Community Socioeconomic Deprivation and Nonalcoholic Fatty Liver Disease Severity

Abstract

Background and Objectives:

Nonalcoholic fatty liver disease (NAFLD) is linked to obesity. Obesity is associated with lower socioeconomic status (SES). An independent link between pediatric NAFLD and SES has not been elucidated. The objective of this study was to evaluate the distribution of socioeconomic deprivation, measured using an area-level proxy, in pediatric patients with known NAFLD and to determine whether deprivation is associated with liver disease severity.

Methods:

Retrospective study of patients <21 years with NAFLD, followed from 2009 to 2018. The patients’ addresses were mapped to census tracts, which were then linked to the community deprivation index (CDI; range 0–1, higher values indicating higher deprivation, calculated from six SES-related variables available publicly in US Census databases).

Results:

Two cohorts were evaluated; 1 with MRI (magnetic resonance imaging) and/or MRE (magnetic resonance elastography) findings indicative of NAFLD (n = 334), and another with biopsy-confirmed NAFLD (n = 245). In the MRI and histology cohorts, the majority were boys (66%), non-Hispanic (77%–78%), severely obese (79%–80%), and publicly insured (55%–56%, respectively). The median CDI for both groups was 0.36 (range 0.15–0.85). In both cohorts, patients living above the median CDI were more likely to be younger at initial presentation, time of MRI, and time of liver biopsy. MRI-measured fat fraction and liver stiffness, as well as histologic characteristics were not different between the high- and low-deprivation groups.

Conclusions:

Children with NAFLD were found across the spectrum of deprivation. Although children from more deprived neighborhoods present at a younger age, they exhibit the same degree of NAFLD severity as their peers from less deprived areas.

The majority of patients with nonalcoholic fatty liver disease (NAFLD) are overweight or obese (1). Additionally, the degree of obesity is associated with liver disease severity in pediatric NAFLD (2). Currently, the only nonsurgical treatment for pediatric NAFLD is lifestyle modifications aimed at obesity reduction, avoidance of sugar-sweetened beverages (SSB), and increased physical activity, all of which can be challenging to implement or sustain (3–5). Barriers to such lifestyle changes may be related to social determinants of health (SDH), defined as “the conditions in which people are born, grow, live, work and age” (6). These conditions, often rooted in socioeconomic status (SES), may include differential access to greenspace, high-quality nutrition, and education, each tangibly linked to ability to make NAFLD-relevant lifestyle changes (7–12).

Although there have been numerous studies detailing associations between increased obesity prevalence among subjects of lower SES (13–19), the literature outlining relationships between NAFLD and SES is limited (20). Lower SES is associated with greater SSB consumption (7), increased insulin resistance and type 2 diabetes mellitus (T2DM) prevalence (21–24), and poorer sleep quality (25,26). These are known risk factors for severe NAFLD, independent of obesity, suggesting that socioeconomic deprivation may be directly associated with the development of severe NAFLD. Additionally, a recent report described a direct link between SES and liver disease. Kivimäki et al (27) performed a prospective cohort study in Finland, which revealed that exposure to disadvantaged neighborhoods in childhood was associated with the development of fatty liver in adulthood.

Across multiple studies, SDH have been approximated using composite, area-level data on factors relating to socioeconomic deprivation. Measured at the neighborhood or census tract level, area-level indices provide a glimpse into those social and environmental conditions in which health and disease occurs. One such index is the Community Deprivation Index (CDI), which generates a standardized score using measures of poverty, income, education, public assistance, housing, and insurance status (28,29). Our first objective was to evaluate the distribution of socioeconomic deprivation, as measured by the CDI, in 2 pediatric cohorts with known NAFLD (Radiographically Determined Cohort and Histologically Determined Cohort; Supplemental Figure 1, Supplemental Digital Content, http://links.lww.com/MPG/B730). Secondly, we sought to assess whether the CDI was linked to measures of NAFLD severity.

METHODS

Study Design

This was a retrospective cross-sectional study performed at the Steatohepatitis Center of Cincinnati Children’s Hospital Medical Center (CCHMC) after Institutional Review Board approval.

