Gender Differences in the Association between Conduct Disorder and Risky Sexual Behavior

Abstract

Despite suggestions that there are gender differences in the association between conduct disorder (CD) and risky sexual behavior, limited empirical research has examined this question. Youth (N = 616) were recruited from four primary care clinics and completed questions related to risky sexual behavior, alcohol and marijuana use, and CD. Results of stratified multivariate models indicated that the association between CD and having four or more lifetime partners, having two or more partners in the last 3 months, and engaging in condomless sex was stronger among female youth. However, association between CD and alcohol and other drug use before sex was stronger in male youth. This is an important contribution to our understanding of gender-specific manifestations of conduct disorder, and has the potential to inform screening and brief intervention efforts for this population.

Conduct disorder is a disorder of childhood and adolescence that is characterized by a pattern of behavior that violates social norms or the rights of others. Youth with conduct disorder engage in a range of problem behaviors, including aggression, destruction of property, deceitfulness, theft, and serious violation of rules (e.g., running away from home) (American Psychiatric Association, 2013). Moreover, youth with conduct disorder may be more likely to participate in other types of risk-taking behaviors, including substance use and risky sexual behavior (Loeber, Burke, Lahey, Winters, & Zera, 2000).

There have been questions as to whether conduct disorder manifests similarly in male and female youth (Berkout, Young, & Gross, 2011; Loeber et al., 2000). Previous studies have shown conduct disorder to be more prevalent among male than female youth (American Psychiatric Association, 2013; Berkout et al., 2011); however, some have questioned whether this is an artifact of the process used to generate and validate the diagnostic criteria, which has relied largely on samples of male youth (Moffitt et al., 2008). For instance, researchers have argued that the diagnostic criteria comprise behaviors that are more common among boys (e.g., physical aggression). Relatedly, some researchers have suggested that there may be gender-specific symptoms of conduct disorder. For instance, Crick and colleagues have focused on the manifestation of aggression as a symptom of conduct disorder, hypothesizing that “attempts to harm others (to aggress) would focus on social issues most salient in their same-gender peer groups” (Crick & Zahn-Waxler, 2003, p. 723). They suggest that for males, the key concern of childhood and adolescence is physical dominance; therefore, boys with conduct disorder engage in physical aggression. For female youth, however, the focus is on the formation of close relationships. As a result, they hypothesize that attempts to harm others will manifest as relational aggression – that is, attempts to damage interpersonal relationships (Crick & Zahn-Waxler, 2003).

Similarly, it has been suggested that symptoms such as risky sexual behavior may be female-specific symptoms of conduct disorder. For instance, there is evidence that females with conduct disorder are more likely to have comorbid externalizing and internalizing disorders (versus males, who generally display heightened rates of externalizing disorders alone) (Keenan, Loeber, & Green, 1999). In turn, research has shown that youth with comorbid externalizing and internalizing disorders are at greater risk for behaviors such as having multiple sexual partners (Berkout et al., 2011; Dishion, 2000), possibly because these youth are seeking to dampen their negative emotional experiences through risky but pleasurable experiences (Dishion, 2000). Thus, it has been hypothesized that female youth with conduct disorder will participate in more sexual risk behaviors (Berkout et al., 2011). In addition, a recent review found that females with conduct disorder were approximately 4.7 times more likely to have a history of sexual abuse than their male peers (Maniglio, 2014), and childhood sexual abuse is often associated with later sexual risk behavior (Senn, Carey, & Vanable, 2008).

There has been limited research examining the association between conduct disorder and risky sexual behavior in adolescence. There is evidence that conduct disorder is related to earlier initiation of sexual behavior (Monuteaux, Faraone, Michelle Gross, & Biederman, 2007; Wymbs et al., 2013), having multiple partners (Monuteaux et al., 2007), and having unprotected sex (Bryan & Stallings, 2002) in early adolescence to young adulthood. However, there have been certain methodological weaknesses that preclude strong conclusions about these associations. For instance, some studies of conduct disorder use only male (Bryan & Stallings, 2002) or only female (Monuteaux et al., 2007) youth. Other studies fail to control for substance use (e.g., Wymbs et al., 2013), despite evidence that alcohol and marijuana use are strongly related to both conduct disorder (Compton, Conway, Stinson, Colliver, & Grant, 2005) and risky sexual behavior (Bryan, Schmiege, & Magnan, 2012; Calvert, Keenan Bucholz, & Steger-May, 2010). In addition, other work has shown that conduct disorder and other externalizing disorders are associated with youth reports of ever having been sexually active, but not with sexual risk behaviors or associated outcomes (e.g., sexually transmitted infections [STI]) (Brown et al., 2010). These discrepant results suggest that there may be variability in the association between conduct disorder and specific aspects of sexual risk taking among boys versus girls.

