It is well established that injecting drug use is a significant factor in the transmission of HIV. In Europe and the United States, injection drug users (IDUs) are the second largest group of reported AIDS cases.1 Beyond the risks of HIV transmission posed by sharing injecting equipment, however, there is growing concern about sexual transmission of HIV among IDUs and a growing interest in IDUs as a potential bridge population that may transmit HIV to noninjectors.
In early studies of IDU populations, most HIV infections seemed to be because of parenteral transmission,2-4 leading to the belief that sexual risks among IDUs were negligible. Later prospective studies of IDUs have gradually accumulated adequate follow-up to separate the effects of drug use vs sexual risk factors.5,6 In Baltimore, MD, Strathdee and Sherman6 examined risk factors for HIV seroconversion among IDUs participating in the ALIVE study7 and found that sexual behaviors were important risk factors for HIV seroconversion among both male and female IDUs. These results were replicated in analyses conducted on participants in another Baltimore-based cohort study of young injectors.6 Given these findings, it is important to focus research and prevention efforts at both injection and sexual behaviors of IDUs.
One limitation of the prior studies of sexual transmission of HIV among IDUs is that the noninjecting partners of IDUs were rarely assessed. This is of particular concern because a large proportion of the sexual encounters of drug injectors were with noninjectors. In one study of drug injectors in England, 62% of the respondent's primary and casual partners were noninjectors, and 68% of the respondents never used condoms with primary partners, although 66% had vaginal intercourse at least once a week.8 These findings suggest that noninjector sex partners, especially primary sex partners, of drug injectors seem to be at a significantly increased risk for HIV/AIDS through sexual transmission.9-12
Only one large-scale study in the United States specifically examined noninjectors who had IDUs as sex partners. Between 1988 and 1991, sex partners of IDUs were recruited through community outreach settings at more than 40 sites across the United States.13 The authors reported that condom use was relatively infrequent (<40%) among noninjectors with only 1 IDU partner. However, an unanswered question we attempted to address in the present article, through network analyses, was the issue of what factors are associated with noninjectors having IDUs as sex partners.
Network analysis focuses on individuals' ties or relationships rather than on individual attributes. Several studies have documented that network factors are associated with HIV risk behaviors, HIV infection, and other STIs. Several studies have found that social network structure is associated with HIV and STI transmission.14 Some network factors such as density and size are likely to be directly associated with disease transmission, whereas other network characteristics may indirectly lead to transmission. For example, in a study of syphilis transmission, Rosenburg et al15 found that friendship ties had as high levels of syphilis as compared with sex partners, and Youm and Laumann16 demonstrated that strength of friendship ties and number of partners were associated with a history of STIs. In network terminology, IDUs can be characterized as a bridge group who may transmit HIV from injectors into noninjectors through sexual transmission. The aim of this article was to examine factors associated with noninjectors having 1 or more injectors as sex partners. We anticipated that noninjectors with more sex partners would be more likely to have injecting sex partners.
The data used in this analysis were collected as a part of the Self-Help in Eliminating Life Threatening Diseases (SHIELD) project, a network-oriented experimental HIV prevention intervention. SHIELD was designed to empirically alter social processes such as norm formation and peer influence to reduce HIV risk behaviors. The study was approved by the Johns Hopkins School of Public Health Committee on Human Research in March 1997. The data presented here were collected at baseline interviews administered between August 1997 and March 1999.
SHIELD participants were recruited through targeted outreach. Recruitment areas in Baltimore City were identified through ethnographic observations, focus groups, and geocoding of the locations of drug-related arrests in Baltimore over a 3-year period. During street outreach in these specific recruitment locations, individuals that seem to be older than 18 years were approached. Potential participants were provided with a written description of the study and a contact telephone number at the study site. Those who contacted the study by telephone or came to the study site were briefly screened. SHIELD criteria for inclusion were the following: 18 years or older, at least weekly contact with drug users, willingness to conduct peer education among current drug users, and the ability to bring in at least 1 member of their social network for a baseline interview. Participants were paid up to $20 for completion of the interview, and indexes were paid up to $25 for bringing in their network members to be interviewed. There were 1070 indexes interviewed, as well as 567 members of their social networks. The after analyses were restricted to SHIELD participants who reported not having injected drugs in the 6 months before being interviewed.
