نقش تعامل مدرسه در پیشگیری از بزهکاری نوجوانان مصرف مواد مخدر و: تجزیه و تحلیل برای بقا

Abstract The present study was designed to examine the effects of school engagement on risky behavior in adolescence. Using data from the 4-H Study of Positive Youth Development (PYD), a longitudinal study of U.S. adolescents, discrete-time survival analyses were conducted to assess the effect of behavioral and emotional school engagement on the initiation of drug use and delinquency. The current analyses used seven years of longitudinal data collected from youth and their parents. Results of discrete-time survival analysis indicated that, controlling for demographic variables, higher degrees of behavioral and emotional school engagement predicted a significantly lower risk of substance use and involvement in delinquency. Substance use prevention programs and other health-risk reduction programs should include components (i.e., adolescents' participation in and emotional attachment to school) to capitalize on the protective role of the school context against youth risk behavior.

Results Despite the theoretical burden placed on school engagement in influencing youth development, little is known about whether school engagement protects youth against involvement in risk behaviors. Accordingly, the goal of this research was to ascertain if levels of behavioral and emotional involvement in school were associated with hazards of initiating substance use and engaging in delinquent acts among youth in Grades 5 through 11. Results are presented in three general steps. First, we described the prevalence rates for substance use and delinquency among these youth. Second, we present unconditional survival models (i.e., no predictor variables entered) for each risk behavior; and third, we add demographic variables and school engagement to the survival model for the behavior in question. Prevalence and risk of substance use and delinquency initiation Table 1 displays the percentages of boys and girls reporting substance use (i.e., cigarettes, tobacco/snuff, alcohol, marijuana, and other illicit drugs) and involvement in delinquent acts (i.e., stealing, hitting, beating up someone, getting trouble with police and vandalism) prior to each assessment. Consistent with developmental expectations and findings of other studies (e.g., Costello et al., 2008 and Schwartz et al., 2010), rates of substance use and delinquency generally increased across grades for each substance and for every type of delinquent behavior through the early high school years and then decreased from 9th or 10th grades to 11th grade. This pattern was demonstrated both for boys and girls. The slight drop during the high school grades may be due to participant attrition; it may be that youth with the most problematic behavior dropped out of the study (e.g., see Schaie & Strother, 1968, for a similar conjecture about the influence of selective attrition on the nature of longitudinal trajectories in mental abilities). It is also possible that youth who belonged to the later cohorts were lower in delinquency and drug use compared to students who entered the study earlier (see Schwartz et al., 2010). Table 1. Prevalence of substance use (cigarette, alcohol, marijuana, and other illicit drugs) and delinquency (Stealing, hitting, beating up, getting in trouble with police and vandalism) by gender and grade level. Grade 5 Grade 6 Grade 7 Grade 8 Grade 9 Grade 10 Grade 11 Boys Girls Boys Girls Boys Girls Boys Girls Boys Girls Boys Girls Boys Girls Substance use 20.36% 9.77% 31.82% 28.10% 29.26% 25.92% 38.15% 36.00% 49.86% 39.07% 43.37% 37.27% 42.46% 35.60% Cigarette use 4.32% 1.72% 6.19% 5.89% 9.36% 9.22% 13.27% 12.40% 19.72% 15.02% 19.13% 13.02% 17.38% 9.66% Chewing tobacco 1.76% 0.49% 3.10% 1.22% 5.64% 3.22% 12.78% 3.59% 15.58% 3.39% 17.09% 2.48% 15.33% 2.89% Alcohol use 19.16% 8.98% 21.27% 18.63% 22.28% 20.06% 32.20% 31.87% 43.91% 35.88% 36.23% 32.98% 37.24% 32.52% Marijuana use 1.37% 0.49% 2.60% 2.15% 5.12% 3.87% 7.42% 6.75% 17.23% 12.01% 13.72% 9.80% 14.50% 8.08% Other drugs 1.23% 0.62% 0.91% 1.13% 2.98% 2.25% 4.15% 2.71% 5.65% 4.57% 5.70% 2.55% 3.53% 