Dissecting disease, race, ethnicity, and socioeconomic factors for hepatocellular carcinoma: An analysis from the United States Safety Net Collaborative

Abstract

Background:

Racial/ethnic and socioeconomic disparities are assumed to negatively affect treatment and outcomes for hepatocellular carcinoma (HCC). Our aim was to investigate the interaction of racial/ethnic and socioeconomic factors with stage of disease and type of treatment facility in receipt of treatment and overall survival (OS) of patients with HCC.

Methods:

All patients with primary HCC in the US Safety-Net Collaborative database (2012–2014) were included. Patients were categorized into “safety-net” or “tertiary referral center” based on where they received treatment. Socioeconomic factors were determined at the zip-code level and included median income and percent of adults who graduated from high-school. Primary outcomes were receipt of treatment and OS.

Results:

On MV Cox regression, neither race/ethnicity, median income, nor care provided at a SNH were associated with decreased OS (all p > 0.05). Independent predictors of decreased OS included lack of insurance (HR 1.34), less educational attainment (HR 1.59) higher MELD score (HR 1.07), higher stage at diagnosis (II:HR 1.34, III:HR 2.87, IV:HR 3.23), and not receiving treatment (HR 3.94) (all p < 0.05). Factors associated with not receiving treatment included history of alcohol abuse (OR 0.682), increasing MELD (OR 0.874), higher stage at diagnosis (III: OR 0.234, IV: OR 0.210) and care at a safety net facility (OR 0.424) There were no racial/ethnic or socioeconomic disparities in receipt of treatment.

Conclusions:

There is no intrinsic or direct association of race/ethnicity, socioeconomic status, or being treated at select safety-net hospitals with worse outcomes. Poor liver function, no insurance, and advanced stage of presentation are the main determinants of not receiving treatment and decreased survival.

1. Introduction

Survival in patients with hepatocellular carcinoma (HCC) is complex and multifactorial. Incidence of HCC in the United States has increased rapidly over the past two decades, likely due to a combination of improved diagnostic techniques, implementation of screening protocols for patients with cirrhosis, and increased prevalence of risk factors for cirrhosis and carcinogenesis, including chronic Hepatitis C infection and non-alcoholic fatty liver disease [1-3]. Though the incidence of HCC has increased, treatment options have improved, with transplantation, surgical resection, and liver directed therapies offering a chance for cure, while novel systemic and immunotherapies are showing promise for patients with more advanced disease [3,4].

Despite this progress, there is some evidence that these advancements and resultant improvements in mortality have not been equally distributed, and that the epidemiology of the disease itself is changing. For example, Ha and colleagues investigated the incidence of HCC in the United States from 2003 to 2011 using the Surveillance, Epidemiology, and End Results (SEER) cancer registry and found the change in incidence to vary amongst different racial/ethnic groups. Hispanic patients had the greatest increase in HCC incidence (+35.8%), followed by non-Hispanic Black and non-Hispanic White patients, while Asian patients experienced a 5.5% decrease in incidence over that time period [5]. Furthermore, the authors found that non-Hispanic Black patients presented with more advanced disease at diagnosis, and when stratified by stage, both non-Hispanic Black and Hispanic patients were less likely to receive curative treatment compared to non-Hispanic White patients [6].

Socioeconomic status has also been implicated in disparities in HCC outcomes. Hoehn and colleagues found improved overall survival in patients with private insurance compared to Medicare, Medicaid, and uninsured patients using the National Cancer Database. Insurance status is often used as a surrogate for income, and patients with non-private insurance are thought to be socioeconomically disadvantaged compared to patients with private insurance. The observed survival disparity persisted in all HCC patients who received surgery and in patients with early stage disease who underwent resection or transplantation [7].

When attempting to explain some of these disparities, the safety net hospital setting has often been implicated. Safety net hospitals treat a disproportionate number of minority and socioeconomically disadvantaged patients and often operate with a paucity of resources and high patient burden. Mouch and colleagues conducted a systematic review investigating the quality of surgical care in safety net hospitals and found several studies citing worse surgical care in multiple Institute of Medicine quality domains at safety net compared to non-safety net facilities. The authors concluded that “targeting [safety net hospitals] for quality improvement could provide a direct method of addressing health disparities among the poorest, most underserved, and most vulnerable populations in the United States health care system” [8]. Patients undergoing treatment for HCC at safety net hospitals have also been shown to receive surgery or curative treatment less often and to have higher procedure specific mortality rates compared to other hospitals [7]. Additionally, Mokdad and colleagues found that patients treated for HCC at hospitals with a high safety net burden in the Texas Cancer Registry have been shown to have decreased overall survival, even when accounting for stage at presentation, race, and poverty index [9].

