Hepatitis C Attributable Healthcare Costs and Mortality among Immigrants: A Population-Based Matched Cohort Study

Background Data on the economic burden of chronic hepatitis C (CHC) among immigrants are limited. Our objective was to estimate the CHC-attributable mortality and healthcare costs among immigrants in Ontario, Canada. Methods We conducted a population-based matched cohort study among immigrants diagnosed with CHC between May 31, 2003, and December 31, 2018, using linked health administrative data. Immigrants with CHC (exposed) were matched 1 : 1 to immigrants without CHC (unexposed) using a combination of hard (index date, sex, and age) and propensity-score matching. Net costs (2020 Canadian dollars) collected from the healthcare payer perspective were calculated using a phase-of-care approach and used to estimate long-term costs adjusted for survival. Results We matched 5,575 exposed individuals with unexposed controls, achieving a balanced match. The mean age was 47 years, and 52% was male. On average, 10.5% of exposed and 3.5% of unexposed individuals died 15 years postindex (relative risk = 2.9; 95% confidence interval (CI): 2.6–3.5). The net 30-day costs per person were $88 (95% CI: 55 to 122) for the prediagnosis, $324 (95% CI: 291 to 356) for the initial phase, $1,016 (95% CI: 900 to 1,132) for the late phase, and $975 (95% CI: −25 to 1,974) for the terminal phase. The mean net healthcare cost adjusted for survival at 15 years was $90,448. Conclusions Compared to unexposed immigrants, immigrants infected with CHC have higher mortality rates and greater healthcare costs. These findings will support the planning of HCV elimination efforts among key risk groups in the province.


Introduction
Untreated chronic hepatitis C (CHC) can lead to advanced liver disease and is one of the major causes of liver cancer and liver-related death.Fortunately, highly efective direct-acting antivirals (DAAs) for the hepatitis C virus (HCV) ofer the opportunity to eliminate HCV as a public health concern.Te World Health Organization (WHO) has set a series of targets for HCV elimination by 2030 [1].However, elimination will require considerable planning and investment in expanded screening and treatment initiatives [2].
Individuals living with CHC are highly heterogeneous.In Canada, a disproportionate number of CHC infections (∼30%) occur among foreign-born individuals [3].Compared to nonimmigrants, immigrants can face long delays in diagnosis and can present with more advanced liver disease [4,5].Indeed, the health and economic consequences of delaying linkage to HCV treatment can be substantial, especially among those with serious liver disease [6,7].
Reliable estimates of CHC-attributable healthcare costs are necessary for making appropriate resource allocation decisions among diferent populations and for healthcare planning to meet WHO targets.Several studies have examined CHC costs in Canada [8][9][10]; however, there are no data on the economic burden of CHC among immigrants.Our objective was to estimate CHC-attributable mortality and healthcare costs among immigrants in Ontario, using real-world population-level data.

Methods
2.1.Study Design, Setting, and Population.We conducted a population-based retrospective matched cohort study to estimate CHC-attributable healthcare costs among immigrants in the Canadian province of Ontario (population ∼14.9 million as of 2022) using an incidence-based costing approach.We included individuals with a record of CHC infection diagnosed between May 31, 2003, and December 31, 2018, among individuals who immigrated to Canada after 1985 with a valid Ontario health insurance plan (OHIP) number whose laboratory testing data could be linked to health administrative records held at ICES (formerly the Institute for Clinical Evaluative Sciences).We followed the study cohort up to March 31st, 2020, or until death.Consistent with the previously published study by our group [11], CHC was identifed by a positive test result for HCV antibody or HCV ribonucleic acid (RNA) during the study timeframe, with the exclusion of cases with acute clearance of HCV defned by negative or undetectable HCV RNA test results within 12 months following the frst positive HCV test with no record of Ontario Drug Beneft (ODB) claim for HCV antiviral treatment identifed by drug identifcation number (DIN) listed in S. Table 1.After excluding 158 individuals with missing data on age, sex, or socioeconomic covariates, and those aged >100 years, our study cohort included 6,914 individuals with CHC (hereafter referred to as "exposed") who met eligibility.For the exposed individuals, the index date was the frst date of HCV diagnosis.
Individuals without CHC (hereafter referred to as "unexposed") were Ontario immigrants with no or negative HCV antibody or HCV RNA testing records who were alive at the start of the study period on May 31, 2003.We selected unexposed individuals by randomly sampling 5% of the Ontario population using the Registered Persons Database (RPDP), a population-based registry of all Ontario residents eligible for provincial health insurance.After excluding individuals with missing covariate data and those aged >100 years, 120,693 unexposed individuals who met the eligibility criteria were included.For unexposed individuals, we randomly assigned an index date based on the index date distribution of individuals with CHC, following their landing date in Ontario and between May 31, 2003, and December 31, 2018.