2.43% Delinquency 48.27% 24.22% 48.23% 23.85% 46.48% 25.81% 55.62% 30.34% 51.65% 27.09% 54.15% 27.24% 42.71% 20.64% Stealing 14.97% 7.55% 11.27% 8.73% 13.95% 9.88% 17.38% 9.07% 18.33% 11.54% 17.41% 10.08% 12.81% 6.63% Hitting 14.42% 3.47% 12.07% 5.54% 13.18% 6.84% 17.56% 8.12% 19.06% 10.07% 18.12% 7.95% 12.63% 6.04% Beating up someone 39.39% 18.20% 40.97% 18.57% 33.45% 16.52% 31.06% 19.21% 30.77% 18.49% 31.66% 14.50% 19.34% 8.03% Getting in trouble with police 11.46% 3.49% 12.22% 5.07% 12.48% 6.52% 31.06% 6.82% 16.62% 7.38% 18.63% 6.84% 11.11% 3.55% Vandalism 23.36% 5.98% 35.73% 9.75% 26.26% 7.55% 35.46% 9.51% 25.63% 6.65% Note: Vandalism was not measured in Grades 5 and 6. Table options According to participants’ self-reports, alcohol was the most frequently used substance, whereas beating up someone else was the most frequently exhibited delinquent act among the adolescents in the current sample. In addition, as shown in Table 1, girls consistently reported lower rates of substance use and of involvement in delinquency compared to boys. When combining rates of use of different types of substances and delinquent behavior, we found that approximately half of the boys and almost 40% of the girls used at least one substance at some point in time prior to the 11th grade assessment. About 55% of the boys and almost one third of the girls were involved in at least one type of delinquent behavior during the course of the study. In other words, with some gender differences, adolescent boys and girls both were exposed to substantial risks to their health or to the health of others, either in the form of substance use or delinquency. Behavioral and emotional school engagement as predictors of substance use initiation Discrete-time survival analysis was used to assess the role of behavioral and emotional school engagement in predicting substance use initiation. Because of sex differences in the prevalence of substance use, we included sex and family SES variables (indexed by mother’s education and per capita family income) as covariates in addition to school engagement variables. To estimate the discrete hazard model, we used PROC LOGISTIC in SAS 9.2 to regress the event indicator on the school engagement and demographic predictors in the reconstructed person-period data set (Singer & Willet, 1993). Youth who had initiated substance use or had been involved in delinquency prior to their participation in the study were excluded from all models for which they were left-censored on the variable of interest (if a student initiated substance use but not delinquency, he or she was still included in the delinquency analyses). Given that some participants joined the study in middle or later adolescence (in high school, for example), the numbers of left-censored cases for delinquency and substance use both were somewhat large (1681 initiated substance use and 2054 initiated delinquency prior to study participation). In testing these discrete-time hazard models, a taxonomy of models were fitted to assess the effects of grade levels (i.e., time), sex, family SES, and school engagement predictors. Fitted models of increasing complexity were built and then compared as a way of assessing the different influences of these predictors. The parameter estimates and corresponding hazard ratios are presented in Table 2. Table 2. Hazard models of the effect of behavioral and emotional engagement on substance use initiation. Variable Model 1 Model 2 Model 3 Model 4 Combined B Average hazard ratios Combined B Average hazard ratios Combined B Average hazard ratios Combined B Average hazard ratios Grade 5 −19.20 0.00 −18.98 0.00 −18.08∗∗∗ 0.00 −15.72∗∗∗ 0.00 Grade 6 −4.52∗∗∗ 0.01 −4.30∗∗∗ 0.01 −3.40∗∗∗ 0.03 −1.10∗∗ 0.33 Grade 7 −3.66∗∗∗ 0.03 −3.43∗∗∗ 0.03 −2.53∗∗∗ 0.08 −0.24 0.80 Grade 8 −3.01∗∗∗ 0.05 −2.78∗∗∗ 0.06 −1.88∗∗∗ 0.16 0.57 1.78 Grade 9 −2.76∗∗∗ 0.06 −2.53∗∗∗ 0.08 −1.61∗∗∗ 0.20 0.63 1.89 Grade 10 −2.27∗∗∗ 0.10 −2.03∗∗∗ 0.13 −1.11∗∗∗ 0.33 1.41∗∗∗ 4.11 Grade 11 −0.97∗∗∗ 0.38 −0.71∗∗∗ 0.49 0.23 1.28 2.86∗∗∗ 17.67 Sex −0.38∗∗∗ 0.68 −0.38∗∗∗ 0.69 −0.18∗ 0.83 Mother education −0.05∗ 0.95 −0.02 0.98 Per capita family