Needle and syringe programs in Yunnan, China yield health and financial return

Background As a harm reduction strategy in response to HIV epidemics needle and syringes programs (NSPs) were initiated throughout China in 2002. The effectiveness of NSPs in reducing the spread of infection in such an established epidemic is unknown. In this study we use data from Yunnan province, the province most affected by HIV in China, to (1) estimate the population benefits in terms of infections prevented due to the programs; (2) calculate the cost-effectiveness of NSPs. Methods We developed a mathematical transmission model, informed by detailed behavioral and program data, which accurately reflected the unique HIV epidemiology among Yunnan injecting drug users (IDUs) in the presence of NSPs. We then used the model to estimate the likely epidemiological and clinical outcomes without NSPs and conducted a health economics analysis to determine the cost-effectiveness of the program. Results It is estimated that NSPs in Yunnan have averted approximately 16-20% (5,200-7,500 infections) of the expected HIV cases since 2002 and led to gains of 1,300-1,900 DALYs. The total $1.04 million spending on NSPs from 2002 to 2008 has resulted in an estimated cost-saving over this period of $1.38-$1.97 million due to the prevention of HIV and the associated costs of care and management. Conclusion NSPs are not only cost-effective but cost-saving in Yunnan. Significant scale-up of NSPs interventions across China and removal of the societal and political barriers that compromise the effects of NSPs should be a health priority of the Chinese government.


Background
HIV epidemics in Asia were initially driven by injecting drug use and sex work. Waves of infection have historically occurred in these population groups, followed by infection among clients of sex workers and their regular sexual partners [1,2]. The first recognizable HIV outbreak in China occurred among injecting drug users (IDUs) in the city of Ruili, Yunnan province in 1989 [3], following which the epidemic rapidly expanded throughout Yunnan and neighboring provinces. By 2009, an estimated 740,000 people were infected with HIV in China, including 105,000 with AIDS [4]. Yunnan, a multi-ethnic province of China with a long history of opium and heroin trade and high prevalence of illicit drugs [5], have accounted for over one-quarter of all HIV cases in China [6,7]. HIV in Yunnan has primarily spread through intravenous drug use with a high annual incidence rate of 2.2%-8.0% [8]. These intravenous drug users are young and approximately 80% of them are in their 20 s and 30 s [9]. In 2002, over a decade after the epidemic commenced, needle and syringes programs (NSPs) were initiated by various agencies throughout China as a harm reduction strategy. Currently it is estimated that more than 100 needle and syringe exchange sites are operating throughout Yunnan province alone and a total of 2.5 million syringes were distributed in 2008 [10,11]. Despite the large investment in NSPs, it is estimated that less than 25% of IDUs in Yunnan obtain their injecting equipment through NSPs [12,13] and rates of sharing of injecting equipment remains as high as 45% [14]. NSPs have been shown to be a safe and effective means to reduce HIV transmission in some developed and developing country settings [15][16][17][18][19][20]. Therefore, it is important to investigate whether the same degree of effectiveness has been achieved in mitigating the spread of infection in a setting such as China, with an established HIV epidemic among IDUs.
In this study we aim to (1) estimate the population benefits that NSPs in Yunnan have likely had in preventing HIV infections and related health outcomes among IDUs; (2) calculate the cost-effectiveness of NSPs from a governmental perspective. We estimate the expected number of infections averted due to NSPs through the development of a mathematical transmission model. The model uses data on the numbers of units of injecting equipment distributed by Yunnan NSPs and behavioral and clinical data from Yunnan province (such as rates of needle sharing between IDUs, rates of disposal of used needles, frequency of injecting, natural history of HIV disease progression in infected IDUs, initiation of antiretroviral therapy), coupled with biological data from the international literature and local HIV epidemiological data. The model is calibrated to accurately reflect the unique HIV epidemiology (incidence, prevalence and aggregate reported number of relevant clinical outcomes including deaths) among Yunnan IDUs and then it is used to estimate the likely epidemiological and clinical outcomes if NSPs were not present. We also conduct a health economics analysis by comparing the financial costs associated with implementing NSPs with the costs saved due to averting infections to determine how cost-effective the program has been from a societal perspective. We consider the cost-effectiveness over the last 7 years and also lifetime value of the program.

