THE LIFE CIRCUMSTANCES OF street-sex workers in urban centers in the United States make them highly vulnerable to HIV infection. Many of these sex workers live in acute poverty, are homeless, drug and alcohol dependent, and have suffered multiple life traumas, including childhood abuse.1–4 Poor mental health and issues of drug dependence may weaken their motivation and ability to practice safer sex. Economic, social, and gender inequalities make it difficult for women to get their regular partners to use male condoms.5–8 Although most commercial sex workers are aware of HIV risks and report using male condoms with their commercial partners, many will engage in risky behavior if the customer offers to compensate them with more money or drugs. This is especially true if business has been slow or if sex work is their main source of income and they are desperate for money or drugs.9 To provide an additional source of protection for these women, greater access to alternative female-initiated barrier methods, such as the female condom, is needed.
The female condom is an alternative barrier method of HIV and sexually transmitted disease (STD) prevention that is acceptable to sex workers. Results from research in Uganda, Senegal, Cote d'Ivoire, Malawi, Zimbabwe, Indonesia, Papua New Guinea, Singapore, Thailand, France, Costa Rica, Mexico, and the United States show that sex workers report high levels of female condom acceptability.10–11 Female condom use is associated with an increase in protected sexual acts and a reduction of STDs among sex workers. Setiadi et al12 reported a 25% increase in condom use among sex workers after an intervention that included free access to male and female devices. Fontanet et al13 found that sex workers supplied with both male and female condoms reported a 34% lower incidence of STDs and a 25% reduction in unprotected sex acts compared with those supplied with male condoms alone. Mason et al14 found that among 147 sex workers randomly assigned to receive male and female condoms with education or male condoms only with education, the former group was more likely to be free of STDs for a longer period.
Most female condom studies to date have focused on women's attitudes toward and opinions of the female condom, or on its efficacy in preventing STDs or pregnancy.9,11–21 Few studies have addressed correlates of female condom use,10,22–24 and even fewer have addressed predictors of female condom use25,26; therefore, we know little about who is more likely to accept or reject and use or refuse to use the female condom. As with any new device, time and practice are needed to become comfortable with and proficient at its use. A critical first step to continued female condom use is availability and first-time use of the device. A clearer understanding of the predictors of initial female condom use is needed to target street sex workers most likely to try the device, and to develop effective HIV-prevention programs that may enhance a sex worker's ability to sustain use of this device. The purpose of this study was to explore the characteristics of street-sex workers (e.g., sociodemographic, drug-use, sexual risk, victimization history) that will predict their initial use of female condoms.
A convenience sample was recruited among street-sex workers receiving services from the Foundation for Research on Sexually Transmitted Diseases (FROST'D), a nonprofit organization based in New York City. Since 1989, FROST'D has been providing HIV counseling and testing, condom, and bleach kit distribution services to commercial sex workers in New York City. FROST'D operates an outreach program through a mobile van that makes weekly stops at identified sex work “strolls.”
Two trained female interviewers approached street-sex workers attending the mobile van for services, and invited them to participate in the project. After a brief description of the project, interviewers clarified any questions or concerns expressed by the women and obtained informed consent. Of approximately 350 women approached to participate in the study, 113 (32%) agreed to participate. Face-to-face, structured baseline interviews were administered.
A 30-minute baseline interview covered sociodemographic characteristics, drug-use history, sexual risk behavior with both regular and commercial partners, and victimization history. After the interview, the interviewer gave a brief explanation and demonstration of proper female condom use, and gave each participant 10 female condoms, asking her to return in 2 weeks for a similarly structured follow-up interview. To assure confidentiality, women were assigned an identification number and given an appointment card for the follow-up visit.
Predictor variables were selected based on earlier research,11,23 and included sociodemographic, drug-use, sexual risk behavior, and victimization variables.
