Efficacy of an HIV Prevention Program Among Female Adolescents Experiencing Gender-Based Violence

Abstract

Objectives. We examined the efficacy of an HIV prevention intervention among African American female adolescents reporting a history of gender-based violence.

Methods. In this analysis of a subgroup of participants involved in a randomized controlled trial, consistent condom use, psychosocial mediators associated with HIV-preventive behaviors, and presence of sexually transmitted diseases were assessed at 6- and 12-month follow-ups. The intervention emphasized ethnic and gender pride, HIV knowledge, condom attitudes, healthy relationships, communication, and condom use skills.

Results. Relative to the comparison condition, participants randomized to the intervention reported using condoms more consistently, had fewer episodes of unprotected vaginal sex, engaged in a greater proportion of protected intercourse acts, were more likely to have used a condom during their most recent intercourse, were less likely to have a new sexual partner, were less likely to have a sexually transmitted disease, and demonstrated more proficient condom skills.

Conclusions. Given the substantial prevalence of gender-based violence among female adolescents and the associations observed between gender-based violence, HIV risk, and HIV infection, it is essential that HIV interventions involving young women address partner violence.

HIV prevention interventions have been developed for several at-risk adolescent populations, including homeless and HIV-positive adolescents, and these programs have been conducted at a range of venues such as schools and clinics.1^–4 However, 1 subgroup of adolescents at considerable risk for HIV—young women who have experienced gender-based violence—has been relatively neglected by HIV prevention programs.

Research has shown that the prevalence of gender-based violence among adolescents ranges from 9% to 39%5^–9 and that rates may be higher among African American female adolescents than among female adolescents from other racial/ethnic groups.5^,7 In addition, a seroepidemiological study of Job Corps applicants showed that HIV prevalence rates among African American female adolescents were significantly higher than rates among their White or Hispanic female peers (4.9, 0.7, and 0.6 per 1000, respectively) and exceeded those observed among White, Hispanic, and African American male adolescents (0.8, 1.5, and 3.2 per 1000, respectively).10

Several studies have shown that adolescents with sexually or physically violent partners engage in HIV risk-taking practices, including having multiple sexual partners, initiating sexual intercourse at an early age, and abusing drugs.11^–17 Moreover, studies have revealed associations between gender-based violence and HIV infection.18^–21 According to this body of research, gender-based violence may either operate indirectly, through social mechanisms (e.g., women are predisposed to engage in HIV risk taking because they perceive that they lack the power to negotiate and practice safe sex), or operate directly, through biological mechanisms (e.g., having nonvolitional intercourse with an HIV-infected partner). Given the relation between gender-based violence, HIV risk, and HIV infection, designing HIV prevention programs for this vulnerable population is critical.

In an earlier study, DiClemente et al.22 described the efficacy of an HIV prevention program, conducted among sexually experienced African American women, in reducing risky sexual behaviors, decreasing the incidence of sexually transmitted diseases (STDs), and enhancing psychosocial mediators (e.g., HIV prevention knowledge, condom use self-efficacy) associated with HIV-preventive behaviors. We examined the efficacy of the intervention among a subsample of participants in the same study who reported a history of gender-based violence.

METHODS

The study methods have been described in detail elsewhere.16 Briefly, from December 1996 through April 1999, recruiters screened 1130 African American female adolescents seeking services at community health agencies in Birmingham, Ala. Of these young women, 609 (53.9%) met the study’s eligibility criteria, which included being female and African American, being 14 to 18 years of age, reporting vaginal intercourse in the preceding 6 months, and providing written informed consent. Eighty-seven young women were eligible but did not participate, most of whom had conflicts with other activities or were not interested. Thus, 522 adolescents agreed to participate in the original study, completed baseline assessments, and were randomized to study conditions.

