Cost-effectiveness of mass screening for Hepatitis C virus among all inmates in an Irish prison

Abstract

Background:

In Irish prisons, there is a high proportion of people who inject drugs (PWID; 26%) and a high prevalence of HCV (16%), making prison a high priority setting for HCV testing and treatment. We evaluate the cost-effectiveness of a mass HCV screening intervention in Mountjoy Prison, Dublin, compared to the standard-of-care of intermittent screening on committal.

Methods:

Primary cost data was collected from the intervention using an overall provider perspective. Standard-of-care (SOC) costs were estimated through interview. All costs were inflated to 2020 Euros. An HCV transmission and disease progression model among incarcerated and community PWID and ex-injectors was calibrated to the Dublin HCV epidemic, allowing inclusion of population-level health benefits. The model used intervention data, suggesting 419 individuals were screened, 50 HCV infections diagnosed and 32 individuals initiated treatment, to project the resulting costs and health benefits (quality adjusted life years or QALYs) over 50 years with 5% discounting. The incremental cost effectiveness ratio (ICER), cost per QALY gained, was estimated for the screening intervention compared to the standard-of-care. Probabilistic sensitivity analyses (PSA) determined the probability that the intervention was cost-effective compared to a willingness-to-pay threshold of €30,000/QALY as used in Ireland. The ICER for 1- or 3-yearly mass screening in all Dublin prisons was also calculated.

Results:

The total direct costs of the intervention (not including treatment drug costs) was €82,392, with most costs being due to staff (43%) and overhead or management costs (38%). Despite having little epidemiological impact due to the small numbers treated, over 50 years the incremental cost of the intervention was €36,592 and 3.8 QALYs were gained, giving a mean ICER of €9,552/QALY. The majority (84%) of PSA runs were below the willingness-to-pay threshold. Yearly mass screening had an ICER of €2,729/QALY compared to SOC and gave a higher net monetary benefit (€7,393,382) than screening every 3 years (€6,252,816).

Conclusion:

Prison mass screening could be a cost-effective initiative for increasing testing and treatment of HCV in Ireland.

Introduction

Hepatitis C virus (HCV) is a blood borne virus that causes significant morbidity worldwide. Since the development of highly curative direct acting antiviral (DAA) drugs for HCV in 2015, the World Health Organisation has developed a Global Health Strategy to eliminate HCV as a public health threat[1]. This has led to many countries developing national elimination strategies and scaling up testing and treatment[2,3].

As in other high-income settings[4], the majority of new HCV infections in Ireland are acquired through injecting drug use (IDU)[5]. Because of the criminalisation of IDU, a high proportion of people who inject drugs (PWID) globally have ever been imprisoned (58%)[6], with incarceration being linked to heightened overdose risk[7]and elevated risk of blood borne virus transmission[8]. In Ireland, 70% of PWID have ever been incarcerated while 26% of prisoners report injecting drug use[9]. Because of the high prevalence of PWID in prison, a high proportion of those incarcerated in Ireland are antibody-positive for HCV (16% in 2011 survey from Dublin[9]), with this being higher (40%) among prisoners with a history of injecting drug use[9]. This high prevalence of HCV highlights the importance of undertaking HCV screening and treatment interventions in Irish prisons as part of their strategy to eliminate HCV[10].

Currently, HCV screening in Irish prisons occurs intermittently on prison entry with treatment occurring on site since 2018[11]. To help develop strategies for improving the coverage of treatment in prisons, the HepCare Europe program ran a mass HCV screening and linkage to care intervention at Mountjoy Prison in Dublin in 2017. This study evaluates the cost-effectiveness of this intervention compared to the standard-of-care (SOC) comparator which was low coverage of screening on committal and screening in the community.

Methods

We used a dynamic HCV transmission and disease progression model among PWID (in and out of prison) with primary cost and outcome data to determine the cost-effectiveness of the HepCare mass screening and linkage to care intervention in Mountjoy Prison (Dublin, Ireland) compared to the current SOC. An overall payers’ perspective was used for costs (and includes costs incurred by the prison) with health benefits being tracked in the full population. A time horizon of 50 years with a 5% per year discount rate was used to incorporate the long-term health and prevention benefits of successful HCV treatment. A willingness-to-pay threshold of €30,000 per QALY was used to determine cost-effectiveness as per Irish guidelines[12]. In addition, the model projected the impact of the intervention on the prevalence and incidence of HCV and the number of new infections averted.

