Remple, Valencia P. BSN, MSN, PhD; Patrick, David M. MD, MHSc, FRCPC; Johnston, Caitlin BA; Tyndall, Mark W. MD, DPH; Jolly, Ann M. PhD
FEMALE SEX WORKERS (FSW) are viewed as a high-risk group susceptible to acquiring HIV and sexually transmitted infections (STI) as well as transmitting them to others.1–3 Because many FSW work in public areas—such as on tracks (street), in red light districts, and in brothels—they may be located and therefore are a primary target of intervention and education messages. In Canada, although up to 80% of FSW work in indoor environments (e.g., massage parlors, escort agencies, private in- or out-call, private indoor microbrothels),4 virtually all research and most intervention is targeted at FSW in the more visible street-based sex trade.
Clients of FSW are a more elusive population. With the exception of when they are attending a sex establishment (e.g., visiting a brothel) or in police custody (e.g., enrolled in “John School”) they are invisible within the general population.5–7 Although clients are thought to be key to the spread of infection from the commercial sex core to the general population,2,8 HIV and STI prevalence, sexual behaviors, and FSW patronage patterns vary, and therefore it may not be reasonable or efficient to lump them all into 1 high-risk category. Furthermore, little is known about the structure and context of the sexual networks that form between clients and FSW who work in the indoor commercial sex industry.
Social network analysis (SNA) is increasingly being used to help explain the spread of STI through focusing on the relationships among individuals in a group. SNA produces statistics that describe the quality, density, position, and structure of relationships.9 Certain network structures provide the potential pathways for rapid propagation should shifts in behavioral patterns occur (e.g., reduction in condom use) or if a pathogen is introduced. Table 1 provides a summary and definitions of selected network concepts. Cohesive subgroups within a network are subsets of individuals who are relatively frequently linked.9 For example, 2-cores are subgroups in which all individuals are connected to at least 2 others in the group. Cliques are another important subgroup and are defined as 3 or more people who are fully connected. These more densely connected subgroups provide the pathways for transmission of HIV and other STI through sexual contact or needle sharing and have been empirically demonstrated to be associated with HIV and STI.11–13
The position of individuals in a network also has implications for HIV and STI spread and acquisition. Centrally located infected individuals are more likely to result in disease propagation than if they are located in the periphery, and likewise, centrally located susceptible individuals may be at greater risk for infection because of greater opportunity.11,12,14
The importance of sexual networks to HIV and STI epidemiology and control is becoming widely accepted. However, conducting network studies can be logistically challenging, particularly when targeting hidden populations. It was the purpose of this study, therefore, to determine whether a classification of higher-risk clients, which could be validated by network analysis, was possible through egocentric interviewing of FSW that would enable straightforward identification for the purposes of targeted intervention. The research questions we examined in this paper were whether so-called high-risk clients occupied important network positions, such as high centrality or membership in 2-cores or cliques. We also tentatively explored the question of whether these high-risk clients were more likely to be HIV and STI infected.
Materials and Methods
This research was conducted in 7 indoor commercial sex establishments in the Canadian cities of Vancouver (population ∼550,000), Richmond (population ∼182,000), Surrey (population ∼350,000), and Burnaby (population ∼197,000). All 4 cities are located in southwestern British Columbia in a region containing more than 50% of the total provincial population and are between zero (i.e., share a border) and 20 km of each other.
A detailed description of the approach used for accessing the establishments and recruiting participants is described elsewhere.15 Briefly, research recruitment occurred concurrent to a community-based outreach service program for indoor FSW that was being conducted by a community-based AIDS service organization. Outreach teams used the phone directory, the Internet, the local newspapers, and word-of-mouth to compile a list of establishments. For each establishment, the team attempted to identify the location, the number of workers, and the primary races or ethnicities of the workers. In consultation with outreach team members, the authors used adapted purposive, targeted sampling16 to select ethnically and geographically diverse sex establishments from which participants would be recruited. The authors consulted with the outreach team members to determine the perceived feasibility of recruiting from specific establishments. The feasibility issue was very important. For example, the managers and staff in some of the establishments appeared very nervous when teams visited, and we opted not to recruit from those establishments because we were unwilling to introduce any potential risk to sex establishment or study personnel. An experienced female ethnographic interviewer and a second female observer accompanied outreach teams to meet the workers and management in the establishments to explain the study, to recruit participants, and to conduct interviews. The observer's role was to document ethnographic field notes during the interview.