All eligible participants were administered a detailed face-to-face interview on their sociodemographic background, patterns of drug use, HIV prevention and risk behaviors, and social networks.
Information about sex partners was ascertained through a personal support network survey in which participants were asked to give the first name and first letter of the last name or pseudonyms of individuals whom they had known for at least 1 month. A series of questions, described in the succeeding sections, were asked about the type of support provided, relationships between individuals, and network members' drug-using practices. Once the list of names was completed, participants provided demographic information about network members such as age, sex, and employment status. The individuals named in the survey formed the participants' social networks, which comprised smaller networks such as sex and drug support networks. This network inventory has demonstrated concurrent and predictive validity.17
Sex partners were identified by the after questions: "Of the people that you listed so far, who did you have sex with in the last 3 months? Who else did you have sex with in the last 3 months?"
Network Members Drug Use
Participants reported their network members, including sex partners' frequency of drug use and their drugs of choice in the past 6 months.
Index Drug Use
Participants were asked their use of an extensive list of drugs in the past 6 months, including injecting heroin and cocaine, smoking crack, and sniffing heroin, as well as their frequency of use. There were 863 individuals identifying themselves as not having injected in the past 6 months, and these individuals were selected as the study population for this analysis.
Social Context and Socioeconomic Status
Socioeconomic status was assessed with the after questions: "Have you received public assistance at any point in the past 10 years? Have you been unemployed for 1 or more years in the past 10 years? Have you been homeless at any point in the past 10 years?" Participants also reported if they were currently "employed full time," "employed part time," "unemployed and seeking work," or "unemployed and not seeking work." To assess drug availability in the neighborhood, participants were asked, "Is the selling of drugs a problem in your neighborhood?" The response categories were "not a problem," "somewhat of a problem," and "a big problem."
The responses to the question about sex partners and network member drug use were used to construct a dependent variable called "Number of injecting sex partners" that was then dichotomized, based on the distribution, into 2 categories: 0 and 1 or more injecting sex partners in the network.
Study Design and Analyses
The first series of analyses were descriptive; we conducted bivariate tests of association between our dependent variable and the independent variables. Our selection of which independent variables to test was based primarily on existing literature and an ecological model perspective, which focuses on social and physical environmental factors that promote risk behaviors. For candidate variables that were categorical, we used both the χ2 test and Fisher exact test; variables whose P value was ≤0.15 in these analyses were considered for inclusion in the final model. For continuous variables, we obtained smoothed plots of the log odds of having 1 or more injecting sex partner against the candidate variable and visually observed whether there seemed to be any significant trend. If 2 or more variables were highly correlated (our benchmark being a correlation coefficient of 0.5), we included only 1 of these variables in our models to avoid the problems associated with autocorrelation; we used standard likelihood ratio tests to help determine which one contributed most to the regression model and was most deserving of inclusion.
The variables considered for inclusion in multivariate regression models are listed and described in Table 1. Although included in our descriptive analyses, noninjectors who reported having zero sexual encounters in the 90 days before their interview were not included in our multivariate analyses. Our multivariate analyses were therefore focused specifically on sexually active noninjectors, that is, "Among sexually active noninjectors, what factors were associated with having 1 or more injecting sex partners?"
Furthermore, because of the skewed distribution of the variable representing the number of sexual encounters in the past 90 days, we categorized the responses as 0, 1, 2, 3, and 4 or more. One candidate variable, total network size, was subjected to linear spline transformation with one cut point at 15 before being placed in our multivariate models. This cut point was based on plots of total network size vs the crude odds of having 1 or more sex partners; the plots were constructed using the LOWESS smoothing procedure in STATA (STATA Corp, College Station, TX).