The present analyses focused on the subset of 146 participants who reported a history of gender-based violence at the baseline assessment. We categorized young women who had ever been coerced into having intercourse against their will by their boyfriend or who had been physically abused (i.e., they had been kicked, slapped, hit, or pushed or had had something thrown at them) by their boyfriend as having experienced gender-based violence.16

Procedures

We used a randomized controlled trial design. Participants were randomly assigned to either the 4-session HIV prevention intervention or a 4-session general health promotion condition. The HIV prevention intervention consisted of 4 interactive group sessions, each 4 hours in duration, conducted on consecutive Saturdays. Sessions included 10 to 12 participants on average, and they were implemented by a trained African American female health educator assisted by 2 African American female peer educators. Participants randomized to the general health promotion condition also attended 4 interactive group sessions (again, 4 hours in duration) taking place on consecutive Saturdays. These sessions focused on the importance of exercise and proper nutrition, provision of information on local venues for purchasing healthy foods, and the health consequences of a poor diet.

Two complementary theoretical frameworks, social cognitive theory23 and the theory of gender and power,24^,25 guided the design and implementation of the HIV prevention condition. Session 1 of the intervention emphasized ethnic and gender pride, self-actualization, and self-worth by discussing the joys and challenges of being a young African American woman, acknowledging the accomplishments of African American women, reading poetry written by African American women, and framing artwork created by African American female artists. The goals of session 2 were to increase awareness of strategies for reducing the risk of HIV and other STDs and to explore common myths regarding HIV prevention.

Session 3 built on the knowledge base established in session 2 to enhance adolescents’ condom use and communication skills and examined gender roles, partner influences, and societal myths that contribute to attitudes toward using condoms. Session 4 sought to promote attitudes toward having healthy relationships, define how imbalances of power and control could influence having a healthy dating or sexual relationship, provide information about community resources for adolescents in unhealthy relationships, and discuss adolescents’ choices in selecting dating partners. These activities were designed to enhance participants’ perception that they had greater control in their sexual relationships (see DiClemente et al.22 for further discussion of the intervention).

Data Collection

Data were collected at baseline and at 6- and 12-month follow-ups. Recruitment occurred 1 month before randomization. Specifically, the final baseline cohort was recruited in April 1999, and follow-ups were conducted in November 1999 and May 2000. During each assessment, participants completed a self-administered survey that addressed sociodemographic characteristics and psychosocial mediators of HIV-preventive behaviors. Also, a trained African American female interviewer conducted an interview assessing participants’ sexual behaviors and, using a structured scoring protocol, rated participants on their skill in correctly applying condoms. Finally, participants provided (self-administered) vaginal swab specimens that were analyzed to determine the presence of STDs.

Measures

Self-reported consistent condom use, the primary outcome assessed, was defined as use of a condom during every episode of vaginal intercourse in the preceding 30 days. We calculated consistent condom use by dividing the number of condom-protected episodes of vaginal intercourse by the total number of episodes of vaginal intercourse. Consistent condom use in the preceding 30 days was measured at the 6-month and 12-month assessments and computed over the entire 12-month study period (from baseline to the 12-month follow-up).

Other outcomes measured at the 6- and 12-month assessments and computed over the entire period (from baseline to 12 months) included (1) condom use at most recent sexual intercourse, (2) proportion of condom-protected vaginal intercourse episodes in the preceding 30 days, (3) proportion of condom-protected vaginal intercourse episodes in the previous 6 months, (4) number of unprotected vaginal intercourse acts in the preceding 30 days, (5) number of unprotected vaginal intercourse episodes in the preceding 6 months, (6) whether participants had had a new sexual partner in the previous 30 days, (7) demonstrated proficiency in applying condoms, and (8) acquisition of an STD (Neisseria gonorrhoeae, Chlamydia trachomatis, or Trichomonas vaginalis) over the 12-month follow-up period.

As mentioned, participants provided vaginal swab specimens that were analyzed to determined the presence of STDs. One vaginal swab was placed in a specimen transport tube and evaluated for C trachomatis and N gonorrohoeae via ligase-chain reaction assay; newly developed DNA amplification technology (Abbott LCx Probe System, Abbott Labs, Abbott Park, Ill)26^,27 was used in conducting these analyses. A second swab was used to inoculate culture medium for T vaginalis (In Pouch TV test; BioMed Diagnostics Inc, Santa Clara, Calif). This culture was incubated at 37°C and examined daily, via light microscopy (100× magnification), for 5 days to determine the presence of motile trichomonads.28

All STD assays were conducted at the research laboratory of the Division of Infectious Diseases at the University of Alabama, Birmingham. Adolescents identified with an STD were provided directly observable single-dose treatment (i.e., they were observed taking their medication), received appropriate risk reduction counseling in accord with the recommendations of the Centers for Disease Control and Prevention, and were encouraged to refer their sexual partners for treatment.

