Tobacco and Alcohol Use as an Explanation for the ... - CiteSeerX

3 downloads 0 Views 110KB Size Report
Amy Helstrom,1 Angela Bryan,1,2,4 Kent E. Hutchison,1 Paula D. Riggs,3 and Elaine A. ...... Tarter, R. E., Susman, E. J., Mezzich, A., & Clark, D. B. (2000).

C 2004) Prevention Science, Vol. 5, No. 4, December 2004 (

Tobacco and Alcohol Use as an Explanation for the Association Between Externalizing Behavior and Illicit Drug Use Among Delinquent Adolescents Amy Helstrom,1 Angela Bryan,1,2,4 Kent E. Hutchison,1 Paula D. Riggs,3 and Elaine A. Blechman1

The prevalence and persistence of adolescent substance use and abuse is a national health issue, and substance use among adolescents is frequently comorbid with other psychiatric disorders. Most studies in this area utilize samples of middle or high school students or from inpatient settings. Less is known about substance use and psychiatric comorbidity among delinquent adolescents. The present study examined data from two cohorts of juvenile offenders collected over a 2-year period (n = 245, n = 299). Participants reported frequency of cigarette, alcohol, marijuana, and other substance use. Participants’ parents completed a measure of behavior problems. Path analyses suggested that parental reports of externalizing problems were significantly related to self-reported substance use while parental reports of internalizing problems were not. Results also suggested that smoking and alcohol use act as mediators between externalizing problems and marijuana and other drug use. Although there were some mean differences by gender, the pattern of relationships amongst the variables did not differ by gender. Implications of the findings and future directions are discussed. KEY WORDS: juvenile offenders; externalizing behavior; internalizing problems; adolescent substance use.

High rates of adolescent substance use continue to be a significant national health concern. Adolescents who begin using substances by age 15 are at a higher risk for antisocial behavior (Van Kammen & Loeber, 1994) and substance abuse (Kandel & Davies, 1992; Robins & Przybeck, 1985). Adolescents who use substances are at higher risk for high-risk sexual behavior (Bryan & Stallings, 2002) and general maladaptation in young adulthood (Kandel et al., 1986). The prevalence of adolescent substance use is high, with approximately 74.8% of high school seniors report-

ing alcohol use in the past year, 34% reporting using cigarettes, and 42.4% reporting past illicit drug use (Johnston, 1996; Johnston et al., 1998). Results from the 1998 National Household Survey on Drug Abuse (NHSDA) indicated that 4.1 million 12–17-year-olds reported smoking at least one cigarette in the past month and almost 2 million reported past month marijuana use (Office of Applied Studies, 1999). Tobacco use is both particularly prevalent and associated with a host of health and behavior problems. Nearly two thirds of all adolescents smoke at least one cigarette before the age of 18, greatly increasing their risk for regular adult smoking (Chassin et al., 1996; U.S. Department of Health and Human Services [USDHHS], 1994), and 75% of daily adolescent smokers will smoke as adults (Johnston et al., 1992). Ellickson et al. (2001) found that seventh graders who had smoked three or more times in the past year or any time in the past month were more likely to regularly use tobacco, use hard drugs, sell drugs, have multiple


Department of Psychology, University of Colorado, Boulder, Colorado. 2 Institute of Behavioral Science, University of Colorado, Boulder, Colorado. 3 University of Colorado Health Sciences Center. 4 Correspondence should be directed to Angela Bryan, PhD, Department of Psychology, University of Colorado, Campus Box 345, Boulder, Colorado 80309-0345; e-mail: [email protected]

267 C 2004 Society for Prevention Research 1389-4986/04/1200-0267/1 

268 health problems, drop out of school, and experience early pregnancy and parenthood than nonsmokers. Many adolescents experiment with drugs and alcohol, but questions remain about which adolescents go on to develop regular use and how this progression occurs. Retrospective reports suggest that initial experimentation with substances occurs in the seventh and eighth grades and that frequency and quantity of use increases over time (Johnston et al., 1998; Newcomb, 1995). Several reviews have identified risk and protective factors for the escalation of adolescent substance use (e.g., Dawes et al., 2000; Swadi, 1999; Tarter et al., 1999; Weinberg et al., 1998). Many risk factors (e.g., economic deprivation, neighborhood disorganization, poor parenting) are comorbid with externalizing problems (e.g., conduct disorder, delinquency, antisocial behavior) and internalizing problems (e.g., depression, anxiety, withdrawal; Kazdin & Weisz, 1998; McConaughy & Skiba, 1993). Conduct disorder is a reliable predictor of substance use and problems (Bryan & Stallings, 2002; Dawes et al., 2000; Disney et al., 1999; Kandel et al., 1999), as well as rapid progression from experimentation to substance use disorders (Crowley & Riggs, 1995; Robins & McEvoy, 1990). Externalizing behavior problems appear to increase the risk for early substance use, and comorbidity with internalizing problems may add further risk (Milberger et al., 1997). Riggs and colleagues found that among conduct-disordered adolescent boys, those with comorbid major depressive disorder had more substance dependence diagnoses than boys without depression (Riggs et al., 1995). Whitmore et al. (1997) found that conduct-disordered adolescent girls with comorbid depression showed greater substance use severity, and Patton et al. (1996) found that adolescents with depressive and anxiety symptoms showed higher risk for smoking initiation than adolescents without such symptoms. Although previous studies have reported that externalizing behavior and internalizing problems are independently or additively associated with adolescent substance use, none of these investigations have examined the association between internalizing problems and substance use while controlling for externalizing behavior. In addition, it is unclear whether externalizing and internalizing problems are directly associated with drug use or whether the association follows a specific progressive pattern of use. Given the reliable association between externalizing and internalizing problems (e.g., McConaughy & Skiba, 1993), it is possible that internalizing problems may show a

