بزهکاری به عنوان یک واسطه از رابطه بین عاطفه منفی و نوجوانان مبتلا به اختلال مصرف الکل
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|38563||2007||19 صفحه PDF||سفارش دهید||8990 کلمه|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Delinquency as a mediator of the relation between negative affectivity and adolescent alcohol use disorder, Volume 32, Issue 12, December 2007, Pages 2747–2765
Abstract This investigation examined mediators of the longitudinal relation between negative affectivity and the development of problematic drinking behavior in adolescent boys and girls. In the present study, 499 early adolescents completed inventories of negative affectivity, attitudes toward delinquency, personal delinquency, and affiliation with delinquent peers. Positive attitudes toward delinquency emerged as the most consistent mediator and strongly predicted drinking frequency in various situations. Compared with personal delinquency, both attitudes toward delinquency and peer delinquency were superior predictors of affect-related drinking. Our results also demonstrated that positive attitudes toward delinquency mediated the relation between negative affectivity and later development of an alcohol use disorder. These findings suggest that a proneness to unpleasant affect impacts adolescent drinking by heightening risk for general rejection of normative behavior, rather than by increasing drinking as a means of managing affect. The importance and implications of testing delinquency variables together in the same model are discussed.
. Introduction Numerous studies have linked the personality variable of negative affectivity, defined as the predisposition to aversive emotional states, to greater amounts of drug and alcohol use in adolescents (e.g. Colder and Chassin, 1993, Krueger et al., 1996, Labouvie et al., 1990 and Shoal and Giancola, 2003). However, efforts to apply traditional negative affect regulation models (Conger, 1956) to adolescent drinking have yielded mixed results (see Shoal & Giancola, 2003 for review). A more powerful and empirically validated predictor of substance use is delinquent behavior. Delinquency has manifested a strong association with substance use across a variety of studies (Brook et al., 1996, Giancola and Parker, 2001 and Kingery et al., 1992). As such, some researchers argue that delinquency, peer delinquency (Chassin, Pillow, Curran, Molina, & Barrera, 1993), or simply positive attitudes toward delinquency (Kaplan, 1980) may be more important proximal predictors of adolescent alcohol use, compared with the potentially more distal effects such as negative affectivity. Kaplan (1980) proposed a specific model of adolescent substance use in which individuals who repeatedly experience negative affect in a given social environment will begin to experience substantial frustration with the parameters of that environment. One response to this frustration might be to reject culturally prescribed values and increasingly embrace more “deviant” attitudes and behaviors in the effort to increase positive reinforcement. Kaplan emphasized that the development of positive attitudes toward deviancy occurs primarily in individuals who lack the instrumental resources (i.e. coping ability) to manage affect in a constructive way within the normative environment. In the final step of the Kaplan's basic model, positive attitudes toward deviance lead to substance use as the individual drinks alcohol or uses other drugs as an expression of a more general deviancy. Models from the criminology literature suggest that attitudes toward delinquency and association with delinquent peers are intimately linked as proximal precursors of adolescent substance use. Differential association theory (Akers, 1977 and Sutherland, 1939) contends that delinquent behavior is learned from close peer groups. According to this theory, the developing individual's exposure to attitudes and motives that promote non-normative behavior are weighed against exposure to factors encouraging more lawful behavior. Moreover, Cairns and Cairns (1994) found that friendships among adolescents are most likely to form between individuals similar on the dimensions of social class, popularity, aggression, and achievement. Once individuals who are already similar in personality and attitudes group together, a type of contagious reciprocity may