Pathways to Substance Use and Sexual Risk Behavior Among College Students with ADHD Symptomatology

While growing numbers of students with attention deficit hyperactivity disorder (ADHD) symptomatology are pursuing postsecondary education, there is a dearth of information concerning the social functioning of these students. ADHD symptomatology has been strongly linked with risk behaviors that contribute to chronic health problems, including substance use and risky sexual behavior, resulting in twice the health care costs for these students in the United States. Despite such critical findings, specific pathways between ADHD and substance use and sexual risk, have not been identified. A large body of literature has demonstrated that individuals with ADHD are at greater risk for developing externalizing behavior problems, which in turn appear to predict substance use and sexual risk behavior. Evidence also suggests that individuals with ADHD symptomatology often exhibit executive function (EF) deficits, and several studies have linked executive dysfunction to substance use problems and sexual risk behavior. Therefore, the purpose of the present study was to: a) examine the relationship among ADHD symptomatology, externalizing symptomatology, EF deficits, substance use, and sexual risk behavior among N=411 college students; b) propose and test three nested, latent variable models (i.e., a mediation, full, and a direct effects model) and identify significant paths between the variables; and c) examine the three latent variable models and determine which model best represents the relationship between the variables. Overall, results revealed significant correlations among ADHD symptomatology, externalizing symptomatology, EF deficits, substance use, and sexual risk behavior. While the mediation and full models demonstrated specficiation errors that could not be resolved into meaningful solutions, significant pathways were identified within the direct latent variable model, including paths between ADHD symptomatology, externalizing symptomatology, EF deficits, substance use, and sexual risk behavior, respectively. Furthermore, the direct model proved to best represent the data, over and above the two other latent variable models. The present findings have implications for public health policy, particularly as it relates to the college population. Limitations of the study and suggestions for future research are discussed.


Introduction
Attention-deficit/hyperactivity disorder (ADHD) is a chronic neurodevelopmental disorder, characterized by clinically significant symptoms of inattention, hyperactivity, and impulsivity, affecting approximately 2-7% of individuals in the United States (American Psychiatric Association, 2013). ADHD has been associated with significant difficulties in psychosocial and academic adjustment, including disruptive behavior, lower grade point averages, academic underachievement, school dropout, and higher rates of comorbid psychopathology (American Psychiatric Association, 2013; Barkley, 2008;Bussing, Mason, Bell, Porter, & Garvan, 2010). Despite these increased risks, recent research has found that increasing numbers of high school students with ADHD are pursuing higher education (Weyandt & DuPaul, 2013;Wolf, Simkowitz, & Carlson, 2009). Although the exact prevalence of the disorder in the college population is unknown, a recent national survey indicates that approximately 6% of first-year college students report being diagnosed with ADHD (Eagan et al., 2014). Weyandt and DuPaul (2013) emphasized the dearth of information available and stressed the need for studies to explore the prevalence, nature, and academic and social functioning of college students with ADHD.
ADHD has been associated with risk behaviors that contribute to chronic health problems, including substance use and risky sexual behavior (Flory, Molina, Pelham, Gnagy, & Smith, 2006;Harty, Galanopoulos, Newcorn, & Halperin, 2013;Schoenfelder & Kollins, 2015). Young people with ADHD, for example, have twice the health care costs in the United States (Leibson, Katusic, Barbaresi, Ransom, & O'Brien, 2001) and increased mortality rates compared to those without the disorder (Dalsgaard, Ostergaard, Leckman, Mortensen, & Pedersen, 2015). Despite such critical findings, specific pathways between ADHD and substance use and sexual risk, have not been identified (Flory et al., 2006;Molina & Pelham, 2014). A large body of literature, however, has demonstrated that children and adolescents with ADHD are at a greater risk for developing behavior problems, including oppositional defiant disorder (ODD) and conduct disorder (CD; Barkley, Murphy, & Fischer, 2008), which in turn appear to predict substance use (Zucker, 2006) and risky sexual behavior (Flory et al., 2006).
Difficulties with impulsivity in conjunction with EF deficits have also been associated with greater sexual risk behavior (Barkley et al., 2008;Quinn & Fromme, 2010). In summary, although research has identified ADHD symptomatology, externalizing symptomatology, and EF deficits as increasing the risk for substance use and sexual risk behavior, to date no studies have systematically examined potential pathways between these variables. The present study addressed this void in the literature by proposing and testing three latent variable models concerning the relationship between ADHD symptomatology, externalizing symptomatology, EF deficits, substance use, and sexual risk behavior, in a sample of college students with and without ADHD symptomatology. Specifically, goodness of fit indices were hypothesized to be strongest for the mediational model relative to the full and direct effects models.

