Past and new victimization among African American female drug users who participated in an HIV risk-reduction intervention.

Negative experiences triggered by the ecological concentration of social and health problems are common among disadvantaged women of color residing in inner-city communities. Among the threats to their individual well-being are victimization experiences, substance abuse, and the HIV/AIDS epidemic, and a lack of community-level resources or social capital (Coyle, 1998; Dancy, 2000; Sterk, 1999; Wechsberg, Dennis, & Stevens, 1998). For women to enhance their health and welfare they need to acquire the skills to reduce their individual threat levels as well as strategies to successfully navigate the social infrastructure.

In this study, we focused on women who face additional challenges due to their illicit drug use (Koenig & Moore, 2000; Maman, Campbell, Sweat, & Gielen, 2000; Miller & Neaigus, 2001; Sterk, 1999; Wechsberg, Craddock, & Hubbard, 1998). Researchers have shown that being involved in illegal activities increases the likelihood that individuals will experience victimization (Jensen & Brownfield, 1986; Sampson & Lauritsen, 1990). The lifestyle of female drug users often places them in high-risk situations, thereby making them an easy target for victimization.

Illegal drug use has been associated with emotional problems such as depression, anxiety, and post-traumatic stress disorder (PTSD; McKusker, Goldstein, Bigelow, & Zorn, 1995; Morrill, Kasten, Urato, & Larson, 2001) and with childhood and adult physical, emotional, and sexual abuse (El-Bassel, Simoni, Cooper, Gilbert, & Schilling, 2001; Morrill et al., 2001; Nyamathi, Bayley, Anderson, Keenan, & Leake, 1999; Sterk, 1999; Wechsberg, Craddock, et al., 1998: Wilsnack, Vogeltanz, Klassen, & Harris, 1997). Compared to women who do not report sexual victimization experiences, women with such histories are more likely to have sex with casual and new acquaintances, have a higher number of lifetime sexual partners, be involved with sexual partners who themselves are at risk, and engage in prostitution (El-Bassel et al., 2001; Morrill et al., 2001; Nyamathi et al., 1999; Sterk, 1999; Wechsberg, Dennis, et al., 1998; Wilsnack et al., 1997). The links between victimization, drug use, and HIV sexual risk-taking are complex, as there are many potential pathways leading from one to the other (El-Bassel et al., 2001 ; Sterk, 1999).

As women continue to be proportionally more affected by the devastating HIV/AIDS epidemic, risk reduction prevention interventions have been designed to assist them in reducing their sexual and (when applicable) drug use-related risk. Many interventions show success. Among those that have targeted female drug users, however, drug-related risk reduction tends to exceed sexual risk reduction (Cottler et al., 1998; Sloboda, 1998). Attention to the social context of the women's lives beyond sex and drugs seems key to risk reduction (Ehrhardt & Exner, 2000; Ickovics & Yoshikawa, 1998; Sterk, Theall, & Elifson, 2003: Sterk, Theall, Elifson, & Kidder, 2003). We were unable to identify any studies to date that directly addressed women's victimization experiences following HIV or other behavioral interventions.

In this research we sought to determine whether a woman-oriented and culturally specific HIV intervention targeting African American women who use illegal drugs influenced their new victimization experiences, meaning those that occurred during the 6 months after enrollment into the study. In addition, we identified factors associated with new victimization.

METHODS

Study Procedures

We recruited women in Atlanta, Georgia, using street outreach techniques, including ethnographic mapping and targeted sampling. Recruitment communities were identified using the community identification (CID) process, a mapping method to record social and health epidemiological indicators (e.g., from emergency rooms, law enforcement, and drug treatment facilities), expert opinions (e.g., local political leaders and public health officials), and ethnographic information from local researchers and members of the target population (Tashima, Cram, O'Reilly, & Sterk-Elifson, 1996).

Participants enrolled between June 1998 and January 2001 as part of an HIV risk reduction trial among HIV-negative, heterosexually active African American women who use drugs. To be eligible, women had to be 18 years of age or older, reside in one of the study communities, be out of drug treatment or any other institutional setting, be proficient in English, be HIV-negative, have had vaginal sex with a man at least once during the month prior to the interview, and be an active illegal drug user (measured as having smoked crack cocaine or having injected drugs at least three times 30 days prior to the baseline interview).

The Emory University Human Investigations Committee and the Georgia State University Internal Review Board reviewed the consent procedures. Eligible women were assigned to one of the three intervention conditions: (a) a National institute on Drug Abuse (NIDA) standard intervention for drug users (NIDA, 1992); (b) a four-session enhanced motivation intervention; and (c) a four-session enhanced negotiation intervention (Sterk, 2002). We considered women in the NIDA standard intervention the control group.

We derived the enhanced intervention conditions from a 1-year formative research study (Elifson & Sterk, 2003), and the content was empirically as well as theoretically based. Theories employed included social cognitive theory (Bandura, 1977), the theory of reasoned action (Fishbein & Middlestadt, 1989), the theory of planned behavior (Ajzen & Madden, 1986), the transtheoretical model of change (Prochaska & DiClemente, 1983), and the theory of gender and power (Connell, 1985). Hence, the intervention conditions addressed topics such as decision making, weighing advantages and disadvantages of actions, ambivalence, readiness and motivation to change certain behaviors, and gender role expectations and role strain.