Furthermore, despite the theoretical arguments that there are gender differences in the prevalence of risky sexual behavior among youth with conduct disorder, there has been limited empirical evidence to support this claim (Moffitt et al., 2008). There have been studies examining gender differences in risky sexual behavior among youth involved in the juvenile justice system; however, results of these studies have been mixed. Some research indicates that male adolescents involved in the juvenile justice system report more lifetime sexual partners than female adolescents (Robillard, Conerly, Braithwaite, Stephens, & Woodring, 2005), and that male youth with conduct disorder initiate sex at a younger age than female youth (Galéra et al., 2010). However, other studies have shown that females in the juvenile justice system are less likely to use condoms and have higher rates of STIs than males (Kahn et al., 2005; Robillard et al., 2005). Some research has also looked at gender differences in the association between various developmental trajectories of delinquent behavior (e.g., adolescent-limited, life course persistent, late onset), finding that the association between delinquent behavior/conduct problems and risky sexual behavior is largely similar for male and female youth across trajectories (Aalsma, Tong, Wiehe, & Tu, 2010; Miller, Malone, Dodge, & Conduct Problems Prevention Research, 2010). However, most studies have examined this association with at-risk youth, such as those in the juvenile justice system. Research is needed with racially and ethnically diverse youth sampled from the general population.

A better understanding of the relationship between conduct disorder and risky sexual behavior is critical. First, risky sexual behavior in adolescence is associated with an increased risk of STIs, HIV, and unplanned pregnancy (Capaldi, Stoolmiller, Clark, & Owen, 2002; Malhotra, 2008). Therefore, understanding the correlates of risky sexual behavior has the potential to inform screening and targeted prevention. Second, examining the role of gender may help to shed light on the theories suggesting that there are gender-specific manifestations of conduct disorder. Knowing whether there are gender differences in the symptoms of conduct disorder is important for the accurate assessment and diagnosis of this disorder, which will increase the likelihood that girls who meet the diagnostic criteria for conduct disorder receive needed services (Moffitt et al., 2008). The present study adds to the existing literature in this area by examining whether there are gender differences in the association between conduct disorder and risky sexual behavior in a large and diverse sample of youth, age 12–18, accessing routine healthcare through primary care clinics.

This study had two primary aims. First, we aimed to examine the association between conduct disorder and risky sexual behavior in a sample of youth recruited through primary care clinics during routine appointments. We hypothesized that there would be a significant association between conduct disorder and risky sexual behavior, controlling for demographic factors and other adolescent risk behaviors (specifically, alcohol and drug use). Second, we aimed to determine whether there were gender differences in this association. It was hypothesized that gender would significantly moderate the association between conduct disorder and risky sexual behavior. Because prior research has been mixed and often dependent upon the sexual behavior (e.g., number of partners versus condom use), we expected that we would also see differences with males reporting more partners than females, and with females less likely to report condom use than males.

Method

Procedures

The current study involved four primary care clinic sites (one in Los Angeles and three in Pittsburgh). We approached and invited every youth between the ages of 12 and 18, inclusive, that came to the clinics for any type of primary care appointment. Recruitment in the clinics occurred over a 20 month period in Los Angeles and a 17 month period in Pittsburgh. For youth interested in participating, we obtained parental consent and youth assent (for youth under age 18), or youth consent (for youth aged 18). We obtained a certificate of confidentiality and procedures were approved by the institution’s Institutional Review Board and each of the clinics. For additional detail, see D’Amico and colleagues (2016).