With the formation of the United States Safety Net Collaborative, the unique opportunity presents itself to further dissect disparities in HCC treatment and outcomes in the United States. Our aim was to investigate the interaction of racial/ethnic and socioeconomic factors with stage of disease and type of treatment facility in overall survival (OS) and receipt of treatment in patients with HCC.

2. Materials and methods

The United States Safety Net Collaborative (USSNC) is a collaboration of five major safety net hospitals and four sister tertiary referral center counterparts: Grady Memorial Hospital and Emory University, Parkland Memorial Hospital and University of Texas Southwestern, Jackson Memorial Hospital and University of Miami, Bellevue Hospital Center and New York University, and Ben Taub Hospital. Institutional Review Board (IRB) approval was obtained at each individual institution prior to data collection. Patients diagnosed with HCC between 2012 and 2014 were evaluated. Pertinent baseline, screening, treatment, intraoperative (if applicable), pathologic, and postoperative (if applicable) data were collected. Staging was based on the American Committee on Cancer (AJCC) 8th edition guidelines. Data regarding neoadjuvant and adjuvant therapy, disease recurrence, and survival were also recorded.

Patients with primary HCC were included. Patients were categorized into “safety net” and “tertiary referral center” groups based on where they received their care. Socioeconomic variables of education and income were determined using the 2008–2012 American Community Survey data by patient zip code. Primary outcomes were receipt of treatment and overall survival (OS), defined as time in months from diagnosis to death or last follow-up.

2.1. Statistical analysis

Statistical analysis was conducted using SPSS 26.0 software (IBM Inc., Armonk, NY) Descriptive analyses were performed for the entire cohort. Chi-squared analysis was used to compare categorical variables, and Student’s t-test or one-way ANOVA was used for continuous variables, where indicated. The univariate and multivariable associations between each covariate including study cohorts and receipt of treatment and overall survival were assessed using binary logistic or Cox logistic regression, where appropriate. Statistical significance was predefined as p < 0.05.

3. Results

3.1. Patient characteristics

After excluding patients with recurrent tumors, 1832 patients were included for analysis. Average age was 61 years and 77% (n = 1405) were male. Median follow up was 43 months. Fifty-seven percent (n = 1039) were diagnosed and treated at a tertiary referral and 43% (n = 793) at a safety-net facility. Overall, 75% (n = 1382) of patients received treatment for their HCC. Patients who received care at a safety-net facility were younger (59 vs 63 years, p < 0.001), less likely to be White (40% vs 65%, p < 0.001), more likely to be Hispanic (33% vs 15%, p < 0.001), more likely to be uninsured (27% vs 4%, p < 0.001), have lower median income ($43,001 vs $52,598, p < 0.001), and more likely to live in areas where fewer adults graduated from high school (84% vs 76%, p < 0.001). Safety-net patients were more likely to have a history of alcohol abuse (51% vs 40%, p < 0.001) and to present with stage IV disease (25% vs 14%, p < 0.001). Safety-net patients were less likely to receive HCC screening within one year of diagnosis (17% vs 27%, p < 0.001) and to receive HCC treatment (65% vs 83%, p < 0.001) (Table 1). Patients receiving care at a safety net hospital also had decreased 5-year OS compared to those cared for at tertiary referral centers (24% vs 37%, p < 0.001) (Fig. 1).

Table 1.

Baseline characteristics and comparative data by facility type.