Both enhanced intervention conditions emphasized the local HIV epidemic, sex and drug-related risk behaviors, safer sex (including the use of male and female condoms), safer injection drug use, issues surrounding HIV risk reduction strategies, and the impact of race and gender on HIV risk and protective behaviors. Central to the enhanced intervention conditions was the social context of the women's daily lives, including the meaning of behaviors and social interactions, gender dynamics, economic stressors, gender-specific norms and values, and power and control. Incorporation of this daily life context allows for prevention interventions that are sensitive to women's needs. The motivation intervention focused on short- and long-term behavioral change, including readiness for change, sense of control, and feelings of ambivalence. The negotiation intervention focused on general communication skills and strategies to develop assertiveness, including negotiation and conflict resolution strategies.

The formative research phase, prior to implementation of the intervention, suggested a very high prevalence of past and current victimization among the women. Therefore, we emphasized experiences of coercion, abuse, and trauma during the intervention sessions. Because most reported abuse was received from men in the participants' lives, the intervention addressed the roles played by all men (e.g., steady partners, paying partners, family members). The same abuse and victimization module was used for both enhanced intervention conditions. Both enhanced conditions gave the women tools to cope with past and current victimization and to prevent potential acts of abuse. The intervention focused on women in their social relationships, addressing their experiences and recognizing their vulnerabilities and strengths (Elifson & Sterk, 2003: Sterk, Theall, & Elifson, 2003).

We interviewed participants immediately following completion of the intervention (postintervention) and 6 months later. Of the women assessed at baseline and who completed all intervention sessions as well as the postintervention interview, 96% were interviewed 6 months later. A total of 58 women did not complete all intervention sessions. Baseline comparisons revealed no significant sociodemographic or behavioral differences between these women and those who did complete all sessions.

Study Sample

The sample in the present analysis consisted of 336 HIV-negative African American female drug users. Ages of respondents ranged from 18 to 59 years, with a mean age of 38 years (SD = 7.5 years). Over one half had less than a complete high school education (53%), and approximately 38% were married (including common law) at the time of baseline assessment. Lifetime abusive experiences were common among the women, with approximately 72% reporting any lifetime physical abuse, 60% sexual abuse, and 87% emotional abuse. Only 13% reported never being in jail or prison, and slightly less than half (48%) worked legally in the year prior to baseline interview. In total, 79% were primarily crack cocaine smokers, 8% were injection drug users (IDUs) who did not smoke crack, and 13% were IDUs who also smoked crack.

Measures

Victimization and abuse. This construct was assessed with three items capturing physical, sexual, and emotional abuse. Lifetime rates for each of these types of abuse as well as for simultaneous abuse were examined descriptively.

Physical abuse included whether the respondent was physically abused, was attacked with a gun, knife, stick, bottle, or other weapon, or was hurt to the point that she had bruises, cuts, or broken bones in the past 6 months (yes/no). Sexual abuse assessed whether the respondent was sexually abused or was pressured or forced to participate in sexual acts against her will in the past 6 months (yes/no). If the respondent was abused emotionally in the past 6 months (yes/no), we counted it as emotional abuse. We also examined all three facets of abuse simultaneously at baseline, capturing whether the respondent experienced any type of victimization versus none in the previous 6 months (yes/no).

Independent variables other than intervention assignment (enhanced vs. standard) included sociodemographic, sexual, drug using, psychosocial, and criminal characteristics.

Sociodemographic characteristics. These included age at baseline (years); education (< high school diploma, high school diploma or GED, at least some college); marital status (married or living as married vs. all others); children under age 18 (yes/no); pregnancy status (pregnant vs. not pregnant); unstable living situation (yes/no for living on the streets or in a car, shelter, boarding house, or hotel vs. in own home, someone else's home, dormitory, or public housing); lifetime prison or jail time (for never incarcerated vs. ever); and whether the respondent worked legally in the past year (yes/no). (1)

Drug using behaviors. We measured a number of drug use behaviors including types of drug use or routes of drug administration, such as reported crack smoking (CCU only), reported injection drug use (IDU only), and both injection drug use and crack smoking (CCU/IDU). We also measured the frequency of illegal drug use (i.e., crack cocaine, powder cocaine, heroin, and/or speedball) in the past 30 days, coded as 0 = not in the last 30 days, 1 = in the last 30 days but not the last week, and 2 = within the past week. We assessed whether participants used drugs in a public setting (yes/no) in the past 30 days, defined as use outside, in a public restroom, or in a car. Finally, we asked about the number of individuals in the respondent's drug using network in the past 30 days.

Sexual behavior and characteristics. In this category we measured sex bartering in the past 30 days, defined as trading sex for drugs or money (yes/no for both). We also asked if participants had sex with an IDU in the past 30 days (yes/no), what their total sex frequency was in the last 30 days (total number of times vaginal, oral, or anal sex was had with a steady, casual nonpaying, or paying male partner), and the number of paying male sex partners in the past 30 days. Additional measures were mean male condom use for vaginal sex with a steady, casual nonpaying, or paying male sex partner in the last 30 days (never to always for each partner type), and whether a casual partner refused condom use in the past 30 days (yes/no).

Criminality. We measured criminality as the number of offensive acts (violent and nonviolent) committed in the past 6 months, including behaviors such as stealing property, shoplifting, physical assault, threatening someone with a weapon, or selling, distributing, or helping to make drugs, among others. The 15 collected dichotomous items were summed to obtain a construct of overall criminality ranging from 0 to 15 (K[R.sub.20] = .90).