Psychosocial mediators of HIV-preventive behaviors were derived from social cognitive theory and the theory of gender and power, and we assessed these constructs using scales that had satisfactory psychometric properties and had previously been used with young African American women.29^–31 The psychometric properties and other elements of these continuous measures have been discussed in detail elsewhere; here we provide only a brief description of each measure.22

HIV prevention knowledge assessed participants’ knowledge of HIV risk reduction practices. Perceived partner-related barriers to condom use assessed attitudes impeding participants’ ability to effectively use condoms. Attitudes toward using condoms assessed hedonistic beliefs about using condoms. Frequency of sexual communication assessed the frequency with which participants discussed HIV-preventive practices with their sexual partners. Condom use self-efficacy assessed participants’ confidence in their ability to properly use condoms. Participants’ demonstrated proficiency in applying condoms was assessed as a measure of condom skills.

Analysis

Analyses included only adolescents reporting a history of gender-based violence at baseline. An intention-to-treat protocol was used in which participants were analyzed according to their original assigned study conditions irrespective of the number of sessions they attended.32^,33 At baseline, descriptive statistics were calculated to summarize differences in sociodemographic variables, psychosocial mediators, sexual behaviors, and STD prevalence rates between study conditions. We used Student t tests (for continuous variables) and χ² analyses (for categorical variables) to assess differences between conditions.34 Variables that involved statistically significant differences (P ≤ .05) or that were theoretically determined to be potential confounders (douching, gang involvement, pregnancy desire in past 30 days) were included as covariates in longitudinal analyses.

We report differences between the study conditions on dichotomous variables as unadjusted percentages at the 6- and 12-month assessment intervals and as odds ratios (ORs) over the 12-month follow-up period; we calculated odds ratios using logistic regression generalized estimating equations. We report differences between the study conditions on continuous variables as unadjusted means at the 6- and 12-month assessment intervals and as adjusted means over the entire 12-month period, calculated via linear regression generalized estimating equation models. To calculate adjusted mean differences, we used a procedure not assuming normality of distributions; in this procedure, models were repeatedly estimated from bootstrap samples drawn with replacement at the level of the participant.

Each model included a time-independent variable (study condition) as well as time-dependent variables (covariates and outcomes). We adjusted all outcome measurement models by the corresponding baseline measure and other identified covariates to obtain adjusted odds ratios and adjusted mean differences. In addition, we included a time period indicator in each model to capture any temporal effects that had not been taken into account.35^,36 In generalized estimating equation analyses, the 95% confidence intervals (CI) around adjusted odds ratios were calculated (intervals including 1.0 were not significant), as were the 95% confidence intervals around adjusted mean differences (intervals including 0.0 were not significant).

RESULTS

Of the 522 adolescents from the original sample who completed baseline assessments, 14% (n=73) reported that their boyfriend had coerced them into having intercourse against their will, and an additional 14% (n=73) reported that their boyfriend had physically abused them. Thus, 28% of the participants (n=146) reported a history of gender-based violence and were randomized to the study conditions, yielding 73 participants in the HIV prevention condition and 73 participants in the general health promotion condition. At baseline (Table 1 ▶), we assessed the comparability of the 2 groups in terms of sociodemographic characteristics, psychosocial mediators linked to HIV-preventive behaviors, sexual behaviors, and presence of STDs. There were no statistically significant differences between the groups.

TABLE 1—

Comparability of the HIV Prevention and General Health Promotion Groups on Selected Characteristics

