Childhood Maltreatment Histories, Alcohol and Other Drug Use Symptoms, and Sexual Risk Behavior in a Treatment Sample of Adolescents

Abstract

Objectives. We tested a structural model of relations among self-reported childhood maltreatment, alcohol and other drug abuse and dependence symptoms, and sexual risk behavior in a sample of adolescents receiving outpatient treatment of substance use problems.

Methods. Structured interviews were administered to an ethnically diverse sample of 394 adolescents (114 girls, 280 boys; mean = 16.30 years; SD = 1.15 years; 44.9% Hispanic, 20.6% African American, 25.4% White non-Hispanic, and 9.1% other) in 2 outpatient treatment settings.

Results. Path analyses yielded findings consistent with a mediation model. Alcohol abuse and dependence symptoms mediated (1) relations between emotional neglect scores and sex with co-occurring alcohol use and (2) relations between sexual abuse scores and sex with co-occurring alcohol use. Drug abuse and dependence symptoms mediated relations between (1) neglect scores and (2) sexual intercourse with co-occurring alcohol or drug use, as well as unprotected sexual intercourse.

Conclusions. Efforts to treat alcohol or drug use problems among adolescents or to prevent transmission of HIV or other sexually transmitted infections among youths with substance use problems may require tailoring treatment or prevention protocols to address client histories of maltreatment.

Sexual risk behaviors among adolescents are of great concern to public health. According to Centers for Disease Control and Prevention¹ estimates, high-risk sexual contact was responsible for approximately 30% of reported HIV infections in 2007, and in 2008, roughly half a million cases of sexually transmitted infections (STIs) occurred among youths aged 15 to 19 years.² Therefore, understanding the distal and proximal factors associated with sexual risk behavior has profound implications for the design of successful intervention programs to reduce behavioral risk factors for STIs.

Significant associations exist between the use of alcohol and other drugs and sexual risk behaviors among adolescents.^3–5 Adolescents receiving treatment for substance use problems have reported earlier ages of onset of sexual activity, more sexual partners, and less consistent condom use than those without alcohol or drug use problems.^6–8 Alcohol and other drug use may pose an immediate risk factor as well; adolescents who engage in substance use before or during sex are less likely to use condoms or other forms of contraception than adolescents who do not engage in substance use.^6,9

Both sexual risk behaviors and alcohol or drug use problems are influenced by distal environmental risk factors, such as exposure to child abuse and neglect. Childhood maltreatment is a multidimensional construct represented by the type, duration, chronicity, and severity of abuse experiences, as well as features of the perpetrator and the specific acts perpetrated.¹⁰ Short- and long-term consequences of childhood maltreatment are well documented,^11–14 including increased risk for the development of internalizing and externalizing problems.^15,16 Childhood maltreatment has been shown to potentiate risk for the onset and development of substance use disorders among adolescents and young adults.^17–22 We tested a structural model of cross-sectional relations among 3 types of child maltreatment, past-year alcohol or drug abuse and dependence symptoms, and sexual risk behavior in a sample of adolescents receiving outpatient treatment for substance use problems.

Youths referred for substance use treatment often present with co-occurring problem behaviors, psychiatric disorders, and significant histories of childhood maltreatment.¹² Adolescents in clinic- and community-based samples who reported experiences of abuse and neglect engaged in higher levels of sexual risk behavior.^16,23 Early onset of sexual activity, multiple sex partners, unprotected sexual intercourse, and involvement in sex trading^{12,16,18,24–26} are separate indices of risk for HIV and STI exposure linked to childhood maltreatment. Less is known, however, regarding relations among the severity of specific types of childhood maltreatment experiences and indices of sexual risk behavior. For example, although childhood sexual abuse is significantly associated with increased sexual risk behavior across gender groups, relations between other types of maltreatment and sexual risk behavior outcomes are not consistent.²⁶