Structured interviews were conducted in a private room at the sex establishments. Although the FSW reported having dozens or even hundreds of clients, they were asked to identify only up to 10 of their sexual partners with whom they had had most recent sexual contact, including clients and noncommercial partners. The interviewer completed a separate proxy questionnaire for each of the up to 10 identified partners. Respondents were asked to describe each of their partners in terms of first names or nicknames, age, appearance, ethnicity, city of residence, marital status, and any other unique characteristics. Detailed information on their sexual relationships was collected, including dates of first and last sexual contact, number of sexual contacts, frequency of sexual contact, types of sex acts, and condom use. Finally, respondents were asked if they were aware of each contact's HIV status, whether the contact had had an STI during their relationship, and whether the contact was known to have sex with other FSW or patronize other commercial sex establishments or street-based FSW.
Participants were compensated with 30 Canadian dollars. This study received ethical approval from the University of British Columbia Clinical Research Ethics Board (certificate no. C04-0044).
Data Management and Analysis
Data entry was conducted by the principal investigator and the interviewer together, which facilitated validation of questionnaire data with ethnographic field notes and identification of unique sexual contacts who were identified by 2 or more respondents. Two apparently independent contacts identified by more than 1 respondent were considered the same individual if all of the following characteristics matched: (a) race/ethnicity; (b) location of residence (if available); (c) physical description; and (d) approximate age. If location of residence was not available, at least 2 of the following additional characteristics were required for matching: name, marital status, profession, make of car, and unusual mannerisms or sexual preferences.
The following individuals were included in the network: (a) the respondents; (b) the up to 10 first-order sex partners of respondents on whom proxy data were collected during the interviews (clients and noncommercial partners); (c) the second-order, non-FSW partners of respondents' partners (wives and girlfriends); and the second-order FSW partners of respondents' partners. If a respondent's client had told her that he patronized another sex establishment, he was randomly assigned (by a random number generator) to be sexually linked to 1 of the n number of FSW known to be employed at that establishment. In this way, relational data were available for the respondents, their first-order sex partners, and the second-order partners of their partners (e.g., sexual links between respondents' clients and their proxy-reported wives, girlfriends, and other FSW).
Respondents' sex partners were coded as HIV or STI infected if the respondent answered “definitely yes” to the questions: “Do you think this person has HIV/AIDS?” or “Do you think this person has had an STD since you've known him?” If a respondent identified a contact as HIV or STI infected, the interviewer probed to find out how she knew, and in each case the respondent was definitely sure because the client had disclosed his status to her or because she had noted that the client had genital sores. Unless respondents were sure of their partner's status they were recoded as “maybe.”
For this analysis, the high-risk category was defined as clients who were known to patronize more than 1 sex establishment or who were known as “house regulars.” House regular was a term used in the industry to describe clients who have sex with all or most of the workers in a particular establishment.
Data were first entered into EpiData 3.0.17 A data conversion program (Network conversion program, Ann Jolly 2003) converted relational and attribute questionnaire data into text files that were imported into Pajek18 and Ucinet 619 for graphic visualization and analysis. The 2-cores and cliques were extracted from the total network using Ucinet. SPlus 7.020 was used to conduct statistical analyses. Bivariate relationships between risk category and variables of interest were analyzed using nonparametric cross-tabulation (χ2 and Fisher exact test) and Mann-Whitney U Test for categorical and continuous data, respectively.
Between June 2004 and February 2006, a total of 215 sex establishments were identified, 118 (54.9%) of which were locatable (i.e., had an advertised address or a receptionist who was willing to provide it over the phone). The outreach teams attempted to visit 50 establishments during the study time period and were allowed to return for ongoing visits at 37 (74.0%). The authors selected 7 establishments from 4 cities for research recruitment. The managers at each location allowed interviewers to recruit on site. Compared to the 118 with known addresses, there were no significant differences in terms of geographical location. Compared with the 50 establishments in which the ethnic distribution of workers was known, there were less primarily Asian establishments (28.5% vs. 54.0%) and more mixed race establishments (71.4% vs. 22.0%). Fifty-one FSW were invited to participate, of whom 49 (96.1%) consented.
