AIDS and Behavior

, Volume 17, Issue 7, pp 2313–2340 | Cite as

Respondent Driven Sampling for HIV Biological and Behavioral Surveillance in Latin America and the Caribbean

  • Jane R. Montealegre
  • Lisa G. Johnston
  • Christopher Murrill
  • Edgar Monterroso
Substantive Review

Abstract

Since 2005, respondent driven sampling (RDS) has been widely used for HIV biological and behavioral surveillance surveys (BBSS) in Latin America and the Caribbean (LAC). In this manuscript, we provide a focused review of RDS among hard-to-reach high-risk populations in LAC and describe their principal operational, design, and analytical considerations. We reviewed published and unpublished reports, protocols, and manuscripts for RDS studies conducted in LAC between January 1, 2005 and December 31, 2011. We abstracted key operational information and generated summary statistics across all studies. Between 2005 and 2011, 87 RDS studies were conducted in 15 countries in LAC (68 % in South America, 18 % in Mexico and Central America, and 14 % in the Caribbean). The target populations were primarily men who have sex with men (43 %), sex workers (29 %), and drug users (26 %). Study considerations included establishing clear eligibility criteria, measuring social network sizes, collecting specimens for biological testing, among others. Most of the reviewed studies are the first in their respective countries to collect data on hard-to-reach populations and the first attempt to use a probability-based sampling method. These RDS studies allowed researchers and public health practitioners in LAC to access hard-to-reach HIV high-risk populations and collect valuable data on the prevalence of HIV and other infections, as well as related risk behaviors.

Keywords

HIV/AIDS Respondent driven sampling Surveillance Latin America Caribbean Key populations at risk 

Resumen

Desde el 2005, el muestreo dirigido por participantes (o respondent driven sampling, RDS) se ha utilizado ampliamente en Latinoamérica y el Caribe (LAC) para las encuestas de vigilancia biológica y de comportamientos de riesgo para el VIH. En este artículo, ofrecemos una revisión enfocada de los estudios RDS que se han realizado en LAC entre las poblaciones de alto riesgo y difícil de acceder y describe retos operacionales, de diseño, y analíticos y sus consideraciones. Examinamos informes publicados e inéditos, protocolos, y manuscriptos sobre los estudios RDS realizados en LAC entre enero 1, 2005 a diciembre 31, 2011. Resumimos la información operacional clave de todos los estudios y generamos estadísticas sumarias. Encontramos que entre el 2005 al 2011 se realizaron 87 estudios RDS en 15 países de LAC (68 % en Sudamérica, 18 % en México y Centroamérica, y 14 % en el Caribe). Las poblaciones objetivas fueron principalmente hombres que tienen sexo con hombres (43 %), trabajadoras sexuales (29 %) y usuarios de drogas (26 %). Los desafíos principales incluyeron el establecimiento de criterios de elegibilidad, la medición del tamaño de la red social, la toma de muestras para las pruebas biológicas, entre otros. La mayoría de los estudios revisados son los primeros en sus respectivos países que recopilan datos sobre las poblaciones de difícil acceso y el primer intento de utilizar un método de muestreo basado en la probabilidad. Estos estudios RDS permitieron a los investigadores y a los profesionales de salud pública accesar a poblaciones de alto riesgo de difícil acceso y recolectar valiosos datos sobre la prevalencia del VIH y de otras infecciones, así como riesgos de conducta relacionados.

Introduction

Latin America is largely characterized as having low-level and concentrated HIV epidemics, with a regional prevalence of 0.5 % among adults (ages 15–49 years). The Caribbean has a mixture of generalized and concentrated HIV epidemics with the second highest level of adult HIV prevalence (1.0 %) after sub-Saharan Africa [1]. In Latin America, transmission through male-to-male sex accounts for the largest proportion of HIV infection, although there is a sizable proportion of infection among injecting drug users (IDU) and female sex workers (SW). In the Caribbean, heterosexual transmission, often related to sex work, accounts for the largest proportion of HIV infections, although there are indications that transmission is increasing among men who have sex with men (MSM). The most recent evidence suggests that the regional rates of HIV infections in both the Latin America and Caribbean (LAC) regions have stabilized [1, 2].

HIV biological and behavioral surveillance surveys (BBSS) used to measure the prevalence of HIV and other infections, as well as behaviors that place persons at risk, are important for national HIV surveillance programs. Information obtained from such surveys is critical for documenting the need for prevention and treatment interventions, advocating for resource allocation, and monitoring the effectiveness of program implementation [3]. Given the concentrated nature of the epidemic in LAC among certain populations such as MSM, IDU, and SW, it is important that these high-risk groups be included in BBSS.

Sampling frames do not often exist for populations at highest risk for HIV, thus precluding the use of simple random sampling techniques. Venue-based sampling methods that access high-risk populations at clinics or public venues such as bars, brothels, or street locations, may result in samples that are not representative of the target population if attendance at such venues is non-random. In response to these challenges, respondent driven sampling (RDS) [4, 5], a form of chain-referral sampling, has been used to access populations that are considered hard-to-reach through venue-based sampling. RDS makes use of existing social networks in the target population and comprises both peer-referral methods used for field recruitment and analytic methods to weight the sample data to account for the peer-recruitment process [6].

In brief, the use of RDS for field recruitment starts with the selection of an initial number of members of the target population, referred to as ‘seeds.’ Upon completion of an interview and biological testing, seeds are provided a specific number of recruitment coupons for distribution to members of their social network. Eligible persons recruited by the seeds are enrolled in the survey and, in turn, provided the opportunity to distribute coupons for subsequent peer referrals. RDS uses incentives to reward those who participate in the survey and those who refer eligible peers who participate. Each set of recruited persons forms a wave and the sample (chain) expands as additional waves are obtained [7]. Once field recruitment is complete, RDS involves the use of analytic procedures to address the biases in parameter estimation that originate from chain-referral sampling methods. These analytic procedures model the recruitment process based on recruitment patterns and participants’ social network sizes. Data from the sample are weighted to derive parameter estimates that may be considered representative of the target population as long as RDS data collection is implemented correctly with adherence to current assumptions. These include (1) respondents know one another as members of the target population, so ties are reciprocal; (2) respondents are linked by a network composed of a single component; (3) sampling occurs with replacement; (4) respondents accurately report their personal network size, defined as the number of relatives, friends, and acquaintances who fall within the target population; (5) peer recruitment is a random selection from the recruiter’s network; and, (6) each respondent recruits at least one peer [6, 8, 9].

RDS has been widely used for BBSS in international settings [10, 11], with 123 studies conducted in 28 countries by 2008, including 14 studies in LAC. Prior to RDS, countries in LAC primarily used convenience and venue-based sampling to conduct behavioral surveillance [12]. Surveys were done sporadically and were limited in their ability to provide estimates of prevalence of HIV and risk behaviors in populations at highest risk. The introduction of RDS methods has allowed for BBSS to be implemented in a more standardized manner with greater attention to hidden populations [13]. Since the twin reviews of RDS in international settings by Malekinejad et al. [10] and Johnston et al. [11] in 2008, RDS has been adopted or continued for BBSS in several LAC countries, warranting a comprehensive description of how RDS has been implemented in the region. Here we provide a focused review of RDS for BBSS among high-risk populations in LAC and describe the studies’ principal operational, design, and analytical considerations.

Methods

Data Sources

Data sources for this review included published and unpublished reports, protocols, manuscripts, and personal communications with study investigators. Reports and protocols were provided by study investigators who were identified and contacted through the co-authors’ professional networks. Published manuscripts were identified through an iterative search conducted in Ovid Medline using combinations of search terms from the following domains: (1) methodology: “respondent driven sampling” or “chain-referral sampling;” (2) target population: “men who have sex with men,” “transgender” “sex workers,” “drug users,” “transgender,” “youth at risk”, “most-at-risk populations”, “HIV high risk populations.” Additionally, study investigators were contacted to enquire about new publications and other studies of which they may have heard. This process was continued until no new studies were encountered, making us confident that our search is exhaustive and covers all surveys conducted in the region.

Eligibility Criteria

RDS studies were eligible for inclusion if they (1) were conducted in LAC; (2) completed data collection between January 1, 2005 and December 31, 2011; (3) collected biological and/or behavioral data related to HIV risk; and (4) met the minimum requirements for RDS studies based on the criteria established by Malekinejad et al. [10]. Briefly, Malekinejad et al. consider that full RDS studies are those that (1) initiated recruitment chains using seeds; (2) used a recruitment quota; (3) systematically collected data on participants’ social network sizes; (4) systematically recorded who recruited whom; and (5) adjusted prevalence estimates to account for differential social network sizes and recruitment patterns. For studies that had only recently completed data collection, we considered that criterion 5 was met if the investigators planned to conduct weighted data analysis.