Several models have attempted to explain relations between concurrent alcohol or other drug use and sexual risk behavior. Alcohol myopia theory²⁷ posits that substance use negatively affects attentional capacity, such that intoxicated people attend primarily to the most salient environmental cues. Significant environmental cues that promote substance use and sex, such as partner's alcohol or other drug use, become more salient than inhibiting cues that discourage unprotected sex. Other models have sought to explain specific links between childhood sexual abuse and subsequent sexual risk behavior.^26,28 Miller and Mancuso²⁸ proposed a conceptual model that hypothesized that childhood sexual abuse promotes sexual risk behavior through 4 paths: victimized youths (1) use alcohol and other drugs as a coping strategy, (2) have higher rates of mental illness because of past abuse, (3) are more likely to be involved in risky social networks, and (4) have poor sexual adjustment. Given that childhood maltreatment often precedes the development of alcohol or other drug use problems and participation in sexual risk behavior during adolescence, one plausible question is whether alcohol or other drug abuse and dependence symptoms mediate relations between childhood maltreatment and sexual risk behavior during adolescence.²⁹ Cross-sectional analyses are an important first step toward informing future longitudinal studies that may use more stringent causal modeling strategies.^29,30 In this study, we used age, gender, and ethnicity as covariates on all of the modeled endogenous variables to control for their potential influence on paths in the mediation model.

Understanding the multivariate associations between types of child maltreatment, alcohol or other drug abuse and dependence symptoms, and sexual risk behavior can inform the design of prevention and treatment efforts for youths at risk for exposure to HIV and STIs. Using a path analysis framework, we evaluated relations among (1) 3 types of childhood maltreatment experiences (i.e., sexual abuse, physical punishment, and negative home environment and neglect), (2) past-year alcohol or drug abuse and dependence symptoms, and (3) 3 indicators of sexual risk behavior over the 180 days before the baseline interview. We hypothesized that (1) higher scores for childhood maltreatment experiences would be associated with higher scores for alcohol and drug abuse and dependence symptoms, and (2) higher scores for alcohol and drug abuse and dependence symptoms would be associated with higher scores for specific indices of sexual risk behavior. Using the path model, we tested the putative cross-sectional mediational role of alcohol or other drug abuse and dependence symptoms in relations between childhood maltreatment experiences and scores for recent sexual risk behavior among adolescents in outpatient treatment of substance use problems.

METHODS

This study is based on data collected in structured interviews of adolescents receiving outpatient substance abuse treatment services. All participants were sexually active, reported substance use and other health risk or problem behaviors. Data were collected at intake into an HIV and STI risk reduction intervention for adolescents with substance abuse problems from September 2004 to September 2008.

Participants and Procedure

The sample consisted of an ethnically diverse sample of adolescents (44.9% Hispanic, 20.6% African American, 25.4% White non-Hispanic, and 9.1% other; 71.1% male, 28.9% female) receiving substance use treatment services at 2 outpatient facilities in south Florida. Demographic variables are summarized in Table 1.

TABLE 1—

Demographic Information for the Sample of Adolescents (n = 394) Receiving Substance Abuse Treatment Services at 2 Outpatient Facilities in South Florida: September 2004—September 2008

Variable	No. (%)
Gender
Female	114 (28.9)
Male	280 (71.1)
Race/ethnicity
White non-Hispanic	100 (25.4)
Hispanic	177 (44.9)
African American	81 (20.6)
Other	36 (9.1)
Court mandated	120 (30.5)
Repeated grade	202 (51.3)
Father unemployed	84 (21.3)
Mother unemployed	92 (23.4)
Father born in United States	178 (45.2)
Mother born in United States	205 (52.0)

