Social-Environmental Factors and Protective Sexual Behavior Among Sex Workers: The Encontros Intervention in Brazil

Abstract

Objectives. We sought to determine the association of social–environmental factors with condom use and sexually transmitted infections (STIs) among 420 sex workers participating in an STI/HIV prevention study in Corumbá, Brazil, to inform future intervention efforts.

Methods. Participants provided urine samples for polymerase chain reaction testing of chlamydia and gonorrhea and responded to multi-item scales addressing perceived social cohesion, participation in networks, and access to and management of resources. We conducted multivariate log-linear and negative binomial regression analyses of these data.

Results. Increased social cohesion was inversely associated with number of unprotected sex acts in the preceding week among women (adjusted incidence rate ratio [IRR] = 0.80; P < .01), and there was a marginal association among men (adjusted IRR = 0.41; P = .08). Women's increased participation in social networks was associated with a decrease in frequency of unprotected sex acts (adjusted IRR = 0.83; P = .04), as was men's access to and management of social and material resources (IRR = 0.15; P = .01). Social–environmental factors were not associated with STIs.

Conclusions. The social context within which populations negotiate sexual behaviors is associated with condom use. Future efforts to prevent STI/HIV should incorporate strategies to modify the social environment.

With growing agreement about the importance of economic, political, and social contexts in shaping sexual behaviors and the course of the HIV epidemic, HIV prevention researchers have escalated calls for implementation of multilevel structural and social interventions that modify social environments and empower communities.^1–9 Interventions that seek to create social cohesion, improve access to resources, create networks, ensure community participation, mobilize communities, and otherwise establish human rights are especially important for marginalized groups that experience discrimination and exclusion from public life.

Improving social–environmental factors has become an integral element in HIV prevention^2,5,10,11 and health promotion generally.^12–15 The United Nations Joint Programme on HIV/AIDS and the World Health Organization now recommend that local governmental and nongovernmental efforts to promote health among marginalized groups incorporate strategies to change social environments at the community level, including social mobilization.^16–18

Challenges to the successful implementation of such interventions include identifying the factors in the social environment that are key to shaping health and behavior and are amenable to change, how to change these factors, and how to measure these intricate processes.^7,10,19,20 Although qualitative studies have described social intervention processes and experiences,²¹ most efforts to change social environments have not included rigorous empirical components such as measurement of the factors targeted for change.

This gap is especially evident when the objective is to improve sexual behaviors.^3,19 Only limited causal evidence from longitudinal studies has documented that changes in the social environment may translate into improved behaviors and reduced rates of HIV and other sexually transmitted infections (STIs).²² However, growing cross-sectional evidence points to associations between sexual behavior and STIs, including HIV, and various measures of social participation or perceived community cohesion.^23–26

In an attempt to fill these gaps in social intervention research, a multidisciplinary team of epidemiologists, physicians, social scientists, sex worker advocates, and governmental representatives developed a combined clinical and social intervention among sex workers in Corumbá, Brazil, called Encontros (Coming Together). Encontros was designed to improve health behaviors and outcomes by offering clinic-based services in combination with an array of strategies to create an environment enabling and reinforcing healthy behaviors.

The social–environmental aspects of the project were predicated on the idea that sex workers' negotiations with clients, including negotiation of condom use, are embedded not only in the psychology of risk taking and the economics of access to resources but also in social factors such as gender inequality, discrimination, violence, competition, and social marginalization. A major component of the Encontros intervention was the development of community-based workshops and events designed to build trust and inclusiveness among sex workers and between sex workers and the Corumbá community.

We briefly describe the Encontros intervention and use baseline data from the intervention cohort to provide further evidence of an association between aspects of the social environment and sexual behavior.

METHODS

Corumbá is located in the Pantanal on Brazil's western border with Bolivia. In addition to approximately 100 000 inhabitants, Corumbá is home to Brazilian and Bolivian migrants and attracts approximately 75 000 tourists each year. Sex commerce in Corumbá caters to tourists and locals at brothels, at local bars and restaurants, on board fishing vessels, and on the street.

