Modelling the spread of hepatitis C of hepatitis C via tattoo parlors

Modelling the spread of hepatitis C via commercial tattoo parlours: Implications for public health interventions Doris A. Behrens OR and Dynamical Systems Research Unit Department of Mathematical Methods in Economics, Vienna University of Technology, Argentinierstrasse 8/105-4, 1040 Vienna, Austria. Email: [email protected] Marion S. Rauner* School of Business, Economics, and Computer Science, Institute of Business Studies, Department of Innovation and Technology Management, University of Vienna, Bruenner Strasse 72, 1210 Vienna, Austria, Email: [email protected] Phone: 00431 4277 38150, Fax: 00431 4277 38144. Jonathan P. Caulkins Carnegie Mellon University Qatar Campus and Heinz School of Public Policy and Management, 5000 Forbes Avenue, Pittsburgh, PA 15213-3890, USA and Qatar Campus PO Box 24866, Doha Qatar. Email: [email protected]

Abstract: Hepatitis C (HCV) is a serious infection caused by a blood-borne virus. It is a contagious disease spreading via a variety of transmission mechanisms including contaminated tattoo equipment. Effectively regulating commercial tattoo parlours can greatly reduce this risk. This paper models the cost-effectiveness and optimal timing of such interventions, and parameterises the model with data for Vienna, Austria. This dynamic model of the contagious spread of HCV via tattooing and other mechanisms accounts for secondary infections and shows that regulating tattoo parlours such as was done in Vienna, Austria in 2003, is a cost-saving intervention. Keywords: HCV epidemic, commercial tattoo parlours, cost-effectiveness, dynamic modelling, policy analysis

*

Corresponding author

1

Introduction The last decades have witnessed rapid spread of hepatitis C (HCV), a highly contagious blood-borne pathogen which is characterised by substantial morbidity and mortality [9]. One important vector of transmission is tattooing. This paper models HCV infection via this and other means to help answer the question of whether and when commercial tattoo parlours should be regulated to reduce the spread of HCV. Altogether, 3.1% of the world’s population (170 million people) is infected with HCV, where prevalence substantially varies between regions [10, 44]. It is relatively low in Australia (1%: [12]), the US (1.6%: [6]) and most of the European Union (between 0.003% (Sweden) and 1.5% (Greece): [13]). Eastern Europe exhibits, however, higher rates (up to 4.9%: [13]). The country with the highest infection rate is Egypt with 20% of military recruits being tested positively for HCV antibodies (http://www.pharmaceutical-business-review.com/article_feature.asp?guid=7C440DBF-2112-4E7C-B306-803773387502 accessed 6 June 2006). Yet hepatitis C has not received attention commensurate with the magnitude of the problem, possibly because infection is usually asymptomatic for many years and less rapidly fatal than is HIV/AIDS [2]. Nevertheless, HCV is more virulent, contagious, and prevalent than HIV. It infects four times as many people than HIV, and –– due to high incidence from the late 1970s through the early 1990s –– is expected to kill more Americans than HIV by the year 2010 [2, 14]. Hepatitis C can be spread in a variety of ways, including via (unscreened) blood transfusion and organ transplantation, haemodialysis, use of inadequately sterilised medical, surgical or dental equipment, birth to an infected mother, injecting and other illegal drug use, sexual contact with an infected partner, household contact and via percutaneous exposures caused by, e.g., folk medicine practices, tattooing, body piercing, and commercial barbering [9]. Currently, the most frequent transmission mechanism in developed countries is sharing druginjecting equipment among injection drug users [11], since the development of HCV screening and regulation of organ transplantation and blood transfusion has been so successful that they have gone from ‘one of the most common’ to ‘hardly any’ sources of HCV transmission, e.g. from 10% to no more than 0.004% per unit transfused in the United States [9, 45]. A question motivating this paper is whether tattooing may (have) be(en) similar to blood transfusion in this sense – a significant source of infections that still are preventable via regulation. There are widely divergent opinions concerning the proportions of HCV infections caused by tattooing, with estimates ranging from less than 1% [7], to over 3% [22], and even up to more than 20%: Haley and Fischer’s 1991/92 study [20, 21] of 626 selected patients of an orthopaedic spinal clinic reports that 22% of patients with a tattoo exhibit an HCV infection (as opposed to only 3.5% of patients with no tattoos). Moreover, Haley and Fischer [20] report that commercially-obtained tattoos may be the source of twice as many HCV infections as is injection drug use. Other researchers argue, however, that the Haley-Fischer study [20] misinterpreted the data and that the likelihood of acquiring an HCV infection via tattooing is rather small [1]. Nonetheless, there is some cause for concern that, at least in the United States, tattooing might have been grossly underestimated as a source of HCV infection. This paper investigates whether and when tattoo parlours should be regulated to reduce their role in the spread of HCV. In particular, one might ask ‘At what point in the spread of an HCV epidemic are expensive regulatory interventions a cost-effective way to reduce the spread of HCV?’ Hence we present a (six-state) nonlinear dynamic model of HCV spread that captures the effects of secondary HCV infections. The states of the model correspond to people who are uninfected and hence susceptible to infection and infected people in both the acute or chronic infection states, with all three infection states split between individuals seeking and not seeking tattoos. We parameterise our model for Vienna, Austria by building of a

2

pioneering study of Viennese tattoo parlours by Blaschke and Purkert [8] that provides unique insights into the operational practices of tattoo parlours. In 2003 the Austrian Chamber of Commerce instituted strict regulation of commercial tattoo parlours in order to control the spread of HCV [18], an action that is retrospectively justified by our analysis. This regulation required safe tattooing practices ranging from simple hygiene (e.g., hand-washing and protective clothing) to using one-shot, non-reusable needles and ink disposables and adequately sterilising reusable equipment (by using an autoclave). Before regulation, control practices were often nonexistent or inadequately disinfected the reusable equipment which included needles (reused up to five times) and ink. Other cities and regions have not yet taken similar control action. This paper may inform their deliberations concerning such actions. On the one hand, intervention programs accrue additional cost (e.g., sterilisation cost), but on the other hand infection costs are averted (e.g., treatment costs). Typically, policy makers have to invest additional money to gain additional health for the population, and the intervention is judged to be cost-effective if the cost per unit of health gained is below some willingness-to-pay threshold. In exceptional cases, the costs averted by the intervention may exceed the costs of the intervention, in which case the programs are not only cost-effective, but also cost-saving [19]. We found that regulating tattoo parlours can avert enough HCV infections relative to its low cost that is often a cost-saving intervention. However, for a sufficiently low prevalence of HCV in the population and/or a sufficiently low HCV infection risk per tattoo, and/or a low number of tattoos per person, the total sterilisation costs can be above the total HCV infection cost averted. Mathematical models of the spread of contagious diseases have long been used to inform public health interventions [3, 26]. An important subset developed in response to the HIV/AIDS epidemic [25, 29, 32, 33]. That work is highly relevant to HCV control because many of the transmission vectors are similar. Indeed, interesting papers have been written using essentially identical modelling methods for both diseases, but contrasting the implications that emerge because of differences in parameter values [15, 30]. In brief, HIV has lower infectivity because that virus is less robust outside the human body, but the average cost per HIV infection is significantly higher than with HCV. There are more differences between HCV and HIV in the course of disease progression post infection. A number of authors have developed interesting models of incidence and progression of HCV infection. Such models are important for making medical treatment decisions and for healthcare forecasting and planning purposes, but are less directly relevant here, where the focus is primarily on modelling incidence and its prevention. Likewise, many papers have explicitly modelled the spread of HCV via other mechanisms (e.g., [31, 35, 36]) including injection drug use (e.g., [16, 27, 28, 37]), sexual contact (e.g., [40]), and blood transfusion (e.g., [42]). This paper is, however, to the best of our knowledge, the first that models HCV transmission via tattoos.

