Psychosocial Predictors of Sexual HIV Transmission Risk Behavior among HIV-Positive Adults with a Sexual Abuse History in Childhood

Abstract

Childhood sexual abuse (CSA) is associated with HIV sexual risk behavior. Although many psychosocial correlates of sexual risk among HIV-positive persons have been identified, studies predicting continued risk among HIV-positive adults with histories of CSA are limited. This cross-sectional study identified variables predictive of sexual transmission risk behavior among an ethnically diverse sample of 256 HIV-positive adults (women and men who have sex with men; MSM) with CSA histories. Participants were assessed for trauma symptoms, shame related to HIV and sexual trauma, substance use, coping style, and sexual risk behavior. Logistic regression analyses were conducted to identify variables predictive of unprotected sexual behavior in the past 4 months. Unprotected sex was significantly associated with substance use and trauma-related behavioral difficulties among women and men, and less spiritual coping among men. Unprotected sex with HIV negative or serostatus unknown partners was significantly associated with greater trauma-related behavioral difficulties, more HIV-related shame, and fewer active coping strategies. Thus, trauma symptoms, shame, coping style, and substance use were significantly associated with sexual risk behavior among HIV-positive adults with histories of CSA, with models of prediction differing by gender and partner serostatus. HIV prevention intervention for persons with HIV and CSA histories should address trauma-related behavioral difficulties and enhance coping skills to reduce sexual transmission risk behavior.

Introduction

There are approximately 1.2 million people living with HIV infection in the United States (UNAIDS, 2006). Of the 75% who are aware of their infection (Marks, Crepaz, & Janssen, 2006), the majority remain sexually active after learning of their diagnosis (Bingman, Marks, & Crepaz, 2001; Crepaz & Marks, 2003). While it is estimated that those who are unaware of their infection are 3.5 times more likely to transmit HIV than those who are aware of their infection, nearly half of all new sexually transmitted HIV infections result from HIV-positive persons who know of their infection and engage in sexual risk behavior (Marks et al., 2006). Estimates of the rate of continued sexual risk behavior range from 10% to 60%, depending on the time frame and specific sex act (Kalichman, 2000; Marks, Burris, & Peterman, 1999). Thus, as the number of persons living with HIV/AIDS rises, it is critical that secondary HIV prevention interventions are developed to target sexual risk reduction among HIV-positive persons (CDC, 2003).

Though a growing body of research is available to guide such efforts, many specific groups and predictors of risk remain unexamined. Crepaz and Marks (2002) provided a comprehensive review of the empirical literature on the psychosocial and medical correlates of continued sexual risk behavior among HIV-positive persons. Research has substantiated several constructs central to prevailing health behavior models, including HIV/AIDS knowledge, behavioral intentions, self-efficacy, communication skills, and perceived barriers (de Vroome, de Wit, Stroebe, Sandfort, & van Griensven, 1998; Godin, Savard, Kok, Fortin, & Boyer, 1996; Hays et al., 1997; Kalichman, Kelly, & Rompa, 1997). Additionally, partner serostatus is an important factor, as studies consistently find that HIV-positive persons are more likely to have unprotected sex with an HIV-positive partner compared to an HIV-negative partner (Kalichman, Roffman, Picciano, & Bolan, 1998; Lansky, Nakashima, & Jones, 2000; Sobel, Shine, DiPietro, & Rabinowitz, 1996). Substance use has been linked to continued sexual risk among people living with HIV across a number of studies (O'Leary et al., 2005; Purcell, Parsons, Halkitis, Mizuno, & Woods, 2001; Stein, Rotheram-Borus, Swendeman, & Milburn, 2005). Other predictors have yielded inconsistent results. Associations between psychiatric distress, such as depression and anxiety, and continued sexual risk behavior have been mixed and, in fact, in a meta-analysis of 34 studies, the average effect size was not significant (Crepaz & Marks, 2001). The range of effect sizes, however, was -.41 to .55, and displayed a fairly even distribution, with a third of studies showing little effect, a third showing significant negative effects, and a third showing significant positive effects. Given the existing evidence, the role of affective states in influencing sexual risk behavior is unclear, although anger and impulsivity have shown more consistent relationships with increased sexual risk (Crepaz & Marks, 2001; Hays et al., 1997; Kalichman, Greenberg, & Abel, 1997; Wolitski, Parsons, & Gomez, 2004). Further, studies examining coping strategies have also yielded mixed results (Brook et al., 1998; Clement, 1992; Kalichman, 1999; Robins, Dew, Kingsley, & Becker, 1997; Semple, Patterson, & Grant, 2000). These discrepant results may be partially explained by the diverse characteristics defining groups of HIV-positive persons (e.g., injection drug users, men who have sex with men, pregnant women).

One group deserving of greater attention is HIV-positive persons with histories of childhood sexual abuse (CSA). Definitions of CSA vary between studies, but generally, the research summarized here utilized broad definitions, including any forced or coerced sexual contact (not just penetration) between a child or adolescent (under age 18) with an adult or someone at least 5 years older than the child when the abuse occurred. Researchers have documented a strong association between CSA and later engagement in sexual risk behaviors (Greenberg, 2001; Koenig, Doll, O'Leary, & Pequegnat, 2004). Correspondingly, the prevalence of CSA among HIV-positive persons is strikingly high, with rates ranging from 32% to 76% across studies (Bedimo, Kissinger, & Bessinger, 1997; Kalichman, Sikkema, DiFonzo, Luke, & Austin, 2002; Liebschutz, Feinman, Sullivan, Stein, & Samet, 2000; Schiff, El-Bassel, Eng-strom, & Gilbert, 2002). CSA rates in the general population range from 8% to 32% in females and from 1% to 16% in males (Allers, Benjack, White, & Rousey, 1993; Briere & Elliott, 2003; Finkelhor, 1994; Kimerling, Armistead, & Forehand, 1999). Although we are unaware of any studies that compare rates of CSA in matched samples of HIV-positive and HIV-negative participants, and differences in research methodologies and operational definitions limit our ability to make direct comparisons in CSA rates between HIV infected samples and the general population, it seems apparent that rates of CSA are higher among HIV-positive adults. Additionally, rates of both CSA and HIV infection are high in clinical and high-risk populations, such as the homeless (Benda, 2006), the severely mentally ill (Van Dorn et al., 2005), and substance abusers (Boles, Joshi, Grella, & Wellisch, 2005). Given the recurring association between CSA history and HIV infection across diverse studies and samples, a pressing concern is identifying variables related to continued sexual transmission risk behavior among those who are living with HIV infection and have experienced CSA.

