The importance of personal mentoring relationships is consistent with the literature on positive youth development, which sees building adolescent competence and self-image as core aspects of health promotion. In fact, studies suggest adult mentoring might reduce risky health behaviors in part through changes in adolescent self-concept. Education policy-makers have explored structural aspects of a school environment, such as smaller schools and lower student load on teachers, that might foster better teacher-student relationships and ultimately raise school performance.38 Such strategies may also prove useful in impacting student health by supporting behavioral self-concept and reducing risky health behaviors. In addition, preserving or even expanding opportunities for students to engage with adults through sports programs and other extracurricular activities might further support healthy adolescent self-concept. Child advocates, including pediatricians and educators, might work together to buttress these school-related assets, particularly during times of economic crisis and intensifying pressure on schools to narrow their focus. This study is limited by the cross-sectional design, which makes it impossible to determine whether factors present in the social networks precede self concept or are causally related to self-concept. While it is conceivable that self-concept impacts relationships with adults at school rather than vice versa, we found that self-concept mediated the associations between relationships with teachers and coaches and substance use much more strongly than teacher or coach relationships mediated the associations between self-concept and substance use, cannabis grow equipment suggesting that reverse causality is less likely. Further, our analysis controls for academic performance, which would likely mediate a causal pathway from self-concept to perceived teacher support or relationships with coaches.

Given that relationships with coaches was not consistently associated with substance use, it is possible that the association between number of coaches named in the network and marijuana use is spurious, and should be confirmed in future studies. The large proportion of charter school students in our sample may have led to less exposure to coaches, as some of the sample schools do not support large athletics programs, and it is possible that the ability of coaches to change health behaviors depends on unmeasured factors, including strength and size of the athletic programs, cultural emphasis on sports in schools, and the visibility of coaches on campus. In addition, although our analysis controls for many likely confounders, we cannot exclude the possibility that there remains unmeasured confounding variables. In particular, we could not account for personality traits or social status, which might influence self-concept, relationships with adults, and substance use. In addition, all data are self reported and potentially biased by social desirability. Finally, given the largely low-income, Latino sample, from a single urban center, our findings might not generalize to other sociodemographic, ethnic, or regional populations.The Multimodal Treatment Study of Children with Attention-Deficit/Hyperactivity Disorder , known as the MTA, was initiated in 1994 as a randomized clinical trial designed to evaluate effects of intensive treatments provided by state-of-the-art protocols with and without stimulant medication. After baseline diagnosis of 579 children with ADHD-Combined Type, cases were randomly assigned to treatment-by-protocol for medication management , Behavior Modification , or their combination provided by MTA staff for 14-months, or assigned to treatment-as-usual obtained from non-MTA clinicians for a Community-Comparison treatment. On the primary outcome measure of the study the groups with medication provided by protocol showed greater benefits than the other two groups .

Ten months after the end of treatment-by-protocol in the 14-month RCT, the MTA transitioned into an observational long-term follow-up study. At this point , 289 randomly selected classmates were added as a local normative comparison group . During the LTF, both groups were assessed 3, 6, 8, 10, 12, 14, and 16 years after baseline. The LTF was intended to track the developmental course of ADHD from childhood to adulthood, to monitor extended use of stimulant medication provided by non-randomized, self-selected treatment in community settings, to perform exploratory analyses and generate hypotheses about possible long-term effects of stimulant medication, and to evaluate outcomes in early adulthood. The initial findings the LTF showed that the relative benefits of medication in the RCT dissipated during childhood , and subsequent findings indicated relative benefits of medication did not reappear in adolescence . A recent report indicated the no significant residual benefits of extended self-selected treatment-asusual with stimulant medication on symptom severity in adulthood . In addition to ADHD symptom severity, outcomes related to substance use have been reported. The SU outcome measure was based on the Substance Use Questionnaire that was administered in childhood , adolescence , and adulthood . The SUQ provided a quantitative measure of SU , which revealed important ADHD-LNCG differences at each developmental stage, including earlier emergence and greater continuation of SU in the ADHD group compared to the LNCG. However, these analyses did not detect significant effects of medication on SU outcomes based on comparisons of the randomized treatments of the RCT or the self-selected patterns of extended treatment in the LTF. The Qualitative Interview Study was developed for further exploration of SU in the MTA. Innovative assessment methods were used based on the Ecocultural Family Interview approach adapted for the MTA .