take effect whereby similarities in behaviors become even more pronounced. As this would predict, one adolescent being in the presence of other adolescents who drink has been shown to escalate both the drinking behavior of the individual and that of the group (Curran, Stice, & Chassin, 1997). As these arguments demonstrate, the effects of peer association and attitudes toward delinquency upon adolescent drinking are multifaceted. Most social scientists believe that the behavioral similarities of group members are a result of a combination between socialization and group selection processes (Reed & Rountree, 1997). What these conceptualizations have in common with Kaplan's (1980) theory is that attitudes favorable toward delinquency and association with delinquent peers are highly influential in determining the extent to which the adolescent becomes involved in alcohol use. As such, attitudes toward delinquency and association with delinquent peers merit examination in models of affect-related adolescent drinking. 1.1. Empirical evidence: personal delinquency and attitudes toward delinquency as mediators Underage alcohol use represents a form of delinquency. As such, tests of negative affect regulation models should address the possibility that drinking is simply one facet of an overall syndrome of non-normative or problematic behaviors. Caspi et al. (1997) demonstrated that adolescents high in negative affectivity are significantly more likely than controls to engage in a broad array of high-risk behaviors including violent offending, sexual risk-taking, dangerous driving, and problematic alcohol use. They found these behaviors to covary significantly, indicating that adolescents high in negative affectivity are prone to engage in clusters of dangerous and antisocial activities as detailed in Jessor and Jessor's (1977) problem behavior theory. Evidence from other longitudinal studies indicates that the experience of negative affect is mediated in its relation to later drug use by general delinquency (Shoal & Giancola, 2003). Likewise, “difficult temperament,” which includes irritability, intense reactions to stimuli, and general negative mood (Thomas & Chess, 1977) appears to be mediated in its relation to later drug use by general antisocial behavior (Giancola & Parker, 2001). Cooper, Wood, Orcutt, and Albino (2003) recently extended this trend of linking problem behaviors as a general syndrome of delinquency by showing covariation between educational underachievement, substance abuse, and risky sexual behavior. Given this evidence, it is important for negative affect regulation investigations to examine the possibility that drinking simply represents one aspect of personal delinquency. An additional possibility that must be considered is that this clustering of delinquent behavior is driven by an underlying acceptance of delinquency. Tolerance of deviant behavior has been shown to be related to quantity and frequency of drinking (Jessor, Graves, Hanson, & Jessor, 1968), and positive attitudes toward delinquency have been shown to mediate the relation between some forms of negative affect (self-derogation) and adolescent substance use (Kaplan, Johnson, & Bailey, 1988). Together, these findings raise the possibility that high negative affectivity simply predisposes the adolescent to be more accepting of delinquency, and this greater acceptance leads to a number of potentially harmful actions, including underage drinking. 1.2. Empirical evidence: affiliation with delinquent peers as a mediator Several studies have revealed that being a part of a delinquent peer group may mediate the relation between personality and adolescent drug use. Affiliation with drug using peers has been shown to mediate the relation between negative affect and adolescent substance use in both cross-sectional (Chassin et al., 1993) and longitudinal investigations (Chassin, Curran, Hussong, & Colder, 1996). In a more specific investigation of the effects of anxiety and anger on drug use among high school students, Swaim, Oetting, Edwards, and Beauvais (1989) found that association with drug using peers fully mediated the relation between affect and drug use. They interpreted their findings as evidence that peer influence is a much more important risk factor for drinking than emotional distress. Interestingly, there is some evidence that