Childhood sexual abuse had been found to be associated with a wide variety of HIV risk behaviors. For example, researchers have linked CSA with sex work (Bartholow et al., 1994; Cunningham, Stiffman, Dore, & Earls, 1994; Senn, Carey, Vanable, Coury-Doniger, & Urban, 2006, 2007; Zierler et al., 1991), risky sexual behavior (Rotheram-Borus, Mahler, Koopman, & Langabeer, 1996; Senn et al., 2007; Wingood & DiClemente, 1997), multiple sex partners (Rotheram-Borus et al., 1996; Senn et al., 2007; Thompson, Potter, Sanderson, & Maibach, 1997; Wyatt et al., 2002), substance abuse (Kalichman, Adair, Somlai, & Weir, 1995; Longshore & Anglin, 1995), and sexual victimization in adulthood (Briere, Woo, McRae, Foltz, & Sitzman, 1997; Classen, Palesh, & Aggarwal, 2005; Rich, Combs-Lane, Resnick, & Kilpatrick, 2004). Additionally, CSA often results in psychological difficulties, such as helplessness, hopelessness, low self-esteem, dissociation, denial, avoidance, and self-destructiveness, which have also been linked to HIV-risk behavior (Briere, 2004). Further, among HIV-positive adults, those with sexual assault histories were more likely than those without to have abused substances and reported recent unprotected intercourse (Kalichman et al., 2002). Finally, in the only study we are aware of that has examined the association between CSA and transmission risk behavior specific to partner serostatus, those with histories of CSA were more likely to have engaged in unprotected insertive and receptive sex with HIV-negative or serostatus unknown partners (O'Leary, Purcell, Remien, & Gomez, 2003). Thus, CSA itself is clearly a predictor of continued sexual risk behavior for people living with HIV.

In order to develop and tailor effective secondary HIV prevention interventions for HIV-positive persons with CSA histories, it is critical to understand the factors that are associated with sexual risk behavior in this population. These predictors may include variables related to CSA itself (e.g., age when abuse occurred, chronicity of abuse, relationship to perpetrator), as well as consequences of abuse (e.g., traumatic symptoms, shame, poor affect regulation). Briere's (2002) self-trauma theory suggests that “early and severe child maltreatment interrupts normal child development, conditions negative affect to abuse-related stimuli, and interferes with the usual acquisition of self capacities” (p. 185). From this perspective, the posttraumatic experiences of intrusion and avoidance are viewed as natural mechanisms for healing, with intrusive symptoms serving to expose the individual to traumatic material and avoidant symptoms titrating exposure to manageable levels. In the context of severe and/or chronic trauma, and particularly in the presence of under-developed or diminished self-capacities, however, traumatic stress overwhelms the individual's abilities to handle its effects, and the intrusion-desensitization process fails. Thus, many instances of sexual risk behavior can be categorized as: (1) maladaptive attempts to regulate affect or cope with traumatic symptoms; (2) learned maladaptive behavior patterns; or (3) behaviors following from the negative assumptions, expectations, and self-evaluations that result from early trauma (Briere, 2004). Unfortunately, HIV-positive persons with CSA also face stigma associated with their illness (Comer, Kenker, Kemeny, & Wyatt, 2000), which may intensify preexisting feelings of powerlessness, betrayal, and victimization and contribute to continued sexual risk behavior. In addition, the cognitive and behavioral strategies utilized by individuals to cope with both CSA and HIV infection may be important predictors of continued sexual risk behavior (Lazarus & Folkman, 1984). Finally, comorbid psychiatric conditions and substance abuse are high in this population, which may also contribute to risk behavior. Thus, within the framework of self-trauma theory, CSA specific predictors of continued sexual risk should be examined within samples of HIV-infected individuals who have experienced CSA. These variables related to: (1) the traumatic stressor (age when abuse started, frequency of abuse, number of perpetrators, abuse involving oral, anal, or vaginal penetration, and distress at time of abuse); (2) self-capacities (shame related to HIV infection and sexual abuse, and cognitive and behavioral strategies for coping with CSA); and (3) traumatic symptoms and substance abuse.

The primary purpose of the current study was to identify predictors of sexual transmission risk behavior among a diverse sample of HIV-positive adults with histories of CSA. We focused on two subgroups who continue to have high rates of HIV infection: women and men who have sex with men (MSM). We hypothesized that sexual risk behavior would be associated with sexual abuse characteristics (greater frequency of abuse, more perpetrators, greater distress at the time of the abuse, and abuse involving penetration), greater traumatic symptoms, greater HIV- and abuse-related shame, and greater maladaptive coping strategies. Specifically, more avoidant (e.g., denial, withdrawal) and self-destructive (e.g., substance use, suicidal ideation) coping strategies and less active, spiritual, and social support seeking coping strategies were expected to be associated with sexual risk behavior. Based on prior research, we also hypothesized that sexual risk behavior would be associated with current substance use. Specifically, participants indicating current use of alcohol, marijuana, and/or cocaine/crack were predicted to exhibit more sexual risk behavior. Because transmission risk behavior with an HIV-negative partner is of public health concern, we conducted the analyses in two steps. First, we examined predictors of sexual risk behavior irrespective of partner serostatus in order to examine unprotected sexual activity across all partner types. As few studies have examined predictors of sexual risk behavior among both men and women, little is known about common and unique risk factors across genders. Therefore, we analyzed the data separately by gender, hypothesizing that the predictive models for women and men would differ. Second, we examined predictors of unprotected vaginal or anal intercourse with partners who were HIV-negative or of unknown serostatus, as this sexual behavior is specifically associated with high HIV transmission risk.

