Objectives Research suggests that individual and environmental resilience protective factors may be associated with adolescent substance use; however, the associations between a broad range of such factors and use of various types of substances have not been examined. The study aimed to determine the association between a comprehensive range of adolescent individual and environmental resilience protective factors and measures of tobacco, alcohol and illicit substance use.
Design Cross-sectional study.
Setting 32 Australian secondary schools.
Participants Grade 7–10 students (aged 11–17 years).
Measures Data regarding 14 student individual and environmental resilience protective factors and seven substance use measures (tobacco, alcohol, marijuana, other illicit drug use) were obtained via an online self-report survey. Adjusted multivariate logistic regression analyses examined the association between all student resilience protective factors and seven substance use measures.
Results Inverse univariate associations were found for 94 of 98 relationships examined (n=10 092). Multivariate analyses found: consistent inverse associations between 2 of 14 protective factors and all substance use measures (‘goals and aspirations’, ‘prosocial peers’); inverse associations between 4 protective factors with multiple substance use measures (‘home support’ (5 of 7), ‘school support’ (3 of 7), ‘self-awareness’ (2 of 7), ‘community meaningful participation’ (2 of 7)); positive associations between 2 resilience protective factors with multiple measures of substance use (‘community support’ (3 of 7), ‘peer caring relationships’ (5 of 7)) and 6 protective factors not to be associated with any substance use measure.
Conclusions Despite individual relationships between the majority of resilience protective factors and substance use types, the protective benefit of such factors for adolescent substance use was limited to only a small number of such factors when considered collectively. Such results suggest that interventions seeking to reduce adolescent substance use may need to target specific protective factors to address specific types of substance use.
Trial registration number ACTRN12611000606987, Results.
- PREVENTIVE MEDICINE
- PUBLIC HEALTH
This is an Open Access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/
Statistics from Altmetric.com
Strengths and limitations of this study
This study represents the most comprehensive examination of the associations between a broad range of individual and environmental resilience protective factors and multiple measures of adolescent substance use.
Design strengths of the study include: a large sample of adolescents, comprehensive measurement of individual and environmental resilience protective factors using a tool validated in an Australian population, use of multiple accepted measures of substance use, and analyses that accounted for a range of known confounders, potential clustering effects within schools and sensitivity analyses of data with imputation for missing data.
Although the study was reliant on adolescent self-report of substance use and subject to the known limitations of self-report in this population, strategies to increase the validity of adolescent report were used including a web-based survey and confidential participation by students.
While a non-response bias may exist, consistency of results with comparative national data suggests that the likelihood of such bias may be limited.
The cross-sectional design of the study does not allow for investigation of causal pathways of the associations between resilience protective factors and adolescent substance use.
Tobacco, alcohol and illicit substance use are responsible for more than 12% of deaths worldwide1 and cost more than US$600 billion annually in the USA2–5 and US$46.5 billion in Australia.6 Initiation of tobacco, alcohol and illicit substance use in high-income countries primarily occurs during adolescence.7–9 The younger the age of initiation of substance use, the greater the likelihood of ongoing use, dependence and harm in later life.7 ,10–12 In the USA, the UK and Australia, between 23% and 45% of adolescents (aged 11–17 years) have smoked a cigarette,9 ,13 ,14 43–74% have consumed an alcoholic drink,9 ,13 ,14 22–29% have consumed at least five alcoholic drinks on one occasion9 ,13 and between 15% and 40% have taken an illicit substance.9 ,13 ,14 The prevention of substance use among adolescents is a recommended strategy for reducing substance use-related harms throughout the lifecourse.15–17