Data Sources.
We obtained information on HCV exposure using HCV antibody and RNA testing records in the Public Health Ontario (PHO) laboratory dataset and linked them to health administrative, demographic, and immigration data held at the ICES to identify immigrants living with HCV and to collect demographic and healthcare utilization data.We accessed information on immigration status using the Immigration, Refugees, and Citizenship Canada (IRCC) Permanent Resident Database, which holds data on individuals who have been granted permanent resident status in Canada since 1985.We accessed demographic information on birth year, sex, rurality, and neighborhood income quintile from the RPDB.Markers of social marginalization, such as residential instability quintiles, material deprivation quintiles, and ethnic concentration quintiles, were obtained from Ontario's Marginalization Index Database (ONMARG).Information on human immunodefciency virus (HIV) status, cirrhosis, decompensated cirrhosis (DC), hepatocellular carcinoma (HCC), liver transplant, death, substance use disorders related to drug and/or alcohol use, and comorbidities (Aggregated Diagnosis Groups (ADG) comorbidity classifcation scheme) were collected from the Canadian Institute for Health Information's Discharge Abstract Database (DAD) and National Ambulatory Care Reporting System (NACRS), OHIP, Ontario Mental Health Reporting System (OMHRS), Ontario Cancer Registry (OCR), and Ofce of the Registrar General Death registry (ORGD) datasets, using the diagnostic, procedure-related, and death codes listed in S. Table 2-3.We used the Johns Hopkins Adjusted Clinical Group ® (ACG ® ) system to determine the total number of ADGs over the year preceding the index date [12].We compiled all baseline characteristics at the index date.All records were linked using unique identifers and analyzed at ICES.Healthcare utilization and the corresponding costs were retrieved from the above-mentioned administrative databases and are described in more detail in the"Outcomes" section.

Phase-of-Care
Approach.We used a phase-based approach to estimate CHC-attributable costs from diagnosis to death [8,13,14].To do so, we allocated the total observation time of each individual to phases that represent the natural history of disease consistent with previous CHC costing studies [8]: (1) prediagnosis phase, defned as 6 months prior to index date, which capture costs associated with diagnostic testing and work-up to establish HCV diagnosis; (2) initial care phase, covers the time between CHC diagnosis up to start of either the late phase or the terminal phase or until the end of follow-up for those remaining alive throughout follow-up and without advanced liver disease; this phase captures the cost of active treatment or monitoring following initial diagnosis; (3) late phase starts at 3 months prior to the frst diagnosis of advanced liver disease(decompensated cirrhosis, hepatocellular carcinoma, and liver transplantation) and captures ongoing surveillance for CHC-related complications and lasts up to the start of the terminal phase for those who died or up to end of follow-up for those remaining alive, and (4) terminal phase, representing the last 6 months prior to death for individuals who have died and captures the care received during the endof-life period.Te length of each phase was based on a previous study focused on CHC [8] and was validated by a clinical expert (CG).Te starting points of each phase were indicative of clinically meaningful shifts in care patterns and costs.

2
Canadian Journal of Gastroenterology and Hepatology 2.4.Matched Cohort.We matched each immigrant with CHC to one immigrant without CHC using nearestneighbor matching without replacement.Matching was performed using a combination of hard matching (sex, birth year, index year, presence of alcohol-or drug-related substance use disorder at index, and overall morbidity burden categories (within one year prior to index as measured by ADGs)) and propensity-score matching using a caliper width equal to 0.2 standard deviations of the logit of the propensity score [15].Te propensity algorithm included covariates collected at indexes such as age at diagnosis, neighborhood income quintile, residential instability quintile, material deprivation quintile, ethnic concentration quintile, dependency quintile, rurality, HIV coinfection, landing year, years of schooling at landing, and the total number of comorbidities within the one year prior to index diagnosis based on the total number of ADGs.Exposed individuals were rematched to unexposed controls to estimate the phase-specifc healthcare costs.For the late phase, an index date of three months prior to the date of advanced liver disease diagnosis was selected for exposed individuals, whereas for unexposed individuals, it was randomly assigned.Te variables used for hard and propensityscore matching were the same as above.To estimate the terminal phase cost, exposed individuals who died during the study were rematched to unexposed individuals who also died during the study period.Te index date was the date of death.Hard-match variables included index year, sex, and birth year.For unexposed individuals, the birth year was set to be earlier than or equal to that of the exposed individuals.Te variables used for propensity-score matching were the same as described above, except for age at the time of diagnosis, which was replaced by age at death.