Respondents reported a median of 30 (IQR 20–50) unique sex partners per month. They provided detailed proxy data on 256 sexual partners. Respondents reported consistent condom use with 217 (84.8%) of their partners. Twelve clients were identified by more than 1 FSW. Eleven (4%) of the original partners were possible matches but did not meet the matching criteria. Of the 234 unique partners, 205 (87.6%) were male commercial clients and were included in the analysis.
The demographic and behavioral characteristics and HIV and STI status of the 205 clients are summarized in Table 2. Because we elicited proxy data on respondents' most recent sexual contacts only (vs. all of their sexual contacts), there was a high proportion of clients whom the respondents had met for the first time immediately preceding the interviews. Nonetheless, they were able to provide detailed information on most of their clients with the exception of their infection status and their street-based FSW patronage. Two (1.0%) and 8 (3.9%) of the clients were known to be HIV infected or have had an STI during the partnership, respectively. The respondents were unaware of the HIV and STI status of approximately half of the clients.
The complete sexual network was composed of a fully connected component of 553 people, including 49 FSW respondents, 205 commercial sex partners, 29 noncommercial sex partners (i.e., husbands and boyfriends), 153 second-order FSW partners (i.e., other FSW partners of respondents' clients), and 117 second-order non-FSW partners (i.e., wives and girlfriends of respondents' partners). Of the 1502 links in the network, 151 (10.1%) were randomly assigned. Table 3 summarizes the median, the interquartile range, and the range of the degree distribution of the 5 types of network members. There were 115 cliques (n = 3) and a single 2-core of 230 individuals comprised of FSW and clients.
Although most of the 205 clients (72%) were known to have had sex with more than 1 FSW, 2 unique categories of client emerged from the interview data that we defined as “higher risk” because of their apparent role in creating multiple sexual linkages between FSW and between sex establishments. The first, which FSWs referred to as “house regulars,” were defined as men who did not restrict their patronage to 1 or 2 workers, but who had sex with most or all the workers at the establishment. Anecdotally, these men were also viewed as the clients who attended the establishments most frequently—many several times per week and in some cases even daily. Several house regulars also had the reputation of seeking out higher-risk activities, such as anal sex, sex without a condom, and “girlfriend experience” (GFE) (GFE is a sexual encounter with higher contact, which may include kissing, cunnilingus, or oral and vaginal sex without a condom). The median number of times respondents reported having sex with the house regular clients was 3 (IQR 1–3), ranged between 1 and 200 times, and occurred during a period of months or years.
The second category, dubbed “establishment bridgers,” comprised clients who were known to patronize more than 1 sex establishment or segment of the sex industry. Thirty-five percent of the clients were known to patronize more than 1 establishment, and 6.3% were known to patronize street-based FSW. The median number of times respondents reported having sex with establishment bridgers was 4 (IQR 2–10), and ranged between 1 and 100 times (data not shown). Seventeen (8%) clients were known both as a house regular and establishment bridger (data not shown).
In total, there were 111 (54%) clients that were categorized as high risk through their house regular or establishment bridging patronage patterns (Fig. 1). The demographic and network characteristics of the high- and low-risk clients are compared in Table 4. High-risk clients were significantly more likely to be members of the large 2-core, BE members of a 3-clique, be a cut vertex (an individual whose removal would result in an increase in the number of components), and had higher centrality measures than the low-risk clients. Furthermore, all of the HIV and STI were reported in high-risk clients, although for HIV this did not achieve statistical significance. Although more high-risk clients created sexual bridges to the general population through wives or girlfriends, this difference was also not statistically significant (58.6% vs. 38.7%, P = 0.07).
It was the goal of this study to determine if clients of indoor FSW could be categorized in a meaningful way that would allow for identification and targeting of those who represented higher sociometric risk. Additionally, we sought to explore the hypothesis that higher-risk clients experience a higher prevalence of HIV and STI. Despite sampling from 4 distinct cities, we found that the indoor commercial sex network was a dense, fully connected network that was characterized by sexual bridges, both between commercial sex establishments and between the commercial sex networks to the general population. On the basis of simple egocentric interviewing of FSW, we were able to ascertain that the client population was not homogeneous in terms of sex-buying behavior and that behaviors existed that could introduce substantial risk to the overall network through creating myriad transmission pathways.