Method
A mathematical transmission model was developed to describe the HIV epidemic among IDUs in Yunnan. The model is based on a system of ordinary differential equations [21] and is used to track the rates of HIV transmission from infected IDUs to susceptible IDUs through injecting-related risk behavior and the natural history of infection for HIV-infected individuals as presented in the schematic diagram in Figure 1. This is well-developed epidemiological model that has been used in numerous occasions and has been shown to accurately reflect the dynamic of HIV epidemic [21]. This standard schematic diagram is mathematically translated into 4 ordinary differential equations, one equation for each health state. The health states represented in the model are: uninfected and potentially susceptible individuals (S), and HIV-infected individuals that are either in the asymptomatic chronic stage (I), treatment-eligible stage (CD4 count < 200 cells/μl) (T E ) or are receiving antiretroviral treatment (T). The differential equations to describe the change in the number of people in each of these disease states are: As such, the model tracks the numbers of HIVinfected people as they progress in disease through the long chronic/asymptomatic infection to late-stage infection; treatment-eligibility in China is based on progression to late-stage infection (CD4 count < 200 cells/ μl [22]) and this is what is described in this model. The rates of disease progression are assumed to be constant over time and represent population average rates. Effective antiretroviral therapy (ART) is assumed to reduce viral load in treated individuals and consequently decrease the likelihood of transmitting the virus (by 95% on average, assuming the same relative reduction in infectiousness as in other modes of transmission [23]).
The number of IDUs in the population is defined as N and it is assumed that each IDU injects an average of n times per year. A proportion, s, of all IDUs may share their syringes with others and they do so in a proportion, q, of their injections. It is assumed that if sharing of injecting equipment occurs then it happens between two IDUs. The probability of infection from a contaminated syringe per use is denoted as b I , but people who are on treatment are expected to have a lower probability of transmission b T due to suppressed viral load [23]. In the model, syringe cleaning has effectiveness ε C and cleaning occurs before an average proportion of p C shared injections. Given these definitions, the proportion of sharing events in total per person per year is nsq/2 and each susceptible person could acquire infection with probability (1p C ε C ) b I or (1p C ε C )b I depending on the HIV stage of his/her injecting partner. Therefore, considering the proportion of the entire population in each disease stage, the force of infection l, which estimates the average per-capita risk of infection [21], can then be calculated as Official figures over the past decade indicate that the number of registered drug users in Yunnan varies between 50,000-70,000, among which~55% are IDUs [14,24,25]. The actual number of IDUs may be much higher as it is widely accepted that in China there are approximately 4 implicit IDUs behind every registered IDU [26,27]. In this study a sensitivity analysis is carried out on the IDU population size such that the number of IDUs is (i) 2.5 times the registered number; or (ii) 4 times the registered number.
IDUs in Yunnan may obtain syringes through both commercial means and NSPs, but the majority (~75-95%) are obtained commercially [12,13]). Therefore, we assume that the majority of NSP-based distribution of sterile needles and syringes adds to the total numbers that are in circulation. If P M and P N commercial and NSP syringes are distributed each year through these means, respectively, and a proportion ω of all syringes are not used, then the number of syringes distributed that are used is (P M + P N )(1 -ω). In this model it is assumed that syringe distribution by NSPs (P N ) increases linearly since its initiation but commercial syringe distribution (P M ) remained constant throughout these years. If a non-shared needle is used δ p times on average before disposal and a shared needle is used an average of δ s times before disposal then defines a relationship between the total number of needles distributed and their use. This relationship was defined previously by Kwon et al [15]. Several factors may change due to changes in the number of needles distributed. These include: the proportion of syringes that remain unused (ω), the proportion of IDUs who share injecting equipment (s) and the prevalence of sharing injecting equipment (q), or the average number of times each syringe is used before it is disposed (δ p , δ s ). Changes in ω and δ p will not influence transmission levels of the virus but changes in s, q and δ s could potentially result in significant changes in HIV incidence. Previous studies have demonstrated that the average number of times a syringe is used before it is disposed varies very little [28][29][30][31] and that IDUs are unlikely to share if they have access to clean injecting equipment [32]; therefore, it is assumed that sharing rates (q) change with syringe distribution (according to equation (6)) and other values remain constant. Parameter estimates used in the model are presented in Table 1.