Project staff first administered the National Institute on Alcohol Abuse and Alcoholism Screening Guide (NIAAA SG) to participants, who then completed a web-based survey that included demographic characteristics, additional measures of substance use, and participation in delinquent behaviors and sexual behavior. Surveys were completed in a private space without a parent present to ensure confidentiality. Youth were paid $25 for their participation.

Setting and participants

The Pittsburgh and Los Angeles area clinics provide care for a highly ethnically and racially diverse population of largely-underserved youth. These four sites offer both ongoing, continuity-based care and episode-based urgent care to their patients. Clinics in both cities have a large percentage of minority patients and serve a high proportion of low-income patients who do not have insurance or who utilize Medicare and/or Medicaid/Medi-Cal. A total of 3,309 youth were approached for study recruitment. Of these youth, 27% (n = 892) were ineligible due to age, lack of English proficiency, being present at the clinic for an appointment other than their own, or disability status; 18.5% (n = 614) declined to participate, mostly due to time constraints or youth being at the clinic for a family planning appointment and not wanting their parent to know they were at the clinic. This yielded a total sample of 1803 youth who enrolled or provided consent to contact with further information about study participation. Of the 1803 youth, 12.8% (n = 230) of youth did not complete the baseline within the field period or had poor contact information. Thus, the final enrolled sample included 1573 youth (Pittsburgh Site 1: n = 297; Pittsburgh Site 2: n = 254; Pittsburgh Site 3: n = 161; Los Angeles Site 1: n = 861). Because the present analyses focus on risky sexual behavior, we only included participants in our analytic sample that reported ever having sex in their lifetime (39%; n = 616; Pittsburgh Site 1: n = 141, Pittsburgh Site 2: n = 98; Pittsburgh Site 3: n = 63; Los Angeles Site 1: n = 314). On average, youth in this sample were 16.75 years old. The sample was 61.9% female and diverse, with approximately 50% Hispanic youth and 31% African American youth (see Table 1).

Table 1.

Descriptive statistics and bivariate relationships with GAIN score

	Overall (n=616) Mean (SE) or %(n)	Bivariate Association with GAIN Score (Mean GAIN (SE) or Correlation)
Demographics
Age	16.75 (0.20)	r = −0.14 **
Gender
Female	61.87% (378)	1.79 (0.10)
Male	38.13% (233)	1.62 (0.17)
Mother Education
≥ College	30.68% (166)	1.64 (0.13)
< College	69.32% (450)	1.73 (0.13)
Race/Ethnicity
White	13.96% (86)	1.60 (0.10) *†
Black	31.17% (192)	1.43 (0.14)
Hispanic	50.32% (310)	1.95 (0.05)
Other	4.55% (28)	1.64 (0.25)
Past Month AOD Use
Alcohol **
Yes	41.95% (258)	2.30 (0.22) *
No	58.05% (357)	1.32 (0.10)
Marijuana **
Yes	41.63% (256)	2.31 (0.20)
No	58.37% (359)	1.31 (0.04)
Sexual Behavior
Lifetime Partners
4+ partners	31.97% (196)	2.15 (0.14)
<4 partners	68.03% (417)	1.51 (0.12)
Partners in Last 3 Months
2+ partners	17.07% (105)	2.29 (0.08) **
<2 partners	82.92% (501)	1.61 (0.09)
AOD Use Before Sex
Yes	17.40% (107)	2.83 (0.27) *
No	82.60% (508)	1.50 (0.04)
Condomless Sex
Yes	36.42% (224)	1.88 (0.14)
No	63.58% (391)	1.63 (0.09)

^*p < .05,

^**p < .01;

^†pairwise comparisons revealed significant differences in GAIN score for Black vs. Hispanic youth, and for White vs. Hispanic youth

Measures

Demographics

Participants indicated age, gender, and parent level of education (for the present analyses, operationalized as mother’s completion of college or greater).

Past month substance use

Past month substance use was assessed with items adapted from the Monitoring the Future study asking respondents how many days during the past month they had used each of several substances (Johnston, O’Malley, Bachman, & Schulenberg, 2013). For this study, responses were dichotomized into “any use” versus “no use” so that we could account for any use in the past month. For the present analyses, we focused on alcohol and marijuana use, as these were the substances most frequently endorsed by youth, and because correlations between these specific substances and both conduct disorder and risky sexual behavior have been demonstrated in previous studies (Bryan et al., 2012; Calvert et al., 2010; Compton et al., 2005).

