Disparities in health utilities among hepatitis C patients receiving care in different settings

Abstract

BACKGROUND: Although chronic hepatitis C (CHC) disproportionately affects marginalized individuals, most health utility studies are conducted in hospital settings which are difficult for marginalized patients to access. We compared health utilities in CHC patients receiving care at hospital-based clinics and socio-economically marginalized CHC patients receiving care through a community-based program. METHODS: We recruited CHC patients from hospital-based clinics at the University Health Network and community-based sites of the Toronto Community Hep C Program, which provides treatment, support, and education to patients who have difficulty accessing mainstream health care. We elicited utilities using six standardized instruments (EuroQol-5D-3L [EQ-5D], Health Utilities Index Mark 2/Mark 3 [HUI2/HUI3], Short Form-6D [SF-6D], time trade-off [TTO], and Visual Analogue Scale [VAS]). Multivariable regression analysis was performed to examine factors associated with differences in health utility. RESULTS: Compared with patients recruited from the hospital setting (n = 190), patients recruited from the community setting (n = 101) had higher unemployment (87% versus 67%), history of injection drug use (88% versus 42%), and history of mental health issue(s) (79% versus 46%). Unadjusted health utilities were lower in community than hospital patients (e.g., EQ-5D: 0.722 [SD 0.209] versus 0.806 [SD 0.195]). Unemployment and a history of mental health issue(s) were significant predictors of low health utility. CONCLUSIONS: Socio-economically marginalized CHC patients have lower health utilities than patients typically represented in the CHC utility literature. Their utilities should be incorporated into future cost-utility analyses to better represent the population living with CHC in health policy decisions.

Introduction

Hepatitis C is estimated to affect 2.5% of the global population and 1.3% of people living in the Americas (1). This includes those who are hepatitis C antibody-positive, indicating current or past infection. The virus disproportionately affects socio-economically marginalized individuals. Published studies have demonstrated associations between hepatitis C antibody positivity and unstable housing, homelessness, unemployment, low educational attainment, low income, and mental health issues (2,3). In Canada, it is estimated that 66% of people who inject drugs, 29% of people who formerly injected drugs, and 24% of people in correctional facilities are hepatitis C antibody-positive (4). Among those who are antibody-positive, 67% are estimated to have chronic hepatitis C (CHC) (1).

Marginalized individuals with CHC have a poorer prognosis (5) and health-related quality of life (6) than CHC patients who are not marginalized. Socio-economically marginalized individuals face many barriers to accessing mainstream CHC treatment and care, including hospital-based clinics. Barriers include clinics’ intolerance for late/missed appointments; lack of resources to accommodate patients with mental health issues; and stigmatization of mental health issues, homelessness, and substance use (7). Because hospital-based clinics are the setting for a high proportion of research studies, marginalized patients are often excluded from clinical research, including health state utility value (hereafter health utility) studies.

Health utility is a global measure of health status, and a useful outcome measure for capturing the burden of diseases including CHC. Utilities are anchored at 0 (equivalent to dead) and 1 (equivalent to full health), although utilities less than zero are possible for health states considered to be worse than dead (8). In contrast to widely used psychometric quality of life measures which describe the presence and severity of symptoms, health utilities provide insight into the impacts of diseases on patients’ overall health status by incorporating preferences for health outcomes (9). Indirect utility measures use societal preferences to value health states, while direct measures use the respondent’s own preferences to value health states. Thus, utilities capture not only a patient’s state of health but also how it is valued.

Health utilities can be used to measure individual-level health, quantify disease burden in a population, and measure the potential benefits and harms of clinical and policy interventions—such as in cost-utility analysis (10). Cost-utility analysis is crucial to guide health policy decisions surrounding hepatitis C screening, treatment, and prevention. Thus, it is essential to capture health utilities from individuals who are representative of the population actually living with CHC, including socio-economically marginalized patients.

In order to explore these issues, we recruited CHC patients from hospital-based liver clinics as well as a community-based hepatitis C program designed to be inclusive of socio-economically marginalized individuals. We compared health utilities between CHC patients from the two settings to provide insight into whether utility studies conducted in hospital settings are potentially unrepresentative of a distinct and important demographic group.

Methods

Ethics

This study received approval from the Research Ethics Boards of the University of Toronto and the University Health Network in Toronto, Ontario, Canada. Written informed consent was obtained from all subjects prior to study participation.

Subject recruitment

Consecutive patients were recruited from three community-based sites of the Toronto Community Hep C Program and three hospital-based clinics at the University Health Network from February 2015 to March 2017. Patients currently infected with the hepatitis C virus (hepatitis C virus RNA-positive) with mild to moderate liver damage (METAVIR F0-F3), as well as patients with past or current infection (hepatitis C virus antibody-positive) and advanced liver disease (METAVIR F4), were eligible for the study. There were no exclusion criteria related to age, coinfection, or comorbidity.