	All patients n = 1832	Tertiary Referral n = 1039	Safety Net n = 793	p-value
	n (%)	n (%)	n (%)
Age mean ± SD	60.8 ± 9.6	62.5 ± 9.7	58.5 ± 9.0	<0.001
Gender
Male	1405 (76.7)	784 (75.5)	621 (78.3)	0.169
Female	427 (23.3)	255 (24.5)	172 (21.7)
Race
White	856 (55.3)	625 (64.7)	231 (39.7)
Black	522 (33.7)	240 (24.8)	282 (48.5)
Asian	144 (9.3)	81 (8.4)	63 (10.8)	<0.001
American Indian/Alaskan Native	2 (0.1)	1 (0.1)	1 (0.2)
Other	10 (0.6)	5 (0.5)	5 (0.9)
Unknown	14 (0.9)	14 (1.40)	0 (0)
Ethnicity
Not Hispanic	1384 (77.3)	864 (84.7)	520 (67.5)	<0.001
Hispanic	405 (22.6)	155 (15.3)	250 (32.5)
Health Insurance
Private	476 (28.2)	384 (38.2)	92 (13.5)
Government	960 (56.8)	581 (57.8)	379 (55.5)	<0.001
Hospital Card	32 (1.9)	5 (0.5)	27 (4.0)
Uninsured	221 (13.1)	36 (3.6)	185 (27.1)
Median Income	$47,862	$52,598	$43,001	<0.001
Median (IQR)	($38,116-$63,682)	($41,830-$67,483)	($35,282-$55,616)
Education* mean ± SD	80.6 ± 11.7	84.2 ± 9.5	75.9 ± 12.8	<0.001
Alcohol Abuse	817 (44.6)	413 (39.8)	404 (50.9)	<0.001
MELD	10 (8–15)	10 (8–15)	10 (8–16)	0.439
Median (IQR)
Screening**	376 (22.3)	247 (26.6)	129 (17.0)	<0.001
Stage at Diagnosis
I	769 (42.0)	488 (48.5)	281 (38.0)
II	349 (20.0)	208 (20.7)	141 (19.1)	<0.001
III	306 (17.5)	169 (16.8)	137 (18.5)
IV	323 (18.5)	142 (14.1)	181 (24.5)
Treatment (yes)	1382 (75.4)	864 (83.2)	518 (65.3)	<0.001
Treatment Type
Transplant	250 (13.6)	186 (17.9)	64 (8.1)
Surgery	149 (8.1)	84 (8.1)	65 (8.2)	<0.001
Liver Directed	819 (44.7)	538 (51.8)	281 (35.4)
Therapy Chemotherapy	164 (9.0)	56 (5.4)	108 (13.6)
None	450 (24.6)	175 (16.8)	275 (34.7)

^*% of adults in patient’s zip code who graduated from high school.

^**Within one year of diagnosis.

Fig. 1.

Overall survival by facility type.

3.2. Survival analysis

On univariate analysis, Black race, Hispanic ethnicity, non-private insurance, below median income, living in areas with higher levels of poverty and lower levels of educational attainment, alcohol abuse, higher MELD at diagnosis, increasing stage at diagnosis, not receiving screening within one year of diagnosis, treatment other than transplant or no treatment, and receiving care at a safety net facility were associated with decreased overall survival (Table 2). On multivariable analysis, lack of insurance (HR 1.337, 95%CI 1.012–1.766, p = 0.041), living in an area of lower educational attainment (44–77.3%: HR 1.496, 95%CI 1.089–2.057, p = 0.013), higher MELD at diagnosis (HR 1.074, 95%CI 1.060–1.088, p < 0.001), higher stage at diagnosis (II: HR 1.343, 95%CI 1.073–1.682, p = 0.010; III: HR 2.874, 95%CI 2.288–3.611, p < 0.001; IV: HR 3.230, 95%CI 2.583–4.038, p < 0.001), and treatment other than transplant (surgical resection: HR 3.369, 95%CI 2.043–5.554; liver directed therapy: HR 8.740, 95%CI 5.945–12.851; chemotherapy: HR 14.306, 95%CI 9.023–22.683, all p < 0.001), or no treatment (HR 28.997, 95%CI 19.039–44.162, p < 0.001) were independently associated with decreased overall survival. There were no racial/ethnic disparities nor was care at a safety net facility associated with decreased overall survival on multivariable analysis (Table 2).

Table 2.

Univariate and multivariable cox regression for overall survival – entire cohort.

