Metabolic Dysfunction–Associated Steatotic Liver Disease Is Associated With Adverse Social Factors: An Analysis Using All ofUs Authors: Manuela Araque, Stephanie M. Garces, Hui Yu, Mengsheng Zha, Sakina Qazi, Yan Guo, David Goldberg, Vatche Agopian, Jacob L. McCauley, Yvonne N. Flores, Patricia D. Jones Published: 2026 ABSTRACT Background and AimsMetabolic dysfunction–associated steatotic liver disease (MASLD) is progressive, with estimated global prevalence exceeding 30%. Social determinants of health (SDOH) may impact MASLD risk and progression. However, this has not been fully characterized. We harnessed the power of the All of Us Research Program (AoU) dataset to conduct a comprehensive analysis examining the association between various SDOH and MASLD.MethodsWe conducted a retrospective cross-sectional analysis of the AoU database. We identified participants with MASLD using ICD-9 and -10 codes and excluded individuals with other chronic liver diseases or self-reported heavy alcohol use. Healthy control participants were devoid of chronic liver diseases, heavy alcohol use, and comorbidities associated with MASLD if obese. We examined SDOH by combining relevant questions from various AoU surveys and conducted univariate and multivariate logistic regression to examine the association between various SDOH and MASLD.ResultsAfter matching cases to controls by age and race/ethnicity, the sample of 57,895 participants had a 1:3 case to control ratio; 16,666 had complete SDOH survey data (MASLD to controls, 1:3.4). Compared to controls, individuals with MASLD were more likely to have less than a high school education (10.7% vs 9.7%; P < .001) and annual income ≤$35,000 (42.4% vs 34.3%; P < .001). Compared to controls, participants with MASLD had significantly higher levels of social isolation, neighborhood disorder, perceived stress, food insecurity, and transportation insecurity (P < .001 for all).ConclusionThe study identified significant associations between MASLD and multiple SDOH. Future studies should investigate how SDOH interact to drive MASLD risk and progression. Introduction Metabolic dysfunction–associated steatotic liver disease (MASLD) is progressive and may lead to liver fibrosis, cirrhosis, and hepatocellular carcinoma. The estimated global prevalence of MASLD is 30%; it is predicted to increase to 55% by 2040.1,2 Latin America has the highest reported prevalence of 44.4%, followed by 36.5% in North Africa and the Middle East, 33.8% in South Asia, 33.1% in Southeast Asia, 31.2% in North America, and 25.1% in Western Europe.3 In the United States, Hispanic and Black patients with nonalcoholic fatty liver disease (NAFLD), now known as MASLD,4 were noted to have 1.44 and 2.13 times higher risk of overall mortality compared to non-Hispanic White patients.5 Known MASLD risk factors, such as obesity, diabetes, hypertension (HTN), dyslipidemia, and metabolic syndrome, disproportionately affect minority populations.6,7 Also, differential expression of genetic polymorphisms (eg, PNPLA3 rs738409 C>G and HSD13B13 rs72613567 TA) may partially explain observed differences in MASLD risk.8, 9, 10 However, the effect of genotype and Hispanic ethnicity on liver fibrosis was no longer significant after adjusting for education, suggesting that socioeconomic factors may influence MASLD progression more than genetics.9 Low socioeconomic status is associated with MASLD risk factors, such as obesity, higher hemoglobin A1c among diabetics, and diabetes-related mortality.11, 12, 13 Black and Hispanic communities are unduly affected by factors impacting diet and lifestyle, including food insecurity, targeted marketing of unhealthy food and alcohol, and limited access to healthcare.2 Due to lack of screening for the 5 domains of social determinants of health (SDOH): economic stability, education access and quality, health care access and quality, neighborhood and built environment, and social and community context, there are sparse data specific to patients with MASLD.6,14 Most studies examining SDOH in the MASLD population focused on economic stability, specifically food insecurity,15 which disproportionately affects individuals with MASLD. In a small sample, n = 210, food insecurity affected 56.8% of patients with MASLD vs 26.1% of controls.16 In larger population-based studies, food insecurity was associated with increased hepatic fat and liver stiffness.2,17,18 There is a known association between education, MASLD prevalence, and fibrosis.9 Patients with lower education levels had higher MASLD prevalence, 40%, compared to 23% in those with higher education, independent of age and sex.19 Individuals with lower educational attainment typically lack access to healthcare and often have poor health literacy, which may influence dietary choices and activity level.19, 20, 21 Likely, education and cultural context interact to inform health behaviors that increase MASLD risk. This reinforces the need for culturally sensitive interventions and policies to eliminate structural barriers to improving health in MASLD.2,8 While data regarding the impact of education and food insecurity on MASLD exist, the role of other SDOH domains remains to be evaluated. To further characterize and understand the interactive nature of the various domains of SDOH in individuals with MASLD, we conducted a comprehensive analysis using the All of Us Research Program (AoU) dataset. Experimental Procedures Data Source The data were obtained via a retrospective cross-sectional analysis of survey results available through the AoU Research Program (www.allofus.nih.gov). A prospective cohort study established by the National Institutes of Health, AoU provides data using self-reported