Helstrom, Bryan, Hutchison, Riggs, and Blechman significant association with substance use due to their association with externalizing behavior. Several models that help to explain the connection between antisocial behavior and adolescent substance use have been developed. Problem behavior theory (e.g., Jessor & Jessor, 1977) suggests that drug and alcohol use and externalizing behavior problems are manifestations of a core, underlying construct of unconventional personality. This theory might logically predict bidirectional associations between externalizing behavior and other problem behaviors such as alcohol, tobacco, marijuana, and hard drug use (Turbin et al., 2000). In contrast, a developmental model of the progression of adolescents’ involvement in substance use was first proposed over 25 years ago (Kandel, 1975). Known as the gateway theory, the general model suggests that adolescents typically use tobacco or alcohol before progressing to illicit substances such as marijuana and hard drugs (Kandel & Davies, 1992; cf., Kandel et al., 1992). The vast majority of work on this developmental progression has been conducted with representative samples of young people, i.e., samples not selected to be particularly at risk for substance use or dependence. Studies have shown gender differences in the rates and severity of substance use, with higher rates of alcohol, marijuana, and hard drug use among males (e.g., Dembo et al., 1991). But the gender gap may be closing, and some studies show that nicotine use among females may exceed use among males (Boyle et al., 1992). There may also be gender variability in the pattern of substance use progression. For females, either cigarette or alcohol use appear to lead to illicit drug use. For males, alcohol use is a more important predictor of progression to illicit drug use (Kandel et al., 1992; Kandel & Davies, 1992). The relationship between substance use, externalizing problems, and internalizing problems has not, however, been examined in a high-risk sample with both genders. In this investigation, we propose to test a crosssectional mediational model of the role of alcohol and cigarette smoking in illicit drug use in a sample of high-risk adolescents. These individuals are expected to have high levels of externalizing and internalizing problems, and to be at extremely high risk for substance use problems. Although the model is crosssectional, we draw on the findings of developmental work on stage theories of drug use progression to hypothesize that the relationships between internalizing and externalizing problems and illicit drug use will be mediated by involvement in cigarette smoking


Adolescent Externalizing and Substance Use

Fig. 1. Hypothesized mediational model of the relationship between internalizing and externalizing problems and drug use.

and alcohol use, consistent with a developmental progression (see Fig. 1). Although this is not a test of the Gateway or stage theory (as that would require prospective data), a demonstration of the validity of this model would provide support for the longitudinal examination of the model in a legally referred population. A secondary goal is to examine the model in male versus female adolescents.

METHODS Participants Participants were juvenile offenders who were taking part in a diversionary program in Boulder County, Colorado. All participants had been arrested or given a ticket for a second misdemeanor or first nonviolent felony, such as vandalism, theft, and trespassing. Each was eligible for voluntary participation in the Juvenile Diversion Program, which offered an alternative to prosecution and provided treatment to participants as needed. The program typically consisted of payment of restitution, maintenance of school attendance and grades, community service, and individual or family therapy. Successful completion of the program allowed participants to avoid prosecution and have their criminal record expunged. Diversion Program participants and their parents were required to complete a questionnaire in order for Diversion Program staff to individually tailor each participant’s 6-month program according to his or her needs. At this point, participants and their parents

were also informed that researchers were interested in using the data from these questionnaires in their studies of adolescent problem behavior. It was made clear to participants that allowing their data to be included in the research was completely voluntary, there was no penalty for not participating in the research, and their decision would not affect their treatment in diversion in any way. For both research participants and nonparticipants, a brief report was written for use by the Diversion Program in determining an individualized program for each adolescent. Participants were 8–18 years old and comprised two 1-year cohorts of 245 participants (1996–1997) and 299 participants (1997–1998). In Cohort 1, the mean age was 14.88 (SD = 1.72), and in Cohort 2, the mean age was 15.00 (SD = 1.76). In both cohorts, approximately 88% were between 13 and 17 years old. A t-test indicated there were no significant differences in age between the two cohorts. Both cohorts were approximately 72% male and 28% female. Participants’ ethnic/racial distributions were as follows: Cohort 1—17.2% Hispanic/Latino, 76.7% Caucasian, 1.2% Native American, 2.9% African American, 0.8% Asian-American, and 1.2% multiracial; Cohort 2—14.1% Hispanic/Latino, 72.6% Caucasian, 2.3% Native American, 1.3% African American, 3.0% Asian-American, 5.7% multiracial, and 1% undisclosed. Chi-square tests indicated no significant differences in the gender or ethnic/racial distributions between the two cohorts. Participants’ parents reported family income with one item, “What is your family income?” Answer choices included (1) less than $12,000; (2) $12,000–24,999;

270 (3) $24,000–35,999; (4) $36,000–48,000; and (5) above $48,000. The median family income was $24,000– 35,999, and a χ 2 test revealed no significant cohort differences in family income.

Design and Procedures Participants and their parents were interviewed at the Boulder County Justice Center approximately 2 weeks following their arrest. Upon entry into the Juvenile Diversion Program, participants and their parents were notified that they would be completing a questionnaire in order to help the program determine each participant’s individual treatment. Adolescents and parents each met with a research assistant who explained the use of the questionnaire for assessment purposes in the Diversion Program, and explained the optional use of the questionnaire for research on adolescent problem behavior. Parents and children were informed that their status in the Diversion Program would not change regardless of their decision about whether or not to participate in the research. Participants received no compensation for their participation, and both parental consent and participant assent were required in order for the responses to be included in the research study. Of the 802 adolescents interviewed as Diversion Program participants, 544 consented to participate in the study. The majority of the families who did not consent to participation indicated that they were not interested in the study because no incentives were being offered. The questionnaire was administered to adolescents by research assistants who read each question aloud to participants to avoid problems with reading level. Parents completed questionnaires independently.

