Population Size Estimation of People Who Inject Drugs in Selected High Priority Countries: Review of Current Knowledge © International Bank for Reconstruction and Development / The World Bank 1818 H Street NW, Washington DC 20433 Internet: www.worldbank.org; Telephone: 202 473 1000 This work is a product of the staff of The World Bank. The findings, interpretations, and conclusions expressed in this work do not necessarily reflect the views of the Executive Directors of The World Bank or other partner institutions or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. Nothing herein shall constitute or be considered to be a limitation upon or waiver of the privileges and immunities of The World Bank, all of which are specifically reserved. Rights and Permissions This work is available under the Creative Commons Attribution 3.0 Unported licence (CC BY 3.0 IGO) https://creativecommons.org/licenses/by/3.0/igo/. Under the Creative Commons Attribution license, you are free to copy, distribute and adapt this work, including for commercial purposes, under the following conditions: Attribution – Please cite the work as follows: The World Bank. 2016. Population Size Estimation of People Who Inject Drugs in Selected High Priority Countries: Review of Current Knowledge. Washington DC: World Bank. License: Creative Commons Attribution CC BY 3.0 IGO. Translations – If you create a translation of this work, please add the following disclaimer along with the attribution: This translation was not created by The World Bank and should not be considered an official World Bank translation. The World Bank shall not be liable for any content or error in its translation. All queries on rights and licenses should be addressed to the Office of the Publisher, The World Bank, 1818 H Street NW, Washington DC, 20433, USA; fax: 202-522-2625; email: pubrights@worldbank.org. i Table of Contents Preface and Acknowledgments ....................................................................................................... iii List of Abbreviations ......................................................................................................................... v Background and Objectives .............................................................................................................. 1 Methods used for estimating PWID populations .............................................................................. 3 Reported Country Estimates and Methodologies ............................................................................. 8 Country-Specific Recommendations................................................................................................. 9 Conclusion ......................................................................................................................................20 ii Preface and Acknowledgments This report investigates the methodologies used to obtain current population size estimates of people who inject drugs (PWID) in China, India, Philippines, Myanmar, Tajikistan, Kyrgyzstan, Uzbekistan, Kazakhstan, Belarus, and Libya. In this synthesis report, current population size estimates are presented, with evaluations of accuracy, based on the type of methodology used. Recommendations are made to build data capture or infrastructure that will allow for the most appropriate methodology to update estimates of the country’s PWID population, if necessary. We hope that this review will be useful in prioritizing and allocating resources to gain a better understanding of the PWID population size and the mobilization of resources to change the dynamics of the epidemic in the respective countries. We gratefully acknowledge support received from Pandu Harimurti (The World Bank) and Riku Lehtovuori (UNODC/Vienna). Waimar Tun (Population Council) Nancy Ralph (Independent consultant) Scott Geibel (Population Council) Henry Fisher Raymond (San Francisco Department of Public Health) Avina Sarna (Population Council) Susie McLean (International HIV/AIDS Alliance) Tetiana Salyuk (International HIV/AIDS Alliance) iii iv List of Abbreviations AIDS Acquired immunodeficiency syndrome HIV Human immunodeficiency virus HRI Harm Reduction International IBBS Integrated Biological and Behavioral Surveillance PWID People who inject drugs RDS Respondent-driven sampling UNODC United Nations Office of Drugs and Crime WDR World Drug Report WOTC Wisdom of the Crowds v Executive Summary This document summarizes the state of knowledge of the injecting drug use population sizes in 10 priority developing or transitioning countries (Belarus, China, India, Kazakhstan, Kyrgyzstan, Libya, Myanmar, Philippines, Tajikistan, and Uzbekistan). Estimates of the size of PWID are critical as they are needed to understand the magnitude of the drug use epidemic, allocate adequate resources, design and implement programs, monitor the coverage and impact of programs, and advocate for better policies and programs. Information was gathered from peer-reviewed and grey literature for information related to size of the PWID populations in these countries, as well as key staff from relevant international organizations (USAID, CDC, UNAIDS, UNODC) and implementing organizations (International HIV/AIDS Alliance, FHI360). Additionally, for the four Central Asian countries (Kazakhstan, Kyrgyzstan, Tajikistan, and Uzbekistan), UNODC convened a stakeholder meeting with representation from government (including Ministry of Health and drug control agencies) and NGOs implementing PWID size estimation data collection in June 2014. The meeting entailed gathering information on current size estimation activities from country representatives and developing concrete recommendations for next steps in each country. Overall, with the exception of India, these countries lack the infrastructure and resources for routine and timely data capture and analysis to inform policy and programs adequately. On the other end of the spectrum from India is Libya, where there has been a recent surge in HIV in PWID and there has been virtually no information on the size estimates of PWID. In such a case, there is an immediate need for population size estimation. While size estimates for many of the priority countries could be found in the literature, generally, the methodologies were not well- specified, which brings into question the accuracy of these estimates. In some cases, estimates were based on Delphi method of expert self-reporting or only one method of estimation had been used, which reduces the robustness of the estimates. Basic needs for data and estimation include sub-national and national estimates and breakdown of population estimates by age and sex. Country-specific recommendations are made on next steps to improve estimates, and include recommendations to standardized estimates for comparisons across countries and to ensure that multiple estimates from multiple methods are used to determine a range for the estimate. Where possible, recommendations for data collection build on existing infrastructure. First, we recommend using routine or sentinel surveillance data registers from multiple PWID services, where possible and relevant, as a cheaper and faster way to get rough estimates of PWID population sizes. Second, we recommend incorporating PWID population size estimation data collection into existing or planned integrated biological-behavioral surveillance (IBBS) surveysas a low-cost opportunity to derive multiple size estimates. This method is considered low-cost because it only requires adding a few questions into the survey and/or adding unique object distributions. Third, in countries where there is no IBBS or where incorporating size estimation with the next IBBS is not feasible, we recommend conducting a simplified respondent-driven sampling (RDS)-based survey, which incorporates various size estimation methods. Though assessments of feasibility and cost-effectiveness of this method might be needed, this simplified survey would forego the normal lengthy behavioral survey and sero-testing, which would greatly reduce the time and expense needed to implement the survey. In addition, some global health organizations have been conducting frequent behavioral tracking surveys (similar to a simplified IBBS survey), and size estimation components can be added into these surveys if planned ahead of time. We also advocate for incorporating size estimation exercises into national surveys where appropriate (e.g., AIDS Indicator Survey, Demographic and Health Survey; Expenditure Survey). This would provide more data in a cost-saving way for triangulation and strengthening the estimates. 1 Background and Objectives Worldwide, according to UNODC World Drug Report, the 2013 estimate for the size of the people who inject drugs (PWID) population is 14 million people between the ages of 15 to 64.1 This is down from 15.9 million in 2008 which may reflect improved availability of more reliable data as well as changes in injecting behavior.1 Globally, PWID are disproportionately affected by HIV, with HIV prevalence in this population being significantly higher than the prevalence in the general population. For example, the 2012 estimates of global new infections for Eastern Europe and Central Asia indicate that HIV infection among PWID and their sexual partners accounted for 40% of new HIV cases; in the Philippines, 36% of new infections were attributed to PWID.2, 3 As a consequence, PWID are a population of interest in HIV prevention, in assessing local and global populations at risk, and in informing prevention and treatment efforts going forward. To date, data on the size of the PWID population in many countries, including countries where HIV prevalence is high or growing, are scarce or inaccurate. The drug-using population is a difficult population to assess. Their illegal and stigmatized behaviors make estimates from surveys of general populations inaccurate, and increase bias in many data sources. Reliable estimates of the size of PWID populations are critical to understanding the magnitude of the drug use epidemic, allocating adequate resources, designing and implementing programs, monitoring the impact of programs, and advocating for better policies and programs for PWID. Population size estimates are also critical to projection and estimation of HIV infections, and HIV infections averted under certain intervention programs, through mathematical modeling. Finally, size estimates are an important part of strategic information as accurate estimates are needed to set appropriate program targets and assess the adequacy of program coverage. Some countries have initiated the recommended Comprehensive Package of HIV services (WHO/UNODC/UNAIDS), which includes interventions among PWID. These interventions require knowledge of sizes for planning purposes, and places where the intervention activities should target. Different methods have different advantages and limitations and each method might be more relevant for certain settings and purposes, which will be described in the next chapter. In addition, without more robust estimates of the PWID population, it would be challenging to determine the coverage and impact of the current programs targeting this population. Size estimations are also needed by some countries to justify initiating PWID interventions in countries where current resources are inadequate. This document summarizes population estimates for PWID in ten priority countries selected by The World Bank and UNODC, and outlines methods used for estimation. Recommendations for additional data collection and analytical methods are provided by country. 2 Overview of Key Methods Used for Estimating the Size of PWID Populations There is no gold standard method for estimating the size of PWID populations and choosing a method depends on several factors, including cost and time, what the data will be used for, what data are already available, and the legal environment. Selection of methodologies should be informed by formative research in the local contexts where size estimations are required. 