%)	0.016

HBV = hepatitis B virus, HCC = hepatocellular carcinoma, HCV = hepatitis C virus, HE = hepatic encephalopathy, HPS = hepatopulmonary syndrome, HRS = hepatorenal syndrome, MASH = metabolic dysfunction-associated steatohepatitis, MELD-Na = model for end-stage liver disease with sodium, SBP = spontaneous bacterial peritonitis, SD = standard deviation

3.2. Time-to-Event Analysis for Mortality, Decompensation, and LT Waitlisting

Patients residing in increasingly deprived areas had lower rates of waitlisting, LT, and spent longer time on the waitlist before receiving a LT (P<0.001) (Table 2). Patients residing in the Most Deprived Areas had faster times to hepatic decompensation complications and death over time (P=0.002, Figure 4).

Figure 4. Kaplan-Meier survival curve stratified by Area Deprivation Index (ADI) quintile.

Figure 4A.) Among ADI quintile categories, there was a difference in survival time between patients living in the 5^th ADI quintile (most deprived) and those living in less deprived ADI quintiles (p = 0.0021). Patients with cirrhosis who live in the 5^th ADI quintile had shorter survival times compared to patients who reside in areas with lower ADI. Figure 4B.) Among ADI quintile categories, there was a difference in time to decompensation (p<0.0001)

Over a five-year follow-up period, patients residing in the Most Deprived Area were associated with a reduction in lifespan by 4.8 months (95%CI 8.04–1.56, P=0.004) compared to those residing in the Least Deprived Areas. Compared to patients in the Least Deprived Areas, patients in the Most Deprived Areas experienced delays in LT waitlist culminating to a 7.58 month delay in 5-years of follow-up (ΔRMST 7.58 months; 95%CI 5.13–10.03; P<0.001). There were delays to LT waitlisting by 2.12 days (ΔRMST 0.07 months; 95%CI 0.04–0.10; P<0.001), 7.44 days (ΔRMST 0.24 months; 95%CI 0.15–0.32; P<0.001), 20.15 days (ΔRMST 0.65 months; 95%CI 0.43–0.88; P<0.001), and 59.21 days (ΔRMST 1.91; 95%CI 1.34–2.47; P<0.001) at 3-months, 6-months, 1-year, and 2-years of follow-up respectively. Comparisons of time to decompensation were not significantly different until 5 years of follow-up where patients residing in the Most Deprived Areas were associated with a faster time to hepatic decompensation complication compared to those residing in the Least Deprived Areas by 1 month (95%CI 0.02–1.97; P<0.001) (Table 3).

Table 3.

Restricted Mean Survival Time (RMST) analysis for all-cause mortality, transplant waitlist, and decompensation.

Time (ΔRMST [95% CI])	All-cause Mortality	Transplant Waitlist	Decompensation
3 months	0.02 (−0.04, 0.08)	0.07 (0.04, 0.10)**	0.00 (−0.05, 0.04)
6 months	0.03 (−0.11, 0.18)	0.24 (0.15, 0.32)**	0.01 (−0.08, 0.11)
1 year	−0.06 (−0.42, 0.29)	0.65 (0.43, 0.88)**	0.09 (−0.11, 0.29)
2 years	−0.75 (−1.59, 0.08)	1.91 (1.34, 2.47)**	0.37 (−0.07, 0.80)
5 years	−4.80 (−8.04, −1.56)*	7.58 (5.13, 10.03)**	1.00 (0.02–1.97)*

Comparison between the Most Deprived Area (ADI 5^th Quintile; reference) and Least Deprived Area (ADI 1^st Quintile).

CI = 95% confidence interval, RMST = restricted mean survival time, ∆RMST = change in restricted mean survival time measured in months

^*= p<0.01

^**= p<0.001

3.3. Competing Risk Multivariable Regression Analyses of Mortality, Decompensation, and LT Waitlisting

Each categorical increase of deprivation (per quintile increase of ADI) was associated with increased subdistribution hazards of hepatic decompensation complications (sHR 1.07; 95%CI 1.04–1.10; P<0.001) and all-cause mortality (sHR 1.09; 95%CI 1.06–1.12; P<0.001). There were various neighborhood-level SDOH covariates that were significantly (P<0.05) associated with hepatic decompensation complications, though their effect sizes may not be clinically significant. Notable specific neighborhood-level SDOH covariates from the ACS associated with increased subdistribution hazards of all-cause mortality were: unemployment (sHR 1.17; 95%CI 1.03–1.10; P<0.001), high school education or more (sHR 1.06; 95%CI 1.03–1.10; P<0.001), and single parent households with children (sHR 1.02; 95%CI 1.00–1.04; P=0.031) (Table 4). Additional sensitivity analyses accounting for the missingness of MELD-Na continued to show that ADI was an independent predictor of all-cause mortality, LT waitlisting, and decompensation (P<0.01) (Table S4). Interaction analyses did not demonstrate any modifying effect between ADI and insurance type (Table S4). There were no significant interactions between race/ethnicity and ADI except for those among non-Hispanic Black (sHR 0.91, 95%CI: 0.84–0.99, P=0.020) and Hispanic (sHR 0.88, 95%CI: 0.80–0.98, P=0.020) patients (Table S4).

