HIV Risk Behavior Patterns, Predictors, and Sexually Transmitted Disease Prevalence in the Social Networks of Young Roma (Gypsy) Men in Sofia, Bulgaria

Kabakchieva, Elena MD*; Vassileva, Sylvia MS*; Kelly, Jeffrey A. PhD†; Amirkhanian, Yuri A. PhD†‡; DiFranceisco, Wayne J. MS†; McAuliffe, Timothy L. PhD†; Antonova, Radostina MS*; Mihaylova, Milena MS*; Vassilev, Boyan MS*; Khoursine, Roman MS‡; Petrova, Elena MD§

Sexually Transmitted Diseases: August 2006 - Volume 33 - Issue 8 - pp 485-490
doi: 10.1097/01.olq.0000204533.20324.56

Objectives and Goal: This research studied predictors of high-risk sexual practices and sexually transmitted disease (STD) prevalence among Roma (Gypsy) men's social networks in Sofia, Bulgaria.

Study Design: Fifty-four socially active individuals, approached in Roma neighborhood venues, recruited members (n = 296) of their own networks into the study. Participants completed sociometric and risk behavior interviews and were tested for chlamydia, gonorrhea, syphilis, and trichomonas.

Results: Men had a mean of 7 partners in the past year. Fifty-nine percent had multiple partners in the past 3 months. Seventy-three percent reported recent unprotected vaginal and 51% unprotected anal intercourse. Fifty-nine percent of men had sex with other men in the past year. Twenty-two percent had one of the STDs. The social network to which an individual belonged accounted for 23% to 27% of variance in predicting sexual risk behavior.

Conclusions: One's social network was the most powerful predictor of HIV risk behavior. HIV/STD prevention interventions directed toward entire social networks are especially promising.

A study among Roma (Gypsy) male social networks revealed high rates of sexually transmitted disease and sexual risk. The network to which one belonged was the major predictor of sexual risk behavior.

From the *Health and Social Development Foundation, Sofia, Bulgaria; the †Center for AIDS Intervention Research (CAIR), Department of Psychiatry and Behavioral Medicine, Medical College of Wisconsin, Milwaukee, Wisconsin; ‡Municipal Hospital for Infectious Diseases No. 30 Named after S.P. Botkin, St. Petersburg, Russia; and the §Department of Dermatology and Venereology, Medical University, Sofia, Bulgaria

The authors thank Nikolay Tsankov, Anastas Mihaylov, Teodora Pencheva, Marin Iliev, Dimitar Sashov, and Pavel Pavlov for their contributions to the study.

This research was supported by grants R01-MH064410 and P30-MH052776 from the U.S. National Institute of Mental Health.

Correspondence: Yuri A. Amirkhanian, PhD, Center for AIDS Intervention Research (CAIR), Medical College of Wisconsin, 2071 North Summit Avenue, Milwaukee, WI 53202. E-mail:

Received for publication August 13, 2005, and accepted December 12, 2005.

Article Outline

ROMA (GYPSIES) ARE THE LARGEST and most disadvantaged ethnic minority population in Eastern Europe. The vulnerability of Roma communities to sexually transmitted diseases (STDs) and HIV has been noted by U.N. agencies and the World Bank1,2 but not systematically studied. Of approximately 15 million Roma in the world, 8 to 10 million live in Eastern Europe. Bulgaria's 750,000 Roma constitute nearly 9% of the population.3 Roma originally migrated from northern India and are identifiable by their language, culture, and customs. The history of the Roma in Europe is filled with struggle and persecution. Although Roma slavery was abolished in 1856, Roma have long been denied social and human rights. The Nazis exterminated 500,000 Roma as part of their genocide. Even today, Roma communities experience stigmatization and ethnic-based negative prejudices, and often lack culturally appropriate education and healthcare services.4 Although Roma historically maintained nomadic lifestyles, most Roma now live in permanent communities.