4 In this section, we present some of the most common methods that have been used to estimate the size of key populations, including PWID. We aim to provide a brief description of select methodologies, data requirements, and advantages and limitations for each of these methods. Details for some of these methods can be found in the UNAIDS and WHO Guidelines on Estimating the Size of Populations Most at Risk to HIV4. A summary table of the methods discussed in this section are presented in Appendix I. CENSUS AND ENUMERATION METHODS Censuses are efforts to count all members of a target population. For PWID, however, it is not possible to count all PWID within population households, and a census of PWID would likely need to be venue-based. Venues where PWID congregate are identified by individuals familiar with the local context at all “hotspots” (i.e., congregation venues, injection sited, etc.) and identified areas are mapped. These identified venues are then accessed and PWID are then counted. Reported numbers may serve as the lowest number in the estimate range, as members of these groups can remain hidden. This method is preferred for sub-locations or small geographical areas. Together with location coordinates collected from mapping, these data are useful for targeted intervention programs. Enumeration is similar to census, but instead of counting every individual at every site or hotspot, a sample of sites are chosen from a sample frame or list of venues, and only individuals within those chosen venues are counted. Final numbers are then inflated to reflect the estimate of all sites (in the selected region or location). For example, from a list of hotspots in a district, select and visit one thirds of the hotspots to get an average number of PWID in these selected hotspots. Then multiply by 3 to get an estimate of PWID in the whole district. Enumeration method is used when it is only possible or feasible to reach a fraction of the target populations due to time, budget, or access constraints. CAPTURE-RECAPTURE METHOD The capture-recapture method derives a population size estimate using two different population counts within a defined area.4-6 In the first count (capture), PWID are counted and “tagged” with a unique object so they may recall easily that they were counted. A second count (recapture) is conducted after a minimal time period, for instance, 1-2 weeks. A mathematical formula is then used to estimate of total population size [Estimated Size = (number in first capture x number in second capture) / number in both captures]. The key steps in capture-recapture are below: • teams visit and map all hotspots • enumerators count all PWID in each hotspot and distribute some card/ gift (something unique and easy to identify)- “tag” • enumerators re-visit the same sites one or two weeks later and count of all PWID and ask whether they received the token/card/gift (count both previously tagged and untagged) Capture-recapture assumes: 3 • No major migration between the two captures (a closed population). • It is possible to identify and match individuals in the two captures. • Every member has an equal chance of being included in either count. • The two samples are independent from each other, meaning that the probability of identification in the second count is not biased based on their identification in the second count. • The sample size for each capture needs to be large enough to be precise. These assumptions may pose challenges during implementation. For example, PWID may move rapidly between locations, or police may conduct raids in PWID hotspots between counts. This, of course, may affect the probability of being re-captured. In addition, the estimate is only reliable when the number of matched individuals in the two captures is large enough. However, this method is easy to implement and has been used successfully in many countries to estimate the size of PWID and other stigmatized and/or criminalized populations. MAPPING AND INTERVIEWS OF KEY INFORMANTS This method was developed by researchers with the University of Manitoba7, and has been implemented successfully in several countries in Asia and sub-Saharan Africa. It includes a geographic mapping of key locations where target populations congregate. Preliminary steps of the geographic mapping involve developing or acquiring maps of the targeted area, segmentation of the target area into smaller geographic zones to facilitate field work data collection. After the mapping step, interviews with primary and secondary key informants are conducted to verify the list of hotspots, estimate number of target population at each hotspot, and identify peak time and date at each location for the next step. At the second step, enumerators visit each hotspot at the peak time and date, and interview primary key informants to get an estimate of minimum and maxim number of target population. Average, minimum, and maximum numbers are then calculated. Similar to capture-recapture, this method may likely miss segments of the target populations that do not congregate or spend much time in venues. In addition, it relies on informant estimates rather than a actual counts of target population at the spots identified, which may lead to biases in the estimates. A key advantage of this approach is that it is easy to implement and the results tend to be informative for program implementation. The mapping data can be used in HIV programming for a variety of purposes, including planning, outreach activities, peer-education activities, and setting up services to meet the needs of the target population. SERVICES MULTIPLIER This method requires at least two aggregate data sources.4 The first data source is obtained from a program that provides services to PWID. The other is the proportion of PWID receiving services from that particular program obtained through a population-based survey. As an example, Source 1 might be the number of PWID captured in opioid substitution programs. Source 2 might be the prevalence of PWID who enroll in opioid substitution programs, discovered through a survey based on probability-based sampling.a For example, if 100 people are registered in an opioid substitution program, and a representative sample of PWID show a 20% enrolling at the same opioid substitution program, the estimate would be: 100 [people registered based on service statistic] x [the inverse proportion of survey population indicating registration (100/2)] = 5,000 PWID The underlying assumptions for this method are: a The survey should be ideally conducted among PWID using probability-based sampling to reduce biases. 4 • The first data source should be complete-- in our example, the enumerated population must include the count of all PWID in the opioid substitution program. • The second data sources should be derived from a representative sample. • The two data sources must define PWID in the same way. • The definition and the timeframe for the first data source should be the same as the definition and the time frame in the second data source. For example, if the program data counts PWID seeking services for a full year of 2012, then survey should ask participants if they seek services at that program during the full year of 2012. • Both data sources must be independent. This method can be advantageous as researchers can make use of a wide range of available data sources if they are available. Program data can be obtained at relatively low cost; while second source of data (multiplier) can be incorporated in current or up-coming surveys. In addition, several service multipliers (more than two) can be used on the same survey to strengthen the estimates (e.g., producing a median estimate and range). The choice of service multipliers is determined by the availability and quality of the unduplicated count available in each location during the formative phase of the project. UNIQUE OBJECT MULTIPLIER This method uses 2 different data sources and the formula to estimate the size is similar to the capture recapture method. First multiplier: 1 week before the survey, peer educators from community outreach projects distribute simple object to people they identify as PWID. Each person receives only 1 object, and will be asked to keep the object (not to distribute the object to their peer). The goal will be to distribute as many of the unique objects as possible. The second multiplier: will be enumerated during a survey (ideally RDS survey) among PWID in the same catchment areas where objects were distributed. The RDS survey will ask each participant the following questions: o Have you received this object in the past 1 week (or applicable time period)? o From whom did you receive it (if they have received a watch from someone other than a peer educator, they will be eliminated from the multiplier). o Where did you receive it? (To verify the object was received in the catchments area). The unique object identifier method is advantageous in that it is controlled by the survey team and may reduce potential bias. This method is favored when combining with a probability sampling survey (for example with an RDS survey). WISDOM OF THE CROWD Wisdom of the crowd (WOTC) is based on the assumption that, when asked to estimate the number of members in a population of interest, the average response of a surveyed population approximates the actual number in the population of interest.8 This method relies on the knowledge of members of the target population. For example, in an IBSS survey among PWID, survey participants were asked to estimate the average number of PWID in a particular location. The median of the responses is considered the size of the PWID population for that particular location. While the most desirable use for WOTC is within a population-based surveillance survey, it is also sometimes may be used among venue-based or non-probability samples (such as with key informants). 5 DELPHI METHOD In Delphi methodology, an estimate is generated by systematically soliciting, sharing and collectively reviewing selected expert knowledge on the size of a target population (i.e. PWID in this case). Often, this process involves a series of expert-panel meetings where these experts are asked to estimate how many PWID are in a particular location. Each estimate or opinion undergoes a series of reviews by the other experts. Authors are encouraged to reconsider their estimates in light of feedback. The accuracy and quality of Delphi estimates may vary widely. This is dependent on the availability (or lack of) data sources. Often the Delphi method is employed when systematic size estimation data is currently lacking, and thus based on limited information. NETWORK SCALE-UP (NSUM) Network scale-up estimates are generated by asking a sample of the general population about people known to them -- their “network” – and how many engage in the behavior of interest, in this case, injection drug use. The below information must be collected: 1. The average network size (person they know/acquaintance) of the survey participants. 2. In their personal network, the percent of people they know who inject drugs. 3. Size of the general population for a particular geographical location where you want to estimate PWID size. For example, if the average network size from a survey is 50, and the average percentage of PWID they know in the networks is 0.1%, then the estimated size of PWID in City A with a total population of 500,000 will be: (0.1 x 500,000) / 50= 1000. This method can be incorporated into ongoing national household surveys such as the Demographic and Health Surveys or AIDS indicator surveys. Advantages include that it generates an estimate from a sample of the general population rather than the target population. This is particularly important because individuals are more comfortable reporting sensitive behaviors of others than their own. The limitations are that in certain places, PWID might not associate with the general population, and thus are not known as PWID to the survey respondents. In addition, estimating network sizes can be challenging. Additionally, since the prevalence of injection drug is often low among the general population, this method requires a large sample size in order to capture PWID. GENERAL POPULATION SURVEY This method is similar to the network scale up method that uses general population surveys. However, instead of asking drug injecting behavior of people in a participant’s personal network, it asks participants about their own drug injecting behavior. This method is easy to use and social-desirability bias can be reduced by using computer assisted personal interview or online surveys. Some PWID, however, may not be easily reached through traditional household sampling methods employed by these surveys. NATIONAL LEVEL POPULATION SIZE ESTIMATES Often size estimates of key populations are produced for major urban or other specific areas of a country. While helpful in local planning or estimate adjustments, these estimates and the prevalence they generate cannot in themselves provide data-supported estimates for a whole country. Using statistical methods to extrapolate population size from areas where estimation has been performed to areas where it has not is one approach to constructing country-wide estimates. However, careful review and understanding of the regional differences in drug-using characteristics is important so that programmatic decisions can be adjusted accordingly. For example, in some countries there may be a high prevalence of 6 PWID in rural communities, or in border areas, or along drug trafficking routes but very low prevalence in other areas. The quality of national PWID population size estimates can be highly dependent on the quality of each regional or local estimate. For this reason, review of the method used to have a sense of data quality and credibility is important in every national estimation effort. Caution should also be taken if only urban estimates exist, as applying urban population prevalence to other areas could overestimate the size of the PWID population. 