Adolescents were approached within 1 week of enrollment in outpatient substance use treatment services and invited to participate in the brief motivational HIV and STI risk reduction intervention from which data for this study were drawn. Clients were read the eligibility criteria and, if interested, were invited to contact a project staff member to have their eligibility confirmed and begin the informed consent process. Adolescents were excluded if they were not sexually active during the previous 6 months, did not provide assent in addition to parental consent, exhibited significant cognitive deficits or developmental delays (by case manager report), or reported current suicidality. Adolescents were excluded from participation for the last 2 criteria because of (1) the cognitive abilities required for the psychotherapeutic intervention delivered in the treatment arm of the larger study and (2) ethical concerns about client safety. Next, adolescents were assessed for Diagnostic and Statistical Manual of Mental Disorders (4th ed.)³¹ psychiatric symptoms and were administered a battery of questionnaires before being enrolled in the HIV and STI risk reduction intervention.

At entry, participants completed a 60- to 90-minute assessment that included measures of substance use, sexual risk behaviors, and demographics. Graduate students collected data using a computer-assisted structured interview at the facilities at which clients were receiving treatment services. Assent was obtained from adolescents along with active consent from a primary caregiver. Procedures were approved by the institutional review board. Participants were paid $25.00.

Measures

Psychiatric symptoms.

Lifetime and past-year Diagnostic and Statistical Manual of Mental Disorders (4th ed.) psychiatric symptoms and diagnoses were assessed via the brief University of Michigan version of the Composite International Diagnostic Interview,³² which is a computer-assisted structured interview administered by lay interviewers. The computerized delivery of items on the University of Michigan version included appropriate skip patterns and probe questions and did not allow out-of-range responses, thus simplifying the administration. This version of the Composite International Diagnostic Interview was developed to standardize the assessment of disorders in community settings and has excellent interrater reliability, good test–retest reliability, and sufficient concordance with clinical judgments and structured clinical interviews.^32–34

Timeline Follow Back for sexual risk behavior.

We modified the standard Timeline Follow Back instrument³⁵ to collect data regarding adolescents’ self-reported sexual risk behavior, including unprotected intercourse, number of partners, and co-occurring substance use and sexual behavior.³⁶ We used an adapted calendar format to assist in the recall of days when target sexual risk behaviors occurred. Participants completed this version of the Timeline Follow Back for the 180 days immediately before the baseline assessment. Similarly adapted Timeline Follow Back calendar methodology has been used in published research to assess sexual risk behavior among people with alcohol or drug use problems and among psychiatric inpatients^36,37 who have difficulty with self-report measures. Carey et al.³⁶ concluded that empirical evidence supported the feasibility, reliability, and validity of the Timeline Follow Back for the assessment of sexual risk behavior.

In this study, we used a 2-item latent construct for unprotected intercourse. This construct included the ratio of unprotected intercourse during the past 6 months (via the Timeline Follow Back) and a 5-point Likert item assessing the frequency of unprotected sex during the past 12 months, ranging from never (1) to always (5). A latent construct based on multiple measures was intended to provide a more robust and reliable measure of unprotected intercourse. This latent factor also included more than 1 method of collecting data on this variable. To make sure the latent factor did not change the results, we reran the model using 1 observed variable of unprotected intercourse. Results did not show significant statistical differences whether a latent construct or 1 observed variable was included in the model tested.

We used 2 additional items as indicators of sexual risk for HIV and STI exposure. Participants reported how often during the past 12 months they or a partner (1) drank alcohol before or during sex or (2) used any drugs to get high or intoxicated before or during sex. The response formats for these 2 items were always (5), usually (4), sometimes (3), rarely (2), or never (1). We included these items as recognized risk factors for HIV and STI exposure.⁹

Child Abuse and Trauma Scale.