Measures Participants’ parents completed The Child Behavior Checklist (CBCL; Achenbach, 1991). Parents’ ratings on the CBCL are widely used, particularly among clinically referred populations of children and adolescents, and have shown medium to large stabilities over 3-year intervals (McConaughy et al., 1992). The CBCL includes items about a child’s relationship with others, his or her behavior at home and at school, delinquent acts, anxious and depressed feelings, and attention and thought problems. Parents are presented with 118 items to which they could answer, “not true of the child,” “somewhat or some-

Helstrom, Bryan, Hutchison, Riggs, and Blechman times true of the child,” or “very true or often true of the child.” Ratings are based on the 6 months period prior to the testing. Items from the quantitative portion of the CBCL are clustered into eight syndromes, including anxious/depressed, withdrawn, somatic complaints, social problems, thought problems, attention problems, delinquent behavior, and aggressive behavior. Scales are normed separately for gender and age ranges (ages 4–12 or ages 13–18). The test-retest reliability for the CBCL is .95 (Achenbach, 1991). For the present study we used the second-order scales, defined as the internalizing and externalizing problem scales. The internalizing problems scale is comprised of the anxious/depressed, withdrawn, and somatic complaints subscales. The externalizing behaviors subscale includes the delinquent and aggressive behavior subscales. Participants completed a subset of items from the American Drug and Alcohol Use Survey (ADAS; Oetting & Beauvais, 1990), which includes subscales concerning drug involvement, liking, peer use, perceived harm, and access to substances. Previous research has shown that parents’ reports about their child’s substance use tends to be consistently lower than their child’s report (Kandel et al., 1997), thus we used child report as the basis for our measure of substance use quantity and frequency. For the present study we selected items from the ADAS from the drug involvement subscales. These items included frequency and quantity of use of tobacco, alcohol, marijuana, and hard drugs. Tobacco use was measured by one item (i.e., What kind of smoker are you?) with answer choices on a scale that included non-smoker, once in a while, 1–2 times/day, 6–10 times/day, and almost all the time. Alcohol and marijuana use were each measured by one item, which asked how often the respondent had used alcohol and marijuana, respectively, in the last 12 months. Answer choices included 0,1–2 times, 3–9 times, 10–19 times, 20–49 times, and 50 or more times. Hard drug use was measured with nine items targeting experimentation with hard drug use. Participants were asked if they had ever tried amphetamines, cocaine, crack, sniffing something, LSD, other psychedelic drugs, heroin, quaaludes, and methamphetamines. Answer choices were 1 = yes and 0 = no. Because the frequency of endorsement of hard drug use experimentation was fairly low (29.8% in Cohort 1 and 33.8% in Cohort 2 reported that they had tried at least one of the above drugs one time or more), for data analysis purposes, we summed all nine hard drug use variables into a summary index of number of hard drugs tried.


Adolescent Externalizing and Substance Use Data Analysis In order to test our mediational model of the relationship between internalizing/externalizing and substance use, we utilized path analysis within the Mplus 2.12 structural equation modeling program (Muthen ´ & Muthen, ´ 2001). Mplus allows for the specification of endogenous variables as categorical, and thus we were able to retain the categorical nature of our substance use variables. After assessing the fit of our theoretical model in each cohort, we combined the data and conducted tests of mediation on the model in the combined sample. First, we included a direct path from the exogenous psychological predictor to the hard drug use outcome of interest to determine whether or not a significant relationship remained after controlling for the mediator. Second, we assessed the significance of the two parts of the indirect path. Third, we calculated the parameter estimates for the indirect effects, and used the Sobel (1982) method to calculate the standard error of the mediated effect in order to test it for significance. The presence of a

nonsignificant direct path in combination with a significant indirect path is evidence for at least a partial mediated effect. Finally, we examined gender differences using a multiple groups modeling approach in Mplus.

RESULTS The two cohorts were first examined for comparability on measures of internalizing, externalizing, and substance use. The only significant differences between the two groups were slightly higher levels of alcohol use and marijuana use in Cohort 2. Raw means and standard deviations for continuous variables and proportions for categorical variables appear in Table 1. Tetrachoric correlations provided by Mplus between all pairs of model variables are given in Table 2, with the correlations for Cohort 1 appearing below the diagonal, and those for Cohort 2 appearing above the diagonal. As expected, the correlations were high among the various types of substance use as

Table 1. Means/Proportions and Standard Deviations for Means Variable

Cohort 1

Cohort 2

Test for Difference

Externalizing behavior problems [M (SD)] Internalizing problems [M (SD)] Cigarette use Nonsmoker Once in a while 1–2 times/day 6–10 times/day Almost all the time Alcohol use in past 12 months None 1–2 times 3–9 times 10–19 times 20–49 times 50 or more times Marijuana use in past 12 months None 1–2 times 3–9 times 10–19 times 20–49 times 50 or more times Hard drug use Used no hard drugs Used 1 hard drug Used 2 hard drugs Used 3 hard drugs Used 4 hard drugs

17.50 (12.60)

16.82 (11.86)

t(536) < 1, ns

10.54 (8.58)

10.16 (7.84)

t (536) < 1, ns χ 2 (4, n = 541) = 5.16, ns

43% 23% 14% 10% 10%

46% 16% 15% 12% 11% χ 2 (5, n = 540) = 14.35, p < .05

18% 38% 21% 15% 5% 3%

23% 28% 20% 11% 13% 5%

38% 25% 12% 7% 6% 12%

41% 16% 12% 4% 8% 18%

72% 13% 10% 4% 1%

69% 12% 12% 6% 1%

χ 2 (5, n = 538) = 11.22, p < .05

χ 2 (4, n = 540) = 2.45, ns


Helstrom, Bryan, Hutchison, Riggs, and Blechman Table 2. Correlations for Cohort 1 (Below Diagonal, n = 233) and Cohort 2 (Above Diagonal, n = 290)