Method

Participants

HIV-positive adults with CSA histories were recruited between March 2002 and January 2004 in New York City to participate in an intervention study on coping with HIV and sexual abuse. All participants were HIV-positive, had experienced sexual abuse prior to turning 18 years of age, and were interested in obtaining group psychotherapy to cope with CSA in the context of HIV infection (for details, see Sikkema et al., 2007). The sample included 132 women and 124 MSM. Participants were ethnically diverse (67.3% African-American, 16.9% Hispanic/Latino, 10.0% Caucasian, 5.8% other), had a mean age of 42.2 years (SD = 7.0), a mean education of 12.2 years (SD = 2.4), and 92.7% had an annual income less than $20,000. They had been HIV positive for a mean of 9.6 years (SD = 5.2).

Procedure

Participants were recruited from community organizations, including public health and AIDS service organizations, and health care clinics. Brochures were distributed to these locations and providers were asked to refer clients to the study. A total of 333 individuals participated in the screening process. Twenty-one of these were ineligible (seven were severely depressed, seven were not sexually abused in childhood, six were cognitively impaired, and one had experienced acute sexual revictimization) and 41 did not return for the baseline assessment (23 could not be located, 12 were uninterested/unable to continue, three were incarcerated or institutionalized, two were deceased/hospitalized, one felt emotionally unprepared to participate in the study). Thirteen heterosexual men were excluded from the analysis due to inadequate numbers to form a comparison group. Data from two participants were incomplete on several measures and hence dropped from the analyses, leaving a total sample of 256.

All interested participants were individually screened using a structured clinical interview that assessed demographics, sexual abuse history, depression, mental status, and participant risk to self or others. Abuse history was assessed using a modified version of the Traumatic Experiences Questionnaire (TEQ; Kaplan, Asnis, Lipschitz, & Chorney, 1995). Participants were asked about types of abuse and perpetrator of the abuse, including experiences throughout childhood, adolescence, and adulthood. Screening interviews where scheduled for a 60 min time slot. Inclusion criteria were: (1) sexual abuse as a child (age 12 and under) and/or adolescent (age 13–17), defined as any unwanted touching of a sexual nature by an adult or by someone at least 5 years older than the participant when the incident occurred; (2) current age of 18 or older; and (3) HIV-positive serostatus. Exclusion criteria were: (1) acute distress due to sexual revictimization experienced within the past month; (2) presence of acute psychosis or impaired mental status; and (3) extreme distress evidenced by suicidal intention or severe depressive symptomatology as suggested by a score of ≥30 on the Beck Depression Inventory (BDI; Beck & Steer, 1987). Exclusion for extreme distress or depression was due to the possibility of participants being randomized into a waitlist control condition during the intervention study and thus having a delay before receiving treatment. Individuals (n = 7) excluded for these reasons were provided with immediate referrals for appropriate mental health services. All participants provided informed consent. Eligible participants then completed a 90-min baseline assessment using a self-administered computer assisted interview. All procedures were approved by an institutional review board.

Measures

Predictor Variables

Sexual Abuse Characteristics

Briere and Elliott (2003) identified five characteristics of childhood sexual abuse that were related to traumatic distress in adults who had been abused: (1) abuse at a later age; (2) a greater number of abuse incidents; (3) multiple perpetrators; (4) abuse involving oral, anal, or vaginal penetration; and (5) greater distress at the time of the abuse. Based on these findings, participant abuse data from the modified TEQ administered at screening were used to create the following six variables reflecting these domains: (1) Age of first abuse, (2) Frequency of abuse (on a Likert scale ranging from 1 (one time) to 5 (constantly)), (3) Number of perpetrators, (4) Distress at time of abuse (on a Likert scale ranging from 0 (not at all distressing) to 4 (extremely distressing)), and (5) Penetrative abuse (two dichotomous variables indicating any oral, anal, or vaginal penetrative abuse during childhood (age 0–12) and adolescence (age 13–17)).

Trauma Symptom Inventory (TSI)

The TSI is a 100-item self-report measure that assesses acute and chronic post-traumatic symptomatology, including the effects of rape, partner abuse, physical assault, and the lasting sequelae of childhood abuse and other early traumatic events (Briere, 1995). Each item reflects a potential trauma symptom and was rated according to its frequency of occurrence over the preceding 6 months on a 4-point scale (0 = never, 3 = often). The TSI consists of 10 clinical scales that can be categorized into three broad symptom clusters labeled: trauma-related symptoms, mood and anxiety symptoms, and behavioral difficulties. The trauma-related symptoms cluster assesses trauma-specific symptom domains and contains the following four scales: Intrusive Experiences (“nightmares or bad dreams”), Defensive Avoidance (“pushing painful memories out of your mind”), Dissociation (“feeling like things weren't real”), and Impaired Self-Reference (“your feelings or thoughts changing when you were with other people”). The mood and anxiety symptoms cluster assesses symptom domains that are frequently associated with trauma and contains the following three scales: Anger/Irritability (“becoming angry for little or no reason”), Depression (“feeling hopeless”), and Anxious Arousal (“feeling tense or ‘on edge”’). The behavioral difficulties cluster assesses common behavioral disruptions that are associated with trauma, particularly sexual trauma, and contains the following three scales: Sexual Concerns (“confusion about your sexual feelings”), Dysfunctional Sexual Behavior (“having sex or being sexual to keep from feeling sad or lonely”), and Tension Reduction Behavior (“intentionally hurting yourself (for example, by scratching, cutting, or burning) even though you weren't trying to commit suicide”). It should be noted that the behavioral difficulties cluster does not specifically assess sexual behavior, including neither the frequency nor risk level of sexual behavior, nor sexual partner-related information, and thus it does not overlap with outcome variables used in this study. The clinical scales of the TSI are internally consistent (α = .81–.91, current sample).

