Triple Jeopardy for HIV: Substance Using Severely Mentally Ill Adults

SUMMARY

Severely Mentally Ill (SMI) adults have disproportionately high HIV seroprevalence rates. Abuse of alcohol and other substances (AOD) and lifetime exposure to trauma by others are particularly potent risk factors, which, in combination with psychiatric disabilities, create triple jeopardy for HIV infection. This study examined the predictive utility of demographic characteristics; history of physical, emotional, or sexual abuse; extent of drug and alcohol abuse; knowledge about HIV/AIDS; sexual self-efficacy; and condom attitudes toward explaining the variance in a composite of HIV high-risk behavior among 188 SMI women and 158 SMI men. History of sexual abuse, engaging in sexual activities while high on substances, and lower cannabis use were the most significant predictors of HIV sexual risk behaviors. Given the triple jeopardy for HIV risk in this population, a triple barreled approach that simultaneously addresses multiple health risks within an integrated treatment setting is warranted.

Severely mentally ill (SMI) adults or individuals suffering from persistent and serious psychiatric symptoms and conditions show disproportionately high and escalating HIV seroprevalence rates (Carey, Weinhardt, & Carey, 1995; Rosenberg et al., 2001). These rates have been linked to a nexus of interacting clinical, behavioral, and contextual factors (Cournos & McKinnon, 1997; Kelly, 1997; Otto-Salaj, Kelly, Stevenson, Hoffman, & Kalichman, 2001).

Abuse of alcohol and other substances (AOD) have clearly been identified as potent risk factors for the development of HIV infection (Dausey & Desai, 2003). Nearly half of SMI individuals will develop some type of substance use disorder within their lifetime (RachBeisel, Scott, & Dixon, 1999; Regier et al., 1990). Those SMI adults who are dually diagnosed with a substance use disorder are at significantly greater risk for HIV infection than individuals with a single psychiatric diagnosis (Carey et al., 2004; McKinnon & Cournos, 1998). In combination with psychiatric disability, AOD can exacerbate psychotic symptoms and impede judgment and impulse control, leading to higher risk for HIV by increasing sexual desire, lowering sexual inhibition, and/or disturbing the consistent practice of safer sex.

Lifetime traumatic abuse by others has a substantial impact on risk behaviors for both genders. Sexual and/or physical assault histories have been associated with high-risk sexual practices in AOD abusing women (Dansky, Saladin, Brady, Kilpatrick, & Resnick, 1995; Hien & Scheier, 1996), in SMI women (Rosenberg, Drake, & Mueser, 1996), and in a community sample of both males and females (Bensley, Eenwyk, & Simmons, 1999). AOD abusing women who are also severely mentally ill have high rates of traumatic abuse histories (Malow et al., in press). Traumatic abuse severity is also positively correlated with HIV risk (Malow et al., in press). Thus, by virtue of the heightened risk for HIV that is associated with membership in three separate but underserved groups (SMI, AOD, victims of traumatic abuse by others), SMI adults appear to be at triple jeopardy for HIV infection. Within the cognitive-behavioral realm, information processing deficits that are an inherent component of the symptoms associated with SMI substance abusers, coupled with lack of knowledge and misconceptions about HIV transmission, may be mediating factors that contribute to HIV risk among SMI adults (Aruffo, Coverdale, Chacko, & Dworkin, 1990; Kalichman, Kelly, Johnson, & Bulto, 1994; Kelly & Murphy, 1992). However, other studies with SMI adults have shown that adequate amounts of HIV related knowledge is partially independent of HIV risk behavior but that knowledge by itself is not sufficient to explain behavior (McKinnon, Cournos, Sugden, Guido, & Herman, 1996).

SMI substance abusers present distinctive risks by virtue of their comorbidities and the attendant contextual challenges to cognitive and behavioral change. However, there is a paucity of information on the linkages between these risks. In order to determine what constitutes high-risk behavior in this population, we sought to identify a composite of experiences related to the practice of unsafe sex and sexual relations with high-risk partners. Thus, we examined the predictive utility of demographic characteristics; history of physical, emotional, or sexual abuse; extent of drug and alcohol abuse; knowledge of HIV/AIDS; sexual self-efficacy; and condom attitudes toward explaining the variance in HIV high-risk behavior among inner-city SMI adults in outpatient treatment.