The M-EFI was an openended, conversational interview with the participants in adulthood. It provided unconstrained narrative accounts reflecting perceptions, attitudes, and opinions about 11 topics : General Functioning, ADHD, SU, Work, Future, Family, Peers, School, Turning Points, Self-knowledge, and Conclusion. Also, an innovative design was used to focus on a subset of the full MTA sample, with enrichment by strategic selection of cases with early emergence of SU in adolescence and continued manifestation of SU in adulthood . The cases with a history of Persistent SU were expected to comprise a small proportion of the longitudinal MTA sample, but this subgroup has been described as theoretically important and associated with increased major health risks . Thus, the QIS was an exploratory study of SU and ADHD, using an innovative assessment of outcome and an innovative design . These aspects of the QIS were intended to broaden previous evaluations and to explore additional associations of SU with ADHD. This paper has four purposes. The first two are methodological and use data from the standard MTA assessment battery: to provide details about the origin of the QIS sample and the selection of ADHD and LNCG cases with histories of Persistent SU and Non Persistent SU and to compare the strategically selected QIS subgroups on demographic factors, treatment histories, and frequency of use of different substances in adulthood. The other two purposes are empirical and use qualitative data from the M-EFI: to compare the Diagnostic groups and the Persistence groups on self-assessment of Frequent use expressed in the unconstrained narratives and to describe and evaluate the Reasons for use of different substances mentioned spontaneously during the M-EFI.There were several steps in the formation of the QIS sample. First, the study was conducted when the MTA cases were in adulthood, which restricted recruitment to the cases retained up to this point. Second, only four of the original sites collected data for the QIS add-on study, vertical grow rack which excluded participants from the other 3 sites . Third, strategic recruitment utilized the SUQ data from the MTA assessments conducted at 8 assessment points of the LTF to define SU history integrated across multiple assessment points and across multiple substances. Developmentally sensitive thresholds were used to define SU appropriately for the range of ages of the participants at assessment points over the course of the LTF . This provided a target for strategic selection of cases with histories of Persistent SIU, defined as any SU by early adolescence , monthly SU during adolescence , and weekly SU in adulthood . Because a small percentage of the MTA cases were expected to meet these rigorous criteria, all ADHD and LNCG participants who met these criteria for Persistent SU were recruited to enrich the QIS sample.The remaining MTA participants were assumed to have a history of no SU or SU restricted to some but not all developmental stages.

Random selection from these ADHD and LNCG cases was used to establish the Non-Persistent SU subgroups with about twice the number of cases as those strategically selected for the Persistent SU subgroups. Weisner et al described some general demographic characteristics of the QIS sample , and found no significant differences between the ADHD group and LNCG . To provide additional context for interpretation of the findings from the QIS study, further evaluation is provided for additional demographic variables shown in epidemiologic studies to be associated with SU , as well as a refinement of the evaluation of Sex and Race/Ethnicity provided by Weisner et al . In addition to comparisons based the full MTA sample for Diagnosis in childhood , comparisons were made of the successive subsets of cases from the full MTA sample based on retention into adulthood, site participation in data collection, and selection by strategic and random processes. This produced 4 subgroups , which were compared on the 5 key demographic variables to assess possible confounding in the QIS sample.As shown in Table 2-A, at recruitment of the full MTA sample, the ADHD group and the LNCG did not differ in percentage of male and female participants, as expected due to group-matching at baseline. The recruitment of the LNCG participants from the same schools as the ADHD cases was intended to match the groups on other demographic factors, but random selection of a small subset of the volunteers resulted in some differences: the LNCG compared to the ADHD group had a slightly but significantly lower percentage of households with public assistance and higher percentage of households with socio-economic advantages . Both of these differences are consistent with the hypothesis that the ADHD group had increased risk for SU relative to the LNCG. However, other hypotheses could be proposed, such as that classmate volunteers for the LNCG had decreased risk for SU compared to the non-volunteers in the schools where random selection was used to form the LNCG. As shown in Table 2-B, retention of the MTA sample was high, with 476 ADHD and 267 LNCG cases having at least one observation in adulthood . Significant differences associated with retention were observed, which were similar to those reported by Howard et al. for comparisons of cases with complete and incomplete observations for the full ADHD group and full LNCG. The retained compared to non-retained subgroup had a higher percentage of female participants and higher percentage of households with high parental education and socioeconomic advantages, and a lower percentage of participants from households with public assistance and from racial/ethnic minorities. These differences suggest retention was associated with protection against SU.As shown in Table 2-C, 325 ADHD and 159 LNCG cases were retained in adulthood from the 4 QIS sites. Compared to the 3 non-QIS sites, these retained cases had a higher percentages of households with high parental education and socioeconomic advantages, and a lower percentage of households with public assistance and non-white racial/ethnic status. These differences suggest site participation was associated with protection against SU. In summary, ADHD diagnosis was associated with increased sociodemographic risk for SU, while participant retention in the MTA and site participation in QIS-data collection were associated with decreased sociodemographic risk.As shown in Table 3-A, statistical comparisons of the ADHD group and LNCG revealed that in the retained cases available for the QIS study, Diagnosis of ADHD in childhood continued to be associated with household characteristics conveying risk for SU . As shown in Table 3-B, the strategic selection of cases with Persistent SU identified a small proportion of the sample . By the QIS definition, the remaining cases had Non-Persistent SU, constituting a large proportion of the sample . Random selection was employed to identify and recruit a subset of these Non Persistent SU cases for the QIS subgroups . Thus, in the available sample, the ratio of Non-Persistent to Persistent SU cases was about 7.50 to 1, but by recruiting all of the cases with Persistent SU and only a subset of the cases with Non-Persistent SU, the ratio was reduced to 2.55 to 1. The resulted in about a 3-fold enrichment of Persistent SU in the QIS sample. To evaluate the main effect of Diagnosis, the ADHD group and the LNCG were compared on the 5 key demographic variables. In these cases selected for the QIS sample, the demographic differences manifested in the larger available groups remained significant: the ADHD group had a significantly lower percentage of cases from households with socio-economic advantages and higher percentage with public assistance.