Coping Measures

Two measures of coping were used for the study: The Ways of Coping Questionnaire (WOC; Folkman & Lazarus, 1988) is a 66-item inventory that measures coping style in response to a particular stressor (for the current study, CSA was identified as the specific stressor). Participants indicated the extent to which they used each coping strategy using a 4-point scale (0 = not at all, 3 = most of the time). The Coping With Illness (CWI; Murphy, Rotheram-Borus, & Marelich, 2003) scale is a 31-item instrument that measures participants’ coping responses related to their HIV illness. Participants rated how often they used different coping behaviors to help them deal with being HIV-positive in the past month using a 4-point scale (0 = never, 3 = often).

Due to the instability of the eight WOC subscales across different samples and stressors (van Heck & de Ridder, 2001), most studies using this scale utilize alternative scoring methods to better reflect their particular sample and stressor under examination (Simoni, Demas, Mason, Drossman, & Davis, 2000). A five-factor solution combining the WOC and CWI identified in a previous study (Tarakeshwar, Hansen, Kochman, & Sikkema, 2005) provided adequate internal consistency in the current sample: Active Coping (29 items; e.g., “formed a plan of action in your mind,” α = .93); Avoidant Coping (16 items; “I tried to forget the whole thing,” α = .91); Social Support Seeking (8 items; “talked with others with problems like yours,” α = .82); Self-Destructive Behavior (8 items; “tried to reduce tension by drinking more than usual,” α = .82); and Spiritual Coping (5 items; “trusted your belief in God,” α = .80). It should be noted that Avoidant Coping reflects a passive approach to coping with stress, whereas the Defensive Avoidance subscale of the TSI assesses active methods of blocking or avoiding thoughts, places, or people associated with trauma.

Shame Concerning CSA and HIV Infection

This is a 33-item measure composed of three scales: Shame Concerning HIV-Infection, Shame Concerning Sexual Abuse, and Impact of Shame on Behavior. The Shame Concerning HIV Infection scale has 14 items that assess the current experience of shame about being HIV positive (α = .92). Sample items include: “Being HIV-positive makes me feel defective, like there's something wrong with me,” and “I hide my infection from others.” The Shame Concerning Sexual Abuse scale has 9 items that assess the current experience of shame regarding childhood sexual abuse (α = .93). Sample items include: “I struggle with feeling worthless because I've been sexually abused/raped,” and “When others find out about the sexual abuse/rape, I expect them to reject me.” The Impact of Shame on Behavior has 10 items that assess how behaviors relating to HIV-infection are influenced by shame (α = .78). Sample items include: “Feelings of shame that I'm HIV-positive have kept me from wearing a condom or asking a partner to wear a condom,” and “Feelings of shame that I'm HIV-positive have kept me from telling my HIV status to a sex partner.” Participants reported agreement with each statement using a 4-point scale (0 = not at all, 4 = very much). These scales assess whether shame associated with HIV-infection and sexual trauma inhibit communication with peers, caregivers, and sexual partners, not specific frequency or risk level of sexual behavior.

Substance Use (Sikkema, Kochman, DiFranceisco, Kelly, & Hoffman, 2003)

This scale assessed the frequency of use of a number of substances, including alcohol, marijuana, cocaine, tranquilizers, narcotics, amphetamines, and injection drugs over the past 4 months. For the current analyses, only data on alcohol, marijuana, cocaine, and crack were utilized, as use of other substances was infrequent. Additionally, due to the high correlation between cocaine and crack use (r = .67, p < .001), these variables were summed into a variables reflecting cocaine and/or crack use. Thus, three substance abuse variables were formed: alcohol use, marijuana use, cocaine/crack use. Because 46.9% of participants reported no substance use in the previous 4 months, and data were highly skewed, for each of the three substance use variables, dichotomous coding was used, yielding three dichotomous indicators coded as “no substance use” and “any substance use.”

Outcome Variables

Sexual Behavior (Kalichman et al.,2002; Sikkema et al., 2000)

Participants reported the number of times they engaged in oral, anal, and vaginal intercourse with and without condoms in the past 4 months. They also reported the HIV status of each partner with whom they engaged in each behavior in the previous 4 months. As with substance abuse, sexual behavior data formed highly skewed and zero-inflated distributions. Thus dichotomous coding was used to categorize participant responses into two dichotomous outcome variables. The first sexual outcome variable represented “any recent unprotected sex,” and was coded affirmatively if the participant had engaged in any oral, anal, or vaginal sexual behavior without a condom in the previous 4 months. The second sexual outcome variable represented “high sexual transmission risk behavior” and was coded affirmatively if the participant had engaged in any unprotected anal or vaginal sexual behavior with a HIV-negative or serostatus unknown partner in the previous 4 months.