	Univariate analysis		Multivariable analysis
Variable	HR (95% CI)	p-value	HR (95% CI)	p-value
Age	1.003 (0.997–1.010)	0.291	–	–
Gender			–	–
Female	Reference	0.060
Male	1.152 (0.994–1.334)
Race
White	Reference		Reference
Black	1.259 (1.092–1.451)	0.001	0.744 (0.611–0.905)	0.003
Asian	0.715 (0.543–0.941)	0.017	0.624 (0.445–0.875)	0.006
Ethnicity
Not Hispanic	Reference		Reference
Hispanic	1.192 (1.029–1.380)	0.019	0.713 (.547–0.928)	0.012
Insurance
Private	Reference		Reference
Government	1.323 (1.135–1.542)	<0.001	0.976 (0.807–1.180)	0.785
Uninsured	2.490 (2.032–3.051)	<0.001	1.337 (1.012–1.766)	0.041
Income Quartile
63,683+	Reference		Reference
47,863–63681	1.111 (0.928–1.331)	0.253	1.003 (0.732–1.373)	0.987
38,116–47862	1.299 (1.088–1.551)	0.004	1.036 (0.702–1.528)	0.860
0–38115	1.578 (1.324–1.881)	<0.001	0.960 (0.629–1.465)	0.850
Poverty*
0–11.8%	Reference		Reference
11.9–18.6%	1.207 (1.010–1.442)	0.039	0.839 (0.618–1.139)	0.260
18.7%+	1.436 (1.229–1.678)	<0.001	0.925 (0.627–1.364)	0.693
Education**
89.6%+	Reference		Reference
73.4–89.5%	1.303 (1.114–1.524)	0.001	1.198 (0.946–1.517)	0.133
44–73.3%	1.695 (1.421–2.024)	<0.001	1.496 (1.089–2.057)	0.013
Alcohol Abuse	1.379 (1.219–1.559)	<0.001	0.965 (0.817–1.139)	0.671
MELD	1.027 (1.023–1.031)	<0.001	1.074 (1.060–1.088)	<0.001
Stage at diagnosis
1	Reference		Reference
2	1.528 (1.278–1.828)	<0.001	1.343 (1.073–1.682)	0.010
3	3.454 (2.901–4.111)	<0.001	2.874 (2.288–3.611)	<0.001
4	4.631 (3.913–5.480)	<0.001	3.230 (2.583–4.038)	<0.001
Screening***	0.585 (0.497–0.688)	<0.001	1.030 (0.842–1.261)	0.773
Treatment Type
Transplant	Reference		Reference
Surgery	2.795 (1.864–4.189)	<0.001	3.369 (2.043–5.554)	<0.001
Liver Directed	7.381 (5.357–10.171)	<0.001	8.740 (5.945–12.851)	<0.001
Therapy Chemotherapy	19.494 (13.640–27.860)	<0.001	14.306 (9.023–22.683)	<0.001
No treatment	31.224 (22.398–43.528)	<0.001	28.997 (19.039–44.162)	<0.001
Facility Type
Tertiary	Reference		Reference
Referral
Safety Net	1.501 (1.327–1.697)	<0.001	1.138 (0.949–1.364)	0.163

^*% of households in patient’s zip code below the federal poverty level.

^**% of adults in patient’s zip code who graduated from high school.

^***Within one year of diagnosis.

3.3. Receipt of treatment

Factors associated with decreased odds of receiving treatment on univariate analysis included Black race, Hispanic ethnicity, lack of insurance, lower income, living in areas with higher rates of poverty and lower educational attainment, alcohol abuse, higher MELD at diagnosis, increasing stage at diagnosis, and treatment at a safety net facility. Screening within one year of diagnosis was associated with increased odds of receiving treatment (Table 3). On multivariable analysis however, there were no racial/ethnic or socioeconomic disparities in receipt of treatment for HCC in our cohort. Factors independently associated with decreased odds of receiving treatment included alcohol abuse (OR 0.682, 95%CI 0.485–0.959, p = 0.028), higher MELD at diagnosis (OR 0.874, 95%CI 0.851–0.898, p < 0.001), higher stage at diagnosis (II: OR 0.967, 95%CI 0.581–1.611, p = 0.899; III: OR 0.234, 95%CI 0.149–0.469, p < 0.001); IV: OR 0.210, 95%CI 0.135–0.327, p < 0.001), and care at a safety net facility (OR 0.424, 95%CI 0.289–0.622, p < 0.001) (Table 3).

Table 3.

Univariate and multivariable binary logistic regression for receipt of treatment – entire cohort.