	HIV Prevention (n = 73)		General Health Promotion (n = 73)
	Mean (SD)	No. (%)	Mean (SD)	No. (%)	P
Sociodemographic characteristics
Age, y	16.18 (1.23)		15.97 (1.19)		.31
Did not complete 10th grade		28 (38.4)		34 (46.6)	.32
Not attending school		7 (9.6)		7 (9.6)	1.00
Receiving public assistance		19 (26.0)		4 (19.2)	.32
Living in a single-parent home		34 (61.8)		42 (70.0)	.36
Employed		11 (15.1)		14 (19.4)	.49
Has children		19 (26.0)		21 (28.8)	.71
Consumed alcohol in previous 30 d		32 (43.8)		24 (32.9)	.17
Age at first nonvolitional intercourse, y	13.42 (2.53)		13.67 (2.81)		.69
Mediators
HIV knowledge	9.25 (3.25)		9.26 (3.40)		.99
Condom attitudes	36.03 (4.40)		34.75 (4.82)		.10
Partner-related condom barriers	42.97 (13.56)		45.43 (14.86)		.31
Condom use self-efficacy	31.92 (9.16)		31.01 (8.83)		.55
Demonstrated skill in applying condoms	3.01 (1.29)		3.20 (1.19)		.38
Sexual behaviors and STDs
No. of times used condoms in past 30 d	0.72 (0.41)		0.76 (0.37)		.53
No. of times used condoms in past 6 mo	0.65 (0.38)		0.61 (0.42)		.57
No. of times had unprotected vaginal sex in past 30 d	1.89 (3.81)		0.90 (1.72)		.07
No. of times had unprotected vaginal sex in past 6 mo	7.34 (17.18)		6.32 (11.15)		.69
Consistent condom use in past 30 d		42 (64.6)		41 (65.1)	.96
Condom use during most recent sexual intercourse		36 (55.4)		39 (61.9)	.45
New partner in past 30 d		5 (6.8)		9 (12.3)	.26
Positive for any STD		20 (28.2)		19 (26.4)	.81
Covariates
Douching		21 (29.1)		26 (36.2)	.10
Gang involvement		12 (16.8)		9 (12.0)	.12
Pregnancy desire		24 (32.7)		20 (26.9)	.15

Note. STD = sexually transmitted disease.

Of the 73 participants assigned to the HIV prevention condition, 63 (86.3%) completed the 6-month assessment, and 61 (83.6%) completed the 12-month assessment. Of the 73 participants allocated to the general health promotion condition, 65 (89.0%) completed the 6-month assessment and 65 (89.0%) completed the 12-month assessment. There were no differences in attrition between participants in the 2 study conditions at either the 6-month (P=.62) or 12-month (P=.34) assessment.

Effects of the HIV intervention on dichotomous measures of safe sex and STDs are shown in Table 2 ▶. Over the entire 12-month period, participants in the HIV prevention condition were more likely than participants in the general health promotion condition to report using condoms consistently (OR = 2.71; 95% CI = 1.24, 5.93; P = .01) and to report using a condom at their most recent sexual intercourse (OR = 3.69; 95% CI = 1.78, 7.65; P = .0001); conversely, they were less likely to have acquired a sexually transmitted infection (OR = 0.47; 95% CI = 0.25, 0.87; P = .02), and (although this result was not statistically significant) and were less likely to have a new sexual partner (OR = 0.31; 95% CI = 0.09, 1.08; P = .07).

TABLE 2—

Effects of HIV Intervention on Measures of Safe Sex and Frequency of Sexually Transmitted Infection

	6-Month Assessment		12-Month Assessment
	Intervention Group, Unadjusted %	Comparison Group, Unadjusted %	Intervention Group, Unadjusted %	Comparison Group, Unadjusted %	GEE Model Odds Ratioa (95% Confidence Interval)	P
Consistent condom use in past 30 d	61.3	38.3	72.6	48.9	2.71 (1.24, 5.93)	.01
Condom use during most recent sexual intercourse	66.0	34.0	64.7	48.9	3.69 (1.78, 7.65)	.0001
New sex partner in past 30 d	3.1	11.1	4.9	6.8	0.31 (0.09, 1.08)	.07
Positive for any sexually transmitted infection	16.9	39.7	18.5	23.0	0.47 (0.25, 0.87)	.02

Note. GEE = generalized estimating equation.

^aAdjusted by baseline value of outcome variable and theoretical covariates (douching, gang involvement, and pregnancy desire in past 30 days).