Three to 5 times more Roma than majority population members live below poverty lines of eastern European countries.5,6 Health indicators among Roma have significantly deteriorated during the past decade,1 and infectious diseases such as tuberculosis and hepatitis are widespread. Roma receive lower-quality health care as a result of discrimination and their poverty.7 Across the region, Roma unemployment rates are 70% to 90%,5 and employed Roma typically do menial work. In Bulgaria, only 10% of Roma complete primary education compared with 72% of the majority population,6 and approximately 80% are illiterate.1 Bulgarian Roma life expectancy is approximately 10 years less than the country's majority population.8

Few studies have examined HIV/STD risk among Roma. However, almost 60% of Bulgarian Roma become sexually active before age 16, 54% of pregnancies are unplanned, abortion is common, and condom use rates vary from 4% to 21%.9 A recent survey in Sofia showed that the majority of Roma men do not regularly use condoms, often have anal intercourse with women, and frequently report same-sex anal intercourse.10

Gender determines Roma HIV risk sexual behavior patterns. Men have greater sexual freedom than women before and during marriage, often engaging in a wide range of unprotected practices with multiple partners.11 Men have much more power and control in relationships, with women expected to maintain virginity before marriage and then sexual exclusivity to their husbands. Because HIV risk in Roma communities is largely determined by men's behavior, understanding patterns and predictors of Roma men's sexual risk practices can assist in developing interventions to reduce HIV/STD risk in Roma communities.10,11

Distrustful of governmental institutions and other outsiders,12 Roma often rely on their own friends and social networks as sources of credible information.13 Networks are environments of shared peer norms and values that serve as sources for trusted advice and psychosocial support.14,15 Interventions for Roma and other marginalized populations are especially promising if carried out within naturally existing social structures.16 For these reasons, the current study recruited entire social networks of Roma men and specifically studied the role of their social networks in predicting HIV risk.

This study was carried out in Fakulteta, a Roma neighborhood in Sofia. With 20,000 residents, Fakulteta is among the 3 largest and oldest Roma neighborhoods in Bulgaria. The area is heavily overpopulated often without water, plumbing, and sewers. Many neighborhood dwellings are located near dung hills, and communicable disease outbreaks are common. Fakulteta's houses are usually very small and overcrowded by large families who frequently share a single room.17 The neighborhood harbors drug dealers and users, and both hidden and open prostitution. Only 8% of children in Fakulteta complete secondary education.17

This study's aims were to 1) establish prevalence of predominant STDs, 2) characterize HIV risk behavior practices, and 3) identify predictors of HIV risk behavior and STDs in a community sample of young Roma men's social networks.

Back to Top | Article Outline


Sample Recruitment and Demographic Characteristics

The study was carried out in 2003–2004. Using methods we have described elsewhere,18 ethnographers carried out field observations in Fakulteta community venues—disco clubs, cafés, pubs, and on streets—where social circles were present. “Social circles”19 were defined as socializing groups of at least 3 persons characterized by stability, close proximity of group members, and frequent conversational interactions with positive affect. Each circle was further observed to identify its center of attention (the “index”) whose ties with others in the circle appeared strongest and who appeared to be its social and emotional leader. Indices were then approached, provided with a brief study description, and asked to indicate the first names of the friends with whom they most liked to spend time, talk, and felt close. Each index and all named members of his social network participated in individual, confidential assessment interviews. The social networks of 54 indices (n = 296 participants) were interviewed, 90% of all named network members. All were males.*

Network mean size was 5.5 members (range, 3–9). Participants' mean age was 19.6 years (range, 16–33 years). A total of 68.9% of respondents were single/never married, 7.8% (n = 23) worked full-time, and 14.9% (n = 44) were students.

Back to Top | Article Outline

Sociometrically Identifying Network Leadership

We sought to empirically establish each social network's leader. To do so, participants were given a list of all other members of the same network and specified who they most, and least, preferred in five leadership areas (for example, discussion of important matters). Using sociometric analysis software,20 a “social status” indicator was calculated that ranked network members on their balance of positive and negative nominations received from others in the same network. The network member with the highest social status indicator score across leadership areas was designated as its leader.

Back to Top | Article Outline

HIV Risk Behavior Interviews With Network Members

Measures, adapted from our previous HIV behavioral epidemiology studies in Central Europe,10,16,20 were administered verbally in individual interviews lasting less than 1 hour.

Back to Top | Article Outline

Psychosocial Scales.

