|
Article Excerpt
This 4-year study of 698 young adolescents examined the covariates of early onset substance use from Grade 6 through Grade 9. The youth were randomly assigned to a family-centered Adolescent Transitions Program (ATP) condition. Variable-centered (zero-inflated Poisson growth model) and person-centered (latent growth mixture model) approaches were taken to examine treatment effects on patterns of substance-use development across early adolescence. Variable-centered analyses revealed treatment effects both on decreasing the likelihood of initiating substance use and on the rate of growth in substance use among those who initiated use. Person-centered analyses revealed the following five trajectories of early substance use: (1) no use, (2) low/rare use, (3) early accelerating use, (4) late-accelerating use, and (5) early high but decreasing use. Of note, random assignment to the ATP intervention was strongly predictive of following the decreasing-use trajectory. In addition, the early high but decreasing group was most likely to engage in the Family Check-Up and linked intervention services. These findings suggest that covariates of early adolescent substance us, as well as the effectiveness of prevention strategies, vary as a function of the developmental pattern underlying early adolescent risk.
**********
Despite popular beliefs, the majority of young adolescents develop socially and emotionally without significant drug or alcohol involvement, deviant peer clustering, or serious mental health problems. Only a relatively small group of young adolescents can be labeled "high risk" (Bry, 1982). Statistically speaking, about 5% to 6% of adolescents within an age cohort commit 50% of the delinquent acts (Farrington, Barnes, & Lambert, 1996). The same percentage can be thought of as multiple-problem youth, who tend to initiate drug use early, engage in delinquent behavior, and experience high levels of emotional distress (Biglan et al., 2004). Reams of journal articles delineate the multiple risk factors of these youngsters, their families, and, more recently, their genetic pedigrees.
The approach typically taken to understand risk for substance use is to develop a nomothetic model that accounts for variation in substance use. Such a variable-centered approach has its advantages with respect to providing an empirical model that can then guide intervention and policy. The effort to systematically construct developmental models that guide interventions has been referred to as model building; it was originally described by Patterson, Reid, and Dishion (1992) and was later elaborated by Dishion and Patterson (1999). The variable-centered approach to model building was inspired by innovations in statistical methods that allowed the joint consideration of factor loadings and causal structure in a framework of structural equation modeling (Bentler, 1980; Dwyer, 1983; Joreskog & Sorbom, 1983). Model building has provided an important framework for guiding the development of effective intervention and prevention strategies (Dishion, Kavanagh, Schneiger, Nelson, & Kaufman, 2002; Dishion, Nelson, & Kavanagh, 2003; Forgatch, 1991; Forgatch, Patterson, & DeGarmo, 2005).
The model that has guided our understanding of early adolescent substance use has been relatively straightforward. In the 1970s there was a cohort burst in the use of psychoactive substances among U.S. youth. Along with this increase was a corresponding increase in research defining promising variables that predicted adolescent substance use. For instance, influential research by Kandel (1973) demonstrated that when one compares the influence of parent and peer substance use on adolescent substance use, peer use is considerably more important. Work by Jessor and colleagues (Jessor, 1976; Jessor & Jessor, 1975) revealed that Kandel's conclusion stood up when considering alcohol and marijuana use as distinct drug-using behaviors, leading Jessor to conclude that substance use and delinquency were really part of a single underlying dimension referred to as problem behavior. However, in our own analysis we found that the combination of delinquency and drug use was indicative of heightened risk of parental monitoring and deviant peer affiliation (Dishion & Loeber, 1985). In fact, youth who used substances but were not delinquent were statistically similar to abstinent adolescents on parenting practices and deviant peer affiliation. The analysis of adolescent substance use quickly became more sophisticated, with Chassin and colleagues (1986) examining longitudinal variation in parent and peer effects. The conclusion held: peer effects were dominant.