2.5.
Outcomes.We estimated the cumulative incidence of all-cause mortality at 1 year, 5 years, 10 years, and 15 years postindex.Direct healthcare costs adjusted to 2020 Canadian dollars were estimated from the provincial public payer perspective.We used the ICES costing algorithm to obtain healthcare costs (standardized to 30 days)of the study cohort from the index date to the end of follow-up or death [16].Phase-specifc costs were calculated from the beginning of each phase to the start of the subsequent relevant phase or end of follow-up.Next, we combined phase-specifc costs with survival data to estimate the long-term costs.

Statistical Analyses.
We assessed the quality of the match using standardized diferences (SD), with SD less than 0.1 indicating negligible diferences between covariates.We estimated CHC-attributable all-cause mortality by calculating the absolute risk reduction (ARR) and relative risks (RR) using conditional Poisson regression with robust variance estimators.
We calculated the attributable (net) costs by determining the mean diference between the total healthcare costs for the matched exposed and unexposed individuals using regression analysis.Te 95% confdence intervals (CIs) were estimated using the standard errors of the regression coefcients.To estimate mean 1-year, 5-year, 10-year, and 15year net costs, we combined phase-specifc costs with crude survival data using the Yabrof equation (17).Te margins of the 95% CIs of phase-specifc costs were used to estimate the plausible range of mean net costs.Costs beyond one year were discounted by 1.5% annually [18].Additionally, we stratifed the results by age and sex.SAS Enterprise Guide 7.15 (SAS Institute, Inc.) was used for all statistical analyses.

Study Cohort.
Between May 2003 and December 2018, we identifed 6,914 Ontario immigrants with CHC who met eligibility, and we were able to match 81% (N � 5,575) of immigrants with CHC to immigrants without CHC (Table 1).Compared to unexposed individuals, exposed individuals were older, had slightly earlier landing dates, were more likely to have markers of social marginalization (e.g., lower income and higher material deprivation levels), and were more likely to have drug-or alcohol-related substance use disorders and other comorbidities prior to matching.Following matching, we were able to achieve a balance in all characteristics with standardized diferences of less than 0.1, except for the landing year (standardized diference: 0.38).On average, matched individuals were aged 46 years at the index date and had received approximately 11 years of education at the time of landing, 44% were baby boomers(born between 1945 and 1963), 52% were male, 99% lived in urban settings, ∼1% experienced drug-or alcohol-related substance use, <1% were infected with HIV, ∼35% experienced the highest levels of material deprivation, residential instability, and lowest income levels, and 59% experienced a low comorbidity burden (0-3 ADGs) 1 year prior to the index date.Unmatched immigrants with CHC were older at the time of diagnosis and had a higher comorbidity burden (>8 ADGs) and a higher proportion of drug-or alcohol-related substance use than matched individuals.
Te demographic characteristics of the study cohort in the late and terminal phases, before and after matching, are shown in S. Table 4 and S. Table 5.In total, we identifed 1,582 and 1,070 exposed individuals in the late and terminal phases, respectively.We were able to match 97% (N � 1,535) and ∼71% (N � 756) of the exposed individuals in the late and terminal phases, respectively, to achieve balanced matches.