1. Internalizing 2. Externalizing 3. Smoking 4. Alcohol 5. Marijuana 6. Hard drugs 7. Age




— .68∗∗∗ .19∗∗ .04 .14∗ .11 −.14∗

.63∗∗∗ — .31∗∗∗ .09 .24∗∗∗ .11 −.28∗∗∗

.29∗∗∗ .33∗∗∗ — .43∗∗∗ .48∗∗∗ .40∗∗∗ −.01

4 .04 .08 .30∗∗∗ — .63∗∗∗ .49∗∗∗ .26∗∗∗

5 .05 .16∗∗ .35∗∗∗ .68∗∗∗ — .74∗∗∗ .18∗

6 .15∗ .24∗∗ .44∗∗∗ .55∗∗∗ .67∗∗∗ — .30∗∗∗

7 −.12∗ −.18∗∗ .12 .39∗∗∗ .20∗∗ .27∗∗∗ —

Note. Correlations between continuous variables (e.g., age and internalizing behavior) are Pearson product moment correlations, correlations between one continuous variable and one ordered categorical variable (e.g., age and alcohol use) are polyserial correlations, and correlations between ordered categorical variables (e.g., alcohol use and marijuana use) are polychoric correlations. ∗ p < .05. ∗∗ p < .01. ∗∗∗ p < .001.

well as between externalizing and internalizing problems. Age was negatively related to both internalizing and externalizing problems. In addition, age was positively related to many of the substance use variables. In order to control for the effects of age, it was included in the model as an additional continuous exogenous predictor (see Fig. 2).

Path Analysis The model was estimated in Mplus, specifying that all four substance use variables were categorical in nature, while internalizing, externalizing, and

age were continuous in nature. The recommended estimation procedure in models with categorical endogenous variables in Mplus is WLSMV (refers to estimating the weighted least square parameter estimates using a diagonal weight matrix with robust standard errors and mean- and variance-adjusted χ 2 test statistic; Muthen ´ & Muthen, ´ 2001). Model fit was thus evaluated with the mean- and variance-adjusted χ 2 provided by WLSMV estimation, the Comparative Fit Index (CFI), and the Root Mean Square Error of Approximation (RMSEA). The CFI is a measure that compares a baseline model in which no relationships are estimated between the variables to a theoretical model in which hypothesized paths are estimated.

Fig. 2. Mediational model including age as an exogenous predictor of substance use as estimated in both cohorts. Coefficients are standardized path coefficients. Coefficients for Cohort 1 appear first; overall model fit in Cohort 1: χ 2 (2, n = 234) = 4.79, p = .11, CFI = .99, RMSEA = .08. Coefficients for Cohort 2 appear second (after the “/”); overall model fit in Cohort 2: χ 2 (3, n = 290) = 5.61, p = .13, CFI = .99, RMSEA = .06. In both models, the correlations between alcohol use and cigarette use, and between marijuana use and hard drug use were estimated. For simplicity of presentation, these correlations are not shown.


Adolescent Externalizing and Substance Use This index ranges from 0 to 1, where .9 indicates adequate fit, and .8 is considered marginal fit (Bentler, 1990). The RMSEA ranges from 0 to ∞, with fit values less than .05 indicating close fit, and values less than .10 indicating reasonable fit (Browne & Cudeck, 1993). The RMSEA is sensitive to overfit, that is, it begins to increase when too many paths have been included (Rigdon, 1996). These cutoffs apply to models with continuous outcomes, however Yu and Muthen ´ (2001) report that they are reasonable for models with categorical outcomes as well. For both models, listwise deletion was employed, resulting in a sample size for path analyses of 233 in Cohort 1 and 290 in Cohort 2.

relationships emerged, with all hypothesized effects being significant except those between internalizing and both cigarette and alcohol use. The pattern of relationships with age and the substance use variables was slightly different. In Cohort 2, as in Cohort 1, age was significantly and positively related to alcohol use but was unrelated to marijuana use. Unlike Cohort 1, there was a significant and positive relationship between age and cigarette use and no relationship between age and hard drug use. This model accounted for 49% of the variance in marijuana use and 42% of the variance in hard drug use. Standardized path coefficients as well as significance levels for all paths appear in Fig. 2, with values for Cohort 2 appearing second (after the “/”).

Test of the Model in Cohort 1 The theoretical model in Fig. 1 was first estimated in Cohort 1. In this model, we hypothesized that smoking and alcohol use would mediate the relationship between externalizing behavior and marijuana and hard drug use. The model was a good fit to the data, χ 2 (2, n = 233) = 4.79, p = .11, CFI = .99, RMSEA = .08. As expected on the basis of the first-order correlations, the relationships between internalizing and both smoking and alcohol use were nonsignificant. The relationships between externalizing and both smoking and alcohol use were significant, and in turn the relationships between smoking and alcohol use and both marijuana use and hard drug use were significant. The significance of the two segments of the indirect paths from externalizing to hard drug use and marijuana use is suggestive of mediation. Age was positively and significantly related to alcohol use and hard drug use, but showed no direct relationship to marijuana use or cigarette use. This model accounted for 47% of the variance in marijuana use and 33% of the variance in hard drug use. All standardized path coefficients and significance levels are presented in Fig. 2, with values for Cohort 1 appearing first (before the “/”).