Inclusion/Exclusion Criteria

Patients younger than 21 years of age were included if they had been seen at the Steatohepatitis Center for histologically confirmed or presumed NAFLD (magnetic resonance imaging [MRI]-based evidence of steatosis in the context of a negative work up for other liver diseases) from 1 April 2009 through 30 July 2018. Exclusion criteria were secondary NAFLD (eg, because of underlying genetic disorder including mitochondrial myopathy and Duchenne Muscular Dystrophy), concurrent liver diseases (eg, autoimmune liver disease, Wilson’s disease, alpha-1 antitrypsin deficiency, lysosomal acid lipase deficiency, hemochromatosis), other chronic inflammatory or endocrinologic conditions (such as inflammatory bowel disease, type 1 diabetes mellitus, hypothyroidism, malignancy, psoriasis, and juvenile idiopathic arthritis), preceding weight loss surgery, or liver transplantation.

Variables

Clinical characteristics, concurrent diagnoses and medications, and imaging study results at the time of biopsy were collected. All parameters of interest were only deemed relevant for collection if obtained within 90 days of either liver biopsy or imaging. Previously outlined criteria were used to define BMI categories and T2DM status (2,30).

Cohorts

Radiographically Determined Nonalcoholic Fatty Liver Disease Cohort

Patients who had had MRI-PDFF and MRE performed for clinical indications and who met the aforementioned inclusion/exclusion criteria were studied in this cohort. Indications to obtain an MRI-PDFF/MRE clinically include persistent elevations in serum alanine aminotransferase (ALT >50 U/L) for at least 3 to 6 months typically in conjunction with a nonimproving BMI or metabolic comorbidities (eg, T2DM, dyslipidemia). For patients with multiple MRI-PDFF/MRE studies, the first available study with complete information (either MRI-PDFF or MRE) was included. Hepatic fat fraction (HFF) was only recorded if it had been measured via MRI-PDFF. MRE had been performed using an active-passive driver system at 60 Hz with 4 axial slices obtained through the mid-liver, as previously described (31). Liver stiffness was expressed as the weighted (by measurement region of interest) average of the stiffness values for each of the 4 elastograms.

Histologically Determined Nonalcoholic Fatty Liver Disease Cohort

Patients who had had a liver biopsy performed for clinical indications were studied in this cohort. Indications to obtain a liver biopsy in the clinical setting include a concern for more severe NAFLD (eg, persistent ALT elevation [>50 U/L] for at least 3 to 6 months without improvement despite lifestyle modifications, particularly in conjunction with a rising BMI or metabolic comorbidities, or concern regarding advanced fibrosis based on findings, such as splenomegaly or elevated liver stiffness on MRE) or a concern for an alternate underlying liver disease (eg, Wilson’s disease).

Outcomes of Interest

Radiographically Determined Nonalcoholic Fatty Liver Disease Cohort

Key outcomes were liver stiffness (kPA) and HFF. Elevated liver stiffness was defined as greater than 2.71 kPa, as this has previously been shown to detect advanced fibrosis in pediatric patients with 85% specificity (32).

Histologically Determined Nonalcoholic Fatty Liver Disease Cohort

The outcome was histologic NAFLD severity, scored by expert pathologists using the methodology previously validated by Nonalcoholic Steatohepatitis Clinical Research Network (NASH CRN) (33). The NAFLD Activity Score (NAS; score 0–8) was calculated for each patient as the sum of scores for hepatic steatosis (0–3), ballooning (0–2), and lobular inflammation (0–3). Portal inflammation (score 0–2) and fibrosis stage (0–4) were also documented.