Total and CHC-Attributable Healthcare Costs by Phase of
Care.Mean (median) lengths of initial and late phases were 8.5 (8.8) and 7.1 (6.7) years for matched exposed individuals and 7.4 (5.3) and 7.1 (7.7) years for unexposed individuals, respectively.Te mean lengths of the prediagnostic and terminal phases were predefned, with a duration of 180 days each.
For the prediagnostic phase, the total 30-day healthcare costs were $247 per person for exposed and $159 per person for unexposed individuals, while CHC-attributable healthcare costs were $88 (95% CI: $55 to $122) (Table 3).In the prediagnostic phase, physician and laboratory services accounted for the main diferences in health care costs for exposed compared to unexposed individuals.For the initial care phase, the total 30-day healthcare costs were $510 per person for exposed and $187 for unexposed individuals, translating to CHC-attributable costs of $324 (95% CI: $291 to $356) per person.In this phase, outpatient prescriptions accounted for important diferences in costs between matched pairs.For the late phase, the total 30-day healthcare costs were $1,290 per person for exposed and $274 per person for unexposed individuals, translating to CHCattributable costs of $1,016(95% CI: $901 to $1,132) per person.Diferences in outpatient prescriptions followed by inpatient care and physician services accounted for the important diferences in cost between matched pairs in this phase.For the terminal phase, 30-day healthcare costs were $8,967 per person for exposed and $7,992 for unexposed individuals, translating to CHC-attributable costs of $975 (95% CI: −25 to 1,974) per person, with main diferences attributable to inpatient care.
Net 30-day costs were slightly lower for females than for males in the prediagnosis and initial phases but higher in the late ($1,092 vs $952) and terminal phases ($1,752 vs. $392), primarily because of higher net costs for inpatient care (Figure 1, S. Table 6).In the terminal phase, net 30-day costs were also higher for those born before 1945 than for those born during 1945-1965 or after 1965 (-$1,617 vs. $406 vs. $1,261, respectively) (Figure 2, S. Table 6).

