Skip to main content
SearchLoginLogin or Signup

Substance Use Profiles Among Juvenile Offenders: A Lifestyles Theoretical Perspective

Published onJan 01, 2017
Substance Use Profiles Among Juvenile Offenders: A Lifestyles Theoretical Perspective
·

Abstract

Base rates of illicit substances such as cocaine, crack cocaine, and heroin are typically low in community-based studies, which often inhibits more complex multivariate analysis. Also, single-item measures and aggregate scales mask within-group differences among those showing versatility in their substance use users. Latent class analysis was used to model the substance use profiles of adjudicated female (n =98) and male (n = 378) youth. Alcohol, marijuana, acid, mushrooms, ecstasy, cocaine, crack cocaine, heroin, crystal methamphetamine, and non-medical use of prescription pills were used to define latent profiles of substance use. Three latent classes were identified that were qualitatively different across males and females. Multinomial logistic regression analyses indicated that time spent outside of the home of the biological parents, early substance use, and parental substance abuse were informative of the use of substances such as cocaine, crack cocaine, and heroin. Implications for more individualized treatment strategies are discussed.

Keywords. Gender; latent class analysis; risky lifestyles; substance use


Although prevalence rates of substance use vary across different samples, overall, substance use is a relatively normative part of adolescent development (Doherty, Green, & Ensminger, 2007; Patrick & O’Malley, 2016; Peiper, Ridenour, Hochwalt, & Coyne-Beasley, 2016; Johnston, O’Malley, Bachman, & Schulenberg, 2006). For a small proportion of the adolescent population, however, substance use begins earlier in the life course and is more frequent and versatile compared to other adolescent substance users. For this small group, understanding their substance use profiles, risk factors, and appropriate treatment strategies is important for practitioners within the criminal justice, public health, and social services sectors to, for example, help inform case management and treatment planning (see Baron, 2006; Doherty, et al., 2007; Vaughn, Salas-Wright, DeLisi, & Piquero, 2014). Despite recommendations for more research on adolescents that use substances such as crack cocaine and heroin (Vaughn & Howard, 2004), relatively little is known about this group because the prevalence of the use of such substances is extremely low within high school samples (e.g., Patrick & O’Malley, 2016). As well, such substances are more likely to be used among young offenders, youth that dropped out of school, left home, were placed in foster care, or were living on the street (see Baron, 2006; Vaughn, Ollie, McMillen, Scott, & Munson, 2007), and these types of individuals are rarely found in studies that sample from schools or require caregiver consent to participate (e.g., Jessor & Jessor, 1978; Johnson et al., 1978; Johnson et al., 2006). In effect, the type of research design commonly used to study adolescent substance use, although valuable for describing the characteristics of adolescent marijuana and alcohol use, is unlikely to include a sufficient subsample of adolescents whose substance use is more versatile and frequent. Gilreath et al. (2014) noted the specific lack of attention to nuanced differences in the substance use profiles of those using multiple illicit substances. Consequently, there is also a lack of research on correlates associated with adolescents characterized by a more versatile pattern of substance use.

The development of intervention and treatment strategies would seem particularly important for adolescents with a versatile substance use pattern given this group is at a disproportionate risk of experiencing a variety of negative outcomes associated with general substance use, including family problems, health problems, and involvement in serious and violet offending (Baron, 2006; DeLisi, Vaughn, Salas-Wright, & Jennings, 2015; Newcomb & Bentler, 1988). In terms of the types of factors that intervention strategies should target, adolescent substance users are more likely than non-substance users to be involved in a variety of other risky behaviors such as early sexual behavior, offending, and running away from home/living on the street (e.g., Baron, 2006; Chassin, Flora, & King, 2004; Elliott & Morse, 1989; Raskin-White, Loeber, & Farrington, 2008; Welte, Zhang, & Wieczorek, 2001). However, less is known about within-group differences among adolescent substance users and whether involvement in risky behaviors and an associated lifestyle help explain within-group heterogeneity in substance use profiles. A better understanding of the correlates of a more frequent or more versatile substance use pattern is important for helping prevent negative outcomes (e.g., serious and violent offending) for the types of substance users most likely to experience such outcomes (DeLisi et al., 2015).

Using a sample of adjudicated Canadian adolescent males and females (n = 476), the current study examined heterogeneity in adolescent substance use profiles using latent class analysis. There is a noted absence of females within studies of substance use within at-risk adolescent populations (e.g., Baron, 2006; Chassin, Knight, Vargas-Chanes, Losoya, & Naranjo, 2009) and there is some evidence from community studies that substance use patterns differ across males and females (Newcomb, Maddahian, Skager, & Bentler, 1987). Thus, a specific purpose of this analysis was to evaluate whether female and male substance use profiles differed. Another aim of this study involved examining whether negative lifestyle factors were associated with specific substance use profiles. Within this sample, drug use was common (i.e., approximately 99%) and participants used a wide range of different substances (i.e., approximately 75% used at least three different substances). In effect, these were the types of adolescent substance users that are concerning to treatment providers and policymakers because they are more likely to be associated with several other treatment needs (e.g., homelessness, mental disorder, offending, parental conflict; see Baron, 2006; DeLisi et al., 2015; Newcomb & Bentler, 1988).

SUBSTANCE USE AND THE LOW BASE RATE PROBLEM

As noted by Baron (2006) and others (e.g., Hagan & McCarthy, 1997), most research concerns community or student populations and the relationship between substance use and minor delinquency. Within these studies, substance use is often aggregated into different categories (e.g., ‘hard drug use’), combined into a cumulative measure, or based on single-item indicators of substances more common to the general population (e.g., alcohol, marijuana). Using these broader measures is necessary because the base rate of more concerning types of substances (e.g., cocaine, heroin, crystal methamphetamine) is too low for statistical analysis (e.g., Agrawal, Lynskey, Madden, Bucholz, & Heath, 2007; Johnson et al., 2006; Kandel & Logan, 1984; Newcomb, Maddahian, & Bentler, 1986). However, aggregate scales of substance use are inappropriate for evaluating person-centered substance use profiles because such scales do not capture, for example, the difference between two individuals both using three substances, the first using alcohol, marijuana, and acid, and the second using heroin, crack cocaine, and crystal methamphetamine. Arguably, these are two different types of substance users, but an aggregate scale does not capture this difference. A person-centered approach is useful for seeing patterns of development within individuals as opposed to between groups (e.g., Magnusson & Bergman, 1988). In the context of research on substance use, this allows for more of an individualized focus on substance use patterns.

Low base rates are therefore an impediment to addressing important research questions regarding more versatile substance use profiles and their association with other risky behaviors/lifestyles. For example, the variance of variables that have a low base-rate is restricted, which reduces the maximum correlation possible with other variables, thereby limiting the statistical techniques that can be used (Copas & Tarling, 1986; MacLennan, 1988). As such, within a group of substance users, important differences cannot be extrapolated. By way of illustration, in a sample of approximately 15,000 grade twelve students, lifetime prevalence of heroin, ecstasy, crystal methamphetamine, hallucinogens, cocaine, and crack cocaine use were all less than 13% and therefore it was necessary to aggregate these substances into one measure (Johnston et al., 2006). Indeed, aggregation of different substances into a single measure is a common theme with the substance use research (see Baron, 2006; Chassin et al. 2004; Chassin et al., 2009; Thornberry et al., 1995; Milloy, Kerr, Buxton, Montaner, & Wood, 2009; Torok, Darke, Kaye, & Ross, 2011; van Kammen & Loeber, 1994). Aggregate measures of substance use (e.g., minor drug use, hard drug use, polysubstance use) may be problematic because such measures assume that individuals using different drugs have the same treatment needs (Braker, Gobel, Scheithauer, & Soellner, 2015).