Ecological Predictor

Our primary predictor, socioeconomic deprivation, was approximated using the CDI at the census tract level (28,29). A census tract is a geography defined to be sociodemographically homogeneous and composed of generally between 1200 and 8000 people (34). Using the appropriate software (DeGAUSS), we were able to link the patient’s home address at the time of liver biopsy or imaging study to a census tract. Linkage to a census tract in turn allowed for connection to a variety of measures publicly available from the 2011 to 2015 U.S Census American Community Survey. Six of these individual metrics have been compiled using principal components analysis to create 1 numerical representation of the local socioeconomic environment via the CDI (Supplemental Figure 2, Supplemental Digital Content, http://links.lww.com/MPG/B730) (28,29). The CDI ranges from 0 to 1, with a higher score reflecting a more deprived community. The detailed methodology of this process as well as data regarding the CDI for every census tract in the nation is publicly available (35,36). The median CDI for Hamilton County, Ohio (site of CCHMC) was recently found to be 0.38 (range: 0.12–0.85) (28).

Participants were excluded if their address was not available, could not be linked to a census tract (eg, PO Box), was from a foreign country, was listed as that of the hospital or Ronald McDonald House, or was the address of Jobs and Family Services Headquarters, which corresponds to the facility overseeing our local foster care system.

Statistical Analyses

Descriptive statistics enumerated the distribution of key variables, highlighting, for continuous variables, means with standard deviation (SD) or medians with ranges. The median CDI of our cohort was then used to subdivide the imaging and histology cohorts into 2 subgroups for comparison. Low- and high-CDI groups, split at the sample median, were compared using Student t test, Wilcoxon-Mann Whitney, Kruskal-Wallis, and chi-square testing as appropriate. Stata MP version 13.0 (College Station, TX) and SAS® version 9.4 (Cary, NC) were used for statistical analyses. Analysis of Covariance (ANCOVA) analyses were used to build prediction models for HFF and stiffness separately. The stepwise model selection procedures based on the predicted residual sum of squares (PRESS) was done through SAS GLMSELECT procedure. The candidate variables subject to selection were: age at MRE, ethnicity, T2DM status, CDI, and their 2-way interactions. CDI was tried both as a continuous variable, and as a binary variable (high and low) in the model selection. A similar process was followed for the histology cohort with NAS and fibrosis as the outcome variables.

RESULTS

Radiographically Determined Nonalcoholic Fatty Liver Disease Cohort

Of the 878 unique patients seen at the Steatohepatitis Center within the study period, 334 (38%) had an MRI performed. This cohort of patients was predominantly boys, non-Hispanic, publicly insured, and severely obese. The mean age at the time of MRI was 14 ± 3 years. Demographic and clinical characteristics of both study cohorts are summarized in Table 1.

TABLE 1.

Baseline clinical and laboratory characteristics of the study cohorts at the time of assessment

Variable	Result
Study cohort	Radiographic	Histologic
	N = 334	N = 245
Health insurance, n (%)
Public	184 (55%)	136 (56%)
Private	132 (40%)	91 (37%)
Unknown	18 (5%)	18 (7%)
Age at first clinic visit, mean years (SD)	13 (±3)	13 (±3)
Age at MRI/Bx, mean years (SD)	14 (±3)	14 (±3)
Time from first clinic visit to MRI/Bx, mean days (SD)	(447 ± 552)	332 (±402)
Obesity severity, n (%)
Normal weight	4 (<1%)	1 (<1%)
Overweight	8 (2%)	4 (2%)
Class I obesity	59 (18%)	45 (18%)
Class II obesity	120 (36%)	83 (34%)
Class III obesity	143 (43%)	112 (46%)
Comorbidities, n (%)
Type 2 diabetes mellitus	16 (5%)	24 (10%)
Serum biochemistries, median (range)
ALT, U/L	72 (18–538)	93 (22–459)
AST, U/L	36 (10–281)	49 (13–271)
GGT, U/L	36 (3–349)	48 (10–710)
ALP, U/L	167 (18–614)	171 (43–512)
HDL, mg/dL	39 (15–86)	37 (15–64)
LDL, mg/dL	86 (9–218)	92 (20–202)
Triglycerides, mg/dL	132 (13–1316)	144 (49–721)
Total cholesterol, mg/dL	156 (65–268)	162 (63–262)

ALP = alkaline phosphatase; ALT = alanine aminotransferase; AST = aspartate aminotransferase; GGT = gamma- glutamyltransferase; HDL = high-density lipoprotein; LDL = low-density lipoprotein; MRI = magnetic resonance imaging; SD = standard deviation.