Discussion
We conducted a population-level propensity-score-matched cohort study of incident cases of CHC among Ontario immigrants diagnosed between May 31, 2003, and December 31, 2018, to estimate CHC-attributable mortality, as well as phasespecifc and long-term net healthcare costs.Our fndings suggest that, compared to unexposed individuals, immigrants with CHC experience a signifcantly higher mortality rate and incur substantial healthcare costs that vary by phase of care and time horizon of analysis; the highest CHC-attributable costs were observed for the late phase.During the initial and late phases, outpatient prescriptions were the largest cost component, whereas, in the terminal phase, the costs were primarily driven by inpatient care.Te average annual cost for immigrants with CHC was $6,120 during the initial phase and rose to $15,480 in the late phase, three times higher than Ontario's per capita healthcare spending of $5,400 in 2020 [19].Te substantial increase in costs highlights the role of early diagnosis and treatment in preventing CHC sequelae and reducing healthcare expenditure.
Our fndings are generally consistent with those of previous HCV costing studies in Ontario, Canada.For instance, prior Ontario-based costing studies among the First Nations also found similar efects, with the largest burden of CHCattributable healthcare costs in the late phase of care.Te reported 30-day net costs (2018 CAD) incurred during the initial and terminal phases were consistent with our estimates but were found to be higher in the late phase of care ($1,768 ($1,814 in 2020 CAD) vs. $1,016) [8].Another populationlevel costing study in Ontario by Wong et al. estimated CHCrelated costs by disease stage but did not report net costs.Tey found the mean 30-day costs (2018 CAD) to vary by disease severity from $798 ($819 in 2020 CAD) for individuals with  no cirrhosis to $1,487 ($1,526 in 2020 CAD) for those with compensated cirrhosis and to $3,659 ($3,755 in 2020 CAD) for decompensated cirrhosis [10].Te higher costs observed in advanced disease stages in those two studies could be attributed to a higher level of comorbidity and drug-or alcohol-related substance use disorder in the study populations compared to our study.Tese factors were found to be important predictors of CHC costs [10].Although we did not evaluate the diferences in costs between immigrants and non-immigrants in the current study, Wong et al. found signifcantly lower healthcare costs among immigrants experiencing non-advanced and advanced liver disease, but higher costs during the terminal care phase compared to nonimmigrants [10].
In the present study, the 15-year net costs amounted to $102,881 CAD (undiscounted) and were higher for females and those born after 1965, possibly due to their longer life expectancy.Previous analyses have demonstrated an association between life expectancy and healthcare costs in the CHC population [20].Notable, our long-term costs were ∼30% higher than the estimated lifetime costs of $64,694 ($70,654 in 2020 CAD) reported in a Canadian study of individuals with CHC in 2013 [21].Tese diferences may be attributed to variations in costing methods and the use of novel treatments.In Ontario, the cost of publicly funded antivirals is estimated to be over $50,000 per completed treatment course [22].
Our study has several limitations.First, given that the IRCC dataset starts in 1985, individuals identifed as immigrants consist of only those who immigrated to Canada after 1985; thus, results may not be generalizable to earlier immigrants [23].Second, although we were able to match a large proportion of HCV-exposed immigrants in Ontario, in general, the matched population tended to have lower levels of drug-or alcohol-related substance use disorder (1% vs. 36%) and lower comorbidity level (ADG >8 : 5.8% vs. 49%) compared to the unmatched population.Tus, our fndings are less generalizable to immigrants with multiple comorbidities or a history of substance use disorders at the time of diagnosis.Indeed, in previous phase-based costing studies using similar methods among populations with a much higher proportion of substance use disorders, prediagnosis costs were found to be higher than in our study [8].Tird, we did not conduct stratifed analysis by DAA era.However, it is important to note that the prior populationlevel costing study by Wong et  8 Canadian Journal of Gastroenterology and Hepatology individuals with fbrosis and compensated cirrhosis [10].In contrast, our analysis revealed even higher percentages, with outpatient costs accounting for 65% and 48% of the total CHC-attributable costs during the initial (fbrosis, compensated cirrhosis) and late phases (decompensated cirrhosis, HCC) respectively.Tis notable diference suggests that a signifcant portion of treatment in our study may have occurred during the DAA era, given the high proportion of outpatient costs.Nonetheless, it is reasonable to anticipate that stratifed analysis by DAA era would likely reveal an even greater proportion of outpatient costs in the DAA era.Lastly, as the current study is from the provincial healthcare payer perspective, the analysis does not include drug costs covered by private drug plans.
Our study has important strengths.Using populationlevel health administrative datasets and propensity-scorematching controls provided the frst real-world estimate of CHC-attributable costs among immigrants in Ontario, Canada.We were also able to explore all important resource components, including inpatient and outpatient visits, emergency department visits, prescribed medications, laboratory and diagnostic services, and other services that account for CHC-attributable costs.Furthermore, using the phase-of-care approach, we estimated the cumulative longterm costs adjusted for survival.
Te existing literature on costing in Ontario is limited, especially on diverse populations.To our knowledge, this is the frst costing study to estimate CHC-attributable costs among immigrants in Ontario, Canada.Future studies should also report the results among diferent risk groups to gain a better understanding of the target population for whom interventions should be prioritized.
In conclusion, this study provides information on the economic burden of CHC among key risk groups in Ontario to help support provincial policymaking.Efective implementation of programs facilitating early diagnosis, Canadian Journal of Gastroenterology and Hepatology treatment, and continuity of care among at-risk immigrant populations can signifcantly reduce the long-term health consequences and associated costs of untreated CHC, thereby alleviating the economic burden on the healthcare system.Our fndings on CHC-attributable mortality and healthcare costs are important for planning and guiding economic evaluations of various HCV elimination eforts and public health interventions among Ontario immigrants living with hepatitis C.

Table 1 :
Characteristics of the study cohort.
*Te index year is the year of CHC diagnosis.ADG: aggregated diagnostic groups; CHC: chronic hepatitis C; HIV: human immunodefciency virus; N: number of observations; sd: standard deviation; SMD: standardized diference in means.

Table 2 :
Total and CHC-attributable mortality among immigrants by phase of care.

Table 3 :
Total and CHC-attributable healthcare costs (in 2020 CAD) per 30 days among immigrants by phase of care.
al. demonstrated that outpatient drug costs constituted a substantial portion, ranging from 27% to 40% of total costs in the DAA era for Figure1: Net healthcare costs (in 2020 CAD) by phase of care for males and females.Note.Negative attributable cost observed, e.g., in the terminal phase for "other services," often occurs because exposed individuals are hospitalized and do not incur costs outside the hospital, but unexposed do.
Figure2: Net healthcare costs (in 2020 CAD) by phase of care by birth cohorts.Note.Negative attributable cost observed, e.g., in the terminal phase for "other services," often occurs because exposed individuals are hospitalized and do not incur costs outside the hospital, but unexposed do.