Statistical Analysis

The statistical analysis involved two steps. First, logistic regression analyses were conducted to identify predictors of unprotected sex among men and women using the outcome variable “any recent unprotected sex,” defined as one or more episodes of unprotected oral, vaginal, or anal sex during the previous 4 months. Participants who were sexually abstinent or used condoms consistently in the past 4 months, and thus reported no unprotected sex, were defined as having “no risk.” Due to the number of variables and issues of collinearity, the three symptom clusters of the TSI were included instead of the 10 clinical subscales. All predictors that were associated with “any recent unprotected sex” in univariate logistic regression analyses at p < .10 were included in the multivariate model. Variables were then removed from the model using backward deletion.

Second, a logistic regression analysis was conducted using only participants who had engaged in anal or vaginal intercourse with an HIV-negative or HIV serostatus unknown partner within the past 4 months. Participants who consistently used condoms during vaginal or anal sex were defined as “low sexual transmission risk.” Participants who engaged in one or more episodes of unprotected vaginal or anal intercourse were defined as “high sexual transmission risk.” Participants who were sexually abstinent or engaged only in oral sex were excluded from this analysis. Due to the reduction in sample size and subsequent reduction in power, this analysis was not conducted separately by gender. All predictor variables that were associated with “high sexual transmission risk” in univariate logistic regression analyses at p < .10 were included in the multivariate model. Variables were then removed from the model using backward deletion.

As an odds ratio reflects the percentage of change associated with a one-unit change on the predictor variable, continuous variables with a large range can produce significant results with extremely small odds ratios. Therefore, to aid in the interpretation of odds ratios, all continuous scale scores, including coping, shame, and TSI, were divided by the number of items in the scale to reduce the score to reflect the average score in the item level metric for that scale.

Results

Participants reported extensive sexual trauma histories. As per the inclusion criteria, all had experienced unwanted sexual contact prior to age 18. The vast majority (90%) of the sample had experienced penetrative vaginal or anal sexual abuse as a child or adolescent. Additionally, 87% experienced sexual revictimization at some point in their lives and, on average, participants were abused by two different perpetrators.

The majority (69%) of participants reported sexual activity during the previous 4 months. Table 1 shows participants’ sexual risk behavior by gender. Men reported significantly more partners and they were more likely to have engaged in any unprotected sex and unprotected anal/vaginal sex with an HIV-negative or serostatus unknown partner.

Table 1.

Sexual behavior by gender, past 4 months

	Women (N = 132)		Men (N = 124)		t
Continuous variables	M	SD	M	SD
Number of sex partners	1.0	1.1	4.5	13.4	2.91a**
Dichotomous variables	N		N	%	χ 2
Sexually abstinent	45	34.1	35	28.2	1.02
Any unprotected sex	39	29.5	63	50.8	12.06**
Unprotected anal/vaginal sex	33	25.0	43	34.7	2.87
Unprotected anal/vaginal sex with HIV-negative partner	19	14.4	34	27.4	6.61*

^aZ Mann Whitney U = –3.97, p < .001

^*p < .05

^**p < .01

Table 2 lists each of the predictor variables by gender. Compared to women, men were younger, more educated, less likely to be of minority ethnicity, less distressed at the time of abuse, more likely to have experienced penetrative oral sexual abuse, more likely to have used substances in the previous 4 months, had lower social support seeking and spiritual coping, and had higher scores on trauma-related behavioral difficulties.

Table 2.

Descriptive statistics for predictive variables

	Women (N = 132)		Men (N = 124)		t
Continuous variables	M	SD	M	SD
Age (in years)	43.1	7.1	41.2	6.8	2.20*
Education (years)	11.6	2.1	13.0	2.6	4.65**
Childhood abuse variables
Age of first abuse	8.9	3.1	8.8	3.2	<1
Frequency of abuse	3.6a	1.4	3.5a	1.3	<1
Number of perpetrators	2.0	1.2	2.3	1.8	1.49
Distress at time of abuse	3.3b	1.1	2.9b	1.3	2.58b
Coping scales
Active coping	44.3	17.0	42.7	16.2	<1
Social support seeking	13.5	5.6	11.4	5.7	2.90**
Spiritual coping	10.0	3.8	8.7	3.6	2.86**
Self-destructive behavior	5.3	5.5	5.5	4.8	<1
Avoidant coping	20.6	11.2	20.0	9.4	<1
HIV-related shame
Shame concerning HIV-infection	17.2	13.7	18.5	13.2	<1
Impact of shame on behavior	3.3	5.3	4.2	5.0	1.33
Shame concerning sexual abuse	13.5	10.8	13.7	9.7	<1
Trauma symptom inventory
Mood and anxiety symptoms	29.8	19.1	28.2	16.3	<1
Trauma-related symptoms	40.4	24.8	38.9	22.2	<1
Behavioral difficulties	18.6	17.2	24.1	16.7	2.55*
Categorical variables	N	%	N	%	χ 2
Ethnicity					11.14**
Caucasian	7	5.3	19	15.3
African-American	100	75.8	75	60.5
Hispanic/Latino	18	13.6	26	21.0
Other	7	5.3	4	3.2
Alcohol use: Past 4 months	42	31.8	66	53.2	12.01**
Marijuana use: Past 4 months	24	18.2	45	36.3	10.65**
Cocaine and/or Crack use, past 4 months	25	18.9	41	33.1	6.67**
Penetrative childhood abuse variables
Oral sexual: Age 0-12	56	42.4	82	66.1	14.46**
Anal/vaginal sex: Age 0-12	81	61.4	89	71.8	3.11
Oral sex: Age 13-18	58	43.9	71	57.3	4.54*
Anal/vaginal sex: Age 13-18	93	70.5	74	59.7	3.27

^*p < .05

^**p < .01

^aThese value fall between a score of 3 (abuse occurred “several times”) and 4 (abuse occurred “quite a few times”) on the Likert scale used for this question

^bThese values round to a score of 3 (abuse was “quite distressing” when it occurred) on the Likert scale used for this question