Mean liver stiffness for the cohort was 2.5 ± 0.82 kPa. HFF was available for 153 (46%) of all MRIs, where the mean HFF was 21 ± 11%. The baseline radiographic findings are summarized in Supplemental Table 1 (Supplemental Digital Content, http://links.lww.com/MPG/B730).

The median CDI of the imaging cohort was 0.36 (range 0.14–0.85) (Fig. 1A). There was no association between CDI and serum ALT levels, HFF, or liver stiffness. Using the median CDI, the imaging cohort was further divided into low deprivation (<0.36) and high deprivation (≥0.36) cohorts (Table 2). Patients in the higher deprivation group were younger at time of referral to the Steatohepatitis clinic (12 vs 13 years, P = 0.01) and at time of imaging (13 vs 14 years, P < 0.01). Those from census tracts with higher deprivation were also more likely to be publicly insured (65% vs 35%, P < 0.01; Supplemental Figure 3, Supplemental Digital Content, http://links.lww.com/MPG/B730).

FIGURE 1.

(A) Community deprivation index at radiographic assessment. (B) Community deprivation index at liver biopsy.

TABLE 2.

Characteristics in low versus high-deprivation imaging cohort

	Variables	Low deprivation index (<0.36)	High deprivation index (≥0.36)	P value
Clinical characteristics	Sex; n (% male)	112 (51)	109 (49)	0.62
	Ethnicity; n (% Hispanic)	29 (39)	46 (61)	0.1
	Class III obesity, n, (%)	63 (44)	80 (56)	0.07
	Age at MRE; years	14 (± 3)	13 (± 3)	< 0.01
	Age at first clinic visit; years	13 (± 3)	12 (± 3)	0.01
	Time from first clinic visit to MRE; days	438 (± 557)	457 (± 448)	0.75
	Publicly insured; n (%)	64 (35)	120 (65)	< 0.01
	T2DM diagnosis; n (%)	7 (4)	9 (6)	0.64
	Metformin use; n, %	44 (27)	37 (22)	0.34
	Vitamin E use; n, %	11 (7)	13 (8)	0.69
Biochemical values	AST, U/L	47 (± 39)	45 (± 32)	0.65
	ALT, U/L	88 (± 66)	88 (± 60)	0.97
	GGT, U/L	50 (± 45)	49 (± 44)	0.72
	Alkaline phosphatase, U/L	186 (± 110)	201 (± 114)	0.25
	HDL, mg/dL	39 (± 9)	40 (± 10)	0.09
	LDL, mg/dL	88 (±29)	89 (±31)	0.88
	Triglycerides, mg/dL	160 (± 92)	148 (± 119)	0.29
	Cholesterol, mg/dL	158 (± 35)	157 (± 34)	0.75
	HgbA1c, %	5.4 (± 0.8)	5.4 (± 0.8)	0.99
Radiographic findings	Stiffness score, kPa	2.45 (± 0.5)	2.57 (± 1.0)	0.18
	Volume, mL	2200 (± 700)	2224 (± 695)	0.75
	High stiffness (>2.71), n (%)	43 (47)	49 (53)	0.53
	HFF % PDFF only (n = 151)	20 (± 11)	22 (± 12)	0.22

T-test, chi-square and Mann-Whitney U test. Reported values are mean (SD), unless otherwise indicated.

ALT = alanine aminotransferase; AST = aspartate aminotransferase; GGT = gamma-glutamyltransferase; HDL = high density lipoprotein; HgbA1c = glycated hemoglobin; LDL = low density lipoprotein; MRE = magnetic resonance elasticity; NAS = NAFLD activity score.

There was no difference in serum biochemistries by CDI grouping (Table 2). There was also no difference in HFF, liver stiffness or the proportion of patients with elevated liver stiffness (> 2.71 kPa), even after controlling for age, ethnicity, and diagnosis of T2DM across the 2 groups (P = 0.23).