Replication in Cohort 2 Analyses conducted with Cohort 2 were identical to those used in Cohort 1. The theoretical model again yielded acceptable fit, χ 2 (3, n = 290) = 5.61, p = .13, CFI = .99, RMSEA = .06.5 The same pattern of 5

The estimation procedure used by Mplus in a model that contains continuous and categorical variables is robust weighted least

Tests of Mediation In both models, the significant paths from externalizing to both cigarette use and alcohol use and from these variables to marijuana and hard drug use are suggestive of mediation of the effects of externalizing on marijuana and hard drug use through cigarette and alcohol use. Given the identical pattern of findings with regard to our theoretical variables (i.e., externalizing, internalizing, and substance use) in the two cohorts, we combined the data sets in order to obtain more reliable assessments of mediation. The larger sample size reduces the possibility of Type II error in assessing the direct effect (i.e., finding a nonsignificant direct effect supporting mediation when a direct effect does indeed exist). The first test involved adding the direct path from externalizing to hard drug use and externalizing to marijuana use and assessing the significance of these paths; a nonsignificant direct path implies at least partial mediation. The second test involved assessing the significance of the indirect paths from externalizing to marijuana use and from externalizing to hard drug use with the Sobel (1982) test. When a direct path was added from externalizing to marijuana use, this path was not significant (B = .07, ns). Calculation of the total indirect effect (i.e.,

squares (WLSMV). When this procedure is utilized the degrees of freedom for the model are actually estimated as the closest integer ´ & to d*, where d* = (tr(U))2 /tr((U)2 ) (formula 110; Muthen Muthen, ´ 2001, p. 358). The model is estimated identically in the two cohorts, but because of sample fluctuations in the covariance matrix used in the computation of the WLSMV chi-square, the estimated d* differs in the two groups, and thus the degrees of freedom for the chi-square test also differ.

274 through both cigarette use and alcohol use) of externalizing on marijuana use yielded an unstandardized parameter estimate of 0.058 with a standard error of 0.0068. This was a statistically significant indirect effect, z = 8.49, p < .001. Equivalently, the direct path from externalizing to hard drug use was not significant. Calculation of the total indirect effect of externalizing on hard drug use yielded an unstandardized parameter estimate of 0.019992 with a standard error of 0.0060219, a statistically significant indirect effect, z = 3.32, p < .05. These results are consistent with a model in which the effects of externalizing on hard drug use and marijuana use are at least partially mediated through cigarette use and alcohol use. Gender Differences Our final analysis involved the comparison of the model among girls versus boys. As reviewed in the introduction, in tests of the progression of substance use among average young people (i.e., those not involved in the criminal justice system), there were some gender differences in the strength of the pathways. We sought to test this assertion in the current study using the collapsed data set combining Cohorts 1 and 2 from the earlier analyses. We first examined the differences between levels of substance use, internalizing problems, and externalizing behavior in boys versus girls. Examining the substance use data via χ 2 -tests, we observed no differences in alcohol or hard drug use by gender (ps > .17). We did find, however, that girls were more likely to smoke cigarettes (p < .001) and more likely to use marijuana (p < .01), consistent with the findings of Boyle et al. (1992). In addition, girls had higher levels of both internalizing (p < .001) and externalizing (p < .01) problems, which is consistent with literature indicating that girls presenting in clinical or juvenile justice settings are likely to be more severely disturbed than boys, i.e., there may be a higher threshold of problem behavior necessary for girls to enter the system (e.g., Duclos et al., 1998). While these mean differences are of interest, our primary goal in this analysis was to test whether the relationships among the variables showed a different pattern across groups. We estimated a multiple-groups model in Mplus, constraining all structural paths in the model to be identical in the two groups. The fit of the model was adequate, χ 2 (12, n = 523) = 35.109, p = .0004, CFI = .97, RMSEA = .09,6 suggesting no dif6

See footnote 1.

Helstrom, Bryan, Hutchison, Riggs, and Blechman ferences in the pattern of relationships for girls versus boys.

DISCUSSION Researchers have long posed the question of whether a particular behavioral profile leads adolescents to engage in all types of substance use or whether there may be indirect pathways that flow from risk factors, through less severe forms of substance use, to illicit drug use. In this study we tested a cross-sectional model consistent with the idea that the influence of a particular risk factor (externalizing behavior problems) on illicit substance use could be conceptualized as an indirect one, with smoking and alcohol serving as the mediators. In line with our hypothesis, the data indicate that the association between externalizing problems and hard drug use was significantly mediated by alcohol use and smoking. This study is particularly informative as the sample was a group of young people who are at extremely high risk for substance use and abuse (i.e., juvenile offenders). Our results emphasize the need to focus on methods to intervene with alcohol use and smoking among adolescents with externalizing problems in order to prevent an escalation to illicit drug use. Moreover, since externalizing problems generally precede even experimentation with cigarettes and alcohol, early identification of these high-risk youth may diminish risk of substance involvement (e.g., Forgatch & Patterson, 1998). Internalizing problems demonstrated a significant, albeit weak, bivariate relationship only with smoking. In addition, internalizing behavior was generally not associated with substance use after controlling for its association with externalizing problems. Consistent with other reports in the literature (e.g., Young et al., 1995), these results suggest that externalizing behavior is the more important risk factor for substance use among juvenile offenders. Our finding that internalizing problems are not strongly associated with substance severity is inconsistent with some previous research. One possible explanation for the discrepancy between the current investigation and other studies is related to methodological differences. For example, the study by Riggs et al. (1995) used DSM-III-R clinician-evaluated diagnostic criteria for depression and conduct disorder derived from a structured diagnostic interview. In the current study the measure of internalizing behavior was parent-reported symptoms. It is possible that there is a qualitative difference in the relationship