The 38-item Child Abuse and Trauma Scale (CATS)³⁸ is a self-report inventory that assesses experiences of maltreatment during childhood or adolescence. Prior research has suggested that official records underestimate the true prevalence of child maltreatment, particularly among boys and young men.³⁹ The CATS includes subscales that assess 3 dimensions of childhood maltreatment: Neglect/Negative Home Environment, Punishment, and Sexual Abuse. Sample CATS items include, for Neglect/Negative Home Environment, “Were you left home alone as a child?” for Punishment, “Were you physically abused?” and for Sexual Abuse, “Have you experienced sexual abuse? The response format for CATS items was always (5), usually (4), sometimes (3), rarely (2), or never (1). We found among our sample that the CATS subscales had acceptable internal consistency: for Neglect/Negative Home Environment (14 items), α = .87; for Punishment (6 items), α = .65; and for Sexual Abuse (6 items), α = .74. The CATS demonstrates significant convergent validity with similar measures of child maltreatment.⁴⁰ The Neglect/Negative Home Environment subscale assesses retrospective perceptions of childhood emotional maltreatment, that is, parental emotional neglect, as well as recollections of global negative emotional responses to hostile or negative aspects of the childhood home environment. Although this measurement strategy has limitations, it does appear to be a useful composite index of adversities experienced in the context of parent–child relations or the broader family environment.

Statistical Analyses

We used path analyses to evaluate associations between observed and latent constructs of interest. Analyses were performed using Mplus version 6.00.⁴¹ Missing data were minimal (as much as 5%) and were determined to be missing at random,^42,43 and analyses used all available data. To account for non-normality presented in the unprotected sex variable as well as the sexual abuse variables, we used a maximum likelihood estimator with robust standard errors using a numerical integration algorithm.⁴¹ Maximum likelihood parameter estimates with standard errors and a χ² test statistic are robust to non-normality and nonindependence of observations. Traditional maximum likelihood methods assume that the distributions of the continuous variables in the model are multivariate normal. The normal distribution assumption is problematic in mediation models because the product coefficients used to evaluate mediation rarely meet this assumption.^44–46 Thus, in this study, we assessed mediation models using 2000 bootstrap replicates to obtain bias-corrected bootstrap confidence intervals for the indirect effects.^44,45,47

We used several global fit indices to evaluate the fit of the path model, including the traditional overall χ² test of model fit (which should be statistically nonsignificant) and these statistical criteria^{48, 49}: the root-mean-square error of approximation (< 0.08), the test of close fit (P > .05), the comparative fit index (> 0.95), and the standardized root-mean-square residual (< 0.07). In addition to the global fit indices, more focused tests included examination of the results for the absence of negative residual variance values, as well as restricting standardized residual covariances to values ranging between −2.00 and 2.00 and nonsignificant modification indices (< 3.84).

We considered multiple group analysis as a way to address possible gender variations in the proposed mediation model. However, the significantly different numbers of boys and girls in the sample (114 girls, 280 boys; n = 394) was problematic for this approach. For a variance–covariance matrix containing more than 8 variables and 22 freely estimated parameters, a limited sample size per group, and each maltreatment type not reported by each study participant, we found it difficult to make definite empirical conclusions about the role of gender in the model. Statistical power was undermined, and the risk of creating bias in the parameter estimates and standard errors was increased.⁴¹ Thus, we decided that handling gender conventionally would be more justifiable, that is, using gender as a covariate in the analysis to evaluate whether controlling for gender changed any of the structural relations.

RESULTS

Table 2 summarizes bivariate correlations among variables included in the structural equation model evaluated in this study. Figure 1 represents the final structural equation model–based multivariate path analysis model resulting from iterative model-fitting procedures. Standardized parameter estimates for indirect effects are displayed in Table 3. This model tested paths between (1) 3 separate dimensions of childhood maltreatment (i.e., the CATS Neglect/Negative Home Environment, Punishment, and Sexual Abuse subscales), (2) past-year substance abuse or dependence symptoms (i.e., for alcohol or drug use), and (3) 3 sexual risk behavior variables during the 180 days before the baseline interview (e.g., a latent construct for unprotected intercourse, frequency of sex with co-occurring alcohol use, and frequency of sex with co-occurring drug use). Paths were defined from each exogenous variable to each of the 5 endogenous variables, creating a saturated model, with the exception of the patterning of paths among the 5 endogenous variables. Residual variances were assumed to be uncorrelated, with the exception of alcohol and drug abuse or dependence symptoms and alcohol or drug use co-occurring with sex. Correlations were assumed for all exogenous variables in the model.