Historically, research regarding the determinants of adolescent substance use has focused on risk factors such as access to substances, socioeconomic disadvantage and substance use by parents, peers and siblings.18–22 More recent research has explored a range of factors that may be protective of adolescent substance use,23 including individual factors such as self-esteem23–26 and problem-solving ability,27 and environmental factors such as connection to school,23 ,26 ,28–33 family,19 ,23 ,26 ,28 ,34 and prosocial peers.34 ,35 Such factors have been considered to be factors protective of an adolescents' ‘resilience’,25 ,36–41 broadly described as a process, capacity or outcome of successfully adapting to challenging or threatening life circumstances.42–44 As a consequence, enhancement of such protective factors is recommended as a strategy for reducing adolescent substance use.15–17 The specific protective factors to be addressed by such interventions, however, are only broadly defined or are limited to a few examples in such recommendations.15–17
Although considerable research has been reported regarding the association between adolescent resilience protective factors and adolescent substance use,23 ,27–30 ,32 ,45–65 such research using multivariate analyses has only considered a limited number of resilience protective factors (six at most46) or created aggregate scores of such factors,65 with the latter precluding assessment of associations for particular factors. In such studies, inconsistency of findings is apparent in terms of the presence and direction of the associations between resilience protective factors and substance use. For example, adolescents have been reported to be either less, more, or no more or less likely to use a substance if they have low self-esteem,23 ,54 ,63 low school connectedness23 ,28 ,29 ,32 or low academic aspirations.28 ,47 ,50 ,58 ,61 Inconsistency is similarly evident between substances in their reported association with specific protective factors. For example, in one study a significant negative association was reported between educational aspirations and alcohol and marijuana use, but not tobacco use.50 In other studies, significant negative associations have been reported between community involvement and tobacco and marijuana use, but not alcohol use.49 Such contrasting findings between studies may be attributable, in part, to the different measures of such factors across studies, and to the inconsistent inclusion of protective factors.
To date, no peer-reviewed study has reported the associations between a comprehensive range of adolescent individual and environmental resilience protective factors and multiple types of substance use. To address this gap and provide information that may guide future development of interventions targeting adolescent substance use, a study was conducted to determine the association between 14 adolescent individual and environmental resilience protective factors and seven measures of tobacco, alcohol and illicit substance use in a population of Australian adolescents.
Study design and setting
A cross-sectional study was conducted in one Health District of New South Wales, Australia. The District encompasses metropolitan, regional, rural and remote areas with a population of ∼114 000 people aged 10–19 years.66 The data were collected as baseline data for a randomised controlled trial for which ethics and other study approvals were obtained. The methods of the larger study are described in detail elsewhere.67
Participants and recruitment
Eligible schools were either Government or Catholic secondary schools located in a disadvantaged Local Government Area,68 with enrolments in Grades 7–10 (typically aged 12–16 years) on one campus, and with more than 400 total student enrolments. Independent, special needs, selective, central (schools catering for children aged 4–18 years), boarding schools or schools that were not coeducational were ineligible.
Eligible schools were approached for study participation according to a randomly ordered list of schools. If a school declined, the next school was invited to participate until a quota of 32 schools was recruited.
All students enrolled in Grade 7 (usually aged 12–13 years) to Grade 10 (usually aged 15–16 years) in the 32 selected schools were eligible to participate (n=18 310). Parents of students were mailed a study information sheet, a consent form and a reply paid envelope. Two weeks following, non-responding parents were telephoned by school-affiliated staff to prompt return of the consent form.
Data collection procedures
Students with parental consent were invited to complete a self-report anonymous web-based survey in class time (August–November 2011).
Student and school characteristics
Student age, school grade, sex, Aboriginal and Torres Strait Islander status and residential postcode were collected via the student survey.