Model description

The ordinary differential equation HCV transmission model used in this analysis included stratification according to incarceration status (never in prison, current imprisonment, and recently or ever incarcerated but not currently) and injecting status (current and ever, figure 1a), with the model assuming only people with an injecting history have HCV. The model also included the cascade of care (figure 1b) and HCV disease progression (figure 1c). Model equations are in the supplementary information (table S1–7).

Figure 1:

Schematic of different components of the model, including injecting and incarceration stratifications (1a); Epidemiology and cascade of care (1b); and HCV disease progression (1c)

An individual enters the model on initiation of injecting as susceptible to infection, and enters into any of the community injecting categories (never incarcerated, recently incarcerated or ever incarcerated). Individuals can become incarcerated at specific rates with re-incarceration occurring at a higher rate than initial incarceration. Individuals remain incarcerated for a fixed duration.

Current PWID become infected at rates dependent on the prevalence of HCV in their setting (community or prison) with the model allowing differences in the risk of infection among currently and recently incarcerated PWID (compared to other community PWID). New infections can either spontaneously clear and move to the exposed category or develop chronic infection, leading to liver disease progression over time. Chronically infected individuals can be diagnosed, and then either enter the treatment pathway or are lost to follow-up. Engagement in the treatment pathway includes all appointments between diagnosis and initiating treatment (e.g. for fibroscan, liver function tests). Only those engaged in the treatment pathway can start treatment. Upon successful treatment individuals enter a susceptible SVR state with no or reduced disease progression[13,14]. Re-infection can occur if they have not ceased injecting. Those who fail treatment can be retreated.

Upon cessation of injecting, ex-injectors can still become incarcerated at a lower rate. PWID experience drug-related mortality, whereas ex-injectors experience mortality according to their life expectancy. Individuals progressing beyond compensated cirrhosis also experience mortality due to liver disease.

Model parameterisation and calibration

The population modelled includes Dublin based PWID (and ex-injectors) and their incarceration dynamics. The first step was to parameterise and calibrate the model for the SOC comparator in 2014, and following that, incorporate the effects of the intervention. Data for the model came from various sources as detailed in Table 1. The following model components were parameterised and calibrated using a multi-step process:

Table 1:

Calibration data for the mathematical model

HCV Transmission Data	Value	Source
Chronic HCV Prevalence in current injectors	51% (45–59%)	Adjust 69% (60–78%) seroprevalence (meta-analysis on data reviewed in [5], see supplementary materials figure S2) assuming 75% have spontaneously cleared their infection [42]. Used in calibration step 2.
Testing Data
Percentage of current injectors diagnosed in standard-of care pathway	Mean 63% (55–71%)	[16]HepLink enrolment data. Used in calibration step 2.
Engagement and Treatment Data
Referral rate from diagnosis for standard-of-care pathway in all compartments	Mean 1.21 (0.94–1.56)	70% per year (61–79%) (data from HepLink study [16]). Assume this percentage referral is all within the last year to obtain engagement rate [16]
Demographic Data
Population size of Injectors	7,000–15,000	Assumption based on a wide range around the number of people on OST in Dublin in 2016 (9,844) allowing for those not in drug treatment [21](Health Service Executive data)
Population size in prison (current and ever injectors)	550–650	Weighted average (606) of proportion of prisoners who reported a history of injecting in each of the Dublin prisons using data from prison survey [9] and Irish Prison Service inmate numbers. See supplementary information for the calculation.
Proportion of community injectors that have never been incarcerated	0.32 (0.21–0.43)	[16]HepLink study data (used in calibration step 1)
Proportion of those PWID in prison who have been incarcerated twice or more	0.80 (0.73–0.87)	[15]HepCheck study data (used in calibration step 1)
Length of incarceration episode for PWID	0.25–2.75 years	[15]HepCheck study data gave average length of incarceration episodes being 2 years, we assumed a wide interval around this.
Proportion of incarcerated individuals who were current injectors when last in the community	0.40 (0.33–0.46)	2011 Prison survey [9]. This is used to estimate the relative reduction in incarceration rate for ex-injectors (calibration step 2)
Proportion of incarcerated individuals that began injecting after first incarceration episode	0.69 (0.59–0.79)	[15]HepCheck study data - this is used in calibration step 1.