Table 3 summarizes the univariate and multivariate logistic regression analyses examining any recent unprotected sex by gender. Variables associated with any recent unprotected sex in univariate analyses at p < .10 (Wald χ² values not shown) are marked in bold type and were included in the multivariate analyses. Behavioral difficulties, alcohol use, marijuana use, and cocaine and/or crack use were associated with any unprotected sex for both women and men. Self-destructive behavior was also associated with any unprotected sex for women only, and social support seeking, spiritual coping, avoidant coping, and trauma-related symptoms were associated with less unprotected sex for men only. In multivariate analyses, both models significantly improved prediction over a constant only model (Model χ² = 9.74 for women and 32.28 for men). The improvement in prediction was less for women than men (3.8% and 21.7%, respectively), possibly due to the smaller number of women who engaged in sexual risk behavior. After backward removal of non-contributing variables, the model for women was reduced to two variables, with cocaine and/or crack use in the past 4 months and greater behavioral difficulties predictive of any unprotected sex (Nagelkerke R² = .10). The final model for men included four variables, including marijuana use in the past 4 months, less spiritual coping, fewer trauma-related symptoms, and greater behavioral difficulties. However, collinaerity diagnostics revealed that trauma-related symptoms were redundant with other variables in the model, thus it was dropped from the final model (Negelkerke R² = .32).

Table 3.

Summary of logistic regression analyses predicting any recent unprotected sex, past 4 months

	Women (N = 131)				Men (N = 124)
	Univariate		Multivariatea		Univariate		Multivariateb
Predictor variable	OR	CI	OR	CI	OR	CI	OR	CI
Age (in years)	.98	(.92-1.03)			.99	(.94-1.05)
Education (years)	.91	(.76-1.09)			1.08	(.94-1.25)
Ethnicity	1.85	(.40-8.70)			1.40	(.52-3.76)
Alcohol use: Past 4 months	2.71*	(1.25-5.85)	-	-	4.11**	(1.94-8.69)	-	-
Marijuana use: Past 4 months	2.42⊥	(.97-6.01)	-	-	3.81**	(1.74-8.37)	3.22**	(1.32-7.85)
Cocaine and/or crack use: Past 4 months	3.38**	(1.37-8.31)	2.67*	(1.13-5.39)	2.52*	(1.16-5.48)	-	-
Abuse variables
Age of first abuse	.99	(.87-1.12)			.94	(.84-1.05)
Frequency of abuse	1.09	(.82-1.45)			1.12	(.86-1.48)
Number of perpetrators	1.09	(.79-1.50)			.93	(.76-1.14)
Distress at time of abuse	1.01	(.72-1.41)			.89	(.68-1.17)
Penetrative abuse: Age 0-12	1.36	(.57-3.25)			.58	(.22-1.52)
Penetrative abuse: Age 13-18	1.43	(.58-3.51)			.98	(.46-2.06)
Coping scales
Active coping	1.00	(.54-1.85)			.68	(.36-1.26)
Social support seeking	1.08	(.63-1.84)			.49**	(.29-.84)	-	-
Spiritual coping	.91	(.56-1.47)			.31**	(.17-.55)	.30**	(.16-.57)
Self-destructive behavior	1.96*	(1.14-3.34)	-	-	1.20	(.66-2.16)
Avoidant coping	1.17	(.71-1.95)			.48*	(.27-.88)	-	-
HIV-related shame
Shame concerning HIV-infection	.94	(.64-1.39)			1.01	(.69-1.47)
Impact of shame on behavior	1.16	(.58-2.32)			1.75	(.80-3.80)
Shame concerning sexual abuse	1.21	(.88-1.64)			.94	(.68-1.31)
Trauma symptom inventory
Mood & anxiety symptoms	1.05	(.64-1.72)			.65	(.37-1.15)
Trauma-related symptoms	1.38	(.82-2.31)			.57⊥	(.32-1.01)	-c	-
Behavioral difficulties	1.97*	(1.12-3.46)	1.70⊥	(.94-3.06)	2.44a	(1.31-4.57)	2.21*	(1.11-4.42)

^*p < .05

^**p < .01

^⊥p < .10

^aNagelkerke R² = .10, 73.8% correctly classified

^bNagelkerke R² = .32, 74.2% correctly classified

^cCollinearity diagnostics indicated that Trauma-Related Symptoms, while contributing to the multivariate model, was too highly correlated with both Behavioral Difficulties and Avoidant Coping. Thus, this variable was removed as being redundant in the model

Table 4 presents the final analysis predicting high sexual transmission risk behavior among the 124 participants (48% of the sample) who had sex with HIV-negative or serostatus unknown partners. Because this analysis was conducted for males and females together, gender was first entered into the regression model using forced entry (Block 1). Next, variables univariately associated with high sexual transmission risk behavior at p < .10 (Wald χ² values not shown) were entered into a multivariate logistic regression analysis (Block 2). This model significantly improved prediction over a constant only model (14.5%; Model χ² = 30.71), with three variables contributing significantly to the final prediction model: Marijuana use in the past 4 months, greater impact of HIV-related shame on behavior, and less use of active coping. Additionally, the trauma-related behavioral difficulties cluster was retained in the model, though this variable was not significant in the presence of the other three variables.

Table 4.