Five AIDS-related psychosocial characteristics were assessed. A 15-item scale measured knowledge and misconceptions about AIDS risk behavior and risk reduction steps (e.g., “If a man pulls out before orgasm, it protects from getting AIDS and venereal diseases”). Participants responded “true, “false,” or “don't know” to each statement (score range, 0–15). A 7-item scale measured perceptions about safer sex peer norms (e.g., “Condom use is accepted by my friends”). For each statement, respondents answered “yes,” “somewhat,” or “no” (scale range, 0–14; Cronbach's α = 0.71, current sample). A 10-item scale with the same format measured attitudes toward condom use and safer sex (e.g., “Using condoms interrupts the pleasure of sex” (score range, 0–20; Cronbach's α = 0.61). A 12-item scale with the same response options measured strength of risk reduction behavioral intentions (e.g., “A condom will be used if I have sexual intercourse with a casual partner”; score range, 0–24; Cronbach's α = 0.59). Finally, perceived risk reduction self-efficacy or self-confidence was assessed using a 9-item scale with the same response format (e.g., “I am sure that I can overcome my partner's objections to safer sex or condoms”; score range, 0–18; Cronbach's α = 0.53).

Back to Top | Article Outline

Lifetime, Past Year, and Past 3 Months Sexual Risk Background.

The assessment elicited information on respondents' number of male and female sexual partners, and giving or receiving money or valuables for sex in the past year and lifetime. Detailed partner-by-partner questions inquired about participants' specific sexual behavior practices with up to 5 male, and up to 5 female, of their most recent partners during the past 3 months. For female partners, participants reported how many times with each partner they had vaginal or anal intercourse and how many of these acts were condom-protected. Men were asked how many times with each male partner they had engaged in anal intercourse and how many acts were condom-protected. Participants with over 5 partners of a gender summarized their behavior with all additional partners.

Back to Top | Article Outline

Substance Use During the Past 3 Months.

Participants were asked on how many days in the past 3 months they injected any drug or used alcohol, marijuana, hallucinogens, ecstasy, heroin or opiates, cocaine, or other substances.

Back to Top | Article Outline

Sexually Transmitted Disease Specimen Collection and Analysis


Five-milliliter blood serum samples were tested by VDRL with positive cases confirmed by TPHA (Randox, U.K.) or enzyme-linked immunoassay (ELISA) for antitreponemal IgM/IgG antibodies (ETI-treponeme screen; Dia Sorin, Spain).

Back to Top | Article Outline

Chlamydia and Gonorrhea.

Detection of Chlamydia trachomatis (CT) and Neisseria gonorrhoeae (NG) was carried out by testing urine samples with polymerase chain reaction. After not urinating at least 4 hours, 10 to 15 mL of the first-catch urine was collected in sterile polypropylene tubes and stored at 4° to 8°C for up to 3 days. DNA was extracted using Amplicor CT/NG specimen preparation kits (Roche Molecular Systems, Inc.). Amplification and detection steps were carried out by Amplicor CT/NG amplification using detection kits with thermocycler GeneAmp 9,600 (Perkin Elmer) and a system for micro-ELISA (Labsystems).

Back to Top | Article Outline

Trichomonas vaginalis.

Detection was carried out from 10 to 15 mL of first-catch urine analyzed within 5 minutes of collection. Urine was centrifuged for 2 minutes at 2000 to 2500 revolutions, the separated sediment placed on a slide, covered with a glass, and observed through a light microscope magnifying ×40 to detect trichomonas.

Appropriate antibiotic treatment was provided to participants who had any infection.

Back to Top | Article Outline

Statistical Procedures

Descriptive analyses examined STD prevalence and sexual risk characteristics. Next, we examined the degree to which variations in sexual risk behavior were associated with social networks. General linear models (GLM) tested the univariate relationship (eta) between networks, entered as a random factor, and sexual risk outcome measures. A series of intraclass correlations (r), comparing leaders versus other network members on each sexual risk variable, estimated within-network consistency in risk behavior. To reduce skewness, sexual partners and unprotected intercourse occasion counts were transformed using the formula log10+ 1. Lastly, we examined the relative contributions of individual demographic background, substance use, and psychosocial risk-related characteristics and social network membership in predicting sexual risk behavior. The social network variable was a unique identifying code for an index and the members of his social network. Stepwise regression models (linear or logistic) were fit to the data for each sexual risk variable. All predictors achieving a significance level of at least P <0.10 in the regression analysis were then included as covariates in a GLM for the same sexual risk outcome. The social network identifier variable was entered as a random effect because indices were randomly selected such that networks within the study represent a random sample. Partial eta2 coefficients measured the variance in sexual risk explained by each predictor. This 2-stage approach limited the number of covariates in the final predictive models in relation to the sample size.