Histologically Determined Nonalcoholic Fatty Liver Disease Cohort

Of the 878 unique patients seen at the Steatohepatitis Center within the study period, 245 (28%) underwent a liver biopsy (Supplemental Figure 1, Supplemental Digital Content, http://links.lww.com/MPG/B730). In this cohort, the majority were boys and non-Hispanic, with a mean age of at time of biopsy of 14 ± 3 years. The baseline clinical and laboratory characteristics of this cohort are summarized in Table 1.

The mean steatosis score was 2.0, lobular inflammation score was 1.2, ballooning score was 0.7, and NAS was 3.9. Most patients had grade 2 or 3 steatosis (n=168, 69%), NAS <5 (n=157, 64%), and stage 0 or 1 fibrosis (n=191, 78%). The baseline histologic findings are summarized in Supplemental Table 2 (Supplemental Digital Content, http://links.lww.com/MPG/B730).

CDI was possible to calculate for 233 of the patients included in the histologic cohort. Of the remaining 12 patients, 11 addresses corresponded to PO boxes, and 1 was an international address. These 12 patients did not differ with respect to age, sex, ethnicity, T2DM status, biochemical markers, or histologic severity from the 233 patients for whom CDI could be calculated. Across the included 233 patients, the median deprivation score was 0.36 (range: 0.15–0.85; Fig. 1B). As with the Radiographic cohort, the median CDI was used to divide Cohort 2 into low (<0.36) and high (≥ 0.36) deprivation groups (Table 3). Those from more deprived communities were younger at the time of liver biopsy (13 vs 14 years, P = 0.02) and more likely to have public health insurance (80% vs 38%, P < 0.01, Supplemental Figure 4, Supplemental Digital Content, http://links.lww.com/MPG/B730).

TABLE 3.

Characteristics in low versus high deprivation histology cohorts

	Variables	Low deprivation index (<0.36)	High deprivation index (≥0.36)	P value
Clinical characteristics	Sex, n (% male)	74 (66)	79 (66)	0.96
	Ethnicity, (% Hispanic)	23 (21)	29 (24)	0.48
	Severe obesity, n (%)	83 (74)	102 (86)	0.10
	Age at liver biopsy, years	14 (±3)	13 (±4)	0.02
	Age at first clinic visit, years	13 (±3)	12 (±4)	0.02
	Time from first clinic visit to liver biopsy, days	370 (±414)	306 (±400)	0.12
	Publicly insured, n (%)	38 (38)	92 (80)	<0.01
	T2DM diagnosis, n (%)	11 (10)	11 (10)	0.88
	Metformin use, n (%)	39 (35)	31 (27)	0.19
	Vitamin E use; n (%)	4 (4)	4 (4)	0.96
Biochemical values	AST, U/L	63 (±48)	58 (±34)	0.16
	ALT, U/L	117 (±80)	107 (±76)	0.15
	GGT, U/L	61 (±52)	64 (±75)	0.64
	Alkaline phosphatase, U/L	185 (±100)	203 (±115)	0.89
	HDL, mg/dL	37 (±9)	38 (±9)	0.89
	LDL, mg/dL	94 (±33)	97 (±31)	0.76
	Triglycerides, mg/dL	165 (±97)	154 (±74)	0.19
	Cholesterol, mg/dL	163 (±38)	166 (±33)	0.72
	Fasting glucose, mg/dL	96 (±29)	92 (±17)	0.10
	Insulin level, mcIU/mL	31 (±18)	29 (±17)	0.22
	A1c, %	5.5 (±0.9)	5.3 (±0.5)	0.05
Histologic findings	Steatosis score	2.1 (±0.8)	2.0 (±0.9)	0.34
	Lobular inflammation	1.3 (±0.7)	1.2 (±0.7)	0.09
	Ballooning	0.7 (±0.6)	0.6 (±0.6)	0.12
	NAS	4.1 (±1.5)	3.8 (±1.4)	0.11
	Patients with NAS ≥5, n (%)	43 (38)	38 (32)	0.30
	Portal inflammation	0.8 (±0.6)	0.8 (±0.6)	0.65
	Fibrosis score	0.9 (±1.0)	0.9 (±1.0)	0.57
	Patients with any fibrosis; n (%)	65 (58)	68 (57)	0.89
	Patients with advanced fibrosis (3–4), n (%)	9 (8)	12 (10)	0.59

T-test, chi-square and Mann-Whitney U test. Reported values are mean (SD), unless otherwise indicated.