Students reported tobacco, alcohol, marijuana and other illicit drug use (seven outcomes) via the web-based survey (table 1). The substance use items were sourced from a national triennial survey of school students' health behaviours.9
Resilience protective factors
The Resilience and Youth Development module of the California Healthy Kids Survey, a measure of 14 adolescent individual and environmental resilience protective factors (termed internal and environmental assets), was used to measure protective factors (51 items: 4-point Likert Scale—‘1: Never true’ to ‘4: True all of the time’).69 The survey incorporates items that addressed six individual factor subscales and eight environmental factor subscales (table 1).69 Two minor modifications were made to the survey for use in an Australian population. First, the survey item ‘I plan to go to college…’ was modified to state ‘I plan to go to university or Technical and Further Education …’. Second, the response options for all of the items from the survey were modified from ‘Not at all true, a little true, pretty much true, very much true’ to ‘Never true, true some of the time, true most of the time, true all of the time’. Consistent with reports from the tool developers,69 the data from the current study confirms that the survey tool is an internally consistent and valid measure (Cronbach α coefficients for individual factor subscales: 0.55–0.81; environmental factor subscales: 0.71–0.91). Additionally, confirmatory factor analysis using data from this study demonstrates the individual and environmental subscale factor structure to be a good model fit (Comparative Fit Index 0.92, Standardised Root Mean Square Residual 0.04, Root Mean Square Error of Approximation 0.04, Adjusted Goodness of Fit 0.90); with such results being similar to those reported by the tool developers.69
Student characteristics and substance use
Participants who did not answer any substance use items (ie, they started the online survey but dropped out of the survey before getting to the substance use items) were excluded from all analyses (n=16). Participants who did not answer items for a particular substance were excluded from analyses for that particular substance. Consent and participation rates, demographic and substance use data were examined using descriptive statistics. Socioeconomic status and remoteness of residential location were calculated from student-reported residential postcode using the Australia Bureau of Statistics Socio-Economic Indexes for Areas68 and the Accessibility/Remoteness Index of Australia,70 respectively.
The response options for ‘risky’ alcohol use were collapsed—‘none’ vs ‘once’/‘twice’/‘3–6 times'/‘7 or more times', as were the response options for marijuana and other illicit drug use—‘none’ vs ‘once or twice’/‘3–5 times'/‘6–9 times'/‘10–19 times'/‘20–39 times'/‘40 or more times'.
Differences by sex and grade for each of the seven substance use outcomes were assessed through logistic regression analysis via a Generalised Estimating Equation (GEE) framework71 ,72 to account for potential clustering of students within schools.
Resilience protective factors
Fourteen protective factor scores (six individual factor subscales, eight environmental factor subscales) were created. Protective factor subscale scores were calculated by averaging the responses to all items in a subscale for each student. All such scores ranged from 1 to 4.
Correlation between resilience protective factors
Correlation analysis was undertaken to determine the correlations between all individual and environmental resilience protective factors scores. Pearson correlation coefficients were calculated for each.
Associations between resilience protective factors and substance use
To examine the univariate and multivariate associations between resilience protective factors and student substance use, logistic regression analyses were conducted within a GEE framework71 ,72 to account for potential clustering of students within schools. Individual backward stepwise logistic regression models were conducted for each of the seven substance use outcomes (dependent variables) and each factor measure (independent variables: 14 protective factor subscales) to assess whether a decrease in each mean factor score was associated univariately with substance use (98 models). Multivariate logistic regression analyses explored the association between all individual and environmental protective factor subscales (14 in all, 6 individual and 8 environmental) and the seven substance use outcomes (seven models). In all models, factor score was used as a continuous variable (mean score). All models included potential demographic confounders of substance use, including: school size (400–800 medium, >800 large), school type (government/Catholic school) and student characteristics (sex, grade, remoteness of residential location, socioeconomic and Aboriginal/Torres Strait Islander status). ORs and 95% CIs were calculated for each model. In addition, the odds and probability of use of each substance was derived from the models for specific values of factors (factor score of 2 and 3), in order to calculate the difference in the probability of substance use for a one-unit change in factor score.
Missing data from substance use items were imputed using the recommended method for cross-sectional data in single-item measures—‘hot deck’ imputation.73 Logistic regression analyses were repeated using the imputed data set and any differential results reported.
To account for multiple testing, a criterion for statistical significance of p≤0.0005 was used (Bonferroni-corrected).
Of the 172 eligible secondary schools in the study area, 47 schools were eligible to participate. Across the 32 participating schools (73% school consent rate), parental consent was granted for 13 440 students (73.4%) of which 10 244 students completed at least part of the student survey (participation rate: 55.9% of total enrolled students; 76.2% of students with parental consent). Those students who completed at least one substance use item (n=10 092; 55.1%) are reported in the analysis, the demographic characteristics of whom are shown in table 2.
Twenty-three per cent of students had ever used tobacco and 7% had recently used tobacco, with both forms of use increasing by grade (table 3), with no difference by sex.
For alcohol use, 51%, 14% and 15% of students reported having ever used, recently used or ‘risky’ use of alcohol, respectively. Alcohol use significantly increased by grade across all three measures. A higher proportion of males reported use of each of the three alcohol measures.