Latent class analysis (LCA) has been used to create a multidimensional measure of adolescent substance use that provides a more individualized description of adolescent substance use profile. For example, rather than comparing presence/absence of substance use, Jackson et al.’s (2000) person-centered approach allowed typologies of the severity of alcohol and tobacco use profiles to emerge within their sample of male and female freshmen at a Midwestern university (n = 489). However, because most of the LCA-based research sampled from college and high-school populations (e.g., Agrawal et al., 2007; Cleveland, Collins, Lanza, Greenberg, & Feinberg, 2010; DeLisi et al., 2015; Evans-Polce, Lanza, & Maggs, 2015; Jackson et al., 2000; Shin, Hong, & Hazen, 2010; Vaughn, Freedenthal, Jenson, & Howard, 2007), similar base rate problems emerged. Indeed, a common theme within the LCA literature is that latent classes characterized by use of more than one illicit substance represent less than ten percent of the sample (e.g., Agrawal et al., 2007; Cranford et al., 2013; DeLisi et al., 2015; Evans-Polce et al., 2015), and this is true even within populations where substance use is expected to be more common, such as homeless youth and youth in foster care (e.g., Snyder & Smith, 2015). Consequently, more sophisticated multivariate models were not possible within these studies. Sampling from populations where substance use is highly prevalent is one strategy for addressing this low base rate problem and for avoiding aggregate measures that potentially mask important distinctions in substance use profiles.

INCARCERATED POPULATIONS AS AN ANSWER TO THE BASE RATE PROBLEM

Teplin and colleagues (Teplin, Abram, McClelland, Dulcan, & Mericle, 2002; Teplin et al., 2006) observed that substance use was especially prevalent within samples of adolescent offenders. In their review of epidemiological studies of adolescents, the prevalence of substance use disorders ranged from 22 to 88 percent (Teplin et al., 2006). The observed variation in the prevalence of substance use disorders was likely due to the comparison of studies that sampled from different offender populations. Indeed, incarcerated offenders were more likely to have a substance use disorder compared to offenders on probation (Teplin et al., 2006). Corrado (2002) described incarcerated youth as posing particular treatment and intervention challenges relative to other offenders because of the multi-problem risk factors characterizing the former. These multi-problem risk factor profiles include histories of physical and sexual abuse, impulsivity, exposure to delinquent models, negative parent-child relationships, and homelessness/foster care placement. That these risk factors overlap with correlates of substance use (see Hawkins, Catalano, & Miller, 1992; Loeber, 1988) might help explain why the prevalence of substance use is higher in incarcerated samples compared to samples of youth on probation and other at-risk samples (e.g., Baron, 2006). There also appear to be differences in the prevalence of substance use within samples of incarcerated youth. In Teplin et al.’s (2002) study based on data from the prospective longitudinal Northwestern Juvenile Project (NJP), the prevalence of substance use disorders was higher for White offenders compared to Hispanic and African American offenders. Males and females, however, showed a similar (approximately 50%) prevalence rat of substance use disorder. Whether similarities in the prevalence of a substance use disorder reflected similarities in the types of substances use remained unclear. Given their tendency for a more frequent or versatile substance use pattern compared to other groups (Teplin et al., 2002), understanding correlates of incarcerated adolescent offender substance use may be helpful for the future development of treatment and intervention strategies.

CORRELATES OF SUBSTANCE USE

Risk factors related to adolescent substance use include macro-level factors such as social disorganization at the neighbourhood level, meso-level factors such as school, family, and peer structure, and micro-level factors such as low levels of self-control and a risky lifestyle (see Hawkins et al., 1992; Loeber, 1988). A risky lifestyle was described as particularly important for adolescent substance use (Baron, 2006) and is of central focus in the current study. Lifestyle risk factors such as homelessness or placement in foster care increase an individual’s level of exposure to opportunities or circumstances conducive to substance use such as exposure to street conditions and other substance users (e.g., Baron, 2006). Whether individuals associated with this risky lifestyle can be linked to more versatile substance use profiles that move beyond aggregate scales, and whether these risk factors are important for both male and female substance users may be useful for developing more individualized and specific intervention/treatment programs (Braker et al., 2015.

METHODS

SAMPLE

Adjudicated female (n = 98) and male (n = 378) youth were interviewed in open and secure custody facilities in British Columbia, Canada between 1998 and 2001 as part of the first phase of the Incarcerated Serious and Violent Young Offender Study. Although only 4.9% of the population of British Columbia self-identifies as Indigenous, 22.1% (n = 105) of participants in the current study self-identified as such. The over-representation of Indigenous participants is dissimilar from most incarcerated samples in the United States (e.g., Teplin et al., 2013), although the over-representation of adjudicated Indigenous participants in this sample mirrors the over-representation of Black and Hispanic participants found within incarcerated samples in the United States (Teplin et al., 2013). Additionally, all participants were incarcerated at the time of their interview which meant that the sample likely differed from other adjudicated youth who received a less punitive sentence (e.g., probation, community work service). Consequently, the sample is particularly specific and cannot be generalized to the average adolescent female or male in the community. However, the sample does resemble the proportion of female versus male youth on probation in British Columbia (Calverley, Cotter, & Halla, 2010). Characteristics of the sample can be found in Table 1.

PROCEDURE

Informed consent was first provided by the British Columbia Ministry of Child and Family Development (MCFD). MCFD serves as the legal guardian to all youth in custody. Consent from MCFD allowed this project to approach youth in various custody centres throughout the province. Research assistants (RAs) approached subjects at their unit within the custody centre and asked if they wanted to participate in a research study. Refusal rates were only recorded for a portion of the entire study period. Over the period in which information on refusal rates was recorded, approximately five percent of youth declined to participate. To help protect confidentiality, those that wished to participate were interviewed in a room away from their living unit, other youth, and custody staff. All participants were read and given a copy of an information sheet which explained the purpose of the study, which was to collect self-report and official file information on the risk factors associated with the development of serious and violent offending. This sheet also explained how information would be collected (e.g. interview and file information) and that information would be kept confidential unless the participant made a direct threat against themselves or to someone else. Participants gave their assent by signing a form indicating that they had been read and understood the details of the study as described in the information sheet.