Effects of the HIV intervention on continuous safe sex measures are shown in Table 3 ▶. Over the entire 12-month period, participants in the HIV prevention condition reported significantly fewer episodes of unprotected vaginal intercourse in the preceding 30 days (adjusted mean difference = −1.48; 95% CI = −3.19, 0.23; P = .04) than participants in the general health promotion condition, as well as significantly fewer episodes of unprotected vaginal intercourse in the previous 6 months (adjusted mean difference = −13.34; 95% CI = −25.07, −1.61; P = .008). Also, the percentage of condom-protected sexual episodes in the previous 30 days was significantly higher in the HIV prevention group (adjusted mean difference = 17.00; 95% CI = 2.01, 32.00; P = .03). Finally, participants in the HIV prevention group reported a slightly higher (nonsignificant) percentage of condom-protected sex acts in the preceding 6 months (adjusted mean difference = 16.00; 95% CI = 4.00, 28.00; P = .08).

TABLE 3—

Effects of HIV Intervention on Continuous Measures of Self-Reported Sexual Behaviors

	6-Month Assessment		12-Month Assessment
	Intervention Group, Unadjusted Mean (SD)	Comparison Group, Unadjusted Mean (SD)	Intervention Group, Unadjusted Mean (SD)	Comparison Group, Unadjusted Mean (SD)	GEE Model Adjusted Mean Difference (95% Confidence Interval)	P
No. of episodes of unprotected vaginal sex in past 30 d	2.15 (5.13)	3.28 (5.71)	0.94 (2.13)	3.04 (5.75)	−1.48 (−3.19, 0.23)	.04
No. of episodes of unprotected vaginal sex in past 6 mo	5.89 (14.84)	17.50 (38.54)	6.20 (14.16)	17.51 (36.74)	−13.34 (−25.07, −1.61)	.008
Condom use in past 30 d, %	74.25 (38.91)	53.91 (42.52)	76.72 (40.61)	54.96 (47.64)	17.00 (2.01, 32.00)	.03
Condom use in past 6 mo, %	73.57 (36.12)	52.20 (39.62)	60.90 (41.76)	53.33 (42.47)	16.00 (4.00, 28.00)	.08

Note. GEE = generalized estimating equation. Mean differences were adjusted by baseline value of outcome variable and theoretical covariates (douching, gang involvement, and pregnancy desire in previous 30 days).

Effects of the HIV intervention on psychosocial mediators of safe sex are shown in Table 4 ▶. In comparison with participants in the general health promotion condition, participants in the HIV intervention had higher HIV prevention knowledge scores (adjusted mean difference=1.79; 95% CI=1.05, 2.57; P= .0001), had more favorable attitudes toward using condoms (adjusted mean difference= 1.81; 95% CI=0.79, 2.83; P=.005), reported fewer perceived partner-related condom barriers (adjusted mean difference= −5.51; 95% CI=−9.21, −1.82; P=.05), demonstrated greater proficiency in applying condoms (adjusted mean difference=1.03; 95% CI=0.68, 1.38; P=.0001), and had higher condom use self-efficacy scores (adjusted mean difference=3.21; 95% CI=0.39, 6.04; P=.04). The frequency with which they negotiated safe sex was not higher than the frequency observed among participants in the general health promotion condition.

TABLE 4—

Effects of HIV Intervention on Psychosocial Mediators

	6-Month Assessment		12-Month Assessment
Measure	Intervention Group, Unadjusted Mean (SD)	Comparison Group, Unadjusted Mean (SD)	Intervention Group, Unadjusted Mean (SD)	Comparison Group, Unadjusted Mean (SD)	GEE Model Adjusted Mean Difference (95% Confidence Interval)	P
HIV knowledge (range: 0–18)	11.65 (2.73)	10.00 (3.27)	11.75 (2.55)	10.21 (3.32)	1.79 (1.00, 2.57)	.0001
Condom attitudes (range: 8–40)	37.69 (3.11)	35.56 (4.54)	36.42 (4.26)	35.29 (4.50)	1.81 (0.79, 2.83)	.005
Partner-related condom barriers (range: 10–50)	38.62 (15.19)	45.32 (16.43)	40.49 (15.41)	42.85 (14.48)	−5.51 (−9.21, −1.82)	.05
Condom use self-efficacy (range: 9–45)	35.66 (9.47)	32.08 (9.83)	36.42 (9.54)	32.92 (9.50)	3.21 (0.39, 6.04)	.04
Demonstrated skill in applying condoms (range: 0–6)	4.38 (0.91)	3.30 (1.40)	4.17 (1.28)	3.68 (1.28)	1.03 (0.68, 1.38)	.0001

Note. GEE = generalized estimating equation. Mean differences were adjusted by baseline value of outcome variable and theoretical covariates (douching, gang involvement, and pregnancy desire in previous 30 days).