The model was solved in Matlab and numerical solutions yielded estimates of: the annual number of new HIV incident cases, the number of treatment-eligible HIV patients, the number of people on ART, the number of HIV/AIDS-related deaths, and the total size of the HIV-infected population under conditions of actual distribution of needles and syringes by NSPs and predicted estimates if NSPs had not existed. The model was calibrated to reflect available epidemiological data, mainly HIV prevalence. The calibration was performed by adjusting biological and epidemiological parameters inside the model within their corresponding uncertainty bounds until the model-generated HIV prevalence curve closely mimics the trend of actual data. Retrospectively the model was then used to estimate the expected trajectories of the HIV epidemic in China in the absence of NSPs under the assumption that the reduced number of units of injecting equipment affected sharing rates (according to the supply-demand relationship of equation 6). We estimated the number of disability-adjusted life years (DALYs) gained, and number of HIV cases and deaths averted due to NSPs in Yunnan.
A health economics analysis was also carried out to estimate the cost-effectiveness of the programs for both periods of 2002-2008 and lifetime of IDUs. The lifetime economic impacts of NSPs are evaluated by summing the future costs of currently survived HIV-infected IDUs beyond 2008, without consideration of new future infections. The epidemiological model results became inputs in the health economic analysis. The amount of money invested in NSPs was compared with the estimated costs saved due to averting infections that would have resulted in expenses for monitoring, care, treatment. The costs of NSPs during the period 2002-2008 is calculated by multiplying the average unit expense of distributing a syringe ($0.11 USD), which incorporates and averages over all necessary infrastructure, personnel, marketing and recurring service costs [33], to the estimated total number of syringes distributed. A conservative approach was taken with the inclusion of costs  [73][74][75]. Among these registered drug users, 55% are intravenous drug users period [14,24,25], which corresponds to an increase of~30,000 to~40,000 registered IDUs in the last decade. It is widely accepted that in China behind every registered IDUs there is about 2.5-4.0 implicit IDUs that are unregistered [26,27]. The total number of IDUs in Yunnan is estimated to lie between 95,000 and 125,000 if the scaling factor equals to 2.5, whereas the population size is between 150,000 and 200,000 if scaling factor equals to 4. c. The entrance rate of IDUs is calculated from the variation of population size, whose minimum and maximum bounds are estimated to be 30,000 and 50,000 over the period 2002-2008 based on the above scenarios. Therefore, an average entrance rate is estimated to be 6000. d. The cumulative number of syringes distributed in Yunnan during the period 2002-2008 is approximately 875,000. This number is obtained through personal communication with stakeholders from China CDC. e. Given that the average usage of a syringe (δ t ) among Yunnan IDUs is about 3 [28][29][30][31], the average usage of a non-shared syringe (δ P ) is assumed to be less than 3 and greater than 1. Therefore, the average usage of a shared syringe (δ S ) can then be estimated to be 3-15 by equation δ t = s·q·δ s + (1 -sq)·δ P .
related to routine HIV tests, viral load and CD4 testing and any antiretroviral treatment programs that were supported or subsidized by the Chinese government under its "four free one care" policy [34]. The costs of the programs were calculated by multiplying the unit price of service items, which are determined from published literature and policy documents, to the quantities of provision (Table 2). All costs associated with program investment, infection, treatment and care are listed in Table 2. Undiscounted and 3% discounting analyses were performed.

Results
The epidemiological model accurately resembles the epidemiological trends of HIV prevalence in Yunnan during the past decade (Figure 2, 3). The model estimates that according to scaling factor (SF) assumptions of the total IDU population size being 2.5-times ( Figure 2) or 4-times ( Figure 3)   were spent on treatment-eligible patients who are not yet receiving antiretroviral therapy (ART), whereas the remainder was spent on asymptomatic patients (24%) and those on ART (20%). However, if NSPs did not exist, the model estimates that the Chinese government would have spent an extra USD$1.44 m (SF = 2.5; USD $2.06 m for SF = 4) on HIV/AIDS patients care and management over the past decade due to the higher HIV incidence. Since this saving of healthcare expenses is greater than the initial investment, these programs are not only very cost-effective but cost-saving even over this short time period. The investment in NSPs in Yunnan since 2002 will yield even greater savings in the long term. In the lifetime of the current population, the benefits of averting infections over the last 10 years will lead to large future savings, of USD$27 m (SF = 2.5; $39 m for SF = 4) (3% discounting) over the population lifetime.