MEASURES

SUBSTANCE USE AND ITS CORRELATES. Participants self-reported whether they used each of a total of ten substances: alcohol, marijuana, acid, mushrooms, cocaine, ecstasy, crack cocaine, heroin, crystal methamphetamine, and non-medical use of prescription pills (e.g., taking more of their medication than required, using someone else’s prescription, or buying pills from someone else). These ten substances were included in the LCA. Frequency of use was not measured, but using a versatility scale is helpful in describing profiles of substance users and is commensurate with prior research using LCA (Agrawal et al., 2007; Cleveland et al., 2010; DeLisi et al., 2015; Evans-Polce et al., 2015; Jackson et al., 2000; Shin et al., 2010; Vaughn et al., 2007). Participants were also asked to report the age at which they began using drugs and the age at which they began using alcohol. Early drug use and early alcohol use were defined as use prior to age 12. Individuals using illicit substances at this age do so before entry to high school, a time where substance use is more normative (Patrick & O’Malley, 2016) and provides social rewards (e.g., Gallupe & Bouchard, 2013). Of the sample, 1.0% (n = 1) of females and 1.6% (n = 6) of males reported using neither alcohol nor drugs. The average participant used 4.72 (SD = 2.51) of the 10 substances measured and 76.2% (n = 365) of the sample reported use of three or more different illicit substances. Females were significantly more likely than males to report using substances such as heroin and crack cocaine (see Figure 1) and also reported a significantly greater number of different substances use compared to males (t[472] = 2.24, p < .05).

--Insert Figure 1 about Here--

Different factors associated with an at-risk lifestyle were measured through self-report interviews and included whether the participant had been kicked out of their home for more than a day, had left their home to live somewhere else for more than a day, were not attending school at the time of interview, and had been placed in foster care. Other risk factors known to be associated with a risky lifestyle were also measured in this study, including physical and sexual abuse experiences, negative sense of self, a biological parent with an alcohol abuse problem, and a biological parent with a drug abuse problem. All variables were dichotomously measured, except for negative sense of self, which was measured via Schneider’s (1990) Good Citizen Scale. This scale includes 15 items scored from 1-7, with higher scores used to indicate a more positive sense of self-identity. Example items include whether the youth saw themselves as more dishonest or honest, bad or good, unattractive or attractive, lazy or hardworking. Cronbach’s alpha for the 15 items measured in the current study resulted in a relatively low value (0.663). After removing one item (tough or weak), the Cronbach’s alpha value was acceptable (.700). The 14-item scale was used in subsequent analyses. These different risk factors as well as substance use measures were compared across females and males and are shown in Table 1.

--Insert Table 1 about Here--

ANALYTIC STRATEGY

Proc LCA for SAS 9.4 was used to conduct the LCA (Lanza, Collins, Lemmon, & Schafer, 2007). LCA creates homogenous classes, meaning that individuals within each class are highly similar, at least with respect to the items used to define the latent classes (Collins & Lanza, 2010). A growing number of studies used LCA to elucidate different substance use profiles based on the expectation that substance users are a heterogeneous group (Agrawal et al., 2007; Cleveland et al., 2010; DeLisi et al., 2015; Evans-Polce et al., 2015; Jackson et al., 2000; Shin et al., 2010; Vaughn et al., 2007). The appropriate number of latent classes underlying the data is determined by running successive latent models, beginning with a one class solution, and then comparing changes in penalized log likelihood values represented by Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC) values. When comparing models, AIC and BIC values closer to zero indicate an improvement in model fit (Lanza et al., 2007). For the current study, substance use profiles represented the latent construct of interest and were measured using the ten measures of illicit substances shown in Figure 1. Given that substance use profiles can vary across gender (Shin et al., 2010), the initial LCA included gender as a grouping variable to measure whether the qualitative meanings of the latent classes differed across levels of group membership (i.e., across females and males). A series of ANOVA analyses were used to help validate the latent classes identified by comparing the average number of substances used across the different latent classes. Bonferonni and Tamhane post-hoc tests were performed depending on whether the assumption of equal variance was violated. Another bivariate analysis was performed to compare latent class membership across demographic characteristics and risk factors. Following this, all marginally significant (p < .10) and significant (p < .05) risk factors in the bivariate analysis were included in a multinomial logistic regression analysis with the latent groups of substance users as the outcome of interest. Marginally significant covariates were included to favor the avoidance of excluding potentially important risk factors.

RESULTS

The first step in the LCA involved identifying the number of classes that best fit the data on the ten substances. AIC values indicated that a five-class solution (AIC = 375.90) fit the data better than a four (AIC = 380.57) and six (AIC = 378.72) class solution. BIC values indicated that a three-class solution (BIC = 536.63) fit the data better than a two (BIC = 585.02) and four (BIC = 559.69) class solution. Considering both the principle of parsimony and that BIC values are adjusted for the number of items included in the model relative to sample size (e.g., Collins & Lanza, 2010), a three class solution was retained. An analysis with gender as a grouping variable was performed using this baseline model. This analysis compared one model where item-response probilities were allowed to vary across gender to another model where item-response probabilities were constrained to be equal across groups. Significant qualitative diffences in female and male substance use profiles were observed (χ2 [30] = 80.61, p < .001). Accordingly, the final model retained was a three-group solution which allowed item-response probabilities to vary across females and males. The maximum probability of assignment rule was used to assign participants to the latent class that best described their own substance use profile. Too much uncertainty in the probability of assignment can lead to concerns regarding the fit between the individual and the latent class to which they were assigned. In this model, entropy values exceeded 0.70 (0.79), which indicated good classification accuracy. Average probability of correct assignment, per posterior probabilities, ranged from 0.85-0.93 across the three latent classes. Odds of correct classification (OCC) helped provide further confidence that individuals were assigned to the appropriate latent class. OCC values were calculated as:

OCCg = (AvePPg/ (1-AvePPg)) / (∏g/ (1-∏g))

where ∏g is the estimated size of group g (see Skardhamar, 2010).

OCC values for the three latent classes ranged from 6.04-13.20, which was higher than Skardhamar’s (2010) recommendation that OCC values of at least five be used to indicate high classification accuracy.

INTERPRETATION OF FEMALE AND MALE LATENT CLASSES

Based on the item-response probabilities for the female analysis (see Figure 2a), the three latent classes of substance use profiles for females were named ‘Low Versatility’ (n = 28; 28.6% of females in the sample), ‘Hallucinogen Users’ (n = 15; 15.3% of females), and ‘Street Drug Users’ (n = 55; 56.1% of females). For the Low Versatility group, item response probabilities were high for alcohol (0.84) and marijuana (0.76) use but relatively low for all other substances. One participant in this group did not use any substances (3.6% of the group). Females in this latent class had used an average of 2.29 (SD = 1.01) different substances. For the Hallucinogen Users latent class, like the Low Versatility group, item response probabilities were high for alcohol and marijuana use (1.00 and 0.82, respectively). Per the item response probabilities, conscious-altering substances such as use of mushrooms and acid were also common (0.98 and 0.56, respectively). Females in the Hallucinogen Users latent class had used an average of 4.40 (SD = 0.99) different substances. For Street Drug Users, item response probabilities were relatively high for a wide range of substances, but the defining feature of this group were the high item response probabilities for heroin (0.76), cocaine (0.96), and crack cocaine (0.83). Females in this latent class had used an average of 6.98 (SD = 1.42) different substances. Based on an ANOVA analysis, significant differences in the average number of different substances used were observed when comparing across each latent class (see Table 2). More specifically, females in the Street Drug User latent class averaged a significantly greater number of different substances used compared to the other two latent classes. All females in the Street Drug User latent class used at least one of heroin, cocaine, crack cocaine, or crystal methamphetamine