METHOD

Participants

Of the 346 participants, 188 (54.3%) were women and 158 (45.7%) were men attending day and residential mental health treatment at 16 psychiatric or combined addiction mental health programs in the metropolitan Miami-Dade area. Participants were identified from 444 individuals who volunteered to be screened for the study; there were 126 exclusions. The study exclusion criteria were: (a) residency outside of Miami-Dade FL (n = 31); (b) transportation difficulties that would preclude participation; (c) non-English speaking primary language (n = 5); (d) denial of drug or alcohol use in the past 6 months (n = 36); (e) did not meet criteria for severe mental illness (n = 3); and, (f) evidence of severe cognitive dysfunction as indicated by inability to concentrate or focus during screening session or intake assessment (n = 14). Seven participants left their respective treatment centers and 30 of those eligible refused to participate in the study.

Assessment Procedures

All assessment procedures were conducted by seasoned interviewers, trained to create a process sensitive to gender and cultural issues as well as to adopt a nonjudgmental attitude to establish rapport and build trust. To prevent interviewer drift and other contaminating factors, assessors received ongoing supervision from a clinical psychologist throughout the study. Following informed consent, assessment measures were verbally administered to facilitate full completion and to compensate for any literacy difficulties. In the event that a respondent showed any confusion, an assessor would repeat or elaborate on questions.

A manual and a procedural checklist were used to ensure standard administration of the assessment measures. Interviews were performed in a private room to guarantee confidentiality and enhance compliance. Interviewers used key events and calendar time lines to facilitate accurate reporting of the participants’ behavior over the recall period.

Measures

Psychiatric diagnoses. Researchers abstracted current Axis I and Axis II diagnoses based on the fourth edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV) (APA, 1994) from participants’ clinical records at study sites. Most current chart diagnoses were taken as the best approximation of diagnosis at the time of data collection.

Abuse History Questionnaire

This instrument evaluated lifetime prevalence of abuse. All respondents were asked, “In your lifetime, have you ever been (a) sexually abused, (b) emotionally abused, and/or (c) physically abused?” Direct questions were used to assess abuse history to allow participants to self-define abuse situations rather than narrowly defining specific abuse events.

The Addiction Severity Index (ASI)

This comprehensive, structured clinical research interview was selected to gather sociodemographic information (i.e., age, self-identified ethnic background, years of education, marital status, and employment status), psychiatric history and current symptoms, and alcohol and drug abuse history (McLellan, Metzger, & Peters, 1992). The ASI has shown high concurrent and inter-rater reliability (.74 to .93) and validity among a similar sample of SMI adults. Furthermore, drug treatment outcome research has repeatedly used the ASI and has found it to be reliable, valid, and effective in evaluating adults with serious mental illness (Zanis, McLellan, Cnaan, & Randall, 1994).

The AIDS Risk Reduction Model Questionnaire Revised (ARRM-QR)

The ARRM-QR was created to reliably and validly measure the cognitive behavioral constructs that the ARRM model hypothesizes to be predictive of HIV-risk behaviors (Catania, Kegeles, & Coates, 1990). Subscales used in the present study to predict level of HIV-risk included: knowledge about HIV and AIDS, sexual self-efficacy, anxiety about contracting HIV, and condom attitudes. Scoring for the knowledge about HIV and AIDS subscale was based on the summed total of correct answers with a maximum total score of 18. The scales for condom attitudes, sexual self-efficacy, and anxiety subscale had a 4-point Likert format with response options ranging from 1 (strongly disagree) to 4 (strongly agree). For each scale, higher scores reflect a more positive response.

Risk Behavior Assessment (RBA) (National Institute on Drug Abuse, 1991)

Participants reported the number of male and female sexual partners, intercourse frequency with men and women, and number of times condoms were used during vaginal and anal sex in the past three months. A three-month time frame has been shown to generate reliable recall of sexual behaviors (Kauth, St. Lawrence, & Kelly, 1991). In addition, we asked participants to report (a) number of sex partners, (b) number of occasions of sexual activity, and (c) consistency of condom use during intercourse. Participants described their sexual behavior with primary, casual, and/or “trading” sex partners. A primary sex partner was defined as a person with whom the respondent had a current or long-term relationship. Casual partners were defined as people with whom the respondent had sex, but with limited emotional involvement. A “trading” relationship was characterized by encounters in which sex was exchanged for money or drugs, whether received or given. The RBA has shown adequate test-retest reliability for sexual risk behaviors (Needle et al., 1995) and is similar to instruments used in past research to assess sexual behavior in adults diagnosed with SMI (Kalichman et al., 1994).