Psychosocial measures. We examined sensation seeking or low self-control, anxiety symptoms, and self-esteem. Thirteen summed dichotomous items on thrill and adventure seeking and hostility measured low self-control. Higher scores represent higher levels of low self-control (K[R.sub.20] = .84). We derived items assessing self-control from Zuckerman's scale (Zuckerman, Kuhlman, Thornquist, & Kiers, 1991) but modified them slightly for the present research to exclude alcohol or other drug items (Huba, Newcomb, & Bentler, 1981). We derived indictors of anxiety from Beck's Brief Symptom Inventory (Derogatis & Melisaratos, 1983); the composite item representing overall anxiety ranged from 0 to 10 (Cronbach's alpha = .80). Our measure of self-esteem was based on 10 items from Rosenberg's (1965) Self-Esteem Scale, used to measure general feelings of self-worth. Scores ranged from 1 to 40, with higher scores representing higher levels of reported self-esteem (Cronbach's alpha = .78).

Statistical Analyses

To assess significant bivariate associations between the independent measures and reported victimization at baseline, we performed log likelihood chi-square or Fisher's Exact test of independence and the Mann-Whitney U test or one-way ANOVA where appropriate. Second, we used generalized estimating equations (GEE) logistic regression models to identify significant changes in reported physical, sexual, and emotional abuse over time and to identify factors associated with abuse. This method allows for intrasubject correlation, often treated as nuisance parameters, among repeated measurements on the same subject, accounting for correlation due to repeated observations (Zeiger & Liang, 1986). We used the SAS procedure PROC GENMOD application for estimation and testing in the GEE framework.

We calculated crude logistic regression (GEE) models for each independent variable. All crude models controlled for baseline abuse and the effect of time. Lifetime abuse was not included due to its strong collinear relationship with new (past 6 months) abuse (r = .84). Significant variables at [alpha] = 0.10 identified in bivariate analyses were considered for inclusion in multivariate analysis. We used a backward elimination approach to develop a best-fit multivariate regression model using variables identified as important contributors in bivariate analysis (Kleinbaum, 1994). Adjusted [chi square] estimates, odds ratios, and 95% confidence intervals were used to identify factors significantly (p < 0.05) associated with change in abuse or victimization. We considered it confounding when exclusion of a variable (or combination of variables) in the initial multivariate models resulted in a change of more than 10% in estimates of key factors (e.g., drag use) already present in the model. Due to the potential relationships between drug use and sexual behaviors, psychosocial characteristics, and the intervention assignment, we decided to test for interaction between these items and abuse.

Because both enhanced intervention conditions focused on victimization experiences and provided potential tools to prevent or handle these experiences, we contrasted the enhanced interventions with the NIDA standard in all analyses.

RESULTS

Victimization at Baseline

Characteristics of respondents according to victimization history are presented in Table 1. More than three fourths (77%) of the sample reported a past victimization, referring to victimization during the 6 months prior to enrolling in the study. In terms of past victimization, 53% reported physical abuse, 67% sexual abuse, and 67% emotional abuse: approximately 43% reported being abused physically, sexually, and emotionally; 45% physically and sexually: 44% physically and emotionally: and 62% sexually and emotionally.

Past victimization was significantly more common among women who were younger, on average. Younger women were also more likely to be in an unstable living situation (27.5% vs. 17.4%, p < 0.05) and to have been in jail or prison at some time during their lives (p < 0.001). Criminal involvement was significantly higher (p < 0.001) among women reporting past victimization than among those who lacked such a history. Crack cocaine users reported the highest level of victimization. Furthermore, low self-control (p < 0.001) and lower self-esteem (p < 0.05) were more likely to be reported among those who reported past victimization.

Victimization Over Time

When we compared past victimization with new victimization, we found a significant decrease (p < 0.001) in the probability of new victimization (OR = 0.72, 95% CI = 0.63-0.83). This applied to all three forms of victimization: physical (OR = 0.61, 95% CI = 0.51-0.72; p < 0.001), sexual (OR = 0.75, 95% CI = 0.65-0.85: p < 0.001), and emotional (OR = 0.83, 95% CI = 0.73-0.95; p < 0.01).

The results of analyses examining demographic and individual characteristics associated with new victimization are presented in Table 2. Increased age was significantly (p < 0.01) associated with decreased likelihood of all forms of abuse. Women who had given birth were significantly more likely to experience physical and sexual abuse (p < 0.001 for each), and they showed a trend toward more emotional abuse (p < 0.10). Pregnant women (who reported being pregnant at baseline or during the study) were nearly 2.5 times more likely to have experienced sexual abuse than their nonpregnant counterparts (OR = 2.45, p < 0.001), and we observed a trend for more physical abuse among pregnant respondents. An unstable living situation also contributed to new victimization, as did criminal involvement (p < 0.005 for each). Those who reported lower self-control and more symptoms of anxiety were also more likely (p < 0.001) to report all forms of new victimization, compared to women who reported higher levels of self-control and lower levels of anxiety (p < 0.001). We observed a trend for women with a greater social support system to have a lower probability of new victimization (p < 0.10).

Table 3 presents the results of crude GEE logistic regression analyses examining drug using and sexual characteristics associated with all forms of new victimization. New victimization differed by drug use pattern: Compared to users who only smoke crack cocaine and those who only inject drugs, those who inject drags and smoke crack cocaine were more likely to report new victimization experiences (p < 0.001). The frequency of illegal drug use also was significantly associated with all forms of abuse, with respondents who continued to use more frequently having an increased likelihood of victimization (see Table 3). Women who used drugs in a public setting were 2.5 times more likely to experience physical or sexual abuse than those who did not (p < 0.001 for each) and they were also more likely to be abused emotionally. The likelihood of emotional, physical, and sexual abuse was significantly higher among women whose social network largely consisted of other users, who traded sex for drugs or money, who had sex with an injection drug user, or who had sex more frequently (see Table 3). Furthermore, there was less physical and sexual victimization among those who used condoms more frequently with casual (p < 0.01) and steady (p < 0.05) male partners.