The three male substance use latent classes (see Figure 2b) were also referred to as Low Versatility (n = 137; 36.2% of males in the sample), Hallucinogen Users (n = 136; 30.0% of males), and Street Drug Users (n = 105; 27.8% of the sample). For the male Low Versatility latent class, high item response probabilities were observed for alcohol (0.88) and marijuana (0.79) use; however, males in this class were associated with low probabilities of using other substances. Six of the males in this latent class (4.4% of the group) reported using neither drugs nor alcohol. Males in this latent class had used on average 2.04 (SD = 0.87) different substances. For the Hallucinogen Users latent class, high item response probabilities were observed for alcohol (0.97) and marijuana (0.98) use, but individuals in this group also had a relatively high probability of endorsing use of conscious-altering substances such as mushrooms (0.77) and acid (0.82). Males in this latent class had used on average 4.63 (SD = 0.95) different substances. For Street Drug Users, like their female counterpart and unlike the other two male latent classes of substance users, this group had relatively high probabilities of also endorsing heroin (0.69), cocaine (0.90), and crack cocaine (0.85). Males in this latent class had used on average 7.90 (SD = 1.09) different substances. Based on an ANOVA analysis (see Table 3), significant differences in the average number of different substances used were observed when comparing across each latent class. Post-hoc analyses showed that the Street Drug User latent class averaged a significantly greater number of different substances used compared to the other two latent classes. All males in the Street Drug User latent class reported using at least one of heroin, cocaine, crack cocaine, or crystal methamphetamine

­--Insert Figure 2a and 2b about Here--

CORRELATES OF SUBSTANCE USE PROFILES

Given that the meanings of the latent classes varied across gender, bivariate associations between substance use profiles and risk factors for substance use were compared separately across females (Table 2) and males (Table 3). For females, early drug use, a history of leaving home to live somewhere else, and having a biological parent with a drug abuse problem were significantly associated with a more versatile substance use profile. For males, being White or Indigenous, early alcohol use, early drug use, a history of leaving home to live somewhere else, a history of being kicked out of the home, not attending school, having a more negative self-identity, and having a biological parent with a drug use problem were significantly associated with a more problematic substance use profile.

--Insert Table 2 about Here--

--Insert Table 3 about Here--

Variables that were significant (p < .05) or marginally significant (p < .10) in the bivariate analyses were included in the male and female MLR analyses to identify predictors of substance use profiles. For the female MLR model (left-hand side of Table 4), the Low Versatility latent class was treated as the reference group to which the other two latent classes were compared. With the Hallucinogen Users latent class, only one individual reported that they did not leave their home to live somewhere else. Consequently, a high standard error and wide-ranging 95% confidence interval were observed and so this variable was excluded from the model. The overall MLR model for females was significant, but only early drug use significantly differentiated latent class membership. More specifically, controlling for age, ethnicity, and biological parent drug use, early drug users were approximately 15 times more likely to be in the Hallucinogen Users latent class compared to the Low Versatility latent class and were approximately 14 times more likely to be in the Street Drug User latent class compared to the Low Versatility latent class.

--Insert Table 4 about Here--

For the male MLR model (see right-hand side of Table 4), the Low Versatility latent class was treated as the reference group to which the other two latent classes were compared. The overall model was significant. For the Hallucinogen Users, being a non-Indigenous ethnic minority decreased the odds of membership in this latent class relative to the Low Versatility latent class by approximately 68% (p < .01). Male participants that reported leaving home for more than 24 hours to live somewhere else were twice as likely to be in the Hallucinogen Users latent class (p < .05). Finally, male participants that attended school were approximately 45% less likely to be in the Hallucinogen Users latent class (p < .10). For the Street Drug User latent class, males that used alcohol before age 12 were nearly three times more likely to be associated with this latent class compared to Low Versatility latent class (p < .05). The odds of being in the Street Drug User latent class were twice as high for male participants that reported that they had been kicked out of their home (p < .05) and were three times as high for participants that had left their home for more than 24 hours to live somewhere else (p < .01). Male participants that were attending school were 58% less likely to be associated with the Street Drug User latent class (p < .05). Finally, although only marginally significant, those with a more positive self-identity were less likely to be associated with the Street Drug User latent class.

DISCUSSION

Most studies of adolescent substance users sample from community-based populations where the base rate of more illicit substances such as cocaine, crack cocaine, heroin, and crystal methamphetamine is relatively low. Consequently, studies often examine the use of more normative adolescent substances (e.g., alcohol, marijuana) or rely on aggregate measures that cannot describe more nuanced substance use profiles. In the current study, data were used that included a sample of adjudicated male and female youth; among this group, substance use was extremely common (approximately 99% of the sample). The lifetime usage of ten substances (alcohol, marijuana, acid, mushrooms, ecstasy, cocaine, crack cocaine, heroin, crystal methamphetamine, and non-medical use of prescription pills) were self-reported by participants. Defining substance use by lifetime prevalence was consistent with prior research (e.g., Mitchell & Plunkett, 2000; Snyder & Smith, 2015; White et al., 2013). These ten items were entered in a latent class analysis to evaluate the number and nature of substance use profiles that best fit the data, and whether these profiles qualitatively differed between male and female participants. This person-oriented approach to identifying a participant’s substance use profile addressed a need for research that measured substance use along multiple dimensions (e.g., Shin et al., 2010) and among a sample of individuals where substance use was highly prevalent. The use of LCA helped expand on work using additive scales of substance use. Specifically, additive scales lose the specificity of individual substances through aggregation and thus cannot allow for interpretation of qualitative differences in the versatility of substance use. Here, individuals using the same number of different types of substances could also be distinguished based on the type of substance used as well. In effect, unlike measures of polysubstance use or versatility of substance use captured by prior studies using additive scales, it was possible in the current study to examine versatility in the number of substances used without giving equal weight to each of the ten substances measured. Further, by including a variety of risk factors for substance use, the current study identified correlates associated with youth characterized by a more problematic substance use profile.

Using LCA, three different profiles of adolescent substance use were identified: (1) non-use or use of only minor substances like alcohol and marijuana (i.e., Low Versatility), (2) use of hallucinogenic drugs (i.e., Hallucinogen Users), and (3) elevated probabilities of using street drugs such as crack cocaine, heroin, and crystal methamphetamine (i.e., Street Drug Users). The findings add to the literature supporting the assertion that adolescent substance users are not a homogenous group. The nature of the latent classes was also similar to observations from prior studies using LCA (e.g., DeLisi et al., 2015; Snyder & Smith, 2015; White et al., 2013). Where the current study differed was in the prevalence of those characterized by use of illicit substances such as cocaine, crack cocaine, heroin, and crystal methamphetamine. Other studies, including those of more at-risk samples such as youth in foster care (e.g., Snyder & Smith, 2015), found that only approximately three percent of their sample showed historical use of three or more different illicit substances. Consequently, unlike the current study, these prior studies did not include a multivariate analysis to examine the correlates of substance use profiles because of the low base rate of adolescents with a versatile substance use pattern that included illicit substances such as cocaine, crack cocaine, heroin, and crystal methamphetamine.

The substance abuse profiles observed were also gendered. That is, profiles were qualitatively (per the measurement invariance test) and quantitatively different across gender. Regarding quantitative differences, the prevalence of females in the Street Drug User latent class was twice as high as the prevalence of males in the Street Drug User latent class. As a caution, item response probabilities differed slightly between the female Street Drug User latent classes and the male Street Drug User latent class, which makes direct comparisons in the prevalence of females and males in each latent class difficult. For example, although not statistically compared, ecstasy and crystal methamphetamine appeared to be more likely to be endorsed among male Street Drug Users compared to female Street Drug Users. Nevertheless, the general finding in this study was that females were associated with a more versatile substance use pattern. As shown in Figure 1, the prevalence of cocaine, crack cocaine, and heroin, were all more common for females compared to males. Gendered differences in substance use profiles may be related to the large number of female youth incarcerated specifically because of their use of street drugs (Corrado, Odgers, & Cohen, 2000).