Adolescent Externalizing and Substance Use between substance use and depression between adolescents who meet criteria for major depressive disorder and adolescents who have been rated on internalizing symptoms on a checklist by their parents. Previous research suggests that cigarette and alcohol use appear to play a significant role in the initiation and maintenance of marijuana and other drug use (e.g., Ellickson et al., 1992; Kandel & Yamaguchi, 2002). Our results lend support to this assertion, as both alcohol use and smoking emerged as significant mediators of the relationship between externalizing behavior and marijuana and hard drug use. We would not argue that this is the only form the progression can take, and certainly other models should be tested. For example, Young et al. (1995) found that among boys referred for substance use treatment, marijuana was the first substance for 42% of the sample. Our data are nevertheless consistent with recent findings by Merrill et al. (1999), who showed that among high school seniors cigarette and alcohol use were associated with the likelihood of marijuana use, and marijuana use was associated with the likelihood of other drug use over and above other problem behaviors. Despite their empirical support, stage theories of substance abuse need not be mutually exclusive with problem behavior theory (Donovan & Jessor, 1985). Our finding that adolescents higher in externalizing behaviors are more likely to use alcohol and cigarettes suggests that a general proclivity for deviance may account for an adolescent’s entree ´ into substance use, and thus facilitate the progression into hard drug use. A general proclivity for deviance may be associated with an underlying disposition for risk taking or impulsivity that leads to greater experimentation with drug use. Several studies have suggested that females tend to have higher rates of internalizing problems and exhibit a slightly different pattern of drug use escalation. Although our findings concur that females have higher rates of internalizing and externalizing problems, the multiple group model showed that the pattern of relationships is very similar across gender. This is consistent with prior work with noncriminally involved samples. Farrell et al. (2000) found that their model in sixth and seventh graders including physical and nonphysical aggression, drug use, and delinquent behaviors did not differ by gender. A study of nonreferred 17-year-olds by Disney et al. (1999) showed that the risk of substance use and abuse did not differ by gender, and was best predicted by the presence of conduct disorder. Our findings, in conjunction with those of others, suggest that a determination of the

influence of gender on substance use initiation and progression cannot yet be made, and theorizing about and empirical study of this question are warranted.

Limitations and Conclusions Our investigation was limited by the use of a cross-sectional design, and thus we are unable to make unequivocal statements about causation or temporal precedence. In addition, our sample represents only a portion of the population of potential participants. Approximately 68% of the adolescents we recruited consented to participate. It is possible that this group who consented may be qualitatively different from the group who did not. The choice of measures may also limit our findings. The CBCL, for example, is a measure based only on parent report. Despite its limitations, the present study adds to the body of knowledge about adolescent substance use and its relationship to internalizing and externalizing behavior problems. While the present study does not allow conclusions about the direction of the progression, the pattern of findings is consistent with the idea that alcohol and cigarette use play a significant role in the relationship between delinquency, aggression, and marijuana and other drug use, and that for a population with high levels of externalizing behavior, internalizing problems may play less of a role in substance use. Bearing the limitation of the crosssectional design in mind, these results suggest at least preliminary support for our hypothesized model in a legally referred population, and serve as a basis for future longitudinal work with this high-risk population. First, longitudinal studies with juvenile offenders are imperative, and would add an important component to the rich body of literature about adolescent substance use (e.g., Chassin et al., 1992, 1996, 2000). Second, it will be necessary for future studies to include or focus on female adolescents. The majority of research on adolescent externalizing behavior and substance use has been conducted with males. As females begin to represent an increasing proportion of the juvenile offender population, research will be needed in order to understand etiological models for females and how they may differ from males. Finally, additional research is needed in the area of treatment for adolescent substance use. Although such studies with high-risk youth are extremely difficult to conduct due to often overwhelming problems (e.g., difficulty of follow-up, participants failing to present for


Helstrom, Bryan, Hutchison, Riggs, and Blechman

treatment), it is particularly important to direct intervention efforts at juvenile offenders and other highrisk youth. A goal of basic research on the etiology of problem behaviors is, of course, the development of effective prevention strategies. Many tobacco, alcohol, and illicit drug use prevention efforts are school-based and may be unable to reach the most at-risk adolescents who do not attend school regularly or who drop out (Flay, 1993). Young people involved in the criminal justice system are particularly vulnerable to school dropout, and thus are likely to be missed by schoolbased prevention efforts. It may be that prevention efforts for this high-risk group need to be directed toward nontraditional settings. Programs such as juvenile diversion and juvenile detention allow for the possibility of targeting these young people. The present study suggests that the effects of externalizing behavior on hard drug use are significantly mediated through alcohol and cigarette use. Although longitudinal data are required, our data support a theoretical conceptualization in which alcohol and cigarette use may lead to hard drug use. Within this model, several gender differences were found, including the finding that girls are more likely to smoke cigarettes and marijuana. These results add a piece of the puzzle to the broad question about the nature of adolescent behavior problems and substance use which will lead to answers about how to prevent and treat adolescent substance use.

ACKNOWLEDGMENTS This research was supported by grants from NIDA (Blechman, K01DA000316-04 and Helstrom, R03DA013182-02).