ADHD Symptomatology and Substance Use
Substance use disorders are health outcomes well recognized to co-occur with ADHD. For example, adults with ADHD have been found to use alcohol and other drugs at higher rates than those in the general population (Lee, Humphreys, Flory, Liu, & Glass, 2011). Indeed, an estimated 15.2% of adults with ADHD meet criteria for a substance use disorder, a rate that is almost three times greater than among adults without ADHD (Kessler et al., 2006). A recent meta-analysis documented that one in four substance dependent persons had an ADHD diagnosis during their lifetime (van Emmerik-van Oortmerssen et al., 2012), while another study by Lee and colleagues (2011) suggested that youth with ADHD have at least 1.5 times the average risk of developing dependence on nicotine, alcohol, marijuana, cocaine, and other drugs.
Interestingly, young adult ADHD research has yielded mixed results with regard to the relationship between ADHD and substance use problems. Some studies have documented that young adults with ADHD tend to report higher rates of underage consumption of alcohol, marijuana use, and experimentation with other illicit drugs compared to their non-ADHD peers (Bidwell, Henry, Willcutt, Kinnear, & Ito, 2014;Dunne, Hearn, Rose, & Latimer, 2014;Langley et al., 2010;Lee et al., 2011), in addition to a faster progression and less odds at recovering from substance use disorders (Fuemmeler, Kollins, & McClernon, 2007;Molina et al., 2009). Additionally, Upadhyaya and Carpenter (2008) reported a positive correlation between ADHD symptom severity and alcohol and marijuana use among a group of young adults. Likewise, Upadhyaya et al. (2005) found that college students with ADHD had more past-year tobacco and marijuana use than their peers. Alternatively, other research has demonstrated little to no alcohol or substance use differences between young adults with and without ADHD (Baker, Prevatt, & Proctor, 2012;Bussing et al., 2010;Rabiner, Anastopoulos, Costello, Hoyle, & Swartzwelder, 2008). For example, Rabiner and colleagues (2008) found that students with past or current ADHD were not more likely than other college students to report consuming alcohol. Likewise, Baker and colleagues (2012) reported no illicit substance use differences among college students with and without ADHD. While research linking ADHD and substance use in college students is inconsistent, perhaps discrepant findings are due to the lack of thoroughly confirmed ADHD diagnoses within the samples. The present study, however, implemented comprehensive clinic-based ADHD criteria and rigorous methodological strategies, and may therefore help address the inconsistent findings in the literature.
While it remains unclear whether young adults with ADHD consume more alcohol and take part in greater rates of substance use compared to their non-ADHD peers, several studies have suggested that college students with ADHD engage in more problematic drinking behaviors, resulting in greater alcohol-related consequences (Baker et al., 2012;Glass & Flory, 2012;Lee et al., 2011;Rooney, Chronis-Tuscano, & Huggins, 2012;Rooney, Chronis-Tuscano, & Yoon, 2011;Wilens & Biederman, 2006).
More specifically, studies report that college students with ADHD are more likely than their peers to have difficulty limiting their alcohol consumption, consume alcohol until they "black out", drive under the influence of alcohol or illicit substances, experience injuries resulting from fights while under the influence, and have more alcohol-related conflict with their significant other (Baker et al., 2012;Glass & Flory, 2012;Lee et al., 2011;Rooney et al., 2012;Wilens & Biederman, 2006). Research also suggests that increased alcohol use among college students with ADHD is linked to greater impairment in daily activities, social relationships, and sexual interactions (Langberg, Dvorsky, Kipperman, Molitor, & Eddy, 2014).
In summary, empirical evidence suggests that college students with ADHD symptomatology are at elevated risk for negative health consequences associated with substance use, and greater research is needed to develop effective substance use prevention and intervention programs. In particular, research is needed to identify the specific factors that predict substance use in college and the potential mediators in such relationships that, in turn, could be targeted via prevention and intervention.

ADHD Symptomatology and Sexual Risk Behavior
Although studies examining the effects of ADHD on risky sexual behavior are generally lacking, preliminary data suggest that individuals with ADHD lead higher-risk sexual lifestyles (Brown et al., 2010;Flory et al., 2006;Hosain, Berenson, Tennen, Bauer, & Wu, 2012;Nigg, 2013). For example, in the Milwaukee Young Adult Outcome Study, Barkley (2006) reported that young adults with ADHD tended to have sexual intercourse at an earlier age, had more sexual partners, used less contraception, which may lead to teen pregnancy and sexually transmitted infections (STIs). By the time adolescents in the study turned 20, the ratio of births by the ADHD group to the control was 42:1 (Barkley, 2006). Flory and colleagues (2006) found similar results among a group of young men with ADHD who reported a number of unsafe sexual behaviors including earlier debut of sexual activity and intercourse, an increased number of sexual partners, and more casual sex. Male students with ADHD were also less likely to use contraception, resulting in greater rates of STIs and partner pregnancies (Flory et al., 2006).
More recently, a retrospective study of college students found that women with ADHD reported more unprotected sex not only than women without ADHD but also more so than men on average (Huggins, Rooney, and Chronis-Tuscano, 2015). Hosain and colleagues (2012) also found that young adult women with ADHD symptomatology reported risky sexual behaviors, including sex before 15 years of age, more risky sexual partners in their lifetime, greater numbers of sex partners in the last 12 months, less condom use in the last 12 months, alcohol use before sex in the last 12 months, having traded sex in their lifetime, and having been diagnosed with sexually transmitted infection (STI) in their lifetime.
In summary, while preliminary evidence suggests that both college student men and women with ADHD are at elevated risk for the negative health consequences associated with risky sexual behavior (e.g., sexually transmitted infections), more research is needed to better understand the specific factors that predict different types of risky sexual behavior in addition to potential mediator variables. Such research would substantially aid future health promotion efforts.