Conduct disorder

Participants completed the Global Appraisal of Individual Needs (GAIN) Conduct Disorder Scale (Dennis, White, Titus, & Unsicker, 2008). This scale comprises 15 items, each of which represents a specific problem behavior (e.g., “Used a weapon in fights;” “Set fires;” “Taken things from a store or written bad checks to buy things”; α = 0.73). Participants indicate whether they have done a behavior two or more times in the past year, with response options including “yes” or “no”. The number of behaviors is summed to yield a total score. Consistent with previous studies and research supporting a dimensional conceptualization of conduct disorder (e.g., Bryan & Stallings, 2002; Moffitt et al., 2008), we analyzed this variable as a continuous measure.

Risky sexual behavior

Youth completed several questions about sexual behavior. This included asking, “Have you ever had sexual intercourse?” The present study focuses on the 616 youth who answered “yes” to this question. Youth who reported a history of sexual activity were then prompted to respond to four additional questions focused on risky sexual practices: (1) number of sexual partners they have had in their lifetime, dichotomized as 4 or more partners versus less than 4 partners (“4+ lifetime partners”) (Centers for Disease Control and Prevention, 2014; O’Donnell, O’Donnell, & Stueve, 2001); (2) number of sexual partners they had in the past three months, dichotomized as 2 or more versus less than 2 partners (“2+ partners last 3 months”) (Markham et al., 2012; Taliaferro, Rienzo, & Donovan, 2010); (3) use of alcohol or other drugs (AOD) immediately prior to engaging in sexual intercourse the last time the youth had sexual intercourse (“AOD before last sex”); and (4) not using a condom the last time the youth had engaged sexual intercourse (“condomless sex”). Each question was analyzed individually as an outcome for the present study.

Data analysis

Data analysis was conducted using SAS (SAS Institute, Cary, NC). We first calculated descriptive statistics and examined bivariate relationships between the GAIN score and all predictors and outcomes. We then fit multivariate logistic regression models to examine the association of the GAIN score with each risky sexual behavior outcome. Due to missingness in the mother’s education variable (n = 75), this variable was imputed using a logistic regression model with race/ethnicity, city, and an indicator of living in a two parent household (Gelman & Hill, 2007; Rubin & Little, 2002). We first examined models to evaluate the main effects of gender and conduct disorder on risky sexual behavior. These models controlled for age, race/ethnicity, mother’s education, and alcohol and marijunana use. To this basic model, interaction terms of gender and the GAIN score were added to evaluate how conduct disorder might differentially affect the risky sexual behavior outcomes for males and females. All models were clustered by clinic to account for similarities among youth seen at each clinic. To interpret any differential effects by gender, models stratified by gender were conducted as a follow-up to these analyses.

Results

Bivariate associations

The mean GAIN score in the sample was 1.73 (SD = 0.09). There were statistically significant bivariate associations between conduct disorder and age, race/ethnicity, past month alcohol and marijuana use, number of past 3 month sexual partners, and AOD use before sex (see Table 1). Specifically, younger age was associated with more conduct disorder symptoms, although the magnitude of this correlation was small. Regarding race/ethnicity, Hispanic youth reported somewhat higher levels of conduct disorder symptoms than Black or White youth. Youth who used alcohol or marijuana in the past month, had more sexual partners in the past three months, and reported AOD use before last sex reported higher levels of conduct disorder symptoms.

Main effect models

We examined multivariate models predicting the four risky sexual behavior outcomes. In our main effect models, there were significant main effects of gender on three outcomes (see Table 2). Females were 1.73 times more likely to report condomless sex; by contrast, they were significantly less likely to report having 4+ lifetime partners or 2+ partners in the last three months. There were also significant main effects of conduct disorder, such that youth who reported higher levels of conduct disorder symptoms were also more likely to report having 4+ lifetime partners, having 2+ partners in the last 3 months, having used AOD before last sex, and having had condomless sex. In addition to the effect of gender and conduct disorder, both alcohol and marijuana use were independent predictors of lifetime number of partners, last three month partners, and AOD use before last sex, such that youth who reported using alcohol and marijuana were more likely to report engaging in these risky sexual behaviors.