REFERENCES Achenbach, T. M. (1991). Manual for the Child Behavior Checklist/4–18 and 1991 Profile. Burlington: University of Vermont, Department of Psychiatry. Bentler, P. M. (1990). Comparative fit indices in structural models. Psychological Bulletin, 107, 238–246. Boyle, M. H., Offord, D. R., Racine, Y. A., Szatmari, P., Fleming, J. E., & Links, P. S. (1992). Predicting substance use in late adolescence: Results from the Ontario Child Health Study follow-up. American Journal of Psychiatry, 149, 761–767. Browne, M. W., & Cudeck, R. (1993). Alternative ways of assessing model fit. In K. Bollen & K. Long (Eds.), Testing structural equation models (pp. 136–162). Newbury Park, CA: Sage. Bryan, A., & Stallings, M. C. (2002). A case control study of adolescent risky sexual behavior and its relationship to personality

dimensions, conduct disorder, and substance use. Journal of Youth and Adolescence, 31, 387–396. Chassin, L., Presson, C. C., Sherman, S. J., & Edwards, D. A. (1992). The natural history of cigarette smoking and young adult social roles. Journal of Health and Social Behavior, 33, 328–347. Chassin, L., Presson, C. C., Pitts, S. C., & Sherman, S. J. (2000). The natural history of cigarette smoking from adolescence to adulthood in a midwestern community sample: Multiple trajectories and their psychosocial correlates. Health Psychology, 19, 223–231. Chassin, L., Presson, C. C., Sherman, S. J., & Edwards, D. A. (1996). The natural history of cigarette smoking from adolescence to adulthood: Demographic predictors of continuity and change. Health Psychology, 15, 478–484. Crowley, T. J., & Riggs, P. D. (1995). Adolescent substance use disorder with conduct disorder and comorbid conditions. NIDA Research Monograph, 156, 49–111. Dawes, M. A., Antelman, S. M., Vanykov, M. M., Giancola, P., Tarter, R. E., Susman, E. J., Mezzich, A., & Clark, D. B. (2000). Developmental sources of variation in liability to adolescent substance use disorders. Drug and Alcohol Dependence, 61, 3–14. Dembo, R., Williams, L., Getreu, A., Genung, L., Schmeidler, J., Berry, E., Wish, E. D., & LaVoie, L. (1991). A longitudinal study of the relationship among marijuana/hashish use, cocaine use, and delinquency in a cohort of high-risk youths. Journal of Drug Issues, 21, 271–312. Disney, E. R., Elkins, I. J., McGue, M., & Iacono, W. G. (1999). Effects of ADHD, conduct disorder, and gender on substance use and abuse in adolescence. American Journal of Psychiatry, 156, 1515–1521. Donovan, J. E., & Jessor, R. (1985). Structure of problem behavior in adolescence and young adulthood. Journal of Consulting and Clinical Psychology, 53, 890–904. Duclos, C. W., Beals, J., Novins, D. K., Martin, C., Jewett, C., & Manson, S. M. (1998). Prevalence of common psychiatric disorders among American Indian adolescent detainees. Journal of the American Academy of Child and Adolescent Psychiatry, 37, 866–873. Ellickson, P. L., Hays, R. D., & Bell, R. M. (1992). Stepping through the drug-use sequence: Longitudinal scalogram analysis of initiation and regular use. Journal of Abnormal Psychology, 101, 441–451. Ellickson, P. L., Tucker, J. S., & Klein, D. J. (2001). High-risk behaviors associated with early smoking: Results from a 5-year follow-up. Journal of Adolescent Health, 28, 465–473. Farrell, A. D., Kung, E. M., White, K. S., & Valois, F. R. (2000). The structure of self-reported aggression, drug use, and delinquent behaviors during early adolescence. Journal of Clinical Child Psychology, 29, 282–292. Flay, B. R. (1993). Youth tobacco use: Risks, patterns and control. In C. T. Orleans & J. D. Slade (Eds.), Nicotine addiction: Principles and Management. New York: Oxford University Press. Forgatch, M., & Patterson, G. (1998). Behavioral family therapy. In F. M. Dattilio (Ed.), Case studies in couple and family therapy: Systemic and cognitive perspectives (pp. 85–107). New York: Guilford Press. Jessor, R., & Jessor, S. L. (1977). Problem behavior and psychosocial development: A longitudinal study of youth. New York: Academic Press. Johnston, L. D. (1996). The rise in drug use among American teens continues in 1996. Press release from the Monitoring the Future Study, December 19. Ann Arbor: University of Michigan. Johnston, L. D., O’Malley, P. M., & Bachman, J. G. (1992). Smoking, drinking, and illicit drug use among American secondary school college students and young adults, 1975–1991: Volume I. Secondary school students. Washington, DC: U.S. Government Printing Office. (DHHS publication number 933840.)