The majority of women who work in higher-end brothels come from other states in Brazil, tend to be young and relatively well-educated, and have little contact with the local population. The male, female, and transvestite workers in the streets, bars, and lower-end brothels are mostly from the local area and tend to be poor, with little formal education. The clients of sex workers in Corumbá range from wealthy national and international tourists to lower-income truck drivers.²⁷

The Encontros Intervention

From 2003 to 2005, Encontros aimed to decrease cases of HIV and other STIs and encourage adoption of consistent condom use among female, male, and transvestite sex workers in Corumbá through the use of joint clinical and social intervention strategies. The project was designed to engage sex workers on an individual level via participation in counseling on and testing for HIV and other STIs, on an interpersonal level via peer education, and on a community level via outreach and social activities. Sex workers were offered enhanced access to clinic-based holistic sexual health care, including psychological counseling services. Clinic staff members were trained in provision of high-quality services that emphasized confidentiality and respect, reinforcing the Brazilian human rights framework on HIV prevention and destigmatization of sex work.²⁸

Community activities were designed with input from project participants; the goal was to extend and strengthen peer relationships by providing sex workers with opportunities to engage in dialogue around sex work, discrimination, human rights, and prevention of HIV and other STIs. Community activities included peer-educator outreach; workshops, training sessions, and events; and distribution of condoms and educational materials designed by the Ministry of Health as part of a national campaign to reduce the stigma associated with sex work. Project-sponsored workshops included various artisanal workshops (soap, candle, and chocolate making), a theater workshop, and a fashion design workshop.

To showcase workshop efforts, the project staff, including the peer education team, organized “hot pink” parties, which were essentially cultural showcases for the sex workers that occurred at the city's cultural center, a public space where attendants included project participants and the general public. Project staff ensured that performance messages focused on HIV prevention and stigma reduction; at the same time, the experience of celebrating culture and performance in a community space was hypothesized to facilitate sex workers' social cohesion and social integration with the community and reduce stigma against sex workers.

The project also sought to forge broad partnerships between the project participants and representatives of other entities, beginning with the establishment of a project advisory committee and continuing on to the formation of partnerships with the city's family health program, university professors and students (some of whom interned with the project), and the community cultural center. The goals of these partnerships with the community at large were to ensure that sex workers gained recognition as partners involved in health and human rights efforts, to integrate sex workers into the community dialogue on health and strengthen their community identity, and to sustain activities initiated by the project with and for the sex workers.

Data Collection Procedures

Rolling enrollment was initiated in July 2003 and concluded in January 2005. All people 18 years or older who self-identified as sex workers, spoke Portuguese or Spanish, and did not plan on leaving the study area permanently in the coming month were invited to participate. Overall, 420 sex workers were enrolled. Participation included an enrollment visit and 4 scheduled follow-up visits at 3, 6, 9, and 12 months.

Each data collection visit included administration of a structured, interviewer-administered questionnaire; counseling on HIV and other STIs; collection of urine and blood samples for STI testing; and treatment of STIs if indicated. Women underwent a mammary, pelvic, and bimanual exam; men underwent a genital exam, including inspection of the penis and scrota and palpation of the inguinal area.

The COBAS AMPLICOR polymerase chain reaction system (Roche Diagnostics, Branchburg, NJ) was used to test first-catch urine specimens for chlamydia and gonorrhea. Specimens were frozen and stored at −20°C at the study clinic before transport on dry ice to a Ministry of Health reference laboratory in Fortaleza, Brazil, for immediate processing according to the manufacturer's instructions.

Interviews collected information on participants' sociodemographic characteristics; history of sex work and current sex work practices; sexual behavior with new clients, regular clients, and nonpaying partners; sexual health; condom negotiation; and participation in sexual health education and project social and cultural activities. Also addressed were sex workers' perceptions of their social environment, including multi-item indexes of social cohesion, participation in social networks, and access to and management of material and social resources.

Measures

The independent variables assessed in our analyses included social–environmental factors hypothesized to facilitate protective sexual behaviors: perception of mutual aid, trust, and support (i.e., social cohesion); participation in social networks; and access to and management of social and material resources (Table 1 ). To assess social cohesion, participants were asked to indicate their level of agreement with a series of statements; response options ranged from strongly agree to strongly disagree.

TABLE 1.