Data Abstraction and Analysis

We created a master table in Microsoft Excel and abstracted key operational information for each study using a data abstraction strategy adapted from Malekinejad et al. Operational data were then coded to facilitate the analysis of summary statistics. Target populations were categorized as MSM [which may include male to female transgender (MFT)], drug users [DU, including injecting (IDU) and non-injecting (NIDU)], SW (including male and female), or other. Eligibility criteria were categorized as full if they included all parameters for eligibility (sex, age, geographic area, and a definition of the population’s specific risk behavior) or partial if it failed to include one or more of these parameters. Types of biological tests were categorized as none, HIV only, other infection(s) only, and HIV and other infections. HIV tests were categorized as rapid if results were given to participants during their initial study visit. Type of interview was categorized as face-to-face (conducted with or without a computer or handheld device), computer-assisted self-interview (CASI), or audio computer-assisted self-interview (ACASI). Recruitment sites were categorized as health centers [including public health and sexually transmitted infection (STI) clinics], non-governmental organizations, rented store fronts, or other. Incentives included those used to promote participation and completion of the interview (primary incentives) and those used to promote peer recruitment (secondary incentives). Primary and secondary incentives were categorized as cash, items with minimal monetary value (e.g., beauty products, gift cards, phone cards), or condoms and/or lubricant. We also noted whether other services (e.g., HIV pre- and post-test counseling, referral for HIV treatment, clinical examinations for STIs, condoms and lubricant, and educational materials) were offered to participants. Project duration was defined as the number of weeks of active participant recruitment. The maximum number of waves was defined as the number of waves attained in the study’s longest recruitment chain. Summary statistics were generated using STATA version 12.0 (Stata Corporation, College Station, TX).

Accounts of implementation considerations are based on co-authors first hand experiences, discussions with study investigators, and deliberations from a regional RDS workshop held in Santo Domingo, Dominican Republic in October 2009 [14]. Descriptions of implementation considerations were thematically coded and collapsed into the following categories: operational, design, and analytical considerations.

Results

Study Characteristics

We identified 87 studies in 15 countries in LAC that used RDS for biological and/or behavioral surveys among high-risk groups. Table 1 describes the distribution of studies by region and target population. The majority of RDS studies were conducted in South America (68 %), primarily in Brazil, where 40 studies were conducted between 2005 and 2011. In all regions, the majority of studies were conducted among MSM (43 %), followed by SW (29 %), and DU (26 %). Additionally, there was one study that specifically targeted transgender individuals and another that targeted at-risk youth. Among studies that reported inclusion criteria, 14 % of 37 MSM studies included transgender individuals; 12 % of 25 SW studies included males and females, while the remainder limited inclusion to females; and 74 % of 23 DU studies included NIDU as well as IDU, while the remainder limited inclusion to IDU (Table 2, 3, 4).
Table 1

HIV biological and behavioral studies in Latin America and the Caribbean (LAC) that used respondent driven sampling (RDS), by risk group, 2005–2011

Region

Risk group

Total

MSM

SW

DU

Othera

Caribbean

4

4

4

0

12 (13.8 %)

Mexico and Central America

7

6

3

0

16 (18.4 %)

South America

26

15

16

2

59 (67.8 %)

Total

37 (42.5 %)

25 (28.7 %)

23 (26.4 %)

2 (2.3 %)

87 (100.0 %)

MSM men who have sex with men, SW sex workers, DU drug users

aOther risk groups are male to female transgenders and adolescents living on the street

Table 2

HIV biological and behavioral studies that used respondent driven sampling (RDS), Caribbean, 2005–2011

Country, city, referencea

Years

Population

Eligibility criteria

Biological sample

Formative research

Sites (n)

Interview type

Seeds

Incentives

Sample size

Waves (n)

Duration (weeks)

Total

Failed

Added

Primary

Secondary

Target

Actual

Dominican Republic, Santo Domingo [1]

2008

DU

(1) Male or female; (2) age ≥ 15 years; (3) used drugs in past 6 months; (4) lives in study area

Venous blood for syphilis, HBV, HCV, HIV confirmation. Blood for rapid HIV test

Yes

1

Face-to-face

8

3

0

US $9

US $3

300

309

10

9

Dominican Republic, Santiago [1]

2008

DU

(1) Male or female; (2) age ≥ 15 years; (3) used drugs in past 6 months; (4) lives in study area

Venous blood for syphilis, HBV, HCV, HIV confirmation. Blood for rapid HIV test

Yes

1

Face-to-face

6

0

0

US $9

US $3

300

310

16

16

Dominican Republic, La Altagracia [1]

2008

DU

(1) Male or female; (2) age ≥ 15 years; (3) used drugs in past 6 months; (4) lives in study area

Venous blood for syphilis, HBV, HCV, HIV confirmation. Blood for rapid HIV test

Yes

1

Face-to-face

8

0

0

US $9

US $3

300

285

10

10

Dominican Republic, Barahona [1]

2008

DU

(1) Male or female; (2) age ≥ 15 years; (3) used drugs in past 6 months; (4) lives in study area

Venous blood for syphilis, HBV, HCV, HIV confirmation. Blood for rapid HIV test

Yes

1

Face-to-face

8

1

0

US $9

US $3

300

299

9

9

Dominican Republic, Santo Domingo [1]

2008

MSM

(1) Male; (2) age ≥ 15 years; (3) had oral or anal sex with another man in past 6 months; (4) lives in the study area.

Venous blood for syphilis, HBV, HCV, HIV confirmation. Blood for rapid HIV test

Yes

1

Face-to-face

7

0

0

US $9

US $3

500

510

14

6

Dominican Republic, Santiago [1]

2008

MSM

(1) Male; (2) age ≥ 15 years; (3) had oral or anal sex with another man in past 6 months; (4) lives in the study area.

Venous blood for syphilis, HBV, HCV, HIV confirmation. Blood for rapid HIV test

Yes

1

Face-to-face

10

4

4

US $9

US $3

300

327

13

13

Dominican Republic, La Altagracia [1]

2008

MSM

(1) Male; (2) age ≥ 15 years; (3) had oral or anal sex with another man in past 6 months; (4) lives in study area

Venous blood for syphilis, HBV, HCV, HIV confirmation. Blood for rapid HIV test

Yes

1

Face-to-face

7

0

0

US $9

US $3

300

270

12

9

Dominican Republic, Barahona [1]

2008

MSM

(1) Male; (2) age ≥ 15 years; (3) had oral or anal sex with another man in past 6 months; (4) lives in study area

Venous blood for syphilis, HBV, HCV, HIV confirmation. Blood for rapid HIV test

Yes

1

Face-to-face

9

1

0

US $9

US $3

300

280

11

10

Dominican Republic, Santo Domingo [1]

2008

SW

(1) Female; (2) age ≥ 15 years; (3) exchanged sex for money in past 6 months; (4) lives in study area

Venous blood for syphilis, HBV, HCV, HIV confirmation. Blood for rapid HIV test

Yes

1

Face-to-face

7

1

0

US $9

US $3

400

410

10

14

Dominican Republic, Santiago [1]

2008

SW

(1) Female; (2) age ≥ 15 years; (3) exchanged sex for money in past 6 months; (4) lives in study area

Venous blood for syphilis, HBV, HCV, HIV confirmation. Blood for rapid HIV test

Yes

1

Face-to-face

10

0

0

US $9

US $3

300

301

9

16

Dominican Republic, La Altagracia [1]

2008

SW

(1) Female; (2) age ≥ 15 years; (3) exchanged sex for money in past 6 months; (4) lives in study area

Venous blood for syphilis, HBV, HCV, HIV confirmation. Blood for rapid HIV test

Yes

1

Face-to-face

8

0

0

US $9

US $3

300

297

12

7

Dominican Republic, Barahona [1]

2008

SW

(1) Female; (2) age ≥ 15 years; (3) exchanged sex for money in past 6 months; (4) lives in study area

Venous blood for syphilis, HBV, HCV, HIV confirmation. Blood for rapid HIV test

Yes

1

Face-to-face

9

5

0

US $9

US $3

300

243

13

10

NR Not reported, MSM men who have sex with men, SW sex workers, DU drug users, HBV Hepatitis B virus, HCV Hepatitis C virus, HIV Human immunodeficiency virus

aData source references are found in Table 5

Table 3

HIV biological and behavioral studies that used respondent driven sampling (RDS), Central America and Mexico, 2005–2011

Country, city, referencea

Years

Population

Eligibility criteria

Biological sample

Formative research

Sites (n)

Interview type

Seeds

Incentives

Sample size

Waves (n)

Duration (weeks)

Total

Failed

Added

Primary

Secon-dary

Target

Actual

El Salvador, San Salvador [2]

2008

MSM

(1) Male; (2) age ≥ 18 years; (3) had anal sex with another man during the past 12 months

Blood for rapid HIV test; serum blood for syphilis, HSV-2, HIV ELISA and Western blot; urine for CT, NG, MG, TV; rectal swab for CT and NG

Yes

1

Face-to-face and ACASI

11

Yes

No

Items valued at US$4

Items valued at US$2.70

600

624

18

22

El Salvador, San Miguel [2]

2008

MSM

(1) Male; (2) age ≥ 18 years; (3) had anal sex with another man during the past 12 months

Blood for rapid HIV test; serum blood for syphilis, HSV-2, HIV ELISA and Western blot; urine for CT, NG, MG; rectal swab for CT and NG

Yes

1

Face-to-face and ACASI

5

No

No

Items valued at US$4

Items valued at US$2.70

200

200

10

16

El Salvador, San Salvador [3]

2008

SW

(1) Female; (2) age ≥ 18 years; (3) had vaginal or anal sex in exchange for money during the past 30 days

Blood for rapid HIV test; serum blood for syphilis, HSV-2, HIV ELISA and Western blot; vaginal swab for CT, NG, MG, TV and BV; rectal swab for CT and NG

Yes

1

Face-to-face and ACASI

10

Yes

No

Items valued at US$4

Items valued at US$2.70

600

663

18

22

El Salvador, Sonsonate [3]

2008

SW

(1) Female; (2) age ≥ 18 years; (3) had vaginal or anal sex in exchange for money during the past 30 days