7 Reported Estimates and Methodologies Current sub-national and national estimates, along with their estimation methodologies, as available, are shown in Appendix 2. The information was gathered through literature review (published and grey) and communications with various local and international stakeholders (i.e., regional representatives of USAID, CDC, UNAIDS, and UNODC as well as some implementing organizations such as International HIV/AIDS Alliance and FHI360). Additionally, for the four Central Asian countries (Kazakhstan, Kyrgyzstan, Tajikistan, and Uzbekistan), UNODC convened a stakeholder meeting with representation from government (including Ministry of Health and drug control agencies)b and NGOs implementing PWID size estimation data collection in June 2014. The meeting entailed gathering information on current size estimation activities from country representatives and developing concrete recommendations for next steps in each country. With the major exception of India, a large part of these estimates are based on two comprehensive systematic literature reviews (Aceijas, 2004 and Mathers, 2008) summarizing global drug injection prevalence by country. These papers utilize a vast number of grey and research literature, and thus provide a broader regional understanding of sizes and the estimation methods. Many estimates are reported with insufficient details about methodology, preventing further evaluation of the data quality. Some estimates show wide variation between analogous populations that raises questions about the accuracy of estimates, and warrant further investigations. Many estimates are old and provide no information on demographics or subpopulations. Because most of these estimates are not part of routine data reporting, they provide little information on trends. RECOMMENDATIONS FOR COUNTRY PWID ESTIMATES In addition to country-specific data review, synthesis, and recommendations, some overarching principles for estimating the size of PWID populations are recommended below: • Capacity building is needed on assessing the quality of existing population size estimates in countries and conducting data collection to obtain population size estimates, especially at the local or regional levels. • Multiple methods of population size estimates should be obtained to triangulate estimates and obtain a median of the estimates. The use of multiple methods strengthens confidence in estimates, provides upper and lower plausibility bounds, and reduces the likelihood that biases from any single method will substantially alter results. • Integration of methods into planned national or regional IBBS activities is encouraged to maximize efficiencies in cost and resource use. Since the regular use of IBBS is increasing in many countries, incorporating population size estimation activities as part of the IBBS (e.g., various multiplier methods, unique object identifiers, and WOTC) represents an opportunity to establish regular updates in PWID population size estimates. • For countries that do not conduct regular IBBS or if the PWID population size estimates are desired sooner than the next round of IBBS, we recommend a variation of this protocol (a simplified RDS-based survey). This would produce (a) PWID population-based size estimates and (b) be reliable/comparable if replicated in future IBBS surveys using RDS recruitment. To adapt the protocol, the behavioral survey and the HIV testing procedures are eliminated from the IBBS. The formative assessment and recruitment of participants using RDS would remain the same. Some cost savings (compared to a full b Government representatives from Uzbekistan were not present at the stakeholders meeting in Astana, Kazakhstan. 8 IBBS using RDS) would be achieved by eliminating medical supplies and serotesting and hiring of less staff, but the implementation time may be similar to a full IBBS. Instead of the lengthy behavioral survey, participants would be administered a brief questionnaire (approximately 10 questions) that includes screening questions, demographics, peer network size, and population size estimation-related questions. The following population size estimation methodologies may be incorporated into the simplified RDS-based survey: various multiplier methods, unique object identifiers, and WOTC. • In most countries, there may be existing data sources (such as drug treatment data, arrests, and registries, perhaps even from sentinel surveillance sites) on drug use. These should be considered when using service multiplier method or for capture-recapture method. Formative assessment should be conducted to ensure that these sources are complete and accurate. • PWID size estimation activities should try to disaggregate by sex and age where possible. Females who inject drugs have not been well-assessed in these countries, and more concrete data is needed. This information is crucial as evidence shows that compared to men who inject drugs, women who inject drugs have typically higher rates of HIV, higher levels of risky injecting and sexual behaviors, higher mortality rates, and faster progression to drug dependence. Without sex-disaggregated data, it is difficult to ascertain whether there are sufficient services for women who inject drugs and if there are disparities in access to services.9 LIMITATIONS IN THE ASSESSMENT OF COUNTRY PWID ESTIMATES The review of country-level PWID population size estimates presented several challenges. Primary barriers in gathering data for this assessment included the paucity of published estimates, lack of availability of translated reports, delays in communication responses from in-country officials, and lack of final dissemination or government approvals to publish draft estimates. Many countries, notably in Central Asia, have actually conducted recent PWID size estimation activities, but reports had not been completed, translated, or disseminated. Therefore, it was difficult to systematically assess the quality of timely or current PWID size estimation activities and methods. A standard systematic review would have (a) yielded only outdated information, (b) been difficult to replicate given the need for several personal communications and independent translations of unofficial presentations, and/or (c) not reflected several current ongoing activities or identified timely relevant needs. Also, the size estimations were often not described or described inadequately in a number of papers and documents. Thus, we were not able to grade or score the quality of estimates in a number of countries due to lack of available details. Another limitation is the lack of estimates disaggregated by sex or other important mediating factors such as age, urban residence, or incarceration/residence in institutional settings. 9 Country-Specific Recommendations Details of the review of the ten countries are provided in the Appendix. Tables include sources and methodologies used and the limitations and strengths of the methods. We provide our interpretation and recommendations below. Belarus Overview of the HIV epidemic Belarus has a concentrated HIV epidemic with an average prevalence of 0.1%.10 Among PWID, the prevalence is 17.1%.2 Regional HIV prevalence varies, exceeding 40% among PWID in Svetlogorsk. High PWID prevalence are also found in Mogilev, Minsk, and Gomel Regions (Personal communications with International HIV/AIDS Alliance, Ukraine). Female PWID are highly involved in commercial sex. Harm reduction activities including opioid substitution programs are provided across the whole country. Review of the estimation data The estimates of PWID population in Belarus range from approximately 50,000 to 76,000 nationally.11-17 The officially reported estimate to UNAIDS is 75,000 in 2012.2 A reported registry count from 2005 was only 6,308.15 The registry or registries on which this estimate was based were not specified in the reference. One reference that details the methodology used for estimation was conducted in 2008 and was based on using capture-recapture with two different national level registries.13 The first registry was the Narcological Registerc as of the beginning of 2008, and the second register consisted of HIV positive persons who were identified to be drug users by the Department for HIV/AIDS Prevention of the National Centre for Hygiene, Epidemiology and Public Health as of January 2008. This yielded a national estimate of 76,281 (69,200 – 83,400). The estimation based on two data sources alone is likely subject to bias, especially due to the likely correlation between the two sources (i.e., being in one data source would influence that same person being captured in the other data source). In addition, both data are not representative of the PWID population. A more recent estimate (2009) that was reported with some level of detail of the methodology was based on triangulation of multiple methods including service multiplier incorporated into a sentinel surveillance survey with PWID and the NSUM.11 The sentinel sites for PWID included 17 cities in Belarus. The service multiplier method used multiple services data as the multiplier (narcological service, law enforcement agency, hospital admissions for overdose in the last six months, and anonymous counseling centers for drug users), which increases the robustness of the population size estimate. National estimates obtained from the multiplier and NSUM were triangulated, resulting in an estimate of 45,247 (lower bound) and 65,246 (upper bound). Results from each of the different methodologies were not provided. An RDS-based IBBS survey for PWID was implemented in 2013 in Minsk, Soligorsk, Gomel, Svetlogorsk, Pinsk, and Polotsk. Population size estimation was not incorporated into the survey but the survey included questions related to registration at narcological and psychiatric clinics and number of overdose cases. In addition, a client tracking database of a harm reduction program cThe Narcological Register consists of the Dispensary Narcological Register and a Prevention Narcological Register. The Dispensary Narcological Register includes persons diagnosed with an addiction according to ICD-10 criteria. The Prevention Narcological Register includes non-addicted patients who are suspected of using drugs. 10 provides the number of clients at local and national levels. These data sources have untapped potential for providing additional estimates. Recommendations It will be important to convene a stakeholder meeting in Belarus that includes government officials and drug control agency as well as relevant implementers. The stakeholder meeting will be important for understanding what size estimation exercise have been done, what data are available, and how useful the data are. Furthermore, it will also be worth determining if PWID population sizes can be estimated using the data collected in the 2013 IBBS (i.e., via service multiplier) given that questions related to registration at various service points were included in this survey. There is also a need for technical assistance on publishing the existing size estimation data through a writing support.. China Overview of the HIV epidemic China has a concentrated HIV epidemic with an estimated prevalence of 0.1% among the general population of adults aged 15-49,1. HIV prevalence is much higher among key populations including MSM, FSWs, and PWID.18 Regionally, the HIV prevalence is markedly higher in southwest and northwest China, especially in the provinces Yunnan, Sichuan, Guangxi, Guizhou, Guangdong, and Xinjiang19. Among PWID, the HIV prevalence is estimated at 6.3% (2012).20 China has diverse populations of PWID and their risks also vary significantly across regions and sub-groups. Injection drug use is estimated to be highest near drug trafficking routes (i.e. border towns and ports) and in some major cities.21 Review of the estimation data National estimates of the size of the PWID population range from 1 million to 2.35 million; with the highest numbers of PWID estimated to be in Yunnan, Guangxi, Henan, Xinjiang, Guangdong provinces (>50,000 PWID per province). Provinces which include estimates ranging from 10,000 to <50,000 PWID include Sichuan, Guizhou, Hunan, Chongqing, Zhejiang, Hubei, Anhui, and Jiangsu.1, 14, 15, 21-28 China conducts regular size estimations of PWID, in which they use a combination of methods including service multipliers and others. PWID data in law enforcement registers have been used as a benchmark to estimate the size of PWID. Multiple data sources have been used to strengthen the estimates. Overall, the size estimate of PWID in China is fairly robust because there is a national process and they use multiple methods in many cities. Recommendations As China is a large and diverse country, we suggest the estimation efforts should continue to focus on regional levels, especially where the drug use are reportedly high. The regional estimates ensure its relevancy as well as directly inform programming and policies in each region. To supplement current efforts where relevant and possible, RDS-based surveys should incorporate size estimation components. India Overview of the HIV epidemic India has a concentrated HIV epidemic with an estimated prevalence of 0.27% in 2011.29 The 11 epidemic is concentrated among sex workers, men who have sex with men, and PWID. The HIV prevalence also varies across different states. In particular, the North Eastern States have typically had much higher HIV prevalence compared to other areas and is largely driven by PWID. Other parts of the countries, which were previously perceived as having low prevalence, are experiencing increasing prevalence rates of infection; these include Punjab, Chandigarh, Delhi, Mumbai, Kerala, Odisha, Madhya Pradesh, Uttar Pradesh and Bihar.29 The estimated prevalence of HIV among PWID in Inida has remained stable around 7% since 2007.29 Review of the estimation data Estimates on the size of the PWID in India are very much up to date as well as comprehensive. The latest national estimate of PWID stands at 177,000 as reported in National AIDS Control Organization 2012-2013 Annual Report.29 This estimate is based on the state level epidemiological profiling project in 25 states (India has 28 states and one capital region). These estimates are based on mapping and supplemented by information obtained regularly through the Targeted Interventions Program partners, whose program staff has regular contacts with PWID. Information on mapping and updating through the Targeted Interventions were not specified. However, one of the positive features of the epidemiological profiling is that data is obtained directly at the district level as opposed to extrapolating to the district level from a few local hotspots. It is unknown whether there was any inflation factor used to account for non-venue based PWID. This estimate is consistent with the estimate of 164,820 