income 0.00 1.00 0.00 1.00 Behavioral engagement −0.39∗∗∗ 0.68 Emotional engagement −0.52∗∗∗ 0.60 AIC 5796.70 5773.85 5748.16 5543.85 SC 5852.01 5837.04 5827.17 5638.67 −2LL 5782.70 5757.85 5728.16 5519.85 Note. ∗ p < .05; ∗∗ p < .01; ∗∗∗ p < .001. Table options Unconditional survival model We started by fitting an unconditional survival model for substance use, that is, a survival model that included only the time indicators (one time indicator for each grade). The unconditional hazard function for first use of substances (i.e., tobacco, alcohol, marijuana, and other illicit drugs) is presented in Fig. 1. The survival function takes into account the risk of initiation at each grade to estimate the probability (i.e., the survival function) that a given adolescent will survive (i.e., not initiate use of any substances) through grade 11. Fig. 2, on the other hand, demonstrates the unique risk of initiation at each grade level by presenting the probability that an adolescent initiated use of at least one type of substance, given that he or she had never engaged in substance use at or prior to the time of the Grade 5 assessment. As shown in Fig. 2, a slower increase in the hazard of initiation was observed from Grade 5 to Grade 10 and then a more rapid increase between Grades 10 and 11. Fitted survival functions for onset of substance use and delinquency among ... Fig. 1. Fitted survival functions for onset of substance use and delinquency among adolescents between grade 5 to grade 11. Figure options Hazard functions for onset of substance use and delinquency among youth from ... Fig. 2. Hazard functions for onset of substance use and delinquency among youth from Grade 5 to Grade 11. Figure options Multivariate survival model After testing an unconditional discrete-time survival model for substance use, we added sex, mother education, and average per capita family income as time-invariant covariates, and behavioral and emotional school engagement as time-varying covariates. First, we tested the univariate effect of sex without any other covariates being involved in the analysis. The results indicated that boys were more likely than girls to initiate substance use. The estimated odds of substance use initiation for girls were about one-third lower than the odds of initiation for boys (hazard ratio = 0.68, see Model 2 in Table 2). Fig. 3 presents the fitted hazard functions for boys and girls, computed from the parameter estimates for Model 2 in Table 2. Fitted hazard functions describing the risks of substance use initiation by sex, ... Fig. 3. Fitted hazard functions describing the risks of substance use initiation by sex, from a hazard model containing the main effect of the grade level indicators and the main effect of sex as predictors. Figure options In the second step, we tested another model that included sex, mother’s education, and family income as covariates (Model 3). Girls and youth whose mothers received higher levels of education tended to have lower hazards of substance initiation than boys (hazard ratio = 0.69) and than youth whose mothers had completed fewer years of education (hazard ratio = 0.95). However, per capita family income did not have a significant effect on initiation. Next, we added behavioral and emotional school engagement as time-varying covariates to the model (Model 4). A comparison of goodness-of-fit statistics for this model and Model 3 confirmed that adding the school engagement variables in the model significantly improved prediction of the risk profile. The contribution of mothers’ education and per capita income became nonsignificant with the addition of behavioral and emotional school engagement. Behavioral and emotional school engagement were both significantly associated with decreased odds of initiation of substance use (hazard ratios are 0.68 and 0.60 respectively). After controlling for sex and family SES, the odds of initiating substance use for a given student were about one-third lower than the odds for a student whose behavioral engagement was one point lower. Similarly, the odds of initiating substance use for a given student were 40% lower