affiliating with generally delinquent peers, not just specifically drug using peers, may serve the mediating function in question. For example, Giancola and Parker (2001) found that peer delinquency mediates the relation between difficult temperament (which includes negative mood and intense reactions to stimuli) and drug use later in adolescence. Additionally, Shoal and Giancola (2003) showed that affiliation with delinquent peers mediates the relation between negative affectivity and overall substance use two years later. A significant limitation of this research, however, is that it has not pitted different delinquency variables against one another in order to determine which aspects of delinquency are most strongly related to affect and drinking. Including these other delinquency variables in a model would allow a more specific discrimination of which factors are the most influential in adolescent drinking and which serve the most meaningful mediating function for negative affectivity. 1.3. The present investigation The current study had 2 aims. First, we explored the extent to which positive attitudes toward delinquency, involvement in overall delinquent behavior, and affiliation with delinquent peers mediate the predictive relationship between negative affectivity in early adolescence and drinking in late adolescence (see Fig. 1). It was hypothesized that the group of three delinquency variables would mediate the relation between negative affectivity and drinking. Additionally, because drinking is most directly conceptually related to personal delinquency, it was hypothesized that this variable would exhibit a stronger mediating effect, compared with attitudes toward delinquency and peer delinquency. Mediating role of delinquency on adolescent drinking. Fig. 1. Mediating role of delinquency on adolescent drinking. Figure options Next, we examined the degree to which the variables illustrated in Fig. 1 were related to the actual development of an alcohol use disorder (abuse or dependence) in late adolescence (see Fig. 2). Although drinking quantity and frequency are important indicators of potential problems, they alone do not indicate the degree to which the adolescent is suffering immediate difficulties as a result of drinking. While the incidence of alcohol use disorders in adolescents is low, diagnosis by this age is associated with significant neurocognitive and social problems that extend into adulthood (Brown and Tapert, 2004 and Chung et al., 2005). As such, from a clinical and practical standpoint, diagnosis of alcohol use disorders is where the most important prediction lies. Mediating role of delinquency on adolescent alcohol use disorder. Fig. 2. Mediating role of delinquency on adolescent alcohol use disorder.
نتیجه گیری انگلیسی
Results 3.1. Aim 1: mediating role of delinquency on adolescent drinking 3.1.1. Descriptive statistics and variable distributions Prior to testing the hypothesis that attitudes toward delinquency, actual delinquent behavior, and peer delinquency would mediate the relation between negative affectivity and drinking, the distributions of these variables were examined. All independent variables were found to be within acceptable ranges for skew and kurtosis, and scores on these variables within this sample were similar to those reported elsewhere (Caprara et al., 1985, Ivarsson et al., 2002, Loeber, 1989a, Loeber, 1989b and Loeber et al., 1998). As with the assessment at T3, drinking in the context of unpleasant emotions was somewhat positively skewed, but in this case a square root transformation was successful in reducing this skew to a tolerable level. At T3, 155 (55.1%) of the 282 participants in this study reported at that it is alright to drink alcohol at least sometimes. Two hundred and one (71.5%) reported that at least a few of their friends had consumed alcohol in the last 6 months, with 61 (21.7%) reporting that most or all of their friends drank in the last 6 months. With regard to T4 alcohol use, 178 (63.3%) of the 282 participants reported drinking at least once a month on average in the previous year, and 31 (11.0%) reported drinking on at least 10 occasions in the average month. One hundred and three (36.7%) reported having consumed alcohol in the context of negative emotions. Additional descriptive statistics are presented in Table 1. Table 1. Descriptive