Table 2.

Multivariate models – main effects only

	4+ Lifetime Partners	2+ Partners Last 3 Months	AOD Before Last Sex	Condomless Sex
	OR (95% CI)	OR (95% CI)	OR (95% CI)	OR (95% CI)
Age	1.52 (1.39–1.66) **	1.24 (1.01–1.52) *	1.09 (0.98–1.20)	1.09 (0.98–1.21)
Race/ethnicity
Black	1.72 (0.72–4.11)	2.38 (1.84–3.07) **	1.51 (0.49–4.69)	1.36 (0.92–2.02)
Hispanic	0.98 (0.40–2.40)	1.76 (1.16–2.67) **	1.53 (0.71–3.31)	1.18 (0.69–2.03)
Other	1.25 (0.26–6.11)	1.60 (0.70–3.66)	2.24 (0.70–7.24)	1.95 (1.41–2.69) **
White (ref)
Female	0.44 (0.22–0.87) *	0.44 (0.22–0.90) *	0.68 (0.36–1.28)	1.73 (1.35–2.22) **
Mother Education ≥ College	1.25 (0.59–2.66)	1.15 (0.56–2.34)	0.72 (0.56–0.93) *	0.73 (0.56–0.94) *
GAIN	1.19 (1.07–1.32) **	1.13 (1.11–1.16) **	1.22 (1.10–1.34) **	1.05 (1.01–1.09) *
Past Month Alcohol Use	1.68 (1.31–2.15) **	2.07 (1.87–2.29) **	2.81 (2.00–3.95) **	1.26 (0.86–1.83)
Past Month Marijuana Use	1.82 (1.48–2.24) **	1.56 (1.16–2.08) **	2.69 (2.03–3.57) **	1.14 (1.00–1.30)
Wald Chi Square†	1036.94 **	20971.96 **	1201.81 **	6.32
AIC	697.71	532.32	493.00	795.02

^*p < .05,

^**p < .01;

^†df = 9

Interaction effects

There were no significant gender by conduct disorder interactions for AOD use before last sex or condomless sex (see Table 3). There were significant gender by conduct disorder interactions for the number of sexual partners in the past three months and the number of lifetime sexual partners (see Table 3).

Table 3.

Multivariate models with gender by conduct disorder interactions

	4+ Lifetime Partners	2+ Partners Last 3 Months	AOD Before Last Sex	Condomless Sex
	OR (95% CI)	OR (95% CI)	OR (95% CI)	OR (95% CI)
Age	1.52 (1.40–1.65) **	1.24 (1.01–1.51) *	1.08 (0.97–1.20)	1.09 (0.98–1.21)
Race/ethnicity
Black	1.75 (0.70–4.35)	2.41 (1.89–3.06) **	1.48 (0.47–4.62)	1.37 (0.91–2.06)
Hispanic	1.00 (0.39–2.52)	1.78 (1.17–2.72) **	1.50 (0.69–3.24)	1.19 (0.68–2.09)
Other	1.28 (0.27–6.06)	1.63 (0.72–3.66)	2.18 (0.71–6.67)	1.98 (1.38–2.83) **
White (ref)
Female	0.37 (0.19–0.74) **	0.39 (0.18–0.84) *	0.87 (0.41–1.82)	1.53 (1.10–2.13) *
Mother Education ≥ College	1.26 (0.59–2.67)	1.15 (0.57–2.32)	0.71 (0.54–0.94) *	0.73 (0.56–0.94) *
GAIN	1.12 (0.98–1.27)	1.09 (1.07–1.11) **	1.31 (1.22–1.40) **	1.00 (0.95–1.05)
Past Month Alcohol Use	1.69 (1.31–2.18) **	2.09 (1.86–2.33) **	2.77 (1.92–4.00) **	1.26 (0.87–1.85)
Past Month Marijuana Use	1.83 (1.49–2.24) **	1.56 (1.17–2.08) **	2.68 (2.04–3.51) **	1.14 (0.99–1.31)
Female * GAIN	1.10 (1.00–1.20) *	1.06 (1.00–1.13) *	0.90 (0.80–1.01)	1.08 (1.00–1.16)
Wald Chi Square†	38438.61 **	473289.88 **	227.83 **	4.90
AIC	698.76	533.96	494.08	796.39