reported by Mathers (2008).15 The latest round of IBBS for PWID was conducted for May – July 2014, with another round likely in 2017/18. The 2014 IBBS did not incorporate a component of size estimation; the reason is unknown. The next round of IBBS for PWID will not include a national size estimation exercise. However, they will select a few districts in certain states where there is substantial PWID population, and within these districts, they will validate all hotspots based on information provided by the Targeted Intervention partners and from the wisdom of the crowd method with PWID and gatekeepers at each hotspot. Recommendations India currently has multiple existing estimation data on the size of PWID. However, the methods are heavily dependent on mapping of hotspots. This might underestimate the actual size of PWID population because hidden segments of the population are likely missed from the mapping. In the future round of PWID IBBS (2017/2018), NACO should consider obtaining population size estimates via additional methods to validate current estimates. The round of IBBS should incorporate multiplier methods and WOTC). Obtaining service multipliers should be feasible given the availability of PWID registries maintained by the Targeted Intervention programs. Size estimates from this service multiplier method could be triangulated with the size estimates obtained from the state-level epidemiological profiling project to provide a plausible range for the national estimate. This would provide additional estimates while being cost-saving. Kazakhstan Overview of the HIV epidemic Kazakhstan has a concentrated epidemic with an estimated HIV prevalence of 0.2% in 2011.20 The HIV prevalence is much higher among key populations, especially PWID, which is estimated at 4.8%.30 12 The epidemic is considered to be driven by PWID and the transmission has been bridged into the general population through heterosexual contacts with female partners of PWID. Review of the estimation data Several national estimates of PWID population has been conducted in Kazakhstan, ranging from 100,000 to 200,000.14, 15, 22, 31-35 Kazakhstan has a relatively extensive set of epi-behavioral data (Personal communications, UNAIDS). Areas covered by survey and sentinel surveillance include over 20 sites. The periodicity of epi-behavioral data collection, including PWID size estimations, is regulated by the government and implemented by governmental organizations, based on the decree approved back in 2005. Prior to 2013, the PWID size estimations were carried out on biannual basis at regional level by making use of mainly multiplier and capture-recapture methods. The last estimation was done in 2013, and the last published estimate is from the 2010/11 estimation exercise. The national figure is obtained by summing up the regional level data. The limitations is that many documents have not been transcribed and widely disseminated, and do not describe detailed estimation methods. Kazakhstan has developed national guidelines for estimating the size of most-at-risk population size estimation, which will be used starting from 2014. Methods previously utilized and currently considered for PWID size estimation are service multipliers, capture-recapture using record databases, and NSUM. In fact, the multipliers are already incorporated in the 2014 IBBS, which uses RDS-survey, sentinel survey, and service multipliers available from prevention and treatment programs. Some regions, where independent databases using similar anonymous unique identifiers are in place, will apply a capture-recapture method based on matching identifiers in the databases. NSUM was piloted by UNAIDS in 2012 in two regions of Kazakhstan (Almaty city and Almaty region) although the unadjusted estimates were found to be too low due to reporting bias and other factors. Results were not published or used internally, and further modifications for use of NSUM in the future are under consideration (Personal communications with UNAIDS country office). Recommendations Given the extensive size estimation activities currently on-going, we do not recommend additional data collection in Kazakhstan. Size estimates for multiple sites/cities were reported in 2011, making it unlikely that other priority sites have not been addressed. Additionally, size estimation using multiplier methods were incorporated into the implementation of the 2014 IBBS. Continued efforts to incorporate size estimation methods into ongoing planned IBBS activities are encouraged. At the same time, the implementers of the IBBS and sentinel surveys in Kazakhstan have identified gaps in their resources for analysis, interpretation, and dissemination of the 2014 and 2011 IBBS size estimation data. Based on the Astana Meeting (June 2014), the stakeholders in Kazakhstan have requested support for and endorsed the following recommendations: 1. Support a national seminar in late 2014 to achieve following objectives: a. Build the capacity of regional staff in Kazakhstan: i. Training to implement higher-quality size estimation methods ii. Discuss and assess the quality of data previously and currently collected 2. Provide technical assistance to produce publications on size estimation data, including: a. Technical writing support and expertise, including possible workshop to produce papers, including peer-reviewed journal articles b. Translation and editing services to facilitate wider publication dissemination of existing and future reports 3. Provide technical support as needed to review population size estimation questionnaires and data collection tools for upcoming and future size estimation assessments 13 4. Provide technical support to review and advise on analysis and interpretation of results as needed, including training and support to implement statistical software related to size estimation methodologies, such as the RDS Analysis Tool (RDSAT). Kyrgyzstan Overview of the HIV epidemic Kyrgyzstan has a concentrated HIV epidemic with an estimated prevalence of 0.3% in 2012.20 Key populations, including PWID suffer a much higher HIV burden. PWID has a prevalence of 14.6 (estimated in 2013).2 Review of the estimation data The first PWID population estimate was conducted in 2006 by UONDC and the reported number was 25,000. This estimation exercise utilized a population-based nominationd/multiplier method, at the national level. Other estimates which are much higher seem to be based primarily on expert opinion and not primary data sources. There are several recent estimates. In 2011, a study was conducted with technical support from UNAIDS and the Central Asia Drug Abuse Programme (CADAP). It used multiplier/nomination methods integrated within the 2011 IBBS. The most recent round of IBBS was conducted in 2013 where some multipliers were integrated, but PWID were not assessed in this round. In parallel, however, a PWID size estimation exercise was carried out in 2013 through a project funded by GFATM with UNAIDS technical support. Eight service multiplier and a modified capture-recapture method were used to produce PWID size estimates in eight locations throughout the country. A combined report of the results from the 2011 and 2013 data collections is in progress, but results have not been officially released. The next IBBS is scheduled for 2015, and the protocol will be modified to account for the results and lessons learned from the 2013 size estimation. Recommendations Since the 2011 and 2013 size estimation data has not been released, we recommend providing the necessary support to assist in finalization of the report, as well as support translation and dissemination to a broader audience. However, given that there is a plan to include a size estimation study in to the 2015 IBBS, we recommend expert reviews of this study to ensure that multiple relevant methods for size estimation will be incorporated to yield best estimate possible. In addition, the stakeholders in Kazakhstan have requested support and endorsed the following recommendations: 1. Support a national seminar in late 2014 to achieve following objectives: a. Build the capacity of regional staff in Kyrgyzstan: i. Training within the seminar to implement higher-quality size estimation methods ii. Discuss and assess the quality of data previously/currently collected 2. Provide technical assistance to produce publications on size estimation data, including: d Nomination method starts with a small and accessible subgroup of the target population who are then asked to refer others in the same target population. Those referred individuals are asked to refer others and so forth. This method, however, is not recommended for population size estimations due to selection biases. 14 a. Technical writing support and expertise, including possible workshop to produce papers, including peer-reviewed journal articles b. Translation and editing services to facilitate wider publication dissemination of existing and future reports 3. Provide technical support as needed to review population size questionnaires and data collection tools for upcoming and future assessments 4. Provide technical support to review and advise analysis and interpretation of results as needed, including additional capacity-building trainings as needed and requested. Libya Overview of the HIV epidemic Limited information is available on the HIV situation in Libya. UNAIDS estimated HIV prevalence in Libya at 0.13% (2005).36 The epidemic is concentrated among PWID, with injection drug use accounting for as much as 90% of all new infections in 2010. A range of post-conflict conditions have added and exacerbated risk factors in HIV transmission overall, including a lack of antiretroviral drugs, disrupted infection control and blood safety systems, and a rise in sexual and gender-based violence. Among PWID, HIV prevalence estimates range from 15% to 32%.37 In Mathers 2008, the estimate of HIV prevalence among registered PWID was cited at 22%. There is a more recent (2013) estimate of HIV prevalence among PWID in Tripoli; it was estimated at 87.1% (95% confidence interval: 81.5- 91.9).38 This estimate is from an IBBSS using RDS with a sample size of 294. Review of the estimation data Years of civil unrest have hindered any estimation of injection drug use in Libya; from the research literature, reliable national estimates for Libya do not exist. Some limited national estimates show a wide range: from 1,685 (from registry data, representing only a verified lower bound) to 7,206 (adjusted estimate from Mathers 2008.14, 15, 37). All estimates are derived from narcological registry data, which is likely an underestimate of the true number of PWID. There was no information on how the data was obtained for the narcological registry. IBBS has been conducted in Libya but no population size estimation was incorporated in it. There is currently no PWID population size estimation activities planned. Recommendations There is an urgent need to estimate the population size of PWID for Libya. However, this may depend on the current political situation and security issues in Libya. Methods chosen for size estimation in Libya should be determined based on formative research and experience from other countries. We recommend the use of multiple methods, and where relevant, incorporate size estimation in probability surveys, including IBBS. WOTC, unique objective multipliers, and network scale up methods can be incorporated into probability surveys. It may also be worth exploring the use of the multiplier method with the 2013 IBBS that was conducted in Tripoli. This would require reviewing the questionnaire to see if any specific service use questions were asked (with a timeframe) and investigating the quality of the respective service registry. Myanmar Overview of the HIV epidemic Myanmar has a concentrated HIV epidemic with an estimated prevalence of 0.5% (in 2011).2 Key populations such as female sex workers, MSM, and PWID have a much higher reported prevalence. 