Sociodemographic characteristics. Sociodemographic characteristics included continuous variables (age, years of education), categorical variables (ethnicity, marital status), and dichotomous variables (current homelessness, living with children, living with anyone having drug or alcohol problem, trading sex as main source of income, incarcerated during the past year, and having a regular sexual partner).
Drug history. Participants were asked whether they currently used a variety of drugs, including marijuana, cocaine, intravenous or nasal heroin, crack, alcohol, and methadone.
Sexual risk behavior. Sexual risk variables included continuous variables (e.g., the frequency of commercial sex exchanges), a categorical variable of HIV status (positive, worried, not worried, no answer), and dichotomous variables (e.g., always using a male condom with commercial partners, having heard of the female condom before this project).
Victimization. Victimization variables included physical or sexual abuse by commercial partners during the past year and childhood physical or sexual abuse. Physical abuse was defined as any instance in which a woman suffered painful physical injuries at the hands of someone else. Sexual abuse was defined as any instance in which a woman was forced to have intercourse.
Dependent variables. The dependent variables included (1) female condom use for any reason (e.g., regular or commercial sexual partners, for practice on themselves without having sex); and (2) female condom use specifically with commercial sexual partners. For each dependent variable, participants were asked whether they had tried the device at least once, and the total number of female condoms used out of the 10 initially distributed.
Given the highly exploratory nature of the study, stepwise multiple logistic regressions (backward elimination) were used to build models that best fit the observed data. Some criticize this model-building method27 because it capitalizes on random variations in the data and produces results that are difficult to replicate. However, others28 recommend the stepwise procedure as a useful tool for exploratory research in which “the phenomenon is so new or so little studied that existing ‘theory’ amounts to little more than empirically unsupported hunches about explanations for the phenomenon.”29 At the time of this study, there was little empirical guidance for a theoretically driven set of female condom use predictors.
Two liberal significance criteria for elimination (P < 0.10 and P < 0.20) were used to avoid loss of a potentially significant predictor. The conventional criterion (5%) for statistical significance may be too strict, and often ends up eliminating important predictors from the model. Bendel and Afifi30 recommend that the statistical significance criterion be set in a range from 15% to 20%. Although this results in an increase in Type I error and a decrease in Type II error, the exploratory nature of this research called for placing greater emphasis on finding prospective predictors rather than the elimination of suspicious ones.
Description of Variables
Of the 113 original participants, 101 returned for the second interview, a follow-up rate of 89%. Of these 101 cases, 96 had no missing values for the dependent and predictor variables, and were included in the analysis.
Tables 1 and 2 describe participant and female condom-use characteristics. The majority of women were single, African American, and crack users. One third of the women had histories of either childhood abuse or partner abuse. Two fifths of the women had heard of the female condom before the baseline interview, and about 60% of the women reported always using male condoms with commercial partners. Sixty-six women (68.7%) tried at least 1 of the 10 female condoms for sex with regular or commercial partners or for practice on themselves without having sex. Fifty-seven women (59.4%) tried at least one female condom with a commercial partner.
Table 3 shows the models predicting the profiles of sex workers who tried the female condom for practice or for sex with regular and commercial partners. The different models are labeled as A1 (full model), A2 (20% criteria), and A3 (10% criteria); all models are statistically significant in terms of the model chi-square (P < 0.000). Examination of the change in the model chi-square indicates that A2 is better than A1 and A3. The change from A1 to A2 is not significant (P < 0.928); even after losing 19 degrees of freedom, A2 still fits the data almost as equally as A1. The difference between models A2 and A3 are almost significant (P < 0.056), indicating that 2 degrees of freedom (or two predictors) lost in A3 have contributed to improve the model fit to a significant degree in A2. In short, A2 with 8 degrees of freedom, fits the data as well as A1 with all predictors, whereas A2 fits better than A3 with 6 degrees of freedom. For this reason, A2 was chosen as the best model.
In a similar manner, Table 4 shows that model B2 is selected as the best model predicting the profiles of sex workers who tried the female condom for sex with commercial partners.