1. Demographic and incarceration component (figure S1),

2. HCV infection and diagnosis component (figure 1b),

3. Disease progression and chronic HCV population size component (figure 1c).

At each step, some parameters were randomly sampled from prior ranges, while non-linear least squares fitting was used to estimate unknown parameters that fitted the model to specific calibration data.

The unknown rates of incarceration, re-incarceration and proportion of new injectors who have previously been incarcerated were estimated through calibrating an incarceration sub-model. Assuming the incarceration and IDU dynamics were stable (with a stable PWID population size), this sub-model was calibrated to data on the proportion of PWID who have never been incarcerated (21–43%)[14][15], proportion of incarcerated PWID who have been incarcerated twice or more (70–90%[15]), and proportion of incarcerated PWID who started injecting after being in prison (60–80%)[15]. Runs were rejected if they projected over 4.3% of current injectors are in prison (estimated through prison survey[15]). Other demographic parameters for this sub-model, including the injecting cessation rate, mortality rates and average prison stay came from the literature (Table 2), including uncertainty ranges. This model calibration was repeated for 20,000 random draws of these parameters and the calibration data. This produced 3010 model fits, with the other parameter sets being rejected because they could not fit the proportion of incarcerated PWID who started injecting after being in prison, or projected too high a percentage of current injectors in prison.

Table 2:

Parameters used in mathematical model

HCV Transmission Parameters	Value	Source
Risk ratio for HCV transmission and acquisition when recently incarcerated	1.62 (1.28–2.05)	[8]
Testing Parameters
Testing rate in all compartments for standard-of-care pathway (posterior)	Mean 0.21 (0.11–0.34)	Estimated in calibration step 2 to give proportion diagnosed (63%). Assume proportion of injectors diagnosed is stable.
Engagement and Treatment Parameters for standard-of-care
Referral rate from diagnosis for standard-of-care pathway in all compartments	Mean 1.21 (0.94–1.56)	Assume percentage referral 70% (61–79%) is all within the last year to obtain engagement rate (data from HepLink study [16])
Treatment rate from engaged per year for standard-of-care pathway in ex-injector community compartments prior to 2015	Mean 0.24 (0.15–0.37)	39% (26–53%) treated of those engaged. Convert this percentage to a rate (HepLink data [16]) assuming these treatments occur within the last 2 years (current PWID not treated until after 2014)
Treatment numbers in Dublin with number done in Mountjoy prison in brackets 2015 2016 2017 2018	92 (+5 prison) 438 (+7 prison) 749 (+9 prison) 1288 (+20 prison)	HCV infection in Irish drug users and prisoners scoping review [5]
Percentage of treatments that attain Sustained virological response	93% (88%−97%)	SIMPLIFY trial among PWID with recent injection drug use (last 6 months) [43]
Rate at which individuals complete treatment per year	1/(12/52) community 1/(8/52) prison	Treatment assumed to last 12 weeks in community and 8 weeks in prison (to maximise number treated as per protocol)
Demographic Parameters
Cessation rate of injectors	0.088 (0.053–0.125) per year	Assume mean duration 11 (8–15) years as in UK[44].
Rate of initiation of injecting	Used to fit population size estimates in prison	Assume stable population size
Mortality rate among active PWID	Poisson distribution, 73 (divided by 10,000) per year	Meta-analysis of mortality in homeless individuals, prisoners and individuals with substance use disorders [44]
Ex-injector mortality rate	0.023 (0017–0.032) per year	1/(life expectancy – age at initiation of injecting – injecting duration)
Incarceration rate among active PWID (posterior)	Mean 0.03 (0.008–0.12) per year	Estimated through model calibration in step 1
Re-incarceration rate among active PWID (posterior)	Mean 0.06 (0.01–0.29) per year	Estimated through model calibration in step 1
Proportion of new injectors that have been previously incarcerated (posterior)	Mean 0.61 (0.42–0.80)	Estimated through model calibration in step 1
Factor difference in incarceration rates for ex-injectors (posterior)	Mean 0.43 (0.23–0.78)	Estimated through model calibration in step 2

Posteriors are the ranges of the parameters in the final accepted parameter set of 746 fits.

