Monday 17 December 2018

M28: Design and Implementation of Respondent-Driven Sampling (RDS) and RDS Data Analysis


Let’s say you want to collect quantitative data from sex workers or men who have sex with men to learn more about their sexual behaviours and to collect biological specimens to test for HIV, about people who inject drugs to learn about their high risk injecting practices and their access to programs, young people living on the street to learn about their lifestyle habits in order to plan programs, Syrian migrants in Istanbul to learn about their political opinions, or former survivors or perpetrators of the Rwanda massacre living in Europe to learn more about their psychological status. Because of their particular circumstances, it is difficult to generate a sampling frame from which to gather a representative sample from these populations. They are considered hard-to-reach populations for survey research purposes. At the same time, they are potentially networked (i.e., they know each other), such that you may be able to find a handful of group members through their contacts with governmental and non- governmental organizations and other sources. What is the best sampling method available to capture these types of populations? Can you sample them with something other than a convenience sampling method such as snowball sampling? 

Because most hard-to-reach populations do not have sampling frames from which to draw a probability sample, researchers often rely on convenience sampling methods. However, these can provide biased data. Over the past decade, respondent driven sampling (RDS) has been highlighted as a robust and effective method to recruit large samples (100+) of hard-to-reach populations that are connected through social networks. RDS begins with the researcher purposefully recruiting a set number of eligible individuals, called “seeds”, who themselves then recruit a set number of other eligible individuals from among their network members. This recruitment process produces “recruitment chains”, with several “waves” of recruits. Participants receive compensation for both being interviewed and bringing in new participants. When all of the assumptions are met, the growing sample is hypothesized to eventually reach ‘convergence’, whereby the prevalence values of the characteristics of interest stabilize. An important part of RDS is a statistical analysis process based on Markov chain and biased network theories, where estimates are adjusted using each participant’s social network size and information about who recruited whom.

RDS has been most widely used to collection HIV biological and behavioural data on populations of people who inject drugs, men who have sex with men and people who inject drugs. However, RDS has also been used successfully to gather data on the standards of work and living among migrant populations, salaries and union membership among jazz musicians in the United States, political affiliations and life experiences among Vietnam veterans, school enrolment and income earning among youth who live on the streets, pregnancies related to sexual violence among women in the Congo, sexual practices among high risk heterosexual men and women in South Africa, and numerous other outcomes among populations that are networked through social affiliations.

This course will provide participants with practical and relevant up-to-date information about the methodological and theoretical issues and analytical concerns. It will draw on a variety of lectures, presentations of actual field research, hands-on analysis and practical experience in planning and implementing surveys using RDS. 

Key topics of the course are:

  • Steps in RDS survey design and implementation including formative research, choosing survey sites, selection of staff, selection of seeds, incentives, survey flow, and ending RDS
  • Data collection forms necessary for RDS implementation
  • Coupons and coupon reduction and management
  • Overview and structure of RDS-Analyst (RDS-A)
  • The meaning and importance of technical concepts such as homophily, social network sizes, differential activity, coarsened data and convergence
  • Preparation of data files for RDS-A (Excel, STATA, SPSS, Text, SAS, etc.)
  • Overview of the types of analysis performed by RDS-A, including weighted population estimates, stratified analyses, cross tabs and regression analyses
  • Export of individualized weights to conduct regression analyses

Teaching methods

The course consists of lectures, exercises and individual work on computers so that participants can get hands-on experience in RDS design and data management and analysis. Participants are encouraged to bring their own RDS data sets to work on, particularly if they need to write up the results from RDS surveys for surveillance reports or scientific manuscripts. Those without actual data will be provided with datasets. Participants should be familiar with basic statistics and Excel software. They will be given statistical programmes at the course to work with.

An important part of the course is group or individual work during which participants can either:

  • develop protocols for a planned RDS survey
  • conduct analysis using RDS data (participants can bring their own data or will be given data by facilitators).

Target Audience

Professionals and students considering or currently working on designing and implementing an RDS survey and/or have data from an RDS survey and need to assess and analyse it.

Lecturers and facilitators

  • Lisa Johnston, PhD, Independent Consultant; Senior analyst for University of California, San Francisco, Global Health Sciences; Assistant/adjunct professor, Tulane University, School of International Public Health and Tropical Medicine
  • Ivana Božicević, MD, DrPH, WHO Collaborating Centre for HIV Strategic Information, School of Medicine, University of Zagreb, Croatia
  • Zoran Dominković, WHO Collaborating Centre for HIV Strategic Information, School of Medicine, University of Zagreb, Croatia

Course organizer:

Lucija Šikić, WHO Collaborating Centre for HIV Strategic Information

Duration and site: 

The course takes place in five days, 24 – 28 October 2016 at the Andrija Stampar School of Public Health in Zagreb, Croatia.

To apply for the course please contact Ms Lucija Šikić at:  or apply on-line at

The course is organized and will be held at:

WHO Collaborating Centre for HIV Strategic Information
Andrija Stampar School of Public Health
School of Medicine, University of Zagreb
Rockefellerova 4
10 000 Zagreb, Croatia
Phone: + 385 1 45 90 142/ 45 90 100
Fax: + 385 1 46 84 212

The course fee is 950 USD and includes training materials and lunch and coffee/tea breaks during the course.



Recommended reading for the course:

Introduction to HIV/AIDS and sexually transmitted infection surveillance: Module 4: Introduction to Respondent Driven Sampling. Geneva, WHO/UNAIDS, 2013. 

Available at:

Introduction to HIV/AIDS and sexually transmitted infection surveillance:   Module 4: A Guide to Using RDS Analyst and NetDraw.  Geneva, WHO/UNAIDS, 2014.

Available at:

Johnston LG, Malekinejad M, Kendall C, Iuppa IM, Rutherford GW. Implementation challenges to using respondent-driven sampling methodology for HIV biological and behavioral surveillance: field experiences in international settings. AIDS Behav 2008;12(4 Suppl):S131-41.

Magnani R, Sabin K, Saidel T, Heckathorn D. Review of sampling hard-to-reach and hidden populations for HIV surveillance. AIDS 2005;19 Suppl 2:S67-72.

Malekinejad M, Johnston LG, Kendall C, Kerr LR, Rifkin MR, Rutherford GW.

Using respondent-driven sampling methodology for HIV biological and behavioral surveillance in international settings: a systematic review.

AIDS Behav 2008;12(4 Suppl):S105-30.

Gile KJ, Handcock MS. Respondent-Driven Sampling: An Assessment of Current Methodology. Sociol Methodol 2010; 40(1):285-327.



Download the Programme

Programme - Design and Analysis of RDS Surveys.docx

Created by