The Toronto Community Hep C Program (11) is a collaboration between three sites in downtown Toronto: Sherbourne Health, South Riverdale Community Health Centre, and Regent Park Community Health Centre. This community-based interprofessional program provides hepatitis C treatment, support, and education to marginalized patients who have difficulty accessing mainstream health care. It promotes de-stigmatization of mental health and substance use issues, offers harm reduction services and provides food and public transit fares to reduce financial barriers to attending. Patients were recruited at the weekly group support sessions (12) associated with this program.

The University Health Network is a network of research and teaching hospitals affiliated with the University of Toronto. Patients were recruited from the outpatient liver clinic and liver transplant clinic at Toronto General Hospital, as well as the gastrointestinal cancer clinic at Princess Margaret Cancer Centre. These clinics serve patients referred by physicians from Toronto and the surrounding area. Patients were recruited at these clinics when they came for a scheduled appointment.

Assessments

We collected patients’ health utilities, sociodemographic characteristics, and clinical information using a pencil-and-paper interviewer-administered survey.

We asked patients to rate their current health using six commonly used, validated health utility instruments. We used four indirect utility instruments: EuroQol-5D-3L (EQ-5D) (10), Health Utilities Index Mark 2 and Mark 3 (HUI2 and HUI3) (8), and Short Form-6D (SF-6D) (13); and two direct utility instruments: time trade-off (TTO) (14) and Visual Analogue Scale (VAS) (using the EQ-VAS included with the EQ-5D instrument) (10). Indirect utility instruments consisted of a number of multiple-choice questions pertaining to various domains of health (e.g., mobility, self-care, pain) and asked patients to describe their current health state. Societal preferences for that health state, obtained from a general population sample, were used to generate a health utility. (See Supplemental Table 1 for sources of societal preference weights used.) In contrast, direct utility instruments reflected patients’ own preferences. The TTO asked patients to value their own current health state by asking how much time they would be willing to trade to avoid that health state (15). The EQ-VAS asked patients to rate their current health on a scale from 0 (worst imaginable health state) to 100 (best imaginable health state) (10).

Sociodemographic information collected included age, sex, race, immigrant status (born in Canada or not), marital status, employment status, and education. Clinical information included self-reported comorbidities, mental health history, substance use history, and hepatitis C antiviral treatment history.

Additional clinical information was collected through a medical chart review, including hepatitis C genotype, METAVIR liver fibrosis stage (based on liver biopsy or mapped from FibroScan or FibroTest), presence of decompensated cirrhosis or hepatocellular carcinoma, and Child-Pugh class.

Analysis

Baseline characteristics, stratified by hospital or community recruitment setting, were described using means and standard deviations for continuous variables and frequency counts and proportions for categorical variables. Patients were grouped into the following liver disease health states: no cirrhosis (METAVIR F0-F3), compensated cirrhosis (F4), decompensated cirrhosis (DC) (defined by current or past ascites, variceal bleeding, or hepatic encephalopathy), hepatocellular carcinoma (HCC), and HCC with DC (HCC + DC).

For each utility instrument, mean (SD) health utilities were calculated stratified by age, sex, race, marital status, immigrant status, education, employment status, Charlson Comorbidity Index, substance use history, mental health history, liver disease health state, Child-Pugh class (for patients with cirrhosis), hospital or community setting, and setting plus liver disease health state.

Responses to the EQ-5D and HUI3 instruments were presented in radar charts (16) to visualize scores in the five quality of life domains of the EQ-5D (mobility, self-care, usual activities, pain/discomfort, and anxiety/depression) and the eight attributes of the HUI3 (vision, hearing, speech, ambulation, dexterity, emotion, cognition, and pain), stratified by setting.

Two multivariable linear regression models were fitted for each utility instrument to analyze the effect of setting on health utility. The purpose of the first regression model was to examine whether setting had a significant effect on health utility. This model adjusted for key variables that were not intrinsic to the setting: age, sex, liver disease severity, and Charlson Comorbidity Index. In the second regression model, we included these intrinsic characteristics—namely, socio-economic factors—to determine which factors contributed most to any differences found between settings. This model included the above covariates as well as education, employment status, history of injection drug use, and history of mental health issues.

Multiple Imputation by Chained Equations (MICE) was used to impute missing covariate and utility values for all regression analyses, to avoid potential bias from excluding incomplete observations (17). Data were assumed to be missing at random. The MICE method generates multiple sets of imputed data to account for uncertainty in the imputations. Twenty sets of imputed data were generated. Regression analysis was performed on each set of imputed data; then, the results were pooled into a single set of estimates, 95% CIs, and p values.