Using the parameter sets for these 3,010 model fits, a second sub-model (including HCV transmission and cascade of care) was then parameterised and calibrated to a stable chronic HCV prevalence of 51% (45–59%) among community PWID in Dublin[5], as well as a stable population of PWID. This calibration produced estimates for the unknown HCV transmission risk among never incarcerated PWID, while assuming HCV transmission risk is elevated among recently released PWID[8] and negligible in prison, based on no new infections being identified in HCV retesting (99 individuals retested after 2 years) undertaken in Mountjoy prison in 2019 (Personal communication Dr Desmond Crowley). The baseline modelled HCV testing rate was also estimated, such that the model projected 63% (55–71%) of chronically infected PWID were currently diagnosed[16]. We assumed 70% (61–79%) of diagnosed individuals were engaged with hospital care, of which 39% (26–53%) commenced treatment within 2 years[16]. Due to little data, testing and engagement rates were assumed same regardless of injecting and incarceration status, but only non-incarcerated ex-injectors were treated before 2015[17]. All 3,010 model runs could be fit to the HCV prevalence and diagnosis data.

These 3,010 parameter sets were then used in the full model (including disease progression and ex-injectors) and the initiation rate of new injectors and the factor reduction in the incarceration rate in ex-injectors were varied to calibrate the model to the estimated number of people who have ever injected in the Dublin prison population (sampled between 550–650)[9]. Parameter estimates for HCV disease progression rates were taken from the literature[13,14,18,19], including among PWID[19], with some disease progression rates being reduced post SVR[13,14,20] (table S8). Only those model runs which then gave a Dublin PWID population of 7,000–15,000[21] and 9,000–21,000 chronic HCV infections[22] were accepted. This resulted in 746 baseline model fits, which were used as the initial conditions for the model in 2014.

Standard of Care scenario

From 2015, current injectors started being treated[17]. Treatment numbers from Ireland were used assuming that 85% of treatments occurred in Dublin[5] and 80% were among current or ex-injectors split 40%/60% respectively (Table 3). Mountjoy prison treatment numbers were used assuming that all treatments prior to the intervention (up to 2017) happened in both the SOC comparator and intervention scenarios. From 2018 onwards, it was assumed that 20 treatments per year were possible in Mountjoy prison in the SOC scenario.

Table 3:

Cost of HCV care cascade in standard-of-care comparator arm

Cascade step	Cost in Euros (to nearest Euro)	Source
Diagnosis in community
Ab−	51	HepLink nurse liaison in primary care costing analysis assuming two visits for testing (one for Ab test and one for RNA if Ab is positive) [24]
Ab+ RNA−	184
Ab+ RNA+	223
Previous SVR	95
Diagnosis in prison	As for community but €5 less	Cost of building space replaced by prison staff cost which is €5 less for each testing scenario
Engagement with treatment pathway in community	178	HepLink costing analysis [24]
Engagement with treatment pathway in prison	206	Based on pre-treatment appointments in prison and includes specialist nurse and prison staff costs (see supplementary information)
Treatment monitoring in hospital (community categories)	393	Based on 12-week treatment protocol at Mater Hospital including monitoring tests and staff costs for clinic visits (see supplementary information)
Treatment monitoring in prison	486	Based on 8-week treatment protocol used in Mountjoy prison and includes prison staff costs (see supplementary information)
Treatment cost	€15,180 in community €10,120 in prison	Treatment in prison is cheaper as it is 8 weeks rather than 12

HepCheck intervention scenario

The intervention screened 419 individuals within Mountjoy prison over six days in May 2017 (76% of prison population) using phlebotomy testing for HCV antibody and reflex RNA testing for those individuals antibody positive[11]. Fifty chronically infected individuals were identified, 30 being newly identified, with results being given to prisoners over 3 days (figure 2). This allowed for fibrosis staging (fibroscan) of 40 infected individuals. Of 50 chronically infected patients, 86% (43/50) were linked to the hepatology clinic and 32 (74% of linked) were treated over 18 months within the prison outreach programme run by a specialist nurse (2 in 2017 and 30 in 2018). Of those not linked to hepatology clinics, 71% (5/7) were linked to opioid substitution therapy (OST) services in the community. Only the 12 treatments among newly diagnosed individuals were assumed to be associated with the intervention, which is a conservative estimate based on the number of treatments in years prior to the intervention (2 treatments in 2017 and 10 in 2018). The remaining 20 treatments in 2018 were associated with the SOC (previously diagnosed individuals). The intervention scenario simulated these treatment numbers with treatment rates from 2019 reducing to the SOC rate for 2018 (20 per year). The SOC and intervention scenarios therefore differ from 2017 onwards.