survey results, electronic health records (EHRs), biophysical sensors, and genomic profiling from individuals 18 years or older living in the United States. Participants complete “The Basics” survey prior to other surveys22 and may omit any questions within a survey. EHR data are extracted at enrollment; when participants consent to use of EHR data, AoU obtains retrospective and prospective data: their future clinic visits, treatments, and procedures after AoU enrollment date are obtained for as long as the participant continues to be enrolled. Thus, individuals’ EHR data vary by the length of time that the participant has received care at the affiliated healthcare organization. We used data from version 7-controlled tier, which included 413,457 individuals who enrolled for participation between May 30, 2017, and July 1, 2022. This study was approved by the University of Miami Institutional Review Board (IRB# 20231220). The AoU protocol and respective materials have already been approved by the AoU IRB. We followed rigorous procedures to identify participants with MASLD and healthy participants as a control cohort (Figure 1).Figure 1Flow diagram outlining steps taken to select participants with metabolic-associated steatotic liver disease and age- and race-matched healthy controls in All of Us version 7. AUD, alcohol use disorder. Flow diagram outlining steps taken to select participants with metabolic-associated steatotic liver disease and age- and race-matched healthy controls in All of Us version 7. AUD, alcohol use disorder. Participants With MASLD Participants with a historical EHR condition code mapping to International Classification of Diseases (ICD)-9 codes “571.8” and “571.9” and ICD-10 code “K76.0” as the defining EHR codes for MASLD were considered the primary set of participants with MASLD (Supplementary Table 1). Using ICD-10 codes to identify NAFLD/NASH has been validated, showing a high positive predictive value of 91% with a moderate sensitivity of 57.1%.23,24 The diagnostic overlap between NAFLD and MASLD is estimated to be around 95%–99%; therefore, we would expect similar positive predictive value and sensitivity.25 We subsequently excluded participants with a historical diagnosis of other liver diseases based on ICD codes, including alcohol-related liver disease, hepatitis B, C, or D virus, Wilson’s disease, alpha-1-antitrypsin deficiency, hereditary hemochromatosis, autoimmune hepatitis, primary biliary cholangitis, or primary sclerosing cholangitis. This study group was further refined by excluding heavy drinkers identified via either the patient-disclosed “Lifestyle” survey or EHR condition codes (Supplementary Table 1), leaving a raw set of 15,301 participants. Among those included in the MASLD study set, we retained participants of major self-reported races/ethnicities (Hispanic, non-Hispanic Black, non-Hispanic White, non-Hispanic Asian) and major self-reported sexes (those participants indicating male and female). After applying these additional exclusion criteria, the MASLD cohort consisted of 14,236 individuals. Healthy Control Participants Of the 413,457 participants in AoU, we first excluded 3738 participants who reported liver disease conditions via the “Personal and Family Health History” survey. We excluded all patients with an EHR history of chronic liver disease or other conditions (eg, transplant, hepatocellular carcinoma, cirrhosis, and/or complications of cirrhosis). To minimize the risk of including individuals with undiagnosed MASLD, we excluded obese individuals with one occurrence of ICD-9/10 codes for additional metabolic comorbidities4 (ie, diabetes mellitus [DM], hyperlipidemia [HLD], hypertriglyceridemia, coronary artery disease [CAD], or obstructive sleep apnea [OSA]). We excluded participants with abnormal values calculated for the Fibrosis-4 index26 or the NAFLD fibrosis score,27 those with 2 instances of elevated alanine aminotransferase or aspartate aminotransferase, defined as >20 IU/L in women or >30 IU/L in men, or heavy drinkers as above. After applying the exclusion criteria, there were 236,928 participants in the raw control group (Supplementary Table 2, step 3). We applied the sex and race limitations used for the MASLD study cohort to the controls, retaining 218,308 participants. The raw MASLD patients were substantially older than healthy participants, and Hispanics were unsurprisingly over-represented. To prevent analyses from being driven by inherent age and ethnicity differences, we conducted a frequency matching process sequentially28 (Supplementary Figure 1), with the final age- and sex-matched control cohort including 43,659 participants. Data Processing and Statistical Analysis Demographic, EHR records, and AoU survey responses for the overall cohort were accessed from the AoU Researcher Workbench via the R package bigrquery in Jupyter Notebooks. Basic demographic features included age, sex, and race/ethnicity. Age was determined at the freezing date of the AoU database (July 1, 2022) and was categorized into these age brackets: [18-40], [41-55], [56-70], and [70-120]. Sex was taken as self-identified. A broad race variable was determined by absorbing both self-identified race and self-identified ethnicity: the Hispanic group was foremost formulated for those individuals selecting “Hispanic, Latino, or Spanish,” and 3 additional racial categories (Black, White, and Asian) were based on self-identified race among non-Hispanic participants; we excluded all participants whose self-identified racial or ethnic category fell beyond these 4 technically defined racial groups. Eight comorbidities (CAD, DM, human immunodeficiency virus [HIV], HLD/hypertriglyceridemia, HTN, OSA, and