Blood for rapid HIV test; serum blood for syphilis, HSV-2, HIV ELISA and Western blot; vaginal swab for CT, NG, MG, TV and BV; rectal swab for CT and NG

Yes

1

Face-to-face and ACASI

5

No

No

Items valued at US$4

Items valued at US$2.70

200

185

10

10

Guatemala, Guatemala City [4]

2011

DU

(1) Male or Female; (2) age ≥ 18 years; (3) consumed cocaine, crack, heroin, LSD, or ecstasy in past 30 days; (4) lives in study area

Blood for HIV rapid test

Yes

1

Face-to-face

11

2

Yes

US$6

US$6

367

299

11

12

Honduras, San Pedro Sula [5]

2006

MSM

(1) Male; (2) age ≥ 18 years; (3) had anal or oral sex with another man within past 12 months; (4) lives in study area

DBS, venous blood, rectal swab and urine for HIV, Syphilis, CT, TV, GC, HSV-2, MG, LGV

Yes

1

ACASI

7

0

Yes

Condoms, lubricants

Items valued at up to US$2.51

200

204

9

8

Honduras, La Ceiba [5]

2006

MSM

(1) Male; (2) age ≥ 18 years; (3) had anal or oral sex with another man within past 12 months; (4) lives in study area

DBS, venous blood, rectal swab and urine for HIV, Syphilis, CT, GC, HSV-2, MG, LGV

Yes

1

ACASI

9

0

Yes

Condoms, lubricants

Items valued at up to US$2.51

200

197

6

8

Honduras, Tegucigalpa [6]

2006

SW

(1) Female; (2) age ≥ 18 years; (3) had vaginal or anal sex for money and work within past 12 months; (4) lives in study area

DBS, venous blood, vaginal swab for HIV, Syphilis, CT, TV, NG, HSV-2, BV, and MG

Yes

1

ACASI

6

1

Yes

Items valued at US$2.00

Items valued at up to US$2.51

200

205

9

10

Honduras, Comayagua [6]

2006

SW

(1) Female; (2) age ≥ 18 years; (3) had vaginal or anal sex for money and work within past 12 months; (4) lives in study area

DBS, venous blood, vaginal swab for HIV, Syphilis, CT, TV, NG, HSV-2, BV, and MG

Yes

1

ACASI

7

1

Yes

Items valued at US$2.00

Items valued at up to US$2.50

200

212

8

8

Honduras, San Pedro Sula [6]

2006

SW

(1) Female; (2) age ≥ 18 years; (3) had vaginal or anal sex for money and work within past 12 months; (4) lives in study area

DBS, venous blood, vaginal swab for HIV, Syphilis, CT, TV, NG, HSV-2, BV, and MG

Yes

1

ACASI

9

1

Yes

Items valued at US$2.00

Items valued at up to US$2.50

200

210

13

8

Honduras, La Ceiba [6]

2006

SW

(1) Female; (2) age ≥ 18 years; (3) had vaginal or anal sex for money and work within past 12 months; (4) lives in study area

DBS, venous blood, vaginal swab for HIV, Syphilis, CT, TV, NG, HSV-2, BV, and MG

Yes

1

ACASI

6

1

Yes

Items valued at US$2.00

Items valued at up to US$2.50

200

182

6

8

Mexico, Tijuana [7]

2005

DU

(1) Male or female; (2) age ≥ 18 years; (3) injected drugs within past month; (4) has track marks

Venous blood for HIV

NR

3

Face-to-face

15

9

No

US$10

US$5

200

207

8

8

Mexico, Ciudad Juarez [7]

2005

DU

(1) Male or female; (2) age ≥ 18 years; (3) injected drugs within past month; (4) has track marks

Venous blood for HIV

NR

1

Face-to-face

9

8

No

US$20

US$5

200

197

8

2

Nicaragua, Managua [8]

2009-2010

MSM and MTFT

(1) Born male; (2) age ≥ 18 years; (3) had anal sex with another man within past 12 months

Blood for rapid HIV test; serum blood for syphilis, HSV-2, and HIV ELISA and Western Blot HIV test; urine and anal swabs for CT and NG

Yes

NR

ACASI

9

NR

NR

NR

NR

600

644

NR

24

Nicaragua, Chinandega [8]

2009-2010

MSM and MTFT

(1) Born male; (2) age ≥ 18 years; (3) had anal sex with another man within past 12 months

Blood for rapid HIV test; serum blood for syphilis and HIV ELISA and Western Blot

Yes

NR

ACASI

5

NR

NR

NR

NR

300

315

NR

12

Costa Rica, San Jose [9]

2009

MSM

(1) Male; (2) age ≥ 18 years; (3) had anal sex with another man within past three months; (4) resident of study area for at least 6 months; (5) Costa Rican nationality

Venous blood for HIV test and syphilis

Yes

1

Face-to-face and CASI

11

2

No

Condoms, education-al mate-rials

Condoms, education-al mate-rials

300

300

15

NR

NR Not reported, MSM men who have sex with men, SW sex workers, DU drug users, MTFT Male to female transgenders, HSV-2 Herpes simplex virus type 2, HIV Human immunodeficiency virus, CTChlamydia trachomatis, NGNeisseria gonorrhea, MGMycoplasma genitalium, TVTrichomonas vaginalis, HBV Hepatitis B virus, HCV Hepatitis C virus, BV Bacterial vaginosis, DBA Dry blood spot, ELISA Enzyme-linked immunobsorbent assay, CASI Computer-assisted structured interviews, ACASI Audio computer-assisted structured interviews

aData source references are found in Table 5

Table 4

HIV biological and behavioral studies that used respondent driven sampling (RDS), South America, 2005–2011

Country, city, referencea

Years

Population

Eligibility criteria

Biological sample

Formative research

Sites (n)

Interview type

Seeds

 

Incentives

 

Sample size

Waves (n)

Duration (weeks)

Total

Failed

Added

 

Primary

Secondary

 

Target

Actual

Argentina, Buenos Aires [10]

2007–2010

MSM

(1) Self-identifies as male; (2) age ≥ 18 years; (3) had sex with a man/transvestite within past 6 months; (3) at least 10 sexual episodes with man/transvestite in lifetime; (5) lives in study area

Blood for HIV, HCV, HBV, Syphilis; anal swab for HPV; urine for CT

Yes

1

Web-based CASI

16

7

No

 

US$20

US$5

 

500

500

22

36

Bolivia, La Paz [11]

2008

MSM

(1) male; (2) age ≥ 18 years; (3) had anal sex with a man within past 12 months; (4) lives in study area

Blood for rapid HIV test; serum blood for HIV ELISA and Western Blot, HSV-2, and Syphilis

Yes

2

Face-to-face

20

NR

NR

 

Items valued at US$5

Items valued at US$3

 

400

360

11

20

Bolivia, Cocha-bamba [11]

2008

MSM

(1) male; (2) age ≥ 18 years; (3) had anal sex with a man within past 12 months; (4) lives in study area

Blood for rapid HIV test; serum blood for HIV ELISA and Western Blot, HSV-2, and Syphilis

Yes

1

Face-to-face

10

NR

NR

 

Items valued at US$5

Items valued at US$3

 

400

286

15

20

Bolivia, Santa Cruz [11]

2008

MSM

(1) male; (2) age ≥ 18 years; (3) had anal sex with a man within past 12 months; (4) lives in study area

Blood for rapid HIV test; serum blood for HIV ELISA and Western Blot, HSV-2, and Syphilis

Yes

2

Face-to-face

17

NR

NR

 

Items valued at US$5

Items valued at US$3

 

400

373

11

20

Brazil, Curitiba [12]

 

ALS

(1) age 15–19 years; (2) consumed illicit drugs; (3) lived in slum during past 3 months

NR

Yes

1

Face-to-face

NR

NR

NR

 

NR

NR

 

NR

NR

NR

NR

Brazil, Campinas [13]

2006

DU

NR

NR

NR

NR

NR

NR

NR

NR

 

NR

NR

 

330

60

NR

NR

Brazil, Curitiba [13]

2006

DU

NR

NR

NR

NR

NR

NR

NR

NR

 

NR

NR

 

392

295

NR

NR

Brazil, Manaus [13]

2006

DU

NR

NR

NR

NR

NR

NR

NR

NR

 

NR

NR

 

245

499

NR

NR

Brazil, Recife [13]

2006

DU

NR

NR

NR

NR

NR

NR

NR

NR

 

NR

NR

 

291

184

NR

NR

Brazil, Belo Horizonte [14]

2009

DU

(1) age ≥ 18 years; (2) injected drugs ≥ one time within past 6 months; (3) used cocaine, crack, opiates, or illicit drugs (other than marijuana or hashish) by other routes (snorted, smoked, ingested) for ≥25 days within past 6 months; (4) lives in study area

Blood for rapid HIV test and Syphilis

Yes

NR

ACASI

NR

NR

NR

 

NR

NR

 

NR

NR

NR

NR

Brazil, Brazilia [14]

2009

DU

(1) age ≥ 18 years; (2) injected drugs ≥ one time within past 6 months; (3) used cocaine, crack, opiates, or illicit drugs (other than marijuana or hashish) by other routes (snorted, smoked, ingested) for ≥25 days within past 6 months; (4) lives in study area

Blood for rapid HIV test and Syphilis

Yes

NR

ACASI

NR

NR

NR

 

NR

NR

 

NR

NR

NR

NR

Brazil, Campo Grande [14]