Figure 2:

Flow diagram of intervention outcomes

Costing

The costing analysis was undertaken with costs presented in 2020 Euros. Healthcare costs relating to HCV disease stages came from a recent Irish economic analysis[23] (Table S9). In Ireland there is a free at point of use health service for secondary care and a medical card scheme for those on low income for primary care. Diagnostic tests, such as for HCV, are paid for by the health service. Prison healthcare costs are paid for by the Irish Prison Service while HCV treatment is paid for by the health service.

Community Based Screening and Treatment Costs

Community screening costs (same for SOC and intervention) were obtained through interview with three general practitioners and HCV nurse specialist who undertake screening of patients accessing opioid substitution therapy in primary care as part of a previously published HCV case finding intervention (details in[24]). These include phlebotomy for HCV Ab and RNA tests and a specialist nurse appointment to determine whether referral to hospital is necessary. Costs relating to DAA treatment at hospital (for those in the community) were based on the treatment protocol in a Dublin hospital (12-week treatment regime see supplementary information).

Prison Based Screening Costs

Cost data for the testing and linkage to care intervention was gathered via University College Dublin (UCD) financial records and interviews with UCD and prison staff involved in the intervention. Intervention costs were collected and classified as capital or recurrent. Costs were allocated to set-up and management, screening or linkage to care. A bottom-up micro-costing approach was used to determine the cost of a screening day, a results day and managerial meetings at the prison relating to the intervention, first estimating resource use and then attaching cost values[25]. A top-down approach was used for estimating set-up and implementation costs incurred by UCD, with research costs excluded (see supplementary information and tables S11 and S12).

Screening costs in prison for the SOC comparator were assumed to be comparable to the community where the cost of a consulting room is replaced by the cost of a prison officer escort (€5 less).

Prison Based Treatment Costs

Prison-based treatment (same for SOC and intervention) was costed through interviews with the specialist nurse who provides HCV clinics in the prison, including staff costs for the nurse, consultant, pharmacist and prison officers who escort patients to appointments. There are generally seven treatment-associated appointments with the specialist nurse, two are pre-treatment visits, three are associated with the 8-week treatment course, an end of treatment visit and lastly the SVR24 visit (Table 1 and supplementary information tables S13–14). Prison treatment is shorter than community treatment (8 weeks rather than 12) to allow more people to complete treatment prior to release.

Utility values

Utility estimates for different HCV disease stages came from the literature[26]. PWID that cease injecting were assumed to have a higher uninfected health utility than active PWID (mean 0.85 versus 0.73)[27]. For each model state, we applied the minimum of the relevant PWID status utility or their disease progression stage health utility (table S10). Health utilities were sampled for each of the 746 model fits and attached to each model state[12]. Health utilities were summed for the population over the time horizon (discounted at 5% per year[12]) to give quality adjusted life years (QALYs).

Cost-effectiveness analyses

The mean incremental cost effectiveness ratio (ICER) was calculated as the mean difference in costs divided by the mean difference in QALYs between the HepCheck intervention and the SOC comparator. Cost-effectiveness was determined using the willingness-to-pay threshold of €30,000 per QALY gained for Ireland[12]. Net monetary benefit was calculated as the mean incremental QALYs multiplied by the willingness-to-pay threshold minus the mean incremental costs. The number of infections and disease related deaths averted were calculated over the same time period.

Sensitivity Analysis

The variation in model projections across the 746 model fits was used to undertake a probabilistic sensitivity analysis (PSA), so producing a joint distribution of the incremental cost and QALYs for the intervention. The impact of parameter uncertainty on the incremental costs, QALYs and net monetary benefit was assessed using an analysis of covariance (ANCOVA)[28].

Matched univariate sensitivity analyses examined the effect of the following changes on the ICER:

1. Varying the time horizon to 25 years (compared to 50 years) and discount rate to 0% or 3.5% (compared to 5%)

2. Halving the cost of HCV drugs to €7,500 per 12-week course

3. Multiplying health utilities when combining them instead of using the minimum utility[29]

4. Standard-of-care testing and engagement rates halved in prison (instead of assuming same as in community) or halved in current PWID (in prison and community) compared to ex-injectors.

5. Fifteen treatments allotted to SOC in 2018 (instead of 20) so 15 treatments allotted to intervention (instead of 10 in base case).