Table 4.

Multivariate competing risk analyses^* for risk for all-cause mortality, transplant waitlist, and decompensation.

	All-cause Mortality			Transplant Waitlist			Decompensation
Risk Factor	HR	95% CI	p-value	HR	95% CI	p-value	HR	95% CI	p-value
Income
Per capita income for every $10,000 increase	0.96	0.94–0.98	<0.001	1.08	1.05–1.12	<0.001	1.00	1.00–1.00	<0.001
Median family income in US dollars per $10,000 increase	0.98	0.97–1.00	0.022	1.12	1.09–1.14	<0.001	1.00	1.00–1.00	<0.001
% Families below federal poverty level per 10% increase	1.03	0.99–1.06	0.150	0.62	0.57–0.68	<0.001	1.00	1.00–1.01	0.002
Income disparity per 0.1 increase of Gini Index	0.93	0.88–0.98	<0.01	0.73	0.65–0.81	<0.001	0.76	0.50–1.17	0.217
Education
% Population aged 25 years or older with less than 9 years of education per 10% increase	1.02	0.98–1.05	0.340	0.72	0.67–0.78	<0.001	1.00	1.00–1.01	0.029
% Population aged 25 years or older with at least a high school diploma per 10% increase	1.06	1.03–1.10	<0.001	0.98	0.92–1.05	0.601	1.01	1.00–1.01	<0.001
% Civilian labor force population aged 16 years and older who are unemployed per 10% increase	1.17	1.05–1.31	<0.01	0.65	0.49–0.84	<0.001	1.01	1.00–1.02	0.169
Social Support
% Single-parent households with children younger than 18 per 10% increase	1.02	1.00–1.04	0.031	0.80	0.76–0.84	<0.001	1.00	1.00–1.00	0.002
Housing Conditions
Median home value in US dollars per $10,000	0.97	0.96–0.99	<0.001	1.01	0.98–1.04	0.640	1.00	1.00–1.00	<0.001
Median gross rent in US dollars per $1,000	0.61	0.47–0.80	<0.001	2.38	1.48–3.83	<0.001	1.00	1.00–1.00	<0.001
Median monthly mortgage in US dollars per $1,000	0.87	0.81–0.93	<0.001	1.01	0.88–1.15	0.917	1.00	1.00–1.00	<0.001
% Owner-occupied housing units per 10% increase	1.00	0.98–1.02	0.906	1.23	1.18–1.28	<0.001	1.00	1.00–1.00	0.603
% Households with more than 1 person per room per 10% increase	0.96	0.87–1.07	0.466	0.43	0.34–0.55	<0.001	1.00	0.99–1.01	0.532
% Households without a motor vehicle per 10% increase	0.96	0.93–0.99	0.020	0.66	0.59–0.75	<0.001	1.00	1.00–1.00	0.772
% Occupied housing units without complete plumbing per 10% increase	0.98	0.85–1.14	0.802	0.15	0.08–0.26	<0.001	1.02	1.01–1.03	0.004

sHR = subdistribution hazard ratio, CI = confidence interval

^*Controlled for age, race, gender, insurance status, etiology of cirrhosis (HCV, HBV, alcohol, MASH, cholestasis), decompensation status (ascites, HE, esophageal varices, variceal bleeding, spontaneous bacterial peritonitis, hepatorenal syndrome, hepatopulmonary syndrome), HCC, MELD-Na, Charlson Comorbidity Index

^**A new model was constructed for each individual SDOH covariate and this table is an aggregate of multiple models. ADI and each individual SDOH risk factor was included into the model separately while adjusting for clinicodemographic covariates.