2009

DU

(1) age ≥ 18 years; (2) injected drugs ≥ one time within past 6 months; (3) used cocaine, crack, opiates, or illicit drugs (other than marijuana or hashish) by other routes (snorted, smoked, ingested) for ≥ 25days within past 6 months; (4) lives in study area

Blood for rapid HIV test and Syphilis

Yes

NR

ACASI

NR

NR

NR

 

NR

NR

 

NR

NR

NR

NR

Brazil, Curitiba [14]

2009

DU

(1) age ≥ 18 years; (2) injected drugs ≥ one time within past 6 months; (3) used cocaine, crack, opiates, or illicit drugs (other than marijuana or hashish) by other routes (snorted, smoked, ingested) for ≥25 days within past 6 months; (4) lives in study area

Blood for rapid HIV test and Syphilis

Yes

NR

ACASI

NR

NR

NR

 

NR

NR

 

NR

NR

NR

NR

Brazil, Itajai [14]

2009

DU

(1) age ≥ 18 years; (2) injected drugs ≥ one time within past 6 months; (3) used cocaine, crack, opiates, or illicit drugs (other than marijuana or hashish) by other routes (snorted, smoked, ingested) for ≥25 days within past 6 months; (4) lives in study area

Blood for rapid HIV test and Syphilis

Yes

NR

ACASI

NR

NR

NR

 

NR

NR

 

NR

NR

NR

NR

Brazil, Manaus [14]

2009

DU

(1) age ≥ 18 years; (2) injected drugs ≥ one time within past 6 months; (3) used cocaine, crack, opiates, or illicit drugs (other than marijuana or hashish) by other routes (snorted, smoked, ingested) for ≥25 days within past 6 months; (4) lives in study area

Blood for rapid HIV test and Syphilis

Yes

NR

ACASI

NR

NR

NR

 

NR

NR

 

NR

NR

NR

NR

Brazil, Recife [14]

2009

DU

(1) age ≥ 18 years; (2) injected drugs ≥ one time within past 6 months; (3) used cocaine, crack, opiates, or illicit drugs (other than marijuana or hashish) by other routes (snorted, smoked, ingested) for ≥25 days within past 6 months; (4) lives in study area

Blood for rapid HIV test and Syphilis

Yes

NR

ACASI

NR

NR

NR

 

NR

NR

 

NR

NR

NR

NR

Brazil, Rio de Janeiro [14]

2009

DU

(1) age ≥ 18 years; (2) injected drugs ≥ one time within past 6 months; (3) used cocaine, crack, opiates, or illicit drugs (other than marijuana or hashish) by other routes (snorted, smoked, ingested) for ≥25 days within past 6 months; (4) lives in study area

Blood for rapid HIV test and Syphilis

Yes

1

ACASI

6

1

No

 

NR

NR

 

600

605

11

28

Brazil, Salvador [14]

2009

DU

(1) age ≥ 18 years; (2) injected drugs ≥ one time within past 6 months; (3) used cocaine, crack, opiates, or illicit drugs (other than marijuana or hashish) by other routes (snorted, smoked, ingested) for ≥25 days within past 6 months; (4) lives in study area

Blood for rapid HIV test and Syphilis

Yes

NR

ACASI

NR

NR

NR

 

NR

NR

 

NR

NR

NR

NR

Brazil, Santos [14]

2009

DU

(1) age ≥ 18 years; (2) injected drugs ≥ one time within past 6 months; (3) used cocaine, crack, opiates, or illicit drugs (other than marijuana or hashish) by other routes (snorted, smoked, ingested) for ≥ 25days within past 6 months; (4) lives in study area

Blood for rapid HIV test and Syphilis

Yes

NR

ACASI

NR

NR

NR

 

NR

NR

 

NR

NR

NR

NR

Brazil, Campinas [15]

2005

MSM

(1) Born male; (2) age ≥ 15 years; (3) had oral or anal sex with a man/transvestite within past 6 months; (3) age 15+ ; (4) lives in study area; (5) willing to be tested for Syphilis

Venous blood for HIV, Syphilis

Yes

1

ACASI

30

8

Yes

 

Items valued at US$10

Items valued at US$5

 

800

658

23

56

Brazil, Fortaleza [16]

2005

MSM

(1) Male; (2) age ≥ 14 years; (3) had anal or oral sex with man within past 12 months; (4) lives in study area

Venous blood for HIV

Yes

2

Face-to-face

10

0

No

 

US$5

US$5

 

380

406

11

6

Brazil, Belo Horizonte [17]

2008–2009

MSM

(1) Male; (2) age ≥ 18 years; (3) had sex with a man within past 12 months; (4) lives in study area

Blood for rapid HIV test

NR

NR

Hand- held assisted face-to-face

NR

NR

NR

 

US$8

US$5

 

350

NR

NR

NR

Brazil, Campo Grande [17]

2008–2009

MSM

(1) Male; (2) age ≥ 18 years; (3) had sex with a man within past 12 months; (4) lives in study area

Blood for rapid HIV test

NR

NR

Hand- held assisted face-to-face

NR

NR

NR

 

US$8

US$5

 

350

NR

NR

NR

Brazil, Curitiba [17]

2008–2009

MSM

(1) Male; (2) age ≥ 18 years; (3) had sex with a man within past 12 months; (4) lives in study area

Blood for rapid HIV test

NR

NR

Hand- held assisted face-to-face

NR

NR

NR

 

US$8

US$5

 

350

NR

NR

NR

Brazil, Itajai [17]

2008–2009

MSM

(1) Male; (2) age ≥ 18 years; (3) had sex with a man within past 12 months; (4) lives in study area

Blood for rapid HIV test

NR

NR

Hand- held assisted face-to-face

NR

NR

NR

 

US$8

US$5

 

350

NR

NR

NR

Brazil, Manaus [17]

2008–2009

MSM

(1) Male; (2) age ≥ 18 years; (3) had sex with a man within past 12 months; (4) lives in study area

Blood for rapid HIV test

NR

NR

Hand- held assisted face-to-face

NR

NR

NR

 

US$8

US$5

 

350

NR

NR

NR

Brazil, Recife [17]

2008–2009

MSM

(1) Male; (2) age ≥ 18 years; (3) had sex with a man within past 12 months; (4) lives in study area

Blood for rapid HIV test

NR

NR

Hand- held assisted face-to-face

NR

NR

NR

 

US$8

US$5

 

350

NR

NR

NR

Brazil, Rio de Janeiro [17]

2008–2009

MSM

(1) Male; (2) age ≥ 18 years; (3) had sex with a man within past 12 months; (4) lives in study area

Blood for rapid HIV test

NR

NR

Hand- held assisted face-to-face

NR

NR

NR

 

US$8

US$5

 

350

NR

NR

NR

Brazil, Salvador [17]

2008–2009

MSM

(1) Male; (2) age ≥ 18 years; (3) had sex with a man within past 12 months; (4) lives in study area

Blood for rapid HIV test

NR

NR

Hand- held assisted face-to-face

NR

NR

NR

 

US$8

US$5

 

350

NR

NR

NR

Brazil, Santos [17]

2008–2009

MSM

(1) Male; (2) age ≥ 18 years; (3) had sex with a man within past 12 months; (4) lives in study area

Blood for rapid HIV test

NR

NR

Hand- held assisted face-to-face

NR

NR

NR

 

US$8

US$5

 

350

NR

NR

NR

Brazil, São Paulo [18]

2005

SW

(1) Female; (2) age ≥ 18 years; (3) had commercial sex within past month; (4) works in Santos; (5) willing to be tested for HIV

Blood for rapid HIV test

Yes

3

Face-to-face

10

2

Yes

 

Items valued at US$5; transpor-tation

Items valued at US$5

 

200

303

5

20

Brazil, Manaus [19]

2006

SW

(1) Male, female or transgender; (2) age ≥ 18 years; (3) had commercial sex within past month; (4) works in study area

None

NR

NR

NR

NR

NR

NR

 

NR

NR

 

568

210

NR

NR

Brazil, Porto Alegre [20]

2006

SW

(1) Male, female or transgender; (2) age ≥ 18 years; (3) had commercial sex within past month; (4) works in study area

None

Yes

NR

Face-to-face

13

1

Yes

 

US$5; transpor-tation

US$5; transpor-tation

 

381

190

6

20

Brazil, Santos [21]

2006

SW

(1) Male, female or transgender; (2) age ≥ 18 years; (3) had commercial sex within past month; (4) works in study area

None

NR

NR

NR

NR

NR

NR

 

NR

NR

 

200

175

NR

NR

Brazil, Belo Horizonte [22]

2008–2009

SW

(1) Female; (2) age ≥ 18 years; (3) had sexual intercourse for money within past 4 months; (4) work in study area

Blood for rapid HIV test and Syphilis

Yes

NR

Face-to-face and ACASI

NR

NR

NR

 

Beauty products, lunch, transpor-tation

US$6

 

300

289

NR

NR

Brazil, Brazilia [22]

2008–2009

SW

(1) Female; (2) age ≥ 18 years; (3) had sexual intercourse for money within past 4 months; (4) work in study area

Blood for rapid HIV test and Syphilis

Yes

NR

Face-to-face and ACASI

NR

NR

NR

 

Beauty products, lunch, transpor-tation

US$6

 

300

308

NR

NR

Brazil, Campo Grande [22]

2008–2009

SW

(1) Female; (2) age ≥ 18 years; (3) had sexual intercourse for money within past 4 months; (4) work in study area