We placed candidate variables in a single multiple logistic regression model and then retained variables that were significant at the 10% level. Once we trimmed down our multivariate model by removing variables that were no longer marginally or statistically significant, we then proceeded to test for interactions between each predictor variable remaining and sex, past injection drug-using status, and HIV/AIDS status, respectively, and in that order. Interaction terms whose coefficient was significant at the 10% level were retained in the final model.
Our sample overall consisted largely of individuals between the ages of 28 and 47 years (85%), single (68%), and who smoked tobacco (82%). Almost everyone (96%) had used some type of drug at some point in the past, and most (71%) had family members who had had a drug problem. A little more than half of our sample (57%) was male and completed high school (58%). Only 26% of our sample was employed, and 11% had been in prison in the prior 6 months. About half (48%) the sample lived on their own, and 30% had children younger than 17 years living with them.
Men were less likely than women to have 1 or more injecting sex partners (Table 2). Those who reported having received public assistance in the previous 6 months, having been homeless at any point in the past 10 years, and having been unemployed for 1 or more years in the past 10 years all were more likely to have 1 or more injecting sex partners. Those with a family member or friend who had died of AIDS were themselves much more likely to have 1 or more injecting sex partners (odds ratio [OR], 3.60; 95% confidence interval [CI], 0.98-13.28).
As seen in Table 2, relative to those who had only 1 sex partner in the past 90 days, those who had had sex with 2 and with 3 different individuals, respectively, had more than a 2-fold higher odds of having 1 or more injecting sex partners. Noninjectors with 1 injecting non-sex partner in their network had more than a 2-fold higher odds of having 1 or more injecting sex partners relative to those with no injecting non-sex partners in their network. Those with 2 or more injecting non-sex partners in their network had an almost 4-fold higher odds of having 1 or more injecting sex partners. Finally, there was a positive association between total network size and the odds of having 1 or more injecting sex partner, with each 1-person increase in total network size associated with a 5% higher odds of having 1 or more injecting sex partners.
We present 3 sets of multivariate models in Table 3. The first model represents our final multivariate model without any interaction terms. Noninjectors who reported being unemployed for more than a year in the past 10 years had an almost 3-fold higher odds of having 1 or more injecting sex partners (OR, 2.92; 95% CI, 1.26-6.77). Noninjectors with 1 or more injecting non-sex partners (relative to those with none) and those having sex with 2 or more different individuals in the past 90 days (relative to those with 1 sex partner) were also much more likely to have 1 or more injecting sex partners. There was a 24% increase in the odds of having 1 or more injecting sex partners, with each unit increase in the variable representing the proportion of women in the network (OR, 1.24; 95% CI, 1.05-1.46); there was an 18% decrease in the odds of having 1 or more injecting sex partners, with each unit decrease in the variable representing the proportion of those who snort heroin in the subject's network (OR, 0.82; 95% CI, 0.66-1.01). Finally, men had 42% lower odds of having 1 or more injecting sex partners relative to women (OR, 0.52; 95% CI, 0.33-1.03).
When we tested for interactions with sex, the only interaction that was significant was between sex and the proportion of women in the network. The association between the proportion of women in the network and the odds of having 1 or more injecting sex partners was stronger among men than among women. We then tested for interactions between each variable in the model and past injection drug-using status, that is, ever vs. never injected drugs. The only interaction that was statistically significant was that between past injection drug-using status and the number of different individuals the subject had sex with in the past 90 days. The association between having more than 1 sex partner and having 1 or more injecting sex partners was much stronger among individuals who were former injectors than among individuals who had never injected. Finally, we tested for interactions between each variable in our final model and HIV/AIDS status but found no significant interactions.
We found in this study that long-term unemployment, increasing network size, increasing proportions of women in the network, increasing numbers of injecting non-sex partners in the network, and increasing numbers of sex partners were all associated with an increased odds of having 1 or more injecting sex partners. We found, in turn, that the association between the proportion of women in the network and having 1 or more injecting sex partners was much stronger for men than for women and that the association between the number of sex partners and having 1 or more injecting sex partners was stronger for former injectors than for never injectors.