	Univariate analysis		Multivariable analysis
Variable	OR (95% CI)	p-value	OR (95% CI)	p-value
Age	1.002 (0.991–1.013)	0.694	–	–
Gender			–	–
Female	Reference	0.711
Male	0.953 (0.740–1.228)
Race
White	Reference		Reference
Black	0.617 (0.479–0.796)	<0.001	1.146 (0.762–1.724)	0.512
Asian	0.928 (0.600–1.435)	0.737	0.733 (0.389–1.381)	0.336
Ethnicity
Not Hispanic	Reference		Reference
Hispanic	0.726 (0.567–0.930)	0.011	1.093 (0.617–1.936)	0.760
Insurance
Private	Reference		Reference
Government	0.783 (0.592–1.035)	0.085	0.995 (0.659–1.501)	0.980
Uninsured	0.252 (0.180–0.354)	<0.001	0.588 (0.339–1.020)	0.059
Income Quartile
63,683+	Reference		Reference
47,863–63681	1.192 (0.855–1.663)	0.301	1.928 (1.003–3.707)	0.049
38,116–47862	0.731 (0.536–0.996)	0.047	1.197 (0.544–2.634)	0.655
0–38115	0.563 (0.415–0.762)	<0.001	1.052 (0.451–2.453)	0.906
Poverty*
0–11.8%	Reference		Reference
11.9–18.6%	0.885 (0.636–1.231)	0.169	0.895 (0.467–1.717)	0.740
18.7%+	0.564 (0.427–0.746)	<0.001	0.761 (0.341–1.699)	0.505
Education**
89.6%+	Reference		Reference
73.4–89.5%	0.823 (0.623–1.087)	0.169	0.728 (0.429–1.236)	0.240
44–73.3%	0.553 (0.407–0.752)	<0.001	0.795 (0.403–1.569)	0.509
Alcohol Abuse	0.603 (0.407–0.752)	<0.001	0.682 (0.485–0.959)	0.028
MELD	0.883 (0.867–0.899)	<0.001	0.874 (0.851–0.898)	<0.001
Stage at diagnosis
1	Reference		Reference
2	0.864 (0.604–1.236)	0.423	0.967 (0.581–1.611)	0.899
3	0.250 (0.183–0.341)	<0.001	0.234 (0.149–0.369)	<0.001
4	0.192 (0.142–0.260)	<0.001	0.210 (0.135–0.327)	<0.001
Screening***	2.303 (1.685–3.148)	<0.001	1.583 (0.990–2.531)	0.055
Facility Type
Tertiary	Reference		Reference
Referral
Safety Net	0.382 (0.307–0.475)	<0.001	0.424 (0.289–0.622)	<0.001

^*% of households in patient’s zip code below the federal poverty level.

^**% of adults in patient’s zip code who graduated from high school.

^***Within one year of diagnosis.

3.4. Survival analysis in treated patients

When considering only patients who received treatment of any type, there were no racial/ethnic or socioeconomic disparities in overall survival (Table 4). Factors independently associated with decreased overall survival in treated patients included higher MELD at diagnosis (HR 1.051, 95%CI 1.034–1.069, p < 0.001), higher stage at diagnosis (II: HR 1.308, 95%CI 1.028–1.664, p = 0.029; III: HR 2.655, 2.026–3.480, p < 0.001; IV: HR 2.406, 95%CI 1.844–3.140, p < 0.001), and treatment type other than transplant (surgical resection: HR 3.317, 95%CI 2.005–5.486; liver directed therapy: HR 8.961, 95%CI 6.091–13.182; chemotherapy: HR 21.124, 95%CI 13.199–33.807, all p < 0.001) (Table 4). Notably, receiving care at a safety net facility was not associated with decreased overall survival in patients who received treatment (HR 1.049, 95%CI 0.844–1.303).

Table 4.

Univariate and multivariable cox regression for overall survival – patients receiving HCC treatment.