Levels of HIV Risk Behavior Composite

This scale was derived from six dichotomous items from the RBA that assessed the presence or absence of HIV risk factors for the preceding three months. Risk factors included: (a) having injection drug using partner, (b) sex with homosexual man, (c) engaging in sexual trade (e.g., sex for shelter, money, drugs), (d) unprotected vaginal or anal sex, (e) more than one partner, and (f) occurrence of STDs. These factors were summed to estimate HIV risk based on a continuous index, in which multiple factors indicate higher risk.

Data Analysis

Multiple regression analysis was utilized to test for the influence of the predictors on the levels of HIV risk behavior composite. A hierarchical analysis with sets of variables was used with predictors entered sequentially in four blocks (Cohen & Cohen, 1983). Sets of variables were added cumulatively to the equation in an order that approximated their temporal relationship to the HIV high-risk related behaviors. The variables of gender, age, and ethnicity were entered in the first block to examine preexisting demographic differences that could account for the variance in the risk composite. Lifetime history of emotional, physical, and sexual abuse was entered in the second block. Drug use variables were entered into the third block including number of times alcohol, marijuana, heroin, and cocaine were used in the last 30 days, as well as frequency of sex while high. Knowledge of HIV/AIDS, anxiety regarding contraction of HIV/AIDS, condom attitudes, and sexual self-efficacy were entered in the fourth block. This order controls for demographics, introduces traumatic abuse and substance abuse as independent blocks of indicators, and finally enters variables that are the target of cognitive-behavioral risk reduction interventions.

RESULTS

Description of the Sample

Based on the distributions of participant self-report and clinic charts, the modal sample participant was an indigent, ethnic minority individual who was diagnosed with schizophrenia, abused alcohol, non-injection “crack” cocaine and/or marijuana, and lived in the urban inner city. The participants ranged in age from 19 to 81 years (M = 58.01, SD = 12.18), and years of formal education ranged from 0 to 18 years, with 41.3% having completed high school. Their marital status distribution was as follows: single, 52%; married, 7%; separated/divorced, 36%; and widowed, 5%. Approximately 55% of participants were unemployed. Participants were 50.9% African American; 26.3% Non-Hispanic White; 20.8% Hispanic of Cuban, South American, Mexican or Puerto Rican origin; and 2% of other ethnic origin (i.e. Asian, Native American). The most common psychiatric diagnosis for the sample was schizophrenia (23%), followed by major depressive disorder (19%). Clinic chart diagnostic data were missing on 28 participants. Of the remaining participants, 92% had a major psychiatric disorder, while 46% of participants met criteria for both substance abuse/dependence and psychiatric disorder. Emotional, physical, or sexual abuse was a common phenomenon in this sample with 76% (n = 262) of participants reporting abuse of some form. For the entire sample, 68% (n = 236) had been emotionally abused, 53% (n = 186) had been physically abused, and 41% (n = 141) had been sexually abused. Multiple abuse experiences were frequent, with 26% (n = 91) of participants reporting two forms of abuse and 30% (n = 105) of participants reporting three forms of abuse.

Average use of substances in the previous 30 days was reported as follows: alcohol, 6.77 times (SD = 11.18); cannabis, 2.40 times (SD = 7.32); cocaine, 2.18 times (SD = 6.85); and heroin, 0.62 times (SD = 4.05). Mean scores on the cognitive behavioral constructs were as follows: knowledge of HIV/AIDS (M = 13.27, SD = 3.07), attitudes towards using condoms (M = 3.23, SD = .42), sexual self-efficacy (M = 3.02, SD = .53), and anxiety about contacting AIDS (M = 3.01, SD = .74). There were no significant differences between treatment sites on the levels of risk composite (F(3,219) = .302, p = .82).