All analyses were performed in R (R Foundation for Statistical Computing, Vienna, Austria)(18). Radar charts were generated using the ‘fmsb’ package (19). MICE was performed using the ‘mice’ package (20), and its ‘pool’ function was used to pool regression analyses of the imputed datasets.

Results

Patient characteristics

One hundred one patients were recruited from the community-based sites and 190 from the hospital-based sites.

Hospital patients were older (mean age: 58 versus 51 years), and more likely to be female (41% versus 33%), immigrants (42% versus 13%), and married (54% versus 8%) (Table 1). These patients also had higher comorbidity (mean Charlson Comorbidity Index: 1.45 versus 1.31).

Table 1:

Patient characteristics by setting

	All; n = 291	Community; n = 101	Hospital; n = 190
Age	56 (SD 11)	51 (SD 10)	58 (SD 11)
Sex (male)	179 (62%)	67 (67%)	112 (59%)
Race*
White	184 (68%)	77 (77%)	107 (63%)
Black	17 (6%)	8 (8%)	9 (5%)
South Asian	15 (6%)	0 (0%)	15 (9%)
Indigenous	12 (4%)	7 (7%)	5 (3%)
Other	43 (16%)	8 (8%)	35 (20%)
Immigrant	89 (33%)	11 (13%)	78 (42%)
Married	99 (37%)	8 (8%)	91 (54%)
Education
Did not finish high school	93 (37%)	44 (47%)	49 (31%)
Finished high school only	52 (21%)	19 (20%)	33 (21%)
Attended post-secondary	108 (43%)	31 (33%)	77 (48%)
Unemployed	202 (75%)	88 (87%)	114 (67%)
On social support†	164 (64%)	94 (97%)	70 (44%)
Charlson Comorbidity Index
Mean	1.39 (SD 1.60)	1.31 (SD 1.29)	1.45 (SD 1.81)
0	70 (34%)	25 (28%)	45 (40%)
1	59 (29%)	32 (36%)	27 (24%)
2	41 (20%)	24 (27%)	17 (15%)
3+	33 (16%)	9 (10%)	24 (21%)
HIV-positive	10 (4%)	5 (5%)	5 (3%)
Substance use
History of alcohol dependence	86 (36%)	63 (62%)	23 (17%)
History of injection drug use	152 (61%)	89 (88%)	63 (42%)
History of intranasal drug use	166 (68%)	92 (91%)	74 (51%)
History of drug dependence	123 (51%)	83 (83%)	40 (28%)
Current injection or intranasal drug use	29 (12%)	25 (25%)	4 (3%)
Currently taking methadone	31 (13%)	25 (25%)	6 (4%)
History of mental health issues	146 (60%)	80 (79%)	66 (46%)
Liver disease severity
No cirrhosis	153 (53%)	76 (75%)	77 (41%)
Compensated cirrhosis	58 (20%)	22 (22%)	36 (19%)
Decompensated cirrhosis (DC)	38 (13%)	3 (3%)	35 (18%)
Hepatocellular carcinoma (HCC)	27 (9%)	0 (0%)	27 (14%)
HCC + DC	15 (5%)	0 (0%)	15 (8%)
Child-Pugh class‡
A	85 (29%)	21 (21%)	64 (34%)
B	40 (14%)	3 (3%)	37 (20%)
C	11 (4%)	0 (0%)	11 (6%)
Genotype
1	154 (62%)	62 (67%)	92 (58%)
2	26 (10%)	8 (9%)	18 (11%)
3	59 (24%)	22 (24%)	37 (23%)
Other	11 (4%)	0 (0%)	11 (7%)
Hepatitis C treatment history
History of treatment	104 (37%)	8 (9%)	96 (51%)
History of successful treatment (SVR)	33 (13%)	0 (0%)	33 (19%)

^* Races with at least n = 10 are shown separately

^† Includes welfare, disability insurance, unemployment insurance

^‡ Applies to cirrhotic patients only (including those with compensated cirrhosis, decompensated cirrhosis, hepatocellular carcinoma, and HCC+DC)

HIV = Human immunodeficiency virus; SVR = Sustained virologic response

Higher percentages of the community patients had a history of mental health issue(s) (79% versus 46%), history of alcohol dependence (62% versus 17%), low level of educational attainment (47% versus 31% did not complete high school), and were unemployed (87% versus 67%).

The main risk factors for CHC reported by patients in both subgroups were injection drug use and intranasal drug use. A higher proportion of patients from the community subgroup had a history of injection (88% versus 42%) or intranasal (91% versus 51%) drug use, and a higher proportion reported that they were currently using drugs compared with the hospital subgroup (25% versus 3%).