The different substance use profiles observed in the current study were associated with different risk factor profiles and thus adolescents associated with different substance use profiles may require different treatment and intervention strategies. Specifically, for both females and males, living a riskier lifestyle increased the odds of also being associated with a more versatile substance use pattern characterized specifically by an increased probability of the use of cocaine, crack cocaine, heroin, and crystal methamphetamine. Early use of illicit substances (drug use for females, alcohol use for males) and having residential mobility issues (i.e., getting kicked out of their home or leaving the home voluntarily to live somewhere else) increased the odds of membership in the Hallucinogen and Street Drug User latent classes compared to the Low Versatility latent class. That an earlier onset of substance use was associated with a more versatile profile of substance use for both males and females was consistent with the criminal career literature showing that age of onset of offending is informative of other criminal career dimensions including frequent and versatile offending (e.g., Le Blanc & Loeber, 1998). Among more versatile substance users, especially those associated with the Street Drug User latent class, the lack of a stable home life and the tendency, at least for males, to not be attending school, are indicative of a lifestyle characterized by unstructured activities and unsupervised time in the community. That some of these youth may have been living on the street while outside of the home could potentially exacerbate or influence their use of illicit substances. Although effect sizes at the bivariate level were small, having a biological parent with a drug use problem appeared to differentiate male and female participants in the Street Drug User latent class from those in the Low Versatility latent class. Previous studies indicated that youth with substance-using parents are also likely to use drugs (e.g., Baron, 2006; Hawkins et al., 1992), and the current study adds to this literature by showing, at least at the bivariate level, that a substance-abusing parent is a correlate of a more versatile substance use profile. This relationship was not observed in the multivariate analysis. Future research should examine whether the effect of a substance-abusing parent on their child’s own substance use is mediated by the tendency for adolescents with a substance-abusing parent to leave the home and spend more time on the street.

Treatment programs in institutional settings have had limited success because the drug use of individuals within these contexts is typically quite frequent and versatile (Vanderwaal, McBride, Terry-McElrath, & Van Buren, 2001). A movement towards more individualized programs can be effective in treating substance use issues, but at the same time is costly and requires a lot of resources and case-planning (Schackman, Rojas, Gans, Falco, & Millman, 2007; French et al., 2008). LCA is described as especially useful for identifying subgroups of homogenous individuals that may differ in their response to a standardized treatment program (Lanza & Rhoades, 2013). The continued use of this analytic strategy may help provide a treatment solution that is both highly specialized without being highly individualized and costly (also see Braker, 2015). For example, in the current study, three different types of substance use treatment programming might be warranted (i.e., distinct programs for Low Versatility, Hallucinogen Users, and Street Drug Users) and it is likely that the individuals assigned to a given group will appear highly similar in terms of their substance use profile. Moreover, by examining the factors surrounding substance use latent profiles, it was observed that a more versatile substance use profile was associated with a more tumultuous home environment. Therefore, it may also be necessary to develop multisystemic therapy and similar treatment programs to help youth and family reconcile in a way that provides a supportive home environment (Henggeler, Clingempeel, Brondino, & Pickrel, 2002).

LIMITATIONS AND FUTURE RESEARCH

The current study uses a very specific sample of serious and violent young offenders, which means that the latent classes should not be generalized to same-aged youth who are not involved in the legal system. The item response probabilities for all substances included in the current study are much higher than what past research suggests because past research sampled from high schools where substance use is less prevalent compared to custody facilities (Johnston et al., 2006; Cleveland et al., 2010; Shin et al., 2010). Although the study findings may be less generalizable to community populations, it is also atypical for individuals within such populations to be characterized by these more versatile substance use profiles that were observed within the sample in the current study. In other words, the study design was necessary to learn about the types of adolescents involved in use of substances such as cocaine, crack cocaine, heroin, and crystal methamphetamine. The research design used resembled prior substance use literature sampling from homeless or street-involved youth (e.g., Baron, 2006). The research design was also cross-sectional, which prevented the use of latent transition analysis to elucidate within-individual change or stability in substance use profiles over time and would help clarify, for example, the order in which Street Drug Users began using cocaine, crack cocaine, and heroin (see Baggio et al., 2014; Maldonado-Molina & Lanza, 2010). It would also be helpful to look beyond lifestyle risk factors to more fully account for the characteristics of more versatile substance use profiles. Future research should consider the temporal role of internalizing and externalizing disorders in the development of problematic substance use (Carney et al., 2016).

Each latent class was created based on whether an individual had used each of the ten substances. The frequency with which an individual uses these substances was not included when creating the latent classes. Future research should examine whether youth frequency of substance use was associated with versatility profiles to validate assertions that more versatile profiles are most likely to be found within youth that also use substances more frequently than youth with less versatile profiles of use. Future research should also use more recent data on substance use profiles. Using data from the Monitoring the Future Study, Patrick and O’Malley (2016) showed that substance use among high school students from the United States declined at approximately the same time as the initiation of the current study. It is possible that the prevalence of more versatile substance users would be lower within a more recent sample of incarcerated young offenders. Fortunately for the interpretation of the findings from the current study, Patrick and O’Malley (2016) observed declines in the prevalence of use across a wide range of illicit substances, and these declines were rank-ordered. In other words, they did not typically observe changes in the popularity of different illicit substances. For example, the prevalence of cocaine declined, but still remained more prevalent in use compared to other key substances measured in the current study such as crack cocaine, heroin, and crystal methamphetamine (Patrick & O’Malley, 2016). Therefore, knowledge about the versatility of illicit substance use, specifically the use of the abovementioned four illicit substances, is likely to still have utility for treatment providers. One exception to consider for future research is that the illegal use of prescription pills increased in prevalence in use among high school students over the last decade (Patrick & O’Malley, 2016).

REFERENCES

Agrawal, A., Lynskey, M. T., Madden, P. A., Bucholz, K. K., & Heath, A. C. (2007). A latent class analysis of illicit drug abuse/dependence: results from the National Epidemiological Survey on Alcohol and Related Conditions. Addiction102(1), 94-104.

Baggio, S., Studer, J., Deline, S., N’Goran, A., Dupuis, M., Henchoz, Y., Mohler-Kuo, M, Daeppen, J.B., & Gmel, G. (2014). Patterns and transitions in substance use among young Swiss men: A latent transition analysis approach. Journal of Drug Issues44(4), 381-393.

Baron, S. W. (2006). Street youth, strain theory, and crime. Journal of Criminal Justice34(2), 209-223.

Bräker, A. B., Göbel, K., Scheithauer, H., & Soellner, R. (2015). Adolescent alcohol use patterns from 25 European countries. Journal of Drug Issues45(4), 336-350.

Carney, R., Yung, A. R., Amminger, G. P., Bradshaw, T., Glozier, N., Hermens, D. F., ... & Wood, S. J. (2016). Substance use in youth at risk for psychosis. Schizophrenia Research, (online first), 1-7.