pre-DM) were identified in the overall MASLD and healthy cohorts via 2 channels: diagnosis codes (ICD-9, ICD-10, and Systematized Nomenclature of Medicine codes) in EHR records or self-affirmation in the “Personal and Family Health History” survey. Combining relevant questions from AoU Surveys22 (Basics, Overall Health, Social Factors of Health, Health Care Access, and Utilization) (Supplementary Table 3), we examined several SDOH domains. Social isolation was assessed via the University of California Los Angeles (UCLA) Loneliness Scale29 and neighborhood disorder using the Ross-Mirowsky Perceived Neighborhood Disorder Scale.30 Stress was measured using the Perceived Stress Scale,31 and health literacy was ascertained using the Chew health literacy score.32 Food insecurity was measured using the Hunger Vital Sign.33 Housing insecurity was assessed using the Veteran’s Administration Homeless Clinical Screener Reminder and residential mobility via the number of years in the same residence.34 Emotional health was assessed using one question from the PROMIS Global Health Scale.35 Delays in care related to transportation were ascertained.33 Healthcare-related perceptions of discrimination were assessed using the Discrimination in Medical Settings Scale.36,37 Based on a Venn diagram analysis, 28.8% of the overall sample responded to questions on social isolation, neighborhood disorder, perceived stress, and food insecurity, and responders tended to address all 4 variables rather than questions from one SDOH domain (Figure 2). Hence, we took the congregate subset of 16,666 individuals with complete responses to these 4 variables and designated it the SDOH subset. Consisting of 3776 cases and 12,890 controls, the SDOH subset accounts for 28.8% of the overall cohort of 57,895 participants (Figure 1).Figure 2The Venn diagram illustrates overlap and missingness for social isolation, neighborhood disorder, perceived stress, and food insecurity in All of Us version 7. Of 57,895 individuals in the overall cohort, 16,666 participants had data from these different SDOH domains. However, 40,351 did not have SDOH survey data. Numbers inside the circles represent participants with overlapping responses. The Venn diagram illustrates overlap and missingness for social isolation, neighborhood disorder, perceived stress, and food insecurity in All of Us version 7. Of 57,895 individuals in the overall cohort, 16,666 participants had data from these different SDOH domains. However, 40,351 did not have SDOH survey data. Numbers inside the circles represent participants with overlapping responses. Binary and categorical variables were assessed via a 2-sample proportion chi-squared test. Numerical variables including social isolation scale, neighborhood disorder, and perceived medical discrimination, were assessed via Wilcoxon rank-sum test. When applicable, answers from a set of SDOH questions were summarized and reported as is. Raw values were used directly in univariable analyses. Perceived stress31 and health literacy32 were originally summed integer scores but were converted to a binary and tri-group categorical variable, respectively, based on published category ranges. Missing responses were excluded from summary tabulation. Certain factors, such as insurance type and education, used categories merged from miscellaneous, granular original groups. Finally, we conducted multivariable logistic regression to examine the association between SDOH and MASLD in the SDOH cohort. Since our case and control groups were matched for age and race/ethnicity, these 2 demographic variables were not considered in the regression model. It is known that MASLD is more prevalent in men,7 but the inherent over-representation of women in AoU participants (the ratio of female to male participants in AoU version 7 is 1.5:1) and especially in the MASLD group could erroneously categorize female sex as a risk factor for MASLD. Therefore, we excluded sex from the model. All 18 aforementioned SDOH variables, namely social isolation, neighborhood disorder, perceived stress, health literacy, food insecurity, housing instability, emotional problems, perceived medical discrimination, transportation insecurity, birth in the United States, non-English use, employment, education, housing type, built environment, income, insurance type, and residential mobility, were combined to model the probability of developing MASLD. Social isolation, neighborhood disorder, and perceived medical discrimination had diverse original value scales, but they were scaled to 0-1 in the regression analysis. All other 15 variables were either binary or categorical variables. The multivariable logistic regression model was executed once without further efforts to narrow down predictors. SDOH variables with P < .05 were deemed more credible MASLD association factors than the others, with the companion odds ratio (OR) and 95% confidence interval (CI) indicative of the direction of association: protective (OR < 1) or aggressive (OR > 1). In the summary tables, binary and categorical variables are represented in numbers and percentages, while numeric variables are reported in median and interquartile range. For all analyses, P values were 2-sided, and a P value of .05 or less was considered statistically significant. Any P-values below .001 are shown as P < .001.38 All statistical analyses were performed using the R statistical computing environment in AoU Researcher Workbench empowered by Google computational platform. Results Baseline Characteristics We included data from 57,895 AoU participants, of whom 14,236 participants were defined to have MASLD and 43,659 participants were defined as healthy controls (ie, 1: 3 