A1c = glycated hemoglobin; ALT = alanine aminotransferase; AST = aspartate aminotransferase; GGT = gamma- glutamyltransferase; HDL = high-density lipoprotein; LDL = low-density lipoprotein; NAS = NAFLD activity score.

With respect to histology, there were no differences in the scores for steatosis, lobular inflammation, ballooning, NAS, and fibrosis between the CDI groups (Table 3). When investigating the CDI as a continuous variable, there was also neither association between CDI and NAS (P = 0.63), nor was there a difference in CDI scores between those with no/mild (stage 0–2) versus advanced (stage 3–4) fibrosis.

Biochemical markers of liver injury, dyslipidemia, and insulin resistance were not different between the high and low deprivation groups in the histology cohort (Table 3).

DISCUSSION

This is the first pediatric study to evaluate associations between social determinants of health and NAFLD and between community deprivation and liver disease severity. We showed that the distribution of the CDI in this cohort of patients with NAFLD was reflective of the region, suggesting that NAFLD is widely distributed geographically. We also demonstrated that children with NAFLD were found across the spectrum of deprivation, as measured using an area-based proxy. Patients living in more deprived communities were more likely to be publicly insured and younger at presentation to the steatohepatitis clinic. Although possession of public health insurance is typically associated with a lower SES, clinically it does not appear that public insurance coverage impacted care as these patients presented for evaluation earlier than their privately insured counterparts. Despite a younger age at both first clinic visit and at investigation (liver biopsy or imaging), patients living in more deprived communities had similar histologic and radiographic disease severity in the presence of comparable risk factors and comorbidities, including Hispanic ethnicity, obesity severity, T2DM diagnosis, and dyslipidemia.

Prior literature linking SDH to obesity provided a sound rationale to investigate whether such factors were linked to the severity of NAFLD. In that literature, the diminished availability of healthy food options and diminished access to an environment that encourages physical activity (eg, access to greenspace)—both challenges often encountered in the context of socioeconomic deprivation—have been hypothesized to increase the risk of obesity (9,15,37–40). Still, although links between such contextual measures and obesity have been well established, limited research has addressed specific socioeconomic aspects of adults with NAFLD. Using NHANES data, Rosenblatt et al (41) assessed the impact of certain SDH factors on liver disease severity in nearly 3.8 million adults with NAFLD and T2DM. They found that food insecurity, assessed using a validated survey, was associated with an increased likelihood of advanced fibrosis as determined using noninvasive biomarkers (ALT>40 and either elevated NAFLD fibrosis score, FIB-4 score, or AST-to-platelet ratio index). Although they did not find an association with other socioeconomic factors (education, insurance status, poverty level), the cross-sectional nature of their design, like ours, was not able to take into account duration of exposure to socioeconomic factors. The latter is important, as it has previously been associated with the risk of NAFLD development in adulthood (27).

Using prospective, population-based data collected on over 2000 individuals in Finland, Kivimäki et al (27) demonstrated an association between cumulative neighborhood disadvantage and risk of NAFLD development in adulthood. Kivimäki et al found that the prevalence of metabolic risk factors that were associated with NAFLD development (eg, obesity, dyslipidemia, insulin resistance. etc) increased over time in those who were more disadvantaged during childhood. More importantly, the cumulative time spent in a disadvantaged environment increased the risk of NAFLD development in adulthood by 73%, suggesting a complex interplay between early environmental exposures and adulthood risk of fatty liver development. Our study findings also suggest that exposure to community deprivation may play a role in the natural history of NAFLD, as we detected a significant difference in age at first clinic visit and at the time of MRI and biopsy, with children from more deprived neighborhoods presenting earlier. Interestingly, the NAFLD cohort included in our study was not more deprived than the general population living in the same area. In a study by Brokamp et al (28), approximately 27,000 children born in Hamilton County from 2013 to 2015 were followed to determine the association between the CDI and cost of hospital utilization in their first year of life. In this population, the median CDI was 0.38, whereas that of our patients was 0.36. This suggests that community factors that could hypothetically affect liver disease severity (eg, limited access to healthy diet, a safe environment for physical activity, and/or to outpatient clinic visits) were not more prevalent in our NAFLD population at the time of MRI or biopsy. However, given that the deprived cohort had similar disease severity at an earlier age, it is possible that socioeconomic deprivation is associated with earlier disease onset.