Blood for rapid HIV test and Syphilis

Yes

NR

Face-to-face and ACASI

NR

NR

NR

 

Beauty products, lunch, transpor-tation

US$6

 

150

147

NR

NR

Brazil, Curitiba [22]

2008–2009

SW

(1) Female; (2) age ≥ 18 years; (3) had sexual intercourse for money within past 4 months; (4) work in study area

Blood for rapid HIV test and Syphilis

Yes

NR

Face-to-face and ACASI

NR

NR

NR

 

Beauty products, lunch, transpor-tation

US$6

 

200

201

NR

NR

Brazil, Itajai [22]

2008–2009

SW

(1) Female; (2) age ≥ 18 years; (3) had sexual intercourse for money within past 4 months; (4) work in study area

Blood for rapid HIV test and Syphilis

Yes

NR

Face-to-face and ACASI

NR

NR

NR

 

Beauty products, lunch, transpor-tation

US$6

 

100

90

NR

NR

Brazil, Manaus [22]

2008–2009

SW

(1) Female; (2) age ≥ 18 years; (3) had sexual intercourse for money within past 4 months; (4) work in study area

Blood for rapid HIV test and Syphilis

Yes

NR

Face-to-face and ACASI

NR

NR

NR

 

Beauty products, lunch, transpor-tation

US$6

 

200

199

NR

NR

Brazil, Recife [22]

2008–2009

SW

(1) Female; (2) age ≥ 18 years; (3) had sexual intercourse for money within past 4 months; (4) work in study area

Blood for rapid HIV test and Syphilis

Yes

NR

Face-to-face and ACASI

NR

NR

NR

 

Beauty products, lunch, transpor-tation

US$6

 

200

237

NR

NR

Brazil, Rio de Janeiro [22]

2008–2009

SW

(1) Female; (2) age ≥ 18 years; (3) had sexual intercourse for money within past 4 months; (4) work in study area

Blood for rapid HIV test and Syphilis

Yes

NR

Face-to-face and ACASI

NR

NR

NR

 

Beauty products, lunch, transpor-tation

US$6

 

600

601

NR

NR

Brazil, Salvador [22]

2008–2009

SW

(1) Female; (2) age ≥ 18 years; (3) had sexual intercourse for money within past 4 months; (4) work in study area

Blood for rapid HIV test and Syphilis

Yes

NR

Face-to-face and ACASI

NR

NR

NR

 

Beauty products, lunch, transpor-tation

US$6

 

300

260

NR

NR

Brazil, Santos [22]

2008–2009

SW

(1) Female; (2) age ≥ 18 years; (3) had sexual intercourse for money within past 4 months; (4) work in study area

Blood for rapid HIV test and Syphilis

Yes

NR

Face-to-face and ACASI

NR

NR

NR

 

Beauty products, lunch, transpor-tation

US$6

 

150

191

NR

NR

Chile, Santiago [23]

2009

MSM

(1) Male; (2) ages 18-45 years; (3) had insertive anal sex with a man within past 12 months; (4) resides or works in study area; (5) presents photo identification

Venous blood for HIV

NR

3

Face-to-face

7

0

No

 

US$9.70

US$5.80

 

300

317

10

15

Chile, Valparaiso [23]

2009

MSM

(1) Male; (2) ages 18–45 years; (3) had insertive anal sex with a man within past 12 months; (4) lives or works in study area; (5) presents photo identification

Venous blood for HIV

NR

1

Face-to-face

3

1

No

 

US$9.70

US$5.80

 

100

154

8

15

Colombia, Barranquilla [24]

2010

MSM

(1) Male; (2) age ≥ 18 years; (3) had oral/anal/hand sex with man within past 12 months; (4) lives in study area; (5) presents valid identification

Blood for rapid HIV test

Yes

NR

Computer assisted face-to-face

NR

NR

NR

 

Giftcard US$10.5

Condoms, lubricants

 

350

NR

NR

NR

Colombia, Bogotá [24]

2010

MSM

(1) Male; (2) age ≥ 18 years; (3) had oral/anal/hand sex with man within past 12 months; (4) lives in study area; (5) presents valid identification

Blood for rapid HIV test

Yes

NR

Computer assisted face-to-face

NR

NR

NR

 

Giftcard US$10.5

Condoms, lubricants

 

504

NR

NR

NR

Colombia, Cali [24]

2010

MSM

(1) Male; (2) age ≥ 18 years; (3) had oral/anal/hand sex with man within past 12 months; (4) lives in study area; (5) presents valid identification

Blood for rapid HIV test

Yes

NR

Computer assisted face-to-face

NR

NR

NR

 

Giftcard US$10.5

Condoms, lubricants

 

350

NR

NR

NR

Colombia, Cartagena [24]

2010

MSM

(1) Male; (2) age ≥ 18 years; (3) had oral/anal/hand sex with man within past 12 months; (4) lives in study area; (5) presents valid identification

Blood for rapid HIV test

Yes

NR

Computer assisted face-to-face

NR

NR

NR

 

Giftcard US$10.5

Condoms, lubricants

 

350

NR

NR

NR

Colombia, Cúcuta [24]

2010

MSM

(1) Male; (2) age ≥ 18 years; (3) had oral/anal/hand sex with man within past 12 months; (4) lives in study area; (5) presents valid identification

Blood for rapid HIV test

Yes

NR

Computer assisted face-to-face

NR

NR

NR

 

Giftcard US$10.5

Condoms, lubricants

 

350

NR

NR

NR

Colombia, Medellin [24]

2010

MSM

(1) Male; (2) age ≥ 18 years; (3) had oral/anal/hand sex with man within past 12 months; (4) lives in study area; (5) presents valid identification

Blood for rapid HIV test

Yes

NR

Computer assisted face-to-face

NR

NR

NR

 

Giftcard US$10.5

Condoms, lubricants

 

350

NR

NR

NR

Colombia, Pereira [24]

2010

MSM

(1) Male; (2) age ≥ 18 years; (3) had oral/anal/hand sex with man within past 12 months; (4) lives in study area; (5) presents valid identification

Blood for rapid HIV test

Yes

NR

Computer assisted face-to-face

NR

NR

NR

 

Giftcard US$10.5

Condoms, lubricants

 

350

NR

NR

NR

Colombia, Medellin [25]

2010

IDU

(1) age 18–65 years; (2) injected drugs ≥ one time within past 6 months; (4) consumed drugs in study area

Venous blood for HIV

Yes

NR

Face-to-face

3

0

0

 

US$5

US$4

 

240

237

8

8

Colombia, Pereira [25]

2010

IDU

(1) age 18-65 years; (2) injected drugs ≥ one time within past 6 months; (4) consumed drugs in study area

Venous blood for HIV

Yes

NR

Face-to-face

3

0

0

 

US$5

US$4

 

240

297

6

8

Ecuador, Quito [26]

2011

MSM

(1) Born male; (2) age ≥ 15 years; (3) had anal sex with a man within past 12 months; (4) lives or works in study area

Venous blood for HIV, Syphilis, HSV-2, and HBV; urine and anal & oral swaps for future CT and GC tests

Yes

1

CASI

6

2

2

 

US$5

Items valued at US$5

 

400

423

14

22

Paraguay, Ciudad del Este [27]

2006

MSM

(1) Male; (2) age ≥ 16 years; (3) had sex with man at least once within past 12 months; (4) lives in study area; (5) able to understand/speak Spanish; (6) willing to be tested for syphilis

Venous blood for HIV and Syphilis

Yes

1

Hand- held assisted face-to-face

6

1

Yes

 

US$6.50; condoms, lubricant, eduacatio-nal materials

US$3.50,

condoms

 

403

296

22

24

Paraguay, Ciudad del Este [27]

2006

SW

(1) Female; (2) age ≥ 16 years; (3) exchanged sex for money, drugs, or gifts within past 12 months; (4) lives in study area; (5) able to understand/speak Spanish; (6) willing to be tested for syphilis

Venous blood for HIV and Syphilis

Yes

1

Hand- held assisted face-to-face

9

3

Yes

 

US$6.50; condoms, lubricant, eduacatio-nal materials

US$3.50, condoms

 

425

160

13

24

Peru, Lima [28]

2009

MTFT

(1) Born male; (2) age ≥ 18 years; (3) self-identifies as trans-woman

Venous blood for HIV, Syphilis, and HSV-2

Yes

6

Face-to-face

11

5

Yes

 

US$7, condoms, lubricant

Condoms, lubricants

 

420

450

12

12

NR Not reported, MSM men who have sex with men, SW sex workers, DU drug users, ALS Adolescents living on the streets, MTFT Male-to-female transgenders, HSV-2 Herpes simplex virus type 2, HIV Human immunodeficiency virus, CTChlamydia trachomatis, NGNeisseria gonorrhea, MGMycoplasma genitalium, TVTrichomonas vaginalis, HBV Hepatitis B virus, HCV Hepatitis C virus, BV Bacterial vaginosis, DBA Dry blood spot, ELISA Enzyme-linked immunobsorbent assay, CASI Computer-assisted structured interviews, ACASI Audio computer-assisted structured interviews

aData source references are found in Table 5

Operational Considerations

Formative Research

Formative research is useful for evaluating the social networks of target populations, assessing the feasibility of RDS, and planning study logistics, including selecting incentives, seeds, and interview locations [15]. All studies with available data (n = 58) reported that they conducted formative research, although the content and approaches to formative research varied across studies.