6. Treatment length in prison set to 12 weeks (instead of 8 weeks in base case).

Mass Screening Scenario Analysis

We also modelled the potential impact of scaling up the mass screening intervention to all five prisons in Dublin, and repeating it every 1 or 3 years from 2017 onwards. Based on what was achieved in Mountjoy, we assumed the intervention would screen 76% of the prison population, 86% of diagnosed individuals would engage with the treatment pathway in that year, and 74% of those engaged would initiate treatment in 18 months. This is compared to the SOC, which has a screening rate of 21%, engagement rate of 70% and treatments fixed at 20 per year overall (see Table 3 for details). The intervention costs were either assumed to scale-up with the numbers of people screened, or alternatively the UCD costs were fixed and the remaining costs were scaled up with the numbers of people screened.

Results

Epidemiological model projections

Under the SOC scenario, the model projects a 56.0% (95%CrI 43.1–68.1%) decrease in chronic HCV prevalence in Dublin over 2017–2030, from a median of 52.9% (95%CrI 46.3–58.9%) to 23.0% (95%CrI 15.7–32.7%). There is a corresponding 55.9% (95%CrI 43.1–67.9%) decrease in incidence over the same period from 15.1 (95%CrI 10.8–21.7) to 6.5 (95%CrI 3.9–11.6) per 100 person years (py) (figure 3). The intervention has little additional impact on these projections because of its low coverage, with a median of 1.0 (95%CrI 0.6–1.5) disease related death and 6.3 (95%CrI 3.2–9.0) infections averted over 50 years due to the 12 additional individuals treated compared to the SOC scenario.

Figure 3:

Model projections of chronic Prevalence and incidence of HCV among all people who inject drugs (community and in prison) in Dublin (does not include ex-injectors) including projections for the baseline prison intervention and the scaled up yearly mass screening intervention. The thick lines give the median model projections while the regions bounded by the thin lines show the 95% credibility intervals across the model projections (dark grey for mass screening and light grey for baseline).

In Dublin, our model suggests that 2.1% of PWID are currently incarcerated, with PWID estimated to be incarcerated on average 0.12 times or 1.9% of their injecting career.

Intervention costs

Not including treatment costs, the total direct costs of the intervention came to €82,392 (Table 4; supplementary information for full breakdown). Dividing by the number of individuals tested (n=461) gives a cost per person screened of €179. Similarly, the cost per patient diagnosed (n=50) was €1,648 and the cost per patient linked to care (n=40) was €2,059. Staff costs for the screening and results days constituted the largest cost component (43%, Table 4) with UCD setup and implementation costs making up 38% and clinical tests making up 13%.

Table 4:

Breakdown of Intervention costs

Activities	Cost in Euros	Percentage
Strategic and operational Meetings (4 per year)	€ 2,263	3%
Screening days (staff costs)	€ 25,508	31%
Results days (staff costs)	€ 10,043	12%
Irish Red Cross training of peer supporters	€ 2,402	3%
UCD costs for setup and implementation	€ 31,538	38%
Per patient costs
Phlebotomy and Antibody test costs (419)	€ 5,042	6%
RNA test costs (89)	€ 4,878	6%
Fibroscan test costs (40)	€ 809	1%
Total	€ 82,392

UCD: University College Dublin

Other than the direct intervention costs, our modelling suggests the intervention resulted in a large increase in treatment costs in the prison (€110,173; Table 5), with this being offset by savings (€82,035) in testing, engagement and HCV treatment in the community as well as disease management costs (€73,939) over the 50-year time horizon. Overall, the intervention arm cost on average an additional €36,592 (95%CrI €13,766-€60,547) than the SOC comparator (Table 5).

Table 5:

Breakdown of discounted costs over 50-year time horizon

	Standard of care comparator	HepCheck Intervention plus standard of care	Difference
Disease management	€ 140,815,613	€ 140,741,675	−€ 73,939
HCV treatment in community*	€ 220,648,555	€ 220,571,701	−€ 76,854
HCV treatment in prison*	€ 2,080,095	€ 2,190,268	€ 110,173
Testing	€ 13,308,155	€ 13,305,207	−€ 2,948
Engagement	€ 1,648,184	€ 1,645,951	−€ 2,233
Intervention	€ 0	€ 82,392	€ 82,392
Total	€ 378,500,602	€ 378,537,194	€ 36,592

^*HCV treatment costs includes course of direct acting antivirals and treatment monitoring up to 12 weeks SVR.