15 Among PWID, the estimated prevalence is 18.0%)39, and concentrated in the regions that cultivate opium or near drug trafficking routes. Review of the estimation data National estimates range from 60,000 to 195,00014, 15, 40, 41, and a consensus meeting of in-country stakeholders agreed to an estimated number of 75,000 PWID as the national estimate. This number of 75,000 was arrived at from the estimate of 0.5% of the male population 15-49 years of age being PWID.42 No justification for the use of 0.5% was provided. There are a number of harm reduction programs currently operated by the government, UNODC and Burnet Institute Centre for Harm Reduction that would potentially provide services data for size estimation purposes. UNAIDS/CDC/WHO is currently providing technical assistance to the National AIDS Program for the current IBBS (data collection completed May 2014), which includes size estimation activities using the service multiplier and unique object multiplier methods. (Personal communications, UNAIDS/Myanmar). The IBBS for PWID was conducted in a total of 16 sites: Lashio, Naung Mon, Nampoung, Muse, Kukkhai, Nam Pha Kar, Myitkyina, Waimo, Ay Myint Tha, Bamaw, Mandalay (Aung Myae Thar Zan, Pyi Gyi Tha Gon), Yangon (Thin Gangyun, Kyimyndaing), Kalay, and Tamu, that would give regional estimates as well as a credible benchmark to result in a new national estimate. Recommendations Progress is being made in that population size estimations of PWID using sound methodology have been incorporated in the current round of the IBBS in 16 sites. National estimate based on this survey would have to carefully consider the regional differences. The profile of these 16 sites are somewhat unique in that they represent two major urban centers (Yangon and Mandalay) or sites in high opium growing states (Kachin and Shan) and their estimated sizes may not be applicable to other parts of Myanmar which are not urban centers or in opium growing states. Some formative assessment should be conducted to determine if additional data on PWID population size should be obtained from other sites in Myanmar (the north and the south). It may also be worthwhile exploring the need for conducting capacity building workshops in PSE implementation, data analysis and interpretation, and technical writing. Philippines Overview of the HIV epidemic Although the Philippines has a relatively low HIV prevalence (0.036% in 2011), the HIV epidemic has changed drastically in the past five years with an exponential increase in newly reported HIV cases.43 Similar to other concentrated epidemics, key populations including PWID also suffer a higher burden of HIV compared to the general population. HIV prevalence among PWID was less than one percent prior to 2010. However, as of 2012, the estimated HIV prevalence among PWID was as high as 13.6% nationally.44 The 2011 Integrated HIV Bio-Behavioral Surveillance Survey (IHBSS) in the province of Cebu indicated a much higher prevalence (53.8%).45 Review of the estimation data The National Epidemiology Center/Department of Health (NEC-DOH) has recognized the need for PWID size estimation and has conducted numerous size estimation exercises since 2011. Population size estimations of key populations are systematically determined every two years by NEC-DOH, with technical assistance from various agencies including WHO. (Personal communications, Ricardo Mateo MD, FHI360, Philippines) 16 The latest national estimate was 16,578 PWID (2011).42 PWID is a priority population for HIV prevention in 70 selected areas/cities. These areas were selected based on data from the IHBSS, rapid assessment of HIV vulnerability in 2010 and 2011, and the estimated number of key populations in each of the area for 2011. There are a total of 122 cities in the Philippines. In the 70 priority cities together, the estimated number of PWID was reported to be 8,111, and the rest of the country was estimated to be 8,467, yielding a total estimate of 16,578 nationally. However, the methodology for this is not specified. A previous reported estimate is from the 2009 consensus meeting, which took into account the Dangerous Drug Board’s survey which reported that 2-4% of the general population use illegal drugs, and of these, 0.089% were PWID.46 The consensus meeting also took into account estimated figures from three cities (Cebu City, General Santos City, and Zamboanga City) with identified PWID networks (methodology not specified). This consensus meeting reported a figure of 12,705-21,567 PWID nationally. The latest IHBSS survey was conducted by NEC-DOH in 2013 in Cebu and Mandaue with technical assistance from WHO.47 They survey incorporated population size estimation using multiple methods: i) unique object distribution implemented one week prior to the PWID IHBSS data collection; ii) service multiplier using service data from the Social Hygiene Clinics; and iii) services multipliers using data from the Cebu City and Mandaue City prison/jail. The results will be available in September 2014. (Personal communications, NEC-DOH) Policy changes to legalize syringe exchange could provide improved multipliers for benchmark- multiplier methods in the Philippines. Comprehensive harm reduction interventions, including needle and syringe programs, are planned to be implemented in Barangay Kamagayan in Cebu City as part of the Big Cities Projecte. (Personal communications, Sutayut Osornprasop, World Bank/Thailand) Recommendations The Philippines appear to have a national process in place to estimate the size of PWID. The methodology seems to be robust with the use of multiple size estimation methodologies. The upcoming comprehensive harm reduction interventions that will be initiated should ensure that service data is of high quality and that it can be used for the service multiplier method (i.e., ensure that duplicate clients can be identified). We recommend these continuous size estimation efforts and that estimated figures, detailed methods, and lessons learned from the 2013 and 2014 size estimation to be shared broadly with national and international stakeholders. There might be the need for technical assistance in using multiple service data when they become available and to help publish size estimation data. Tajikistan Overview of the HIV epidemic Tajikistan has a concentrated HIV epidemic with a national estimated prevalence of 0.3% (2012).20 Female sex workers and PWID and their sexual partners suffer a much higher prevalence. The prevalence among PWID is 13.5% (2012).20 Injecting drug use among men has largely driven the HIV epidemic in Tajikistan and heterosexual transmission to the sex partners of male PWID is also e This project is supported by the Asian Development Bank and World Bank under Department of Health guidance and leadership. This is being implemented jointly by the Population Services International (PSI) and the Philippine NGO Council on Population Health and Welfare Inc. 17 common. Recent routine surveillance data showed an increase in the proportion of newly registered HIV cases due to heterosexual transmission (57.5%); of which more than half was women. Review of estimation data Few size estimation studies are available and the estimates range from 15,000 to 50,000.14, 15, 34, 48, 49 Most of these estimates come from in-country expert opinion only—possibly extrapolated from drug registry data. The estimate of 25,000 of PWID that comes from the 2009 AIDS Project Management Group study is widely used [REF] using multiplier methods. The most recent estimates are from 2013, when a USAID project conducted a situational analysis and estimated the PWID size by making use of existing program data and expert opinion in assigning a multiplier coefficient to each area. It resulted in a national estimate of 26,500 PWID. (Personal communications, UNAIDS/Tajikistan). The last IBBS was done in 2011, and a new round of RDS-based IBBS is planned for late 2014. In the current version of the 2014 IBBS protocol that was approved in 2013, the PWID size estimation component is not integrated. However, the data collection tool includes questions on overdose, narcological or police registration, TB registry and opioid substitution treatment registration. In addition, there is a separate small survey (supported by UNAIDS) using RDS methodology being concurrently implemented for the purposes of population size estimation using multiplier methods. Recommendations Since size estimation of PWID using multiplier methods is currently being implemented, we did not recommend further size estimation to be implemented at this point. However, we recommend investigating the potential of using services data as well as data from the 2014 IBBS to produce more data to strengthen the national estimate. Based on the Astana Meeting (June 2014), the implementers and supporters of the size estimations in Tajikistan have requested support and endorsed the following recommendations: 1. Support a national seminar in late 2014 to achieve following objectives: a. Review the draft results of the 2014 RDS survey size estimation b. Build the capacity of regional staff in Kyrgyzstan: i. Training to implement higher-quality size estimation methods ii. Training in interpretation and utilization of data and results c. Session focusing on identifying needs and linking services to utilization of 2014 results d. Session on monitoring of service quality 2. Provide technical assistance to produce publications on size estimation data, including: a. Technical writing training. b. Translation and editing services to facilitate wider publication dissemination of existing and future reports. c. Training on dissemination of results. Uzbekistan Overview of the HIV epidemic Uzbekistan has a concentrated HIV epidemic with the national estimated prevalence of 0.1% (2012).20 Similar to other countries in the Central Asian Region, the HIV epidemic is particularly concentrated in PWID with an HIV prevalence of 7.3% (2013).30 Transmission between PWID and their female partners is also considered be a driver of the HIV epidemic in the country. 18 Review of the estimation data National PWID estimates range from 79,300 to over 100,000.14, 49-51 Most of these estimates come from in-country expert opinion only—possibly extrapolated from prevalence estimates and drug registry data. The first government-led PWID size estimation exercise was carried out in 2011 with technical support from UNAIDS and the Central Asia Drug Action Programme (CADAP) by making use of multiplier/nomination method integrated within the 2011 IBBS. (Personal communications, UNAIDS). Draft results are not available for dissemination outside of the country. The results from these size estimations are also not currently accepted and adopted by Uzbekistan government officials. Based on the lessons learned from the 2011 round, the 2013 IBBS data collection tool was adjusted accordingly. The data collection has been completed, but the analysis has not yet performed. Recommendations It is difficult to make feasible recommendations to support current and future recommendations for PWID size estimation in Uzbekistan, given that recent estimates are likely to remain unpublished. It is also noted that implementation of UN- or outside-funded assessments is not easily approved in Uzbekistan; thus complementary PWID population size research may not be feasible. If possible, the first and logical next step towards moving PWID size estimation—both past estimates and future—in Uzbekistan is to organize a national stakeholders meeting on size estimation to review what has been done and what would be the relevant next steps for sharing the findings and lessons learned. This consultation would have to be held in Uzbekistan, ensure key government representation, and be based on a strategic agenda. REGIONAL DYNAMICS Seven of the ten priority countries share borders. Therefore, regional coordination in size estimation efforts may be key to understanding the impact of drug trafficking routes on PWID population estimates. Regional studies could also help inform HIV prevention and harm reduction activities in the region. Furthermore, regional estimates would also facilitate the standardization of several recommended approaches in population size estimation, including using multiple methods, incorporating size estimation into probability-type surveys, or using multiple existing service or drug registry data. 19 Conclusion Table 2 below summarizes the key recommendations and next steps for the population size estimation of PWID in the ten priority countries. A number of the priority countries have conducted population size estimation exercises using multiple methods and incorporating it into the IBBS surveys. It is important to note that there is no gold standard method for population size estimation of PWID and therefore multiple methods are recommended. Where feasible, size estimation should be conducted to triangulate the data and strengthen the estimates. Given many of these ten countries are rolling out IBBS or other probability- based surveys, efforts should ensure that there is enough local capacity in population size estimation incorporated within these surveys. This would potentially produce more credible data and at the same time can be cost-saving. Although several of these countries have conducted population size estimates of PWID, there is still a need for technical assistance in analysis, reporting and publications of these data. Regardless of the fact that size estimation has been done in these countries, this review has faced critical challenges in gathering size estimation data from these countries. In addition, much of the recommendations generated by in-country stakeholders in the Central Asian countries indicate the need for capacity building workshops in implementation, analysis and disseminations of size estimation exercises. Lastly, we emphasize the importance of convening country-level stakeholder meetings as was conducted in Astana for the four Central Asian countries for this project. The stakeholder meetings were crucial in evaluating the quality of the existing estimates and assessing the need for future size estimation needs as well as technical assistance needs. We recommend that national level stakeholder meetings be conducted in the remaining countries. Table 2. Summary of recommendations and next steps in the ten priority countries Country Recommendation/Next Steps Belarus Size estimation may not be immediately required because current estimates have incorporated multiple methods. Stakeholder meeting should be convened to determine if there is a need for technical assistance for producing publication on size estimation data such as through technical writing workshop and translation and editing services to facilitate wider publication dissemination of existing reports. China Given that there is already an existing national process for PWID population size estimations, it may be very difficult to obtain buy-in from the government to conduct any externally-funded assessments. We, however, recommend a more rigorous assessment in coordination with China officials of any gaps in regional estimates for high drug use provinces, and to support future activities if needed. Sizes of PWID can be changed overtime and thus efforts to continue the estimation is needed. India Given there is national estimate effort