Although we had difficulty finding existing literature concerning long-term unemployment and sexual behavior among noninjectors specifically, there are studies that have highlighted the negative effect of long-term unemployment on health behaviors in general. Hammarstrom and Janlert 18 followed 1060 Swedish adolescents from their last year of high school until the age of 21 years and found, particularly among men, that long-term unemployment was associated with increased rates of unprotected sex (without the desire to conceive). Unprotected sex without the desire to conceive can be considered as sexual risk taking-in the same way that having sex with an IDU can be considered as sexual risk taking-and therefore, one can argue that the findings of our study that long-term unemployment increased the odds of having 1 or more injecting sex partners, in a very broad sense, is consistent with the findings of the study by Hammarstrom and Janlert.18
We observed that among male noninjectors, there was a positive association between the percentage of women in the network and the odds of having 1 or more injecting sex partners. However, among female noninjectors, there was no corresponding negative association. In our sample, and consistent with other reports,19,20 female participants were much more likely than male subjects to have had sex with only 1 person in the 90 days before being interviewed (data not shown). Because women are more likely, at a given point in time, to have only 1 sex partner, the odds of a female noninjector having an injecting sex partner would be less likely to vary as the percentage of men in her network increases (ie, as the percentage of women in her network decreases).
We found an association between the number of sexual partners and the odds of having 1 or more sexual partner, with the association being much stronger among former injectors than among never injectors. This result could partially be explained as a statistical artifact in that the greater the number of sex partners, the greater the probability that at least 1 of these partners will have a given characteristic-in this case, injecting drug use. Alternatively, having sex with more than 1 person in the prior 90 days could be an indicator of the propensity to interact with network members the former injectors affiliated with when they were actively injecting.
On the other hand, we found that the presence of 1 or more injecting non-sex partners in the network was associated with an increased odds of having 1 or more injecting sex partners in the network. The presence of injecting non-sex partners in the network may reflect the possibility that noninjectors who have injecting non-sexual network members either begin to affiliate with other injectors or that the IDU non-sex partner introduces them to injector sex partners. It is important to note however, on account of the cross-sectional nature of this study, that the reverse might be true-IDU sex partners of noninjectors could be introducing them to other IDUs who then become a part of the noninjector's network, hence the association between the presence of 1 or more injecting non-sex partners in the network and an increased odds of having 1 or more injecting sex partners.
Our results of an association between network size and having an injecting sex partner were also intriguing. In our multivariate analyses, we found that among those with network sizes below 15, there was a marginally significant 9% decrease in the odds of having 1 or more injecting sex partner with each 1-person increase in total network size (OR, 0.91; 95% CI, 0.81-1.00). However, among those with total network sizes above 15, there was an increase, albeit not statistically significant, of 28% in the odds of having 1 or more injecting sex partners with each 1-person increase in total network size. This suggests that, up to a certain point, larger network sizes may be providing noninjectors with more stable support systems and contributing to less risky behavior. These results also suggest, however, that after a certain point, increasingly larger network sizes may actually have a risk-enhancing effect. The idea that network characteristics can have both risk-enhancing and risk-decreasing properties has been demonstrated elsewhere. For example, Ennett et al21 reported in a study of 327 runaway or homeless youths in Washington, DC, that youths without a social network were significantly more likely to report current illicit drug use, multiple sex partners, and survival sex than youths with a network.
This study is not without limitations. One potential weakness was that all sex partners were not reported-it is likely that participants reported on more stable relationships. We also do not know exactly how accurate were the reports on the risk behaviors of network members. In summary, the study's generalizability was limited by the recruitment criteria, self-report data, and cross-sectional study design.
This study represents one of the first attempts to examine factors associated with noninjectors having 1 or more injecting sex partners. In summary, long-term unemployment, increasing network sizes above 15, increasing proportions of women in the network (among men), increasing numbers of sexual partners within the past 90 days (particularly among former injectors), and the presence of 1 or more injecting non-sex partners are all associated with and increased odds of having 1 or more injecting sex partners.
The authors thank our study participants, study staff, and Melissa Davey.