Table 4 presents the results of the final multivariate GEE logistic regression models predicting new victimization. After taking into account additional covariates, we found that the enhanced intervention conditions were significantly associated with new sexual and emotional abuse experiences. New victimization also was significantly and negatively associated with psychosocial characteristics such as increased self-esteem, lower anxiety, and higher levels of self-control. Furthermore, the enhanced intervention conditions emerged as an important confounder in the multivariate models for all forms of new abuse, including physical.

As expected, participants who used illegal drugs more frequently were more likely to experience physical victimization (aOR = 1.29, 95% CI = 1.10-1.52; p < 0.01), as were those who used drugs in a public setting (p < 0.05) or lived in an unstable situation (p < 0.01). Conversely, women who reported higher levels of male condom use for vaginal sex were significantly less likely to be abused physically (aOR = 0.87, 95% CI = 0.79-0.95; p < 0.01). Women who reported more criminal activity were significantly (p < 0.05) more likely to report physical victimization than those who reported less criminal activity. Additional factors significantly (p < 0.05) associated with physical abuse in the final model included age and low self-control.

With respect to sexual abuse, women in the enhanced intervention conditions were significantly less likely to report new victimization (aOR = 0.71, 95% CI = 0.52-0.97; p < 0.05). Greater self-esteem (aOR = 0.95, 95% CI = 0.92-0.98; p < 0.001) and more frequent male condom use (aOR = 0.90, 95% CI = 0.82-0.99; p < 0.05) were also associated with less sexual abuse. Participants with lower sell-control (p < 0.001) and an increased frequency of illegal drug use (p < 0.05) were more likely to report sexual abuse. Along with age, these sexual behaviors were also important confounders in the final multivariate model for sexual abuse.

For each type of abuse, all interaction terms remained insignificant in the examined multivariate models. As expected, past victimization was strongly and positively associated with new victimization for all types of abuse.

DISCUSSION

A central goal of this study was to determine if participation in an HIV risk-reduction intervention that also considered the context of the participating women's daily lives reduced their victimization experiences during the 6 months following their completion of the intervention. Results showed a significant decline in new victimization experiences (p < 0.001). Emotional abuse fell from 67% to 58% (p < 0.01), sexual abuse from 67% to 53% (p < 0.001), and physical abuse from 53% to 32% (p < 0.001).

After taking into account other factors, we found that the enhanced interventions were associated with fewer new victimization experiences, particularly sexual and emotional abuse. Due to the nature of the enhanced interventions, women in these conditions were expected to improve their negotiation and conflict resolution skills, to advance their motivation to decrease or cease risky drug using or sexual behaviors, to focus on enhancing their general health and quality of life, and to fare better psychologically. The precise relationship between such factors, HIV risk, and victimization are difficult to determine from this study and in general. However, we can hypothesize that such skills may aid in decreasing the likelihood of abuse.

For example, emotional and psychological factors can complicate a person's decision to use condoms. Many women face the risk of abuse if condom use is proposed and especially if the woman insists (Koenig & Moore, 2000). Women in our study who reported a greater frequency of condom use for vaginal sex were significantly less likely to report physical and sexual abuse, even after we controlled for important covariates. This may be because the women also had developed conflict resolution and negotiation skills beyond situations of persuading a partner to use a condom. It is also important to recognize that women who do not report condom use may not use condoms because of recurrent victimization by a resistant sex partner.

Even after we controlled for other covariates, we found that low self-control and lifestyle factors (Jensen & Brownfield, 1986)--such as criminal behavior (including expressed violence), drug use in a public setting, and living in an unstable situation--were both positively and significantly associated with physical victimization. Low self-control was also significantly associated with emotional and sexual abuse, again with the exact relationship being difficult to determine. Does abuse lead to lower self-control, or are women with lower levels of self-control more likely to put themselves in abusive situations?

As expected, drug use was significantly associated with victimization and remained associated with physical and sexual abuse after other covariates were taken into account. As with other salient factors associated with victimization, the exact relationship between drug use and received victimization should be determined. However, the direction of the relationship is likely to be nonrecursive: Victimization leads to continued drug use and continued drug use leads to repeated victimization, especially physical abuse.

Despite important findings, limitations to our study should be considered. Our data came from self-reports, which may be affected by recall, social desirability, or additional types of bias. However, the types of drug using and sexual behaviors measured here have been consistently reported as both valid and reliable (Dowling-Guyer, Johnson, Fisher, & Needle, 1994; Weatherby et al., 1994). Researchers have also demonstrated the validity and consistency in such measures over time among individuals who use drugs (Adair, Craddock, Miller, & Turner, 1996). We used a past-30-day time frame to minimize the possibility of recall bias. Nonetheless, the method of recruitment (due to an unknown population of drug users), the small sample size, the global measures of abuse, and the cultural and geographic characteristics of this sample limit the generalizability of findings to a wider population of women.

Our overall findings suggest that culturally appropriate, woman-tailored, and theoretically based interventions may be effective at enhancing not only HIV preventive behavior among African American women who use illegal drugs, but also the likelihood of decreased victimization. Teaching women to stand up for themselves, to solve con flict, and to negotiate to achieve their goals should not be limited to main HIV risk behaviors such as unsafe drug use and sex. Instead, women should be encouraged to apply these skills to all aspects of their lives. The high prevalence of past and new victimization among these women is only one indication of the continued need to incorporate violence prevention measures in behavioral and other risk-reduction interventions. HIV intervention programs targeting women who use drugs may benefit from including strategies that strengthen sexual negotiation skills, including partner support, as well as structural factors that may reduce the likelihood of victimization through improved lifestyle factors.