	Univariate analysis		Multivariable analysis
Variable	HR (95% CI)	p-value	HR (95% CI)	p-value
Age	1.005 (0.998–1.013)	0.182	–	–
Gender			–	–
Female	Reference	0.141
Male	1.140 (0.957–1.358)
Race
White	Reference		Reference
Black	1.127 (0.952–1.335)	0.165	0.729 (0.585–0.909)	0.005
Asian	0.637 (0.462–0.880)	0.006	0.570 (0.384–0.847)	0.005
Ethnicity			–	–
Not Hispanic	Reference
Hispanic	1.092 (0.911–1.310)	0.341
Insurance
Private	Reference		Reference
Government	1.348 (1.130–1.609)	0.001	1.012 (0.816–1.501)	0.915
Uninsured	1.884 (1.437–2.470)	<0.001	1.109 (0.776–1.585)	0.569
Income Quartile
63,683+	Reference		Reference
47,863–63681	1.212 (0.983–1.493)	0.072	0.990 (0.702–1.396)	0.955
38,116–47862	1.273 (1.030–1.575)	0.026	1.119 (0.729–1.717)	0.608
0–38115	1.573 (1.273–1.944)	<0.001	1.079 (0.671–1.737)	0.753
Poverty*
0–11.8%	Reference		Reference
11.9–18.6%	1.208 (0.984–1.482)	0.071	0.815 (0.585–1.134)	0.225
18.7%+	1.332 (1.110–1.599)	0.002	0.829 (0.540–1.273)	0.392
Education**
89.6%+	Reference		Reference
73.4–89.5%	1.382 (1.150–1.661)	0.001	1.179 (0.898–1.547)	0.236
44–73.3%	1.618 (1.306–2.006)	<0.001	1.399 (0.960–2.039)	0.080
Alcohol Abuse	1.343 (1.160–1.556)	<0.001	1.185 (0.983–1.429)	0.075
MELD	1.034 (1.020–1.048)	<0.001	1.051 (1.034–1.069)	<0.001
Stage at diagnosis
1	Reference		Reference
2	1.606 (1.321–1.953)	<0.001	1.308 (1.028–1.664)	0.029
3	2.973 (2.402–3.681)	<0.001	2.655 (2.026–3.480)	<0.001
4	3.965 (3.219–4.883)	<0.001	2.406 (1.844–3.140)	<0.001
Screening***	0.634 (0.527–0.763)	<0.001	0.940 (0.754–1.172)	0.584
Treatment Type
Transplant	Reference		Reference
Surgery	2.848 (1.900–4.269)	<0.001	3.317 (2.005–5.486)	<0.001
Liver Directed	7.923 (5.748–10.922)	<0.001	8.961 (6.091–13.182)	<0.001
Therapy Chemotherapy	24.469 (17.098–35.017)	<0.001	21.124 (13.199–33.807)	<0.001
Facility Type
Tertiary	Reference		Reference	0.669
Referral
Safety Net	1.267 (1.092–1.472)	0.002	1.049 (0.844–1.303)

^*% of households in patient’s zip code below the federal poverty level.

^**% of adults in patient’s zip code who graduated from high school.

^***Within one year of diagnosis.

4. Discussion

In our unique study cohort, comprised of five of the largest safety net hospitals in the United States and their sister tertiary referral centers which promoted a robust and diverse patient population, we found that tumor and host liver factors, as well as treatment type, largely drove survival outcomes, rather than race/ethnicity, socioeconomic factors, or facility type. When examining the entire cohort, lack of insurance, less educational attainment, higher MELD at diagnosis, higher stage at diagnosis, and treatment type or lack of treatment were associated with decreased overall survival. There were no racial/ethnic or socioeconomic disparities in receipt of treatment, however care at a safety net facility was associated with decreased odds of receiving treatment, a phenomenon that has been previously documented in the literature [9, 10]. Otherwise, receipt of treatment was driven by a history of alcohol abuse, MELD, and stage at diagnosis. In patients who did receive treatment, there were no racial/ethnic or socioeconomic disparities in overall survival, nor was treatment at a safety net facility associated with decreased survival. Outcomes were again driven by MELD, stage at diagnosis, and treatment type.

The interplay of race/ethnicity, socioeconomic factors, disease biology, and the condition of the host liver is complex as it relates to treatment and outcomes of patients with HCC. Race has been associated with disparate outcomes in a number of diseases, and early explanations for these differences centered around genetics and biology [11]. However, self-reported racial categories are not accurate indicators of biologic differences, and, in fact, the majority of genetic variation occurs amongst individuals, rather than racial groups [11]. Thus, racial disparities in the United States likely represent other factors captured by racial categories, including differences in socioeconomic factors and environmental exposures. The fact that socioeconomic factors vary by race and tend to favor non-Hispanic White and Asian populations has also been well documented. In the most recent survey conducted by the United States Census Bureau, the median household income for all races was $61,372 [12]. The variation by racial group is quite striking; the median income for Asian households was $81,331, compared to $68, 145 for non-Hispanic White, $50,486 for Hispanic, and $40,258 for non-Hispanic Black households [12]. Similar trends are seen in educational attainment, in 2018 24.2% of non-Hispanic White citizens obtained a Bachelor’s degree, compared to 16.3% of non-Hispanic Black, 13.0% of Hispanic, and 31.4% of Asian citizens [13]. These differences also carry over into insurance coverage; non-Hispanic Whites had the lowest uninsured rate (6.3% in 2017), followed by Asians (7.3%), non-Hispanic Blacks (10.6%), and Hispanics (16.1%) [14].