The diagnostic measures-condition index and tolerance statistics-did not detect any serious multicolinearity problems in these models. (Because of the unavailability of collinearity detective measures for the logistic regression models, we ran linear regressions for the same set of dependent variable and predictors used in the logistic regression models to obtain the condition index and tolerance statistics. Our concern was with the relationship among the predictors; the functional form of the model for the dependent variable is not relevant to the estimation of collinearity.29) The largest condition indexes for these models were well below the criteria of possible collinearity (> 10) suggested by Gujarati.31 The largest condition indexes were 5.7 for A2 and 7.7 for B2. The smallest tolerance statistics were above the criteria of potential collinearity (< 0.20) suggested by Menard.29 The smallest tolerance were 0.810 (physical or sexual abuse by commercial partner, A2) and 0.679 (current homelessness, B2).
Predictors of Female Condom Use
The odds ratios (ORs) of the selected predictors in the two best models are presented in Table 5. The following three variables had the strongest association with female condom use: (1) living with anyone with drug-alcohol problem; (2) having heard of the female condom; and (3) physical or sexual abuse by a commercial partner. The most striking finding was that sex workers living with anyone having a drug or alcohol problem were more likely to try the female condoms in both models. Sex workers who had heard of the female condom before this study were more likely to try the female condom than those who learned of the device for the first time. This positive association was statistically significant (P < 0.05) for both models. Current physical or sexual abuse by a commercial partner was negatively associated with female condom use in both models (P < 0.01).
Other predictors demonstrated weaker but almost significant associations with the dependent variables. Being homeless was positively related to female condom use with commercial partners (B2) but not to overall use. Childhood physical or sexual abuse was positively associated with female condom use with commercial partners. Married sex workers were less likely than single sex workers to try the device. This negative association was especially strong among women trying the female condom for any reason (OR 0.03; CI = .00, 0.31).
Overall, drug-use variables did not demonstrate strong independent effects on the dependent variables; however, cocaine showed a positive relationship to female condom use with commercial sex partners (OR 9.01; CI = 1.35, 60.20).
Having a regular sexual partner was the only variable that demonstrated contrary effects on the two dependent variables. Its effect was positive for female condom use for practice and for sex (A2; OR = 4.14, CI = 0.98, 17.61), but negative for its use with commercial sex partners (B2; OR = 0.45, ci = 0.13, 1.52). The positive result in A2 may simply indicate that, assuming an equal number of commercial partners, a woman with a regular partner has more opportunity to try a new device.
An intriguing finding was the negative association between having a regular sexual partner and using the female condom with commercial partners (B2). Although this variable is not significant in B2, it became almost significant when we repeated the stepwise linear regression analysis using the number of female condoms tried with commercial partners as the dependent variable (b = −1.0, SE = 0.58). (We conducted stepwise multiple linear regressions using the number of female condoms tried by sex workers as the dependent variable. Results were essentially identical to those of logistic regressions, and are available from the authors upon request.)
Why are women with a regular partner less likely to try the female condom when they should, perhaps, have an added incentive to protect their regular partner from HIV or STD exposure during sex exchange? Two possible explanations were tested. First, it may be that sex workers with a regular partner used most of the 10 female condoms provided with their regular partner, leaving fewer for use with commercial partners; however, a review of the data indicated that only five sex workers with a regular partner used all 10 female condoms provided. Second, sex workers with a regular sexual partner tended to always use male condoms with their commercial partners to prevent the transmission of STDs to their regular partner. If this is the case, these sex workers had no incentive or need to try the female condom with commercial partners.