Model According to the Association of Professional Piercers a set of standard guidelines exists that identifies safe venues for getting body piercings and tattoos. ‘Needles must be the disposable and the single-use type, which should be discarded properly upon completion. For tattooing, not only should the needle be disposable and sterile, but the equipment should be autoclaved, and the ink cartridges disposable and new for each person. The piercer or tattooer must wear latex gloves when performing the procedure’ (http://tattoo.about.com/gi/dynamic/offsite. htm?zi=1/XJ&sdn=tattoo&zu=http://www.jhunewsletter.com/vnews/display.v/ART/2004/11/ 05/418ad9a178e7a accessed 25 December 2005). 3

Together with the discussion above the existence of these guidelines for customers strongly suggests that governments may be contemplating regulating tattoo parlours in order to reduce the spread of HCV. Sterilisation is, however, not inexpensive; as we will elaborate below, it could increase the cost per tattoo by ~15%. Hence, one might expect some resistance from business interests, and quantification of the benefits vs. cost of regulation ought to enter the public debate. Such quantification should recognise secondary infections. When people get infected by receiving a tattoo, they may subsequently infect others by having other types of HCV risky contacts. The number of secondary infections is not simply the so-called ‘reproductive number’ of ‘the’ tattoo-born epidemic because there are many HCV transmission mechanisms. An HCV-infected person who gets a tattoo participates in two interrelated epidemics, one spreads by tattooing and the other spreads by other means. Furthermore, HCV is still spreading. So, decisions concerning tattoo parlour regulation pertain to the transient spread, not just steady state conditions. Hence, we need a dynamic model that explicitly tracks both forms of HCV transmission in a manner that supports ‘what if ’ policy experiments to explore the effects of regulating tattoo parlours earlier or later in the spread of the HCV epidemic. Model of uncontrolled HCV spread The model has three pairs of states, with one state in each pair for people who visit tattoo parlours (subscripted by t) and one for others (subscripted by o). The three pairs distinguish between people who are uninfected and hence Susceptible to infection, and infected people in both the Acute and Chronic infection state (see Figure 1). Thus, HCV prevalence is

 :

Ao  At  Co  Ct S o  S t  Ao  At  Co  Ct

.



For HCV transmission other than via tattoos, we make the common assumption in epidemiological models that infection spreads by random mixing based on existing data. Compared to selective mixing, random mixing generates a larger epidemic and thus can be interpreted as a worst-case spread pattern of an epidemic given that the infection risk per contact remains the same for all risk groups [23]. Thus, the coefficient, , can be thought of as a sum over transmission mechanisms other than tattooing of the product of the frequency of contact between individuals and the probability that contact between an uninfected and an infected individual leads to a new infection. Those components never appear individually and cannot be estimated separately, so they are combined into a single parameter . There are many ways of intervening to reduce transmission via these routes that merit study. However, the objective here is to model the benefits of regulating tattoo parlours, so those issues are suppressed. Seeking tattoos is an activity that is usually initiated in early adolescence, at roughly the same time that the other, dominant HCV-transmitting behaviours (e.g., drug use) begin. Hence, letting  represent the fraction of people who will ever be interested in getting a tattoo, we split the inflow of uninfected people reaching adolescence, k, into a stream   k flowing into the population of uninfected individuals who seek tattoos (St) and a stream (1 – )  k of other uninfected people (So). As indicated by Figure 1 tattoo-induced transmission is presumed to occur at a rate that is proportional to the product of the annual rate with which susceptibles who visit tattoo parlours (St) get tattoos ( ), the probability that an infection results from getting a tattoo with equipment that is infectious (), and the probability that the tattoo equipment is infectious (  ). The last term ( ) is the product of the probability that use of an infected tattoo recipient renders the equipment infectious ( ), and the probability the equipment has previously been used on someone who was HCV positive ( f((t)), where t denotes the HCV prevalence among people 4

getting tattoos.1 Blaschke and Purkert’s (2004) direct observations and discussions with tattoo shop owners and equipment producers suggest that before regulation tattoo equipment was typically used five times before being replaced. Hence, customer demand for safe tattoos was not sufficient to motivate tattoo parlours to adopt safe practices. Assuming the prior HCV status (before the tattoos) of successive customers is independent, this last probability, {tattooing equipment exposed to HCV}, can be written as 4

 At  Ct 1  1  f(t)   S t  At  Ct i 0 1 5



i

  . 



k

S tattooed subpopulation

k

susceptibles

S

(1 – )k

g

St

non-tattooed subpopulation

susceptibles

So

rCS

rCS

 + 

rAS chronic inf.

Ct

C

rAC

acute inf.

At

 g

A

rAS

acute inf.

Ao

A

rAC

chronic inf.

Co

C

g

Figure 1: Flow chart of the HCV model (Equation (3)) for    f(t) as defined by Equation (2).

The final model flows reflect people ‘maturing out’ of seeking tattoos. These are ‘within infection state’ flows from those seeking tattoos to those no longer seeking tattoos and are presumed to be at an annual per capita rate g. Hence, the full model can be written:

1

Here we view the probability that equipment becomes infectious as being a function of the operator/shop/day as opposed to the particular tattooing session. That is, that there are some shops/operators/conditions under which giving a tattoo to an infected person renders the needle infectious, and others that do not. E.g., some operators give tattoos in ways that draw blood, and others do not. In particular, for 100 % of the operators/shops/days/conditions that happens, and in the others it does not. When the equipment becomes infectious, it becomes infectious regardless of whether it was used on 1, 2, 3, or 4 infected people.