Summary of logistic regression analyses predicting high sexual HIV transmission risk behavior with HIV-negative or serostatus unknown partners

	Univariate (N = 122)		Multivariate (N = 122)a
Predictor variable	OR	CI	OR	CI
Block 1 (forced entry)
Gender	2.18*	(1.05-4.53)	1.17	(.48-2.83)
Block 2 (backward removal)
Age (in years)	.95*	(.90-1.00)	-	-
Education (years)	1.00	(.86-1.16)
Ethnicity	2.70⊥	(.85-8.59)	-	-
Alcohol use: Past 4 months	3.02**	(1.38-6.62)	-	-
Marijuana use: Past 4 months	3.00*	(1.24-7.28)	3.33**	(1.33-8.30)
Cocaine and/or crack use: Past 4 months	2.44*	(1.13-5.26)	-	-
Abuse variables
Age of first abuse	.94	(.84-1.06)
Frequency of abuse	.90	(.69-1.18)
Number of perpetrators	.92	(.71-1.19)
Distress at time of abuse	.87	(.65-1.17)
Penetrative abuse: Age 0-12	1.35	(.56-3.26)
Penetrative abuse: Age 13-18	.80	(.36-1.80)
Coping scales
Active coping	.44*	(.22-.86)	.38*	(.19-.83)
Social support seeking	.65	(.39-1.06)
Spiritual coping	.55*	(.34-.89)	-	-
Self-destructive behavior	1.69*	(1.02-2.80)	-	-
Avoidant coping	1.12	(.65-1.92)
HIV-related shame
Shame concerning HIV-infection	1.39⊥	(.96-2.03)	-	-
Impact of shame on behavior	4.25**	(1.73-10.44)	3.34*	(1.21-9.24)
Shame concerning sexual abuse	1.19	(.86-1.63)
Trauma symptom inventory
Mood and anxiety symptoms	1.17	(.71-1.94)
Trauma-related symptoms	1.17	(.70-1.96)
Behavioral difficulties	1.97**	(1.18-3.28)	1.68⊥	(.91-3.09)

^*p < .05

^**p < .01

^⊥p < .10

^aNagelkerke R² = .31, 70.9% of participants correctly classified

Discussion

Given the established association between CSA and sexual risk behavior, and the high prevalence of CSA among HIV-positive adults, we examined correlates of sexual risk behavior among a sample of HIV-positive women and men with CSA histories to identify predictive models of sexual transmission risk behavior. Participants reported an extensive range of sexual abuse experiences, with nearly all having experienced forced vaginal or anal intercourse as a child or adolescent, as well as later sexual revictimization. The majority of both women and men reported vaginal or anal penetration prior to age 13. It has been previously documented that HIV-positive women and men have higher rates of CSA than the general population (Allers et al., 1993; Bedimo et al., 1997; Kalichman et al., 2002; Liebschutz et al., 2000; Schiff et al., 2002), and the current findings further suggest that this abuse is repetitive and that adult revictimization is common (see Classen et al., 2005).

Though there was some overlap, predictors of unprotected sexual behavior differed between men and women. Overall, the model for men was stronger than the one for women, most likely due to higher levels of risk behavior among men. Variables related to CSA failed to predict sexual risk behavior. Though somewhat surprising, one potential explanation for this is that the experiences of abuse in this sample were so extensive and prevalent that particular abuse characteristics may not stand out as unique or especially indicative of adult sexual risk. Variables related to self-capacities, however, did predict sexual risk in univariate and multivariate analyses, and predictive variables also differed by gender. For men, more avoidant coping was predictive of less sexual risk, which is the opposite of what was predicted. It is possible that men using avoidant coping strategies were also avoiding sexual situations, though the current data cannot test this premise. Additionally, more spiritual and social support coping was related to less sexual risk behavior in men. Surprisingly, active and avoidant coping strategies and shame concerning HIV and CSA were not associated with recent sexual risk behavior in women. For women, self-destructive behavior was predictive of increased sexual risk behavior in the univariate analysis, though this variable was dropped from the multivariate model. Sexual risk behavior was also predicted by trauma-related behavioral difficulties in both women and men, though this variable was not quite significant in the multivariate model for women. In men, the association between lower levels of trauma-related symptoms and sexual risk behavior approached significance, and this variable was selected as a significant predictor in the multivariate model, though it was dropped from the model due to overlap with avoidant coping and trauma-related behavioral difficulties.

Interestingly, although all substance abuse variables were related to risk for women and men in univariate analyses, these variables differed in multivariate models. For women, cocaine and/or crack use was predictive of risk in multivariate models, while for men marijuana use was predictive of risk. Unfortunately, the current study did not provide sufficient information to test whether the use of these substances itself was related to risk, or whether these substance use variables are markers for personality variables that are related to risk. For instance, it is possible that marijuana use in men is a marker for a lifestyle that includes both recreational drug use and high risk sexual activity. Additionally, it is possible that cocaine/crack use in women is more indicative of substance abuse or dependence and that high risk sexual activity frequently is instrumental in obtaining drugs. This is consistent with findings among male and female STD clinic attendees that substance use mediated the relationship between sexual risk behavior and CSA, and that women were more likely than men to have exchanged sex for money or drugs (Senn et al., 2007). More research is needed to clarify potential differences in the relationship between substance use and sexual risk behavior between gender and sexual orientation.

Due to sample size, the model predicting high transmission risk behavior (i.e., unprotected anal or vaginal sex with an HIV-negative or serostatus unknown partner) included both men and women. A univariate analysis indicated that men were more likely to engage in transmission risk behavior than women, though this relationship did not remain significant in the multivariate model. All substance use variables were univariately related to risk, although only marijuana use was retained in the multivariate model, most likely due to the relationship between marijuana use and sexual risk in men. Although no variables related to CSA were predictive of risk behavior, variables related to self-capacities were identified as potential predictors of HIV transmission risk in univariate analyses. These included: less use of active and spiritual coping, more self-destructive behavior, and more shame concerning HIV-infection and the impact of shame on behavior. In multivariate analyses, participants who reported low levels of active coping (in relation to both HIV and sexual trauma) were more than twice as likely to have engaged in high sexual transmission risk behavior. In addition, those with elevated HIV-related shame, specifically the perceived influence of shame and internalized stigma on behaviors that included condom use and serostatus disclosure, were more than three times as likely to engage in high sexual transmission risk behavior.