Chassin, L., Flora, D. B., & King, K. M. (2004). Trajectories of alcohol and drug use and dependence from adolescence to adulthood: The effects of familial alcoholism and personality. Journal of Abnormal Psychology113(4), 483-498.

Chassin, L., Knight, G., Vargas-Chanes, D., Losoya, S. H., & Naranjo, D. (2009). Substance use treatment outcomes in a sample of male serious juvenile offenders. Journal of Substance Abuse Treatment36(2), 183-194.

Cleveland, M. J., Collins, L. M., Lanza, S. T., Greenberg, M. T., & Feinberg, M. E. (2010). Does individual risk moderate the effect of contextual-level protective factors? A latent class analysis of substance use. Journal of Prevention and Intervention in the Community, 38(3), 213-228.

Collins, L., & Lanza, S. (2010). Latent class and latent transition analysis: With applications in the social, behavioral, and health sciences. Hoboken, NJ: Wiley.

Copas, J. B., & Tarling, R. (1986). Some methodological issues in making predictions. In J.G. Weis (Ed.), Criminal careers and career criminals (v.2), pp. 291-313.

Corrado, R.R. (2002). An introduction to the risk/needs case management instrument for children and youth at risk of violence: The Cracow Instrument. In R.R. Corrado, R. Roesch, S.D. Hart, & J.K Gierowski, J.K. (Eds.). Multi-problem violent youth: A foundation for comparative research on needs, interventions and outcomes. Amsterdam: IOS Press.

Corrado, R. R., Odgers, C., & Cohen, I. M. (2000). The incarceration of female young offenders: Protection for whom. Canadian Journal of Criminology42, 189-207.

Doherty, E. E., Green, K. M., & Ensminger, M. E. (2008). Investigating the long-term influence of adolescent delinquency on drug use initiation. Drug and Alcohol Dependence93(1), 72-84.

Elliott, D. S., & Morse, B. J. (1989). Delinquency and drug use as risk factors in teenage sexual activity. Youth and Society21(1), 32.

French, M.T., Zavala, S.K., McCollister, K.E., Waldron, H.B., Turner, C.W., Ozechowski, T.J. (2008). Cost-effectiveness of four interventions for adolescents with a substance use disorder. Journal of Substance Abuse Treatment 34(2), 272-281.

Gallupe, O., & Bouchard, M. (2013). Adolescent parties and substance use: A situational approach to peer influence. Journal of Criminal Justice41(3), 162-171.

Gilreath, T. D., Astor, R. A., Estrada, J. N., Johnson, R. M., Benbenishty, R., & Unger, J. B. (2014). Substance use among adolescents in California: a latent class analysis. Substance Use & Misuse49, 116-123.

Hagan, J., & McCarthy, B. (1997). Mean streets: Youth crime and homelessness. Cambridge, UK: Cambridge University Press.

Hawkins, J. D., Catalano, R. F., & Miller, J. Y. (1992). Risk and protective factors for alcohol and other drug problems in adolescence and early adulthood: Implications for substance abuse prevention. Psychological Bulletin112(1), 64-105.

Henggeler, S. W., Clingempeel, W. G., Brondino, M. J., & Pickrel, S. G. (2002). Four-year follow-up of multisystemic therapy with substance-abusing and substance-dependent juvenile offenders. Journal of the American Academy of Child & Adolescent Psychiatry41(7), 868-874.

Jackson, K. M., Sher, K. J., & Wood, P. K. (2000). Trajectories of concurrent substance use disorders: A developmental, typological approach to comorbidity. Alcoholism: Clinical and Experimental Research24(6), 902-913.

Jessor, R., & Jessor, S. L. (1978). Theory testing in longitudinal research on marijuana use. In D. B. Kandel (Ed.), Longitudinal research on drug use: Empirical findings and methodological issues (pp. 41–71). Washington, DC: Hemisphere.

Johnston, L.D., O’Malley, P.M., Bachman, J.G., Schulenberg, J.E. (2006). Monitoring the future: National survey results on drug use, 1975–2005: Secondary school students. Bethesda, MD: National Institutes of Health.

Kandel, D. B., & Logan, J. A. (1984). Patterns of drug use from adolescence to young adulthood: I. Periods of risk for initiation, continued use, and discontinuation. American Journal of Public Health74(7), 660-666.

Lanza, S. T., Collins, L. M., Lemmon, D. R., & Schafer, J. L. (2007). PROC LCA: A SAS procedure for latent class analysis. Structural Equation Modeling, 14(4), 671-694.

Lanza, S. T., & Rhoades, B. L. (2013). Latent class analysis: an alternative perspective on subgroup analysis in prevention and treatment. Prevention Science14(2), 157-168.

Le Blanc, M., & Loeber, R. (1998). Developmental criminology updated. Crime and Justice, 23, 115-198.

Loeber, R. (1988). Natural histories of conduct problems, delinquency, and associated substance use: Evidence for developmental progressions. In B.B. Lahey & A.E. Kazdin (Eds.), Advances in clinical child psychology (Vol. 11, pp. 73-124). New York, NY: Plenum Press.

MacLennan, R.N. (1988). Correlation, base-rates, and the predictability of behavior. Personality & Individual Differences, 9(3), 675-684.

Magnusson, D., & Bergman, L.R. (1988). Individual and variable-based approaches to longitudinal research in early risk factors. In M. Rutter (Ed.), Studies of psychosocial risk: The power of longitudinal data (pp. 45–61). New York, NY: Cambridge Press.

Maldonado-Molina, M. M., & Lanza, S. T. (2010). A framework to examine gateway relations in drug use: an application of latent transition analysis. Journal of Drug Issues40(4), 901-924.

Milloy, M. J., Kerr, T., Buxton, J., Montaner, J., & Wood, E. (2009). Methamphetamine use and rates of incarceration among street-involved youth in a Canadian setting: A cross-sectional analysis. Substance Abuse Treatment, Prevention, and Policy4, 17.

Newcomb, M. D., & Bentler, P. M. (1988). Impact of adolescent drug use and social support on problems of young adults: A longitudinal study. Journal of Abnormal Psychology97(1), 64.

Newcomb, M. D., Maddahian, E., & Bentler, P. M. (1986). Risk factors for drug use among adolescents: concurrent and longitudinal analyses. American Journal of Public Health76(5), 525-531.

Patrick, M. E., & O'Malley, P. M. (2016). The epidemiology of substance use among adolescents in the United States. In R. Zucker & S. Brown (Eds.), The Oxford Handbook of Adolescent Substance Abuse (pp 1 – 27), online first.

Peiper, N. C., Ridenour, T. A., Hochwalt, B., & Coyne-Beasley, T. (2016). Overview on prevalence and recent trends in adolescent substance use and abuse. Child and Adolescent Psychiatric Clinics of North America25(3), 349-365.

Plunkett, M., & Mitchell, C. M. (2000). Substance use rates among American Indian adolescents: Regional comparisons with Monitoring the Future high school seniors. Journal of Drug Issues30(3), 575-591.

Raskin-White, H, Loeber, R., & Farrington, D.P. (2008). Substance use, drug dealing, gang membership, and gun carrying and their predictive associations with serious violence and serious theft. In R. Loeber, D.P. Farrington, M. Stouthhamer-Loeber, & H.R. White (Eds.), Violence and serious theft: Development and prediction from childhood to adulthood (pp. 137-166). New York, NY: Routledge.