^*p < .05,

^**p < .01;

^†df = 10

We next conducted stratified analyses by gender to further understand the interaction effects (see Table 4). Conduct disorder was a significant predictor for 4+ lifetime partners in females but not in males, and a similar pattern was observed for condomless sex. Conduct disorder was a significant predictor of 2+ partners in the last 3 months and AOD before last sex for both males and females. However, for 2+ partners in the last 3 months, the magnitude of the effect was larger for females than males, whereas for AOD before last sex, the magnitude of the association was larger in males than females.

Table 4.

Stratified models by gender.

	4+ Lifetime Partners	2+ Partners Last 3 Months	AOD Before Last Sex	Condomless Sex
	OR (95% CI)	OR (95% CI)	OR (95% CI)	OR (95% CI)
	Female Youth
Age	1.65 (1.40–1.93) **	1.40 (1.09–1.81) **	1.10 (0.83–1.45)	1.02 (0.89–1.16)
Race/ethnicity
Black	0.81 (0.47–1.40)	1.72 (0.82–3.59)	1.71 (0.52–5.60)	1.02 (0.75–1.38)
Hispanic	0.61 (0.27–1.36)	1.50 (0.62–3.64)	1.17 (0.50–2.75)	0.84 (0.53–1.31)
Other	0.66 (0.20–2.24)	1.35 (0.48–3.78)	2.48 (1.16–5.32) *	1.69 (0.86–3.34)
White (ref)
Mother Education ≥ College	1.16 (0.53–2.55)	1.37 (0.91–2.07)	1.02 (0.83–1.25)	0.79 (0.47–1.32)
GAIN	1.24 (1.14–1.34) **	1.18 (1.09–1.28) **	1.19 (1.04–1.35) *	1.08 (1.01–1.15) *
Past Month Alcohol Use	1.07 (0.86–1.34)	1.67 (0.88–3.15)	2.44 (1.26–4.70) **	1.28 (1.06–1.56) *
Past Month Marijuana Use	2.47 (1.60–3.83) **	1.77 (1.18–2.65) **	3.59 (2.49–5.17) **	0.86 (0.55–1.34)
Wald Chi Square†	466.36 **	330.92 **	167.74 **	8.21
AIC	453.89	300.02	300.92	525.48
	Male Youth
Age	1.49 (1.27–1.74) **	1.12 (0.80–1.57)	1.03 (0.76–1.40)	1.24 (1.06–1.44) **
Race/ethnicity
Black	6.64 (1.71–25.75) **	3.44 (1.63–7.24) **	0.85 (0.16–4.69)	3.60 (0.88–14.67)
Hispanic	2.31 (0.74–7.22)	1.96 (1.26–3.06) **	1.77 (0.42–7.43)	2.96 (0.57–15.29)
Other	5.95 (0.51–69.64)	1.49 (0.15–14.55)	0.87 (0.06–12.23)	3.19 (1.66–6.13) **
White (ref)
Mother Education ≥ College	1.53 (0.69–3.42)	0.89 (0.28–2.86)	0.43 (0.16–1.21)	0.65 (0.36–1.18)
GAIN	1.09 (0.92–1.28)	1.08 (1.03–1.14) **	1.33 (1.19–1.48) **	0.95 (0.83–1.09)
Past Month Alcohol Use	3.51 (2.28–5.39) **	2.69 (1.25–5.79) *	2.86 (1.44–5.70) **	1.40 (0.76–2.58)
Past Month Marijuana Use	1.37 (0.94–1.99)	1.45 (0.80–2.61)	1.80 (1.14–2.85) *	1.94 (0.56–6.72)
Wald Chi Square†	47.92 **	454912632.00 **	148.66 **	3.87
AIC	278.34	244.56	198.53	272.78

^*p < .05,

^**p < .01;

^†df = 8

Discussion

The current study adds to the literature by examining whether the association between conduct disorder and a range of risky sexual behaviors differed by gender for a large and diverse sample of youth age 12–18 that were recruited during a primary care visit. Researchers have speculated that there are gender differences in the manifestation of conduct disorder and that risky sexual behavior may be specific to female youth with conduct disorder; however, historically, there has been limited empirical evidence to support this hypothesis. Given that prior research on this topic has been sparse, findings from this study can help inform our understanding of which groups may be at highest risk for engaging in risky sexual behavior. This not only has implications for our understanding of the clinical manifestation of conduct disorder, across genders but also for future screening and intervention efforts.