Six per cent of students reported recent marijuana use and 3% reported recent other illicit drug use. Marijuana and other illicit drug use increased by grade, with more males reporting use of marijuana and other illicit drugs.
Resilience protective factors
The mean scores for each measure of student resilience protective factors are shown in table 4. The means varied from 2.36 (SD 0.74) to 3.42 (SD 0.75).
Correlation between resilience protective factors
Significant correlations were found between all resilience protective factor subscale scores. Little to weak positive correlations between all individual (0.25–0.53) and environmental resilience protective factor subscales were found (0.26–0.61) (see online supplementary appendix A).
Associations between resilience protective factor scores and substance use
With four exceptions, all measures of substance use were inversely associated with all individual and environmental resilience protective factor subscale scores (table 5).
The final multivariate model for each substance use measure contained between 4 and 10 of the 14 resilience protective factor subscales (table 6). Of the resilience protective factors that remained in each final model, the majority had an inverse association with substance use (table 6).
Across all substance use models, two protective factors were found to have an inverse association with substance use (‘goals and aspirations’ and ‘prosocial peers’). A one-unit decrease in mean subscale score significantly increased the odds of smoking (both measures), having consumed alcohol (all three measures), having used marijuana or another illicit substance by between 1.20 and 1.65 times for ‘goals and aspirations’, and between 2.30 and 3.64 times for ‘prosocial peers’.
Four resilience protective factors were inversely associated with at least one substance use measure (‘home support’ (5 of 7 substance use measures), ‘school support’ (3 of 7), ‘self-awareness’ (2 of 7), ‘community meaningful participation’ (2 of 7)). Two resilience protective factors were found to have a consistent positive association with at least one substance use measure (‘community support’ (3 of 7 substance use measures), ‘peer caring relationships’ (5 of 7)). The remaining six resilience protective factors were not associated with any substance use outcome.
Data for between 2 and 5 variables were missing for 81 of 10 092 participants, with no identified pattern of missingness. Identical analyses using imputation for missing data did not show a differential pattern of results.
This study explored the associations between 14 adolescent individual and environmental resilience protective factors and seven measures of adolescent substance use. Of the 14 factors examined, 6 had an inverse and 2 had a positive association with at least one type of substance use. Of the resilience protective factors found to be inversely associated with substance use, only two were associated with all substance use measures. Such findings suggest that the protective benefit of resilience protective factors for adolescent substance use may be limited to only a small number of such factors and then, primarily, only for some substances.
No previous single peer-reviewed study has reported the associations between a comprehensive range of individual and environmental protective factor measures and a broad range of adolescent substance use measures. Nonetheless, the inverse associations found between eight of the individual and environmental protective factor measures and substance use are generally consistent with the direction of previous studies of single or small numbers of factors and substance use.23 ,29 ,30 ,47–50 ,58 Similarly, the findings of no association between six such factors and substance use are consistent with previous studies.27 ,48 ,64 In contrast, the consistent positive association found between the protective factors of ‘peer caring relationships’ and use of some substances differs from a previous study that has reported no evidence of an association.30 While the reason for such contrast is unknown, it may be at least partly attributable to the different measurement of resilience protective factors and substance use between studies.30 No previous studies could be identified that examined the association between adolescent substance use and ‘community support’. Further research is required to confirm the contrasting findings and, if confirmed, to understand the mechanisms for such an association such that interventions promoting these factors do not have an untoward effect.
The findings that six factors conferred a protective benefit for adolescent substance use appear to align with models of adolescent substance use prevention, such as the social development model (grounded in Bandura's social learning theory and control theory)19 and models of resilience,23 ,25 ,29 ,36 ,40–44 ,75 which focus on individual capacities or assets, and bonding with family, school and peers as protection against the risk factors for substance use. However, despite the consistency in the direction of such associations, there was a lack of consistency of association across types of substances. For example, lower ‘community meaningful participation’ was only associated with tobacco use (ever and recent) and lower ‘school support’ only associated with some measures of tobacco (ever) and alcohol use (ever and risk). Such findings suggest that the protective benefit of such factors is variable across different types of substance use, rather than being generalised. The finding of no association or a positive association between a number of resilience protective factors and substance use, however, may challenge such models of substance use prevention, or at least the breadth of factors described by such models.