1. Ward JW, Duchin JS. The epidemiology of HIV and AIDS in the United States. AIDS Clin Rev
2. Des Jarlais DC, Friedman SR, Hopkins W. Risk reduction for the acquired immunodeficiency syndrome among intravenous drug users. Ann Intern Med
. November 1985;103(5):755-759.
3. Chaisson RE, Moss AR, Onishi R, et al. Human immunodeficiency virus infection in heterosexual intravenous drug users in San Francisco. Am J Public Health
. February 1987;77(2):169-172.
4. Schoenbaum EE, Hartel D, Selwyn PA, et al. Risk factors for human immunodeficiency virus infection in intravenous drug users. N Engl J Med
. September 28, 1989;321:874-879.
5. Kral AH, Bluthenthal RN, Lorvick J, et al. Sexual transmission of HIV-1 among injection drug users in San Francisco, USA: risk-factor analysis. Lancet
. May 5, 2001;357(9266):1397-1401.
6. Strathdee SA, Sherman SG. The role of sexual transmission of HIV infection among injection and non-injection drug users. J Urban Health
. December 2003;80(4 Suppl 3):iii7-iii14.
7. Vlahov D, Anthony JC, Munoz A, et al. The ALIVE study, a longitudinal study of HIV-1 infection in intravenous drug users: description of methods and characteristics of participants. NIDA Res Monogr
8. Rhodes T, Donoghoe M, Hunter G, et al. Sexual behaviour of drug injectors in London: implications for HIV transmission and HIV prevention. Addiction
. September 1994;89(9):1085-1096.
9. Kall K. The risk of HIV infection for noninjecting sex partners of injecting drug users in Stockholm. AIDS Educ Prev
. August 1994;6(4):351-364.
10. Pisani E, Dadun PK, Sucahya O, et al. Sexual behavior among injection drug users in 3 Indonesian cities carries a high potential for HIV spread to noninjectors. J Acquir Immune Defic Syndr
. December 1, 2003;34(4):403-406.
11. Klee H, Faugier J, Hayes C, et al. Sexual partners of injecting drug users: the risk of HIV infection. Br J Addict
. March 1990;85(3):413-418.
12. Fals-Stewart W, Birchler GR, Hoebbel C, et al. An examination of indirect risk of exposure to HIV among wives of substance-abusing men. Drug Alcohol Depend
. May 1, 2003;70(1):65-76.
13. Weissman G, Brown V, National AIDS Research Consortium. Drug use and sexual behaviors among sex partners of injecting-drug users-United States, 1988-1990. MMWR Morb Mortal Wkly Rep
. December 13, 1991;40(49):855-860.
14. Friedman SR, Neaigus A, Jose B, et al. Sociometric risk networks and risk for HIV infection. Am J Public Health
. August 1997;87(8):1289-1296.
15. Rosenberg D, Moseley K, Kahn R, et al. Networks of persons with syphilis and at risk for syphilis in Louisiana: evidence of core transmitters. Sex Transm Dis
. February 1999;26(2):108-114.
16. Youm Y, Laumann EO. Social network effects on the transmission of sexually transmitted diseases. Sex Transm Dis
. November 2002;29(11):689-697.
17. Latkin C, Mandell W, Vlahov D, et al. People and places: behavioral settings and personal network characteristics as correlates of needle sharing. J Acquir Immune Defic Syndr Hum Retrovirol
. November 1, 1996;13(3):273-280.
18. Hammarstrom A, Janlert U. Unemployment and sexual risk-taking among adolescents. Scand J Soc Med
. December 1997;25(4):266-270.
19. Forrest JD, Singh S. The sexual and reproductive behavior of American women, 1982-1988. Fam Plann Perspect
. September/October 1990;22(5):206-214.
20. Drory Y, Kravetz S, Weingarten M. Comparison of sexual activity of women and men after a first acute myocardial infarction. Am J Cardiol
. June 1, 2000;85(11):1283-1287.
21. Ennett ST, Bailey SL, Federman EB. Social network characteristics associated with risky behaviors among runaway and homeless youth. J Health Soc Behav
. March 1999;40(1):63-78.