TABLE 2—

Descriptive Statistics and Bivariate Correlations Among Variables in a Structural Model Evaluated With Data From a Sample of Adolescents Receiving Outpatient Treatment for Substance Abuse Problems at 2 Outpatient Facilities in South Florida: September 2004—September 2008

Variable	Mean (SD)	1	2	3	4	5	6	7	8	9	10
1. No. unprotected intercourse	12.50 (24.09)	—
2. Frequency of CUa	3.76 (1.27)	0.50**	—
3. Sexual abuse	1.21 (0.47)	0.06	0.18**	—
4. Physical abuse	2.51 (0.73)	−0.01	0.10	0.28**	—
5. Neglect	2.15 (0.77)	0.10*	0.22**	0.45**	0.50**	—
6. Alcohol symptoms	1.56 (2.23)	0.09	0.15**	0.22**	0.16**	0.30**	—
7. Drug symptoms	4.49 (3.51)	0.15**	0.17**	−0.07	0.04	0.19**	0.45**	—
8. Sex with alcohol	2.08 (1.11)	0.17**	0.16**	0.10	0.05	0.06	0.36**	0.32**	—
9. Sex with drugs	2.56 (1.36)	0.16**	0.17**	0.02	0.03	0.12*	0.25**	0.44**	−0.44**	—
10. Age	16.33 (1.15)	0.19**	0.19**	−0.18**	−0.061	−0.021	0.02	0.21**	0.13**	0.22**	—

Note. CU = condom use. Sex with alcohol and Sex with drugs denote sex with co-occurring use of alcohol or drugs, respectively. Age range = 12.78–18.65 years (SD = 5.94).

^aFrequency of CU = a 5-point Likert frequency of unprotected sex during the past 12 months.

*P < .05; ** P < .01.

FIGURE 1—

Structural mediation model of relations among childhood maltreatment experiences, alcohol or other drug abuse and dependence symptoms, and sexual risk behavior outcomes, evaluated with a sample of adolescents receiving outpatient treatment for substance abuse problems at 2 outpatient facilities in south Florida: September 2004—September 2008.

Note. E = residuals (errors). Unstandardized and standardized (in parentheses) path coefficients are presented. Model adjusted for age and ethnicity (not shown).

*P < .05, critical ratio > 2.00; **P < .01, critical ratio > 2.5.

TABLE 3—

Standardized Parameter Estimates for Indirect Effects in the Model of Risky Sexual Behaviors, Alcohol and Other Drug Symptoms, and Childhood Maltreatment Tested With Data From a Sample of Adolescents Receiving Outpatient Treatment for Substance Abuse Problems at 2 Outpatient Facilities in South Florida: September 2004—September 2008

Path	B (SE; 95% CI)a
Neglect to unprotected sex
Via alcohol symptoms	0.014 (0.029; −0.043, 0.071)
Via drug symptoms	0.071* (0.031; 0.014, 0.183)
Sexual abuse to unprotected sex
Via alcohol Symptoms	0.013 (0.015; −0.049, 0.105)
Via drug symptoms	−0.068* (0.021; −0.196, −0.011)
Neglect to sex and alcohol
Via alcohol symptoms	0.083** (0.020; 0.039, 0.151)
Via drug symptoms	0.065** (0.019; 0.027, 0.127)
Sexual abuse to sex and alcohol
Via alcohol symptoms	0.073* (0.016; 0.014, 0.154)
Via drug symptoms	−0.063* (−0.137; −0.137, −0.021)
Neglect to sex and drugs
Via alcohol symptoms	0.013 (0.012; −0.011, 0.036)
Via drug symptoms	0.095* (0.025; 0.047, 0.143)
Sexual abuse to sex and drugs
Via alcohol symptoms	0.023 (0.022; −0.005, 0.082)
Via drug symptoms	−0.165* (0.061; −0.299, −0.067)

Note. CI = confidence interval. Sex and alcohol and sex and drugs denote sex with co-occurring use of alcohol or drugs, respectively. “Via” denotes the mediator between the independent and the outcome variables.