There was a lower likelihood of LT waitlisting (sHR 0.72; 95%CI 0.67–0.76; P<0.001) for each categorical increase of deprivation. Neighborhood-level SDOH covariates associated with decreased subdistribution hazards for LT waitlisting were: population living below the poverty level (sHR 0.62; 95%CI 0.57–0.68; P<0.001); with <9 years of education (sHR 0.72; 95%CI 0.67–0.78; P<0.001), unemployed (sHR 0.65; 95%CI 0.49–0.84; P<0.001), single-parent household with children (sHR 0.80; 95%CI 0.76–0.84; P<0.001), households with >1 person per room (sHR 0.43; 95%CI 0.34–0.55; P<0.001), households without a motor vehicle (sHR 0.66; 95%CI 0.59–0.75; P<0.001), and housing without complete plumbing (sHR 0.15; 95%CI 0.08–0.26; P<0.001). Conversely, per capita income (sHR 1.08; 95%CI 1.05–1.12; P<0.001), family income (sHR 1.12; 95%CI 1.09–1.14; P<0.001) were associated with increased subdistribution hazards for waitlisting (Table 4).

4.0. Discussion

There are gaps in the literature regarding the associations between neighborhood-level SDOH, the influence of its specific domains, and cirrhosis-related outcomes. The present study leverages one of the largest, longitudinal cohort of the general cirrhosis population to provide novel insights into how neighborhood-level SDOH may contribute to mortality and LT waitlisting. First, we observed that patients living in the Most Deprived Areas have increased risks of decompensation and mortality by 40% and 54%, respectively compared to those in the Least Deprived Areas. Moreover, patients residing in the Most Deprived Areas were 81% less likely to be waitlisted for LT compared to those in the Least Deprived Areas. Lastly, though SDOH is multifactorial, poor housing conditions were a notable SDOH risk factor associated with reduced rates of waitlisting.

For every categorical increase in area deprivation, the risks of decompensation and mortality increased by 7% and 9%, respectively. Within a 5-year follow-up, this culminates to an estimated 5-month decrease in lifespan between those residing in the Most and Least Deprived Areas. Our findings confirm prior single center observations by Strauss et al. who found that patients living in the most deprived neighborhood SES who were referred for LT had a 55% increased risk of mortality without LT evaluation compared to those in the least deprived neighborhoods.¹⁸ This association has also been replicated in the post-LT setting, as LT recipients residing in low SES areas had more than a 23% increased risk of mortality.^43–45 Along with the higher risk of mortality, we observed that patients living in more socially deprived neighborhoods were less likely to be waitlisted for LT by 28% and experienced delays to waitlisting of more than 7.5-months. This supports other single center reports of patients from low neighborhood SES who had decreased odds of waitlisting by 14–44%.^18,19 Our findings further advocate that patients with cirrhosis living in deprived areas are in need of attentive, targeted care and access. Disadvantaged areas that lack the resources to invest in the community often have fewer concentrations of primary care and specialty providers which could pose barriers in the referral and LT waitlist process.^5,46–48 The associations of neighborhood-level SES likely affect the entire spectrum of care and public health interventions that address structural barriers are necessary to improve cirrhosis outcomes.

The ADI composite score provides a practical method to account for the multifactorial nature of SDOH, but it is also important to consider the specific factors that drive these associations to develop targeted interventions. In line with prior research^43,45, patients residing in areas with lower median household income and education were associated with poor cirrhosis-related outcomes. However, our analyses also reveal that understudied SDOH risk factors such as poor household conditions were one of the largest drivers of lower likelihood of waitlisting and survival. For every 10% increase in the proportion of patients residing in areas with crowded housing (>1 person per room) or without complete plumbing, there were lower likelihood of waitlist placement by 57% and 85%, respectively. The association of poor household conditions is particularly notable among patients in the US, especially patients living in the Chicagoland area, given the city’s dark history of “redlining”.⁴⁹ “Redlining” was an institutionalized practice by the US Home Owners’ Loan Corporation in the 1930s to indicate “hazardous” areas and “risky” investments based on high concentrations of minority, immigrant, and working-class residents.⁵⁰ It has been nearly 50 years since the practice of redlining has been made illegal, but its enduring impact is still seen in the disproportionately higher prevalence of cardiovascular disease, asthma, obstetric, cancer, and poor health in these areas.^51–58 Although a distant past, considering these disparities within its historical context may help prioritize which communities may benefit most from public health initiatives.