5

S o A o C o S t A t C t

 1    k  gS t  rAS Ao  rCS C o    S   S o   S o  gAt  rAC  rAS   A  Ao  rAC Ao  gCt  rCS   C C o   k  rAS At  rCS Ct   g   S     f  t S t     f  t S t   g  rAC  rAS   A  At  rAC At   g  rCS   C Ct

(

where



k ..........  .......... g .......... rXY ....... X ........  .........  ..........  ..........  ..........

annual inflow to the population (‘births’), all of whom are uninfected, fraction of people who are interested in getting a tattoo, annual rate at which people are no longer interested in getting more tattoos, annual rate of transition from state X to state Y, annual exit rate from death or other causes from state X, annual rate of HCV infections attributed to other mechanisms than tattooing, annual rate of getting tattoos, probability an infection results from infectious equipment, probability an infected tattoo recipient renders the equipment infectious,

and where  is defined by Eq. (1), and f(t) is determined by Eq. (2).

Regulatory Control Practical considerations of how public health regulations are promulgated suggest modelling regulation as a simple, binary condition, not one phased in over time. Hence, we assume any (successful) sterilisation/regulation policy of tattoo parlours can be modelled as reducing  at a specific point in time. Ideally  is reduced to 0, but we explore implications of regulations that reduce  (but do not make it zero) in sensitivity analyses. Model simulations generate two outputs: (a) the number of sterilisations  and (b) the number of HCV infections I over time. As is customary, future outcomes are discounted at a constant annual discount rate r from the present (t0) to the time of regulation ( ) and beyond to the (possibly infinite) end of the planning horizon ( ), so: T

   ,T  

 e

 rt   S

t

 At  Ct  dt ,



I   t 0 , , T  

e

T rt 



S o     f  t S t dt  e rt  S o   S t dt ,

t0

r .......... t0 …….  .......... T ..........





constant discount rate, beginning of planning horizon, point of starting the intervention, end of planning horizon,

where parameters , , , and  are explained above (see Eq. 3) and where  and f(t) are given by Eqs. (1) and (2), respectively. Running the model twice, with two different points ( = t1 and  = t2) of intervention (which includes running the model with and without control, i.e.,  = t1 = t0 and  = t2 = T )

6

generates an incremental cost (increase in the number of sterilisations) and an incremental effectiveness (e.g., reduction in the number of HCV infections). The ratio of these two quantities is the incremental cost-effectiveness of moving the point of intervention forward from t1 to t2 [19]. E.g., setting t1 = T and t2 = 0 assesses the cost-effectiveness of regulating now vs. never. Given estimates of the ‘cost of sterilisation’ ( ) and the cost of a newly acquired HCV infection (r), cost-effectiveness results can be converted into a benefit-cost ratio,

BC  t1 , t 2  

I   t 0 , t 2 , T   I   t 0 , t1 , T   r ,    t1 , T     t 2 , T  





incremental cost due to regulation, including using a new needle, new ink, disposables, sterilisation of the tattooing equipment, r ........ average cost of a newly acquired HCV infection, t1, t2 ..... times of starting an intervention to regulate tattoo parlours, where parameters t0 and T are explained above (see Eq. 5). If the HCV prevalence is low enough, then from a social planner’s perspective the benefits of reduced HCV do not offset the cost of sterilisation, but for a higher prevalence they would. The optimal point in the epidemic for regulation to begin can be calculated as the time when prevalence rises to the point that delaying regulation has an adverse incremental benefit-cost ratio. Policy makers are not only interested in the optimal control point in the epidemic but also in an intervention’s cost-effectiveness, specifically the incremental cost per incremental unit of effectiveness. For example, incremental effectiveness measures include the number of infections averted, number of life years saved, and number of quality-adjusted life years saved.31 We computed the incremental cost-effectiveness ratios (CERs) by calculating the cost of a regulation (CR) compared to a non-regulation (CN) scenario (i.e., the incremental cost due to regulation), divided by the number of infections with the regulation (ER) as compared to the non-regulation (EN) scenario (i.e., the number of infections averted): CER 

C R  C N C  E R  E N E

(7)

The CER can be interpreted as the cost per infection averted by regulation of the tattoo parlours. Programs that obtain a negative ΔC together with a positive ΔE compared to the status quo generate benefits with no net cost; they are referred to as cost-saving. As we’ll see below, the cost of regulating tattoo parlours are so modest compared to not only the social cost but also the direct budgetary cost of an HCV infection, regulating tattoo parlours can be cost-saving in some circumstances. Sensitivity analyses can show how that optimal point and the CERs of intervention change with various parameter values. We turn next to a discussion of those parameter values.

Parameterising the model for Vienna, Austria Population and infection-related parameters Most people begin receiving tattoos on average around age fifteen [22]. There is no clear-cut maximum age for getting a tattoo, but there is enough age-dependence that it does not make