HIV Risk Behaviors

Table 1 shows frequency and relative percentages of HIV sexual risk behavior for men and women in this sample. The most frequent risk behavior for both men (n = 68) and women (n = 105) was having unprotected vaginal or anal sex during the last 30 days. The average number of risk factors per participant was 1.31 (SD = 1.24) and ranged from none (33.2%) to six (.3%). Results were analyzed using bivariate correlations and multiple regression. The highest significant correlations among the variables include alcohol use and cocaine use (r = .40), cannabis use and alcohol use (r = .35), cannabis use and cocaine use (r = .44), sexual self-efficacy and condom attitudes (r = .49), history of physical abuse and history of emotional abuse (r = .50), and history of sexual abuse and history of emotional abuse (r = .40). We found no evidence of multicollinearity among the predictors or violations of other assumptions of regression.

TABLE 1.

HIV Risk Behaviors Composite by Gender

Variable	Males n = 158	Females n = 188
IV Drug Use	2 (1.26%)	3 (1.60 % )
Sex with an MSM	11 (6.96%)	13 (6.91%)
Sex for Trade	37 (23.42%)	52 (27.66%)
Unprotected Vaginal or Anal Intercourse	68 (43.04%)	105 (55.85%)
Sex with Multiple Partners	51 (32.28%)	71 (37.77%)
History of STDs in past 30 days	14 (8.86%)	27 (14.36%)

Predicting HIV Risk Behaviors

Table 2 displays the unstandardized regression coefficients (B), the standardized regression coefficients (β), and R, R² and adjusted R² after entry of all four blocks of predictors. The variables in the first block did not contribute significantly to predicting level of HIV risk. The inclusion of history of abuse variables in the second block contributed significantly to predicting risk, with blocks 1 and 2 combined accounting for approximately 5% of the variance. The standardized regression coefficients show that lifetime history of sexual abuse contributed the most among variables in the second block. The inclusion of drug use variables in block 3 also considerably contributed to predicting HIV risk, accounting for approximately 23% of the variance. Examining the standardized regression coefficients for the third block illustrates that drug use during sexual intercourse and frequency of cannabis use over the last 30 days contributed the most among variables in this block. To identify the role that cognitive-behavioral constructs have on predicting risk level, we entered knowledge of HIV/AIDS, condom attitudes, sexual self-efficacy, and anxiety about contracting HIV/AIDS variables into the fourth block. In the final model, history of sexual abuse (β = .13, t = 2.28, p = .023), being high during sex (β = .47, t = 8.99, p < .001), and lower cannabis use in the last month (β = −.13, t = −2.82, p = .005) significantly predicted a greater level of HIV risk behaviors (R² = .28, F(17, 328) = 7.52, p < .001).

TABLE 2.

Hierarchical Regression Predicting Level of Risk Composite

Variables	B	β	t	ΔR2
Block 1
Sex	.08	.03	.64
Non-Hispanic White Ethnicity	−.23	−.08	−1.43
Hispanic	−.11	−.04	−.67
Other Ethnicity	.08	.01	.20
Age	−.006	−.07	−1.35	.018
Block 2
History of Emotional Abuse	−.05	−.02	−.35
History of Physical Abuse	.14	.05	.96
History of Sexual Abuse	.32	.13	2.28*	.028
Block 3
Sex When High on Drugs	1.24	.46	8.99***
Alcohol Use Last 30 Days	−.009	.09	1.68
Heroin Use Last 30 Days	−.01	−.04	−.88
Cocaine Use Last 30 Days	.009	.06	.95
Cannabis Use Last 30 Days	−.02	−.16	−2.82**	.227
Block 4
Knowledge	−.02	.06	1.03
Condom Attitudes	−.21	−.07	−1.23
Sexual Self-Efficacy	−.04	−.02	−.33
Anxiety	.07	.04	.88	.007
Constant	1.47		2.47*
				R2 = .28
				Adj. R2 = .24 R = .53

^*p ≤.05

^**p ≤.01

^***p ≤.001

DISCUSSION

These findings demonstrate that in a sample of inner city, ethnically diverse SMI adults from a variety of mental health and substance abuse treatment sites, history of sexual abuse, greater frequency of substance use while engaging in sex and lower cannabis use are most predictive of HIV risk. We also found that SMI adults demonstrated a variety of highly risky sexual behaviors including unprotected vaginal or anal sex, sex for money or drugs, sex with multiple partners, and history of STDs. Other correlates of HIV risk behavior such as age, gender, ethnicity, knowledge about HIV/AIDS, anxiety regarding contraction of HIV/AIDS, attitudes toward condom use, and sexual self-efficacy were not predictive.