Among the community patients, 76 had no cirrhosis, 22 had compensated cirrhosis, 3 had DC, and none had HCC. The hospital patients comprised 77 patients with no cirrhosis, 36 with compensated cirrhosis, 35 with DC, 27 with HCC, and 15 with HCC + DC. Hospital patients had more advanced liver disease based on health state (59% versus 25% with advanced liver disease [compensated cirrhosis, HCC, DC, or HCC+DC]) and Child-Pugh class (Class B: 20% versus 3%; Class C: 6% versus 0%). Most patients had genotype 1 infection across both subgroups (62%). Just over half of hospital patients had previously received antiviral treatment for hepatitis C, compared with very few of the community patients (51% versus 9%).

Health utilities by patient characteristics

In simple stratified analyses, older age was associated with lower health utilities across all utility instruments (EQ-5D for age >40: 0.754; age ≤40: 0.787) (Table 2). Female and male patients had similar health utilities (EQ-5D: 0.754 versus 0.760). South Asian patients and those of ‘other’ races had the highest mean utilities on all instruments except the SF-6D, while Indigenous patients had the lowest mean utilities (EQ-5D for South Asian patients: 0.785; Indigenous patients: 0.720) (‘other’ included all races with n <10). Sample sizes were small (<20) for all non-white races.

Table 2:

Health utilities by utility instrument, stratified by clinical and sociodemographic variables