Interview Administration

Mode of interview administration is an important consideration as it may influence participants’ willingness to provide truthful answers to questions regarding private and often stigmatized behaviors. Face-to-face interviews were the most commonly used method to collect behavioral data (74 % of the 81 studies with available data), followed by ACASI (23 %) and CASI (3 %). Seventy-one percent of those that used ACASI or CASI employed an initial face-to-face interview to determine participants’ eligibility and/or the size of their social network.

Specimen Collection and Biologic Testing

Among the 81 studies with available data, all but three collected both HIV biological and behavioral data. In 59 % of these studies, participants were tested using blood- or saliva-based rapid HIV tests. Fifty-nine percent of these studies additionally tested for other infections (most commonly syphilis), and in 38 % of studies, participants were tested for three or more infections. For example, in Honduras, female SW were tested for numerous infections, including Chlamydia trachomatis (CT), Neisseria gonorrhea (NG), Mycoplasma genitalium (MG), Trichomonas vaginalis (TV), and bacterial vaginosis (BV) [16]. More than 90 % of female SW accepted the collection of specimens for these tests using vaginal swabs. In El Salvador, MSM were tested for HIV, syphilis, herpes simplex virus type 2 (HSV-2), CT, NG, and MG. Sixty-four percent of MSM accepted the collection of specimens for CT and NG testing using rectal swabs (personal correspondence, ME Guardado). In all studies, results of biologic tests along with linkages to follow-up care were offered to participants.

Defining Eligibility

Almost all studies with available information (n = 83) had clearly defined criteria related to sex, age, geography, and the population’s membership-defining risk behavior, including the specific type of behavior and the period of time during which this behavior was practiced. Clear definition of eligibility criteria is especially important with RDS because they are used to design the social network size question needed to weight RDS data. We found that 11 % of studies did not define geographic parameters [“the population lives and/or works in a specified area (study catchment area)”], which may result in recruitment outside of the area under investigation [11].

Measuring Social Network Size

In a small number of studies, analytical problems arose due to missing and zero social network size values in the dataset. This should not occur given that participants must at least have their recruiter in their social network. Reasons for missing and zero social network size values included the lack of proper interviewer training and the use of self-administered questionnaires to ascertain social network size. To avoid social network size errors in self-administered interviews, RDS studies in El Salvador (2008) and Brazil (2008–2009) employed a brief face-to-face interview to measure participants’ social network sizes prior to participants’ use of ACASI.

To improve the validity of the social network size question, many studies broke the question down into several smaller questions. For example, in the MSM studies in Dominican Republic (2008), the first social network size question was: “How many men do you know (and they know you) who have had sex with another man in the past 6 months?”; the second question was: “How many of them (repeating the response in the previous question) live in this province?”; the third question was: “How many of them (repeating the response in the previous question) are 15 years of age or older?”; and the forth question was: “How many of them (repeating the response in the previous question) have you seen in the last week?” The fourth response was the social network size measure. While this type of question format usually works well, in one study in Central America problems arose because staff did not understand that the second question was a subset of the first, and the third a subset of the second, and so on.

Using Appropriate Incentives

Details on the types of primary and secondary incentives were reported by 68 studies. Cash or cash equivalents were the most common primary and secondary incentives (59 and 62 %, respectively), followed by items with a minimal monetary value (37 and 25 %, respectively), and condoms and/or lubricant (4 and 13 %, respectively). In addition to the primary and secondary incentives, 84 % of studies with available data reported offering participants at least one of the following additional services: HIV pre- and post-test counseling, referral for HIV treatment, condoms and lubricants, and educational materials. In ten studies, participants received a clinical STI examination.

In some studies there was evidence that the incentives were too low to promote participation and/or recruitment. For example, in Porto Alegre, Brazil (2006), researchers considered that the US$5 cash primary and secondary incentives may have been too low to promote participation among SW, and may have contributed to the study’s inability to reach its target sample size (personal communication, C Germany). A similar situation occurred in São Paulo, Brazil (2005), where food stamps worth approximately US$5 were used as primary and secondary incentives. However, in this case, researchers felt that the HIV test and pre- and post-test counseling proved to be a powerful additional incentive and the study was nonetheless able to reach its target sample size (N Gravato da Silva, personal communication).

Conversely, in some studies, there was evidence that the monetary incentives may have been too high. For example, among DU in Barahona, Dominican Republic (2008), where the primary and secondary incentives were US$9 and US$3, respectively, researchers encountered problems with stealing and selling of coupons. However, in response to a question about reasons why they accepted a coupon and enrolled in the survey, 65 % of DU in Barahona reported doing so for the HIV test and 18 % reported doing so because they were interested in the goals of the study (second author, personal communication). Incentives that are too high may also lead to logistical challenges of managing a high volume of coupon holders eager to participate in the survey. Finally, high incentives may lead to instances of masking, whereby individuals misrepresent themselves as members of the target population in order to participate in the survey and obtain the incentives [4]. Such masking occurred among SW in Manaus, Brazil (2008–2009) and among MSM in Buenos Aires, Argentina (2007–2010), where researchers suspected that participants’ desire for the incentives may have led some people to misrepresent themselves as being part of the target population [17].

Using an Appropriate Study Site

Designating a study site is an important consideration because participants must present at the study site to be interviewed and receive their secondary incentives [4]. It is imperative that the study site be easily accessible to members of the target population, ideally within close proximity to where they live or work, and/or accessible using public transportation. The neighborhood should also be reasonably safe for both study staff and participants. In regard to the number of study sites, having multiple sites may be advantageous if the sampling area is large or public transportation is costly or time-consuming for participants. Eighty-three percent of the 40 studies that reported the number of data collection sites used a single site. Sites used for data collection included rented store fronts (41 %), health centers (35 %), non-governmental organizations (NGO, 20 %), and other spaces such as churches and bakeries (4 %). Only four studies (San Pedro Sula, Honduras, 2006; La Ceiba, Honduras, 2006; Tijuana, Mexico, 2005; and Porto Alegre, Brazil, 2006) used a mobile unit.

Several studies in LAC reported problems related to finding a proper study site that would appeal to different social strata within the target population. For example, in Fortaleza, Brazil (2005), researchers found that upper class MSM were reluctant to visit the study site at an NGO, resulting in low numbers of upper-class participants. In Guatemala City, Guatemala (2011), the study site’s location at an existing NGO was satisfactory for attracting low income DU, but was a barrier for attracting middle- and upper-class DU and may have led to underrepresentation of the upper socio-economic strata of the target population (second author, personal communication). Similar problems may arise when studies are conducted in existing NGOs that already service the target population. This occurred in Valparaiso, Chile (2009) when the network of MSM associated with the partner NGO was so well connected that MSM not associated with the NGO did not enter the sample until the final waves of recruitment (second author, personal communication).

Attaining the Target Sample Size

There were 74 studies that reported a target sample size and 58 that reported a recruited sample size. The median target and recruited sample sizes were 300 (range 100–800) and 292 (range 60–663), respectively. On average, studies that reported both data points (n = 54) attained 99 % of their target sample size, with 50 % meeting or exceeding their target. Only 15 % of studies were unable to attain at least 90 % of their target sample size. Studies with available operational data (n = 42) attained a median of 11 maximum waves (range 5–23) over a median duration of 12 weeks (range 2–56). Longer recruitment chains reduce seed dependence in the sample, while the ability to sample within a minimum number of weeks decreases potential biases due to in- and out-migration and/or changes in the dynamic of social networks.

Design Considerations

Social Networks

As a method that relies on social networks, RDS will not function if the size and structure of the target population’s social networks are not conducive to peer-to-peer recruitment. Ideally, formative assessment activities, conducted prior to RDS implementation, should explore how the target population’s social network properties may enable or impede RDS recruitment, and consequently determine whether RDS is an appropriate sampling strategy [15]. While all studies with available information reported conducting a formative assessment prior to implementing the study, there were numerous studies that reported challenges related to the implementation of RDS given the target populations’ social network characteristics.

One of the social network characteristics that may influence the performance of RDS is the size of the population. In Sonsonate, El Salvador (2008), researchers learned that the female SW population was insufficiently large to sustain recruitment. As the sample grew closer to the target sample size, it became increasingly difficult to obtain participants who had not already participated in the survey. A similar situation occurred in Campinas, Brazil (2006), where recruitment was limited to IDU residing in two neighborhoods of the city. There was evidence of network saturation given that the same participants were repetitively recruited for the study. Because RDS assumes that sampling is with replacement, network saturation may indicate a violation of a key assumption (18). The size of individual social networks is also important for RDS recruitment. As has been previously reported [11], RDS could not be sustained among SW in Ciudad del Este, Paraguay (2006) due to their small and closed social networks that were restricted to those working at the same venue.

Social mixing is another social network characteristic that influences the performance of RDS, as well as the validity and precision of its estimates. RDS assumes that all members of the target population are connected as one social network component. However, there may be barriers to social mixing (e.g., cultural, socioeconomic, and geographic differences) that may prevent members of the target population from socializing and/or recruiting as one network component. Bottlenecks arise when there are barriers that impede recruitment from readily flowing between subgroups, leading recruitment to become “trapped” within a particular subgroup. For example, among DU in Rio de Janeiro, Brazil (2009), geographic bottlenecking prevented DU in several key geographic areas from being included in the sample [19]. Evidence of bottlenecking was also found in Valparaiso, Chile (2009), where recruitment was largely sustained among gay-identifying MSM, leading to a sample that may have underrepresented bisexual and heterosexual-identifying MSM [20].