data for independent and dependent variables (N = 282) Measure M SD Age (T2) 13.11 .94 Education (T2) 6.85 1.23 SES (T2) 42.16 13.65 Negative affectivity (T2) 45.55 24.51 Attitudes toward Delinquency (T3) 31.14 6.14 — Attitudes toward substance use (T3) 1.63 .63 Peer delinquency (T3) 9.09 8.08 — Peer substance use (T3) 1.39 1.25 Personal delinquency (T3) 3.68 2.99 Drinking frequency (T4) 1.12 1.14 Drinking in the context of negative emotions (T4) 23.91 9.20 Note. ⁎p < .05. ⁎⁎p < .01. Table options 3.1.2. Correlations In order to more thoroughly describe the data, a correlation matrix was computed for all of the variables included in this study. The relations between the variables of interest can be seen in Table 2. Table 2. Pearson-product moment correlations (N = 282) Measure 2 3 4 5 6 7 8 9 1) Negative affectivity (T2) .12⁎ .17⁎⁎ .04 .23⁎⁎ .20⁎⁎ .25⁎⁎ .18⁎⁎ .20⁎⁎ 2) Drinking frequency (T2) .16⁎⁎ .25⁎⁎ .18⁎⁎ .28⁎⁎ .17⁎⁎ .17⁎⁎ .16⁎⁎ 3) Attitudes toward delinquency (T3) .60⁎⁎ .59⁎⁎ .55⁎⁎ .56⁎⁎ .34⁎⁎ .41⁎⁎ 4) Attitudes toward substance use (T3) .35⁎⁎ .49⁎⁎ .33⁎⁎ .27⁎⁎ .26⁎⁎ 5) Peer delinquency (T3) .77⁎⁎ .67⁎⁎ .36⁎⁎ .42⁎⁎ 6) Peer substance use (T3) .58⁎⁎ .36⁎⁎ .38⁎⁎ 7) Delinquent behavior (T3) .36⁎⁎ .34⁎⁎ 8) Drinking frequency (T4) .45⁎⁎ 9) Drinking in the context of unpleasant emotions (T4) Note. ⁎p < .05. ⁎⁎p < .01. Table options 3.1.3. Mediation analysis The aim of this study was to determine whether attitudes toward delinquency, personal delinquency, or affiliation with delinquent peers at T3 (ages 15–17 years) would mediate the relation between negative affectivity at T2 (ages 12–14 years) and drinking at T4 (ages 17–20 years). Baron and Kenny (1986) and Holmbeck (1997) have argued that mediation is demonstrated by identifying significant relations between a) the predictor (negative affectivity) and the proposed mediators (attitudes toward delinquency, peer delinquency, and personal delinquency), b) the predictor and the dependent variable (overall drinking frequency or drinking in the context of unpleasant emotions), and c) the proposed mediators and the dependent variable. Finally, the relation between the predictor and the dependent variable should be substantially decreased following the inclusion of the proposed mediators in the model. Regarding the attitudes toward delinquency mediator, satisfaction of the first condition for mediation was demonstrated by regressing attitudes toward delinquency on negative affectivity (β = .16, p < .01), T2 drinking frequency (p = n.s.), and the demographic variables (p = n.s.). Similarly, personal delinquency was regressed on negative affectivity (β = .23, p < .01), T2 drinking frequency (β = .13, p < .05), and the demographic variables (p = n.s.). Finally, affiliation with delinquent peers was regressed on negative affectivity (β = .20, p < .01), T2 drinking frequency (β = .13, p < .05), and the demographic variables (age: β = .16, p < .01; SES: β = − .14, p < .05). These regression procedures demonstrated that negative affectivity was significantly positively related to attitudes toward delinquency, actual delinquent behavior, and peer delinquency. This approach demonstrated that these relations were not an artifact of the correlation between negative affectivity and T2 drinking. The other three conditions for mediation were tested via separate hierarchical multiple regression analyses using T4 overall drinking and T4 drinking in the context of unpleasant emotions criterion variables. For the equation estimating T4 overall drinking, the demographic variables, T2 negative affectivity, and T2 drinking were entered first. This step demonstrated a significant positive relation between T2 negative affectivity and T4 overall drinking (β = .12, p < .05), above and beyond the variance accounted for by T2 overall drinking (β = .16, p < .01). In the second step, attitudes toward delinquency, personal delinquency, and affiliation with delinquent peers were added simultaneously. Examination of the β estimates indicated that attitudes toward delinquency (β = .14, p < .05) and personal delinquency (β = .16, p < .05) accounted for significant amounts of unique variance in T4 overall drinking. This addition of the delinquency variables to the equation reduced the relation between T2 negative affectivity and T4 overall drinking by 83% and rendered it non-significant (β = .02, p = n.s.). These results are presented in Table 3 and depicted in Fig. 3. It should be noted that numbers in parentheses represent associations between variables prior to the inclusion of the delinquency mediators. Table 3. Mediation analyses for T4 drinking frequency regressed on T2 and T3 variables Model and measure β df R2 ΔR2 F for ΔR2 Model 1 (all vars. at T2) 5, 276 .05 .05⁎ 2.65 Age .02 Education .06 SES − .03 Drinking frequency .16⁎⁎ Negative affectivity .12⁎ Model 2: 8, 273 .18 .13⁎⁎ 14.94 Age .05 Education .06 SES − .02 Drinking frequency (T2) .10 Negative affectivity (T2) .02 Attitudes toward delinquency (T3) .15⁎ Peer delinquency (T3) .14 Personal delinquency (T3) .17⁎ Note. ⁎p < .05. ⁎⁎p < .01. Table options Delinquency variables as mediators of the relation between negative affectivity ... Fig. 3. Delinquency variables as mediators of the relation between negative affectivity and overall drinking frequency, with prior drinking included in the model. Figure options Similarly, for the equation estimating T4 drinking in the context of unpleasant emotions, the demographic variables, T2 negative affectivity, and T2 drinking were entered first. This step demonstrated a significant positive relation between T2 negative affectivity and T4 drinking in the context of unpleasant emotions (β = .18, p < .01), above and beyond the variance accounted for by T2 drinking (β = .14, p < .05). In the second step, attitudes toward delinquency, personal delinquency, and affiliation with delinquent peers were added simultaneously. Examination of the β estimates indicated that attitudes toward delinquency (β = .23, p < .001) and peer delinquency (β = .25, p < .01) accounted for significant amounts of unique variance in T4 drinking in the context of unpleasant emotions. Adding the delinquency variables to the equation reduced the relation between T2 negative affectivity and T4 drinking in the context of negative emotions by 61% and rendered it non-significant (β = .07, p = n.s.). These results are presented in Table 4 and depicted in Fig. 4. Again, the numbers in parentheses represent associations between variables prior to the inclusion of the delinquency mediators. Table 4. Mediation analyses for T4 drinking in the context of unpleasant emotions regressed on T2 and T3 variables Model and measure β df R2 ΔR2 F for ΔR2 Model 1 (all variables at T2) 5, 276 .07 .07⁎⁎ 3.89 Age .05 Education − .11 SES − .02 Drinking frequency .14⁎ Negative affectivity .17⁎⁎ Model 2: 8, 273 .24 .17⁎⁎ 20.46 Age .01 Education − .12 SES .00 Drinking Frequency (T2) .07 Negative affectivity (T2) .08 Attitudes toward delinquency (T3) .23⁎⁎ Peer delinquency (T3) .24⁎⁎ Personal delinquency (T3) .03 Note. ⁎p < .05 ⁎⁎p < .01. Table options Delinquency variables as mediators of the relation between negative affectivity ... Fig. 4. Delinquency variables as mediators of the relation between negative affectivity and drinking in the context of unpleasant emotions, with prior drinking included in the model. Figure options It should be noted that the mediation effects noted above pertain to general delinquency T3 variables rather than delinquency variables more specific to substance use. The possibility that T3 attitudes specific to substance use or T3 peer substance use would better mediate the relation between T2 negative affectivity and T4 drinking behavior was also investigated. The first step in testing this possibility was an examination of the correlation table from the previous set of analyses. Because T2 negative affectivity showed no zero-order correlation with T3 attitudes toward substance use, attitudes specifically toward substance use were removed from consideration as a possible mediator between negative affectivity and drinking. However, the significant zero-order correlation between T2 negative affectivity and T3 peer substance use prompted a further analysis. T3 peer substance use was simultaneously regressed on T2 peer substance use (β = .37, p < .01), T2 participant drinking (β = .14, p < .05), the demographic variables (p's = n.s.), and negative affectivity (β = .14, p < .05). This demonstrated a significant relation between T2 negative affectivity and T3 peer substance use, even when accounting for prior peer and participant substance use. The hierarchical analyses to test the mediating effect of T3 peer substance use on the relationships between