7

sense to lump together all adults. Although many middle-aged people get tattoos, few older people do, so we truncate the age range at 50. If there were no mortality, 15-50 year olds would exit the model (by aging) at an annual rate of 1/35 = 0.02857 per year. The average annual death rate for 15-50 year olds in Austria is 0.001475 [39], so we take  S = 0.02857 + 0.00148, or  S = 0.03 in round numbers. There are currently 775,000 15-50 year olds in Vienna. Official statistics project that the population of Vienna will increase by only ~6% by 2020, so for simplicity we assume the population of 15-50 year olds will be constant at 775,000. Age distribution effects stemming from the baby-boom and immigration are not critical for present purposes. If a population of 775,000 individuals is stable despite an exit rate of 3% that implies an inflow of 23,250. Tattoo shops in Vienna serve not only residents of Vienna, but also people in adjoining areas of Lower Austria and Burgenland, effectively expanding the relevant population base by 50%. Hence, we set the annual inflow to the population k = 1.5  23,250 = 34,875. People exit the tattooing states not only by dying or turning 50 but also by no longer being interested in getting more tattoos. In this model that is parameter g. We do not have a way of estimating g directly, but we can estimate the sum of  S and g indirectly from the average age of someone getting a tattoo, which is about 26 (http://www.tattoo.dk/questionnaire/engcyberresults.htm accessed 09 June 2006). To make the average age of people in a 15-50 year old age box with constant inflow and a constant per capita outflow rate equal 26, the constant per capita total annual outflow rate must be ( S + g) = 0.070. This combined exit rate of  S + g = 0.070 implies g = 0.040. Likewise, since the mean of an exponentially distributed dwell time is one over the total exit rate, the average time someone who enters the tattooing population spends willing to get tattoos is about 1/0.070 = 14.3 years. Among those who get a tattoo, the average number of tattoos is 3.5 (http://www.tattoo.dk/questionnaire/eng-cyberresults.htm accessed 09 June 2006), suggesting the average annual rate of getting tattoos is  = 3.5/14.3 = 0.245. We estimate the proportion of people entering the tattooing populations ( ) in two ways. First, Health Canada (2001) finds that 8% of adolescents have tattoos and another 21% want to get some, suggesting  = 29%, and estimates from other sources, countries, and ages generally are also in this range [4, 5]. Second, thinking in steady state terms, dividing Blaschke and Pukert’s (2004) estimate of the number of tattoos given annually in Vienna (36,197) by the average lifetime number of tattoos among those who get tattoos suggests that 10,342 people join the tattoo population each year, which is  = 29.65% of the k = 34,875 people entering the system each year in total. We round these estimates to a base case of  = 30% but do sensitivity analysis for 0.25    0.35. Transition rates from the acute to the susceptible and chronic states are set to rAS = 0.9 and rAC = 1.1 because the acute stage lasts 6 months and about 55% of those who are infected progress to the chronic state [43]. Note that only 25% of the HCV infections are acute and roughly one-third of the others are registered [38]. The transition rate from chronic to susceptible (rCS) reflects both spontaneous remission (0.002 per person per year) and treatment. According to Siebert et al. [38], 61% of people respond to treatment over a 16-year period, suggesting an average annual rate of –ln(1–0.61)/16 = 0.059 for a combined rate of about rCS = 0.059. Chronic HCV can be lethal, but only over an extended period, so most HCV deaths occur after age 50. We reflect those delayed deaths in the social cost per HCV infection, but the death rate for HCV infected 15-50 year olds is not dramatically higher than it is for other 15-50 year olds. The average age of receiving a tattoo is 26 so we are interested in the cumulative HCV death rate over a period of 50 – 26 = 24 years, which is approximately 9.5% [24, 39]. The equivalent annual death rate is 1 – (1 – 9.5%)1/24 = 0.415%. Given that the background exit rate for 15-50 year-olds without HCV was  S = 0.03, we set  C = 0.03415  0.034. Since

8

substantial long-run morbidity and mortality is associated with HCV, the exit rate in the acute state is the same as in the susceptible state, i.e.,  A =  S = 0.03. Tattooing is not the primary source of HCV transmission, so  can be estimated indirectly given a plausible steady-state prevalence in the absence of tattooing as 



rAC  rAS   A  rCS   C   rAC  rCS   C  1  ˆ 



The steady state prevalence ˆ is not known because the HCV epidemic has not stabilised. However, it is probably modest. Hence, given the parameter values cited above, can reasonably be modelled as being a few percent greater than the total exit rate for individuals with chronic infection (rCS +  C), divided by the proportion of infected people who progress to the chronic state, i.e., (0.059 + 0.034) / 0.55 = 0.16909. There is considerable uncertainty concerning . On the one hand, literature suggests that the probability of becoming infected with HCV given one needle stick with an infectious needle is 1.8% (range: 0%–7%) [41]. On the other hand, a tattoo might be thought of as involving multiple pricks over 0.5-2 hours, albeit none to great depth and, therefore, not necessarily exhibiting the contagiousness of intravenous contact. Some Austrian HCV experts have, however, suggested that once a tattoo is bleeding the chance of an HCV infection via an infected needle may quite high. There is likewise no figure in the literature for the probability that tattoo equipment becomes infectious when it is used on someone who is infected ( ). Fortunately  and  only appear together, as a product, so we can do sensitivity analysis with respect to both merely by looking at alternative values for their product. We also do breakeven analysis asking how large  would have to be, for a given current HCV prevalence, in order for regulation to be cost-effective. We choose for a base case value = 0.018 [41], but explore the entire range of possible values from 0-100%, particularly the range between 0% and 5%. Objective function parameters We use an annual discount rate of r = 3% and consider 0% and 5% in sensitivity analysis [19]. Based on Blaschke and Purkert [8], Rauner et al. [34] estimate the incremental materials and other non-labour cost of sterilisation and other hygienic procedures induced by regulation (€ 6.11) and the cost of controlling regulation (annual inspection & certification of the sterilisation machines being equal to € 0.30) would be approximately € 6.41. Assuming staff cost are about € 60 per hour and sterilisation takes about 10 minutes, sterilisation adds € 10 in labour cost for total sterilisation cost of about € 16.41 per tattoo. Before regulation a typical tattoo cost about € 100 so this represents an incremental cost of about 15%. We assume this is not enough of a price increase to drive significant numbers of customers to underground or unregulated tattoo sources. We use Siebert et al.’s [38] model to estimate the average social and health cost per HCV infection, and assign the present value of those cost to the moment of infection [17, 39]. The average social cost includes (a) symptomatic acute care cost of € 1,814 for the 25% who are symptomatic, yielding an average of € 450, (b) chronic health care cost incurred by the 55% of those infected who progress to the chronic stage, and (c) € 26,310 per life year lost, reflecting productivity/consumption losses based on the average yearly Austrian gross income [17]. Lifetime cost are much lower for the roughly 61% of individuals exhibiting a chronic HCV infection who are successfully treated (€ 46,760 per case) than for those for whom treatment fails (€ 267,560), so the total undiscounted cost per infection is r = 0 = €450 + 0.55  (0.61  €46,760 + 0.39  €267,560) = €73,530. Most chronic costs occur considerably after the time of infection. Discounting back to the time of infection at a 3% (5%) annual rate reduces total

9

discounted cost per infection to €37,730 (€25,780). Thus, the discounted cost per infection is quite sensitive to the discount rate because those costs are distributed over 50+ years postinfection. (Discounting calculations are done with detailed projections of the timing of future cost, but in round terms the modal chronic stage cost are about 24 years post-infection, and $450 + ($73,530 – $450 ) / 1.0324 = $36,400, which is close to the $37,730 figure from more detailed calculations). Total social costs are dominated by mortality. Health care costs are less one-quarter of the total, at € 15,450 undiscounted, or € 9,190 and € 6,630 when discounting at 3% and 5%, respectively. Different from the willingness-to-pay approach, where people value ‘health’ versus ‘money’ or ‘survival’ versus ‘consumption’, is the human capital approach, which is used here. Both approaches can be criticised as attributing monetary value to health as consumer good, and the latter for discriminating against the retired, disabled, and unemployed. However, we have chosen the human capital approach for our cost-benefit analysis because policy makers can compare intersectoral projects such as in health care, education, infrastructure, and security. Table 1 summarises all parameter values. Table 1: Base case parameter values. Value k