We were able to classify more than half of the clients identified in our sample as high risk in terms of what respondents knew about their patronage patterns. House regulars may increase the overall level of concurrency (sexual relationships that overlap in time) within the network through their frequent, repeated sexual contact with multiple FSW concurrent with sexual relations with wives or girlfriends. These concurrent partnerships may potentiate rapid spread of STI and HIV because there is less time lost after transmission waiting for the current partnership to dissolve.21–25 This may be particularly salient in the context of STI that have short windows of infectiousness and require either a high rate of new partner acquisition or partnership overlap to spread.
Clients who patronize more than 1 commercial sex establishment create bridges between different local networks (i.e., those centered around a particular establishment) that may differ in terms of risk behavior patterns, HIV and STI prevalences, and ethnic groups. For example, clients who patronize both street-based and indoor FSWs form bridges between the outdoor population where the prevalence of injection drug use, HIV, and hepatitis C is high26 to the theoretically lower-risk, lower-prevalence indoor sex worker population. As well, condom use and GFE practices may vary from establishment to establishment, depending on the socioeconomic status of the workers, the management policies, and the house rules. Clients who create bridges between establishments may also contribute to the overall concurrency levels within the network if they go to different establishments or have sex with the same group of FSW repeatedly, concurrent with sexual relations with wives or girlfriends.
When we compared our categorization of high- and low-risk clients, striking differences in terms of network importance were discovered. First, high-risk clients were more likely to be members of highly connected, dense subgroups such as the 2-core and cliques and to be cut vertices. From the graph theoretical perspective, 2-cores are particularly robust because in the circle of connected members there are always 2 independent transmission pathways for every 2-core member. Network fragmentation is therefore reliant on group behavior change. Three-cliques are subgroups that resemble a fully connected triangle of 3 individuals and are important predictors of HIV and STI spread.12,27 In the context of heterosexual sex, they define a group sex event with 2 participants of 1 gender participating. In the commercial sex network, the 3-cliques occurred primarily through “duos,” where a client would pay for 2 FSW in 1 session.
Second, high-risk clients were significantly more central in the network than lower-risk clients, based on 4 different centrality measures. Although each one is based on a different theoretical framework9 and is computed in different ways, these measures have been shown to be highly correlated.28 High centrality measures put individuals in a better position to both acquire and transmit infection throughout the network, for example, through multiple transmission pathways (e.g., degree centrality) or through providing indirect transmission pathways between individuals (e.g., betweenness centrality).9
All of the proxy-reported HIV and STI infections in the client population were in high-risk clients. Recognizing the limitations of relying on proxy-reported infection status, it is nonetheless important to note the potential implications of this finding. Given the high-risk clients' sociometric prominence, they may play a key role in terms of introducing and spreading infection throughout the network, as well as creating bridges to different establishments and to the general population.
Our findings have several implications in terms of developing HIV and STI prevention strategies. First, theoretical removal of high-risk clients from the network would result in fragmentation of the overall network by reducing the number of potential transmission pathways. From the practical perspective, it is not reasonable to literally remove a large segment of the client population from the commercial sex industry. However, weakening the transmission pathways through consistent, correct condom use may significantly reduce the effectiveness of these pathways. The obvious educational message to FSW of 100% condom use remains very important, especially in the context of regular partners and those who the women know are house regulars or patrons of other establishments. This message cannot be overstated, given the unique context of indoor sex work, in which clients and FSW spend longer periods of time together, have sexual relationships that may extend over months or years, and where condom use may become lax. During this study, some of the respondents disclosed that condom use was more likely to be inconsistent with regular clients whom they thought they knew well or liked, a finding which has been reported by others.29,30 Simplified network diagrams could be used as educational tools to reinforce this message, enabling outreach workers to show FSW how the network is linked and how STI and HIV could travel through it.
Second, although more logistically challenging, providing the same educational messages to high-risk clients may be possible. In addition to the 100% condom message clients could be encouraged to reduce the number of concurrent FSW and establishments they patronize, although this may be viewed as “bad for business” by establishment owners and FSW themselves. Educational messaging could be facilitated by the FSW and establishment owners—for example, by showing and explaining simple network diagrams to their regular clients or allowing educational print materials, such as posters or pamphlets, to be made available in the waiting rooms of escort agencies or massage parlors. The internet is also a powerful dissemination tool in the commercial sex trade. Many of the indoor commercial sex establishments have websites, and it may be possible to post information or links on them. As well, there are several high-volume websites where male sex buyers communicate, such as the Pacific Escort Review Board (http://www.perb.ca), which allows links and postings for health education.