To explore this explanation, an interaction term of having a regular sexual partner and always using the male condom with commercial partners was created, and the stepwise analysis was repeated. The result was that in B2 and the linear regression model, the interaction term became significant at the P = .05 (B2; OR = 0.27, CI 0.08-0.97; linear regression model, coefficient = −1.56, SE = 0.65). At the same time, both original variables lost significance at the P = 0.10 level in two models, which support this explanation (i.e., a woman who always uses male condoms with commercial partners and who has a regular sexual partner is less likely to try the female condom than other women). We need to test this relationship using more cases and other samples.
Multivariate regression analyses suggest that female condom distribution encouraged women exchanging street sex who may be most vulnerable to STD and HIV infection to try female condoms with commercial partners. Sex workers who experienced childhood physical or sexual abuse were more likely to try the female condoms than those without victimization histories. Homeless sex workers were more likely to use female condoms with commercial partners than nonhomeless sex workers. Sex workers using cocaine were more likely than noncocaine users to try the female condoms, especially with commercial partners. These findings emphasize the need to develop STD and HIV prevention interventions that take into account the complex circumstances of street-sex workers in urban centers that make them highly vulnerable to STD and HIV infection. The addition of the female condom as an alternative barrier device for sex workers may broaden their ability to protect themselves from STDs and HIV.
Another interesting finding is that female condoms were more likely to be tried by sex workers living with persons having a drug or alcohol problem (n = 35) than sex workers not living with such persons. Five of 35 sex workers reported that this drug- or alcohol-using individual was a regular sexual partner or spouse; 11 reported that this person was a parent or other relative; 11 reported that this person was a roommate or nonsexual partner, and 8 did not specify the relationship. It may be that living in a household that includes members with drug or alcohol problems reflect an environment of high sexual or drug-risk behavior, and by recognizing this, sex workers may have a heightened awareness of the need for protection and thus respond favorably to female condom use. This association needs to be explored and further clarified in future female condom studies using qualitative and quantitative approaches.
These findings-that childhood abuse, homelessness, cocaine use, and living with anyone with a drug or alcohol problem were related to initial female condom use-suggest the utility of the female condom in providing an alternative barrier method of protection from STDs and HIV among street-sex workers.
Sex workers who reported current physical or sexual abuse by a commercial partner were less likely than those who were not abused to try the female condom. This finding is important, because early promotional messages indicated that the device did not require male cooperation. However, as a female-initiated device, it still requires male involvement and willingness to advance its use. Sex workers may put themselves at risk for physical or sexual abuse if they attempt to use the device without informing their partners. As with the male condom, negotiation between partners is essential; program developers who are concerned with promotion of the female condom must remember that this is not a device that women are able to use without the cooperation of their partners. Additionally, more innovative programs integrating violence prevention and treatment issues targeting this population are needed. Intervention programs targeting female condom use with sex workers should include referrals to shelters, violence prevention and treatment programs, female condom introduction and negotiation skills, or, when appropriate, use of the device with regular sex partners or commercial partners in a couple context.
Inconsistent with one previous study,23 women who had heard of the female condom before the study were more likely to try it than those who had not. Exposure to the female condom may reduce barriers to trying an unfamiliar protective method, implying the need for programmers to enhance awareness and availability of the device to sex workers.
Women with regular sexual partners were more likely to try the device overall, but were were less likely to use the device with commercial partners. The greater likelihood to use the device overall may be a matter of convenience; having two types of partners, regular and commercial, makes it more likely that a woman has the opportunity to try the device, while a woman without a regular partner can only try it with commercial partners. Assuming an equal number of commercial partners, a woman with a regular partner has more opportunity to try a new device. Subanalyses indicate that the negative association between having a regular partner and device use with commercial partners may be because women with regular partners who always use male condoms with commercial partners perceive no need to use an alternative barrier method. This finding is consistent with those of Latka, Gollub, French, and Stein,32 demonstrating that initiating female condom use may not replace male condom use among regular male condom users.