Overall, our study provides evidence that neighborhood-level deprivation were associated with mortality, hepatic decompensation complications, and presents barriers towards LT selection. Identifying specific SDOH, such as poor housing conditions, can help lay the foundation for targeted public health policies and community-engaged interventions. For example, the Housing Choice Voucher program sponsored by the US Department of Housing and Urban Development may help low-income households improve their housing conditions through rent subsidies or provide opportunities to move to less impoverished neighborhoods.⁵⁹ Housing stipends that help families move from lower to higher SES areas have also been shown to improve chronic disease management and patient reported outcomes.^60,61 At the institutional level, outreach programs that engage historically disadvantaged communities, such as Northwestern Medicine’s Hispanic Transplant Program (HTP) and African American Transplant Access Program (AATAP), have successfully mitigated disparities to LT access.^62–64 Other initiatives, like University of California San Francisco’s DeLIVER Care Van or the Cherokee Nation Hepatitis C Virus Elimination program, aim to improve the cascade of cirrhosis care from screening, maintenance, and treatment.^65–67 Similar efforts such as facilitated transportation could improve access to care for patients in deprived neighborhoods.⁶⁸ Lastly, although technological innovation (e.g., telemedicine) have emerged as powerful tools to enhance patient outreach, structural barriers may remain. For example, those living in deprived areas may not have access to smart phone or tablet devices, have limited internet access or cellular data plan, or lack privacy for virtual visits.⁶⁸ Future research should intentionally investigate specific risk factors of SDOH in a community-engaged approach to most effectively deploy efforts to mitigate disparities to care.⁶⁸

There are important limitations to consider when interpreting our results. First, although the ADI is a comprehensive, validated measure of neighborhood SES, it does not include other important variables of SDOH such as those predicting food insecurity, access to care, and the built environment.^69–71 These risk factors may prove to also affect clinical outcomes and may be included in other measures of SES such as the Social Vulnerability Index, Social Deprivation Index, or Community Need Index.^72–74 However, we used the ADI since its aggregate score is not dependent on age or race, thus having more practical applications for public health policy that are not biased by these covariates. Second, because the ADI is a summation of various SDOH factors taken at the ZIP code level, attributing these characteristics to individual patients who reside in these areas should be done with caution to limit ecological fallacy. Compared to census tracts, some ZIP codes can span larger geographic areas where patterns of SDOH may be more heterogenous. Yet, studies using even larger geographic units (e.g., county)^75–77 have continued to elucidate disparities in outcomes. Furthermore, studies demanding more granularity than the census tract level may be challenging as they risk breaching patient anonymity and will likely further highlight observed differences. Third, we excluded patients without ZIP codes from our analysis, thus, we cannot generalize our findings to the homeless population. Yet, the homeless population only makes up an estimated 0.14% of the population in our metropolitan area.⁷⁸ Fourth, our database takes place from 2006–2012 and we recognize that neighborhood conditions in the present day may have changed since 2012. Future research should examine how these relationships may change over time and in a more contemporary cohort. Nevertheless, our findings underscore the significance of how SDOH may potentially impact health outcomes. Next, similar to other large databases, our study was limited by the amount of missingness, particularly when calculating the MELD-Na score. This might be due to the fact that our cohort is composed of the general cirrhosis population in the Chicago metropolitan area, many of whom may not have access to specialty- or liver-centric care where labs of the MELD-Na are routinely collected. Our cohort differs from the Scientific Registry of Transplant Patients (SRTR) or Organ Procurement Transplantation Network (OPTN) cohorts, all of whom are already on the LT waitlist and closely followed by specialists who regularly obtain these labs for waitlist prioritization. In addition to the 7.3% of patients that were merged by the UNOS database, our study also captures an additional 92.7% of patients who were not listed for LT, many of which never undergo evaluation or see a liver specialist. Lastly, similar to other studies that utilizes health databases, our study is retrospective in nature and our observations should not claim causality.^28–31

In conclusion, patients with cirrhosis residing in the Most Deprived Areas had increased risks of hepatic decompensation complications and mortality compared to those in the Least Deprived Areas. These patients were also 81% less likely to be listed for LT. These estimates culminate to a 5 month decrease in lifespan and 7.5 month delay to waitlist placement within 5 years. Specific SDOH domains that drive these associations include: income, education, housing conditions, and social support. Exploration of structural barriers towards LT access or optimizing health outcomes among patients in deprived areas are warranted to mitigate these disparities.

Abbreviations:

ACS

American Community Survey

ADI

Area Deprivation Index

ANOVA

analysis of variance

CI

confidence interval

HBV

hepatitis B virus

HCC

Hepatocellular carcinoma

HCV

hepatitis C virus

HE

hepatic encephalopathy

HPS

hepatopulmonary syndrome

HRS

hepatorenal syndrome

ICD

International Classification of Diseases

LT

liver transplantation

MELD-Na

Model for End-stage Liver Disease with Sodium

MASH

metabolic dysfunction associated steatohepatitis

NASH

non-alcoholic steatohepatitis

RMST

Restricted Mean Survival Time

SBP

spontaneous bacterial peritonitis

SDOH

Social Determinants of Health

SES

Socioeconomic Status

sHR

subdistribution hazard ratio

STROBE

Strengthening the Reporting of Observational Studies in Epidemiology

UNOS

United Network for Organ Sharing

US

United states