This research had several important limitations. First, the study was limited by using a nonprobability sample of FSW to obtain proxy data on commercial clients. Although purposive, targeted sampling is considered more rigorous than simple convenience sampling,16 our sample underrepresented the large population of Asian immigrant women who are purported to be involved in the sex trade in Canada and women who may be working under duress.31 As a result, we may have collected data from FSW who had more power to negotiate condom use, who had more power to choose the sexual activities they were willing to perform, who worked shorter hours, and who saw fewer clients than more vulnerable, hidden women. Furthermore, it is important to point out that our methods precluded a full enumeration of the sexual network because data were collected from only 1 type of node (FSW). However, the study network should be viewed as illustrative, rather than representative, of the types of networks that exist within the commercial sex industry.
Second, proxy reporting on behaviors and attributes of only the most recent sex partners introduced recall bias in the form of clients whom respondents knew less well than some of their more familiar regular clients. Because of this, misclassification of high-risk clients as low risk may have occurred because of respondents simply not knowing some of the first-timers' patronage habits. It is unlikely, however, that misclassification in this manner would result in substantive changes to the differences in terms of network importance, as most of the network characteristics (e.g., centrality measures and membership in cohesive subgroups) are by definition attributed to those who have multiple sex partners and who bridge between establishments.
Third, as previously mentioned, the method used to ascertain HIV and STI status was not rigorous and therefore precluded testing of any hypotheses about the association between network properties and infection. It is likely that the client sample HIV and STI “prevalence” of 1% and 4% was an underestimate, given that for 55% and 46% of the clients the respondents were unaware of their HIV and STI status, respectively. Any generalizations based on these results must be extremely tentative, and these findings should only be used to guide further inquiry.
Fourth, bias may have been introduced through artifactual relationships within the data. The high centrality measures were to some extent dictated by the data collection because the house regular designation resulted in automatically generating links to all FSW at the specific establishment, thereby potentially creating artificially low relative degree measures for those clients who were not identified as house regulars. Furthermore, misclassification of low-risk clients due to missing data resulted in lower centrality scores and a lower likelihood of being included in the cohesive subgroups. The same may be true for establishment bridgers, in that by definition they might be expected to have high information and betweenness centrality because of their bridging behaviors.
Despite these limitations, it should be underscored that the purpose of this analysis was to determine if a feasible, simple method was possible that could effectively dichotomize higher- versus lower-risk clients of FSW. This study has demonstrated that it is possible to identify theoretically high-risk commercial sex clients from the network perspective using straightforward approaches. This is particularly salient given the unique context of indoor commercial sex work, where clients and FSW spend more time together, where the environment is more conducive to vaginal and anal sex, where sexual relationships may continue over months or years, and where condom use may become inconsistent. These findings may assist health providers and educators to provide targeted, contextually appropriate education messages to women who work in the indoor commercial sex industry and to their male clients.
1. Kreiss JK, Koech D, Plummer FA, et al. AIDS virus infection in Nairobi prostitutes. Spread of the epidemic to East Africa. N Engl J Med 1986; 314:414–418.
2. Morris M, Podhisita C, Wawer MJ, et al. Bridge populations in the spread of HIV/AIDS in Thailand. AIDS 1996; 10:1265–1271.
3. Potterat JJ, Rothenberg R, Bross DC. Gonorrhea in street prostitutes: Epidemiologic and legal implications. Sex Transm Dis 1979; 6: 58–63.
4. Benoit C, Millar A. Dispelling myths and understanding realities: Working conditions, health status, and exiting experiences of sex workers [final report]. Victoria, BC: Prostitutes Empowerment, Education and Resources Society; 2001. Available at: http://www.peers.bc.ca/images/DispMythsshort.pdf
5. Lowman J, Atchison C, Fraser L. Men who buy sex, Phase I report. A study funded by the Ministry of Attorney General, Province of BC [final report]. Victoria, BC: Ministry of Attorney General, Province of BC; 1996.