Finally, the influence of trauma-related behavioral difficulties was retained as a predictor in multivariate analyses, though it fell short of statistical significance, with men and women who endorsed more behavioral difficulties almost twice as likely to report high sexual transmission risk behavior. As sex with HIV-negative or serostatus unknown partners is the behavior that is of greatest concern for HIV prevention efforts, these findings have implications for the refinement of prevention interventions targeting persons living with HIV/AIDS. Effective HIV prevention interventions for HIV-positive persons with CSA histories should target the maladaptive behavioral responses used to address negative affect and the consequences of childhood abuse, address the impact of shame on behavior, and enhance the development of active approaches to stress and coping.

It should be noted that this study was exploratory in nature, and did not aim to make comparisons or draw conclusions beyond HIV-positive adults who were sexually abused as children and/or adolescents. The link between CSA and sexual risk behavior has been sufficiently established in the literature to warrant an examination of abuse-related variables (Malow, Devieux, & Lucenko, 2006), including those related to the actual abuse experience (traumatic stressor), self-capacities, and traumatic symptoms and other consequences, such as substance abuse, as predictors of sexual risk behavior. This is particularly relevant among those living with HIV, where sexual risk poses the risk of HIV transmission to others. Given the importance of this issue, and that HIV-positive men with CSA histories have been shown to be more likely to engage in HIV transmission risk behavior than those without CSA histories (O'Leary et al., 2003), the lack of studies identifying trauma-specific predictors of risk was surprising. Future research should focus, however, on comparing HIV positive samples with and without CSA histories to determine if these predictors differ in magnitude between samples or are unique to those with CSA histories.

This study was not without limitations. Like most studies of HIV risk behavior, this research relied on self-reports and may result in underreporting of risk behaviors, especially high sexual transmission risk behavior. To increase accuracy, we used a computerized assessment format, but some participants may have not reported risk behavior or been influenced by social desirability. In addition, this study employed a cross-sectional methodology, and hence, inferences regarding causality are unwarranted. Furthermore, the unique influence of CSA cannot be ascertained without a non-abused comparison group.

Despite these limitations, a major strength of this study was the diversity of the sample with respect to age, gender, and ethnicity, though all of the men self-identified as gay or bisexual. Our sample may reflect the difficulty in finding heterosexual men who are willing to address sexual trauma issues due to fears of being ridiculed as weak or inadequate (McMullen, 1990) or reluctance to raise questions regarding their sexual identity (West, 2000). Thus, in this sample, gender was confounded with sexual orientation, and our findings can only be generalized to HIV-positive women and men who have sex with men, and to persons who are willing to seek treatment or be in a research study related to sexual abuse and HIV disease. An additional strength of this study is that the predictive variables examined were selected based on principles of self-trauma theory (Briere, 2002), a widely used theory of the developmental, cognitive, and behavioral consequences of childhood abuse, which has clear treatment implications and associated guidelines (Briere, 2004; Briere & Scott, 2006). This study was not, however, an attempt to evaluate self-trauma theory; rather, self-trauma theory provided a useful conceptual framework for organizing study variables and findings.

Researchers have noted that the emotional and psychological sequelae of sexual trauma may impact the ability of individuals to fully utilize existing HIV prevention interventions and have begun to call for the creation of new HIV prevention interventions, or the modification of existing interventions, to account for sexual trauma (Briere, 2004; Greenberg, 2001; Parillo, Freeman, Collier, & Young, 2001). Our results suggest the importance of addressing trauma-related symptoms, especially those that manifest in behavioral difficulties related to negative affect, reducing HIV-related shame, and developing adaptive coping styles (including active and spiritual coping strategies) in prevention interventions for persons living with HIV/AIDS and CSA histories. Tailored interventions of this nature are limited; however, findings from two research programs are promising. First, following a 16-session pilot group intervention for coping with HIV/AIDS and CSA among both women and men, clinically significant improvement, based on the Jacobson and Truax (1991) method of normative comparison, was shown in the areas of trauma-related symptoms and behavioral difficulties (Sikkema et al., 2004). A subsequent RCT of this intervention compared a 15-session coping group to both a 15-session support group intervention and a waitlist control (later randomly assigned to an intervention condition). While participants in both group conditions had reduced traumatic symptoms in relation to the waitlist control condition, the coping group condition achieved greater change in tests of both statistical and clinical significance (Sikkema et al., 2007). We anticipate that the effects from this type of intervention will extend to reductions in high sexual transmission risk behavior. Second, a phase I clinical trial was conducted to test the impact of a culturally congruent psychoeducational intervention for HIV-positive women with CSA histories (Wyatt et al., 2004). In comparison to women in a waitlist control, those in the 11-session intervention reported greater sexual risk reduction with main partners and greater medication adherence among women who attended 8 or more sessions.

While targeted interventions such as these require the identification of HIV-positive patients with CSA histories, participants in this study—as well as many individuals with HIV/AIDS without CSA histories—are likely to have experienced a multitude of stressors and nonsexual traumas. Future research should examine the distinct and common influences of various traumas (e.g., childhood physical abuse, discrimination, homelessness, interpersonal violence, violent deaths of family members) that contribute to the social context in which high transmission risk occurs. It is imperative that HIV secondary prevention interventions be developed and evaluated to determine the efficacy of integrating mental health and coping into interventions for reducing high transmission risk behavior among HIV-positive individuals. Especially for those who have experienced CSA, treatment of trauma-related behavioral symptoms, their underlying causes, and the shame and stigma that may result from the “double-trauma” of HIV and CSA, appears critical for improving the quality of life and reducing further transmission to others.