Externalizing Symptomatology and Substance Use
The externalizing pathway is theorized to begin with childhood externalizing symptomatology (e.g., aggression and conduct problems), early onset substance use, increases in antisocial behavior, and the eventual onset of Substance Use Disorders (SUDs; Tarter et al., 2003;Zucker et al., 2006). Externalizing symptoms, therefore, reflect behavioral disinhibition, also referred to as the inability to inhibit undesirable or restricted behaviors (Iacono, Malone, & McGue, 2008). In fact, current risk models suggest that underlying deficits in behavioral inhibition and a high-risk environment may place children at most risk for externalizing behaviors (e.g., Hussong, Curran, & Chassin, 1998;Zucker et al., 2006). Children with ADHD, therefore, may be at greater risk for externalizing symptomatology given that behavioral disinhibition is a core deficit of the disorder (Weyandt & DuPaul, 2006). Indeed, disruptive behavior disorders, such as Oppositional Defiant Disorder (ODD) and Conduct Disorder (CD), are especially common comorbid conditions in children and adolescents with ADHD (Fischer, Barkley, Smallish, & Fletcher, 2002).
Externalizing symptoms are consistently positively correlated with substance use in adolescence and young adulthood (Hawkins, Catalano, & Miller, 1992;Hussong et al., 1998;King, Iacono, & McGue, 2004;Zucker, 2006). In the Minnesota Twin Family Study, externalizing psychopathology predicted experience with alcohol, nicotine and cannabis by age 14, as well as regular and advanced experience with these substances (King et al., 2004). Another study monitored children with ADHD and ODD/CD through adolescence, and found that ODD/CD symptoms were predictive of illicit drug use and CD symptoms in adolescence (Molina & Pelham, 2003). In the same study, children with ADHD reported greater alcohol symptom scores, with childhood inattentive symptom severity being the most predictive of several negative substance use outcomes (Molina & Pelham, 2003). Furthermore, persistence of ADHD and adolescent CD were each associated with elevated substance use behaviors relative to controls (Molina & Pelham 2003). Other studies examining adults with ADHD demonstrate the rate of comorbid conditions, including substance use disorders and antisocial personality disorder (Barkley et al., 2008;Biederman, Petty, Evans, Small, & Faraone, 2010;Garcia et al., 2012).
Interestingly, a behavioral genetics study suggested that disruptive disorder symptoms (i.e., ODD/CD) and substance use may share a common genetic predisposition for disinhibited behavior (Iacono et al., 2008). Collectively, these studies suggest the externalizing pathway may indeed be the primary pathway of risk for the development of SUDs.

Externalizing Symptomatology and Sexual Risk Behavior
Similar to substance use, externalizing symptomatology has been associated with risky sexual behavior (Barkley, 2006;Brown et al., 2010;Sarver, McCart, Sheidow, & Letourneau, 2014). For example, Brown and colleagues (2010) reported that adolescents meeting criteria for an externalizing disorder (i.e., ODD, CD, and ADHD) were significantly more likely to report a lifetime history of vaginal or anal sex. Furthermore, in a sample of adolescents, the relationship between ADHD symptoms and risky sexual behavior emerged only among youth with clinically elevated conduct problems and problematic marijuana use (Sarver et al., 2014), suggesting that early identification and treatment of such conditions may be important for sexual risk prevention. Longitudinal studies have demonstrated that a childhood history of disruptive behaviors is associated with early initiation of intercourse and greater rates of adolescent sexual activity among boys (Barkley, 2006;Ramrakha et al., 2007), and early initiation of intercourse, multiple sex partners, and increased rates of teen pregnancies among adolescent girls (Ramrakha et al., 2007). This line of research has important implications for sexual risk prevention, with accumulating evidence demonstrating support for prevention and treatment programs that address deficits or introduce protective factors important in decreasing externalizing behaviors (e.g., Cutuli et al., 2013;Lochman, Powell, Boxmeyer, & Jimenez-Camargo, 2011).

