Risk and protective factors for adolescent substance use
Sheckter, Marc Eliot
The objective of this study was to assess the degree to which a set of 17 psychosocial variables could predict group membership of adolescent substance users versus adolescent non-users of substances. Past research has shown each of these 17 variables in isolation to be related to adolescent substance use. However, they have not been studied collectively in a manner allowing prioritization of their predictive strength. Variables that accurately predict substance use are labeled "risk factors", whereas variables that accurately predict non-use are labeled "protective factors". The total sample consisted of 429 central-Saskatchewan adolescents in Grades 8-12. A subgroup of the total sample consisted of youth tested in a Saskatoon residential addictions clinic; this subgroup (Group 3) constituted the clinical sample. Youth who were tested in their schools comprised the non-clinical sample; this sample was split into subgroups of substance users (Group 2) and non-users (Group 1), dependent upon participant responses to questions assessing substance use behaviors. A series of discriminant function analyses (DFAs) were conducted, in an attempt: (1) to predict group membership of non-users (Group 1) versus all users (Groups 2 and 3); and (2) to predict group membership of non-clinical users (Group 2) versus clinical users (Group 3). For the first classification (Group 1 versus Groups 2/3), a set of variables emerged which yielded excellent predictive accuracy, both within and across the groups. Conversely, for the second classification (Group 2 versus Group 3), none of the variables demonstrated predictive strength from either a statistical or clinical perspective. Subsequent to these classifications, two sets of post-hoc DFAs were conducted, in an attempt: (1) to predict group membership of non-users (Group 1) versus non-clinical users (Group 2); and (2) to predict group membership of non-users (Group 1) versus clinical users (Group 3). Both sets of post-hoc DFAs yielded good to excellent correct classification rates. All results are discussed both in terms of susceptibility to problematic substance use, and practical implications for addressing this issue.