Schackman, B.R., Rojas, E.G., Gans, J., Falco, M., & Millman, R.B. (2007). Does higher cost mean better quality? Evidence from highly-regarded adolescent drug treatment programs. Substance Abuse Treatment, Prevention, and Policy 23(2), 23-28.

Schneider, A. L. (1990). Deterrence and juvenile crime. New York, NY: Springer.

Shin, S.H., Hong, H.G., & Hazen, A.L. (2010). Childhood sexual abuse and adolescent substance use: A latent class analysis. Drug and Alcohol Dependence 109, 226-235.

Skardhamar, T. (2010). Distinguishing facts and artifacts in group-based modeling. Criminology48(1), 295-320.

Snyder, S. M., & Smith, R. E. (2015). Do physical abuse, depression, and parental substance use influence patterns of substance use among child welfare involved youth? Substance Use & Misuse50(2), 226-235.

Teplin, L. A., Abram, K. M., McClelland, G. M., Dulcan, M. K., & Mericle, A. A. (2002). Psychiatric disorders in youth in juvenile detention. Archives of General Psychiatry59(12), 1133-1143.

Teplin, L. A., Abram, K. M., McClelland, G. M., Mericle, A. A., Dulcan, M. K., & Washburn, J. J. (2006). Psychiatric Disorders of Youth in Detention. Juvenile Justice Bulletin. Office of Juvenile Justice and Delinquency Prevention.

Teplin, L.A., Abram, K.M., Washburn, J.J., Welty, L.J., Hershfield, J.A., & Dulcan, M.K. (2013). The Northwestern Juvenile Project: Overview. OJJDP Bulletin. Washington, DC: US Department of Justice, Office of Justice Programs, Office of Juvenile Justice and Delinquency Prevention.

Thornberry, T. P., Huizinga, D., & Loeber, R. (1995). The prevention of serious delinquency and violence: Implications from the program of research on the causes and correlates of delinquency. In J. C. Howell, B. Krisberg, J. D. Hawkins & J. J. Wilson (Eds.), Sourcebook on serious, violent and chronic juvenile offenders (pp. 213–237). Thousand Oaks, CA: Sage

Torok, M., Darke, S., Kaye, S., & Ross, J. (2011). Conduct disorder as a risk factor for violent victimization and offending among regular illicit drug users. Journal of Drug Issues41, 25-43.

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 Issues24(1), 9-24.

Vanderwaal, C., McBride, D., Terry-McElrath, Y., & Van Buren, H. (2001). Breaking the juvenile drug-crime cycle: A guide for practitioners and policy makers. Retrieved from US Department of Justice, Office of Justice Programs, National Institute of Justice website: http://www.ncjrs.gov/pdffiles1/nij/186156.pdf.

Vaughn, M. G., Freedenthal, S., Jenson, J. M., & Howard, M. O. (2007). Psychiatric symptoms and substance use among juvenile offenders: A latent profile investigation. Criminal Justice and Behavior34(10), 1296-1312.

Vaughn, M. G., & Howard, M. O. (2004). Adolescent substance abuse treatment: A synthesis of controlled evaluations. Research on Social Work Practice14(5), 325-335.

Vaughn, M. G., Ollie, M. T., McMillen, J. C., Scott, L., & Munson, M. (2007). Substance use and abuse among older youth in foster care. Addictive Behaviors32(9), 1929-1935.

Welte, J. W., Zhang, L., & Wieczorek, W. F. (2001). The effects of substance use on specific types of criminal offending in young men. Journal of Research in Crime and Delinquency38, 416-438.

White, A., Chan, G. C., Quek, L. H., Connor, J. P., Saunders, J. B., Baker, P., & Kelly, A. B. (2013). The topography of multiple drug use among adolescent Australians: Findings from the National Drug Strategy Household Survey. Addictive Behaviors38(4), 2068-2073.

TABLES AND FIGURES

Table 1. Comparison of male and female incarcerated offender descriptive statistics.

Females (n = 98)

% (n)/ m (SD)

Males (n = 378)

% (n)/ m (SD)

χ2/t, p, Φ/d

Demographic characteristics

Age

16.07 (1.13)

16.53 (1.31)

t(170.52) = -3.49, p < .01, d = .38

Ethnicity

χ2(1) = 1.1, n.s., Φ = .05

White

62.2% (61)

62.3% (235)

χ2(2) = 5.58, n.s., Φ = .11

Indigenous

28.6% (28)

20.4% (77)

Non-Indigenous Minority

9.2% (9)

17.2% (65)

Risk factors

Alcohol use before 12

23.0% (20)

29.1% (100)

χ2(1) = 1.28, n.s., Φ=.05

Drug use before 12

25.5% (24)

33.7% (117)

χ2(1) = 2.28, n.s., Φ=.07

Kicked out of home

54.2% (45)

45.7% (138)

χ2(1) = 1.90, n.s., Φ=.07

Left home

86.0% (74)

75.4% (230)

χ2(1) = 4.39, p < .05, Φ=.11

Attending school

55.1% (54)

50.5% (191)

χ2(1) = 0.65, n.s., Φ=.04

Living in foster care

34.0% (32)

21.1% (76)

χ2(1) = 6.95, p < .01, Φ=.12

Physical abuse

61.7 (58)

40.6 (152)

χ2(1) = 13.47, p < .001, Φ=.17

Sexual abuse

49.0% (47)

13.7% (51)

χ2(1) = 57.27, p < .001, Φ=.35

Positive self-identity

68.04 (9.44)

69.66 (8.11)

t(460ture of homicide offenseent analysesed in a real of this item (Tough/Weak) resulted in an acceptable Cronbach' ) = 1.49, n.s., d = .18

Alcohol problem - bio parent

63.4% (59)

56.7% (211)

χ2(1) = 1.38, n.s., Φ=.05

Drug problem - bio parent

46.2% (43)

36.9% (137)

χ2(1) = 2.71, n.s., Φ=.08

Note. Φ = phi; d = Cohen’s d; n.s.= not significant (p > .05).

† Levene’s test violated.

Table 2. Association between substance use patterns and other risk factors - Females

Low Versatility (n = 28)

% (n)/ m (SD)

Hallucinogen Users (n = 15)

% (n)/ m (SD)

Street Drug Users (n = 55)

% (n)/ m (SD)

χ2/t, p, Φ/ η2

Demographic characteristics

Age

16.07 (1.33)

15.87 (0.83)

16.13 (1.11)

F(2, 95) = 0.31, n.s., η2 = .01

Ethnicity

χ2(1) = 1.1, n.s., Φ = .05

White

57.1% (16)

73.3% (11)

61.8% (34)

χ2(4) = 4.49, n.s., Φ = .214

Indigenous

25.0% (7)

20.0% (3)

32.7% (18)

Non-Indigenous Minority

17.9% (5)

6.7% (1)

5.5% (3)

Substance use

Alcohol use before 12

17.4% (4)

28.6% (4)

24.0% (12)

χ2(2) = 0.68, n.s., Φ=.09

Drug use before 12

8.0% (2)

26.7% (4)

33.3% (18)