Back to Top | Article Outline


Prevalence of Sexually Transmitted Diseases

Among the 286 participants agreeing to be tested for STDs, 62 men (21.7%) had at least one disease. Twenty-five men (8.7%) had trichomonas and 23 (8.0%) had chlamydia. Thirteen men (4.5%) had gonorrhea and 10 (3.5%) had syphilis These prevalence levels were 1807 and 312 times higher than the 2004 Bulgarian national prevalence of the respective diseases (Bulgarian National Center for Health Information, Sofia, Bulgaria, personal communication, November 3, 2005).

Back to Top | Article Outline

High-Risk Sexual Practices Among Members of Roma Social Networks

Table 1 summarizes participants' high-risk sexual practices reported for lifetime, for the past year, and for the past 3 months. Although on average under 20 years old, Roma men in the sample reported a mean of nearly 27 lifetime partners (median, 13), including 7 (median, 4) in the past year and 3 (median, 2) in the past 3 months. Although almost all men had female partners and most self-identified as heterosexual, a high proportion of men also reported sex with other men in their lifetimes (59%), in the past year (35%), and during the past 3 months (19%). Multiple sexual partnerships were normative. Over 85% of men had more than one partner during the past year and almost 60% did so in the past 3 months. Twenty-six percent of men (n = 77) had a single main partner and 8% of men (n = 23) had one casual partner during the past 3 months. Much more common were patterns in which men had multiple main partners (38%, n = 113) or had multiple casual partners (21%, n = 62) in the past 3 months.

As Table 1 further shows, the prevalence and the frequency of unprotected vaginal and anal intercourse with female partners were also high, including with casual and with multiple partners. Most men had both unprotected vaginal and unprotected heterosexual anal intercourse during the past 3 months, often with nonmain and multiple partners. Fewer men (13%) had unprotected anal intercourse with male partners, although these acts were also frequent during the past 3 months (mean, 14.4; median, 6).

Almost 38% (n = 112) of men were paid for sex in their lifetimes, more than half of them (25.7%, n = 76) in the past year. Over 29% of participants (n = 86) paid someone for sex in their lifetimes, and 20.9% (n = 62) paid for sex in the past year. Only 3.7% of men (n = 11) injected drugs during the past 3 months.

Back to Top | Article Outline

Within- and Between-Network Variations in Sexual Risk Behaviors

Table 2 presents the results of correlational analyses performed to determine the extent to which variability in sexual risk behaviors is associated with social network factors. The first set of analyses assessed differences in the levels of risk behavior found between networks. Significant (eta) coefficients resulted in 8 of the 9 associations presented in the left column of Table 2. There were significant differences among networks in total number of sexual partners during the past year and past 3 months; having had unprotected vaginal intercourse, anal intercourse, or either during the past 3 months; and in the number of unprotected vaginal intercourse acts, anal intercourse acts, and number of either during the past 3 months.

A series of intraclass correlations (Table 2, righthand column) then examined patterns of within-group consistency by comparing the sexual risk characteristics of each social network's of members with the risk characteristics of the network's leader. These coefficients were much more modest. However, significant correlations were found between network leader and network member behavior for number of unprotected vaginal sex acts and number of unprotected vaginal and anal intercourse acts.

Back to Top | Article Outline

Predictors of Sexual HIV Risk Behavior

Table 3 shows results of stepwise regression models predicting whether participants engaged in unprotected vaginal intercourse, anal intercourse, or either during the past 3 months as well as their frequency of engaging in each type of unprotected intercourse act. The social network to which an individual belonged explained from 23% to 27% of the variance in whether individuals had any unprotected vaginal intercourse, anal intercourse, or either during the past 3 months and also in their frequency of unprotected vaginal or anal intercourse acts (P = 0.005 to P = 0.04). Participant age, years of education, safer sex peer norms, AIDS-related attitudes, intentions, and self-efficacy were significant but much weaker predictors of risk behavior outcomes than social network, with partial eta2 coefficients accounting for only 1% to 4% of the variance in each outcome.