Table 1. Characteristics of Participants at Baseline (N = 336)

                                           Victimization in past
                                                  6 months

                                              Yes %           No %
                                            (n = 258)       (n = 78)

Enhanced intervention condition (yes)      58.3           55.8
Lifetime physical abuse (yes)              74.2           25.8
Lifetime sexual abuse (yes)                86.6           10.5
Lifetime emotional abuse (yes)             88.5            1.2

Age category
  18-29                                    14.3            7.0
  30-39                                    48.2           39.5
  40+                                      37.6           53.5
  Mean age ([+ or -] SD)                   37.3           39.6
                                           [+ or -] 7.5   [+ or -] 7.3
Education
  < high school diploma                    51.9           56.6
  High school diploma / GED                33.6           30.1
  Some college or college degree           14.5           13.3
Married or living as married (yes)         37.3           40.7
Children (yes)                             89.5           90.4
Pregnant (yes)                              4.5            2.3
Unstable living situation (yes) (b)        27.5           17.4
Worked legally in last year (yes)          49.8           42.0

Never in prison/jail (yes)                 9.3            24.4
Criminality (past 6 months): mean          4.1            2.1
  violent and nonviolent acts              [+ or -] 3.0   [+ or -] 2.3
    ([+ or -] SD)
Type and route of illegal drug use,
    past 30 days (c)
  Crack cocaine smoking but no
    injection                              81.4           72.1
  Injection drug use but no crack           3.2           12.8
    smoking
  Injection drug use and crack cocaine     15.4           15.1
    smoking
Frequency of illegal drug use in last
    30 days (c)
  Not in past 30 days                       0.0            0.0
  In past 30 days, but not in past week    54.7           60.5
  In past week / last 7 days               45.3           39.5

Mean level of self-esteem ([+ 0r -] SD)    22.6           26.3
                                           [+ or -] 5.8   [+ or -] 5.5
Mean level of low self-control /           4.4            2.9
  hostility ([+ or -] SD)                  [+ or -] 2.1   [+ or -] 2.0

                                             p value (a)

Enhanced intervention condition (yes)           ns
Lifetime physical abuse (yes)                   ***
Lifetime sexual abuse (yes)                     ***
Lifetime emotional abuse (yes)                  ***

Age category                                    *
  18-29
  30-39
  40+
  Mean age ([+ or -] SD)                        **
Education                                       ns
  < high school diploma
  High school diploma / GED
  Some college or college degree
Married or living as married (yes)              ns
Children (yes)                                  ns
Pregnant (yes)                                  ns
Unstable living situation (yes) (b)             *
Worked legally in last year (yes)               ns

Never in prison/jail (yes)                      ***
Criminality (past 6 months): mean               ***
  violent and nonviolent acts
    ([+ or -] SD)
Type and route of illegal drug use,             **
    past 30 days (c)
  Crack cocaine smoking but no
    injection
  Injection drug use but no crack
    smoking
  Injection drug use and crack cocaine
    smoking
Frequency of illegal drug use in last
    30 days (c)                                 ([dagger])
  Not in past 30 days
  In past 30 days, but not in past week
  In past week / last 7 days

Mean level of self-esteem ([+ or -] SD)         **

Mean level of low self-control /                ***
  hostility ([+ or -] SD)

(a) p value from likelihood ratio chi-square or Fisher's exact, or
Mann-Whitney U test or one-way ANOVA test statistics where
appropriate. (b) Defined as living on streets or in car, shelter,
rooming or boarding house, or hotel. (c) Illegal drug use includes
crack cocaine, powder cocaine, heroin, or speed-ball (cocaine in any
form and heroin combined simultaneously or back-to-back). Risky drug
using setting defined as use in crack house, shooting gallery,
public restroom, outside, or in car.

* p <.05. ** p < .01. *** p < .001. ([dagger]) p <.10.

Table 2. Demographic and Individual Characteristics Associated With
Victimization Over Time: Crude GEE Logistic Regression Analysis

                                   Crude odds ratio (95% CI)
                                         Emotional abuse

Enhanced intervention (yes)        0.84 (0.59 - 1.09)
Age (years)                        0.97 (0.94 - 0.99) **
Education                          1.10 (0.85 - 1.43)
Married (yes)                      1.12 (0.78 - 1.62)
Children (yes)                     1.31 (0.97 - 1.78) ([dagger])
Pregnant (yes)                     1.15 (0.60 - 2.21)
Unstable living situation (yes)    1.40 (1.03 - 1.91) *
Work legally, past year (yes)      1.10 (0.78 - 1.56)
Crimes                             1.13 (1.07 - 1.20) ***
Self-esteem (1-40)                 0.92 (0.90 - 0.95) ***
Anxiety (0-10)                     1.28 (1.19 - 1.38) ***
Low self-control (0-13)            1.32 (1.21 - 1.44) ***
Support system                     0.95 (0.89 - 1.02)

                                   Crude odds ratio (95% CI)
                                         Physical abuse

Enhanced intervention (yes)        0.97 (0.76 - 1.15)
Age (years)                        0.96 (0.95 - 0.98) ***
Education                          0.97 (0.80 - 1.17)
Married (yes)                      1.03 (0.78 - 1.36)
Children (yes)                     1.77 (1.31 - 2.40) ***
Pregnant (yes)                     1.88 (0.98 - 3.62) ([dagger])
Unstable living situation (yes)    1.79 (1.35 - 2.38) ***
Work legally, past year (yes)      0.92 (0.71 - 1.21)
Crimes                             1.12 (1.08 - 1.17) ***
Self-esteem (1-40)                 0.96 (0.94 - 0.98) ***
Anxiety (0-10)                     1.16 (1.11 - 1.21) ***
Low self-control (0-13)            1.23 (1.16 - 1.30) ***
Support system                     0.97 (0.92 - 1.03)