We found that youth who reported more sexual partners in the last 3 months and who used AOD prior to sex had higher levels of conduct disorder symptoms. Higher levels of conduct disorder symptoms were also observed among youth who reported alcohol or marijuana use in the past month. In addition, we found that alcohol and marijuana use were independent predictors of most of these sexual risk behaviors. This suggests the importance of screening for both risky sexual behavior and AOD use among youth with conduct disorder. In addition, results highlight the importance of multifaceted prevention programs for youth. Because conduct problems, AOD use, and risky sex may occur in combination, it is essential that prevention efforts address each of these behaviors, as failing to address one risk factor may reduce the overall effectiveness of a prevention program.

Regarding the risk of having sex without a condom at the last sexual encounter, results indicated that there were significant main effects of gender and conduct disorder. In addition, stratified models suggested that the association between conduct disorder and likelihood of engaging in condomless sex was significant among female youth but not male youth, which was consistent with our hypothesis. In part, this result should be interpreted in the context of research indicating that female youth are less likely to be the partner carrying a condom (Levin & Robertson, 2002). It is also important to note that this question asked specifically about condom use, and not about use of other contraceptives (e.g., birth control pills or injectable, implantable forms of birth control). However, this finding is potentially meaningful when designing preventive interventions to reduce both unprotected sex, unintended pregnancies, and STIs. For instance, prevention efforts may focus on how although use of contraceptives can protect against pregnancy, condoms are the only birth control that can decrease chances of STIs.

For number of lifetime partners, the association between conduct disorder and having more lifetime partners was significant for females but not males. In addition, although the association between conduct disorder and having more partners in the past three months was significant in both males and females, the association was stronger among females. We expected that the number of partners would be more strongly associated with conduct disorder symptoms in male youth based on work with justice-involved youth (Aalsma et al., 2010; Miller et al., 2010); however, the current sample was recruited from a population of youth visiting a primary care clinic and therefore would likely not have the same degree of symptoms observed in a higher risk sample. We also found that AOD use before last sex was associated with more conduct disorder symptoms for males and females, though the association was stronger in males.

Overall, the current findings make a unique contribution to the larger body of literature examining gender differences in conduct disorder. For three of the sexual risk behaviors, the association with conduct disorder symptoms was stronger – or only present – among females. On the one hand, these results do not fully support the broader category of “sexual risk behavior” as a female-specific manifestation of conduct disorder, as the association between conduct disorder symptoms and certain outcomes – number of partners in the past three months and AOD use before sex –was also significant among males. However, findings do suggest that females with conduct disorder may be at heightened risk for participating in many of these behaviors. It is important to note, though, that the magnitude of differences observed between males and females in the sample were modest. Currently, the diagnostic criteria for conduct disorder do not include gender-specific criteria (American Psychiatric Association, 2013), and there have been concerns that as a result, current criteria may not accurately detect females with conduct disorder (Moffitt et al., 2008). Because accurate diagnoses are important to predicting prognosis and treatment needs (Moffitt et al., 2008), it will be important for future research to attempt to replicate these findings.