Isolated subpopulations arise when the barriers to social mixing across subgroups are so great as to obstruct cross-over between their social networks. For example, in San Salvador, El Salvador (2008), brothel-based SW with clients of lower socio-economic status did not socially interact with night-club based SW with upper-class clients. Given the lack of cross-over between the two subpopulations, it became necessary to analyze them as two separate samples. A similar situation arose in Porto Alegre, Brazil (2005), where there was lack of cross-over between subpopulations of male and female SW.

Analytical Considerations

Design Effect

Respondent driven sampling’s reliance on social networks means that the observations are not independent, as they would be with a simple random sample (SRS). It is therefore necessary to employ a design effect to account for the greater variance of RDS estimates versus those obtained through SRS. Among the 54 studies with available data, 94 % employed a design effect to calculate their target sample size. The median design effect used was 2, with a range of 1.2–5.0.

Reaching Equilibrium

According to RDS theory, as the sample grows through a successive number of waves, it achieves a state of equilibrium in which the sample composition ceases to change despite additional waves of respondents (indicating that it is independent of the initial non-randomly selected seeds) [4]. Thirty-seven studies reported whether they reached equilibrium on at least one key variable of interest; of these, 95 % attained equilibrium.

Data Analysis

Unless RDS data are analyzed to adjust for differential recruitment and network sizes, the sample is biased and should be considered a convenience sample. Sixty studies reported whether they used statistical weights to analyze their data. Of these, all but one performed some type of statistical adjustment.

Discussion

We identified 87 studies conducted in LAC between 2005 and 2011 that used RDS to access members of high-risk populations for HIV BBSS. Together these studies surveyed over 17,200 MSM, SW, DU, and other high-risk persons in LAC. Additionally, they offered thousands of high-risk persons the opportunity to receive tests for HIV and other infections, as well as pre- and post-test counseling, which are services that are especially important in countries with low coverage.

The reviewed studies encountered operational challenges similar to those described previously [11], albeit often to a lesser extent. For example, in 2008, Johnston et al. reported that 51 % of studies had insufficiently defined eligibility criteria to adequately measure social network size. Among the studies in this review, we found that only 11 % had insufficiently defined eligibility criteria, most commonly missing a geographic parameter. Many studies in LAC improved the accuracy of the social network size measure by breaking down the question into smaller components. Furthermore, most studies ascertained participants’ social network size using a face-to-face interview, regardless of whether the remainder of the interview was administered as face-to-face, CASI, or ACASI. Such techniques improve the accuracy of the social network size measure if interviewers are properly trained to understand the importance and meaning of the social network question, to probe for accurate answers, and to not allow zero as a response. Perhaps as a result of these ascertainment techniques, few studies in LAC failed to systematically record participants’ social network size.

Similar to Johnston et al.’s review, we found some challenges in regard to designating appropriate primary and secondary incentives. However, with the exception of one, most LAC studies did not report increasing or decreasing their primary and secondary incentives during the data collection process. This may be an indicator of the use of formative research (conducted among all studies with available data) to define key operational issues, such as incentives.

An operational consideration described here that has not been described previously is in regard to the collection of biological samples to estimate the prevalence of HIV and other infections. All but three of the studies with available information collected biological specimens. HIV tests were commonly performed with relatively non-invasive sampling techniques, such as saliva or finger blood pricks, which additionally offer the benefit of a rapid turn-around time in providing results to participants. In general, LAC studies reported few challenges in regard to the administration of HIV tests. However, two studies in El Salvador (2008) that tested for other infections using more invasive techniques, namely rectal swabs for NG and CT, encountered some degree of reluctance among participants, decreasing the effective sample size available to estimate the prevalence of such infections. On the other hand, the use of such specimen collection techniques has been well accepted in other countries, including among MSM in Honduras (second author, personal communication). Thus, it is recommended that formative research activities include an inquiry into the target populations’ level of comfort with undergoing such specimen collection procedures.

Another operational consideration relates to the number and type of data collection sites. While the majority of studies with available information used one data collection site, some used two or more. Multiple sites may be useful in that they facilitate access among members of the target population [20]. However, one caveat is that there should be cross-over in recruitment from one site to the other to indicate that the target population is connected as a single network component [6]. This requires active monitoring on behalf of the research team to ensure that participants recruit across sites and that those who present at each site are not different in regard to important characteristics, such as HIV status or risk behaviors [11]. In regard to the type of site, some studies that collected data at an existing NGO encountered challenges in extending recruitment outside of the core group of clients served by the NGO. While it is not always the case that using an existing NGO will limit the attainment of a diverse sample, it is important that researchers anticipate the potential biases and actively monitor recruitment. Other site characteristics may likewise influence recruitment. For example, in Guatemala City, Guatemala (2011), researchers found that higher income DU were unwilling to attend a site in a low-income area, resulting in underrepresentation of high-income DU.

With the exception of possibly SW in Porto Alegre, Brazil (2006) and Ciudad del Este, Paraguay (2006), the target populations in most countries were sufficiently networked to sustain RDS recruitment. However, in some situations important bottlenecks or isolated subpopulations were not identified, resulting in the under-sampling of certain segments of the population. Nonetheless, many studies have been able to better understand the isolation of subpopulations and can take this knowledge into account in future studies. Potential bottlenecks can be mitigated by ensuring that subpopulations are well represented in the selection of seeds. In the case that subpopulations are completely separate (e.g., transgenders in MSM studies), eligibility criteria may be altered to exclude a segment of the population (e.g., exclude transgenders from an MSM study). An independent study that samples the separate segment of the population may also be conducted, as was done in Lima, Peru (2009) with the male to female transgender (MFT) population.

Most studies used design effects of 2 when calculating their sample sizes, which has been the standard recommended design effect for RDS over several years [21]. However, recent data suggest that, although a design effect of 2 may be adequate, it is best to use a design effect closer to 3 or, in some cases, higher [22]. For instance in the study of 450 MTF in Lima, Peru (2009), the design effects of six variables ranged from 2.4 to 5.9, with a mean of 3.7 (LG Johnston, unpublished data, 2012). In studies of MSM, SW, and DU in the Dominican Republic (2008) with approximate sample sizes of 300, the design effects for the HIV prevalence and condom use variables ranged from 1.0 to 5.1, with a mean of 2.3 (second author, personal communication).

Lastly, RDS requires a complex analysis procedure to produce population estimates and confidence intervals. While the statistical methods for conducting appropriate analyses of RDS data are still being refined, it is important that researchers address the biases in parameter estimation that originate from chain-referral sampling. All but one of the studies that reported information about analysis indicated that adjustment was included in final analyses, indicating that researchers in the region are well aware of the special analysis requirements of RDS.

In conclusion, RDS has been widely used for HIV BBSS among hard-to-reach high-risk populations in LAC. Most of the studies reported here are the first in their respective countries to collect data on these populations and the first attempt to use a probability-based sampling method such as RDS. These baseline studies allow researchers and public health practitioners to access hidden populations at high risk for HIV and to collect valuable data on the prevalence of HIV and other infections, as well as related risk behaviors. Unfortunately, the findings from many studies using RDS are never published and those that are published are not consistent in their presentation of the study design. Recommendations for the consistent reporting of RDS findings and study design elements have recently been published and are currently being updated [23]. Our review suggests that future RDS studies may benefit from taking into account recommendations to conduct formative research to explore social network properties and define key operational issues, incorporate strategies to better ascertain social network size, actively monitor recruitment across subgroups, consider bottlenecks and isolated subpopulations, and analyze data to address the biases of chain-referral sampling. We take the opportunity to remind readers that it is best practice to use the same sampling method in the same populations over time in order to monitor HIV trends, evaluate the effectiveness of prevention programs, and better allocate public health resources. Additionally, as the methodology becomes more widely used throughout the world, similar focused reviews are needed to describe the region-specific implementation of RDS for BBSS.

Notes

Acknowledgments

We would like to express our gratitude to the organizations and investigators who generously shared their operational information with us, especially Sonia Morales-Miranda, Maria Elena Guardado, and Jerry Jacobson. We also would like to thank Abraham Miranda, Amy Drake, and Gabriela Paz-Bailey for their assistance in the conception and development of this manuscript. JRM is currently supported by a UTHealth Innovation for Cancer Prevention Research Postdoctoral Fellowship (The University of Texas School of Public Health—Cancer Prevention and Research Institute of Texas Grant # RP101503). The findings and conclusions of this paper are those of the authors and do not necessarily represent the official position of the U.S. Centers for Disease Control and Prevention or the Cancer Prevention and Research Institute of Texas.