case-to-control ratio). Overall, males represented 33.2% of participants with MASLD and 39.9% of healthy controls, consistent with the over-representation of females among AoU participants. Cases and controls were matched for self-reported race and ethnicity; 58.2% were White, 27.8% were Hispanic, 11.4% were Black, and 2.6% were Asian. The percentage of patients identifying as Hispanic did not differ significantly between healthy controls and participants with MASLD, P = 1 (Table 1). In the overall cohort, 80.3% of participants with MASLD were born in the United States, compared to 79.4% of healthy controls, P = .024. As expected, comorbidities including CAD, DM, HIV, HLD, HTN, OSA, and pre-DM were more prevalent (P < .001) in participants with MASLD, as compared to controls (Supplementary Figure 2).Table 1Baseline Demographics of Participants in the Overall and Social Determinants of Health (SDOH) CohortsFactor and categoriesOverall cohort (N = 57,895)Participants with SDOH data (N = 16,666)Healthy controls (n = 43,659)MASLD (n = 14,236)P valueHealthy controls (n = 12,890)MASLD (n = 3776)P valueAge (y) ≤406599 (15.1)1957 (13.7)<.0011351 (10.5)335 (8.9).022 41-5511,221 (25.7)3574 (25.1)2668 (20.7)804 (21.3) 56-7016,364 (37.5)5560 (39.1)5281 (41.0)1601 (42.4) >709475 (21.7)3145 (22.1)3590 (27.9)1036 (27.4)Male sex, n (%)17,408 (39.9)4724 (33.2)<.0014971 (38.6)1262 (33.4)<.001Race/ethnicity, n (%) White25,427 (58.2)8291 (58.2)110,266 (79.6)2905 (76.9)1 Hispanic12,117 (27.8)3951 (27.8)1771 (13.7)546 (14.5) Black4959 (11.4)1617 (11.4)564 (4.4)249 (6.6) Asian1156 (2.6)377 (2.6)289 (2.2)76 (2)Hispanic ethnicity, n (%)12,117 (27.8)3951 (27.8)11771 (13.7)546 (14.5).272 Table 1Baseline Demographics of Participants in the Overall and Social Determinants of Health (SDOH) CohortsFactor and categoriesOverall cohort (N = 57,895)Participants with SDOH data (N = 16,666)Healthy controls (n = 43,659)MASLD (n = 14,236)P valueHealthy controls (n = 12,890)MASLD (n = 3776)P valueAge (y) ≤406599 (15.1)1957 (13.7)<.0011351 (10.5)335 (8.9).022 41-5511,221 (25.7)3574 (25.1)2668 (20.7)804 (21.3) 56-7016,364 (37.5)5560 (39.1)5281 (41.0)1601 (42.4) >709475 (21.7)3145 (22.1)3590 (27.9)1036 (27.4)Male sex, n (%)17,408 (39.9)4724 (33.2)<.0014971 (38.6)1262 (33.4)<.001Race/ethnicity, n (%) White25,427 (58.2)8291 (58.2)110,266 (79.6)2905 (76.9)1 Hispanic12,117 (27.8)3951 (27.8)1771 (13.7)546 (14.5) Black4959 (11.4)1617 (11.4)564 (4.4)249 (6.6) Asian1156 (2.6)377 (2.6)289 (2.2)76 (2)Hispanic ethnicity, n (%)12,117 (27.8)3951 (27.8)11771 (13.7)546 (14.5).272 Baseline Demographics of Participants in the Overall and Social Determinants of Health (SDOH) Cohorts Of the 57,895 participants included, 16,666 had complete SDOH survey data. The SDOH cohort included 3776 participants with MASLD and 12,890 healthy controls, that is, 1:3.4 case-to-control ratio. The age group from 56–70 years was the most represented group in both the overall cohort (37.5% of healthy controls; 39.1% of participants with MASLD) and SDOH cohort (41.0% of healthy controls; 42.4% of participants with MASLD) (Table 1). Males represented 33.4% of participants with MASLD and 38.6% of healthy controls in the SDOH cohort. In contrast to the full cohort, a much higher proportion of participants with complete SDOH data were White: 76.9% of participants with MASLD and 79.6% of healthy controls. There was no significant difference in the percentage of patients identifying as Hispanic between cases and controls in the SDOH cohort, P = .272, or the proportion of participants born in the United States, P = .954. Education Access and Economic Stability in the Overall and SDOH Cohorts Participants with MASLD were more likely to have Medicaid or Medicare compared to healthy controls in both the overall, P < .001, and SDOH cohorts, P < .001 (Table 2). In both cohorts, healthy controls were more likely to report having no insurance. Educational attainment was variable. In the overall cohort, 10.7% of participants with MASLD reported less than high school education compared to 9.7% of healthy controls. Similarly, 73.8% of healthy controls achieved higher than a high school education compared to 69.6% of participants with MASLD, P < .001. The SDOH cohort achieved greater educational attainment than the overall cohort; 89.9% of healthy controls had more than a high school education compared to 82.9% of participants with MASLD, P < .001. In the SDOH cohort, participants with MASLD were more likely to indicate that they were unemployed (56.4%), compared to the healthy controls (48.3%), P < .001 (Table 3, Supplementary Figure 3).Table 2Comparative Social Determinants of Health (SDOH) Data Available in Both the Overall and SDOH CohortsFactor and categories n (%)Overall cohort (N = 57,895)Participants with SDOH data (N = 16,666)Healthy controls (n = 43,659)MASLD (n = 14,236)P valueHealthy controls (n = 12,890)MASLD (n = 3776)P valueInsurance None3197 (9.1)423 (3.7)<.001364 (3.4)61 (1.9)<.001 Medicaid6384 (18.2)2664 (23.1)813 (7.6)411 (13) Medicare6592 (18.7)2447 (21.2)2562 (24)769 (24.4) Multiple4803 (13.7)2088 (18.1)1863 (17.5)685 (21.7) Employment13,385 (38.1)3735 (32.4)4773 (44.7)1159 (36.8) VA/military808 (2.3)180 (1.6)292 (2.7)68 (2.2)Education HS31,572 (73.8)9746 (69.6)11,474 (89.9)3102 (82.9)Living arrangement Own23,765 (56.3)6992 (50.9)<.0019423 (74.1)2472 (66.8)<.001 Rent14,833 (35.2)5688 (41.4)2693 (21.2)1002 (27.1) Other3579 (8.5)1065 (7.7)594 (4.7)229 (6.2)Years living in the same residence ≤1 y6477 (15.0)1583 (11.3)<.0011297 (10.1)295 (7.9).003 >1 y36,717 (85.0)12,454 (88.7)11,527 (89.9)3457 (92.1)Annual income ≤$35,00012,117 (34.3)4723 (42.4)<.0012119 (18.2)933 (28.1)<.001 $35,001-$75,0008156 (23.1)2805 (25.2)2880 (24.8)966 (29.1) >$75,00015,053 (42.6)3605 (32.4)6630 (57)1423 (42.8)HS, high-school; VA, Veteran’s Administration.Table 3Five Domains of Social Determinants of Health (SDOH) Data Available in the SDOH CohortFactor and categoriesHealthy controls (n = 12,890)MASLD (n = 3776)P valueLimited health literacy, n (%)247 (1.9%)120 (3.2%)<.001Perceived stress, n (%) Low7617 (59.1%)1925 (51%)<.001 Moderate4688 (36.4%)1591 (42.1%) High585 (4.5%)260 (6.9%)Median social isolation scale14 (11-19)15 (11-20)<.001Median neighborhood disorder19 (15-25)20 (16-26)<.001Median medical discrimination scale9 (7-13)10 (7-14)<.001Food insecurity, n (%)1185 (9.2%)665 (17.6%)<.001Housing insecurity, n (%)1095 (8.5%)426 (11.3%)<.001Built environment, n (%) At least one issue2385 (18.5%)888 (23.5%)<.001 Bug infestation765 (32.1%)313 (35.2%).016 Inadequate heat265 (11.1%)102 (11.5%) Lead paint or pipes187 (7.8%)52 (5.9%) Mold545 (22.9%)226 (25.5%) Inadequate smoke detector179 (7.5%)56 (6.3%) Oven or stove not working69 (2.9%)32 (3.6%) Water leaks375 (15.7%)107 (12%)Emotional problems, n (%) Never3269 (26%)795 (21.9%)<.001 Rarely4551 (36.3%)1202 (33.1%) Sometimes3524 (28.1%)1190 (32.8%) Often1205 (9.6%)442 (12.2%)Non-English speaker, n (%)1968 (15.3%)635 (16.9%).023Transportation insecurity, n (%)561 (4.9%)266 (8%)<.001Unemployed, n (%)6171 (48.3%)2100 (56.4%)<.001Not born in the United States, n (%)1551 (12.1%)452 (12%).954 Table 2Comparative Social Determinants of Health (SDOH) Data Available in Both the Overall and SDOH CohortsFactor and categories n (%)Overall cohort (N = 57,895)Participants with SDOH data (N = 16,666)Healthy controls (n = 43,659)MASLD (n = 14,236)P valueHealthy controls (n = 12,890)MASLD (n = 3776)P valueInsurance None3197 (9.1)423 (3.7)<.001364 (3.4)61 (1.9)<.001 Medicaid6384 (18.2)2664 (23.1)813 (7.6)411 (13) Medicare6592 (18.7)2447 (21.2)2562 (24)769 (24.4) Multiple4803 (13.7)2088 (18.1)1863 (17.5)685 (21.7) Employment13,385 (38.1)3735 (32.4)4773 (44.7)1159 (36.8) VA/military808 (2.3)180 (1.6)292 (2.7)68 (2.2)Education HS31,572 (73.8)9746 (69.6)11,474 (89.9)3102 (82.9)Living arrangement Own23,765 (56.3)6992 (50.9)<.0019423 (74.1)2472 (66.8)<.001 Rent14,833 (35.2)5688 (41.4)2693 (21.2)1002 (27.1) Other3579 (8.5)1065 (7.7)594 (4.7)229 (6.2)Years living in the same residence ≤1 y6477 (15.0)1583 (11.3)<.0011297 (10.1)295 (7.9).003 >1 y36,717 (85.0)12,454 (88.7)11,527 (89.9)3457 (92.1)Annual income ≤$35,00012,117 (34.3)4723 (42.4)<.0012119 (18.2)933 (28.1)<.001 $35,001-$75,0008156 (23.1)2805 (25.2)2880 (24.8)966 (29.1) >$75,00015,053 (42.6)3605 (32.4)6630 (57)1423 (42.8)HS, high-school; VA, Veteran’s Administration. Comparative Social Determinants of Health (SDOH) Data Available in Both the Overall and SDOH Cohorts HS, high-school; VA, Veteran’s Administration. Table 3Five Domains of Social Determinants of Health (SDOH) Data Available in the SDOH CohortFactor and categoriesHealthy controls (n = 12,890)MASLD (n = 3776)P valueLimited health literacy, n (%)247 (1.9%)120 (3.2%)<.001Perceived stress, n (%) Low7617 (59.1%)1925 (51%)<.001 Moderate4688 (36.4%)1591 (42.1%) High585 (4.5%)260 (6.9%)Median social isolation scale14 (11-19)15 (11-20)<.001Median neighborhood disorder19 (15-25)20 (16-26)<.001Median medical discrimination scale9 (7-13)10 (7-14)<.001Food insecurity, n (%)1185 (9.2%)665 (17.6%)<.001Housing insecurity, n (%)1095 (8.5%)426 (11.3%)<.001Built environment, n (%) At least one issue2385 (18.5%)888 (23.5%)<.001 Bug infestation765 (32.1%)313 (35.2%).016 Inadequate heat265 (11.1%)102 (11.5%) Lead paint or pipes187 (7.8%)52 (5.9%) Mold545 (22.9%)226 (25.5%) Inadequate smoke detector179 (7.5%)56 (6.3%) Oven or stove not working69 (2.9%)32 (3.6%) Water leaks375 (15.7%)107 (12%)Emotional problems, n (%) Never3269 (26%)795 (21.9%)<.001 Rarely4551 (36.3%)1202 (33.1%) Sometimes3524 (28.1%)1190 (32.8%) Often1205 (9.6%)442 (12.2%)Non-English speaker, n (%)1968 (15.3%)635 (16.9%).023Transportation insecurity, n (%)561 (4.9%)266 (8%)<.001Unemployed, n (%)6171 (48.3%)2100 (56.4%)<.001Not born in the United States, n (%)1551 (12.1%)452 (12%).954 Five Domains of Social Determinants of Health (SDOH) Data Available in the SDOH Cohort Home ownership was more common among healthy controls, 56.3%, than participants with MASLD, 50.9%, in the overall cohort, P < .001, with similar findings in the SDOH cohort. Participants with MASLD had less residential mobility than healthy controls; 11.3% of participants with MASLD reported living in the same residence for under 1 year vs 15% of healthy controls in the overall cohort, P < .001, compared to 7.9% of participants with MASLD and 10.1% of healthy controls in the SDOH cohort, P = .003 (Table 2). There were statistically significant differences in reported income for the overall cohort; 42.4% of participants with MASLD reported an annual income ≤$35,000 compared to 34.3% of healthy controls, P < .001. This difference persisted in the SDOH cohort; 28.1% of MASLD participants reported an annual income ≤$35,000 compared to 18.2% of healthy controls, P < .001 (Table 2). Differences in Access to Education, Health Care, Neighborhood Disorder, the Built Environment, and Social and Community Context in the SDOH Cohort Of the 16,666 participants with full SDOH data, participants with MASLD were more likely to have limited health literacy, 3.2%, compared to 1.9% of healthy controls, P < .001. Perceived stress was significantly higher in MASLD participants, with 42.1% and 6.9% reporting moderate or high levels of stress, respectively, compared to 36.4% and 4.5% among healthy controls, P < .001. Participants with MASLD had higher median social isolation scores compared to healthy controls (15 vs 14 on the UCLA loneliness scale; P < .001). Participants with MASLD, as compared to healthy controls, also reported higher levels of neighborhood disorder as determined via the Ross-Mirowsky perceived neighborhood disorder scale (20 [MASLD] vs 19 [control]; P < .001). Participants with MASLD also reported higher discrimination, with a median medical discrimination scale of 10, compared to 9 in healthy controls, P < .001 (Table 3). Participants with MASLD had significantly higher rates of food insecurity as determined via the Hunger Vital Sign (17.6% [MASLD] vs 9.2% [controls]; P < .001). As assessed by the Veteran’s Administration Homeless Clinical Screener Reminder, 11.3% of participants with MASLD were concerned about having a place to live over the past 6 months vs 8.5% of healthy controls, P < .001. Transportation insecurity was also more common in participants with MASLD; 8% reported that they delayed getting health care in the prior 12 months due to lack of transportation compared to 4.9% of healthy controls, P < .001. Regarding the built environment, 23.5% of MASLD participants reported at least one problem with housing, compared to 18.5% of healthy controls, P < .001. Participants with MASLD were more likely than healthy controls to report bug infestation or mold, whereas healthy controls were more likely to report water leaks or lead paint/pipes, P = .016. When asked if they had been bothered by emotional problems in the prior 7 days, 32.8% of participants with MASLD answered “sometimes” and 12.2% reported “often,” compared to 28.1% and 9.6% of healthy control participants, respectively, P < .001. A higher proportion of participants with MASLD reported speaking a non-English language, 16.9%, compared to 15.3% of healthy controls, P = .023 (Table 3). Using a multivariable logistic regression model that included all SDOH variables, we found that food insecurity was associated with increased odds of MASLD, OR 1.53 (95% CI: 1.3-1.81, P < .001). Lower income was also associated with higher odds of MASLD. Compared to those reporting over $75,000 annual income, those reporting less than $35,000 and those reporting $35,000-$75,000 had increased odds of MASLD, OR 1.4 (95% CI: 1.19-1.65, P < .001) and 1.38 (95% CI: 1.23-1.56, P < .001), respectively. Compared to individuals with employer-based insurance, individuals with multiple insurances had increased odds of MASLD, OR 1.18 (95% CI: 1.02-1.37), while those who were uninsured had decreased odds of MASLD, OR 0.49 (95% CI: 0.34-0.69). Perceived medical discrimination was associated with increased odds of MASLD, OR 1.52 (95% CI: 1.09-2.11, P = .013). Unemployment was also associated with increased odds of MASLD, OR 1.2 (95% CI: 1.07-1.36, P = .002). Compared to low stress, moderate stress was associated with increased odds of MASLD, OR 1.13 (95% CI: 1.01-1.28, P = .038), but there was no significant difference between high and low stress. Similarly, having emotional problems sometimes was associated with higher odds of MASLD vs never having emotional problems (referent), OR 1.19 (95% CI: 1.03-1.38, P = .02). However, there was no significant difference between those reporting emotional problems often vs those who reported never having emotional problems, P = .259. Finally, those who lived in the same place for under 1 year had decreased odds of MASLD compared to those with less residential mobility, OR 0.71 (95% CI: 0.59-0.84, P < .001) (Table 4).Table 4Multivariable Logistic Regression to Examine Association of Social Determinants of Health (SDOH) Factors With Metabolic-Associated Steatotic Liver Disease (MASLD)VariableaP valueOdds ratio [95% confidence Interval]Transportation inaccessibility.9851 [0.81, 1.24]Health literacy: limited.8040.95 [0.63, 1.41]Birth: United States.2231.12 [0.93, 1.36]Stress: high vs low.2161.18 [0.91, 1.53]Stress: med vs low.0381.13 [1.01, 1.28]Education: =HS vs HS vs 75k< .0011.4 [1.19, 1.65]Income: ≤ 75k vs >75k< .0011.38 [1.23, 1.56]Insurance: Medicaid vs employment.4461.09 [0.88, 1.34]Insurance: Medicare vs employment.8281.02 [0.88, 1.18]Insurance: multi-insured vs Employment.0291.18 [1.02, 1.37]Insurance: uninsured vs employment< .0010.49 [0.34, 0.69]Insurance: VA/military vs employment.1290.77 [0.55, 1.07]Use of non-English language.4841.06 [0.9, 1.24]Social isolation.2020.84 [0.64, 1.1]Medical discrimination score.0131.52 [1.09, 2.11]Residential mobility within 1 y< .0010.71 [0.59, 0.84]Mood: rare vs never.1031.12 [0.98, 1.28]Mood: sometimes vs never.021.19 [1.03, 1.38]Mood: often vs never.2591.12 [0.92, 1.37]Neighborhood disorder.0781.34 [0.97, 1.84]Bolded values meet the threshold of statistical significance.HS, high school; VA, Veteran’s Administration.aAltogether 18 variables were included in the regression model: MASLD + transportation inaccessibility + health literacy + birth in United States + perceived stress + education + unemployment + food insecurity + housing type + housing issue + income + insurance type + second-language + social isolation + perceived medical discrimination + residential mobility + mood wellness + neighborhood insecurity. Social isolation, neighborhood insecurity, and perceived medical discrimination had diverse original value scales, but they were scaled to 0-1 in the regression analysis. Table 4Multivariable Logistic Regression to Examine Association of Social Determinants of Health (SDOH) Factors With Metabolic-Associated Steatotic Liver Disease (MASLD)VariableaP valueOdds ratio [95% confidence Interval]Transportation inaccessibility.9851 [0.81, 1.24]Health literacy: limited.8040.95 [0.63, 1.41]Birth: United States.2231.12 [0.93, 1.36]Stress: high vs low.2161.18 [0.91, 1.53]Stress: med vs low.0381.13 [1.01, 1.28]Education: =HS vs HS vs 75k< .0011.4 [1.19, 