There is currently no singular method to assess area-level SDH. Examples of methods other than the CDI (28) to ecologically assess deprivation include the Childhood Opportunity Index (42), the Concentrated Disadvantage Index (43), and the Economic Hardship Index (44). Although the literature does not support the superiority of one of these indices over another, we considered the CDI to be an appropriate tool with which to investigate the associations between markers of the SDH and NAFLD severity for a variety of reasons. The individual variables which constitute this index; specifically: poverty, educational level, reliance on public assistance, housing value and vacancy, and insurance coverage status, have all been independently associated with SES at both the individual and area-level (8,17,45–48). In addition, the CDI has been used previously in Hamilton County (the main catchment area for our referrals) and has been found to be reflective of the breadth and range of deprivation present. Moreover, the CDI is available for every census tract within the United States, and thus, is highly generalizable. Lastly, should this index be found to be helpful in predicting disease severity it could be integrated within the electronic health record to enhance clinical care.

This study had several limitations. It is possible that the lack of differences in liver disease severity between those with high and low deprivation were the result of limited power. Still, the cohorts we report on are some of the largest available for children with a diagnosis of NAFLD. In addition, the study of patients who had undergone a liver biopsy to confirm the diagnosis of NAFLD may have introduced selection bias, as in clinical practice, liver biopsies are typically obtained in those with concern for more advanced liver disease. However, the inclusion of the imaging cohort has likely decreased the selection bias, as the thresholds to obtain imaging are lower. Regardless, both cohorts were selected and not representative of the entire spectrum of patients with pediatric NAFLD. Thirdly, this study followed the assumption that the address within the electronic health record at the time of biopsy was associated with sufficient exposure at the census tract level to affect the outcomes of interest. As a recent study has highlighted the importance of total elapsed time in deprivation, a short duration at a specific address may not significantly contribute to changes in disease (27). Furthermore, although we measured deprivation at the neighborhood level, it is important to recognize that CDI does not take into account the contribution of factors from the individual household level. Further investigation into links between household-level SES and pediatric NAFLD is warranted. Relatedly, the deprivation index may not adequately capture those SDH that are of most relevance to the NAFLD population. It is also possible that SDH may still indirectly affect NAFLD severity through its association with comorbidities, such as T2DM and obesity. In our study, the distribution of these risk factors did not differ between the low and high deprivation cohorts at the time of biopsy or imaging, precluding us from exploring this further. Finally, this study was cross-sectional by design, and therefore not able to address if SDH affect outcomes over time.

In conclusion, we demonstrated that patients with NAFLD were found across the spectrum of deprivation. Although we were unable to demonstrate independent associations between community deprivation and liver disease severity, it is possible that increased deprivation is associated with an earlier onset of liver disease considering the younger age at presentation of the more deprived cohorts. This remains to be investigated further. Should childhood deprivation be found to be associated with younger disease onset, the inclusion of neighborhood level data may help risk stratify patients with NAFLD in the future.

What Is Known

• Nonalcoholic fatty liver disease is closely linked to obesity.

• Obesity has been associated with lower socioeconomic status.

• There is limited literature addressing links between socioeconomic status and both nonalcoholic fatty liver disease severity and outcomes.

What Is New

• In large pediatric cohorts with histologically confirmed or radiographically presumed nonalcoholic fatty liver disease, we characterize the distribution of socioeconomic deprivation

• Children from more deprived neighborhoods may be at risk for earlier onset of nonalcoholic fatty liver disease.