Back to Top | Article Outline


The current study shows the high prevalence of HIV/STD risk behavior among social networks of young men in a Bulgarian Roma community. As far as we are aware, these are also the first published data documenting STD prevalence in a community sample of Roma men. The large majority of young men in the sample recently engaged in unprotected vaginal or anal intercourse, usually with multiple partners and in nonexclusive relationships. Rates of condom use were low. To great extent, multiple partnerships and low condom use are the current sexual behavioral norms in the Roma community studied. Approximately one third of men recently had sex not only with women but also with men, and their anal intercourse acts were usually unprotected. Exchange of sex for money or valuables was common. The study illustrates that one can identify and feasibly access high-risk members of the Roma community hidden to outsiders through the social networks to which these persons belong.

In most theoretical models, AIDS-related knowledge, norms, safer sex attitudes, intentions, and self-efficacy are all seen as determinants of an individual's risk behavior. However, the current study established that the social network to which an individual belongs is an even stronger predictor of risk behaviors, including number of sexual partners in the past year and the last 3 months, engaging in recent unprotected intercourse, and exchanging sex for money. Significant variations were found in the risk levels between social networks, and risk behavior levels among members of a given network were correlated. In addition to showing that levels of risk vary across social networks, multivariate analyses established relationships between the risk of network leaders and members about frequency of unprotected intercourse acts. Ellen21 has recently also observed that characteristics of one's sexual and social networks may either contribute to STD risk or protect against it. One avenue for preventive intervention is to change the social norms within high-risk social networks. Another is to assist persons in withdrawing from high-risk networks and instead associating with persons sharing safer norms.21

Our study showed that social networks predict the levels and types of risk behavior in which individuals engage. Social networks provide the basis and social environment for establishing, maintaining, and modifying norms shared by peers, including safer sex norms. We do not know if individuals with similar risk levels tend to bond with one another by establishing or entering the same network or rather if network influences unify members who originally had varying risk levels. Nevertheless, prevention approaches originating from, or endorsed within, the Roma community are especially promising because Roma generally distrust outsiders who are not from their community but trust other members of their immediate social networks.

This phenomenon is probably not unique to Roma. Other community populations vulnerable to HIV/AIDS are disenfranchised and may not trust officials or outsiders. Interventions may have stronger effects if directed toward entire social networks than individuals in isolation.22,23 For example, a recent intervention study conducted by our team with young men who have sex with men in Bulgaria and Russia24 used the same network recruitment strategy as described in the current study. Sociometrically identified network leaders attended 9 intervention training sessions and were taught and encouraged to establish new safer sex norms within their social networks, overcome network members' negative attitudes to condoms, help members' plan steps toward safer sexual behavior, and problem-solve ways to change high-risk behavior. Risk assessment interviews at baseline between 1-year follow up revealed reductions in high-risk behavior among intervention networks relative to a comparison group in which network members received individual risk reduction counseling alone.24 Although network-level interventions are promising, structural interventions to address underlying problems related to lack of employment, housing, education, and health care are also urgently needed in eastern European Roma communities.

Study findings from a single Roma neighborhood may not generalize to other Roma communities. The study did not use HIV testing. Risk behavior questions are sensitive, which may have caused underreporting of stigmatized sexual or substance use behaviors. The sample consisted of Roma male social networks, and further research is needed to examine the HIV risk of Roma females. STD testing may have underrepresented actual rates because gonorrhea and chlamydia were diagnosed only from urine specimens. This would have missed oral and rectal infections.

Young Roma men are at alarming risk for HIV infection and STDs. An HIV/AIDS epidemic may rapidly affect Roma communities unless prevention programs are implemented. Additional research is needed to study how new information or norm spreads within a network, identify mechanisms by which social networks influence their members, study patterns of mixing or overlap between social and sexual networks, and determine the most efficacious strategies of using networks in the HIV prevention arena.