1.65]Income: ≤ 75k vs >75k< .0011.38 [1.23, 1.56]Insurance: Medicaid vs employment.4461.09 [0.88, 1.34]Insurance: Medicare vs employment.8281.02 [0.88, 1.18]Insurance: multi-insured vs Employment.0291.18 [1.02, 1.37]Insurance: uninsured vs employment< .0010.49 [0.34, 0.69]Insurance: VA/military vs employment.1290.77 [0.55, 1.07]Use of non-English language.4841.06 [0.9, 1.24]Social isolation.2020.84 [0.64, 1.1]Medical discrimination score.0131.52 [1.09, 2.11]Residential mobility within 1 y< .0010.71 [0.59, 0.84]Mood: rare vs never.1031.12 [0.98, 1.28]Mood: sometimes vs never.021.19 [1.03, 1.38]Mood: often vs never.2591.12 [0.92, 1.37]Neighborhood disorder.0781.34 [0.97, 1.84]Bolded values meet the threshold of statistical significance.HS, high school; VA, Veteran’s Administration.aAltogether 18 variables were included in the regression model: MASLD + transportation inaccessibility + health literacy + birth in United States + perceived stress + education + unemployment + food insecurity + housing type + housing issue + income + insurance type + second-language + social isolation + perceived medical discrimination + residential mobility + mood wellness + neighborhood insecurity. Social isolation, neighborhood insecurity, and perceived medical discrimination had diverse original value scales, but they were scaled to 0-1 in the regression analysis. Multivariable Logistic Regression to Examine Association of Social Determinants of Health (SDOH) Factors With Metabolic-Associated Steatotic Liver Disease (MASLD) Bolded values meet the threshold of statistical significance. HS, high school; VA, Veteran’s Administration. Altogether 18 variables were included in the regression model: MASLD + transportation inaccessibility + health literacy + birth in United States + perceived stress + education + unemployment + food insecurity + housing type + housing issue + income + insurance type + second-language + social isolation + perceived medical discrimination + residential mobility + mood wellness + neighborhood insecurity. Social isolation, neighborhood insecurity, and perceived medical discrimination had diverse original value scales, but they were scaled to 0-1 in the regression analysis. Discussion Through the available AoU data, the current study further characterizes and strengthens the associations between SDOH and MASLD. Consistent with prior studies, the current study found that participants with MASLD had higher rates of comorbidities, including CAD, DM, HIV, HLD, HTN, OSA, and pre-DM.39 Previous studies have linked SDOH to increased exposure to insulting agents, such as unhealthy foods and tobacco or alcohol use, heightened susceptibility of developing chronic comorbidities, disease course and outcomes.40 Our findings may inform future interventions that target specific SDOH and thereby combat diabetes, obesity, and other MASLD-related comorbidities to support the prevention and treatment of MASLD. In participants who completed the SDOH survey, we assessed variables which were representative of all 5 SDOH domains: economic stability, education access and quality, healthcare access and quality, neighborhood and built environment, and social and community context. Participants with MASLD were more likely to report being unemployed, were less likely to own their home, and had less residential mobility compared to healthy controls, which reinforces that lower socioeconomic status portends increased MASLD risk.41 Participants with MASLD had higher levels of neighborhood disorder, housing insecurity, transportation insecurity, and food insecurity compared to controls. Barriers to lifestyle modifications that prevent MASLD are linked to low socioeconomic status that may limit access to healthy foods, gym memberships, and safe neighborhoods with green spaces that promote physical activity.21 Food insecurity, a known SDOH that impacts MASLD risk, is more common in socioeconomically disadvantaged groups and increases risk of disease progression. In a prior study, food insecurity was associated with 2.82 times the odds of advanced liver fibrosis in adults with HIV and MASLD.42 Diets of food-insecure individuals replace high-quality nutrients with refined foods, added sugars, and saturated fats.43 Hispanic households experience almost twice the rate of food insecurity compared to non-Hispanic White households.17 The disproportionate burden of food insecurity in Hispanic populations may contribute to the higher rates of MASLD observed in Hispanic patients.44 However, food insecurity did not significantly impact diet quality in MASLD vs non-MASLD groups in 1 study,45 highlighting the need to determine how other factors mediate the association between food insecurity and increased risk of MASLD. Previous research revealed that lack of companionship was associated with having DM, a risk factor for MASLD.46 Prior to our study, data regarding the relationship between social connections and MASLD were limited. Within the current study, patients with MASLD felt lonelier, had more perceived stress, and were more likely to report emotional problems than healthy controls. In a prior study, patients with depression had 1.9 times higher chances of having MASLD compared to controls.16 Social isolation results from various social and economic factors including lack of social support, financial, or other stressors. Mood disorders or negative emotions may be associated with overeating or “comfort eating.”47 All together, these data help to contextualize the increased risk of obesity and MASLD observed in patients who report social isolation, stress, and negative mood. Furthermore, participants with MASLD may experience higher levels of discrimination in medical settings. Future