using mapping and key informant interviews, we don’t recommend more immediate estimation exercise. But given there are periodic IBBS being conducted among PWID in India, we recommend to have size estimation to be incorporated in these survey to save money as well as to provide data for triangulation and to enhance the quality of the estimates. This is particularly important as most of the estimates in India are done through mapping, potentially resulting in underestimates of PWID. 20 Kazakhstan No immediate size estimation efforts recommended. However, we recommend having workshops to build local capacity on multiple method use, and data analysis and dissemination. No immediate size estimation efforts recommended. However, we recommend having Kyrgyzstan workshops to build local capacity on multiple method use, and data analysis and dissemination. Libya Population size estimation is needed due to lack of data. Methods chosen for size estimation in Libya should be determined based on formative research and experience from other countries. We recommend the use of multiple methods. Myanmar Efforts to calculate national estimate should take into account regional variance. In addition, formative assessment should be considered in non-opium growing areas to determine if additional PWID size estimations is needed. It may also be worthwhile exploring the need for conducting capacity building workshops in PSE implementation, data analysis and interpretation, and writing. Philippines No population size estimation recommended; estimate based on multiple indirect estimation methods will be released in September 2014. Reach out to national stakeholders to determine if any technical assistance is required for the analysis, reporting and publications. Tajikistan No immediate size estimation efforts recommended. However, we recommend having workshops to build local capacity on multiple method use, and data analysis and dissemination. Uzbekistan Political issues currently inhibit the publishing of previous size estimation results. Implementation of outside-funded assessments is also not easily approved in Uzbekistan, and not recommended at this time. If possible, a national meeting is proposed to review recent results, and work with government officials to identify critical concerns about the data and plan a size estimation agenda for the near future. 21 APPENDIX I SIZE ESTIMATION METHODS WITH LIMITATIONS, ADVANTAGES AND CHARACTERISTICS Method Key attributes Limitations Budget and resource issues Enumeration • Census is a real count, not an estimate; • Require direct interaction that might not perform well if • Census is time-consuming and and Census thus straightforward to calculate population is geographically dispersed or hidden cost-prohibitive if not part of • Easily understood by policy makers since • Care must be taken to avoid risk for participants; community ongoing data collection no special statistics are required guides are needed for access • Enumeration requires fewer • Census produces creditable lower limit • Enumeration requires a reliable sample frame of venues resources and is less • Provide useful information for programming • Method will overestimate if population is mobile and double expensive to conduct than or outreach activities counted; underestimate if populations are well-hidden census Capture- • Quick and relatively easy • It is difficult in practice to not violate some assumptions of • Can be conducted within a Recapture • Does not require much data this method, especially in ensuring (a) the probability of shorter time frame relative to • Does not require statistical expertise selection is the same in both counts, and (b) that members of other survey-related methods, • Capture-recapture with mapping can be the population have an equal chance of selection possibly reducing costs useful to programming • Require direct interaction with PWID, which may not be feasible in hostile environments Services • Quick and relative easy • The two data sources must be independent • Usually low-cost, especially if Multiplier • Uses existing data • The two data sources must define the population in the same added to existing survey • Multiple data sources can be used to way activities strengthen the estimates • Quality service data may not exist, or be difficult to access or • May be integrated into an existing or interpret ongoing survey Unique object • Easy to integrate when combining with a • It may be difficult to distribute the objects to target • Can be low cost, especially if multiplier probability sampling survey like RDS population in drug use settings added to existing survey • Data collection needs to be finished in a short timeframe to activities prevent migration • Depends on the cost/type of object being distributed Mapping and • Cheaper than the traditional capture • May miss hidden segments of the population, especially • Requires field-based data interviews of recapture method those who spend little time at venues collection key informants • Useful for program planning • Estimates can have a wide minimum and maximum range • Costs dependent on number • Easy to implement of locations targeted for enumeration WOTC • Leverages existing surveys targeting PWID, • Difficult to verify the validity of the estimate • Low cost if built into existing such as IBBS surveys • Estimates often have wide reported ranges surveys • Does not require direct interaction or counting of PWID Delphi • Does not require raw data capture or direct • Vulnerable to political influences • Low cost interaction with PWID • Difficult to verify the validity of the estimate • Expert opinion-based • Estimates often have wide confident intervals 1 • May be used in countries with minimal or no data NSUM • Leverages existing surveys • Methodology still under development • Cost of integration into • Social-desirability bias is minimized by • Average personal network size can be difficult to estimate existing surveys can be asking indirect questions • PWID may not associate with members of general population minimal • Respondents may be unaware of someone in his/her personal network injects drugs • Respondents may not admit knowing PWID General • Leverages existing surveys among the • Requires a large sample size in order to capture PWID in the • Cost of integration into population general population population existing surveys can be survey • Straightforward to analyze and explain • Respondents may not report stigmatized and criminalized minimal • Bias can be reduced by using computer behaviors assisted personal interviews • Some PWID may not be reached by household sampling methods employed by large surveys of general populations 2 APPENDIX II ESTIMATES AND METHODOLOGIES BY COUNTRY Belarus Prevalence of HIV among PWID - 17.0* 2012 country population (Age 15-65): 9,625,888 PWID size estimate City/ Male/ (range) Region/ Female [% of 2012 Year /both Age Estimation method(s) population] Source Limitations/biases; Comments National/ Both N/A From 2004 estimates, adjusted 76,500 http://www.idurefgroup.unsw.edu.au/co Underlying data (Vinitskaya) may be 2008 [0.8] untry-data-and-maps/Belarus biased, and adjustment methodology insufficient to capture trends National/ Both N/A simplified ‘capture-recapture’ 76,281 European Monitoring Centre for Drugs The two data sources may be 2008 method: 1) Narco registry** (69,200 – and Drug Addiction (2009). Drug abuse interdependent. Estimate may also be identified PWID 2) Dept for 83,400) and illicit drug trafficking in the Republic biased by age as the likelihood of HIV/AIDS Prevention National [0.8] of Belarus in 2007. ** capture in both sources increases with Centre Registry (cites Vinitskaya, 2008) age. Narco registry has multiple sources and biases may be hard to quantify. National / Both N/A Literature review with 46,000 (41,000 - Aceijas (2004) No details on data or study 2004 midpoint of findings as 51,000) methodologies estimate [0.5] National / Both N/A Registry data 6,308 Mathers B et al (2008) Registry data largely underestimates 2005 [0.1] PWID populations; likelihood of capture may increase with age. National / Both N/A Not disclosed - 50,000– 52,000 Drug Abuse and Illicit Drug Trafficking in Methodologies not provided. 2006 [0.5] the Republic of Belarus in 2007 (2009) [Cites Meleshko LA, et al. Results of HIV Sentinel Surveillance in Belarus, Minsk, Kovcheg (2007) – this report could not be found]*** *World Drug Report 2013. ** From EMCDDA: “The Neurological Register (NR) consists of a Dispensary Narcological Register and a Prevention Narcological Register. The Dispensary Narcological Register includes persons diagnosed with an addiction according to ICD-10 criteria. The Prevention Narcological Register includes non-addicted patients who are suspected of using drugs (i.e., urine tested positive in strip tests). Reasons for including a person on the NR may be the result of an individual’s visit to a doctor; by a request from relatives, police, medical institutions, employers, educational institutions or military service commissions; or the result of inspections for juvenile offenders. All persons who test positive for drug use through a urinalysis, including those conducted at police stations, are then subject by law to observation and regular examination for drug addiction.” *** From report: “The methodology of research based on a complex approach with the use of different sources of information on drug use in the country, results of behavioural researches among IDUs and results of questioning of different groups of population. During the estimate of the number of IDUs a statistical method of frequency coefficients, triangulation and verification of received data were used. The quantitative estimate of IDUs over the country was fulfilled by summation of estimated figures in all regions and Minsk City.” 3 China Prevalence of HIV among PWID – 6.4* 2012 Country Population (15-64years): 1,009,835,100 PWID size estimate Male/ (range) City/ Female [% of 2012 Region/ Year /both Age Estimation method(s) population] Source Comments National / Both N/A Not given 1.93 million (midpoint) World Drug Report, 2014: UNAIDS estimate via 2009 (1.31-2.54 million) CDC China [0.19] National / 80% M N/A Grey and white lit review; Country 2.35 million Needle RH , Zhao L (2010) Delphi, secondary 2010 20% F focal expert responses to a survey (1.8–2.9 million) report instrument [0.23] National / Both N/A Lit review 1,928,200 (.356-3.5 Aceijas (2004) No study 2008 million) methodologies, [0.19] lit review. Not statistical. National / Both N/A Mass screenings by Public Security 1.5–3.0 million Wang L et al. (2009) Workbook 2007 Bureau; Undocumented based on [0.30] Methods to get known HIV prevalence rates among population size not P not registered with the PSB, adequately collected from sentinel surveillance specified. data and from Methadone programs operating in 22 provinces, adjusted by comparing known rates from data on groups participating in Methadone clinics. National/ Both N/A Indirect methods 2.35 (1.8-2.9 million) Mathers, B et al (2008). Lit review not direct 2005 [0.23] National / Both N/A Sum of Public Security Bureau 1.8–2.9 million Lu F, et al. (2006) Workbook. Using 2005 prefecture listings of the numbers [<0.29] estimates from of drug users were combined with security and detox. information from detoxification/ Circular – Detox treatment centers on the follows arrest. proportion of injectors among drug users and local surveys of registration completeness to derive the estimated number of drug injectors. National / Both N/A Not disclosed >3 million Summary Country Profile for HIV/AIDS Treatment No provenance 2004 [0.30] Scale-Up. UNODC. (2005) 4 National / Both N/A National Drug Abuse Surveillance 1.0–3.5 million Qian et al (2005) Secondary data. 2002 Center, [0.35] Original methods not specified. Regional Urumqi, Both N/A Benchmark/ multiplier study. 7,148 Ni et al, (2012) From a public Xinjiang, (68.2% of 10,481) and the maximum security estimate. Peoples’ number was 13,640 (68.2% of Provenance Republic of 20,000). unknown China 2011- Chongqing, Both N/A Network Scale-Up Method 14,975 (95%CI:13,047- Guo et al (2013) Author says China 16,904) probably low. A (2012) common problem with NSUM Dehong Both N/A 7 sentinel sites, 16 community- 14,700 (13,100– Jia Y et al (2008) Workbook. Unclear Prefecture, based surveys, 15 studies 16,200) how they got Yunnan estimates as Province workbook requires (2008) estimates to be imputed. Ruili City; Both NA 7 sentinel sites, 16 community- RC 1650 (1,500– Jia Y et al (2008) (same study as above) Workbook. Unclear Luxi City; based surveys, 15 studies 1,800); Lx C 3,450 how they got Longchuan (3,000–3,900); Lng Cn estimates as Cty; Yingjing 4,600 (4,000-5,200); workbook requires Cty Liaghe YC 4,400 (4,100– estimates to be Cty / 4,700); Li Cty 550 imputed. (2008) (500–600) * World Drug Report 2013. 