Friedman, Lichtenstein, and Biglan (1985), however, considered the situations that provided contexts for the initial smoking episodes of young adolescents and found that 80% of these initial episodes involved other adolescents and were in homes without parental supervision after school. This finding agreed quite well with Steinberg's ecological analysis (1986). Although parental substance use did not appear to be a major direct effect on adolescent substance use, it was clear that parenting was an important factor. If most of the initial substance-use episodes occurred in unsupervised situations with other teenagers, then parental monitoring was clearly an issue (Dishion & Loeber, 1985). In fact, in developing models of adolescent substance use we found that the net effect of parental monitoring was quite large if one accounts for both its direct effect on substance use and its indirect effect on deviant peer affiliation (Dishion, Reid, & Patterson, 1988). We examined the issue more carefully using recursive longitudinal data and found that a construct measuring deviant peer affiliation and parental monitoring correlated at about .75 within a sample of 204 mostly European American males. Moreover, both parental monitoring and deviant peer affiliation were highly predictive of adolescent substance use by age 15.
More recently, we compared the developmental trajectories leading to substance use in males with a history of antisocial behavior with those with no history. To avoid some of the definitional issues endemic to the parental monitoring construct (e.g., Dishion & McMahon, 1999; Kerr & Stattin, 2000), we used a direct observation measure of parent management (relationship quality, monitoring, and behavior management) that was derived from videotaped interactions of the boys with their parents at ages 10, 12, 14, and 16. For boys at age 10, the parents of antisocial boys were very similar to the parents of typically developing boys in terms of parent management. However, at age 12 and beyond there was marked differentiation between the two groups, with parents of the antisocial boys being observed as less effective in parenting than the parents of the typically developing boys. At the same time, direct observations of the boys' friendships revealed increasing levels of deviant talk for the antisocial boys compared to the level of deviant talk for the boys who were typically developing. The interaction term for the slope in parenting and deviant peer affiliation was, in fact, highly significant. Boys who were affiliated with deviant peers and experienced a deterioration in parenting were most at risk for serious drug use by young adulthood. We referred to this process as premature adolescent autonomy (Dishion, Nelson, & Bullock, 2004).
Thus, several researchers who have measured both constructs concur that a variable-centered model that poses both parental monitoring and deviant peer affiliation accounts for young adolescent problem behavior, and the model holds across cultures (Barrera, Castro, & Biglan, 1999). Of particular relevance is the viability of measures of family management in predicting adolescent substance use across ethnic groups. Chilcoat, Dishion, and Anthony (1995) found the same pattern of findings in a sample of African American urban youth. In addition, family management predicted adolescent substance use for both European American and African American youth (Catalano et al., 1992). By its own criteria, therefore, the model-building approach has worked in the effort to develop an empirical account of early onset adolescent substance use that is viable across cultural and gender groups. It is also noteworthy that intervention strategies that target parenting practices in general (Spoth, Redmond, & Lepper, 1999; Spoth et al., 1998) and monitoring practices in particular (Dishion et al., 2002, 2003) are effective in reducing and preventing adolescent substance use.
Although the general associations among family dynamics, peer processes, and adolescent substance use have been well documented from a variable-centered perspective, exploration of subgroups of adolescents with unique developmental trajectories is needed. We suspect that such a person-centered perspective is relevant to both an empirical understanding of adolescent substance use and to the design of effective interventions, as it considers the variability of causal dynamics as a function of the youth's developmental pathway. As several studies indicate, for example, we know that early adolescent substance use does not emerge out of thin air. Rather, early forms of antisocial behavior precede the emergence of substance use in early adolescence for at least a subset of youth (Dishion, Capaldi, Spracken, & Li, 1995; Kellam, Brown, Rubin, and Ensminger, 1983; Smith & Fogg, 1979). Furthermore, this early onset substance-use group has been identified to be at heightened early risk in a variety of domains, including early conduct problems, high risk taking, poor school performance, and exposure to family conflict/divorce (e.g. Colder, Campbell, Ruel, Richardson, & Flay, 2002; Flory, Lynam, Milich, Leukefeld, & Clayton, 2004; Orlando, Tucker, Ellickson, & Klein, 2004; Tucker, Orlando, & Ellickson, 2003). Thus, there may be a developmentally unique set of social processes leading to early adolescent substance use, compared to youth who either never engage in antisocial behavior yet experiment with drugs somewhat later in development, or who never engage in either form of high-risk behavior (Dishion & Loeber, 1985; Moffitt, 1993; Patterson & Yoeger, 1993). If these multiple-problem youth are not distinguished from the rest of the sample, we run the risk of overemphasizing risk dynamics among youth who are otherwise on a normative developmental trajectory.
In the current study we sought to address this possibility using data...
|