The findings establishing a link between sexual abuse and high-risk behaviors are consistent with those from other studies that link past sexual trauma with subsequent sexual maladjustment and increased engagement in sexual risk behaviors (Miller, 1999). History of sexual abuse has also been repeatedly linked to substance use/abuse (Burnam et al., 1988; Finkelhor, Hotaling, Lewis, & Smith, 1990) as well as to HIV risk behavior (Malow et al., in press), further supporting the notion that abuse trauma may be mediated by substance use in subsequent HIV risk behavior. In the aftermath of sexual trauma, it may be that drugs are used as a coping mechanism, which subsequently increases the likelihood of engaging in sexual risk behavior through disinhibition or dissociation (Miller, 1999).

A high prevalence of sexual abuse has been noted in the SMI literature. Mueser, Goodman, and Trumbetta (1998) reported that among a sample of SMI adults, 35% of men and 52% of women had experienced child sexual abuse compared to 16% and 27% respectively in the general population. Forty percent of the study participants in the current study reported a lifetime history of sexual abuse. Of this subgroup, three-fourths were women and one-fourth were men. Sexual trauma has been related to a variety of negative outcomes in SMI adults such as more severe symptoms, greater use of substances, and higher rates of posttraumatic stress disorder (PTSD), a common comorbid disorder in SMI adults (Mueser et al., 1998; Rosenberg et al., 2001). It has been theorized that PTSD may mediate the negative effects of sexual trauma on the course of SMI and subsequent HIV risk both directly, through the effects of specific symptoms of PTSD (i.e., avoidance, overarousal, and re-experiencing the trauma) and indirectly, through the effects of correlates of PTSD (i.e., retraumatization, substance abuse, and difficulties with interpersonal relationships) (Mueser, Rosenberg, Goodman, & Trumbetta, 2002).

Unprotected and risky sexual practices as a long-term consequence of sexual abuse are also found in populations who do not have the cognitive impairments manifested in SMI adults (Bensely et al., 1999; Cunningham, Stiffman, Doré, & Earles, 1994). Thus, it may be that the potency of the trauma may overshadow any other factors that might moderate the relationship between sexual abuse and HIV risk behaviors. This finding also suggests that those who have experienced sexual abuse may have great difficulties in initiating and carrying out safer sexual behaviors. Using drugs as part of their sexual experience may be a form of self-treatment that offers immediate relief from the impact of abuse. Interestingly, in our sample there was a negative relationship between cannabis use and HIV sexual risk behaviors. It could be conjectured that cannabis provides less of a feeling of disinhibition than alcohol or cocaine.

Given the triple jeopardy for HIV risk in this sample, a triple barreled approach that simultaneously addresses multiple risks within an integrated treatment setting is a necessity for HIV prevention efforts. The participants in this study represented a severely impaired group in terms of psychiatric diagnoses, degree of substance use and abuse, and rates of sexual abuse. Any targeted HIV prevention interventions should include both sexual and substance abuse risk reduction approaches (Carey et al., 2004) against the backdrop of psychiatric comorbidities. Because of the lack of association between the cognitive-behavioral constructs and HIV risk behaviors, other factors need to be considered such as specific symptom clusters that are more closely tied to cognitions; differential responses based on diagnosis; or, the impact on information processing by such factors as disinhibition or dissociation. Interventions targeted for this population may need to be tailored specifically to the patient’s level of functioning as well as diagnostic category (Carey, Carey, Maisto, Gordon et al., 2004).

In light of the implications for designing interventions to reduce the spread of HIV and other STDs among SMI adults, certain limitations should be noted in interpretation of the findings. One concern in developing the risk composite was the possibility of correlation between variables, both within the composite and the overlap between predictors. For example, it is very likely that our model did not account for or masked naturally occurring combined practices such as use of cocaine with sex trade because of the need to barter drugs for money/housing (i.e. “survival sex”).

The reliance on clinical records for psychiatric diagnoses and self-report for substance use and sexual behavior raise reliability concerns, and pertinent findings should thus be interpreted with caution. The direct or indirect contribution of specific psychiatric symptoms cannot be determined from these analyses. Future studies addressing risk behavior among SMI adults should focus on specific symptomatology and comorbidity with PTSD in predicting risk and as mediating factors in HIV prevention interventions. This would increase the ecological validity and enhance implications for assessment and risk prevention.