^aStandardized CIs (bias-corrected) obtained on the basis of 2000 bootstrap replicates of the indirect effects.

*P < .05; **P < .01

The selective trimming of paths included the removal of the CATS Punishment subscale from the model because we found no significant paths between the CATS Punishment subscale and any of the other variables in the model. A second run of the model without the CATS Punishment subscale did not reveal significant differences in either global or specific fit indices from the results of the first run of the model. Standardized and unstandardized path coefficients are presented in Figure 1. Standardized residuals and correlations between the exogenous variables were omitted to improve the clarity of the presentation. All path coefficients presented in the model were statistically significant (P < .05; critical ratio > 2.00). To control for chance effects across multiple tests of significance, we adopted a modified Bonferroni criterion for evaluating the statistical significance of a path coefficient, based on the false discovery rate method,⁵⁰ in which a family of tests is defined by the number of path coefficients leading from exogenous variables to a specific endogenous variable. Traditional indices of global fit suggested good fit between the data and the path model tested (comparative fit index = 1.00; root-mean-square error of approximation < 0.001; close fit test, P > .969; 90% CI = 0.00, 0.044; χ² = 3.06, df = 8, P = .878; standardized root-mean-square residual = 0.014).

Age

All paths reported in Figure 1 were significant after adjusting for variance in age. We found statistically significant path coefficients between age and (1) drug abuse and dependence symptoms (B = 0.589; SE = 0.149; P < .01), (2) the unprotected intercourse latent construct (B = 0.407; SE = 0.088; P < .01), and (3) co-occurring sex and drug use (B = 0.188; SE = 0.052; P < .01).

We also found significant age differences in scores for drug abuse and dependence symptoms and 2 sexual risk behaviors such that, on average, older participants reported higher scores for drug abuse and dependence symptoms, unprotected intercourse episodes, and sex with co-occurring use of drugs.

Gender and Ethnicity

All paths were significant after holding gender constant. We found statistically significant path coefficients in the model between gender and co-occurring sex and alcohol use (B = -0.076; SE = 0.030; P < .01). Unstandardized coefficients indicated that, on average, male participants reported higher scores for sex with co-occurring alcohol use.

Ethnicity was operationalized as Hispanic versus others. We found no statistically significant path coefficients in the model between ethnicity and other endogenous variables in the model (e.g., substance use or sexual risk behavior variables).

Direct Paths

Childhood sexual abuse had significant positive coefficient paths (P < .05) to past-year alcohol abuse and dependence symptoms, as well as to co-occurring alcohol use and sexual activity. By contrast, we found a negative coefficient path between sexual abuse experiences and drug abuse and dependence symptoms. To ensure that this unexpected path direction was not a statistical artifact that was distorting overall model fit, we excluded (i.e., constrained) this path and reran the analyses. Very good model fit was retained (comparative fit index = 0.99; root-mean-square error of approximation < 0.031; close fit test, P > .752; 90% CI = 0.00, 0.05; χ² = 12.513, df = 9; P = .186; standardized root-mean-square residual = 0.021).

We found significant direct paths between experiences of neglect and negative home environment and past-year alcohol or drug abuse and dependence symptoms. Significant, positive (P < .05) direct paths were found between past-year drug abuse and dependence symptoms and all 3 indices of sexual risk behavior examined. We identified a single significant, positive (P < .01) direct path between alcohol abuse and dependence symptoms and sex with co-occurring alcohol use.