Consistent with at least two previous studies of female condom use among women at high STD and HIV risk,23,25 married sex workers are less likely than single sex workers to try the female condom. This result should be interpreted with caution, because the number of married women who participated in the study is small (n = 10). Perceptions that regular partners are less risky than commercial partners may explain the lower likelihood of female condom use among married sex workers. Given the increasing incidence of HIV infection among monogamous heterosexual women, this misperception needs to be addressed in programs targeting this population.
This study has several limitations that may reduce the generalizability of the findings to other sex workers living in poor urban neighborhoods. First, the results of the stepwise regression analyses must be taken as tentative and inconclusive. These are data-driven analyses capitalizing on random variations in the data. Further development and testing of theory may be based on these results but also require replication with other data. Second, this study's sample size is small, and subjects were not randomly recruited. Thus, the study may have included motivated sex workers who were ready to reduce their sexual risk behaviors. Third, the study focused on self-reported data. Clearly, biological markers would enhance the quality of the findings. Fourth, the findings at follow-up may be influenced by the subjects' desire to please the interviewers or by incentives received at the follow-up interview. Finally, the measurements used in this study have not been tested for psychometric properties. Research on the acceptability of the female condom among sex workers and other female groups is in its early stages. Studies with large sample sizes using random selection procedures are essential to knowledge building in this area.
This study is the first to examine predictors of female condom use among urban sex workers in the United States. Implications for intervention development include the need to develop innovative programs provided on the street (e.g., through peers) that can access homeless, drug-involved sex workers in the most at-risk environments. Sex workers are the persons best able to inform effective prevention interventions, and should be consulted in the process.33 Information about the female condom should be heavily promoted in areas of sex-work strolls, and issues of accessibility and ease of use among sex workers should be addressed. Skill building, including modeling and practice with the female condom, and communication and negotiation skills, which facilitate female condom use, also need to be incorporated into prevention programs targeting this population. Finally, prevention programs promoting female condoms may be most effective if integrated into drug treatment, housing, violence prevention, and other critical areas of needed services among this high-risk population.
1. Alegria M, Vera M, Freeman DH, et al. HIV infection, risk behaviors, and depressive symptoms among Puerto Rican sex workers. Am J Public Health 1992; 84:2000–2002.
2. Sex work, AIDS and the law: Hearings Before the National Commission on AIDS. Washington, DC, 1992 (testimony of Priscilla Alexander, Director of Research, Foundation for Research on Sexually Transmitted Diseases)
3. El-Bassel N, Schilling RF, Irwin KL, et al. Sex trading and psychological distress among women recruited from the streets of Harlem. Am J Public Health 1997; 87:66–70.
4. Weiner A. Understanding the social needs of streetwalking prostitutes. Soc Work 1996; 41:97–105.
5. Amaro H. Love, sex and power: considering women's realities in HIV prevention. Am Psychol 1995; 50:437–447.
6. Caravano K. More than mothers and whores: redefining the AIDS prevention needs of women. Int J Health Serv 1991; 21:131–142.
7. Heise, L, Elias, C. Transforming AIDS prevention to meet women's needs: a focus on developing countries. Soc Sci Med 1995; 40:931–943.
8. Rao Gupta G, Weiss E. Women and AIDS: developing a healthy strategy. Washington, DC: International Center for Research on Women, 1993. ICRW Policy Series No. 1.
9. Dorfman LE, Derish PE, Cohen JB. Hey girlfriend: an evaluation of AIDS prevention among women in the sex industry. Health Educ Q 1992; 19:25–40.
10. UNDP/UNFA/WHO/World Bank Special Programme of Research, Development, and Research Training in Human Reproduction. The female condom: A review. Geneva: World Health Organization, 1997.
11. Witte S, El-Bassel N, Wada T, Gray O, Wallace J. Acceptability of the female condom among women exchanging street sex in New York City. Int J STD AIDS 1999; 10:162–168.
12. Setiadi BN, Jatiputra I, Santoso G. Sexual negotiation: the empowerment of women and the female condom. Findings cited in: UNDP/UNFA/WHO/World Bank Special Programme of Research, Development, and Research Training in Human Reproduction. The female condom: A review. Geneva: World Health Organization, 1997.