The finding of an inverse association between some resilience protective factors and different forms of adolescent substance use suggests that interventions and programmes that seek to reduce such use by increasing resilience protective factors may be more likely to be effective if they target specific protective factors to address specific types of substance use. A number of studies have demonstrated that interventions can successfully increase resilience protective factors among young people,76–79 including individual factors such as self-efficacy,80 and environmental factors related to family and community connection.77 For example, a randomised controlled trial in Hong Kong examining the efficacy of a positive youth development intervention targeting a range of resilience protective factors (including self-efficacy, beliefs in the future, bonding and prosocial involvement) to reduce adolescent substance use, reported an increase in some targeted protective factors (such as self-efficacy and bonding) and decreases in substance use.80 Further research is required to determine whether intervention approaches focused on the enhancement of the specific resilience protective factors found to be associated with adolescent substance use in this study have a beneficial effect on such use.
As previous research has demonstrated that risk factors such as access to substances, substance use by parents and by peers and siblings18–22 are associated with adolescent substance use; further research exploring the association of resilience protective factors with adolescent substance use in the context of such risk factors is warranted. Such research is needed to identify the specific set of individual and environmental resilience protective factors and risk factors that are associated with each type of adolescent substance use, the relative contribution of each factor, and to determine the consistency of association between such factors and different types of substances. While some studies investigating risk and protective factors generally have found risk factors to be stronger predictors of tobacco, alcohol and marijuana use, such studies have only examined a limited number of resilience protective factors and types of substance use.49 ,81 Additionally, future research investigating the potential of such factors for prevention should be theory driven in an effort to understand the aetiology of substance use, and whether this differs by substance use type.
The findings of this study should be viewed in light of a number of the study characteristics. First, the study included a number of design strengths, including: a large sample of adolescents; use of a tool validated in an Australian adolescent population; comprehensive measurement of individual and environmental resilience protective factors; use of multiple accepted measures of substance use and analyses that accounted for a range of known confounders and potential clustering effects within schools. Although the study was reliant on adolescent self-report of substance use and subject to the known limitations of self-report in this population,81 self-report is an accepted method of measuring substance use by adolescents. To optimise validity of report, a web-based survey was used82 as was confidential participation by students, strategies that have been found to increase the validity of adolescent report of sensitive questions, such as substance use.83
Although a non-response bias may exist, the demographic characteristics and prevalence of adolescent substance use found in this study are consistent with those reported in Australian triennial nationwide surveys,9 and suggests that the likelihood of bias may be limited. The conduct of the study in one local health district of New South Wales, Australia, may limit the generalisability of the results to other adolescent populations, and while the imbalance in the proportion of Aboriginal and Torres Strait Islander and Grade 10 students may be indicative of this, the impact is not known. Finally, the study is limited by its cross-sectional design which does not allow for investigation of the causal pathways of the association findings. Further longitudinal and intervention-based research is required to address these questions.
With thanks to the staff and students from participating schools, the Healthy Schools Healthy Futures project team and Christophe Lecathelinais (Hunter New England Population Health) for statistical support. This work was supported by in-kind support from Hunter New England Population Health and the Hunter Medical Research Institute.
Contributors RKH contributed to the design of the data collection tools, monitored the data collection, led conception of the paper, cleaned and analysed the data, drafted and revised the paper. MF, JB and JW contributed to the design of the data collection tools and conception of the paper, monitored the data collection, provided critical revision and approved the final version of the paper. LW, KG and JD contributed to the critical revision of the manuscript and approved the final manuscript. CL provided statistical advice.
Funding This work was supported by the National Health and Medical Research Council of Australia (grant number NHMRC #1016650) and the nib Foundation.
Competing interests None declared.
Ethics approval Human Research Ethics Committee, Hunter New England Health (Ref:09/11/18/4.01) and Human Research Ethics Committee, University of Newcastle (Ref:H-2010-0029).
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement Requests for additional unpublished data should be forwarded to firstname.lastname@example.org
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.