χ2(2) = 5.78, p < .10, Φ=.25

Number of substances used

2.29 (1.01)b,c

4.40 (0.99)a,c

6.98 (1.42)a,b

F(2, 95) = 132.74, p < .001., η2 = .74

Risk factors

Kicked out of home

54.5% (12)

53.8% (7)

54.2% (26)

χ2(2) = 0.00, n.s., Φ=.01

Left home

63.6% (14)

92.3% (12)

94.1% (48)

χ2(2) = 12.39, p < .01, Φ=.38

Attending school

60.7% (117)

40.0% (6)

56.4% (31)

χ2(2) = 1.78, n.s., Φ=.14

Living in foster care

36.0% (9)

40.0% (6)

31.5% (17)

χ2(2) = 0.44, n.s., Φ=.07

Physical abuse

64.3% (18)

53.8% (7)

62.3% (33)

χ2(2) = 0.43, n.s., Φ=.07

Sexual abuse

46.4% (13)

57.1% (8)

48.1% (26)

χ2(2) = 0.46, n.s., Φ=.07

Positive self-identity

69.63 (7.85)

65.97 (6.98)

70.53 (8.19)

F(2, 86) = 1.50, n.s., η2 = .03

Alcohol problem - bio parent

63.0% (17)

78.6% (11)

59.6% (31)

χ2(2) = 1.71, n.s., Φ=.14

Drug problem - bio parent

44.4% (12)

21.4% (3)

53.8% (28)

χ2(2) = 4.71, p < .10, Φ=.23

Note. Φ = phi; η2 = eta squared; n.s.= not significant (p < .05).

a Indicates significantly different from Low Versatility, b indicates significantly different from Hallucinogen Users, c indicates significantly different from Street Drug Users.

Table 3. Association between substance use patterns and other risk factors - Males

Low Versatility (n = 137)

% (n)/ m (SD)

Hallucinogen Users (n = 136)

% (n)/ m (SD)

Street Drug Users (n = 105)

% (n)/ m (SD)

χ2/t, p, Φ/ η2

Demographic characteristics

Age

16.48 (1.44)

16.29 (1.24)c

16.92 (1.12)b

F(2, 369.6) = 3.98, p < .01, η2 = .04

Ethnicity

χ2(1) = 1.1, n.s., Φ = .05

White

55.1% (75)

68.4% (93)

63.8% (67)

χ2(4) = 20.34, p < .001 Φ = .232

Indigenous

16.2% (22)

22.1% (30)

23.8% (25)

Non-Indigenous Minority

28.7% (39)

9.6% (13)

12.4% (13)

Substance use

Alcohol use before 12

13.0% (15)

26.8% (34)

50.0% (51)

χ2(2) = 36.32, p < .001, Φ=.33

Drug use before 12

20.5% (24)

34.4% (43)

47.6% (50)

χ2(2) = 18.23, p < .001, Φ=.23

Number of substances used†

2.04 (0.87)b,c

4.63 (0.95)a,c

7.90 (1.09)a,b

F(2, 325.28) = 1060.77, p < .001., η2 = .85

Risk factors

Kicked out of home

32.7% (32)

45.4% (49)

59.4% (57)

χ2(2) = 13.96, p < .01, Φ=.22

Left home

64.3% (63)

76.1% (83)

85.7% (84)

χ2(2) = 12.18, p < .01, Φ=.20

Attending school

62.0% (85)

50.7% (69)

35.2% (37)

χ2(2) = 17.09, p < .001, Φ=.21

Living in foster care

19.8% (25)

22.9% (30)

20.2% (21)

χ2(2) = 0.43, n.s., Φ=.03

Physical abuse

34.1% (46)

42.5% (57)

46.7% (49)

χ2(2) = 4.19, n.s., Φ=.11

Sexual abuse

12.7% (17)

14.1% (19)

14.6% (15)

χ2(2) = 0.20, n.s., Φ=.02

Positive self-identity

69.18 (9.48)c

68.42 (9.01)

65.82 (9.55)a

F(2, 368) = 4.00, p < .05, η2 = .02

Alcohol problem - bio parent

52.2% (71)

61.2% (82)

56.9% (58)

χ2(2) = 2.22, n.s., Φ=.08

Drug problem - bio parent

32.3% (43)

34.8% (47)

45.6% (47)

χ2(2) = 4.82, p < .10, Φ=.11

Note. Φ = phi; η2 = eta squared; n.s.= not significant (p < .05).

† Levene’s test violated; Brown Forsythe statistics used.

a Indicates significantly different from Low Versatility, b indicates significantly different from Hallucinogen Users, c indicates significantly different from Street Drug Users.

Table 4. Multinomial logistic regression with substance user latent classes as the outcome of interest

Female Model

Male Model

Hallucinogen Users

OR (95% CI)

Street Drug Users

OR (95% CI)

Hallucinogen Users

OR (95% CI)

Street Drug Users

OR (95% CI)

Demographic characteristics

White

3.36 (0.68-16.57)

1.16 (0.40-3.33)

-

-

Age

0.82 (0.45-1.52)

1.11 (0.70-1.78)

0.81 (0.63-1.04)+

1.30 (0.98-1.71)+

Ethnicity

Indigenous

-

-

1.12 (0.49-2.60)

1.35 (0.53-3.33)

Non-Indigenous Minority

-

-

0.32 (0.14-0.75)**

0.48 (0.19-1.18)

Risk factors

Alcohol use before 12

-

-

1.47 (0.63-3.44)

2.65 (1.11-6.32)*

Drug use before 12

14.66 (1.29-167.26)*

12.73 (1.52-106.51)*

1.55 (0.71-3.39)

1.72 (0.74-3.97)

Kicked out of home

-

-

1.46 (0.75-2.83)

2.27 (1.11-4.64)*

Left home

-

-

2.23 (1.11-4.49)*

3.45 (1.48-8.05)**

Attending school

-

-

0.56 (0.29-1.07)+

0.42 (0.21-0.84)*

Positive self-identity

-

-

1.00 (0.97-1.04)

0.96 (0.93-1.00)+

Drug problem - bio parent

1.09 (0.39-3.15)

0.88 (0.43-1.79)

1.36 (0.64-2.89)

Model fit

LL = -76.25, χ2 (8) = 18.99, p < .05

LL = -257.30, χ2 (20) = 86.20, p < .001

Note. The Low Versatility latent class is the reference group to which Hallucinogen Users and Street Drug Users were compared.

+ p < .10, * p < .05, ** p < .01.

[CHART]

Figure 1. Prevalence of each illicit substance examined in the latent class analysis.

Note. * indicates significant (p < .05) differences in prevalence of use of a substance across gender based on chi-square measures of association.

[CHART]

Figure 2a. Three latent class solution of female substance users

[CHART]

Figure 2b. Three latent class solution of male substance users

Acknowledgements

The sample described in this paper is based on research conducted as part of the ongoing Incarcerated Serious and Violent Young Offender Study, initiated in 1998 under the direction of Raymond R. Corrado. The Social Sciences and Humanities Council of Canada supported this work [410-2004-1875]. In addition, the author gratefully acknowledges the assistance of the British Columbia Ministry of Children and Family Development. The views expressed herein are those of the author and do not necessarily reflect the views or policies of the agencies that funded or supported the research. The author would like to thank the anonymous reviewers and Catherine Shaffer for their comments on earlier drafts of this manuscript.

Comments
0
comment
No comments here
Why not start the discussion?