Back to Top | Article Outline


1. Social Assessment of Roma and HIV/AIDS in Central and Eastern Europe. Bucharest: UNDP Romania Country Office and Center for Health Polices and Services, 2003.
2. World Bank. HIV/AIDS in Southeastern Europe: Case Studies From Bulgaria, Croatia and Romania. Washington, DC: World Bank, 2003.
3. Liégeois J-P. Roma, Gypsies, and Travelers. Strasbourg: Council of Europe, 1994:34.
4. World Bank. Roma and the Transition in Central and Eastern Europe: Trends and Challenges. Washington, DC: World Bank, 2000.
5. National Human Development Report. Geneva: UNDP, 2003.
6. Faces of Poverty, Faces of Hope. Vulnerability Profiles for Decade of Roma Inclusion Countries. Bratislava: UNDP, 2005.
7. Puporka L, Zádori Z. The Health Status of Roma in Hungary. World Bank Regional Office Hungary. NGO Studies 1999:2.
8. Survey on Health Status of the Bulgarian Population. Sofia: National Institute of Statistics, March 2001. Also available at:
9. ASSA-M. Reproductive Behaviour, Family Planning, Contraceptive Usage (empirical sociological survey). Sofia: ASSA-M, 1996.
10. Kabakchieva E, Amirkhanian YA, Kelly JA, et al. High level of sexual HIV/STD risk behaviour among Roma (Gypsy) men in Bulgaria: Patterns and predictors of risk in a representative community sample. Int J STD AIDS 2002; 13:184–191.
11. Kelly JA, Amirkhanian YA, Kabakchieva E, et al. Gender roles and HIV sexual risk vulnerability of Roma (Gypsies) men and women in Bulgaria and Hungary: An ethnographic study. AIDS Care 2004; 16:231–245.
12. Tomova I. Gypsies in the Transition Period. Sofia: IMIR Publications, 1995:81–83.
13. UN Social Development Unit in Bulgaria. Communication in Roma Communities. Sofia: UN Social Development Unit in Bulgaria, 2001.
14. Lee NH. The Search for an Abortionist. Chicago: University of Chicago Press, 1969.
15. Zapka JG, Stoddard AM, McCusker J. Social network, support and influence: Relationship with drug use and protective AIDS behavior. AIDS Educ Prev 1993; 5:352–366.
16. Amirkhanian YA, Kelly JA, Kabakchieva E, et al. Evaluation of a social network HIV prevention intervention program for young men who have sex with men in Russia and Bulgaria. AIDS Educ Prev 2003; 15:205–220.
17. National Human Development Report Bulgaria 1999. Sofia: UNDP, 1999:36.
18. Amirkhanian YA, Kelly JA, McAuliffe TL. Identifying, recruiting, and assessing social networks at high risk for HIV/AIDS: Methodology, practice, and a case study in St. Petersburg, Russia. AIDS Care 2005;17:58–75.
19. Kadushin C. The friends and supporters of psychotherapy: On the social circles of everyday life. Am Soc Rev 1996;31:786–802.
20. Sociometry Pro [computer program]. Version 2.1. Moscow: LeDis Group, 2003. Available at: Accessed November 24, 2005.
21. Ellen J. The next generation of HIV prevention for adolescent females in the United States: Linking behavioral and epidemiologic sciences to reduce incidence of HIV. J Urban Health 2003; 80(suppl 3):40–49.
22. Latkin CA, Sherman S, Knowlton A. HIV prevention among drug users: Outcome of a network-oriented peer outreach intervention. Health Psychol 2003; 22:332–339.
23. Broadhead RS, Heckathorn DD, Weakliem DL, et al. Harnessing peer networks as an instrument for AIDS prevention: Results from a peer-driven intervention. Public Health Rep 1998; 113(suppl 1):42–57.
24. Amirkhanian YA, Kelly JA, Kabakchieva E, et al. A randomized, controlled trial of a social network HIV prevention intervention with young men who have sex with men in Russia and Bulgaria. AIDS 2005; 19:1897–1905.

*All identified social circles had male indices. Females were not excluded from the study, but only one female was identified as a social network member of male indices. Cited Here...

†Procedures used in field observations, index identification, and sociometric analysis calculations to determine leaders of the social networks are described more fully elsewhere.16,18 Cited Here...

© Copyright 2006 American Sexually Transmitted Diseases Association