 g rAS rAC rCS

S A C    r T

 r

Description

Interval

Annual inflow to the population (= ‘births’) – Fraction of people who are interested in getting a tattoo 0.25-0.35 Annual rate at which people are no longer interested in getting more tattoos – Annual rate of transition from ‘acute’ to ‘susceptible’ – Annual rate of transition from ‘acute’ to ‘chronic’ – Annual rate of transition from ‘chronic’ to ‘susceptible’ – Annual exit rate from the pool of susceptibles – Annual exit rate from the acute state – Annual exit rate from the chronic state – Annual rate of getting tattoos – Coefficient of HCV spread via methods other than tattoos 0.16-0.178 Probability of an infection resulting from getting a tattoo with tattoo equip0-1 ment that is infectious 0.030 Discount rate 0-0.05 15 Length of planning horizon (years) 1-15 16.41 Cost of sterilisation process (in €) including using a new needle, new ink, – sterilisation of the tattooing pen, etc. 37,730 Social cost of an HCV infection (in €), discounted to the time of infection for 25,78073,530 r = 0%, 3% and 5% 34,875 0.300 0.040 0.900 1.100 0.059 0.030 0.030 0.034 0.245 0.169 0.018

Results Using the initial conditions for Vienna in 2003 derived by Rauner et al. [34], the model suggests that regulating tattoo shops was a cost-effective action. In particular, the benefit-cost ratio of control for t0 = t1 = 2003 and t2 = T = 2018 when discounting with the usual r = 3% is 19,646  19,297 37,730 BC 2003, 2018    1.82 . 442,052 16.41

10

Discounting at 0% and 5% the benefit-cost ratio becomes 3.59 and 1.23, respectively. This benefit-cost ratio tells us that – discounting at 3% – the social cost averted (due to infections averted) are 1.82 times the cost of regulation. Since health care costs are on the order of one half of the total social cost, regulating tattoo parlours was borderline cost-effective for Vienna even if no monetary value is placed on premature deaths due to HCV. (Note also that truncating the model projection at the end of the 15-year time horizon means that secondary infections prevented after 15 years are not counted in this ratio, so it is conservative.) In net present value terms (subtracting the incremental cost of sterilisation from the social cost savings), over the next 15 years regulation would generate a net (3%) discounted value of social benefits of approximately € 5,942,860. Some might also be interested in a ‘short term’ perspective that focuses on cost and infections over a shorter time horizon. Generally speaking the shorter T is, the lower the benefitcost ratio because some future secondary infections (and their cost) are ignored, but the ratio does not fall precipitously because shortening the horizon not only reduces benefits (credits fewer infections averted) it also reduces the cost (counts up sterilisations only through time T). Reducing the planning horizon to a single Austrian legislative period (T – t0 = 4), for instance, still yields a cost-benefit ratio of BC(2003,2007) = 1 .61. Even if T – t0 = 1, the benefitcost ratio is still favourable (1.57). The choice facing policy makers in 2003 was not so much to regulate in 2003 or never regulate. Not regulating in 2003 still leaves open the possibility of enacting regulations later, say in 2004. For regulating tattoo shops in 2003 instead of 2004, the incremental benefit-cost ratio for a long time horizon with T = 2018 is

BC 2003,2004 

19,349  19,297 37,730   3.32 442,052  405,997 16.41

for

 = 1.8%.

The incremental benefit-cost ratio of regulating now instead of next year is even better than that for regulating ‘now’ vs. ‘never’ because more of the delayed secondary infections averted occur within the planning horizon. The favourable benefit-cost ratio for Vienna is noteworthy inasmuch as in some respects Vienna needs tattoo regulation less than might some other cities. Vienna has a modest overall prevalence (0.8%), relatively low rates of transmission via other mechanisms including injection drug use, a strong social and health security system that offers good treatment (for nondrug users and patients in methadone programs), and we’ve assumed a conservative HCV transmission probability ( = 1.8%). For cities in which these conditions are not true, regulating tattoo parlours may not only be a cost-effective way to make a modest dent in the HCV epidemic. It might make a material difference to the future course of the epidemic. For example, in test runs with a high transmission probability ( = 50%), regulation not only reduced the intensity of the HCV epidemic, it also greatly reduced the endemic level of HCV prevalence. There is considerable uncertainty concerning the parameter constellation , and there is also uncertainty concerning other parameters, notably  and . Does that parametric uncertainty change our results? Holding other parameters at base case values (including a planning horizon of 15 years and a discount rate r = 3%) Table 2 shows the results of varying , and  for the benefit-cost ratio, BC(2003,2004), when starting with initial conditions for Vienna in 2003. Table 2: Illustration of how the cost-benefit ratio BC(2003,2004) of regulate now vs. next year depends on the assumed probability of an infection resulting from getting a tattoo with tattoo equipment that is infectious ( ) and on considerations reflecting HCV infection from reasons other than

11

tattooing (). BC for   = 0.169  = 0.3 0.1% 1.0% 1.8% 5.0% 10.0% 20.0% 50.0% 100.0%

0.18 1.84 3.32 9.28 18.72 38.13 100.75 221.36

BC for  = 0.16  = 0.3

BC for  = 0.178  = 0.3

BC for  = 0.169  = 0.25

BC for  = 0.169  = 0.35

0.17 1.75 3.15 8.81 17.77 36.20 95.64 210.16

0.19 1.94 3.50 9.78 19.74 40.20 106.21 233.33

0.18 1.85 3.34 9.33 18.82 38.34 101.28 222.53

0.18 1.83 3.30 9.23 18.62 37.93 100.21 220.20

As can be seen from Table 2 parameter variation can substantially change the benefit-cost ratio of regulating tattoo parlours, but only quite substantial parameter variation brings that ratio down to the critical level of 1. E.g., the breakeven value for parameter constellation  is only 0.55% (for an initial prevalence of (0) = 0.8% and a planning horizon of T = 15 years). Although the literature does not provide firm estimates of those parameters, it seems plausible that their product exceeds that value. It is interesting to observe that an increase in the fraction of people, who are interested in getting a tattoo () reduces the BC ratio and vice versa. More people getting tattoos means more infections averted, but it also means more expensive sterilisations. However, the smaller the non-tattoo population relative to the tattoo population, the less important secondary infections are relative to primary, tattoo-induced infections. Likewise, increasing the frequency of non-tattoo transmission () increases the BC ratio of tattoo regulation, again because more secondary infections are averted per tattoo-induced infection averted. These results have ”praised” Vienna for regulating tattoo parlours in 2003 when the HCV prevalence was about 0.8%, but perhaps Viennese decision-makers should have acted even sooner, when HCV prevalence was lower than 0.8%. We can investigate this by finding the smallest prevalence such that implementing regulation now is better than waiting another year. That breakeven prevalence is around one quarter of today’s value, i.e. (0) = 0.23%. No one really knows in what year the HCV prevalence was around this value, but it was probably at least a few years earlier. Table 3: Minimum prevalence for which not delaying intervention for another year is cost-justified as a function of planning horizon and whether secondary infections are considered. Planning horizon T