Executive Function Deficits and Substance Use
Executive functioning is a multifaceted construct that has often been defined as the higher-order cognitive abilities that underlie self-regulation, impulse control, decision-making, strategic planning, cognitive flexibility, and goal-directed behavior (Weyandt, 2005;Willcutt, Doyle, Nigg, Faraone, & Pennington, 2005). Although EF deficits are not characteristic of all individuals with ADHD, as noted by Weyandt (2009), a substantial body of research has found that individuals with ADHD often exhibit executive function deficits (Barkley 2012;Murphy, Barkley, & Bush, 2001;Nigg et al., 2006;Weyandt, 2009;2014). For over two decades, deficits in EF have been linked to substance use behaviors (Molina & Pelham, 2003). Such difficulties may begin at an early age, with childhood EF deficits identified as a predictor of drug use in early adolescence (Tarter et al., 2003). Self-regulation (Quinn & Fromme, 2010) and impulse control (i.e., behavioral disinhibition; Nigg et al., 2006) deficits, in particular, have been the major focus of substance use research examining EF.
Poor self-regulation, one underlying construct of EF, is among the strongest personality predictors of alcohol use (Hittner & Swickert, 2006;Hustad, Carey, Carey, & Maisto, 2009;Quinn & Fromme, 2010) and such findings have been replicated among college student samples. For example, Gottfredson and Hussong (2013) examined the role of affective self-regulation on alcohol use, and found that poor self-regulation was predictive of increased drinking frequency and higher levels of self-reported drinking to cope their affect variability. Alternatively, students with high self-regulation inversely predicted heavy episodic drinking and alcohol-related problems (Fromme & Quinn, 2010). Interestingly, difficulties with behavioral self-regulation have been linked to marijuana use, while emotional self-regulation deficits have been predictive of marijuana-related problems (Dvorak & Day, 2014). Perhaps, behavioral self-regulation difficulties place college students at risk for substance use while emotional selfregulation places them at risk for the negative consequences resulting from their substance use.
In addition to self-regulation, difficulties with impulse control have been linked to substance use (Dvorak & Day, 2014). Tarter and colleagues (2003) identified behavioral disinhibition as a stronger predictor of substance use disorders in young adulthood, over and above teenage substance use. In a study of young adults (i.e., 21.1-22.3 years of age) diagnosed with ADHD compared to nonclinical controls, three dimensions of impulse control (i.e., attentional inhibition, response inhibition, and sensation seeking) were predictive of self-reported alcohol use (Weafer, Milich, & Fillmore, 2011). Attentional inhibition, in particular, predicted alcohol consumption in the ADHD group, suggesting that specific types of behavioral disinhibition may contribute to elevated rates of substance use among individuals with ADHD (Weager et al., 2011). In a related study by Rooney and colleagues (2012), impulse control deficits accounted for heightened rates of alcohol use among college students diagnosed with ADHD. College students with EF deficits may take part in greater substance use, and despite the need for work in this area, very few studies have examined the relationship between EF and substance use.

Executive Function Deficits and Sexual Risk Behavior
Similar to the literature examining the relationship between EF and substance use, self-regulation and impulse control have been the major focus of research linking EF deficits to sexual risk behavior (Crockett, Raffaelli, & Shen, 2006;Epstein et al., 2014;Moilanen, 2015;Raffaelli & Crockett, 2003;Quinn & Fromme, 2010). Raffaelli and Crockett (2003) demonstrated an association between self-regulatory skills in early adolescence and risky sexual behavior in late adolescence among a national sample of boys and girls. More recently, college students with high self-regulation inversely predicted sexual risk behavior (e.g., unprotected sex), even when controlling for gender and risk factors (Fromme & Quinn, 2010). In yet another study, Moilanen (2015) found that young adults with long-term self-regulation skills reported fewer sexual risk behaviors including later initiation of oral sex and coitus, fewer lifetime coital partners, increased likelihood of condom and other contraceptive use at last intercourse, and low levels of coitus risk, while participants with short-term self-regulatory skills reported a reduced likelihood of condom use and greater overall coital risk. While deficits in selfregulation have been linked to risky sexual behaviors, a major limitation of studies includes the various ways in which self-regulation and its associated components are identified, defined, and measured (Berger, 2011). For example, previous research has coined several different terminologies (e.g., self-control, vigilance, inhibition) to refer to similar components that overlap with one another (Berger, 2011;Moilanen, 2015;Muraven & Baumeister, 2001 Feldman and Brown (1993) found that boys' self-restraint during childhood was inversely associated with the number of sexual partners years later. Furthermore, in a cross-sectional study of adolescent girls seeking clinic services for either contraceptive advice or termination of a pregnancy, deficits in impulse control significantly predicted membership in the pregnancy group (Rawlings, Boldero, & Wiseman, 1995). More recently, Epstein et al. (2013) demonstrated that adolescent behavioral disinhibition had significant effects on sexual risk taking, which extended into the participants' adulthood. Epstein and colleagues (2014) reported that behavioral disinhibition predicts sexual risk behavior over and above previously identified risk factors.

Purpose of the Study
Given the potentially destructive and life-threatening outcomes of substance use and sexual risk behavior, it is critical that the pathways to such behaviors are identified among college students. To date, however, no study has thoroughly examined the complex relationship between ADHD symptomatology, externalizing symptomatology, EF dysfunction, substance use, and sexual risk behavior, in a sample of college students with and without ADHD. Therefore, the primary purpose of the present study was to propose and test three latent variable models designed to identify the pathways to substance use and sexual risk behavior among college students.