Mediators

Bias-corrected bootstrap confidence intervals for the indirect effects^45,47 were obtained on the basis of 2000 bootstrap replicates. Significant indirect effects (at the 95% confidence interval level) are presented in Table 3. Mediated relations were documented among childhood maltreatment, past-year alcohol or other drug abuse and dependence symptoms, and 3 indices of sexual risk behavior.

First, alcohol abuse and dependence symptoms mediated relations between (1) neglect scores and alcohol use co-occurring with sex and (2) sexual abuse scores and alcohol use co-occurring with sex. Second, drug abuse and dependence symptoms mediated relations between neglect and alcohol or drug use co-occurring with sex, as well as unprotected intercourse. Results suggested that the influence of specific forms of childhood maltreatment (e.g., sexual abuse, emotional neglect) on specific sexual risk behaviors among adolescents receiving substance use treatment services is mediated by past-year alcohol or other drug abuse or dependence symptoms.

DISCUSSION

Similar to previous research that documented significant relations among childhood maltreatment, substance use problems, and sexual risk behaviors,⁵¹ we identified multiple risk factors for HIV and STI exposure among youths undergoing treatment for substance use problems. The evaluation of the structural model revealed partial support for the study's hypothesis in that 2 of 3 forms of childhood maltreatment assessed were significantly associated with substance abuse and dependence symptoms. Findings were consistent with previous research documenting significant relations between childhood sexual abuse or neglect and the development of a wide range of psychopathology among youths, including substance use disorders.^12,52–54

Analyses supported hypothesized relations between alcohol or other drug abuse and dependence symptoms and specific indices of sexual risk behavior, similar to existing studies of youths with multiple problems.^55,56 Associations between severity of alcohol or other drug use problems and sexual risk behaviors have been documented in both treatment^3,57 and community samples of adolescents.⁵⁸ This study extends this work and delineates unique paths between alcohol use problems and alcohol-related sexual risk behavior, whereas drug use problems appeared to be associated with multiple indicators of sexual risk behavior.^56,59 These distinct associations between substance use problems and sexual risk behaviors are supported by cluster analyses of a large community sample of youths, classified into homogeneous multivariate patterns of sexual risk behaviors.^16,60 Particular patterns of sexual risk behaviors (e.g., co-occurring sex and substance use) were significantly associated with alcohol or drug abuse and dependence symptoms. The specificity of the mediating paths from forms of child maltreatment to sexual risk behaviors may also reflect differences in substance use–related coping strategies associated with histories of childhood sexual abuse, for example, distress coping or emotional regulation strategies.^25,59

Consistent with alcohol myopia theory,²⁷ the observed direct and indirect paths from childhood maltreatment to specific indicators of sexual risk behavior are congruent with existing research on short- and long-term responses to prior victimization involving maladaptive patterns of substance use.⁶¹ Significant relations between childhood sexual abuse, substance use, and related sexual risk behaviors may be explained in part by individual-level deficiencies in information processing abilities resulting from alcohol intoxication.⁶² Similarly, our findings also support some of the hypotheses advanced by Miller and Mancuso.²⁸ For example, use of substances to cope with trauma-related symptoms may promote the development of substance abuse and dependence disorders and participation in sexual risk behavior.⁶³ Motives for substance use before or during sex may include anxiety reduction or dissociation from past sexual traumas.^64,65