It is important to also consider the downstream effects of differences in environmental exposures that may accompany racial/ethnic, and socioeconomic differences, for example, exposure to adversity throughout the life course, when considering health outcomes. In general, research regarding socioeconomic status and health indicates that the more advantaged a person, the better their health, specifically regarding chronic diseases. The relationship between socioeconomic factors and cancer outcomes is more complicated and less linear, however the strongest link appears to be in cancers in which local symptoms precede development of metastatic disease, for example cancers of the head and neck, and thus access to healthcare at an earlier stage makes a greater difference in mortality, compared to cancers that are often diagnosed at an advanced stage, such as pancreatic adenocarcinoma where early and frequent access to care does not play such a large role in mortality [15]. We would postulate that this link can be reasonably extrapolated to cancers for which screening is recommended to facilitate early detection of disease, such as hepatocellular carcinoma.

Safety net hospitals, defined by the Institute of Medicine as facilities and providers who “by mandate or mission offer access to care regardless of a patient’s ability to pay and whose patient population includes a substantial share of uninsured, Medicaid, and other vulnerable patients” [16], attempt to address disparities in access to care that ultimately affect health outcomes associated with race and socioeconomic status, however are often hampered by limited resources, including specialty services [9,17]. In our cohort, patients seen at safety net facilities were more likely to be non-White, Hispanic, and uninsured, with lower median income, and less educational attainment than patients seen at tertiary referral centers. These patients also presented with more advanced disease and were less likely to receive treatment, which likely led to the survival difference seen between safety net and tertiary referral center patients on Kaplan Meier analysis. The lack of racial/ethnic and socioeconomic disparities in receipt of treatment and overall survival for HCC in this study is likely due to care at a safety net facility representing the amalgamation of these factors. As previously mentioned, the quality of care provided at safety net hospitals has been called into question, however the lack of disparity in overall survival in the entire cohort and in treated patients between patients cared for at safety net and tertiary referral centers, when accounting for stage at diagnosis and the host liver condition, indicates that similar and appropriate care is being provided. The decreased odds of receiving treatment at a safety net facility is likely multifactorial. Several patient characteristics seen in higher rates at safety net hospitals such as homelessness, alcohol abuse and mental health diagnoses, as well as factors associated with lower income such as lack of reliable transportation, contribute to less stability and reliable follow up which potentially leads to decreased adherence with treatment recommendations. Though not specifically investigated in this study, higher rates of comorbidities and lack of trust in the healthcare system may also contribute to decreased ability to tolerate and willingness to accept treatment, respectively.

While there is overlap amongst race/ethnicity, socioeconomic factors, and care at a safety net facility, there is no intrinsic or direct association with these factors and worse outcomes, but instead, the reality is much more complicated. Ultimately, poor liver function and advanced stage of presentation are the main determinants of decreased survival. And fortunately, one of these factors, stage at diagnosis, is potentially modifiable through screening of patients with cirrhosis. Patients in our cohort who received screening within one year of diagnosis were more likely to present with early stage disease and were more likely to receive treatment, regardless of facility type. Unfortunately, however, screening rates were low in both safety net (17%) and tertiary referral centers (27%) representing a collaborative-wide opportunity for improvement.

There are several limitations to this study, including its retrospective design, which naturally invites selection bias. However, this was a multi-institutional study comprised of geographically and demographically diverse institutions allowing for a robust patient population. This study is both strengthened and limited by the specific safety net facilities included in the collaborative. The five safety net members are some of the largest in the country, however as such, the available resources to these institutions likely exceed those of smaller facilities, and all facilities are high volume, both factors that limit generalizability. Finally, socioeconomic factor data was available only at the zip code and not individual level, and we recognize that wide variation within zip code can and does exist, also introducing bias into our models and results.

5. Conclusion

Though racial/ethnic and socioeconomic disparities in HCC outcomes have previously been reported in the literature, the complex relationship between these factors has not yet been effectively dissected. We found that rather than race, ethnicity, socioeconomic status, or treatment facility type, host liver function and stage at presentation drove outcomes. To improve outcomes moving forward, our focus should involve implementing screening programs, in both safety net and tertiary referral centers and supporting safety net hospitals by provision of resources in caring for vulnerable patient populations.