		Utility Instrument, mean (SD) utility
Subgroup	n	EQ-5D	HUI2	HUI3	SF-6D	TTO	VAS
Age
≤40	25	0.787 (0.206)	0.831 (0.173)	0.705 (0.321)	0.655 (0.121)	0.908 (0.173)	0.764 (0.121)
>40	264	0.754 (0.207)	0.738 (0.206)	0.607 (0.322)	0.618 (0.131)	0.807 (0.242)	0.668 (0.197)
Sex
Female	111	0.754 (0.212)	0.731 (0.218)	0.591 (0.335)	0.613 (0.138)	0.808 (0.243)	0.677 (0.175)
Male	179	0.760 (0.204)	0.757 (0.197)	0.633 (0.315)	0.627 (0.125)	0.822 (0.236)	0.676 (0.204)
Race*
White	184	0.762 (0.204)	0.746 (0.200)	0.621 (0.304)	0.624 (0.128)	0.810 (0.253)	0.675 (0.189)
Black	17	0.734 (0.260)	0.760 (0.193)	0.515 (0.415)	0.615 (0.169)	0.833 (0.223)	0.626 (0.205)
South Asian	15	0.785 (0.192)	0.766 (0.221)	0.669 (0.348)	0.576 (0.124)	0.854 (0.142)	0.700 (0.179)
Indigenous	12	0.720 (0.251)	0.675 (0.289)	0.474 (0.391)	0.569 (0.095)	0.782 (0.272)	0.561 (0.231)
Other	43	0.766 (0.199)	0.768 (0.196)	0.659 (0.319)	0.624 (0.125)	0.845 (0.204)	0.719 (0.200)
Marital status
Not married	169	0.733 (0.224)	0.722 (0.207)	0.572 (0.330)	0.607 (0.128)	0.808 (0.259)	0.655 (0.190)
Married	99	0.803 (0.171)	0.787 (0.201)	0.694 (0.287)	0.641 (0.128)	0.838 (0.202)	0.711 (0.197)
Immigrant status
Not an immigrant	183	0.745 (0.200)	0.725 (0.211)	0.586 (0.317)	0.611 (0.124)	0.809 (0.253)	0.654 (0.192)
Immigrant	89	0.792 (0.212)	0.792 (0.185)	0.691 (0.310)	0.648 (0.140)	0.836 (0.207)	0.722 (0.196)
Education
Did not finish high school	93	0.728 (0.201)	0.706 (0.217)	0.554 (0.337)	0.600 (0.113)	0.816 (0.248)	0.648 (0.185)
Finished high school only	52	0.779 (0.193)	0.753 (0.190)	0.629 (0.308)	0.609 (0.126)	0.814 (0.265)	0.672 (0.195)
Attended post-secondary	108	0.775 (0.220)	0.779 (0.202)	0.668 (0.307)	0.644 (0.142)	0.835 (0.213)	0.700 (0.207)
Employment status
Employed	68	0.861 (0.144)	0.857 (0.130)	0.789 (0.197)	0.698 (0.109)	0.879 (0.200)	0.777 (0.137)
Unemployed	202	0.721 (0.217)	0.706 (0.213)	0.556 (0.331)	0.589 (0.123)	0.796 (0.250)	0.639 (0.199)
Charlson Comorbidity Index
–0	70	0.819 (0.193)	0.813 (0.163)	0.706 (0.270)	0.674 (0.133)	0.881 (0.179)	0.735 (0.173)
–1	59	0.731 (0.194)	0.708 (0.196)	0.547 (0.321)	0.583 (0.115)	0.828 (0.253)	0.646 (0.169)
–2	41	0.708 (0.263)	0.707 (0.227)	0.538 (0.355)	0.583 (0.124)	0.772 (0.270)	0.616 (0.187)
–3+	33	0.706 (0.220)	0.697 (0.217)	0.540 (0.345)	0.591 (0.134)	0.773 (0.249)	0.614 (0.254)
Substance use history
History of injection drug use	151	0.736 (0.215)	0.723 (0.200)	0.574 (0.324)	0.604 (0.122)	0.816 (0.246)	0.648 (0.185)
No history of injection drug use	94	0.793 (0.205)	0.790 (0.201)	0.671 (0.318)	0.639 (0.141)	0.825 (0.229)	0.712 (0.208)
Mental health history
History of mental health issues	146	0.699 (0.219)	0.699 (0.204)	0.529 (0.332)	0.588 (0.118)	0.829 (0.236)	0.643 (0.183)
No history of mental health issues	98	0.848 (0.163)	0.821 (0.179)	0.726 (0.283)	0.667 (0.132)	0.815 (0.234)	0.726 (0.205)
Liver disease health state
No cirrhosis	153	0.784 (0.186)	0.772 (0.184)	0.647 (0.302)	0.637 (0.123)	0.856 (0.238)	0.703 (0.170)
Compensated cirrhosis	58	0.723 (0.245)	0.732 (0.210)	0.588 (0.336)	0.590 (0.144)	0.787 (0.221)	0.645 (0.214)
Decompensated cirrhosis (DC)	38	0.723 (0.230)	0.691 (0.243)	0.557 (0.346)	0.603 (0.124)	0.755 (0.256)	0.657 (0.190)
Hepatocellular carcinoma (HCC)	27	0.776 (0.191)	0.761 (0.221)	0.678 (0.331)	0.633 (0.146)	0.738 (0.232)	0.652 (0.235)
HCC + DC	15	0.676 (0.176)	0.639 (0.238)	0.434 (0.345)	0.597 (0.104)	0.792 (0.215)	0.617 (0.239)
Child-Pugh class
A	85	0.741 (0.226)	0.746 (0.211)	0.614 (0.335)	0.604 (0.138)	0.781 (0.221)	0.655 (0.208)
B	40	0.729 (0.201)	0.681 (0.240)	0.561 (0.329)	0.607 (0.120)	0.763 (0.251)	0.652 (0.200)
C	11	0.649 (0.210)	0.657 (0.221)	0.423 (0.362)	0.552 (0.136)	0.719 (0.259)	0.606 (0.248)
Setting†
Community	98	0.722 (0.209)	0.719 (0.196)	0.548 (0.326)	0.591 (0.117)	0.841 (0.246)	0.644 (0.183)
Hospital	113	0.806 (0.195)	0.797 (0.181)	0.703 (0.281)	0.656 (0.135)	0.835 (0.225)	0.725 (0.179)
Setting and liver disease health state
No cirrhosis, community	76	0.732 (0.200)	0.731 (0.186)	0.564 (0.327)	0.609 (0.115)	0.872 (0.234)	0.664 (0.167)
No cirrhosis, hospital	77	0.835 (0.157)	0.812 (0.173)	0.730 (0.251)	0.668 (0.125)	0.840 (0.242)	0.742 (0.166)
Compensated cirrhosis, community	22	0.686 (0.238)	0.680 (0.225)	0.491 (0.323)	0.525 (0.102)	0.734 (0.264)	0.576 (0.220)
Compensated cirrhosis, hospital	36	0.744 (0.250)	0.764 (0.197)	0.647 (0.334)	0.631 (0.153)	0.823 (0.182)	0.688 (0.202)

^* Races with at least n = 10 are shown separately

^† Since only three patients from the community sites had advanced liver disease (HCC and/or DC), the comparison based on setting was limited to patients with no cirrhosis or compensated cirrhosis

EQ-5D = EuroQol-5D-3L; HUI2 = Health Utilities Index Mark 2; HUI3 = Health Utilities Index Mark 3; SF-6D = Short Form-6D; TTO = time trade-off; VAS = Visual analogue scale

Utilities were lower for patients who were unemployed (EQ-5D: 0.721 versus 0.861) or had a history of injection drug use (EQ-5D: 0.736 versus 0.793) or mental health issues (except on the TTO; EQ-5D: 0.699 versus 0.848). Utilities decreased with increasing Charlson Comorbidity Index (EQ-5D for Charlson Index 0: 0.819; for Charlson Index 3+: 0.706). Higher utilities were observed for patients who were immigrants (EQ-5D: 0.792 versus 0.745), married (EQ-5D: 0.803 versus 0.733), or had a high level of education (EQ-5D for patients who attended post-secondary education: 0.775; patients who did not finish high school: 0.728).