than the odds for someone whose emotional engagement was one point lower. Behavioral and emotional school engagement as predictors of onset of delinquency Unconditional survival model for delinquency A similar set of models were tested for delinquency, following the same steps used in the analysis for substance use. First, we fitted an unconditional survival model including only the time term. As described earlier, Fig. 1 presents the survival probabilities cumulated for the risk of initiation at each grade, to assess the probability that an adolescent would “survive,” or not become involved in delinquent acts, through Grade 11. As the students grew older, the probability of survival (or not becoming engaged in delinquent acts) declined. In other words, increasingly more students became involved in delinquent behaviors. A similar finding is also illustrated in Fig. 2: as students progressed through adolescence, they faced greater and greater hazards of delinquency onset. Multivariate survival model The unconditional survival model for delinquency was then extended by adding sex as a time-invariant covariate (Model 2). A comparison of goodness-of-fit statistics for this model and the unconditional model confirmed that adding sex in the model significantly improved prediction of the risk profile. Boys were more likely than girls to have been involved in delinquent behaviors from Grade 5 to Grade 11. The hazards for boys to be involved in delinquency were more than twice as high as those for girls (see Table 3). Fig. 4 presents the fitted hazard functions for boys and girls computed from the parameter estimates for Model 2 in Table 3. Table 3. Hazard models of the effect of behavioral and emotional engagement on onset of delinquency. Variable Model 1 Model 2 Model 3 Model 4 Combined B Average hazard ratios Combined B Average hazard ratios Combined B Average hazard ratios Combined B Average hazard ratios Grade 5 −19.2 0.00 −18.65 0.00 −17.35∗∗∗ 0.00 −14.86∗∗∗ 0.00 Grade 6 −4.33∗∗∗ 0.01 −3.78∗∗∗ 0.02 −2.49∗∗∗ 0.08 −0.12 0.89 Grade 7 −3.65∗∗∗ 0.03 −3.10∗∗∗ 0.04 −1.80∗∗∗ 0.17 0.58 1.80 Grade 8 −3.03∗∗∗ 0.05 −2.46∗∗∗ 0.09 −1.15∗∗∗ 0.32 1.39∗∗∗ 4.06 Grade 9 −3.04∗∗∗ 0.05 −2.47∗∗∗ 0.08 −1.15∗∗∗ 0.32 1.18∗∗∗ 3.29 Grade 10 −2.51∗∗∗ 0.08 −1.92∗∗∗ 0.15 −0.58 0.57 1.99∗∗∗ 7.40 Grade 11 −1.54∗∗∗ 0.21 −0.91∗∗∗ 0.40 0.45 1.58 3.11∗∗∗ 22.67 Sex −0.89∗∗∗ 0.41 −0.92∗∗∗ 0.40 −0.76∗∗∗ 0.47 Mother education −0.07∗∗ 0.93 −0.04 0.96 Per capita family income −0.00∗∗∗ 1.00 −0.00∗∗ 1.00 Behavioral engagement −0.37∗∗∗ 0.69 Emotional engagement −0.51∗∗∗ 0.60 AIC 5175.08 5060.51 5004.92 4841.08 SC 5230.07 5123.36 5083.48 4935.35 −2LL 5161.08 5044.51 4984.92 4817.08 Note. ∗ p < .05; ∗∗ p < .01; ∗∗∗ p < .001. Table options Fitted hazard functions describing the risks of delinquency onset by sex, from a ... Fig. 4. Fitted hazard functions describing the risks of delinquency onset by sex, from a hazard model containing the main effect of the grade level indicators and the main effect of sex as predictors. Figure options The next model, which assessed the effect of demographic variables, indicated that sex, mother’s education, and family income were all significant predictors of delinquency initiation. Girls, youth whose mothers had higher levels of education, and youth from wealthier families tended to face lower risk of delinquency initiation than boys and youth from lower-SES families. We then added school engagement variables into the model. Sex and per capita income remained significant as predictors, but mother’s education became nonsignificant. After adjusting for the covariates, the protective effect of behavioral and emotional engagement remained significant. In sum, the addition of school engagement variables significantly improved the precision of risk prediction (see Table 3). With other predictors held constant, the odds for a given student were slightly above two-thirds of the odds for a student whose behavioral engagement was one unit lower. Similarly, the odds for a given student to begin use of one or more substances were a bit over half of the odds for someone whose emotional engagement was one unit lower.