Full (inclusive) Model

Naive Model Tracking Direct Infections Only

1 2 3 4 5 10 15

0.500% 0.451% 0.415% 0.385% 0.360% 0.276% 0.229%

0.492% 0.492% 0.492% 0.492% 0.492% 0.492% 0.492%

Table 3 shows this breakeven prevalence for which regulation becomes cost-justified depends on (1) the length of the planning horizon T and (2) whether one considers all HCV infections, including secondary infections as in the full model, or one uses a naive model that 12

considers only primary infections. More specifically, the naïve model compares the cost of sterilisation () to the product of the cost of HCV infection (r) times the probability that someone seeking a tattoo will become infected via tattooing (  f(t)). Since the naïve model does not include secondary infections, it reaches similar conclusions as does the full model for a one-year planning horizon. For T > 1 the full model continuously considers secondary infections and recommends intervening sooner compared to the naïve model that neglects secondary infections. How much sooner depends on the rate of change in prevalence. For the full model Table 3 shows that longer planning horizons T reduce the breakeven prevalence. E.g. extending the planning horizon from one to fifteen years lowers the breakeven prevalence by one half. This information is relevant for decision makers in cities that may not yet have instituted regulations and which may be at different points in the HCV epidemics’ spread. As an additional policy map, Figure 2 provides more information relevant for decision makers. The graph shows current prevalence versus the number of infections averted by not waiting one more year until the regulation is installed per million sterilisations additionally performed (compared to the sterilisations necessary if control was delayed), i.e.,

I   t 0 , t1 , T   I   t 0 , t 0 , T 

  t 0 , T     t1 , T 

.

(9)

We again plan for 15 years, the discount rate amounts to 3%, and all epidemic parameters are chosen as described in Table 1. The solid lines in Figure 2 identify the number of infections averted per million (additional) sterilisations as a function of current prevalence ((0)) for four different probabilities of getting an HCV infection via tattooing (). The amount of this probability  may vary between different cities. The dotted line in the lower part of Figure 2 represents the line where the benefit-cost ratio is equal to one using our base case cost parameters (see Table 1). Numerically, its level (435) is the cost of one million sterilisations (€16.41 * 1,000,000) divided by the social cost of an HCV infection (€37,730). Above the line regulation creates positive net benefits. For a higher cost of control or lower social cost of an HCV infection the threshold line moves up. Following the dotted line from left to right shows how the breakeven prevalence significantly increases with a decreasing likelihood of getting HVC infection via tattooing,  . For the base case parameter value  = 1.8%, not delaying control for one more year (starting with a current prevalence of 0.8%) prevents 52 additional infections at a cost of 36,055 additional sterilisations. This yields 52  1,000,000/36,055 = 1,442 infections averted per million sterilisations as indicated by the black diamond in Figure 2. Following the direction of the grey arrow one would acquire a more ‘optimistic view’ about the probability of getting HVC infection via tattooing ( ). This corresponds to smaller values of  . If that infection risk via tattooing is about half its base case value, i.e.  = 1%, 29 infections are averted (instead of 52). This reduces the number of infections averted per million sterilisations to approximately 800 (indicated by the grey diamond in Figure 2). For a probability of infection as low as  = 0.55%, regulating tattoo parlours Vienna would reach the breakeven number of 435 infections averted per million sterilisations (white diamond in Figure 2). Figure 2 tells us an appropriate policy for any city where our epidemic parameters are a suitable approximation. One locates the current prevalence on the horizontal axis and then moves up to the curve or curves corresponding to what one believes about the probability of infection via tattooing ( ). Moving horizontally back to the vertical axis gives the estimated infections averted per million sterilisations. Comparing that level to the ratio of the sterilisation costs to HCV infection costs suggests what decision to make about regulating tattoo par13

lours. In particular, if the estimated number of infections averted per million sterilisations is greater than the 1,000,000 times the ratio of the cost of a sterilisation divided by the social cost of an HCV infection, then regulating now is cost justified.

Infections averted per million (additional) sterilisations

25,000

20,000

  

15,000

 10,000

5,000

0 0.0%

2.5%

5.0%

7.5%

10.0%

12.5%

15.0%

prevalence 0)

Infections averted per million (additional) sterilisations

1,600 1,400



1,200



Vienna base case scenario



1,000



Vienna optimistic scenario

800 600

Vienna breakeven scenario

400

Threshold level: BC(2003,2004) = 1 200 0 0.0%

0.1%

0.2%

0.3%

0.4%

0.5%

0.6%

0.7%

0.8%

0.9%

1.0%

prevalence (0)

Figure 2: Policy chart for implementing a regulation program for tattoo parlours ‘now or next year’ for the base case parameter values (lower graph is the magnification of the boxed area in the upper graph.

14

r=0

Total HCV Incidence (number of annual infections summed over the time horizon T=15; discounted)

non-tattooing infections  (caused by ) 0.018 24,247 0.05 24,642 0.1 25,318 0.2 26,918

tattooing infections (caused by )

all infections (caused by )

230 669 1,447 3,411

Total Cost (summed over the time horizon T=15; discounted)

24,477 25,312 26,764 30,328

cost of control (in 000s) €0 €0 €0 €0

cost of infections (in 000s) € 1,799,765 € 1,861,159 € 1,967,979 € 2,230,056

total cost (in 000s) € 1,799,765 € 1,861,159 € 1,967,979 € 2,230,056

Final Prevalence (at the end of the time horizon T=15)

CER (incremental cost/averted infection; time horizon T=15)

nontattooing total tattooing population population population 0.88% 0.84% 0.87% 0.89% 0.92% 0.90% 0.92% 1.06% 0.94% 0.99% 1.45% 1.05%

incremental averted CER cost infections (in 000s) (in 000s) -€ 23,374 440 -€ 53 -€ 84,768 1,275 -€ 66 -€ 191,588 2,728 -€ 70 -€ 453,665 6,292 -€ 72

0.5

34,753

14,948

49,700

€0

€ 3,654,467

€ 3,654,467

1.34%

4.02%

1.69%

0

24,036

0

24,036

€ 9,008

€ 1,767,383

€ 1,776,391

0.87%

0.80%

0.86%

r=0.03





191 554 1,191 2,775


19,646 20,307 21,452 24,232


cost of infections (in 000s) € 741,255 € 766,197 € 809,371 € 914,292

total cost (in 000s) € 741,255 € 766,197 € 809,371 € 914,292


incremental averted CER cost infections (in 000s) (in 000s) -€ 5,943 350 -€ 17 -€ 30,885 1,011 -€ 31 -€ 74,060 2,155 -€ 34 -€ 178,980 4,936 -€ 36