Patients with compensated cirrhosis had lower mean health utilities than patients without cirrhosis on all utility instruments (EQ-5D: 0.723 versus 0.784). Patients with decompensated cirrhosis (EQ-5D: 0.723) had similar or lower utilities than those with compensated cirrhosis across all instruments. Utilities for patients with hepatocellular carcinoma (EQ-5D: 0.776) fell in between those for patients with no cirrhosis and compensated cirrhosis across most instruments. Patients in the HCC + DC health state (EQ-5D: 0.676) had the lowest utilities across most instruments except the SF-6D and TTO. A more advanced Child-Pugh class was associated with lower utilities (EQ-5D for Class A: 0.741; Class C: 0.649).

The TTO instrument produced the highest utilities for each health state (range: 0.738–0.856), while the HUI3 and SF-6D produced the lowest utilities (HUI3 range: 0.434–0.678; SF-6D range: 0.590–0.637).

Health utilities by setting

Because only three patients from the community sites had HCC and/or DC, the comparisons based on setting were limited to patients with no cirrhosis or compensated cirrhosis. Health utilities were lower in the community subgroup across all utility instruments except the TTO—in most cases, by a large amount (EQ-5D: community: 0.722; hospital: 0.806) (Table 2). A large difference between settings remained when stratifying utilities by setting and health state.

When comparing results broken down by EQ-5D domain, community patients were more likely to have ‘some’ or ‘severe’ problems with anxiety/depression, mobility, pain/discomfort, and usual activities compared with hospital patients (Figure 1). The most notable difference was observed in the anxiety/depression domain of the EQ-5D. For the HUI3 instrument, community patients were more likely to report moderate or severe disability in the emotion, pain, and cognition attributes—with the greatest disparity between patients from different settings occurring in the pain and cognition attributes.

Figure 1:

Responses to EQ-5D and HUI3 health utility questionnaires by quality of life domain, stratified by setting. (left) Proportion of responses indicating “some problems” or “severe problems” in each EQ-5D domain; (right) Proportion of responses indicating moderate or severe disability in each HUI3 attribute

EQ-5D = EuroQol-5D-3L; HUI3 = Health Utilities Index Mark 3

Regression analyses

The first regression model (adjusting for age, sex, Charlson Comorbidity Index, and liver disease severity) indicated that the effect of setting was large and statistically significant (p <0.05) across all utility instruments except the TTO (EQ-5D coefficient for community setting: -0.076; p = 0.014) (Figure 2; see Supplemental Table 2 for results for all utility instruments). The effects of the compensated cirrhosis health state (EQ-5D coefficient: -0.064; p = 0.043) and Charlson Comorbidity Index score (EQ-5D coefficient for Charlson Comorbidity Index ≥3: –0.111; p = 0.033) were statistically significant across some instruments and were associated with lower health utilities.

Figure 2:

Regression results for EQ-5D health utility instrument: effect of setting and clinical factors on health utilities (Model 1); and effect of setting, clinical factors, and socioeconomic factors on health utilities (Model 2)

Note: Linear regression models were run separately for each utility instrument, using data collected from the study and MICE imputation to estimate missing covariates and utility values.

^*p < 0.05

EQ-5D = EuroQol-5D-3L; MICE = Multiple Imputation by Chained Equations

To explore socio-economic contributors to the difference in health utilities between settings, the following variables were added in the second regression model: education, employment status, history of injection drug use, and history of mental health issues (Figure 2; see Supplemental Table 2 for results for all utility instruments). This model revealed that unemployment (EQ-5D coefficient: –0.091; p = 0.006) and a history of mental health issues (EQ-5D coefficient: –0.123; p <0.001) were associated with substantially lower utilities (p <0.05 across most instruments). Other variables did not reach statistical significance across most instruments, including the setting variable.

Discussion

Our study provides a unique perspective by comparing utilities elicited from CHC patients in two distinct settings. We recruited patients from hospital-based clinics, where utility studies are commonly conducted; as well as community health centres attended by socio-economically marginalized patients who are not well represented in the health utility literature.

We compared unadjusted health utilities for patients from each setting. We then used two regression models to confirm the difference in health utilities between settings and examine potential underlying causes.

Patients recruited from the community setting had lower health utilities

Health utilities were substantially lower in patients recruited in the community setting. A commonly used threshold for a clinically important difference in health utility is 0.03 (21). The differences observed between settings far exceeded this threshold, ranging from 0.065 to 0.155 for all instruments except the TTO.

Utilities from the community patients in our study were also substantially lower than synthesized utility estimates reported in a recent comprehensive meta-analysis of CHC health utilities (22) (Supplemental Table 3). Utilities for most instruments from our hospital subgroup were higher or similar to the meta-analyzed utilities. This suggests that many patients who participated in the meta-analyzed studies were drawn from a less marginalized population.