Notably, the negative association between sexual abuse and drug abuse and dependence symptoms does not necessarily suggest that sexual abuse is not associated with increased drug use. Rather, this finding may suggest that when adjusting for neglect experiences, sexual abuse is more likely to be associated with alcohol use symptoms than with drug use symptoms. For example, the direction of the relation between sexual abuse and alcohol abuse and dependence symptoms may change with the introduction of the neglect variable to the model, a potential reversal paradox effect, complicating the interpretation of the finding.⁶⁶ Yet, the advantage of testing multiple mediation mechanisms⁴⁵ is that it allows one to evaluate the relative contribution of competing mediators. The unexpected difference in the signs of path coefficients between (1) childhood sexual abuse and (2) alcohol or drug abuse and dependence symptoms may also partially reflect meaningful differences in substance use–related strategies for coping with different patterns of childhood maltreatment^63,67,68 or an artifact resulting from bias in the selection of study participants. These issues require further investigation. Other research^12,69 has documented substantial unexplained heterogeneity in relations between childhood sexual abuse and patterns of substance use for male and female adolescents using multivariate classification techniques, indicating the potential for multiple pathways between the 2 constructs.

In general, more severe patterns of alcohol or drug use among adolescents with childhood histories of maltreatment appear to promote participation in co-occurring sex and substance use and reduce the use of self-protective HIV and STI risk reduction behaviors such as condom use. Current understanding of the mechanisms underlying relations among childhood maltreatment, substance use disorders, and sexual risk behaviors is rudimentary and likely to involve complex person–environment interactions.²⁸ Testing for the mediational role that substance use symptoms play in the association between childhood maltreatment and specific indices of sexual risk behavior provides evidence underscoring the importance of substance abuse treatment as a means to reduce risk for HIV and STI exposure.^70,71 However, among youths with substance use problems, additional attention to other putative mediators of relations between childhood maltreatment and specific sexual risk behaviors (e.g., deficits in behavioral skills, socioemotional competencies or coping strategies) is also warranted.

Strengths and Limitations

This study has several limitations. First, measures included sensitive self-reported data from a single source and may reflect response bias. Nonetheless, research has suggested that self-reports may still be more accurate than records in estimating the prevalence of child maltreatment.³⁹ Second, we conducted analyses using retrospectively collected data. A limitation of this practice is the potential risk for violating the specified temporal order of the model.^29,72 Previous longitudinal research has suggested that among youths, substance use problems tend to precede participation in sexual risk behavior.⁹ However, caution must still be exercised with regard to making causal statements about relations among variables in the model, and longitudinal research is necessary to test these relations more rigorously.^29,72,73

Last, our sample consisted of youths receiving outpatient treatment of alcohol or drug use problems who commonly present with co-occurring problem behaviors and psychiatric diagnoses. Therefore, although the findings may generalize to other samples of adolescents undergoing outpatient substance use treatment, they may not generalize to the experiences of adolescents undergoing inpatient treatment or to general population samples of adolescents.

Implications for Prevention Efforts

First, substance use treatment of adolescents with extensive histories of maltreatment must address underlying issues of maltreatment and related emotional distress to enhance relapse prevention efforts and reduce co-occurring health risk behaviors.^74,75 This point highlights the importance of comprehensive assessment of risk behaviors at treatment entry and integrated treatment approaches to enhance both treatment effectiveness and health outcomes in this population. Second, the significant age differences documented for behavioral risk factors for HIV and STI transmission among the adolescents in this study underscore the importance of the early application of selected prevention strategies for substance use and HIV and STI risk reduction in vulnerable youth populations. Lack of access to comprehensive, integrated risk behavior prevention programming increases the likelihood that co-occurring risk behaviors will accelerate and become potentially less amenable to change.⁶⁰ Third, clients reporting specific patterns of childhood maltreatment may benefit from the development of tailored therapeutic approaches that address the impact of different forms of trauma, the perceived benefits of different forms of substance use, and linkages between substance use and specific sexual risk behaviors. Fourth, treatment effectiveness is likely to be enhanced by future research that improves current understanding of (1) motives for co-occurring substance use and sexual behavior or (2) substance use–related coping strategies among adolescents with histories of childhood maltreatment.

Enhanced knowledge in these areas is a first step toward augmenting practitioners’ abilities to design strategies and exercises for improving clients’ knowledge and awareness, interpersonal skills and competencies, readiness to change, and self-efficacy to implement positive health behavior alternatives.