13. Fontanet AL, Saba J, Chandelying V, et al. Protection against sexually transmitted diseases by granting sex workers in Thailand the choice of using the male or female condom: results from a randomized controlled trial. AIDS 1998; 12:1851–1859.
14. Mason P, Ray S, Ndowa F, et al. Sexually transmitted diseases and HIV in commercial sex workers supplied with female and/or male condom. Findings cited in: UNDP/UNFA/WHO/World Bank Special Programme of Research, Development, and Research Training in Human Reproduction. The female condom: A review. Geneva: World Health Orzanization, 1997.
15. Farr G, Gabelnick H, Sturgen K, Dorflinger L. Contraceptive efficacy and acceptability of the female condom. Am J Public Health 1994; 84;1960–1964.
16. Soper DE, Shoupe D, Shangold GA, Shangold MM, Gutmann J, Mercer L. Prevention of vaginal trichomoniasis by compliant use of the female condom. Sex Transm Dis 1993; 20:137–139
17. Trussell J. Comparative contraceptive efficacy of the female condom and other barrier methods. Fam Plann Perspect 1994; 26:66–72.
18. Ford N, Mathie E. The acceptability and experience of the female condom femidom among family planning clinic attenders. Br J Fam Plann 1993; 19:187–192.
19. Shervington D. Acceptance of the female condom among low-income African-American women. J Natl Med Assoc 1993; 85:341–347
20. Gollub EL, Stein Z, El-Sadr W. Short-term acceptability of the female condom among staff and patients at a New York City Hospital. Fam Plann Perspec 1995; 27:155–158.
21. Gil V. The new female condom: attitudes and opinions of low-income Puerto Rican women at risk for HIV/AIDS. Qual Health Res
22. El-Bassel N, Krishnan S, Witte S, Schilling RF, Catan V, Poulin S. Correlates of intention to use the female condom among women on methadone. Womens Health Issues 1998; 8:112–122.
23. Sly DF, Harrison D, Eberstein IW, et al. Factors associated with use of the female condom. Fam Plann Perspec 1997; 29:181–184.
24. Musaba E, Morrison CS, Sunkutu MR, Wong EL. Long-term use of the female condom among couples at high risk of human immunodeficiency virus infection in Zambia. Sex Transm Dis 1998; 25:260–264.
25. Surratt HL, Wechsberg WM, Cottler LB, Leukefeld CG, Klein H, Desmond DP. Acceptability of the female condom among women at risk for HIV infection. Am Behav Sci 1998; 41:1157–1170.
26. Cabral R, Pulley L, Artz L, et al. US women at risk of HIV/STD: who wants and who uses female condoms. Paper presented at: the 12th International Conference on AIDS; July, 1998. Geneva, Switzerland.
27. Studenmund AH, Cassidy HJ. Using Econometrics: A Practical Guide. Boston: Little, Brown & Co., 1995
28. Hosmer DW, Lemeshow S. Applied Logistic Regression. New York: John Wiley & Sons, 1989.
29. Menard SW. Applied Logistic Regression Analysis. Thousand Oaks, CA: Sage Publications, 1995.
30. Bendel RB, Affifi AA. Comparison of stopping rules in forward regression. J Am Stat Assoc 1977; 72:46–53.
31. Gujarati D. Basic Econometrics. 2nd ed. New York: McGraw-Hill, 1988.
32. Latka M, Gollub EL, French P, Stein Z. Male and female condom use among women after full options counseling for STD prevention: Results from an STD clinic sample. Sex Transm Dis 2000 (in press).
© Copyright 2000 American Sexually Transmitted Diseases Association
33. El-Bassel N, Witte S. Designing effective HIV prevention strategies for female street sex workers. AIDS Patient Care STD 1998; 12:599–603.