This is important because cost-utility analyses of CHC screening and treatment programs rely on utility estimates from the literature. Marginalized populations make up a large proportion of CHC patients yet have not been well represented in the CHC health utility literature. This leads to bias in the results of CHC cost-utility analyses. We recommend incorporating marginalized populations into cost-utility analyses by using more representative health utilities as model inputs, weighting health utilities to be more representative of population characteristics, and/or performing sub-analyses focusing on marginalized populations. Our results could also be used to inform equity-weighted cost-utility analyses, which are often hindered by a lack of data on social distributions of key parameters such as health utilities (23).

Potential causes of lower health utilities in patients from the community setting

Our analysis suggests that socio-economic factors—particularly unemployment and a history of mental health issues—are key contributors to the lower health utilities seen in the community setting. It is unclear whether this relationship is independent of CHC, or if there is an interaction between socio-economic factors and CHC.

Comorbid social or health conditions associated with marginalization such as poor nutrition, reduced access to health care, alcohol or substance use, and chronic health conditions may interact with CHC to amplify the physiologic effects of the disease (24–26); potentially leading to an increase in symptoms such as fatigue and mild cognitive impairment, and reducing quality of life. Mental health issues, substance use, and social isolation could also impact patients’ ability to cope with CHC (27,28), for example by increasing worry about the disease or increasing the level of stigma or discrimination experienced.

Alternatively, comorbid health and social conditions may act independently of CHC to reduce quality of life while CHC itself may have a comparatively minimal impact (29–31). Challenges such as mental health issues and unemployment could be responsible for a large negative impact on health-related quality of life that overshadows the effects of CHC.

This is an important question because its answer has a large bearing on the optimal strategy to improve quality of life in marginalized CHC patients. If CHC itself is a major contributor to lower health utilities in marginalized patients, a heavy focus on increasing access to antiviral therapy in this population could have a large beneficial impact. However, if this population’s utilities are more affected by comorbid medical and socio-economic issues than hepatitis C viral status, then antiviral therapy alone might have minimal impact on their health utilities. This would warrant greater focus on addressing mental health and unemployment alongside antiviral therapy.

A longitudinal study on the effect of antiviral therapy on health utilities in marginalized CHC patients is needed to shed light on this question.

Effects of advanced liver disease on health utility

Our study found higher than expected health utilities for the HCC and DC health states, given the severe symptoms that can occur with these conditions. This has also been observed in previous CHC utility literature (22). This is likely due to the fact that the HCC and DC health states each encompass a broad spectrum of disease severity, ranging from a small tumour diagnosed early to severe symptoms requiring hospitalization and/or liver transplantation. It is also worth noting that patients with more advanced HCC or DC are less likely to participate in research studies due to their symptoms and are thus underrepresented in utility estimates for these health states. Grouping mild and severe patients together within the HCC and DC health states dilutes the ability to fully capture the potential harms associated with end-stage liver disease.

A more granular health state classification would better reflect the potential utility decrements associated with end-stage liver disease in CHC. When we grouped patients with cirrhosis by Child-Pugh class instead of by health state, we found that health utilities were lower for patients in advanced Child-Pugh classes, more in keeping with expectations. Most utility studies and cost-utility analyses use a single utility value to represent individuals with HCC and a single value for DC. We recommend subcategorizing the HCC and DC health states using additional measures such as Child-Pugh class, Model for End-Stage Liver Disease (MELD) score, and/or cancer stage in future utility studies and cost-utility analyses.

Effects of utility instrument on health utility

We used six different utility measures including direct and indirect utility instruments. These provide different insights into patients’ quality of life and are useful in different contexts. Indirect utility measures, including the EQ-5D, HUI2/3, and SF-6D, reflect societal preferences and are easy to administer. These factors make indirect utility measures the gold standard for cost-utility analyses to guide public health care funding decisions (32,33).

Direct utility measures, including the TTO, are typically more difficult and time-consuming to administer as they ask the respondent to consider complex trade-offs. (The VAS is an exception as it does not present a choice or trade-off.) In studies such as ours where patients are asked to value their own health state, direct utility measures can provide insights into the impacts of diseases from the patient’s perspective based on their lived experiences (34).

In our study, the TTO instrument produced the highest health utilities. This is consistent with previous health utility studies in hepatitis C (22) and other diseases (35), which show that most direct utility measures (with the exception of the VAS) tend to be higher than indirect measures. Patients tend to rate their own health higher than the general population rates an equivalent hypothetical health state. Potential explanations for this observation relate to how patients cope and adapt to their health condition, as well as shift their conceptualization and valuation of health over time as they live with a health condition (36).