Research Hypotheses
Based on previous empirical findings concerning substance use and sexual risk behavior, it was hypothesized that: 1) ADHD symptomatology, externalizing symptomatology, EF deficits, substance use, and sexual risk behavior would be significantly correlated. A more detailed description of the variables of interest can be found in Appendix A.
2) A mediational latent variable model of ADHD symptomatology, externalizing symptomatology, EF deficits, substance use, and sexual risk behavior would demonstrate statistically significant pathways between the independent variables, mediators, and dependent variables.
3) A mediational latent variable model of ADHD symptomatology, externalizing symptomatology, EF deficits, substance use, and sexual risk behavior would best represent the relationship between the variables, over and above two other nested, latent variable models (i.e., a full model and a direct effects model). More specifically, goodness of fit indices were hypothesized to be strongest for the mediational model relative to the full and direct effects models.

Study Procedure
The present study employed data collected during the initial year of a Information concerning participant demographics can be found in Table 1.

Measures
Demographic Questionnaire. Students completed a demographic form to indicate their gender, age, race, and ethnicity. Additionally, students were asked to self-report their family composition (i.e., number of siblings, parent's marital status, parental educational level, and parental occupation).
Conners' Adult ADHD Rating Scale -Self-Report: Long Version (CAARS). To assess current ADHD symptomatology, the Conners' Adult ADHD Rating Scale (CAARS) was administered. The CAARS is a 66-item standardized symptom rating scale utilized to assess ADHD in adults (Conners, Erhardt, & Sparrow, 1999). Items are rated on a 4-point Likert scale ranging from 0 (i.e., not at all/never) to 3 (i.e., very much/very frequently).
This instrument consists of the following eight subscales with respective reliability coefficients for males and females: 1) inattention/memory problems ( reported to have sufficient factorial, discriminant, and construct validity (Conners et al., 1999). Results have demonstrated the scale's ability to identify ADHD symptomatology (Conners et al., 1999). The DSM-IV inattentive (IA) symptoms (e.g., "I don't plan ahead", "I have trouble listening to what other people are saying") subscale T-score and the DSM-IV hyperactive-impulsive (HI) symptoms ("I am always on the go, as if driven by a motor", "I am a risk-taker or daredevil") subscale T-score served as two continuous independent variables in the present study.

Behavior Rating Inventory of Executive Function-Adult Version (BRIEF-A). The BRIEF-
A was developed by Gioia, Isquith, Guy, and Kentworth to assess executive functioning in adults, aged 18 years and older. A higher score indicates greater executive dysfunction (Gioia, Isquith, Guy, & Kentworth, 2000). Modeled after the ADHD-RS, the EBRS first lists the 8 ODD symptoms, followed by the 12 (out of 15) CD items deemed developmentally appropriate for a college population.
Similar to the ADHD-RS, each EBRS item is rated on a 0 (not at all) to 3 (very much) scale reflecting the degree to which items characterize a participant's behavior over the past six months. Each EBRS item is summed to yield separate ODD and CD symptom severity total scores, which were entered separately as continuous mediator variables.

Sexual Risk Survey (SRS).
The SRS is a 23-item questionnaire, where respondents are asked to report the frequency with which they participated in each of a range of sexual risk behaviors during the preceding 6 months. Frequencies of 0 are coded as "0", and the remaining frequencies are coded into four ordinal categories (i.e., 1 to 4) consistent with the recoding procedure developed by Turchik, Walsh, and Marcus (2015) employed to address positively skewed sexual risk frequency data. For example, item 1 "number of sexual behavioral partners" would be coded as follows, "0= 0", "1= 1", "2-3= 2", "4= 3", and frequencies "5+= 4" (Turchik et al., 2015). Total scores range from 0 to 92 and thus, a higher score indicates greater rates of sexual risk taking (Turchik et al., 2015). has adequate concurrent, construct, and discriminative (i.e. the ability to discriminate between low-, moderate-, and high-risk substance users) validity (Humeniuk et al., 2008).
In the present study, all substances on the scale were assigned to one of three classifications (i.e., stimulants, depressants, or other) as reported by the Australian Government Department of Health (2004). The three categories were entered separately as dependent variables.