References

  1. 1.
    UNAIDS. AIDS epidemic update. 2010. http://www.unaids.org/globalreport/Global_report.htm. Accessed 1 June 2012.
  2. 2.
    US Agency for International Development (USAID). HIV/AIDS Health Profile-Latin America and the Caribbean. 2011. www.usaid.gov/our_work/global_health/.../hiv_summary_lac.pdf. Accessed 22 Mar 2012.
  3. 3.
    Family Health International. Behavioral surveillance surveys (BSS): Guidelines for repeated behavioral surveys in populations at risk for HIV, Arlington. 2000.Google Scholar
  4. 4.
    Heckathorn DD. Respondent-driven sampling: a new approach to the study of hidden populations. Soc Probl. 1997;44(2):174–99.CrossRefGoogle Scholar
  5. 5.
    Heckathorn DD, Semaan S, Broadhead RS, Hughes JJ. Extensions of respondent-driven sampling: a new approach to the study of injection drug users aged 18–25. AIDS Behav. 2002;6(1):55–67.CrossRefGoogle Scholar
  6. 6.
    Salganik MJ, Heckathorn DD. Sampling and estimation of hidden populations using respondent driven sampling. Sociol Methodol. 2004;34(1):193–240.CrossRefGoogle Scholar
  7. 7.
    Johnston LG, Sabin K, Prybylski D. Update for sampling most-at-risk and hidden populations for HIV biological and behavioral surveillance. J HIV/AIDS Surveill Epidemiol. 2010; 2(1).Google Scholar
  8. 8.
    Heckathorn DD. Respondent-driven sampling II: deriving valid population estimates from chain-referral samples of hidden populations. Soc Probl. 2002;49(1):11–34.CrossRefGoogle Scholar
  9. 9.
    Heckathorn DD. Extensions of respondent-driven sampling: analyzing continuous variables and controlling for differential recruitment. Sociol Methodol. 2007;37(1):151–207.CrossRefGoogle Scholar
  10. 10.
    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(Suppl 4):S105–30.PubMedCrossRefGoogle Scholar
  11. 11.
    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(Suppl 4):S131–41.PubMedCrossRefGoogle Scholar
  12. 12.
    Soto RJ, Ghee AE, Nunez CA, et al. Sentinel surveillance of sexually transmitted infections/HIV and risk behaviors in vulnerable populations in 5 Central American countries. J Acquir Immune Defic Syndr. 2007;46(1):101–11.PubMedGoogle Scholar
  13. 13.
    Magis-Rodriguez C, Brouwer KC, Morales S, et al. HIV prevalence and correlates of receptive needle sharing among injection drug users in the Mexican-U.S. border city of Tijuana. J Psychoact Drugs. 2005;37(3):333–9.CrossRefGoogle Scholar
  14. 14.
    Montealegre JR, Johnston LG, Drake A, et al. Overcoming challenges in the implementation and analysis of respondent driven sampling: deliberations from the respondent driven sampling practical implementation. In Workshop for the Dominican Republic and Latin America, Santo Domingo. 2011.Google Scholar
  15. 15.
    Johnston LG, Whitehead S, Simic-Lawson M, Kendall C. Formative research to optimize respondent-driven sampling surveys among hard-to-reach populations in HIV behavioral and biological surveillance: lessons learned from four case studies. AIDS Care. 2010;22(6):784–92.PubMedCrossRefGoogle Scholar
  16. 16.
    Johnston LG, Paz Bailey G, Morales-Miranda S, Morgan M, Alvarez B, Monterroso E. High prevalence of M. genitalium among female sex workers in Honduras: implications for the spread of HIV and other sexually transmitted infections. Int J STD AIDS. 2012;23(1):5–11.PubMedCrossRefGoogle Scholar
  17. 17.
    Carballo-Dieguez A, Balan I, Marone R, et al. Use of respondent driven sampling (RDS) generates a very diverse sample of men who have sex with men (MSM) in Buenos Aires, Argentina. PLoS ONE. 2011;6(11):e277447.CrossRefGoogle Scholar
  18. 18.
    Gile KJ, Handcock MS. Respondent-driven sampling: an assessment of current methodology. Sociol Methodol. 2010;40(1):285–327.PubMedCrossRefGoogle Scholar
  19. 19.
    Toledo L, Codeco CT, Bertoni N, et al. Putting respondent-driven sampling on the map: insights from Rio de Janeiro, Brazil. J Acquir Immune Defic Syndr. 2011;57(Suppl 3):S136–43.PubMedCrossRefGoogle Scholar
  20. 20.
    Johnston LG. Conducting respondent driven sampling (RDS) studies in diverse settings: a training manual for planning RDS studies. Centers for Disease Control and Prevention, Atlanta and Family Health International, Arlington. 2007. http://globalhealthsciences.ucsf.edu/prevention-public-health-group/training-resources/hivaids-epidemiologic-surveillance-trainings. Accessed 1 June 2012.
  21. 21.
    Salganik MJ. Variance estimation, design effects, and sample size calculations for respondent-driven sampling. J Urban Health. 2006;83(Suppl 6):i98–112.PubMedCrossRefGoogle Scholar
  22. 22.
    Wejnert C, Pham H, Krishna N, Le B, DiNenno E. Estimating design effect and calculating sample size for respondent-driven sampling studies of injection drug users in the United States. AIDS Behav. 2012;16(4):797–806.PubMedCrossRefGoogle Scholar
  23. 23.
    White RG, Lansky A, Goel S, et al. Respondent driven sampling—where are we and where should we be going? Sex Transm Infect. 2012;88:397–9.PubMedCrossRefGoogle Scholar
  24. 24.
    Kim EJ, Creswell J, Guardado ME, et al. Correlates of bisexual behaviors among men who have sex with men in El Salvador. AIDS Behav. 2012. doi:10.1007/s10461-012-0152-y.
  25. 25.
    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. 2011;87(4):279–82.PubMedCrossRefGoogle Scholar
  26. 26.
    Frost SD, Brouwer KC, Firestone Cruz MA, et al. Respondent-driven sampling of injection drug users in two U.S.-Mexico border cities: recruitment dynamics and impact on estimates of HIV and syphilis prevalence. J Urban Health. 2006;83(Suppl 6):83–97.CrossRefGoogle Scholar
  27. 27.
    Pando M, Marone R, Balan I, et al. HIV and STI prevalence among men who have sex with men (MSM) recruited through respondent driven sampling (RDS) in Buenos Aires, Argentina. Retrovirology. 2009;6(Suppl 3):P103.CrossRefGoogle Scholar
  28. 28.
    Barbosa Junior A, Pascom AR, Szwarcwald CL, Kendall C, McFarland W. Transfer of sampling methods for studies on most-at-risk populations (MARPs) in Brazil. Cad Saude Publica. 2011;27(Suppl 1):S36–44.PubMedGoogle Scholar
  29. 29.
    Tun W, de Mello M, Pinho A, Chinaglia M, Diaz J. Sexual risk behaviours and HIV seroprevalence among male sex workers who have sex with men and non-sex workers in Campinas, Brazil. Sex Transm Infect. 2008;84(6):455–7.PubMedCrossRefGoogle Scholar
  30. 30.
    Kendall C, Kerr LR, Gondim RC, et al. An empirical comparison of respondent-driven sampling, time location sampling, and snowball sampling for behavioral surveillance in men who have sex with men, Fortaleza, Brazil. AIDS Behav. 2008;12(4 Suppl):S97–104.PubMedCrossRefGoogle Scholar
  31. 31.
    Salani Mota RM, Sansigolo Kerr LR, Kendall C, et al. Reliability of self-report of HIV status among men who have sex with men in Brazil. J Acquir Immune Defic Syndr. 2011;57(Suppl 3):S153–6.PubMedCrossRefGoogle Scholar
  32. 32.
    Bermudez-Aza EH, Kerr LR, Kendall C, et al. Antiretroviral drug resistance in a respondent-driven sample of HIV-infected men who have sex with men in Brazil. J Acquir Immune Defic Syndr. 2011;57(Suppl 3):S186–92.PubMedCrossRefGoogle Scholar
  33. 33.
    Damacena GN, Szwarcwald CL, de Souza Junior PR, Dourado I. Risk factors associated with HIV prevalence among female sex workers in 10 Brazilian cities. J Acquir Immune Defic Syndr. 2011;57(Suppl 3):S144–52.PubMedCrossRefGoogle Scholar
  34. 34.
    Damacena GN, Szwarcwald CL, Barbosa Junior A. Implementation of respondent-driven sampling among female sex workers in Brazil, 2009. Cad Saude Publica. 2011;27(Suppl 1):S45–55.PubMedGoogle Scholar
  35. 35.
    Szwarcwald CL, de Souza Junior PR, Damacena GN, Junior AB, Kendall C. Analysis of data collected by RDS among sex workers in 10 Brazilian cities, 2009: estimation of the prevalence of HIV, variance, and design effect. J Acquir Immune Defic Syndr. 2011;57(Suppl 3):S129–35.PubMedCrossRefGoogle Scholar
  36. 36.
    Silva-Santisteban A, Raymond HF, Salazar X, et al. Understanding the HIV/AIDS epidemic in transgender women of Lima, Peru: results from a sero-epidemiologic study using respondent driven sampling. AIDS Behav. 2011;16(4):872–81.CrossRefGoogle Scholar
  37. 37.
    Johnston LG, Caballero TV, Dolores Y, Vales HM. HIV, Hepatitis B/C and syphilis prevalence and risk behaviors among gay/trans/men who have sex with men, Dominican Republic. Int J STD AIDS. (in press). Google Scholar

Copyright information

© Springer Science+Business Media New York 2013

Authors and Affiliations

  • Jane R. Montealegre
    • 1
    • 2
  • Lisa G. Johnston
    • 3
    • 5
  • Christopher Murrill
    • 4
  • Edgar Monterroso
    • 4
  1. 1.Division of Epidemiology, Human Genetics, and Environmental SciencesThe University of Texas School of Public HealthHoustonUSA
  2. 2.Dan L. Duncan Cancer CenterBaylor College of MedicineHoustonUSA
  3. 3.Department of Global Community Health and Behavioral SciencesTulane University School of Public Health and Tropical MedicineNew OrleansUSA
  4. 4.Division of Global HIV/AIDSU.S. Centers for Disease Control and PreventionAtlantaUSA
  5. 5. Global Health Sciences, University of California, San FranciscoSan FranciscoUSA

Personalised recommendations