We also assessed whether the HIV prevention intervention increased participants’ subsequent risk of experiencing gender-based violence. Over the entire 12-month period of the study, the odds of experiencing gender-based violence did not differ between participants in the HIV prevention intervention and participants in the general health promotion condition (OR = 1.01; 95% CI = 0.19, 5.18; P = .99).

DISCUSSION

This secondary analysis of a randomized controlled HIV prevention trial involving female adolescents demonstrated that, among young women who had a history of gender-based violence, the HIV intervention led to substantial reductions in HIV-associated sexual behaviors, favorable changes in theoretically derived psychosocial mediators, and, most important, reductions in frequencies of sexually transmitted infections. These results are in contrast to previous research suggesting that adolescents with a history of gender-based violence may face significant barriers in enacting many of the safe sex messages of HIV interventions with their male sex partners.16

The efficacy of the HIV intervention assessed in this study might be attributed to its conceptualization within a gender-tailored framework, the theory of gender and power. Specifically, the intervention focused on reducing emotional stressors (e.g., low self-esteem) that can interfere with young women having healthy relationships, reducing partner-related barriers associated with not using condoms, enhancing their competency in using condoms, making informed choices about partner selection, and perceiving control in their sexual relationships. Because female adolescents who have experienced gender-based violence are more likely than those who have not to report having riskier sexual partners,16 lower perceptions of control over safe sex,16 and lower self-esteem,37 addressing these gender-based characteristics as part of the intervention may have encouraged participants to engage in self-protective HIV behaviors.

Another noteworthy finding of this study was that the HIV prevention intervention did not increase the incidence of subsequent abuse during the 12-month follow-up period. Thus, the intervention reduced these young women’s risk of HIV without placing them at harm for further victimization. Also interesting was that, in comparison with participants in the general health promotion condition, participants in the HIV intervention condition did not report increasing the frequency with which they negotiated safe sex. Previous research has shown that female adolescents who report a history of gender-based violence are more likely than female adolescents who do not report such a history to incur abuse as a result of requesting condom use from their male sex partners.16 To reduce further victimization and risk of HIV, participants in the HIV intervention may have elected to use alternative HIV prevention strategies that did not involve negotiation with their partners.

Although our study involved a number of methodological strengths, such as use of a randomized controlled design, use of a comparison condition structurally similar to the intervention condition, and inclusion of laboratory-confirmed data on STDs, it was not without limitations. First, our analyses included only a subgroup of the overall study sample, and statistical power to detect differences between the conditions was limited as a result of the small size of this subgroup. Second, the study population was limited to sexually experienced African American adolescent girls recruited from clinical venues. Thus, our findings may not be applicable to female adolescents who have different risk profiles (i.e., injection drug use history) or who are recruited from nonclinical venues.

Third, although the incidence of sexually transmitted infections was significantly lower among participants in the intervention condition than among those in the comparison condition, future interventions should include relapse prevention strategies to reinforce or amplify educational messages and further reduce frequencies of sexually transmitted infections. Fourth, our limited definition of gender-based violence did not include forms of violence known to be associated with HIV risk taking (e.g., child sexual abuse).38

Finally, a limitation of our measure of abuse is that it did not assess a defined time period. Thus, participants may have experienced recent abuse or abuse that took place many years in the past. Also, because we did not have information on this time frame, we were unable to determine whether the HIV intervention would have had the same effects among adolescent girls reporting a proximal abuse history and those reporting a distal abuse history. Future researchers examining HIV intervention efficacy may want to include participants reporting proximal and distal violence.

Overall, the observed magnitudes, consistency, and scope of the effects observed strengthen our confidence in the efficacy of the HIV intervention assessed. Furthermore, our study contributes new evidence that HIV interventions might be effective among high-risk adolescent populations, including female adolescents with a history of gender-based violence. Given the substantial prevalence of gender-based violence experienced by female adolescents and the significant associations we observed between gender-based violence, HIV risk, and HIV infection, future HIV intervention research involving adolescent girls should address partner violence.