Cost-effectiveness projections

The intervention gained a mean additional 3.8 QALYs (95%CrI −1.2 – 9.3) over the 50-year time horizon giving a mean ICER of €9,552 per QALY gained (Table 7) with 84% of model runs in the probabilistic sensitivity analysis being under the willingness-to-pay threshold for Ireland (€30,000/QALY). Uncertainty in the utility value for uninfected current and ex-injectors accounted for 57% and 52% of the variation in incremental QALYs and net monetary benefit respectively (Figure S3), whereas uncertainty in the chronic HCV prevalence in 2014 accounted for the greatest variation in incremental costs (37%), with a positive correlation between HCV prevalence in 2014 and incremental cost (Figure S4). The uncertainty in HCV prevalence in 2014 only accounted for 4% of the variation in the net monetary benefit reflecting negligible impact on the ICER.

Table 7:

Univariate sensitivity and scenario analysis

Scenario	Incremental Costs	Incremental QALYs	Mean ICER
Base scenario*	€36,592	3.8	€9,552 per QALY
Time horizon 25 years	€ 49,309	0.4	€119,834per QALY
0% discount rate	−€151,325	27.5	cost saving
3.5% discount rate	−€ 1,132	7.3	cost saving
Half drug cost	€ 21,499	3.8	€5,612 per QALY
Multiplied health utilities	€36,592	14.5	€2,530 per QALY
Testing and engagement rates halved in those currently incarcerated	€ 35,969	4.0	€9,076per QALY
Testing and engagement rates halved in those currently injecting	€ 56,056	3.8	€ 14,637per QALY
Standard of care is 15 treatments in prison from 2018 onwards, with 5 more due to intervention in 2018	€ 16,721	5.2	€ 3,243 per QALY
Treatment in prison is 12 week course	€52,877	7	€7,175 per QALY

^*time horizon 50 years, 5% discount rate, drug cost €15,180 for a 12-week course, health utilities multiplied together, 20 treatments under standard of care in 2018, 8 week treatment in prison.

Univariate Sensitivity Analysis

All univariate sensitivity analyses were below the €30,000/QALY willingness-to-pay threshold, except when assuming a shorter time horizon of 25 years (Table 8). Lower discount rates (0 or 3.5%) made the intervention cost-saving, while assuming the intervention resulted in more treatments made the intervention more cost-effective (€3,243 /QALY). Conversely, assuming health utilities are multiplied when combined, rather than using the minimum of the two values, decreased the ICER to €2,530/QALY because more QALYs are gained from treating patients in mild disease stages with this assumption (F0/F1). Other sensitivity analyses at most doubled the ICER.

Mass Screening Scenarios

Scaling-up mass screening to all prisons in Dublin is a cost-effective strategy regardless of whether it is every 1 or 3-years (Table 7). However, yearly screening has a greater net monetary benefit (€7,393,382) suggesting this intervention would be more cost-effective despite 3-yearly screening costing less than the standard-of-care scenario. This is due to yearly screening improving the QALYs gained between the two screening scenarios. Importantly, if we only scale-up the non-UCD costs then both mass screening interventions become cost-saving (Table S15). In epidemiological terms, the impact of yearly screening is small (Figure 3) with a median 122 (95%CrI 83–180) or 1.1% (95%CrI 0.7–1.7%) of HCV infections among PWID being averted by 2030 compared to the SOC (Figure S4).

Discussion

Although this pilot mass screening intervention in Mountjoy prison, Dublin, was fairly modest in scope, only leading to 12 additional individuals being treated, our analysis suggests that it was cost-effective (€9,552/QALY saved), being much lower than the willingness-to-pay threshold for Ireland (€30,000/QALY saved). We also show that it is likely to be highly cost-effective (and possibly cost-saving) to undertake yearly mass screening in all Dublin prisons. However, even with this scale-up, our projections suggest that prison-based HCV screening and treatment is unlikely to have a large prevention impact on the overall HCV epidemic among PWID, with yearly screening only reducing the number of new HCV infections by 1.1% (95%CrI 0.7–1.7%) over 2020–2030. This small impact is due to the low levels of incarceration among PWID in Dublin and negligible HCV risk in prison. This contrasts with our other modelling that suggested considerable impact could be achieved from scaling up treatment in prisons in Scotland[30], a setting with much higher levels of incarceration.