11,739

38,964

€0

€ 1,470,104

€ 1,470,104

1.34%

4.02%

1.69%

0

19,297

0

19,297

€ 7,254

€ 728,058

€ 735,312

0.87%

0.80%

0.86%

r=0.05




170 493 1,055 2,442

17,127 17,698 18,683 21,062


cost of infections (in 000s) € 441,537 € 456,260 € 481,655 € 542,966

total cost (in 000s) € 441,537 € 456,260 € 481,655 € 542,966

-€ 73


27,225


25,664


0.5


-€ 1,878,077

-€ 734,792

19,667

-€ 37




incremental averted CER cost infections (in 000s) (in 000s) -€ 1,469 303 -€ 5 -€ 16,192 874 -€ 19 -€ 41,587 1,859 -€ 22 -€ 102,898 4,237 -€ 24

0.5

23,352

10,085

33,437

€0

€ 862,013

€ 862,013

1.34%

4.02%

1.69%

0

16,824

0

16,824

€ 6,337

€ 433,731

€ 440,068

0.87%

0.80%

0.86%

-€ 421,945

16,613

-€ 25

Table 4: Cost-effectiveness analysis for baseline values of α=0.169 und δ=0.3 varying the discount rate (r) and the tattooing infection probability (.

15

Policy makers are not only interested in the optimal time point for launching intervention but also in an overall cost-effectiveness analysis of regulating tattoo parlours (started at t = 0) for the entire planning horizon (T = 15). The CER calculated for our baseline scenario is robust with respect to substantial variation in the parameter constellation  and discount rate r (cf. Table 4). With other parameters at their base case values, for any infection probability  in the range 0.018, 0.5], regulating tattoo parlours is a cost-saving intervention. The higher the infection probability via tattooing () is, the higher the cost savings per averted infections (CERs). Discounting approximately halves the CERs for  0.05 and r = 0.03 and cuts it by about two-thirds for r = 0.05. The total cost are strongly affected by discounting compared to the total HCV incidence, the effectiveness, because an infection at a certain time point engenders treatment cost over a long period. For our base case value of the infection probability via tattooing ( = 0.018), 230 HCV infections via tattooing and 210 secondary infections (440 averted infections – 230 infections via tattooing) could be averted (r = 0). Furthermore, about 23 Million Euro could be saved. With an increasing risk of acquiring an HCV infection via tattooing (), the prevalence in the tattooing population significantly increases compared to the non-tattooing population. For small values of , the prevalence in the tattooing population is actually a bit lower compared to the non-tattooing population because those getting tattoos are younger, on average, and so have been exposed to HCV risk for fewer years.

Conclusions The analysis above yields two conclusions that are fairly robust with respect to parameter uncertainty. (1) Regulating tattoo parlours was a cost-effective form of HCV control for Vienna in 2003 and, by extension, merits consideration in other jurisdictions. (2) It is possible to create a simple, dynamic, policy simulation that reflects both direct infections from tattoos and also secondary infections spread via additional tattoos and other, non-tattoo mechanisms. Hence, it would seem sensible for jurisdictions with both tattoo parlours and HCV to use this or a related model to inform decisions about whether and when to regulate tattoo parlours. The breakeven threshold prevalence for which regulation becomes cost-justified was low when our model was parameterised for Vienna, Austria. Even with an HCV prevalence as low as 0.8%, regulating tattoo parlours yielded a benefit-cost ratio on the order of 1.82. Moreover, starting regulation now would save policy makers at least 23 Million Euro and avert 440 HCV infections for our base case value of infection risk via tattooing ( = 1.8%) using a discount rate of r = 0% for a time horizon of T = 15. The results may or may not extend to developing countries because most of the social benefit for Vienna stem from avoiding quite expensive medical interventions and from extending the lives of people with high annual earnings. On the other hand, the cost of regulation was also driven by labour rates, specifically additional time required by tattoo parlour staff to sterilise equipment. So in countries with lower wage scales, the cost per sterilisation would also be lower. Various extensions are possible, including: more detailed modelling of subpopulations to see if tattooing may serve as a ‘bridge’ connecting what would otherwise be low- and highrisk subpopulations, crediting tattoo regulation with averting the spread of other blood-born viruses such as HIV/AIDS, and including risk-response of tattoo customers, whereby imperfect regulation might increase the number of people who seek tattoos and, hence, who are exposed to some reduced but still positive risk of infection.

References 1. Alter MJ (2002) Prevention of spread of hepatitis C. Hepatology 36(5 Suppl 1): S93-S98 2. American Gastroenterological Association (2003) Stigma of hepatitis C and lack of awareness stops Americans from getting tested and treated. AGA News Releases, June 2003 3. Anderson RM, May RM (1991) Infectious diseases of humans. Oxford University Press, Oxford 4. Armstrong ML, Murphy KP (1997) Tattooing: another adolescent risk behavior warranting health education. Applied Nursing Research 10(4): 181-189 5. Armstrong ML, Roberts AE, Owen DC, Koch JR (2004) Contemporary college students and body piercing. The Journal of Adolescent Health 35(1): 58-61 6. Armstrong GL, Simard EP, Wasley A, McQuillan GM, Kuhnert WL, Alter MJ (2006) The prevalence of hepatitis C virus (HCV) infection in the United States, 1999-2002. Annals of Internal Medicine 144(10): 705-714 7. Balasekaran R, Bulterys M, Jamal MM, Quinn PG, Johnston DE, Skipper B, Chaturvedi S, Arora S (1999) A case-control study of risk factors for sporadic hepatitis C virus infection in the southwestern United States. The American Journal of Gastroenterology 94(5): 1341-1346 8. Blaschke S, Purkert T (2004) Das ökonomische Potential von Präventionsmaßnahmen im Bereich Hepatitis C: Die Rolle der Tattoo-Studios. Master Thesis. University of Vienna, Austria 9. Centers for Disease Control and Prevention (1998) Recommendations for prevention and control of hepatitis C virus (HCV) infection and HCV-related chronic disease. MMWR Morbidity and Mortality Weekly Report 47(RR-19): 1-33 10. Centers for Disease Control and Prevention (2004) Hepatitis surveillance report no 59. U.S. Department of Health and Human Services, CDC, Atlanta, GA 11. Centers for Disease Control and Prevention (2005) Health Information for International Travel 20052006. Department of Health and Human Services, CDC, Atlanta, GA 12. Commonwealth of Australia (2005) National hepatitis C strategy. 2005-2008. Publication approval number 3642 (JN9005). ISBN 0-642-82656-0 13. Desenclos, JC (2003) The challenge of hepatitis C surveillance in Europe. Euro Surveillance 8(5): 99100 14. Deuffic-Burban S, Poynard T, Valleron AJ (2002) Quantification of fibrosis progression in patients with chronic hepatitis C using a markov model. Journal of Viral Hepatitis 9(2): 114-122 15. Deuffic-Burban S, Wong JB, Valleron AJ, Costagliola D, Delfraissy JF, Poynard T (2004) Comparing the public health burden of chronic hepatitis C and HIV infection in France. Journal of Hepatology 40(2): 319-326 16. Esposito N, Rossi C (2004) A nested-epidemic model for the spread of hepatitis C among injecting drug users. Mathematical Biosciences 188: 29-45 17. Federal Camber of Employees of the Republic of Austria (2004) Wirtschafts- und Sozialstatistisches Taschenbuch 2004. Federal Camber of Employees:.Vienna, Austria 18. Federal Ministry of Economics and Labour of the Republic of Austria (2003) Ausuebungsregeln für das Piercen und Tätowieren durch Kosmetik (Schoenheitspflege)-Gewerbetreibende. BGBl II (Austrian Federal Law Gazette, Part II) 141/2003: 647-651 19. Gold MR, Siegel JE, Russell LB, Weinstein MC (1996) Cost-effectiveness in health and medicine. Oxford University Press, New York 20. Haley RW, Fischer RP (2001) Commercial tattooing as a potentially important source of hepatitis C infection. Clinical epidemiology of 626 consecutive patients unaware of their hepatitis C serologic status. Medicine 80(2): 134-151 21. Haley RW, Fischer RP (2003) The tattooing paradox: are studies of acute hepatitis adequate to identify routes of transmission of subclinical hepatitis C infection? Archives of Internal Medicine 163(9): 10951098

