Estimating the number of people who inject drugs (PWID) in two urban areas in Mozambique using four different methods, 2014 Presented by: Isabel Sathane Coauthors: R. Horth, M. Boothe, C. Baltazar, P. Young, H. F. Raymond, E. Teodoro, M. Goveia, T. Lane and W. McFarland
Introduction PWID are a key population No prior data on HIV prevalence or population size estimation (PSE) in Mozambique National HIV Prevalence is 11.5% in adults aged 15-49 years (2012 National AIDS Indicator Survey) UNAIDS Modes of Transmission model suggests that as much as 10% of HIV infection was attributed to injection drug use (UNAIDS, 2007)
Methodology IBBS* conducted in Maputo and Nampula: How: Respondent driven-sampling When: Oct 2013-Mar 2014 Who: 18 y.o.; any lifetime drug injection; lived, worked or socialized in survey area. How many: 351 in Maputo and 139 in Nampula Pop. size est. methods: No gold standard: must use multiple methods Unique object, wisdom of the crowds, sequential sampling and literature review *Integrated biological and behavioral survey
Methodology: Unique object multiplier Two sources of data: 1. Distributed flashlight keychain to PWID in the community 2. Included question in the IBBS survey tool asking whether participants had received the object Calculated estimate for each survey using: N = n / p N: is the estimated key population size n: is the total number of flashlight keychain distributed p: is the RDSAT adjusted proportion of survey participants having received a flashlight keychain
Methodology: Wisdom of the crowds Ask IBBS respondents their best guess of the number of PWID in the respective city The median of all respondents was used as the best estimate Participants who reported the same number for both personal network size and population size estimate were excluded due to possible confusion between questions.
Methodology: Sequential sampling for PSE Simulation using RDS-A software (Hard-to-Reach Population Methods Research Group) Uses the ordered sequence of self-reported participant network sizes Theory: Participants w/ larger networks more likely to sample first The change in network size distribution in successive waves reflects the depletion of population Use this quantify the PSE
Methodology: Literature review and consensus meeting Literature review: World Drug Report 2014: 0.17% for Africa Applied to 2014 census projection in each city: Maputo: 768,706 inhabitants ages 15-64 Nampula: 339,028 inhabitants ages 15-64 Consensus meeting: Median of four estimates used as best estimate Lowest plausible estimate is IBBS sample size
Results: Unique object multiplier Unique object multiplier inputs: # distributed: 408 Maputo; 447 Nampula # received in IBBS: 92 Maputo; 26 Nampula % received in IBBS: 19.8% Maputo; 11.4% Nampula Pop size estimate based on unique object: Maputo: 2061 Nampula: 3921
Results: Wisdom of the crowds & sequential sampling Wisdom of the crowds Median value of reported best guesses : Maputo: 300 (n=233) Nampula: 70 (n=67) Sequential sampling Mean network sizes: 30 Maputo;14 Nampula Sequential sampling estimate produced: Maputo: 2869 Nampula: 463
Estimated HIV Prevalence among PWID
Number of HIV-positive PWID Applying prevalence of HIV to Population size estimate, gives number of HIV-positive PWID of: 847 in Maputo 191 in Nampula Conclusion Feasibility of using multiple methods to estimate pop. size of PWID in Mozambique Given limited pop size and high prevalence, interventions could be practical and affordable in this population
Acknowledgements & Disclaimer Acknowledgements: The authors would like to acknowledge the contribution of all institutions involved in the preparation and implementation of survey activities, survey participants, field personnel and members of the Mozambique IBBS technical working group, for their important roles in the success of this survey. Disclaimer: This abstract results from research supported by the President s Emergency Plan for AIDS Relief (PEPFAR), through the US Department of Health and Human Services and the Centers for Disease Control and Prevention (CDC) Mozambique Country Office, under the terms of the Cooperative Agreement #U2GPS001468. The views expressed in this poster do not necessarily reflect the views of the CDC or the US Government.
References UNAIDS Reference Group Consultation on Population Size Estimation Based on Respondent-driven Sampling http://wiki.stat.ucla.edu/hpmrg/index.php/unaids_reference_group_cons ultation Handcock, M. S., & Gile, K. J. (2015). sspse: Estimating Hidden Population Size using Respondent Driven Sampling Data. Los Angeles, CA. Retrieved from http://cran.r-project.org/package=sspse Handcock, M. S., Gile, K. J., & Mar, C. M. (2014). Estimating hidden population size using Respondent-Driven Sampling data. Electronic Journal of Statistics, 8(1), 1491 1521. UCSF Global Health Sciences. (2014). Field Implementation: Unique object and unique event multipliers operations manual. Retrieved from http://globalhealthsciences.ucsf.edu/sites/default/files/content/pphg/ibbs/pdf/ UniqueObjectAndEventOpMan.pdf National Institute of Health of Mozambique (INS), National Institute of Statistics of Mozambique (INE) & ICF Macro. (2010). Inquérito Nacional de Prevalência, Riscos Comportamentais e Informação sobre HIV e Sida em Moçambique 2009. Cavelton, Maryland, USA: INS, INE & ICF Macro
Author contact Isabel Sathane (ITECH) isabels@itech-mozambique.org Roberta Horth (UCSF) Roberta.horth@gmail.com
Author contact Isabel Sathane (I-TECH) isabels@itech-mozambique.org Roberta Horth (UCSF) roberta.horth@gmail.com Makini Boothe (UCSF) mboothe.uscf@gmail.com Cynthia Baltazar (INS) cynthiasema@yahoo.com Peter Young (CDC) fqm1@cdc.gov H. F. Raymond (UCSF) hfisherraymond@yahoo.com Eugenia Teodoro (MoH) guelfrida@gmail.com Maria Lidia Gouveia (MoH) vovotetchauque@gmail.com Tim Lane (UCSF) tim.lane@uscf.edu Willi McFarland (UCSF) willi_mcfarland@hotmail.com
Prevalence (%) Number among population aged 15-64 Country low medium high low medium high Year Method Kenya 0.04 0.21 0.52 10,000 49,167 120,000 2011 Indirect Estimate (Respondent Driven Sampling) Mauritius 1.07 10,000 2011 Indirect Estimate Seychelles 2.3 1,283 2011 Indirect Estimate Tanzania 0.08 0.12 0.17 20,000 30,000 42,500 2013 Delphi Method Egypt 0.12 0.18 0.24 57,000 90,000 120,000 2010 Official government estimate with no methodology reported Morocco 0.08 18,000 2011 Official government estimate with no methodology reported Tunisia 0.12 9,000 2011 Official government estimate with no methodology reported South Africa 0.21 67,000 2008 General Population Survey Liberia 0.03 621 2011 Official government estimate with no methodology reported Senegal 0.02 0.02 0.02 1,281 1,324 1,367 2011 Indirect Estimate (Capture-recapture) Source: data.unodc.org, accessed Sept 2015