5 India Prevalence of HIV among PWID – 7.1* 2012 country population (Age 15-64): 808,440,644 Male/ PWID size estimate City/ Region/ Female Estimation (range) [% of 2012 Year /both Age method(s) population] Source Limitations/ biases; Comments National / Both N/A Mapping conducted 177,000 National Aids Control Details of how the mapping was conducted is 2007-09 in 2007-09 and [0.02] Organization– Annual Report minimal. Mapping is conducted as part of the validated yearly 2012-2013 District HIV/AIDS Epidemiological Profile project, through Targeted which is a systematic compilation of available data Intervention from various sources (surveillance, program). program It is possible that this estimate may be an underestimate as mapping likely only captures PWID found in hotspots. However, this is likely to be minimal since a large proportion of PWID in India are street-based (as opposed to home- based). Additionally, these state- and district- level counts of PWID are validated every year as part of the Targeted Intervention program, whose program staff have regular contacts with PWID (both street- and home- based PWID). National / Both N/A Author’s personal 106,518–223,121 Sharma M (2009) Estimate not robust; based on personal 2009 communication: [<0.03] Range from Mathers B, et al, communication. WHO Country Office 2008. and National AIDS Programme Staff. National / Both N/A Adapted from 172,000 (111,000- http://www.idurefgroup.uns 2008 Mathers 2008: 233,000) w.edu.au/country-data-and- prevalence * [0.03] maps/India#idu population from Mathers, 2008 2008 National / Both N/A Indirect methods 164,820 (106,518- Mathers, B et al (2008). No details on what indirect method was used but 2006 223,121) indirect methods are preferred over other [0.02] methods so this estimate is deemed to be fairly strong. National / Male/ N/A Not included Male 96,463 - 189,729 National Aids Control Study methodologies not specified 2006 Female Female 10,055 - 33,392 Program 2006-2011 [<0.02] National / Both N/A Lit review 1,294,000 (563,000- Aceijas (2004) No study methodologies 2004 statistics 2,025,000) [0.16] from 1998- 2003 Regional 6 25 states/ Both N/A Mapping exercise Counts of PWID are Epidemiological profiling of Very comprehensive systematic compilation of 2013 conducted 2007- available for 25 states, HIV/AIDS situation at District program and surveillance data; counts are 2009 including at the district and Sub-district levels using available for 25 out of the 28 Indian states and 1 level Data Triangulation Project Capital region. (NACO 2013) This may underestimate population size; mapping likely only capture PWID found in hotspots. However, this is likely to be minimal since a large proportion of PWID in India are street-based (as opposed to home-based). Additionally, these state- and district- level counts of PWID are validated every year as part of the Targeted Intervention program, whose program staff have regular contacts with PWID (both street- and home- based PWID). 2 states Mapping from Churachandpur (MH): Vadivoo 2008 Multiple methods were used; 2 of 3 program (Maharashtra Avahan Program 2,500 (program estimates fall within 95% CI of estimate through [MH] and compared to cap- estimate); 1,493 (CR) CR; Nagaland recap (CR) (unique Program-based estimates are regularly updated by [NG]) object and Phek (NG): 1,800 program staff who have regular contact with PWID /2008 recapture through (program estimate); pop-based survey 2,439 (CR) using RDS recruitment Wokha (NG): 4,800 (program estimate); 2,941 (CR) 5 cities Stakeholder report Imphal (9,000-1,2000_ Dorabjee and Samson 2000 Delphi method but reported estimates were cross- (Imphal, of estimates and Chennai (15,000-20,000 checked through observations at hotspots. More Chennai, observations at Mumbai (38,000) than 10 years old. Mumbai, hotspots to check Calcutta (10,000-15,000) Calcutta, estimates. Delhi (25,000-30,000) Delhi) /2000 New Delhi Both N/A Lit review New Delhi (35,000) and C Aceijas, 2004 No study methodologies and Mumbai Mumbai (38,000) 1998-2003 Imphal West, Both N/A Service multiplier Imphal West, 7,353 (95% Medhi et al. 2012 Imphal East, method CI: 6,759-8,123); Imphal Thoubal East, 5,806 (95% CI: Churachand- 5,635-6,054); Thoubal pur 3,816 (95% CI: 3,571- Bishenpur 4,139); Churachandpur District 2615 (95% CI:2,528- 7 2,731); Bishenpur district 2,137 (95% CI: 1,979-2,343) Punjab Male >17 Advanced Punjab 2,600-18,148 Ambekar A, Tripathi BM Chandigarh – respondent-driven Chandigarh – Panchkula (UNAIDS) Panchkula – survey – Mohali 762-1,170 Mohali Haryana 2,265-15,858 Haryana / (Sub-regional estimates 2008 also available) * National AIDS Control Organisation (Department of AIDS Control). Annual Report 2012-2013. (India). 8 Kazakhstan Prevalence of HIV among PWID – 4.8* 2012 Country Population (Age 15-64): 11,039,571 Male/ PWID size estimate City/ Female (range) [% of 2012 Limitations/biases; Region/ Year /both Age Estimation method(s) population] Source Comments National/ Both N/A Multiple methods: 123,640 [1.12] Ganina et al., Review of the HIV Epidemiologic Methodology is explained 2011 Quasi capture- Situation and Results of the Integrated Bio-Behevioral vaguely in the report. recapture based on Surveys in the Republic of Kazakhstan 2010-2011, However, information from several lists of names Republican AIDS Center, Almaty, Kazakhstan, 2012. the Republican AIDS Center from different http://www.rcaids.kz/files/000000185.pdf shows that eight service registers; nomination multipliers were utilized (based on information from key informants), and multiplier from RDS-based IBBS. Final estimates were discussed with experts and the report presents averaged estimates; Aggregates from 16 regional estimates. National / Both N/A Sentinel snowball 100,000 http://www.emcdda.europa.eu/publications/country- Methodology not well statistics method; 3 methods: [0.91] overviews/kz documented, in this source. from 2011 Multiple analysis, 124,400 Original description of multiplication and [1.13] methods not locatable. These nomination) and focus Separate estimates Cites: UNODC Regional Office for Central Asia data appear to be a root groups source for other expert estimates ranging from 100,000 to 125,000 National/ Both NA Expert consultation. 122,850 Needle and Zhou, 2010. Cites data from UK Foreign Original source not locatable. 2009 (115,500-130,200) and Commonwealth Office Web site, [1.11] Asia and Oceania: Kazakhstan, http://www.fco. gov.uk/en/about-the-fco/country- profiles/asiaoceania/ kazakhstan/ Link not currently accessible National / Both N/A UNODC national 186,000 Accessibility of HIV Prevention, Treatment and Care Details not available. This 2010 project Officers [1.68] Services for People who use Drugs and incarcerated source seems to make people in Azerbaijan, Kazakhstan, Kyrgyzstan, inferences based on previous Tajikistan, Turkmenistan and Uzbekistan UNODC prevalence estimates from EMCDDA, possibly updated. 9 Cites: C. Cook & N. Kanaef, 2008 Sub cites:: http://www.emcdda.europa.eu/publications/country- overviews/kz National / Both N/A Indirect methods: 102,500 IDU reference Group website This estimate is derived from 2008 Estimated number of [0.93] http://www.idurefgroup.unsw.edu.au/regional-data- the population prevalence PWID (for 2008) and-maps/Central_Asia provided by Mathers et al. derived from Not based on any direct prevalence estimate observation or data. to the 2008 15-64 year old population from Mathers National/ Both N/A Indirect methods 100,000 Mathers B, et al. (2008). No direct methods seem to 2006 [0.91] be utilized. Possible that some documents with direct methods were utilized in lit review, but not specified. Experts consulted on final estimates. Likely a Delphi estimate not based on strong source data. National/ Both N/A Lit review 174,000 (97,000- Aceijas 2004 No study methodologies 2004 250,000) [1.58] National/ Both N/A Nomination, multiplier 1,025-1,156 per Dehne KL and Kobyshcha Y, :2000. Referenced in: Methods not described in 2004 method. 100,000 of “Audit of the number of injecting drug users in Central referencing document. population and Eastern Europe and Central Asia.” New York: Original reference not [<1.16] UNODC 2004. located. National / Both N/A Not disclosed - 200,000 Summary Country Profile for HIV/AIDS Treatment Details not available. 2004 UNODC [1.81] Scale-Up. * UNAIDS 2013. 10 Kyrgyzstan Prevalence of HIV among PWID – 14.6* 2012 Country Population (Age 15-64): 3,590,456 PWID size estimate Male/ (range) City/ Female [% of 2012 Limitations/biases; Region/ Year /both Age Estimation method(s) population] Source Comments National/ Both N/A Indirect Pending Communication with AIDS Republican Centre, Multiplier methods, Global 2013 Ministry of Health, Kyrgyzstan; UNAIDS Fund-supported size communication estimation (independent of IBBS). RDS used for sampling/survey data which were triangulated with service multipliers. National/ Both N/A Indirect Pending AIDS Republican Centre, Ministry of Health, Multiplier method 2011 Kyrgyzstan; UNAIDS communication integrated within the 2011 IBBS round. National / Both N/A Indirect 25,000 [0.70] Mathers B, et al 2008 2006 Original source likely to be UNODC estimation study (EMCDDA 2006). National/ Both N/A ”In 2006, the UNODC 25,000 European Monitoring Centre for Drugs and Drug Methods not well 2006 estimation study using the [0.70] Addiction (EMCDDA) described. Could not locate multiplication (nominative) http://www.emcdda.europa.eu/publications/country source report. method, the number of -overviews/kg problem drug users (PDU) in Kyrgyzstan was Probable source material for Mathers B, et al (2008). estimated at 26 000, of whom 25 000 were intravenous drug users (IDU).” National / Both N/A UNODC national project 44,000 C. Cook & N. Kanaef, 2008 Likely based on the Aceijas 2010 Officers [1.23] 2004 resource cited. National/ Both NA Previous lit review updated 44,398 Aceijas C, et al, 2006. Creates a new midpoint for 2006 with one expert opinion [1.24] Kyrgyzstan from Godinho estimate Cites: Godinho J, et al 2005. 2005 paper. National/ Both NA Expert opinion estimate 70,000 Godinho J et al. 2005. Appears to be Delphi 2005 [2.14] estimate not based on data. 11 National / Both N/A Lit review 21,000 (19,000 Aceijas 2004 No direct methods seem to statistics – 23,000) be utilized. Possible that from 1998- [0.58] some documents with 2003 direct methods were utilized in lit review, but not specified. Experts consulted on final estimates. Likely a Delphi estimate not based on good data. National/ Both N/A unknown 80,000 – United Nations Office on Drugs and Crime, via Source material not 2001 120,000 Summary Country Profile for HIV/AIDS Treatment locatable. May not be data- [<3.34] Scale-Up. based. * UNAIDS Global Report on the Global AIDS Epidemic, 2013. 12 Libya Prevalence of HIV among PWID – 87.0 (Tripoli only)* 2012 Country Population (age 15-64): 4,052,890 Male/ City/ Region Female /Year /both Age Estimation method(s) PWID size estimate (range) Source Limitations/biases; Comments National/ Both All Adapted from Mathers 7,206 (4,633- 9,779) [0.18] Laith J. 2008 No details on adjustment from 2010 registry data. National / Both All Lit review 7,000 (5,000-10,000) [0.17] Aceijas 2004 No details on study 2004 methodologies National / Both All Cumulative registry of 1,685 [0.04] Mathers B, et al (2008). Registry data undercounts PWID. 2001 drug users * Mirzoyan 2013. The HIV prevalence is only for Tripoli. 13 Myanmar Prevalence of HIV among PWID – 18.0* 2012 Country Population (Age 15-64): 36,697,849 Male/ City/ Region/ Female PWID size Year /both Age Estimation method(s) estimate (range) Source Limitations/ biases; Comments National / Male 15-49 Estimated percentage 75,000 [0.50] National AIDS Programme (Strategic Based solely on calculation of 2010 of adult male Information and M&E Working Group): HIV PWID being 0.5% of the male population and Estimates and Projections Asian population aged 15-49 years and consensus Epidemiological Model, Myanmar 2010- agreed upon by in-country 2015 stakeholders; it was not based on enumeration or other indirect methods of estimation. National / Both N/A Lit review 195,000 (90,000, Aceijas 2004 No study methodologies 2004 300,000) [0.53] National / Both N/A Indirect estimates from 75,000 (60,000– Mathers B et al (2008). Methodology not described. Much 2007 registered drug users. 90,000) [0.20] of the HIV and PWID research uses this estimate as a base. National / Both N/A Author’s personal 60,000–90,000 Sharma M 2009 Not based on systematic 2009 communication with [0.24] methodologies. WHO Country Office and National AIDS Programme Staff. * Global AIDS Response Progress Report 2012. 14 Philippines Prevalence of HIV among PWID – 13.6* 2012 Country Population (Age 15-64): 59,657,860 City/ Male/ Region/ Female PWID size estimate Year /both Age Estimation method(s) (range) Source Limitations/biases; Comments National/ 2009 90% N/A Based on national 12,705-21,567 Philippine National AIDS Council/ No indirect estimations are males; consensus meeting [0.030-0.04] Philippine Estimates of the Most At-Risk available for the nation. This is 10% which took into Population and People Living with HIV based on 0.03%-0.04% of the adult females account the Dangerous (2011) male population (15-49 years) Drug Board’s survey being PWID with the exception of indicating that 2-4% of the 3 cities which had reported general population use networks of PWID that were illegal drugs, and 0.89% higher than the estimated 0.03- of them were PWID. 0.04% of the general population This rate was applied to being PWID. all but three sites (Cebu City, General Santos City, and Zamboanga City), where identified networks of PWID were estimated (method not specified). In these three cities, the reported figures from the identified networks were used instead of the percentage of 0.03% of the general population being PWID. National / 2004 Both N/A Lit review 17,000 (10,000- Aceijas 2004 No study methodologies reported 24,000) [0.028] Regional Cebu City, Both N/A Not reported. It only Cebu City estimated Philippine National AIDS Council/ Methodology not reported. Santos City, states that they 2,027; PWID in the Philippine Estimates of the Most At-Risk Zamboanga / “identified networks of city, 627 in General Population and People Living with HIV 2011 IDUs”. Santos City; 1,190 in (2011) Zamboanga City. Pasay, Quezon, Both N/A Data from 10 sentinel 7,239-14,478 2007 HIV Estimates in the Philippines. Appears to aggregated count. Baguio, sites; no method given. [0.012-0.024] Unpublished, Cited by Philippines Country Angeles, Cebu, Profile in Evidence to Action website Iloilo, Aidsdatahub.org; Cagayande 15 Oro, Davao, General Santos, Zamboanga / 2007 * Philippine National AIDS Council. Philippines: 2012 Global AIDS Response Progress Report. 