                                   Crude odds ratio (95% CI)
                                           Sexual abuse

Enhanced intervention (yes)        0.93  (0.69 - 1.03) ([dagger])
Age (years)                        0.96  (0.94 - 0.98) ***
Education                          0.98  (0.79 - 1.21)
Married (yes)                      0.96  (0.70 - 1.31)
Children (yes)                     1.76  (1.27 - 2.43) ***
Pregnant (yes)                     2.45  (1.31 - 4.56) ***
Unstable living situation (yes)    1.55  (1.13 - 2.11) **
Work legally, past year (yes)      1.17  (0.87 - 1.59)
Crimes                             1.11  (1.05 - 1.17) ***
Self-esteem (1-40)                 0.94  (0.92 - 0.96) ***
Anxiety (0-10)                     1.24  (1.17 - 1.32) ***
Low self-control (0-13)            1.32  (1.23 - 1.42) ***
Support system                     0.96  (0.90 - 1.01) ([dagger])

Note. All models controlled for baseline abuse and effect of time.
Education = less than high school diploma to some college or
college degree. Crimes = number of violent and nonviolent criminal
acts in last 6 months. Self-esteem = level of self-esteem.
Anxiety = level of anxiety. Low self-control = level of self-control,
reverse coded. Support system = number of people (e.g., family,
friends) to turn to when there is a problem or when in trouble.

* p < .05. ** p < .01. *** p < .001. ([dagger]) p<.10.

Table 3. Drug-Using and Sexual Characteristics Associated With
Victimization Over Time: Crude GEE Logistic Regression Analysis

                                          Crude odds ratio (95% CI)
                                               Emotional abuse

Type/route of drug administration
  rack cocaine smoking only               1.04 (0.61, 1.76)
  IDU, no crack smoking                   0.67 (0.44, 1.01) ([dagger])
  IDU and crack smoking                   1.00
Frequency of illegal drug use             1.25 (1.04 - 1.51) *
Drug use in public setting (yes)          1.39 (1.03 - 1.88) *
Drug using network                        1.20 (0.97 - 1.48) ([dagger])
Sex for money (yes)                       1.68 (1.25 - 2.28) ***
Sex for drugs (yes)                       1.46 (1.13 - 1.88) ***
Sex with an IDU (yes)                     2.00 (0.95 - 4.23) ([dagger])
Total sex frequency                       1.02 (1.00 - 1.04
Condoms--casual partner(s)                0.73 (0.56 - 0.95) *
Condoms--paying partner(s)                0.98 (0.79 - 1.21)
Condoms--steady partner(s)                0.97 (0.88 - 1.08)
Mean vaginal male condom use              1.00 (0.92 - 1.09)
Casual partner refused condom use (yes)   1.32 (0.46 - 3.76)

                                          Crude odds ratio (95% CI)
                                               Physical abuse

Type/route of drug administration
  rack cocaine smoking only               0.41 (0.23 - 0.71) ***
  IDU, no crack smoking                   0.17 (0.10 - 0.33) ***
  IDU and crack smoking                   1.00
Frequency of illegal drug use             1.86 (1.47 - 2.36) ***
Drug use in public setting (yes)          2.52 (1.87 - 3.41) ***
Drug using network                        1.31 (1.14 - 1.52) ***
Sex for money (yes)                       2.13 (1.60 - 2.83) ***
Sex for drugs (yes)                       3.71 (2.74 - 5.02) ***
Sex with an IDU (yes)                     3.31 (1.62 - 6.77) ***
Total sex frequency                       1.04 (1.03 - 1.06) ***
Condoms--casual partner(s)                0.83 (0.46 - 1.50)
Condoms--paying partner(s)                0.87 (0.54 - 1.40)
Condoms--steady partner(s)                0.91 (0.82 - 1.00) ([dagger])
Mean vaginal male condom use              0.95 (0.87 - 1.00) ([dagger])
Casual partner refused condom use (yes)   1.49 (0.61 - 3.65)

                                          Crude odds ratio (95% CI)
                                               Sexual abuse

Type/route of drug administration
  rack cocaine smoking only               0.75 (0.42 - 1.35)
  IDU, no crack smoking                   0.45 (0.26 - 0.79) **
  IDU and crack smoking                   1.00
Frequency of illegal drug use             1.57 (1.28 - 1.93) ***
Drug use in public setting (yes)          2.55 (1.89 - 3.45) ***
Drug using network                        1.28 (1.08 - 1.52) ***
Sex for money (yes)                       2.27 (1.63 - 3.16) ***
Sex for drugs (yes)                       2.69 (1.96 - 3.68) ***
Sex with an IDU (yes)                     2.31 (1.06 - 4.99)
Total sex frequency                       1.05 (1.02 - 1.09) ***
Condoms--casual partner(s)                0.69 (0.54 - 0.89) **
Condoms--paying partner(s)                0.92 (0.13 - 1.16)
Condoms--steady partner(s)                0.90 (0.81 - 0.99) *
Mean vaginal male condom use              0.95 (0.87 - 1.00) ([dagger])
Casual partner refused condom use (yes)   1.06 (0.39 - 2.90)

Note. All models controlled for baseline abuse and effect of time.
Illegal drug use includes crack cocaine, powder cocaine, heroin, or
speedball (cocaine in any form and heroin combined simultaneously or
back-to-back). Frequency of illegal drug use = not in past 30 days
to in the past week. Drug using network = number of people drugs are
used or shared with. Total sex frequency = total number of times
vaginal, oral, or anal sex (regardless of partner type). Condom use
includes male condom use for vaginal sex with steady, casual
nonpaying, or casual paying sex partners (never to always).