Strengths and weaknesses

The main strength of our analysis is the use of real outcome and cost data from a pilot testing and treatment intervention in a Dublin prison, although the small scale of this pilot was also a potential limitation. Importantly, we included the staff costs for a prison officer escorting a patient to appointments for treatment and the prison staff required for screening, which is not normally included[31,32]. Also, our use of a detailed HCV transmission model that incorporates data uncertainty to estimate the subsequent impact of the intervention improves the robustness of our projections. However, limitations still exist.

The main limitation relates to uncertainty in some model parameters due to a dearth of recent survey data from this setting. Most importantly, little data was available on incarceration history and testing or treatment contact among community PWID, and so enrolment data from a recent HCV case-finding intervention undertaken among OST patients in Dublin was used. This data is likely to be biased towards more stable longer term PWID which may experience greater HCV testing and lower incarceration. To counter this, we included uncertainty for these parameters and our model results were robust despite this. Additional uncertainty exists around the likely treatments that would have occurred in the prison without the intervention. However, even with our conservative baseline assumption that 20 treatments would have occurred without the intervention in 2018, our projections still suggest the intervention is highly cost-effective; with this improving dramatically if more treatments were assumed to be due to the intervention. Lastly, in the standard-of-care scenario we assumed the same level of testing and engagement regardless of injecting and incarceration status because of a lack of data to suggest otherwise. Although these rates are likely to differ by setting, our sensitivity analysis suggests that our projections are robust to this uncertainty.

The costing of testing in the community and prison for the standard-of-care comparator was assumed to be similar to costs we have collected for a case-finding intervention among patients prescribed OST in primary care in Dublin. This simplification could affect the cost-effectiveness of our intervention. A further consideration is that we did not include uncertainty in most aspects of our cost estimates for the intervention, with the main uncertainty being in staff costs. This is unlikely to affect our projections because the intervention could cost twice as much and still be cost-effective compared to the standard-of-care.

The modelling focussed on Dublin which may mean our projections are not directly relevant to other Irish settings or elsewhere in Europe. Although many settings have similar epidemics of HCV among PWID, Dublin’s lower levels of incarceration of PWID (2% currently in prison) means that prison-based testing and treatment is likely to have less impact than other settings with much greater incarceration, such as Scotland where 9% of PWID are currently incarcerated[30].

Comparison with other studies

Numerous other studies have evaluated the cost-effectiveness of prison-based screening and treatment interventions in the US, Switzerland and UK; however only three recent studies have considered real-life prison interventions in Spain[33], Taiwan[34] and the UK[31]. Just considering studies using DAA treatments, all these analyses suggest that screening and treatment in prison should be cost-effective with the cost per QALY being $19,000–30,000 in the US[35–37] and €4,015–23,000 in Europe[31,33,38,39]. Studies using empirical data produced similar results to ones that did not. Although these estimates are comparable, they are at least 2-fold more expensive per QALY saved than in our study. This difference is possibly due to testing and referral rates being lower in other studies, with the most optimistic scenario in the previous UK study being more in line with our projections.

Conclusions and implications

Ours and other recent studies have shown the favourable cost-effectiveness of screening and treatment interventions in a prison setting, highlighting the importance of scaling up interventions in such settings. In some settings, these interventions could have considerable impact and be crucial for achieving HCV elimination among PWID[30,33,35,40]. However, our projections suggest this may not always be the case, as we found in Dublin because of their favourably low levels of incarceration compared to other settings globally[6,41]. This should not be seen as a reason for not doing screening and treatment in prison, because we show that the intervention is still cost-effective. Indeed, it is also possible that these interventions may reach individuals that are not well reached by other testing initiatives.

Table 6:

The incremental cost-effectiveness of the intervention scenarios

			Compared to standard of care scenario
	Total Costs	Total QALYs	Incremental Costs	Incremental QALYs	mean ICER	Net Monetary Benefit
Standard of care	€378,500,602	856,646
HepCheck in addition to standard of care	€378,537,194	856,650	€36,592	3.8	€9,552 per QALY	€78,331
Mass screening every 3 years in Dublin prisons	€377,890,425	856,834	€−610,177	188.1	Cost saving	€6,252,816
Yearly mass screening in Dublin prisons	€379,240,571	856,917	€739,969	271.1	€2,729 per QALY	€7,393,382

Net monetary benefit is calculated as incremental QALYs multiplied by willingness to pay threshold (€30,000) minus incremental cost