T2 negative affectivity and T4 Drinking Frequency and between T2 negative affectivity and T4 Drinking in the Context of Unpleasant Emotions were the same as previously described. The only difference was that T2 peer substance use was included as a control in step one and T3 peer substance use was entered alone in the second step instead of the previously examined more general delinquency variables. For the equation estimating T4 Drinking Frequency, examination of the β estimates after this second step indicated that T3 peer substance use accounted for a significant amount of unique variance in T4 drinking frequency (β = .32, p < .01) and reduced the relation between T2 negative affectivity by 58% (from β = .12, p < .05 to β = .05, p = n.s.), rendering it non-significant. Likewise, for the equation estimating T4 Drinking in the context of unpleasant emotions, examination of the β estimates after this second step indicated that T3 peer substance use accounted for a significant amount of unique variance in T4 drinking in the context of unpleasant emotions (β = .35, p < .01) and reduced the relation between T2 negative affectivity by 35% (from β = .17, p < .01 to β = .11, p = .049), which approached non-significance. The results for the peer drinking specific mediation models are presented in Table 5 and Table 6. Table 5. Mediation analyses for T4 drinking frequency regressed on T2 and T3 variables Model and measure β df R2 ΔR2 F for ΔR2 Model 1 (all variables at T2) 5, 276 .05 .05⁎ 2.58 Age .05 Education .07 SES − .04 Drinking frequency .13⁎ Negative affectivity .12⁎ Peer substance use .10 Model 2: 6, 275 .14 .09⁎⁎ 26.86 Age .05 Education .06 SES − .02 Drinking frequency .09 Negative affectivity .05 Peer substance use (T2) − .03 Peer substance use (T3) .33⁎⁎ Note. ⁎p < .05. ⁎⁎p < .01. All variables measured at T2 unless otherwise indicated. Table options Table 6. Mediation analyses for T4 drinking in the context of unpleasant emotions regressed on T2 and T3 variables Model and measure β df R2 ΔR2 F for ΔR2 Model 1 (all variables at T2) 5, 276 .07 .07⁎⁎ 3.45 Age .03 Education − .11 SES −.02 Drinking frequency .12⁎ Negative affectivity .17⁎⁎ Peer substance use .08 Model 2: 6, 275 .18 .11⁎⁎ 35.12 Age .08 Education − .13 SES .00 Drinking frequency .06 Negative affectivity .11⁎ Peer substance use (T2) .06 Peer substance use (T3) .35⁎⁎ Note. ⁎p < .05. ⁎⁎p < .01. All variables measured at T2 unless otherwise indicated. Table options 3.2. Aim 2: mediating role of delinquency on the development of alcohol use disorder 3.2.1. Descriptive statistics Inasmuch as the sample utilized to address Aim 2 was the same as the sample for Aim 1, Section 3.1.1 can be reviewed for descriptive information regarding the independent variables. The difference for the present aim was that the dependent variable of interest was diagnosis of having an alcohol use disorder by age 17 to 19 years. By this age, 27 (9.5%) of the participants met criteria for alcohol abuse and another 20 (7.1%) met criteria for alcohol dependence. Thus, 47 participants (16.6%) met DSM-III-R criteria for an alcohol use disorder. Although this percentage is in the same range as reported in other studies of individuals in late adolescents using the same set of criteria (Cohen et al., 1993), it is slightly higher. This is most likely due to the fact that individuals with family history of a substance use disorder were oversampled in the CEDAR project. 3.2.2. Logistic regression analyses Binomial logistic regression allows examination of how well a set of independent variables predict the presence or absence of particular characteristic, in this case a diagnosis of alcohol abuse or dependence by age 17–20. This is done by examining the extent to which an equation containing a set of independent variables correctly classifies individuals who will manifest the disorder and distinguishes them from those who will not. Because it is not realistic to expect a model to predict a high proportion of individuals with a disorder that is manifest in under 10% of the sample, rates of alcohol abuse and alcohol dependence were combined into a single alcohol use disorder variable. This resulted in a dichotomous dependent variable that was “affirmative” for 16.6% of the sample. Unlike Ordinary Least Square regression, logistic regression does not assume linearity of relationship between