* p < .05. ** p < .01. *** p < .001. ([dagger]) p < .10.

Table 4. Factors Associated With Victimization Over Time: Final GEE
Logistic Regression Models

                                   Adjusted odds ratio (95% CI)
                                        Emotional abuse

Enhanced intervention (yes)        0.67 (0.46, - 0.98) *
Age (years)                        0.97 (0.94  - 0.99) **
Unstable living situation (yes)            --
Crimes                                     --
Self-esteem                        0.96 (0.92 - 0.98) **
Anxiety                            1.17 (1.07 - 1.30) ***
Low self-control                   1.12 (1.01 - 1.25) *
Frequency of illegal drug use              --
Drug use in public setting (yes)           --
Sex for money (yes)                        --
Total sex frequency                        --
Mean vaginal male condom use               --

                                   Adjusted odds ratio (95% CI)
                                         Physical abuse

Enhanced intervention (yes)        0.83 (0.60 - 1.10)
Age (years)                        0.97 (0.95 - 0.99) *
Unstable living situation (yes)    1.58 (1.15 - 2.18) **
Crimes                             1.05 (1.03 - 1.10) *
Self-esteem                                --
Anxiety                                    --
Low self-control                   1.17 (1.06 - 1.19) **
Frequency of illegal drug use      1.29 (1.10 - 1.52) **
Drug use in public setting (yes)   1.47 (1.03 - 2.10) *
Sex for money (yes)                        --
Total sex frequency                        --
Mean vaginal male condom use       0.87 (0.79 - 0.95) **

                                   Adjusted odds ratio (95% CI)
                                         Sexual abuse

Enhanced intervention (yes)        0.71 (0.52 - 0.97) *
Age (years)                        0.98 (0.96 - 1.00) ([dagger])
Unstable living situation (yes)            --
Crimes                                     --
Self-esteem                        0.95 (0.92 - 0.98) ***
Anxiety                                    --
Low self-control                   1.19 (1.10 - 1.30) ***
Frequency of illegal drug use      1.21 (1.04 - 1.42) *
Drug use in public setting (yes)           --
Sex for money (yes)                1.43 (0.95 - 2.15) ([dagger])
Total sex frequency                1.03 (1.00 - 1.05) ([dagger])
Mean vaginal male condom use       0.90 (0.82 - 0.99) *

Note. All models controlled for baseline abuse and effect of time.
Crimes = number of violent and nonviolent criminal acts in last
6 months. Self-esteem = level of self-esteem. Anxiety = level of
anxiety. Low self-control = level of self-control, reverse coded.
Frequency of illegal drug use = not in past 30 days to in the past
week. Total sex frequency = total number of times vaginal, oral,
or anal sex (regardless of partner type). Mean male condom use
includes average male condom use for vaginal sex with steady,
casual nonpaying, or casual paying sex partners (never to always).

* p < .05. ** p < .01. *** p < .001. ([dagger]) p < .10.

(1) Income was difficult to assess in the present analysis. Qualitative data revealed that the women's income fluctuated greatly and that communicating weekly and yearly income was difficult for most respondents. Distinguishing between legal and illegal income (but not legal and illegal forms of employment or work) was problematic. Therefore, we excluded income in the present analysis.

REFERENCES

Adair, E. B., Craddock, S. G., Miller, H. G., & Turner, C. F. (1996). Quality of treatment data: Reliability over time of self-reports given by clients in treatment for substance abuse. Journal of Substance Abuse Treatment, 13, 145-149.

Ajzen, I., & Madden, T. J. (1986). Prediction of goal-directed behavior: Attitudes, intentions, and perceived behavioral control. Journal of Experimental Social Psychology, 22, 453-474.

Bandura, A. (1977). Self-efficacy. Psychological Review, 84, 191-215.

Connell, R. W. (1985). Theorising gender. Sociology, 19, 260-272.

Cottler, L. B., Leukefeld, C., Hoffman, J., Desmond, D., Wechsberg, W., Inciardi, J. A., et a1.(1998). Effectiveness of HIV risk reduction initiatives among out-of-treatment non-injection drug users. Journal of Psychoactive Drugs, 30, 279-290.

Coyle, S. L. (1998). Women's drug use and HIV risk: Findings from NIDA's Cooperative Agreement for Community-Based Outreach/Intervention Research Program. Women & Health, 27, 1-18.

Dancy, B. (2000). HIV risk reduction strategies for low-income African American women. Nurse Practitioners Forum, 11, 109-115.

Derogatis, L., & Melisaratos, N. (1983). The Brief Symptom Inventory: An introductory report. Psychological Medicine, 13, 595-605.

Dowling-Guyer, S., Johnson, M. E., Fisher, D. G., & Needle, R. (1994). Reliability of drug users' self-reported HIV risk behaviors and validity of self-reported recent drug use. Assessment, 1, 383-392.

Ehrhardt, A. A., & Exner, T. M. (2000). Prevention of sexual risk behavior for HIV infection with women. Aids, 14(Suppl. 2), S53-58.

El-Bassel, N., Simoni, J. M., Cooper, D. K., Gilbert, L., & Schilling, R. F. (2001). Sex trading and psychological distress among women on methadone. Psychology of Addictive Behaviors, 15(3), 177-184.

Elifson, K. W., & Sterk, C. E. (2003). Case study: The Health Intervention Project. In D. Blumenthal & R. DiClemente (Eds.), Community-based research: Issues and methods (pp. 153-170). New York: Springer.