Data Analysis
Five different sets of analyses were conducted to test the hypotheses that 1) ADHD symptomatology, externalizing symptomatology, EF deficits, substance use, and sexual risk behavior would be significantly correlated, 2) A mediational latent variable model of ADHD symptomatology, externalizing symptomatology, EF deficits, substance use, and sexual risk behavior would demonstrate statistically significant pathways between the independent variables, mediators, and dependent variables, and 3) A mediational latent variable model of ADHD symptomatology, externalizing symptomatology, EF deficits, substance use, and sexual risk behavior would best represent the relationship between the variables, over and above two other nested, latent variable models (i.e., a full model and a direct effects model). More specifically, goodness of fit indices were hypothesized to be strongest for the mediational model relative to the full and direct effects models. The first two analyses related to the exploration of the indicator variables in the present study, while the third analysis was conducted to examine the psychometric properties of the measurement model. The remaining two analyses related to the latent variable models and their goodness of fit. More specifically, analyses included: a) a descriptive examination of all indicator variables, b) Pearson correlation analyses among indicator variables, c) confirmatory factor analyses to examine the psychometric adequacy of the hypothesized measurement model, and d) latent variable modeling was conducted to examine whether a mediational latent variable model of ADHD symptomatology, externalizing symptomatology, EF deficits, substance use, and sexual risk behavior would best represent the relationship between the variables, over and above two other nested, latent variable models (i.e., a full model and a direct effects model), further, e) the latent variable models were examined in order to determine goodness of fit, while considering while considering both theory and parsimony (Bentler & Mooijaart, 1989).
More specifically, maximum likelihood estimation was used to generate chi-squared (χ2) values as a measure of goodness of fit, which was expected to be low relative to the degrees of freedom. Additionally, the root mean square error of approximation (RMSEA; Steiger & Lind, 1980) was expected to be at or below 0.05 (Steiger, 1998), the Bentler Comparative Fit Index (CFI; Bentler, 1990) was expected to be at or above 0.90, and standardized residuals below |.20|.

Missing Data
Given that participants completed the study assessments across three separate sessions, and that not all participants completed all three sessions, sample sizes differ across measures/analyses. As for missing data patterns, the percentage of missing data was generally lower for items administered earlier in the survey (e.g., the CAARS) compared to items administered later in the survey (e.g., EBRS, BRIEF, SRS, ASSIST).
The CAARS was completed during the first assessment session, the EBRS was completed during the second assessment, and the BRIEF, SRS, and ASSIST were completed during the third assessment. Thus, due to attrition, the correlation analyses included a sample range of N= 390-411, while the CFA and LVM analyses had an N= 390.

Pearson Correlation Analyses
Prior to conducting confirmatory factor and latent variable modeling analyses, preliminary bivariate correlational analyses were performed. Results, delineated in Table   3

Structural Equation Modeling
Confirmatory factor analyses were performed on the measurement model, testing the sufficiency of the model and associations among the latent variables (Bentler, 2004).  ADHD symptomatology was expected to significantly predict both executive dysfunction and externalizing symptomatology. In turn, executive dysfunction and externalizing symptomatology were expected to significantly predict sexual risk and substance use. The association between ADHD symptomatology, sexual risk, and substance use was expected to become nonsignificant when accounting for the effects of executive dysfunction and externalizing symptomatology, suggesting full mediation (Baron & Kenny, 1986). The second alternative full model suggested that additional paths between the latent variables of ADHD symptomatology (independent variable), sexual risk (dependent variable), and substance use (dependent variable) best represented the data, suggesting partial mediation. In contrast, the third alternative direct model posited that there was a direct path between ADHD symptomatology, externalizing symptomatology, executive dysfunction, sexual risk, and substance use, with no mediating effects.

Results
As stated previously, the mediation model was expected to best represent the data, considering model fit indices, theory, and parsimony. Results of the mediation and full models, however, resulted in specification errors that could not be resolved into Although results were supportive of the direct model, potential problems in the interpretation of these findings should be taken into consideration. First, it is important to note that fit indices of the direct model were relatively poor. For example, the root mean square error of approximation was above the suggested 0.05 level (Steiger & Lind, 1980) while the Bentler Comparative Fit Index (CFI; Bentler, 1990), reached the 0.90 level.
Additionally, 4 out of the 20 largest standardized residuals exceeded the criterion of |.20|., ranging from .204 to .225. This may have contributed to the overall poor model fit.

Discussion
Although research has identified ADHD symptomatology, externalizing symptomatology, and EF deficits as predictors of substance use and sexual risk behavior and their damaging effects, the present study was the first to systematically examine potential pathways between these variables. Specifically, it was hypothesized that: 1) ADHD symptomatology, externalizing symptomatology, EF deficits, substance use, and sexual risk behavior would be significantly correlated (see Appendix A for a more detailed description of the variables of interest); 2) a mediational latent variable model of ADHD symptomatology, externalizing symptomatology, EF deficits, substance use, and sexual risk behavior would demonstrate statistically significant paths between the independent variables, mediators, and dependent variables; 3) a mediational latent variable model of ADHD symptomatology, externalizing symptomatology, EF deficits, substance use, and sexual risk behavior would best represent the relationship between the variables, over and above two other nested, latent variable models (i.e., a full model and a direct effects model). Specifically, goodness of fit indices were hypothesized to be strongest for the mediational model relative to the full and direct effects models.