Social–Environmental Measures: Encontros Cohort, Corumbá, Brazil, 2003–2005

		Properties
Factor	Scale Items	No. of Items	Raw Scores, Mean (Range)	Cronbach α
Social cohesion: perception of mutual aid, trust, connectedness, support among sex workers	Strongly agree, agree, disagree, or strongly disagree:	13; summed and standardized (0, 1, 2, 3 points for ascending responses)b	20.2 (2–36)	0.81
	• You can count on your colleaguesa if you need to borrow money
	• You can count on your colleagues to accompany you to the doctor or hospital
	• You can count on your colleagues if you need to talk about your problems
	• In general, your colleagues in the area where you work only worry about themselves
	• You can count on your colleagues if you need advice
	• You can count on your colleagues if you need somewhere to stay
	• You can count on your colleagues to help deal with a violent or difficult client
	• You can count on your colleagues to help you find clients
	• You can count on your colleagues to support the use of condoms
	• The group of women/men with whom you work is an integrated group
	• In general, the people you work with are always arguing among each other
	• You can trust the majority of the people working in your area
	• In general, the people you work with get along well
Participation in networks: access to and participation in social and other community networks	Do you participate in:	7; summed and standardized scores (no = 0; yes = 1)	2.6 (0–7)	0.61
	• Church or other religious activities
	• Clubs (sports, student groups, parent groups)
	• Neighborhood associations
	• Cultural activities (art, dance, music, samba school)
	• Community center activities
	• Activist organizations/groups
	• Volunteer/help projects
	• Other activities
Access to and management of material and social resources, including goods and services	Do you have a checking account at any bank?	10: summed and standardized scores (no = 0; yes = 1)	3.6 (0–10)	0.61
	Do you have a savings account?
	Have you purchased/paid for health insurance in the last 3 months?
	Have you purchased the following in the last 3 months:
	• Furniture
	• Electrical appliances (radio, TV, blender, other)
	• Clothes
	• Cosmetics
	• CDs
	• Books
	• Cellular phone
	• Other

^aReference is sex worker peers.

^bTruncated lowest score value at 8 and highest at 30 owing to scarcity of observations at the lowest and highest ranges.

We assessed participation in social networks by asking sex workers a series of yes-or-no questions about memberships or participation in various social and group activities. Participation in social networks is often considered a measure or component of social capital; however, because definitions and measurement of social capital vary widely in the sociological and health promotion literatures,²⁹ we opted not to use this terminology.

Finally, to assess access to and management of social and material resources, participants were asked a series of yes-or-no questions regarding purchasing of household and personal goods and their level of access to services such as banking and health insurance. Social–environmental items were drawn from the literature on social interventions with sex workers,^26,30 synthesized based on expert knowledge, pilot tested with sex workers residing outside of the study community, evaluated with respect to response clarity and variation, and adapted to the local context.

Responses were summed to allow bivariate analyses (Table 2 ). They were also standardized in multivariate analyses (Table 3 ) to facilitate interpretation.

TABLE 2.

Social–Environmental Factor Score Means, by Sociodemographic and Clinical Characteristics: Encontros Cohort (n = 420), Corumbá, Brazil, 2003–2005

		Social Cohesion			Participation in Social Networks			Access to and Management of Social and Material Resources
Cohort Characteristics	Cohort, No. (%)	Mean (SD)	F Test	P	Mean (SD)	F Test	P	Mean (SD)	F Test	P
Cohort, mean (SD)		20.2 (4.5)			2.6 (1.7)			3.6 (1.9)
Age, y			0.62	.60		0.77	.51		8.70	<.01
18–24	219 (52.1)	20.4			2.5			3.6
25–30	112 (26.7)	20.0			2.8			4.1
31–35	43 (10.2)	20.3			2.5			3.7
> 35	46 (11.0)	19.6			2.6			2.5
Gender			0.16	.85		7.00	<.01		0.48	.62
Women	385 (91.7)	20.2			2.5			3.6
Men	19 (4.5)	20.7			3.1			3.8
Transvestite	16 (3.8)	20.5			3.9			4.0
Years of schooling			0.77	.51		2.40	.07		6.90	<.01
< 4	52 (12.4)	20.4			2.6			2.8
4–7	226 (53.8)	20.1			2.4			3.6
8–10	99 (23.6)	20.1			2.9			3.7
≥ 11	43 (10.2)	21.2			2.6			4.5
Individual monthly incomea			1.30	.27		0.48	.62		41.90	<.01
Lowest tertile	145 (34.5)	20.7			2.4			2.7
Middle tertile	140 (33.3)	20.1			2.6			3.7
Highest tertile	135 (32.1)	19.8			2.6			4.5
Partnership status			5.40	<.01		1.30	.26		1.20	.30
Married/cohabitating	60 (14.3)	18.5			2.4			3.3
Has partner	172 (40.9)	20.3			2.7			3.8
Single	188 (44.8)	20.6			2.5			3.6
Nationality			3.20	.08		0.01	.92		2.00	.16
Brazilian	405 (96.4)	20.2			2.6			3.6
Bolivian/Paraguayan	15 (3.6)	22.3			2.6			4.3
Primary workplace			2.60	.04		2.00	.10		5.20	<.01
Brothel	92 (21.9)	19.6			2.6			4.4
Bar	207 (49.3)	20.2			2.4			3.3
Street	51 (12.1)	20.8			2.7			3.5
Motel/hotel	28 (6.7)	18.8			3.1			3.5
Telephone/otherb	42 (10.0)	21.8			2.9			3.5
Age at initiation of sex work, y			0.38	.54		4.60	.03		0.55	.46
< 18	178 (42.4)	20.1			2.7			3.5
≥ 18	242 (57.6)	20.3			2.4			3.7
Previous HIV test			0.23	.63		1.90	.17		2.00	.16
Yes	276 (65.7)	20.1			2.5			3.7
No	144 (34.3)	20.4			2.7			3.4
Positive chlamydia or gonorrhea test result			3.00	.08		1.40	.23		1.10	.29
Yes	73 (17.6)	21.0			2.7			3.8
No	342 (82.4)	20.0			2.5			3.6