The return on investment of the NSPs in Yunnan was also assessed with a cost-utility analysis. It is estimated that the cost for each DALY gained since 2002 is approximately USD$753 (SF = 2.5; USD$527 for SF = 4), but the value reduces to USD$82 (SF = 2.5; USD$57 for SF = 4) if the lifetime impacts of the program are considered. Healthcare programs are generally considered cost-effective if the cost per DALY saved is less than the per capita gross national income (GNI) [35]. The Chinese per-capita GNI is USD$3,650 in 2009 [36]. Additionally, with an estimate of NSP investment 1.04 million, the directly gained economic benefit is measured to be 1.4-2.0 million during 2002-2008, which corresponds to a benefit-cost ratio of 1.38-1.97 (SF = 2.5, 4 respectively, Table 2). Thus, this program is extremely cost-effective compared to most healthcare programs. In  [76,78] respectively., b. The costs of NSPs during the period 2002-2008 is calculated by multiplying the average unit expense of distributing a syringe ($0.11 USD), which incorporates and averages over all necessary infrastructure, personnel, marketing and recurring service costs [33], to the estimated total number of syringes distributed (8.75 × 10 6 ). c. Regular viral load monitoring is currently undertaken once a year for HIV/AIDS patients on ART to monitor potential change in viral load. Its cost is hence the product of the number of patients on ART and its unit cost (~USD$300/person [79,80]) and is calculated with 3% value discounting. d. CD4 load tests are performed quarterly for patients on ART [80][81][82] and twice a year for people diagnosed with HIV but not on ART [80,82]. The unit cost of CD4 load test is USD $42 [83]. The total cost is calculated with 3% discounting. e. The annual cost of ART for each AIDS patient is approximately USD $3000 [83][84][85][86], which is multiplied by the number of patients on ART to obtain the total cost. The total cost is calculated with and without value discounting. f. Each year, the government subsidizes approximately USD $340 [87] of healthcare associated with opportunistic infections of symptomatic AIDS patients (those that are in the treatment-eligible stage or potentially on ART (but experienced treatment failure) in our model). Its total yearly spending is calculated as the product of the two with 3% value discounting. g. Each year, the government subsidizes approximately USD $16 [87] of herbal treatment for each symptomatic AIDS patient (treatment-eligible patients and patients on ART). Its total yearly spending is calculated as the product of the two with and without value discounting. h. Each HIV asymptomatic patients are CD4 tested twice a year [80,82], the governmental investment on each individual HIV patients is USD$42 × 2 = USD$84. This amount is multiplied by the total number of asymptomatic patients and then summed with the cost of one-off diagnosis testing of newly infected cases to result in the total governmental investment in this specific target group. i. Each HIV-infected patient yet to receive ART will receive CD4 testing twice each year [80,82]. Treatment of opportunistic infections and Chinese herbal treatment for AIDS are also covered or subsidized by the government [87]. Each patient on ART will receive 1 viral test and 4 CD4 tests each year [79][80][81][82]. In addition to government-subsidized treatment for opportunistic infections and herbal treatment for AIDS, they also receive ART on governmental expenses [85].
fact, not only are NSPs in Yunnan cost-effective, but we have calculated that they are actually cost-saving. That is, the return in healthcare savings is greater than the initial investment. Specifically, for every one dollar the government invested into NSPs, it was estimated the government would gain back that dollar plus an additional USD$0.38 (SF = 2.5; USD$0.97 for SF = 4) due to the reduced number of HIV-related tests and treatments. The return on investment is even greater when the lifetime effects of these programs are taken into account:~USD$15 (SF = 2.5; USD$36 for SF = 4) are returned for every USD$1 invested over the past 7 years. We also estimated that it cost USD$198 (SF = 2.5; USD$138 for SF = 4) of investment into NSPs to avert a new HIV infection.

Discussion
We estimated that the spending of a total USD$1.  This is comparable to a similar investigation carried out in 2006 in Ukraine over its one year harm reduction intervention, which consequently lead to a 22% decrease in IDU HIV incidence and 1% decrease in prevalence [16]. IDU-initiated HIV epidemics remain as confined epidemics and have not yet generalized to other populations in China.