Fishbein, M., & Middlestadt, S. E. (1989). Using the theory of reasoned action as a framework for understanding and changing AIDS-related behaviors. In J. N. Wasserheit (Ed.), Primary prevention of AIDS: Psychological approaches (pp. 93-110). Thousand Oaks, CA: Sage Publications.

Huba, G., Newcomb, M., & Bentler, P. (1981). Comparison of canonical correlation and interbattery factor analysis on sensation seeking and drug use domains. Applied Psychological Measurement, 5, 291-306.

Ickovics, J. R., & Yoshikawa, H. (1998). Preventive interventions to reduce heterosexual HIV risk for women: Current perspectives, future directions. Aids, 12(Suppl. A), S197-208.

Jensen, G. F., & Brownfield, D. (1986). Gender, lifestyles, and victimization: Beyond routine activity. Violence & Victims, 1(2), 85-99.

Kleinbaum, D. (1994). Modeling strategy guidelines. In G. Dietz, K. Krickeberg, & B. Singer (Eds.), Logistic regression: A self-learning text (pp. 161-189). New York: Springer-Verlag.

Koenig, L. J., & Moore, J. (2000). Women, violence and HIV: A critical evaluation with implications for HIV services. Maternal and Child Health Journal, 4, 103-109.

Maman, S., Campbell, J., Sweat, M. D., & Gielen, A. C. (2000). The intersections of HIV and violence: Directions for future research and interventions. Social Science and Medicine, 50, 459-478.

McKusker, J., Goldstein, R., Bigelow, C., & Zorn, M. (1995). Psychiatric status and HIV risk reduction among residential drug abuse treatment clients. Addiction, 90, 1377-1387.

Miller, M., & Neaigus, A. (2001). Networks, resources and risk among women who use drugs. Social Science and Medicine, 52, 967-978.

Morrill, A. C., Kasten, L., Urato, M., & Larson, M. J. (2001). Abuse, addiction, and depression as pathways to sexual risk in women and men with a history of substance abuse. Journal of Substance Abuse, 13, 169-184.

National Institute on Drug Abuse. (1992). The standard intervention of the cooperative agreement program for AIDS community-based outreach/intervention research. Washington, DC: Government Printing Office.

Nyamathi, A., Bayley, L., Anderson, N., Keenan, C., & Leake, B. (1999). Perceived factors influencing the initiation of drug and alcohol use among homeless women and reported consequences of use. Women & Health, 29, 99-114.

Prochaska, J., & DiClemente, C. (1983). Stages and processes of self-change of smoking. Journal of Consulting and Clinical Psychology, 51, 390-395.

Rosenberg, M. (1965). Society and the adolescent self-image. Princeton, NJ: Princeton University Press.

Sampson, R. J., & Lauritsen, J. L. (1990). Deviant lifestyles, proximity to crime, and the offender-victim link in personal violence. Journal of Research in Crime & Delinquency, 27, 110-139.

Sloboda, Z. (1998). What we have learned from research about the prevention of HIV transmission among drug abusers. Public Health Reports, 113(Suppl. 1), 194-204.

Sterk, C. (1999). Fast lives: Women who use crack cocaine. Philadelphia: Temple University Press.

Sterk, C. (2002). The Health Intervention Project: HIV risk reduction among African American women drug users. Public Health Reports, 17, 88-95.

Sterk, C., Theall, K., & Elifson, K. (2003). Effectiveness of an HIV risk reduction intervention for African American women who use crack cocaine. AIDS Education & Prevention, 15, 15-32.

Sterk, C., Theall, K., Elifson, K., & Kidder, D. (2003). HIV risk reduction among African American female injection drug users: A randomized controlled trial. AIDS & Behavior, 7, 73-86.

Tashima, N., Crain, S., O'Reilly, K., & Sterk-Elifson, C. (1996). The community identification process: A discovery model. Qualitative Health Research, 6, 23-48.

Weatherby, N. L., Needle, R., Cesari, H., Booth, R. E., McCoy, C., Watters, J., et al. (1994). Validity of self-reported drug use among injection drug users and crack cocaine users recruited through street outreach. Evaluation & Program Planning, 17, 347-355.

Wechsberg, W. M., Craddock, S. G., & Hubbard, R. L. (1998). How are women who enter substance abuse treatment different than men?: A gender comparison from the drug abuse treatment outcome study (DATOS). Drugs & Society, 13, 1-2.

Wechsberg, W. M., Dennis, M. L., & Stevens, S. J. (1998). Cluster analysis of HIV intervention outcomes among substance-abusing women. American Journal of Drug & Alcohol Abuse, 24, 239-257.

Wilsnack, S. C., Vogeltanz, N. D., Klassen, A. D., & Harris, T. R. (1997). Childhood sexual abuse and women's substance abuse: National survey findings. Journal of Studies on Alcohol, 58, 264-271.

Zeiger, S. L., & Liang, K. Y. (1986). Longitudinal data analysis for discrete and continuous outcomes. Biometrics, 42, 121-130.

Zuckerman, M., Kuhlman, M., Thornquist, M., & Kiers, H. (1991). Five (or three) robust questionnaire scale factors of personality without culture. Personality & Individual Differences, 12, 929-941.

Manuscript accepted June 4, 2004

Katherine P. Theall and Claire E. Sterk

Emory University

Kirk W. Elifson

Georgia State University

This research was supported by NIDA grant R01 DA-10642 and the Emory Center for AIDS Research. The views presented in this paper are those of the authors and do not represent those of the funding agencies.

Address correspondence to Claire E. Sterk, Emory University, Rollins School of Public Health. 1518 Clifton Road N.E., Atlanta, GA 30322: e-mail: [e-mail]