Note. Scores were not standardized in bivariate analyses.

^aMedian incomes in the lowest, middle, and highest tertiles were $260, $600, and $1300, respectively (1 Real = US $0.48).

^bIncludes sex workers whose clients reach them by telephone and arrange meetings at either the client's house or an agent's house as well as sex workers who are hired for private parties.

TABLE 3.

Associations of Social–Environmental Characteristics With STIs Among Women and With Unprotected Anal and Vaginal Sexual Acts Among Women, Men, and Transvestites: Encontros Cohort, Corumbá, Brazil, 2003–2005

		Unprotected Penetrative Sex Acts in Preceding Weekc
Social–Environmental Characteristica	Positive STI Test Result,b Adjusted Prevalence Ratio (95% CI)	Women, Adjusted IRR (95% CI)	Men, Adjusted IRR (95% CI)	Transvestites, Adjusted IRR (95% CI)
Social cohesion	1.10 (0.94, 1.30)de	0.80 (0.68, 0.95)def	0.41 (0.15, 1.10)d	0.04 (0.00, 3.30)d
Participation in social networks	1.10 (0.93, 1.30)de	0.83 (0.70, 0.99)def	1.70 (0.78, 3.80)d	1.60 (0.42, 5.90)d
Access to/management of social and material resources	1.20 (0.98, 1.50)deg	1.10 (0.90, 1.30)deg	0.15 (0.03, 0.68)dg	0.40 (0.05, 2.90)dg

Note. CI = confidence interval; IRR = incidence rate ratio; STI = sexually transmitted infection. Values were derived from multivariate log-linear and negative binomial regressions.

^aScores are continuous in standard deviations.

^bAmong women who provided samples (N = 380; excludes 2 participants with inhibited samples and 3 participants who failed to provide samples), 72 tested positive for chlamydia/gonorrhea; there were too few STIs among men or transvestites to test associations.

^cData are missing for 29 respondents. For women, n = 359; for men, n = 17; for transvestites, n = 15.

^dAdjusted for age.

^eAdjusted for partnership status.

^fAdjusted for primary place of work.

^gAdjusted for income.

Dependent variables included condom use and prevalent STIs. Total number of unprotected penetrative (vaginal or anal) sex acts engaged in during the preceding week was used to assess condom use. Prevalent STIs were defined as positive chlamydia or gonorrhea test results at enrollment. Only data from enrollment visits were included in our analyses.

Data Analysis

Questionnaire data were double entered. Stata version 8.0 (StataCorp, College Station, TX) was used in analyzing the data.³¹ We used Cronbach alpha coefficients to assess the internal consistency of the social–environmental scales. In addition, we entered data into an item response modeling program^32,33 in which a maximum-likelihood function with a Rasch model was used to simultaneously place people and items along a construct continuum. The item response modeling program assigned each participant scores for social cohesion, participation in social networks, and access to and management of material and social resources, and all of the scale items were assigned an endorsability level indicating the likelihood of a positive response.

We tested the hypothesis that participants reporting higher levels of positive social–environmental factors would report fewer unprotected sex acts and be less likely to test positive for an STI. Bivariate associations were assessed via comparisons of mean levels of the social–environmental constructs across sociodemographic, occupational, and health behavior characteristics. We assessed covariates that were associated with the social environment at P < .1 in bivariate analyses and were hypothesized to be confounders or effect modifiers in multivariate regression analyses.