The current study is an important study that evaluated the cost-effectiveness of NSPs for IDUs in China. We have shown the programs to be cost-saving. The fact that the benefit-cost ratio increases substantially over the lifetime of the population reflects the long-term impacts of the programs. Our estimate of USD$138-198 in spending per infection averted is consistent with similar studies in other locations [16,37,38]. By adjusting the cost to account for the purchasing power parity for China (2007 PPP factor was 4.09 relative to 1 USD), the cost translates to $560-810 per infection averted, which is still very cost-effective in comparison to many developed countries ($3,000-$20,000 [39][40][41][42]).
Our analysis involved a conservative calculation of the cost-effectiveness of NSPs by only including governmental investment in HIV/AIDS alone. The actual savings will be much higher for a number of reasons. First, hepatitis C virus (HCV) has a very high prevalence among Chinese IDUs (55-80% [43,44]). A 2009 report on the cost-effectiveness of Australian NSPs demonstrated that the majority of savings by NSPs are actually due to HCVrelated healthcare expenses saved and not HIV [45]. Second, other medical costs such as those due to mental health episodes, injecting related injury, psychosocial benefits, overdose education and prevention may further increase the return on investment. Third, we did not include patient/client costs or productivity losses or gains. AIDS patients are contributing a substantial outof-pocket payment while accessing government-supported services, including transportation to clinic/hospital, consultation fees and hospitalization expenses. Therefore, our calculations should be regarded as a very conservative estimate of the actual expenses saved.
Several limitations of the current study should be noted. First, the cost data used in this study was collected from a number of sources from Yunnan and also other provinces [39][40][41][42][43][44][45]. Treatment-related cost variations may occur due to differences in economic development in different provinces. However, these differences are considered to be small as the treatment prices are standardized in governmental health institutions under the national "four free one care" policy. Second, the current study is based on ART eligibility at a CD4 threshold of 200 cells/μl. China has recently adopted a new guideline to initiate treatment at 350 cells/μl [46]. This will likely significantly increase the potential medical expenses for ART patients in their lifetime. Besides, the current study does not take into consideration of secondary impacts of NSPs, e.g. reduction of infections among sexual partners of IDUs due to decrease of HIV prevalence among IDUs as a result of NSPs. Therefore, our analysis should be regarded as a conservative estimate of the actual costeffectiveness of NSPs in Yunnan. Third, this study used a simple deterministic approach and did not include a detailed uncertainty analysis. However, we took into consideration the variation in IDU population size, which represents the largest and most important uncertain parameter. Results from this analysis should be regarded as an estimation of the average effectiveness and cost-effectiveness of NSPs in Yunnan. Finally, a notable limitation of the current epidemiological model is that it does not include the quitting rate of IDUs, which may vary the population size substantially. We assume that an IDU leaves the population only through death.
The current coverage of NSPs in the province remains low, as less than 25% of IDUs access NSPs. This suggests that it is possible to have even greater health and economic gains by expanding the programs further in the future. Large changes in government attitudes occurred from viewing drug users as criminals to largescale adoption of needle and syringe provision [47]. Ongoing police raids and confinement of drug users cause increasing fear of police arrest and reluctance to access to sterile equipment from NSP sites or peer educators. Compulsory detention of IDUs in China in detoxification centers and labor camps largely limits the accessibility of IDUs to NSPs [48,49]. Hence, enhancing cooperation between multiple institutions including the Ministry of Health and Public Security and Justice is necessary for further scale-up of the programs. Other issues include reducing social stigma against drug users and maintaining users' anonymity in NSP sites. If such barriers can be removed and NSPs expanded then there is strong potential for the large epidemiological and economic benefits to increase even more substantially. Whilst these results are specific to Yunnan, China, the qualitative conclusions are generally applicable to other settings that are considering commencing or expanding NSPs where HIV is endemic among IDUs. NSPs not only save lives and health outcomes but they are also a valuable economic investment.

Conclusion
Conclusively, NSPs are not only cost-effective but costsaving in Yunnan since the implementation of the programs and will have greater epidemiological and economic benefits if in the life span of IDUs. Significant scale-up of NSPs interventions across China and removal of the societal and political barriers that compromise the effects of NSPs should be a health priority of the Chinese government.