Adolescent Externalizing and Substance Use Johnston, L. D., O’Malley, P. M., & Bachman, J. G. (1998). National survey results on drug use from the Monitoring the Future Study, 1975–1997 (National Institute on Drug Abuse NIH Publication No. 98-4345). Washington, DC: US Government Printing Office. Kandel, D. (1975). Stages in adolescent involvement in drug use. Science, 190, 912–914. Kandel, D. B., & Davies, M. (1992). Progression to regular marijuana involvement: Phenomenology and risk factors for neardaily use. In M. Glantz & R. Pickens (Eds.), Vulnerability to drug abuse (pp. 211–253). Washington, DC: American Psychological Association. Kandel, D. B., Davies, M., Karus, D., & Yamaguchi, K. (1986). The consequences in young adulthood of adolescent drug involvement: An overview. Archives of General Psychiatry, 43, 746– 754. Kandel, D. B., Johnson, J. G., & Bird, H. R. (1999). Psychiatric comorbidity among adolescents with substance use disorders: Findings from the MECA study. Journal of the American Academy of Child and Adolescent Psychiatry, 38, 693–699. Kandel, D. B., Johnson, J. G., Bird, H. R., Canino, G., Goodman, S. H., Lahey, B. B., Regier, D. A., & Schwab-Stone, M. (1997). Psychiatric disorders associated with substance use among children and adolescents: Findings from the methods for the epidemiology of child and adolescent mental disorders (MECA) study. Journal of Abnormal Child Psychology, 25, 121–132. Kandel, D. B., & Yamaguchi, K. (2002). Stages of drug involvement in the U.S. population. In D. B. Kandel & K. Yamaguchi (Eds.), Stages and pathways of drug involvement: Examining the gateway hypothesis (pp. 65–89). New York: Cambridge University Press. Kandel, D. B., Yamaguchi, K., & Chen, K. (1992). Stages of progression in drug involvement from adolescence to adulthood: Further evidence for the gateway theory. Journal of Studies on Alcohol, 53, 447–457. Kazdin, A. E., & Weisz, J. R. (1998). Identifying and developing empirically supported child and adolescent treatments. Journal of Consulting and Clinical Psychology, 66, 19–36. McConaughy, S. H., & Skiba, R. J. (1993). Comorbidity of externalizing and internalizing problems. School Psychology Review, 22, 421–436. McConaughy, S. J., Stanger, C., & Achenbach, T. M. (1992). Threeyear course of behavioral/emotional problems in a national sample of 4- to 16-year-olds: I. Agreement among informants. Journal of the American Academy of Child and Adolescent Psychiatry, 31, 932–939. Merrill, J. C., Kleber, H. D., Shwartz, M., Liu, H., & Lewis, S. R. (1999). Cigarettes, alcohol, marijuana, other risk behaviors, and American youth. Drug and Alcohol Dependence, 56, 205– 212. Milberger, S., Biederman, J., Faraone, S. V., Chen, L., & Jones, J. (1997). Further evidence of an association between attentiondeficit/hyperactivity disorder and cigarette smoking: Findings from a high-risk sample of siblings. American Journal on Addictions, 6, 205–217. Muthen, ´ L. K., & Muthen, ´ B. O. (2001). Mplus User’s Guide (2nd ed.). Los Angeles: Muthen ´ & Muthen. ´ Newcomb, M. D. (1995). Identifying high-risk youth: Prevalence and patterns of adolescent drug abuse. In E. Rahdert & D. Czechowicz (Eds.), Adolescent drug abuse: Clinical assessment and therapeutic interventions (NIDA Research Monograph 156, pp.7–38), Rockville, MD: U.S. Department of Health and Human Services.

277 Oetting, E. R., & Beauvais, F. (1990). Adolescent drug use: Findings of national and local surveys. Journal of Consulting and Clinical Psychology, 58, 385–394. Office of Applied Studies. (1999). National Household Survey on Drug Abuse: Main Findings 1997 (DHHS Pub. No. (SMA) 99-3295). Substance Abuse and Mental Health Services Administration, U.S. Department of Health and Human Services. Patton, G. C., Hibbert, M., Rosier, M. J., Carlin, J. B., & Bowes, G. (1996). Is smoking associated with depression and anxiety in teenagers? American Journal of Public Health, 86, 225– 230. Rigdon, E. E. (1996). CFI versus RMSEA: A comparison of two fit indexes for structural equation modeling. Structural Equation Modeling, 3, 369–379. Riggs, P. D., Baker, S., Mikulich, S. K., Young, S. E., & Crowley, T. J. (1995). Depression in substance-dependent delinquents. Journal of the American Academy of Child and Adolescent Psychiatry, 34, 764–771. Robins, L. M., & McEvoy, L. (1990). Sturdy childhood predictors of adult antisocial behaviour: Replications from longitudinal studies. Psychological Medicine, 6, 611–622. Robins, L. N., & Przybeck, T. R. (1985). Age of onset of drug use as a factor in drug and other disorders. National Institute on Drug Abuse: Research Monograph Series, 56, 178–192. Sobel, M. E. (1982). Asymptotic confidence intervals for indirect effects in structural equation models. In S. Leinhardt (Ed.), Sociological Methodology 1982 (pp. 290–312). San Francisco: Jossey-Bass. Swadi, H. (1999). Individual risk factors for adolescent substance use. Drug and Alcohol Dependence, 55, 209–224. Tarter, R. E., Vanyukov, M., Giancola, P., Dawes, M., Blackson, T., & Mezzich, A. (1999). Etiology of early age onset substance use disorder: A maturational perspective. Development and Psychopathology, 11, 657–683. Turbin, M. S., Jessor, R., & Costa, F. M. (2000). Adolescent cigarette smoking: Health-related behavior or normative transgression? Prevention Science, 1, 115–124. U.S. Department of Health and Human Services. (1994). Preventing tobacco use among young people: A report of the Surgeon General. Atlanta, GA: US Department of Health and Human Services, Public Health Service, Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion, Office on Smoking and Health. Van Kammen, W. B., & Loeber, R. (1994). Are fluctuations in delinquent activities related to the onset and offset in juvenile illegal drug use and drug dealing? Journal of Drug Issues, 24, 9–24. Weinberg, N. Z., Rahdert, E., Colliver, J. D., & Glantz, M. D. (1998). Adolescent substance abuse: A review of the past 10 years. Journal of the American Academy of Child and Adolescent Psychiatry, 37, 252–261. Whitmore, E. A., Mikulich, S. K., Thompson, L. L., Riggs, P. D., Aarons, G. A., & Crowley, T. J. (1997). Influences on adolescent substance dependence: Conduct disorder, depression, attention deficit hyperactivity disorder, and gender. Drug and Alcohol Dependence, 47, 87–97. Young, S. E., Mikulich, S. K., Goodwin, M. B., Hardy, J., Martin, C. L., Zoccolillo, M. S., & Crowley, T. J. (1995). Treated delinquent boys’ substance use: Onset, pattern, relationship to conduct and mood disorders. Drug and Alcohol Dependence, 37, 149–162. Yu, C.-Y., & Muthen, ´ B. (2001). Evaluation of model fit indices for latent variable models with categorical and continuous outcomes. Technical report in preparation.

Suggest Documents