The association between standardized social–environmental variable scores and the presence of STIs was estimated via log-linear regression analyses, which yielded prevalence ratios. The association of standardized social–environmental variable scores with number of unprotected sex acts was estimated with negative binomial regression analyses, which yielded incident rate ratios (IRRs). Given effect modification, multivariate analyses were conducted separately for female, male, and transvestite participants.

RESULTS

Of the 474 potential participants screened for enrollment, 51 were ineligible and 3 chose not to participate. Overall, 420 sex workers were enrolled (385 female sex workers, 19 male sex workers, and 16 transvestites). All but 15 participants were Brazilian. Half of the participants were aged 24 years or younger (Table 2), but ages ranged upward to 57, 33, and 32 years among women, men, and transvestites, respectively (data not shown).

Most of the cohort members had completed at least some primary schooling (8 years in Brazil); only 10% had completed at least some technical or university education. Approximately 42% of the participants had entered into sex work before the age of 18 years. Almost 18% had positive chlamydia or gonorrhea test results at baseline. Mean numbers of sex acts reported in the preceding week were 4.7, 3.3, and 5.0 among women, men, and transvestites, respectively (data not shown).

The scales measuring social cohesion, participation in social networks, and access to and management of material and social resources had acceptable Cronbach alpha reliabilities (0.81, 0.61, and 0.61, respectively; Table 1). Item response modeling revealed that none of the individual items in any of the 3 scales varied more or less than expected. In addition, increases in item response endorsability coincided with increases in participants' mean scores, an indication that the items were measuring the same construct.

Multivariate models were run with item response modeling scores as well as summed and standardized scores. Because these 2 scoring approaches led to virtually identical multivariate results, we present only summed and standardized scores. Measures of the social environment were correlated, but no pair of measures had a correlation coefficient above 0.2.

In bivariate analyses, higher levels of perceived social cohesion were associated with being single and with working in bars, on the street, or via the telephone or other means (Table 2). A higher level of participation in social networks was associated with being transvestite or male and being younger at commencement of sex work. A higher level of access to and management of social and material resources was associated with younger age, more years of education, higher income, and working in a brothel.

In multivariate analyses, increased social cohesion among female participants was inversely associated with number of unprotected sex acts in the preceding week (Table 3). The adjusted incidence rate ratio of 0.80 (P < .01) indicated that as levels of perceived social cohesion increased by 1 standard deviation, the number of unprotected sex acts fell at a rate of 0.8; in other words, an increase in social cohesion of 1 standard deviation would result in a typical female sex worker engaging in approximately 1 less unprotected sex act per week. Similarly, increased participation in social networks was inversely associated with number of unprotected sex acts in the preceding week among female participants (adjusted IRR = 0.83; P = .04). Neither chlamydia infection nor gonorrhea infection was associated with social–environmental measures among women.

Among male sex workers, increases in perceived social cohesion were marginally associated with fewer unprotected sexual acts in the preceding week (adjusted IRR = 0.41; P = .08), and increased access to and management of material and social resources were significantly associated with fewer unprotected sex acts in the past week (adjusted IRR = 0.15; P = .01; Table 3). Social–environmental characteristics were not associated with unprotected sex acts among transvestites, and there were too few STI cases to assess associations among men or transvestites.

DISCUSSION

We found that increased perceptions of mutual aid, trust, and connectedness (social cohesion) were associated with fewer reported unprotected sex acts in the preceding week among both female and male sex workers. Also, increased participation in social networks among women was associated with fewer unprotected sex acts in the same time period, as was access to and management of social and material resources among men. Our findings are corroborated by findings from a number of studies. One study showed that perceived neighborhood social cohesion was associated with increased condom use among adolescents in Baltimore, Maryland,²⁵ and another showed that cognitive social capital, or perceived reciprocity and community support, was significantly associated with increased condom use among men and women in rural South Africa.²²

Our finding that women's participation in social networks was associated with their sexual behavior is consistent with results indicating that structural social capital, a measure of group membership and intensity of membership, is related to condom use among South African women.²² Other investigators looking at US state-level data have found that increased social capital is associated with some protective sexual behaviors among adolescents but not with condom use.²³

Among men only, access to and management of social and material resources were significantly associated with fewer unprotected sexual acts. We were unable to locate other studies investigating the association of sexual behavior with access to social and material resources, which is an area that merits further research. Because the observed association with behaviors remained after control for income, this finding may lend itself to Lynch et al.'s hypothesis that neomaterial conditions affect health.³⁴ According to these authors, health inequalities may result from “a combination of negative exposures and lack of resources held by individuals, along with systematic underinvestment across a wide range of human, physical, health, and social infrastructure.”^34(p1202) Responses to our measure may reflect this amalgam of social deficiencies in that the scale items elicit information about access to functional, social infrastructure.