Pearson Correlation Findings
Preliminary correlational analyses were partially supportive of hypothesis 1) that ADHD symptomatology, externalizing symptomatology, EF deficits, substance use, and sexual risk behavior would be significantly correlated. More specifically, results revealed that ADHD-IA symptomatology was significantly and positively correlated with ADHD-HI symptomatology, ODD symptomatology, CD symptomatology, behavioral regulation difficulties, metacognitive dysfunction, sexual risk taking with uncommitted partners, risky sex acts, intent to engage in risky sexual behavior, stimulant use, depressant use, and other drug use. Alternatively, impulsive sexual behaviors and risky anal sex acts were not significantly associated with ADHD-IA symptomatology. While research has linked ADHD symptomatology and sexual risk behavior Flory et al., 2006) Sarver, McCart, Sheidow, andLetourneau (2014) reported that ADHD-HI, but not ADHD-IA symptoms were associated with risky sexual behavior. Therefore, it is plausible that ADHD-IA and ADHD-HI predict different types of behavior, including sexual risk behavior, thereby providing a potential explanation for why ADHD-IA was not associated with impulsive sexual behaviors and risky anal sex acts.
Similarly, ADHD-HI symptomatology was significantly and positively correlated with ODD symptomatology, CD symptomatology, behavioral regulation difficulties, metacognition dysfunction, sexual risk-taking with uncommitted partners, risky sex acts, depressant use, and other drug use. However, unlike ADHD-IA, ADHD-HI symptomatology was also positively associated with impulsive sexual behaviors, which is consistent with previous research (Sarver et al., 2014). Findings also suggested that ADHD-HI symptomatology was not significantly associated with the intent to engage in risky sexual behavior, which conceptually makes sense, as college students with hyperactive-impulsive symptomatology may not consider their risk behavior prior to the time in which it occurs. Furthermore, ADHD-HI was not significantly correlated with risky anal sex acts and stimulant use, as was predicted. One plausible explanation as to why ADHD-HI was not associated with stimulant use, but with depressant and other drug use, is because the substances assigned to the stimulant classification (e.g., cocaine), typically provide increased motor activity, which is already characteristic of students with ADHD-HI symptomatology (Tseng, Henderson, Chow, & Yao, 2004). Despite this plausible explanation, a recent meta-analytic review linked childhood ADHD with stimulant (i.e., cocaine) abuse or dependence in adolescence and young adulthood (Lee et al., 2011). Provided these mixed results, future research must examine the unique relationship between ADHD-HI symptomatology and substance abuse among various stimulants (e.g., cocaine, speed, ecstasy).
Significant positive correlations were also discovered among ODD symptomatology and CD symptomatology, behavioral regulation difficulties, metacognitive dysfunction, sexual risk taking with uncommitted partners, risky sex acts, impulsive sexual behaviors, depressant use, and other drug use. In contrast, intent to engage in risky sexual behaviors, risky anal sex acts, and stimulant use were not significantly correlated with ODD symptomatology. Similarly, CD symptomatology was significantly and positively correlated with behavioral regulation difficulties, metacognition dysfunction, sexual risk taking with uncommitted partners, risky sex acts, impulsive sexual behaviors, depressant use, and other drug use. In contrast to ODD symptomatology, however, CD symptomatology was associated with the intent to engage in risky sexual behaviors and stimulant use. Risky anal sex acts, however, was the only variable not significantly correlated with CD symptomatology. One plausible explanation for why risky anal sex acts were not associated with ADHD-IA symptomatology, ADHD-HI symptomatology, ODD symptomatology, or CD symptomatology relates to the small number of participants reporting at least one anal sex behavior (n=53)  dysfunction was not associated with the intent to engage in risky sexual behavior or stimulant use. Interestingly, metacognitive dysfunction was associated with risky anal sex acts, one of the riskiest sexual behaviors, highlighting the important role of selfawareness, the ability to self-monitor, and problem solve, in order to prevent risk behavior. Therefore, future risk prevention intervention programs with college students should include activities to help build executive function skills, including metacognition.

Indeed, previous HIV prevention intervention research documented that an intervention
increasing self-monitoring resulted in increased protected sex with sexual partners, and changes in attitudes conducive to reducing risk (Lightfoot, Rotheram-Borus, Comulada, Gundersen, & Reddy, 2007). More recently, researchers have begun to identify preferences for mobile health applications used to increase self-monitoring and selfmanagement, providing support for future mobile risk prevention interventions (Ramanathan, Swendeman, Comulada, Estrin, & Rotheram-Borus, 2013). Based on these findings, mobile self-monitoring interventions at the college level may be a feasible and efficacious way to prevent sexual risk behavior and its detrimental outcomes.