6. Lowman J. Prostitution in Canada. In: Jackson M, Griffiths C, eds. Canadian Criminology: Perspectives on Crime and Criminality. Toronto, ON: Harcourt Brace Jovanovich; 1995:333–359.
7. Roberts JM Jr, Brewer DD. Estimating the prevalence of male clients of prostitute women in Vancouver with a simple capture-recapture method. J R Stat Soc A 2006; 169:745–756.
8. Morrison L, Weiss HA, Buve A, et al. Commercial sex and the spread of HIV in four cities in sub-Saharan Africa. AIDS 2001; 15 (Suppl 4):S61–S69.
9. Wasserman S, Faust K. Social Network Analysis: Methods and Applications. Cambridge, MA: Cambridge University Press; 1994.
10. Scott J. Social Network Analysis: A Handbook. London: Sage; 2000.
11. Friedman SR, Neaigus A, Jose B, et al. Sociometric risk networks and risk for HIV infection. Am J Public Health 1997; 87:1289–1296.
12. Rothenberg RB, Sterk C, Toomey KE, et al. Using social network and ethnographic tools to evaluate syphilis transmission. Sex Transm Dis 1998; 25:154–160.
13. Potterat JJ, Rothenberg RB, Muth SQ. Network structural dynamics and infectious disease propagation. Int J STD AIDS 1999; 10:182–185.
14. De P, Singh AE, Wong T, et al. Sexual network analysis of a gonorrhoea outbreak. Sex Transm Infect 2004; 80:280–285.
15. Remple VP, Johnston C, Patrick DM, et al. Conducting HIV/AIDS research with indoor commercial sex workers: Reaching a hidden population. Prog Commun Health Partnerships, In press.
16. Watters JK, Biernacki P. Targeted sampling: Options for the study of hidden populations. Soc Probl 1989; 36:416–430.
17. Lauritsen JM, Bruus M. EpiData. A comprehensive tool for validated entry and documentation of data [computer program]. Version 3. Odense, Denmark: The EpiData Association; 2003.
18. Batagelj V, Mrvar A. Pajek [computer program]. Slovenia: University of Ljubljana; 1996.
19. Borgatti SP, Everett MG, Freeman LC. Ucinet for Windows. Software for social network analysis [computer program]. Version 6. Harvard: Analytic Technologies; 2002.
20. SPLUS for Windows [computer program]. Version 7.0. Seattle: Insightful Corporation; 2005.
21. Watts CH, May RM. The influence of concurrent partnerships on the dynamics of HIV/AIDS. Math Biosci 1992; 108:89–104.
22. Mode CJ, Dietz K. On some formulas in a partnership model from the perspective of a semi-Markov process. J Math Biol 1994; 32:161–169.
23. Hudson CP. Concurrent partnerships could cause AIDS epidemics. Int J STD AIDS 1993; 4:249–253.
24. Morris M, Kretzschmar M. Concurrent partnerships and the spread of HIV. AIDS 1997; 11:641–648.
25. Kretzschmar M, Morris M. Measures of concurrency in networks and the spread of infectious disease. Math Biosci 1996; 133:165–195.
26. Spittal PM, Bruneau J, Craib KJ, et al. Surviving the sex trade: A comparison of HIV risk behaviours among street-involved women in two Canadian cities who inject drugs. AIDS Care 2003; 15:187–195.
27. Potterat JJ, Muth SQ, Rothenberg RB, et al. Sexual network structure as an indicator of epidemic phase. Sex Transm Infect 2002; 78 (Suppl 1):i152–i158.
28. Rothenberg R, Potterat JJ, Woodhouse DE, et al. Choosing a centrality measure: Epidemiologic correlates in the Colorado Springs study of social networks. Soc Netw 1995; 17:273–297.
29. Jackson L, Highcrest A, Coates RA. Varied potential risks of HIV infection among prostitutes. Soc Sci Med 1992; 35:281–286.
30. Vanwesenbeeck I, van Zessen G, de Graaf R, et al. Contextual and interactional factors influencing condom use in heterosexual prostitution contacts. Patient Educ Couns 1994; 24:307–322.
31. Suthibhasilp N, Petroff L, Nipp D. Trafficking in women, including Thai migrant sex workers in Canada [final report]. Toronto, ON: Status of Women Canada; 2000.