Contrary to previous individual-level³⁵ and group-level^24,36 results, we did not find an association between any measured aspects of the social environment and STIs among women at baseline. Results showing that increased condom use was not reflected in decreased STI frequency have been attributed to differences in the transmissibility and infectivity rates of the STI in question, the background prevalence of infection, differences in infection exposures, and measurement issues.³⁷ Although our results may indicate that an enabling environment supports condom use, only 1 unprotected sex act can lead to infection, particularly in the case of highly transmissible STIs. In addition, the time frames for STIs and behaviors measured in this analysis were not equivalent: we measured condom use in the past week and prevalent STIs.

Limitations

The social–environmental measures we used had acceptable alpha reliabilities of 0.60 or higher.³⁸ Although scores in this range may reflect only moderate reliability, they may also reflect the limited suitability of the Cronbach alpha coefficient as an estimate of reliability for binary response data. There are few alternatives for evaluating scale properties; one is item response modeling, which was used in this analysis. Item response modeling results were more robust than those observed with the Cronbach alpha coefficient, providing additional evidence that our scales were reasonable measures of the constructs assessed; however, efforts are warranted to further improve the measurement of these social–environmental factors. (Although presentation of item response modeling results is beyond the scope of this article, Wilson et al.'s introduction to item response modeling in the health sciences provides an excellent overview.³⁹)

Given the cross-sectional nature of our findings, we cannot make causal inferences regarding the relationship between the social environment and protected sex. It is also important that our findings among men and transvestites be interpreted with caution. This study was not originally powered for gender-stratified analyses; these analyses were performed as a result of the presence of effect modification. Because so few previous studies have examined associations between the social environment and sexual behaviors among men and transvestites, our presentation of findings for these groups is merited. In addition, our study provides evidence that relationships between social–environmental variables and behaviors differ according to gender.

Current Evidence

Although the results of most previous cross-sectional studies support an association between the social environment and sexual behavior, causal evidence remains limited. A few community-level projects seeking to change social environments have generated supportive evidence of the positive effects of social interventions. For example, prevention efforts with gay men to change community norms regarding safe sex have led to reductions in unprotected anal intercourse^40–42 and improved community support for safe sex.⁴⁰ However, these efforts have not involved detailed empirical examinations of social–environmental mediators, and thus it is difficult to draw causal inferences between measured changes in the social environment and sexual behavior.

The most well-known community-based HIV prevention and mobilization effort involving sex workers is the project in Sonagachi, India, where 15 years of evolving organizing efforts succeeded in empowering sex workers to prevent HIV transmission.^43–45 This grassroots project did not include rigorous measurement of community development, perceived stigma, or empowerment, and thus again it was difficult to infer the existence of explicit causal associations between the social environment and improved outcomes.

A Sonagachi-style project in Rio de Janeiro measured indicators of social cohesion, social participation, and access to and control over resources and documented their association with condom use among female sex workers. However, no significant increases in protective behaviors were documented over the course of the intervention evaluation, potentially as a result of the restricted time frame of the program.²⁶

A combined social and structural intervention with sex workers in the Dominican Republic was successful in reducing STIs and increasing condom use; this intervention included efforts to establish cohesion among sex workers.³⁰ A community, government, and academic partnership intended to reduce HIV transmission in a South African mining town failed to create an enabling environment for HIV prevention but has provided insight into the processes and barriers involved in building cohesive movements among sex workers and other community members.^21,46,47

Directions for Future Research

Obtaining causal evidence demonstrating that the social environment shapes sexual behaviors and can be modified to increase protective behaviors should be a priority for future research; such studies necessitate the use of longitudinal designs.^22,35 To reach this end, future efforts should also focus on attaining a deeper understanding of the community-level processes that shape behavior, particularly among marginalized populations. Although HIV prevention researchers have made headway in theorizing about contextual processes (see the work of Campbell and colleagues^{10,21,48–50}), there is a need to substantiate existing theory with empirical evidence. Future research should also include efforts to improve the definition and measurement of community-level constructs, expanding the collection of tools available to better assess the reliability and validity of our measures.

We hope that additional longitudinal analyses of the Encontros project will provide insight into the pathways through which participation in intervention activities may lead to improved health through modification of social–environmental mediators.