16 Tajikistan Prevalence of HIV among PWID – 13.5* Country Population 15-64: 4,878,289* City/ Male/ Region/ Female Estimation PWID size Year /both Age method(s) estimate (range) Source Limitations/biases; Comments National/ Both N/A Indirect 26,500 Analysis of the situation and resources to implement a No direct methods seem to be 2013 comprehensive package of services for injecting drug users in the utilized. Multiplication of Republic of Tajikistan, 2013 (in Russian) (Анализ ситуации и benchmarks by the assigned ресурсов для внедрения комплексного пакета услуг для ПИН в coefficient in each area. The Республике Таджикистан, 2013) coefficient was assigned based on expert opinion from 1.5 for areas with good coverage to 5 for areas with low coverage. National / Both N/A Indirect 25,000 (20,000- Study of high-risk groups and the capacity of local organizations in Multiplier methods were used, 2009 30,000)[0.51] the preparation and execution of programs to prevent HIV / AIDS but have not been able to locate among PWID, CSWs and their clients in Tajikistan (personal detailed documentation. communication, Republican Centre on AIDS Prevention and Control, Implemented by the AIDS Ministry of Health and Social Protection, Republic of Tajikistan)). Project Management Group in 2009. National / Both N/A Lit review 53,000 (43,000- Aceijas 2004 No direct methods seem to be statistics 62,000) [1.09] utilized. Possible that some from 1998- documents with direct methods 2003 were utilized in lit review, but not specified. Experts consulted on final estimates. Likely a Delphi estimate not based on good data. National/ Both 15-64 ODC 20,000 use Illicit Drug Trends in Central Asia, UNODC Regional Office for Central It is not well-specified whether 2006 estimate opiates regularly, Asia, 2008, http://www.unodc.org/documents/regional/central- the estimate is only injecting but injection asia/Illicit%20Drug%20Trends_Central%20Asia-final.pdf users, or all opiate users. Not rates are 30-90%) clear if source comes from any [0.41] Cited by Needle and Zhou, 2010. data, or is only Delphi. National/ Both N/A Indirect 17,000 (no range Mathers 2008 No direct methods seem to be 2006 methods given) utilized. Possible that some [0.35] documents with direct methods were utilized in lit review, but not specified. Experts consulted on final estimates. Likely a Delphi estimate not based on strong source data. 17 National / Both N/A Indirect 15,000 [0.31] C. Cook & N. Kanaef, 2008; Not clear if source estimate is 2008 methods Original source: United Nations Office on Drugs and Crime Regional based on data sources, or only Office for Central Asia (2007) Compendium on Drug Related expert opinion. Statistics 1996–2007. National / Both N/A Mapping 34,000 Accessibility of HIV Prevention, Treatment and Care Services for Not clear if source estimate is 2008 [0.70] People who use Drugs and incarcerated people in Azerbaijan, based on data sources, or only Kazakhstan, Kyrgyzstan, Tajikistan, Turkmenistan and Uzbekistan expert opinion. UNODC; Cites: UNAIDS , Tajikistan, country progress report (January 2008). * UNAIDS 2012. 18 Uzbekistan Prevalence of HIV among PWID – 7.3* 2012 Country Population (Age 15-64 years): 19,042,221 PWID size estimate (range) Male/ [proportion City/ Region/ Female of 2012 Year /both Age Estimation method(s) population] Source Limitations/biases Comments National/2011, Multiplier method integrated within 2013 Both N/A Indirect Pending UNAIDS communication the 2011 and 2013 IBBS rounds. National / Both N/A Lit review 87,000 Aceijas 2004 No direct methods seem to be utilized. statistics from (52,000- Possible that some documents with 1998-2003 122,000) direct methods were utilized in lit [0.45] review, but not specified. Experts consulted on final estimates. Likely a Delphi estimate not based on good data` National / 2009 Both N/A 0.5% population prevalence 80,000 (2006) EMCDDA From source: “Such a large estimated based on snowball [0.42] http://www.emcdda.europa.eu/publications number of injecting drug users raises assessment. Size estimate /country-overviews/uz doubts among narcological specialists derived using this and requires specification and prevalence and total confirmation by means of the population. scientific performance of epidemiological surveys.” National / 2008 Both N/A Not given 79,300 Illicit Drug Trends in Central Asia, UNODC UNODC stated as source. Possibly (130,000 Regional Office for Central Asia, 2008, derived from previous “snowball dependent, http://www.unodc.org/ documents/ assessment.” 61% inject) regional/central-Asia/Illicit%20Drug [0.41] %20Trends_Central%20Asia-final.pdf National/ 2001 Both N/A Unknown >100,000 United Nations Office on Drugs and Crime, Source data not found. [>0.53] via Summary Country Profile for HIV/AIDS Treatment Scale-Up. * UNAIDS 2013. 19 APPENDIX II REVIEW OF SELECT LITERATURE ON SIZE ESTIMATION STUDIES References 1. United Nations Office of Drug and Crime. World Drug Report, 2013. New York, NY: United Nations. 2. UNAIDS. Global Report: UNAIDS report on the global AIDS epidemic 2013. 3. Gouws E, Cuchi P, International Collaboration on Estimating H I V Incidence by Modes of Transmission. Focusing the HIV response through estimating the major modes of HIV transmission: a multi-country analysis. Sex Transm Infect. Dec 2012;88 Suppl 2:i76-85. 4. UNAIDS and World Health Organization. Guidelines on estimating the size of populations most at risk to HIV. Geneva, Switzerland: WHO; 2010. 5. Paz-Bailey G, Jacobson JO, Guardado ME , et al. How many men who have sex with men and female sex workers live in El Salvador? Using respondent-driven sampling and capture-recapture to estimate population sizes. Sex Transm Infect. Jun 2011;87(4):279-282. 6. Rocchetti I, Bunge J, Bohning D. Population size estimation based upon ratios of recapture probabilities. Annals of Applied Statistics. 2011;5(2B):1512-1533. 7. Odek WO, Githuka GN, Avery L, et al. Estimating the size of the female sex worker population in Kenya to inform HIV prevention programming. PLoS One. 2014;9(3):e89180. 8. Surowiecki J. The Wisdom of Crowds: Random House; 2005. 9. Roberts A, Mathers B, L. D. Women who inject drugs: A review of their risks, experiences and needs. In: Use RGttUNoHaID, ed. Sydney, Australia: National Drug and Alcohol Research Centre (NDARC), University of New South Wales; 2010. 10. Advisory Council on the monitoring and assessment of the situation on HIV / AIDS. National Report on the progress made in the implementation of global response to AIDS, Republic of Belarus Reporting period: January 2010 - December 2011: National Report on the Declaration of Commitment on HIV / AIDS in 2010-2011, Belarus; 2012. 11. Republican Center for Hygiene, Epidemiology and Public Health. [Behaviors and HIV knowledge among injecting drug users, female sex workers and men who have sex with men: Study Report]. Available at: www.belaids.net/uploads/docs/den_2011.pdf. 12. IDU Reference Group. IDU Reference Group Country data and Maps, Belarus. Available at: http://www.idurefgroup.unsw.edu.au/country-data-and-maps/Belarus, 2014. 13. European Monitoring Centre for Drugs and Drug Addiction (EMCDDA). Country Overview: Belarus. Available at: http://www.emcdda.europa.eu/publications/country-overviews/by. 14. Aceijas C, Stimson GV, Hickman M, et al. Global overview of injecting drug use and HIV infection among injecting drug users. AIDS. Nov 19 2004;18(17):2295-2303. 15. Mathers BM, Degenhardt L, Phillips B, et al. Global epidemiology of injecting drug use and HIV among people who inject drugs: a systematic review. Lancet. Nov 15 2008;372(9651):1733-1745. 16. World Health Organization. Summary Country for HIV/AIDS Treatment Scaleup UNODC, Overview Belarus. 2005. 17. Monitoring Centre for Drugs and Drug Addiction. Drug abuse and illicit drug trafficking in the Republic of Belarus in 2007: Grodno State Medical University; 2009. 18. Ministry of Health of the People's Republic of China. 2012 China AIDS Response Progress Report: Ministry of Health of the People's Republic of China; 2012. 19. Zhang L, Chow EP, Jing J, et al. HIV prevalence in China: integration of surveillance data and a systematic review. Lancet Infect Dis. Nov 2013;13(11):955-963. 20. UNAIDS. AIDS Info Database Available at: http://www.unaids.org/en/dataanalysis/datatools/aidsinfo/. 21. Qian HZ, Vermund SH, Wang N. Risk of HIV/AIDS in China: subpopulations of special importance. Sex Transm Infect. Dec 2005;81(6):442-447. 22. Needle RH, L. Z. HIV Prevention among Injection Drug Users Strengthening U.S. Support for Core Interventions: CSIS Global Health Policy Center; 2010. 23. Wang L, Wang N, Wang L, et al. 2007 Estimates for People at Risk for and Living With HIV in China: Progress and Challenges. J Acquir Immune Defic Syndr 2009;50(4). 24. Lu F, Wang N, Wu Z, et al. Estimating the number of people at risk for and living with HIV in China in 2005: methods and results. Sex Transm Infect. 2006;82(Suppl III):iii87–iii91. 25. World Health Organization. Summary Country for HIV/AIDS Treatment Scaleup UNODC, Overview China. Available at: www.who.int/hiv/HIVCP_CHN.pdf. 26. Ni MJ, Fu LP, Chen XL, Hu XY, Wheeler K. Net financial benefits of averting HIV infections among people who inject drugs in Urumqi, Xinjiang, Peoples Republic of China (2005–2010). BMC Public Health 2012;12(572). 27. Guo W, Bao S, Lin W, et al. Estimating the Size of HIV Key Affected Populations in Chongqing, China, Using the Network Scale-Up Method. PLOS One. 2013;8(8). 28. Jia Y, Sun J, Fan L, et al. Estimates of HIV prevalence in a highly endemic area of China: Dehong Prefecture, Yunnan Province. International Journal of Epidemiology. 2008;37:1287-1296. 29. National AIDS Control Organisation (Department of AIDS Control). Annual Report 2012-2013. (India). Available at: http://www.naco.gov.in/upload/Publication/Annual%20Report/Annual%20report%202012-13_English.pdf. Accessed 4/4/2014. 30. UNAIDS. Global AIDS Response Progress Report 2013. 31. UNODC and Joint United Nations Programme on HIV/AIDS. Audit of the Number of Injecting Drug Users in Central and Eastern Europe and Central Asia. 2004. 32. European Monitoring Centre for Drugs and Drug Addiction (EMCDDA). Country overview: Kazakhstan. Available at: http://www.emcdda.europa.eu/publications/country-overviews/kz. 33. IDU Reference Group. IDU Reference Group Country data and Maps, Kazakhstan. Available at: http://www.emcdda.europa.eu/publications/country-overviews/kz. 1 34. UNODC and Canadian HIV/AIDS Legal Network. Accessibility of HIV Prevention, Treatment and Care Services for People who use Drugs and Incarcerated People in Azerbaijan, Kazakhstan, Kyrgyzstan, Tajikistan, Turkmenistan and Uzbekistan; Legislative and Policy Analysis and Recommendations for Reform 2010. 35. World Health Organization. Summary Country for HIV/AIDS Treatment Scaleup UNODC, Overview Kazakhstan. Available at: www.who.int/hiv/HIVCP_KAZ.pdf. 36. National AIDS Programme. UNGASS Country Progress Report, Lybia, Reporting period: January 2010 – December 2011: National AIDS program; 2011. 37. Laith J, Abu-Raddad F, Akala A, al. e. Characterizing the HIV/AIDS epidemic in the Middle East and North Africa: Time for Action. : World Bank.; 2008. 38. Mirzoyan L, Berendes S, Jeffery C, et al. New Evidence on the HIV Epidemic in Libya: Why Countries Must Implement Prevention Programs Among People Who Inject Drugs. JAIDS Journal of Acquired Immune Deficiency Syndromes. 2013;62(5):577-583. 39. UNAIDS. Global AIDS Response Progress Report 2012. 40. Sharma M, Oppenheimer E, Saidel T, Loo V, R. G. A situation update on HIV epidemics among people who inject drugs and national responses in South-East Asia Region. AIDS. 2009;23:1405-1413. 41. IDU Reference Group. IDU Reference Group Country data and Maps, Myanmar. Available at: http://www.idurefgroup.unsw.edu.au/country-data-and-maps/myanmar. 42. National Epidemiology Center, Department of Health (Philippines). Philippine Priority Areas for HIV lntervention (PAHI) 2012. 43. Samonte GM, Aranjuez E, Bayoneta-Leis N, Dionisio-Libatique R, Z. M. Country Progress Report Philippines: Philippine National AIDS Council; 2011. 44. UNAIDS. HIV in Asia and the Pacific 2013. 45. National Epidemiology Center. 2012 Global AIDS Response Progress Report, Philippines: Department of Health. (Philippines); 2012. 46. Philippine National AIDS Council. Philippine Estimates of the Most At-Risk Population and People Living with HIV. 2011:8-9. 47. Department of Health, National Epidemiology Center, Philippines. Brief report: 2013 Integrated HIV Behavioral and Serologic Surveillance: Males who have Sex with Males (MSM) and Male Injecting Drug Users (IDU). Manila, Philippines; 2013. 48. Cook C, N. K. Global State of Harm Reduction 2008: Mapping the response to drug-related HIV and hepatitis C epidemics In: Association IHR, ed: HR2, the Human Rights Monitoring and Policy Analysis Programme of the International Harm Reduction Association.; 2008. 49. UNODC Regional Office for Central Asia. Illicit Drug Trends in Central Asia : UNODC Regional Office for Central Asia; 2008. 50. European Monitoring Centre for Drugs and Drug Addiction (EMCDDA). Country overview: Uzbekistan. Available at: http://www.emcdda.europa.eu/publications/country-overviews/uz. 51. World Health Organization. Summary Country for HIV/AIDS Treatment Scaleup UNODC, Overview Uzbekistan. 2005. 2