Findings also have important implications for clinical practice. Many risky behaviors and behavioral risk factors for future disease first emerge in adolescence; therefore, detecting these concerns is an important part of the preventive medicine role played by primary care providers (Ham & Allen, 2012). For example, it is recommended that primary care clinicians screen offer interventions to prevent initiation of tobacco use, a behavior that is similar in prevalence to sexual activity in our sample (Moyer, 2013). In addition, when adolescents exhibit conduct or behavioral problems, primary care providers may be one of the first points of contact with the healthcare system (Searight, Rottnek, & Abby, 2001). Therefore, it is useful for primary care providers to understand that conduct problems, AOD use, and risky sexual behavior are likely to co-occur, as found in this study, and that the presence of one of these conditions suggests the importance of screening for the others. Research examining adolescent reports of primary care provider behavior indicates that providers screen for sexual behavior approximately 52–61% of the time, screen for AOD use approximately 60–67% of the time, and provide counseling for these behaviors approximately 50–60% of the time (Ozer et al., 2004; Ozer et al., 2005). Factors such as provider self-efficacy, comfort, and confidence with the topic may impact likelihood of screening for risky behaviors (Boekeloo, 2014; Ozer et al., 2004). There is evidence that training and education can increase rates of screening and counseling by primary care providers (Buckelew, Adams, Irwin Jr, Gee, & Ozer, 2008; Duncan et al., 2012; Ozer et al., 2005), and addressing the correlation between conduct problems and risky sexual behavior as part of these types of programs may help to ensure that providers are targeting those adolescents who are at the highest risk. In addition, there are a number of screening instruments available to providers to facilitate this process and limit provider burden (D’Amico et al., 2016). Although these practices should be implemented with both male and female patients, our results suggest that additional counseling regarding multiple sexual partners and condomless sex may be especially important for females with conduct disorder.

There are certain limitations to this study. First, our conduct disorder measure provides information about behaviors consistent with conduct disorder; however, because we do not have information about frequency or severity of these behaviors or the degree to which they cause psychosocial or functional impairment, it is unknown if these youth meet full DSM-5 diagnostic criteria for conduct disorder. That said, the GAIN is used frequently in clinical settings for diagnostic treatment planning, and outcome monitoring purposes, and this scale was designed to be consistent with the formal diagnostic criteria for conduct disorder (Dennis et al., 2008). Further, although the confidential nature of the study was emphasized to participants, youth were asked to self-report engagement in highly sensitive, illegal and/or often stigmatized behaviors, which could affect willingness to indicate participation. However, given recent evidence for the validity of self-report data (Chan, 2009), our extensive focus on protecting confidentiality as part of our procedures, and the fact that rates of AOD use in the sample match national norms (Johnston et al., 2013), we do not believe this issue affected our results. Moreover, research has suggested that parents report lower estimates of participation in risky behavior than their children, including sexual activity, AOD use, and antisocial behaviors (Jones et al., 2017; Stanton et al., 2000; Yang et al., 2006). Therefore, we believe that basing our analyses on youth-reported data yields a more accurate estimate of the true prevalence of these behaviors.

In addition, our sexual behavior questions did not ask whether the sexual activity was consensual, which is a limitation given evidence for higher rates of victimization seen in youth who engage in risky sexual behavior (Alleyne, Coleman-Cowger, Crown, Gibbons, & Vines, 2011; Champion et al., 2004; Upchurch & Kusunoki, 2004). We also focused broadly on sexual intercourse rather than specific sexual behaviors (e.g., oral or anal sex). Analyses are also based on cross-sectional data; thus, inferences regarding causality or the direction of these associations cannot be made. Future longitudinal work is needed to better understand how these behaviors may unfold over time during this developmental period. Finally, some eligible youth declined to participate because they were visiting a clinic for family planning purposes and did not want their parents to know, and it is possible that the conduct disorder symptoms and sexual behavior of these youth differed systematically in some way from youth who opted to participate.

In sum, this study contributes to the literature examining gender differences in the manifestation of conduct disorder. By focusing on a large sample of diverse youth recruited from primary care clinics, findings are likely more representative of the larger population of youth who engage in behaviors consistent with conduct disorder, but who are not in secure settings (e.g., juvenile justice settings or residential treatment settings). Therefore, results are likely to have clinical utility and to be generalizable to health care settings. Future research should confirm findings of this study with other populations of adolescents with conduct disorder (e.g., those who are seeking mental health treatment, or those who make formal contact with the juvenile justice system). In addition, given research that many youth who participate in risky behaviors as adolescents desist into adulthood (Moffitt, 1993; Mulvey et al., 2010), it will also be important to examine the longitudinal association between conduct disorder and risky sexual behavior across adolescence and as youth transition into young adulthood.