research is needed to identify granular factors within the social and community context SDOH domains that increase MASLD risk. Our results are consistent with studies demonstrating an association between lower education level and increased MASLD prevalence.48 Higher educational attainment develops health literacy, coping strategies, and problem-solving skills, which can influence behaviors.49 Compared to persons with adequate health literacy, limited health literacy significantly increases the likelihood of difficulty estimating calories and fat content of food and underestimating portion size.21 Even in this sample with higher educational attainment than the general US population, more participants with MASLD had limited health literacy. Improving quality and access to education in vulnerable populations could serve as a protective measure against the development or progression of MASLD. Clinic- or population-based interventions to address disparities in health literacy could theoretically decrease the MASLD prevalence and/or prevent disease progression. Participants in the MASLD cohort were more likely to have Medicaid or Medicare, which is consistent with the finding that participants with MASLD are more likely to live in areas where more people had public health insurance, as opposed to private health insurance, and had increased risk of metabolic-dysfunction associated steatohepatitis.50 Although uninsured patients generally experience worse health outcomes, it is unclear how to interpret our findings that healthy controls were more likely to report having no health insurance. Insurance status should be further evaluated to better assess healthcare access and quality for patients with MASLD. Related to the SDOH domains of healthcare access and quality and neighborhood and built environment, participants with MASLD were more likely to report delaying their care in the prior 12 months due to lack of transportation, higher levels of neighborhood disorder, and more problems with housing. A prior study assessing socioeconomic factors within these domains found that lack of transportation and living in renter-occupied and crowded housing units predicted disease severity.50 Furthermore, neighborhood-level SDOH are associated with increased overall mortality in MASLD.41 Importantly, children with higher socioeconomic deprivation were reportedly diagnosed with MASLD at younger ages.51 Population-based interventions tailored to specific age groups and neighborhoods could hamper the anticipated exponential rise in MASLD prevalence. Our findings provide strong evidence of the association between all SDOH domains and MASLD. Based on multivariable modeling, food insecurity, unemployment, income, and medical discrimination are prime areas to target initial interventions. The current study had several strengths. Using the AoU data, this study has a large total sample size of 57,895, one of the largest sample sizes described for an examination of social determinants. Additionally, this study uncovers novel SDOH associations beyond food insecurity and education level in a large sample of 16,666 participants. While other studies examined 1 or 2 domains of SDOH in patients with MASLD,6 most studies involve small sample sizes and did not examine the breadth of social determinants examined herein. In this comprehensive analysis of the association between SDOH and MASLD, we were able to assess all 5 SDOH domains and found significant associations between employment status, residential mobility, home ownership, food insecurity, housing insecurity, transportation insecurity, education level, health literacy, insurance status, neighborhood disorder, lack of transportation, social isolation, perceived stress, discrimination, and emotional burden. This study has limitations including limited racial and ethnic diversity in those who completed the SDOH questionnaire. Additionally, AoU participants are highly educated with adequate health literacy. According to the US census data available at census.gov, 62% of the US population attains more than a high school education compared to 69% in our AoU MASLD cohort and 73% in our healthy control cohort. As there is an association between education level and SDOH, our findings might have been further amplified in a sample with lower educational attainment and more socioeconomic diversity.40 We used a thoughtful, stepwise approach to categorizing patients with MASLD by using highly specific ICD codes23 and excluding participants with other liver diseases or high-risk alcohol use. As with similar database studies, there is a risk of misclassification, for example, we could have inadvertently included healthy controls with MASLD who did not have ICD codes.23 We minimized this risk to the extent possible in the AoU database by excluding individuals with obesity and metabolic comorbidities. Other strategies to ensure that the control group does not include individuals with occult MASLD, for example, natural language processing to review imaging reports and unstructured notes,52 are not possible in the current AoU release. Further, these documents are not transmitted universally by all participating centers. Conclusion To the best of our knowledge, this is the first utilization of the AoU data to assess the associations between SDOH and MASLD. All SDOH parameters assessed in the current analysis were significantly associated with MASLD, suggesting that SDOH factors must be considered when managing patient healthcare. Future prospective studies aimed at understanding how various SDOH factors interact to drive development and progression of MASLD are needed to improve disease measures and outcomes for vulnerable populations.