22. Health Canada (2001) Special report on youth, piercing, tattooing and hepatitis C trendscan findings. Health Canada. Toronto 23. Jacquez J, Koopman J, Perry T, Sattenspiel L, Simon C (1988) Modeling and analyzing HIV transmission: the effect of contact patterns. Mathematical Biosciences 92(2):119-199 24. Jonas S, Jessner W, Rafetseder O, Wild C (2004) Chronische Hepatitis C – Implikationen für Therapie und ökonomischen Ressourceneinsatz in Österreich. Institute for Technology Assessment of the Austrian Academy of Sciences, Vienna 25. Kaplan EH, Brandeau M (eds) (1994) Modeling the AIDS epidemic. Raven Press: New York 26. Layden TJ, Layden JE, Ribeiro RM, Perelson AS (2003). Mathematical modeling of viral kinetics: a tool to understand and optimise therapy. Clinics in Liver Disease 7(1): 163-178 27. Mather D, Crofts N (1999) A computer model of the spread of hepatitis C virus among injecting drug users. European Journal of Epidemiology 15(1): 5-10 28. Mather D (2000) A simulation model of the spread of hepatitis C within a closed cohort. Journal of the Operational Research Society 51(6): 656-66 29. Nelson PW, Perelson AS (2002) Mathematical analysis of delay differential equation models of HIV-1 infection. Mathematical Biosciences 179(1): 73-94 30. Pollack HA (2001) Cost-effectiveness of harm reduction in preventing hepatitis C among injection drug users. Medical Decision Making 21(5): 357-367 31. Pybus OG, Charleston MA, Gupta S, Rambaut A, Holmes EC, Harvey PH (2001) The epidemic behavior of the hepatitis C virus. Science 292(5525): 2323-2325 32. Rauner MS (2002) Using simulation for AIDS policy modeling: benefits for HIV/AIDS prevention policy makers in Vienna, Austria. Health Care Management Science 5(2): 121-134 33. Rauner MS, Brailsford SC, Flessa S (2005) The use of discrete-event simulation to evaluate strategies for the prevention of mother-to-child transmission of HIV in developing countries. Journal of the Operational Research Society 56(2): 222-233 34. Rauner MS, Behrens DA, Caulkins JP (2004) An epidemic model of the spread of hepatitis C via tattoo parlors: implications for the timing of public health interventions. ORDYS working paper 288, Vienna University of Technology, Vienna, Austria (37 pages) 35. Salomon JA, Weinstein MC, Hammitt JK, Goldie SJ (2002) Empirically calibrated model of hepatitis C infection in the United States. American Journal of Epidemiology 156(8): 761-773 36. Salomon JA, Weinstein MC, Hammitt JK, Goldie SJ (2003) Cost-effectiveness of treatment for chronic hepatitis C infection in an evolving patients population. JAMA: The Journal of the American Medical Association 290(2): 228-237 37. Sheerin IG, Green FT, Sellman JD (2003) The cost of not treating hepatitis C virus infection in injecting drug users in New Zealand. Drug and Alcohol Review 22(2): 159-167 38. Siebert U, Sroczynski G, Rossol S, Wasem J, Ravens-Sieberer U, Kurth BM, Manns MP, Hutchison JG, Wong JB; German Hepatitis C Model (GEHMO) Group; International Hepatitis Interventional Therapy (IHIT) Group (2003) Cost-effectiveness of peginterferon -2b plus ribavirin versus interferon -2b plus ribavirin for initial treatment of chronic hepatitis C. Gut 52(3): 425-432 39. Statistics Austria (2004) Statistic Yearbook 2004. Statistics Austria, Vienna, Austria 40. Struve J, Norrbohm O, Stenbeck J, Giesecke J, Weiland O (1995) Risk factors for hepatitis A, B and C virus infection among Swedish expatriates. The Journal of Infection 31(3): 205-209 41. Wasley A, Alter MJ (2000) Epidemiology of hepatitis C: geographic differences and temporal trends. Seminars in Liver Disease 20(1): 1-16 42. Weusten JJ, van Drimmelen HA, Lelie PN (2003) Mathematic modeling of the risk of HBV, HCV, and HIV transmission by window-phase donations not detected by NAT. Transfusion 42(5): 537-348 43. Wiese M, Berr F, Lafrenz M, Porst H, Oesen U (2000) Low frequency of cirrhosis in a hepatitis C (genotype 1b) single-source outbreak in Germany: A 20-year multicenter study. Hepatology 32(1): 9196

44. World Health Organization (1999a) Hepatitis C – global prevalence (update). Weekly Epidemiological Record 74(49): 425-427 45. World Health Organization (1999b) Global surveillance and control of hepatitis C. Report of a WHO consultation organized in collaboration with the Viral Hepatitis Prevention Board: Antwerp, Belgium. Journal of Viral Hepatitis 6(1): 35-47