Structural Equation Modeling Findings
Confirmatory factor analyses were employed to validate the psychometric adequacy of the measurement model. One indicator of sexual risk behavior, risky anal sex acts, was dropped based on it poor factor loading. As described previously, the risky anal sex acts scale has relatively poor internal consistency and in the present study, low reporting, which may in part explain the problems associated with this factor. Once this indicator was dropped, a subsequent confirmatory factor analysis demonstrated statistically significant loadings among all indicators and their respective latent variables, ranging from 0.438 to 0.928. Furthermore, the latent variables demonstrated strong intercorrelations, ranging from 0.186 to 0.913.
Next, latent variable modeling analyses were employed. Results were not supportive of hypothesis 2) that a mediational latent variable model of ADHD symptomatology, externalizing symptomatology, EF deficits, substance use, and sexual risk behavior would demonstrate statistically significant paths between the independent variables, mediators, and dependent variables. However, as depicted in Figure 2 (i.e., the direct model), the ADHD symptomatology latent variable predicted the two proposed latent mediator variables, externalizing symptomatology and executive dysfunction.
Results of the present study and previous research suggest that college students with ADHD symptomatology may have substantial problems managing not only their ADHD symptoms, but also their executive functioning and externalizing behaviors associated with these symptoms.
In addition to the significant pathways between the latent independent variable, ADHD symptomatology, and the two proposed latent mediator variables, executive dysfunction and externalizing symptomatology, the direct model demonstrated significant pathways between ADHD symptomatology, substance use, and sexual risk behavior, respectively. While specification errors did not allow for examination of potential mediators, it is plausible, that ADHD symptomatology underlying and/or comorbid with externalizing problems is most predictive of substance use behavior. In fact, previous research suggests that ADHD symptomatology is an independent risk factor for substance use problems (Frodl, 2010;van Emmerik-van Oortmerssen, 2012;Wilens & Spencer, 2010), however, a combination of ADHD and externalizing problems places individuals at greatest risk for substance use disorders (Flory & Lynham, 2003). Thus, ADHD and externalizing symptomatology appear to play an important role in predicting substance use behavior, however, the way in which the two variables interact requires further investigation.
Latent variable modeling analyses were not supportive of hypothesis 3) that a mediational latent variable model of ADHD symptomatology, externalizing symptomatology, EF deficits, substance use, and sexual risk behavior would best represent the relationship between the variables, over and above two other nested, latent variable models (i.e., a full model and a direct effects model). Results demonstrated the direct model as best fit, with the mediation and full models producing specification errors. While the direct model was preferred, these results must be considered cautiously.
As mentioned previously, the overall model fit was relatively poor, for example, the chisquared statistic was quite large, the root mean square error of approximation was above the suggested 0.05 level (Steiger & Lind, 1980), while the Bentler Comparative Fit Index (CFI; Bentler, 1990) should also consider offering universal substance use and sexual risk prevention programs highlighting effective coping strategies, given the success of these programs among adolescent populations (Griffin & Botvin, 2010).

Limitations and Future Directions
Although this study is the first to rigorously examine the complex relationship between ADHD symptomatology, externalizing symptomatology, EF dysfunction, substance use, and sexual risk behavior, in a sample of college students with and without ADHD, several limitations of the present study should be considered. First, the study was cross-sectional, which disregards the role of time and development in mediation models.
Further, the sample was one of convenience; therefore, participants may differ from the larger population of college students on a number of variables, including ADHD symptomatology, externalizing symptomatology, executive function, substance use, and sexual risk behavior, which may limit the generalizability of the findings. Although the sample was geographically diverse, it was also relatively homogenous with regard to race and ethnicity, which also restricts the generalizability of the findings. Additionally, the present study used data from first-year students only and may underestimate the true prevalence of risk behavior among college students. Future studies should examine college students later in their academic careers. Furthermore, the present study examined college students at 4-year universities, and results may differ with students at 2-year colleges.
Future studies exploring the relationship between ADHD symptomatology, externalizing symptomatology, executive function, substance use, and sexual risk behavior are encouraged to employ a more representative sample of college students.
Ideally, studies would be longitudinal, and include greater focus on the interaction between variables. For example, previous research in conjunction with the present study highlights the important and complex interaction between ADHD, externalizing symptomatology, and its relationship with substance use. This interaction should be studied, provided the detrimental outcomes of substance abuse in college. Furthermore, the pathway between ADHD symptomatology and executive dysfunction should be studied with greater detail, given the increasing numbers of students with ADHD entering college. Such research would inform future prevention intervention programs greatly.

Conclusion
The current study was the first to systematically examine the complex relationship between ADHD symptomatology, externalizing symptomatology, EF dysfunction, substance use, and sexual risk behavior in a sample of college students with and without ADHD symptomatology. interventions should establish effective coping skills to help students manage ADHD symptoms, executive function deficits, and behavioral problems. In addition, universal substance use and sexual risk programs should include similar coping strategies, given the relationship between ADHD symptomatology and risk behavior.
The present findings have important implications for public health policy, particularly as it relates to the college population. Educating college students about the relationship betweem substance use and sexual risk behavior is clearly warranted. Given that many college students who take part in such risky behaviors also experience ADHD symptomatology, externalizing symptomatology, and EF dysfunction, it is crucial that college students be provided with academic and psychosocial supports to help manage their cognitive processes, feelings, and behaviors. Future universal risk prevention intervention programs should incorporate cognitive behavioral and dialectical behavioral coping strategies that have shown promise among college student populations. Notes. * = Correlation is significant at the 0.05 level. ** = Correlation is significant at the 0.01 level