The HUI3 instrument generated the lowest utilities. This is consistent with previous literature finding low health utilities using the HUI3 instrument (22), and may be because the HUI3 was designed to be better able to capture impaired health states and has a lower minimum value compared with its previous iteration (37). The HUI3 also showed the greatest difference between the community and hospital subgroups. This may be because it contains more questions covering more aspects of health than the EQ-5D and SF-6D, allowing it to capture a larger number of health states (972,000 possible health states) (37). The differences seen among these indirect utility measures can also be explained by the fact that each instrument conceptualizes the domains of health differently, has a different source of preference data, and has a different model to link questionnaire responses to preference data to generate health utilities (38).

Limitations

This study recruited marginalized CHC patients who are engaged in care with the Toronto Community Hep C Program. It is likely that these patients are experiencing an improvement in health-related quality of life and health utility owing to the care they are receiving in this program. While this population represents a more marginalized group than is typically represented in health utility literature, our utility values should not be interpreted to represent marginalized CHC patients who are not engaged in care.

Although more prevalent in the community setting, the measures of socio-economic marginalization captured by our study were present in many individuals in the hospital setting as well. Indicators of more profound marginalization could have been used to further differentiate the level of socio-economic disparity between these populations. Previous studies at the Toronto Community Hep C Program have reported that many patients in the program have experienced homelessness, incarceration, and trauma (39); but these factors were not captured by our questionnaire. They may have an important effect on health utility and should be explored in future utility studies.

Additionally, as noted earlier, it is difficult for patients with advanced HCC or DC to participate in research studies. We observed that these patients were more likely to decline to participate in this study. We mitigated the potential bias from recruiting ‘healthy’ HCC and DC patients by stratifying utilities by Child-Pugh class. Ideally, larger samples representing utilities across the spectrum of HCC and DC are needed. This will require study designs that are more accommodating of sicker patients.

Conclusion

Both socio-economic marginalization and liver-related morbidity have large negative impacts on health utility in CHC patients. Utility values corresponding to marginalization and advanced liver disease should be incorporated into future cost-utility analyses in order to better represent the population living with CHC and inform health policy.

Funding Statement

This study was supported by the Canadian Institutes of Health Research (CIHR) (Operating Grant #137490 and #PJT-148970). YA Saeed was supported by a Trainee Fellowship from the Canadian Network on Hepatitis C (CanHepC). CanHepC is funded by a joint initiative of the Canadian Institutes of Health Research (#NHC-142832) and the Public Health Agency of Canada (PHAC). These organizations were not involved in the design, conduct, or writing of this study.

Contributions:

Conceptualization, YA Saeed, N Mitsakakis, JJ Feld, MD Krahn, WWL Wong; Data Curation, YA Saeed, K Mason, A Phoon, A Fried, JF Wong; Formal Analysis, YA Saeed, N Mitsakakis, JJ Feld, MD Krahn, WWL Wong; Funding Acquisition, MD Krahn, WWL Wong; Investigation, YA Saeed, K Mason, N Mitsakakis, JJ Feld, KE Bremner, A Phoon, A Fried, JF Wong, J Powis, MD Krahn, WWL Wong; Methodology, YA Saeed, N Mitsakakis; Project Administration, K Mason, JF Wong; Resources, JJ Feld, J Powis; Supervision, MD Krahn, WWL Wong; Validation, YA Saeed, K Mason, KE Bremner, JF Wong; Visualization, YA Saeed, N Mitsakakis; Writing – Original Draft, YA Saeed, N Mitsakakis, JJ Feld, KE Bremner, MD Krahn, WWL Wong; Writing – Review & Editing, YA Saeed, K Mason, N Mitsakakis, JJ Feld, KE Bremner, A Phoon, A Fried, JF Wong, J Powis, MD Krahn, WWL Wong.

Ethics Approval:

This study received approval from the Research Ethics Boards of the University of Toronto and the University Health Network in Toronto, Ontario, Canada.

Informed Consent:

Written informed consent was obtained from all subjects prior to study participation.

Registry and the Registration No. of the Study/Trial:

N/A

Funding:

This study was supported by the Canadian Institutes of Health Research (CIHR) (Operating Grant #137490 and #PJT-148970). YA Saeed was supported by a Trainee Fellowship from the Canadian Network on Hepatitis C (CanHepC). CanHepC is funded by a joint initiative of the Canadian Institutes of Health Research (#NHC-142832) and the Public Health Agency of Canada (PHAC). These organizations were not involved in the design, conduct, or writing of this study.

Disclosures:

WWL Wong and MD Krahn have received research support from the Canadian Liver Foundation; JJ Feld has received research support and/or consulting fees from AbbVie, Antios, Eiger, Enanta, Finch, Gilead, Janssen, and Wako/Fujifilm; J Powis has received a research grant from Gilead Sciences Canada. All other authors have no conflict of interest to declare.

Peer Review:

This manuscript has been peer reviewed.

Animal Studies:

N/A