the independent variables and the dependent, does not require normally distributed variables, does not assume homoscedasticity, and in general has less stringent requirements. Examination of the chi-square statistic functions as an indicator of the model's fit or appropriateness; Nagelkerke's R2 roughly indicates amount of variance in the dependent variable that is accounted for by the independent variables currently in the model; and odds ratios and the Wald statistic reveal the significance of individual independent variables in predicting the dichotomous dependent variable. Examination of a classification table reveals the percent of cases correctly and incorrectly classified by the variables included in each equation. Independent variables for inclusion in the regression procedure were selected based upon whether they were found to be associated with drinking in the previous analyses for Aim 1. These variables were included in a stepwise fashion based upon the previously forwarded and supported hypothesis that delinquency variables will mediate the relation between negative affectivity and drinking. In the first step, the demographic variables of age, education, and SES were entered. Examination of the chi-square and Negelkerke's R2 values after this first step indicated a poor fit and a weak association between demographic variables and alcohol use disorder diagnosis. None of the 47 cases of alcohol use disorder were correctly classified after this step. In the second step, T2 alcohol use frequency, negative affectivity, and constructive coping were entered to determine whether or not they enhance prediction of an alcohol use disorder. The chi-square value for the model improved substantially (from 3.99 to 21.76, p < .001) and became significant (p < .01). After this step, 85% of the cases were correctly classified, with 12 (25.9%) of the observed positives (individuals diagnosed with alcohol use disorder) predicted as such. At this point, odds ratios were examined to determine the extent to which individual independent variables contribute to the ability of the model to classify cases. The odds ratio represents the factor by which the odds of being diagnosed with an alcohol use disorder changes (i.e. is multiplied by) with a 1 unit change in the predictor variable. Examination of Table 7 reveals that T2 drinking frequency and negative affectivity were positively associated with alcohol use disorder diagnosis, while constructive coping ability was negatively associated with alcohol use disorder diagnosis. Table 7. Logistic regression on T4 alcohol use disorder Model and measure χ2 Δχ2 N. R2 % correct Exp (B) Model 1: (all variables at T2) .377 .377 .00 81.21 Age .95 Education 1.05 SES .99 Model 2: (all variables at T2) 27.24⁎⁎ 26.86⁎⁎ .16 85.00 Age .89 Education 1.12 SES 1.00 Drinking frequency 2.92⁎⁎ Negative affectivity 1.05⁎ Constructive coping .96⁎⁎ Model 3: (including T3 variables) 62.06⁎⁎ 34.81⁎⁎ .33 86.66 Age .68 Education 1.14 SES 1.00 Drinking frequency 2.44⁎⁎ Negative affectivity 1.01 Constructive coping .97 Attitudes toward Del. (T3) 1.12⁎⁎ Peer delinquency (T3) 1.02 Personal delinquency (T3) 1.13 Note. ⁎p < .05. ⁎⁎p < .01; N. R2 = Negelkerke's R2. Table options In the third and final step, T3 attitudes toward delinquency, personal delinquency, and peer delinquency were entered to examine their impact as mediators. This step resulted in another substantial increase in Chi-square (from 27.2 to 62.1, p < .001) indicating an improvement in goodness of fit. Overall correct classification was 87%, with 17 (36%) of the 47 observed positives predicted as such. Only 10 (4%) of the 235 individuals who were not diagnosed as having an alcohol use disorder where predicted to have one. Examination of final odds ratios with all variables in the model revealed that T2 drinking frequency remained a strong predictor of T4 alcohol use disorder (OR = 2.44; p < .001). T3 attitudes toward delinquency was also a significant predictor of T4 alcohol use disorder diagnosis (OR = 1.12; p < .001), but T3 personal delinquency and T3 peer delinquency were not. It is noteworthy that with the T3 delinquency variables included in the model, the association between negative affectivity and alcohol use disorder and between constructive coping and alcohol use disorder became insignificant. These results are presented in Table 7.