A universal harm-minimisation approach to preventing psychostimulant and cannabis use in adolescents: a cluster randomised controlled trial

Twenty-one schools with existing relationships with the researchers agreed to participate in the study. Written informed consent was required from students, their parent/guardians and teachers to participate in the study. Eleven schools were randomly allocated to the Climate Schools intervention group and ten schools were assigned to the control group. Written informed consent was provided by parents of 1839 Year 10 students, and students were required to provide informed consent themselves to participate. The final sample consisted of 1734 students (n = 906 Climate Schools intervention; n = 828 Control) at baseline. The sample had a mean age of 15.44 years (SD = 0.41) and 66.2% were male.

The Climate Schools: Psychostimulant and Cannabis Module is a six-lesson program aimed at decreasing cannabis and psychostimulant use and related harms. Each lesson is approximately 40-minutes long and consists of a 20-minute computer component, completed individually by students, followed by 20-minutes of teacher-delivered class activities. Consistent with the theoretical underpinnings of the existing Climate Schools modules, the program content was based on a social influence approach [24]. Full details of the topics covered in each lesson are listed in Table 1. All teachers were given a Teacher Manual which provided guidelines for implementing the program, but no formal training was given. The manual also included teacher and student summaries and five pre-planned classroom activities per lesson. Teachers were able to choose which activities to implement to ensure the activity suited the needs of their class.

All students completed a 40-minute self-report questionnaire at baseline, immediately post-intervention, and at a 5- and 10-month follow-up. An overview of the intervention and assessment times is presented in Table 2. Student data was linked across time using a unique identification code based on easily remembered fragments of personal information, adapted from the School Health and Alcohol Harm Reduction Project (SHAHRP) [25]. Research staff administering the survey followed a standardised protocol and emphasised the confidential nature of the survey to students to encourage honest responding.

Attrition and differential attrition between groups were examined utilising the software program SPSS through a series of Analysis of Variance (ANOVA) for normally distributed data, Chi-squared and logistic regression for binomial and categorical data and Mann–Whitney U and Kruskal-Wallis for non-normally distributed data. To assess the effectiveness of the intervention, analyses were conducted on an Intention-to-Treat basis, that is, all students with baseline data were included in the analysis irrespective of the number of program lessons they completed. The HLM procedure used for the analyses accounted for missing data. This approach makes estimations based on available data points, meaning there was no need to remove students who have incomplete survey data at a given time point.

To determine the level of variability between schools, intraclass correlations (ICCs) were assessed on all outcomes. To account for clustering at the school level, hierarchical linear modelling (HLM) utilising HLM 6 [28, 29] was conducted for normally distributed data (psychostimulant knowledge) while hierarchical generalised linear modelling (HGLM) using Poisson sampling was used for count data (lifetime use of cannabis and meth/amphetamine, frequency of cannabis, ecstasy and meth/amphetamine use and intention to use cannabis). All outcome variables were centred at post-test, allowing comparison between groups immediately after the completion of the intervention. For each outcome, intervention effects were explored in models which utilized linear and quadratic growth terms to characterise the pattern of change on the outcome over time. As recommended by Lee [30], HLM / HGLM procedures were abandoned in favour of single-level analyses (ANCOVA, logistic regression and hierarchical regression utilising SPSS) when the unconditional hierarchical model revealed that less than 10% of systematic variance existed at the between-school level. The outcome variables with ICCs below 10% were cannabis knowledge, cannabis and psychostimulant attitudes, lifetime use of ecstasy and intention to use meth/amphetamine and ecstasy in the next 12 months. For each outcome variable, to minimise the effects of loss to follow-up, three separate analyses were conducted to assess if there was a significant difference between the groups from baseline to each follow-up occasion. Bonferroni adjustments were made for multiple comparisons to control the type 1 error rate. That is, to adjust for the fact that separate analyses were conducted for baseline to immediate post-test, baseline to the 5-month follow-up and baseline to 10-month follow-up. Corresponding baseline scores were entered as a covariate into the models to account for differences between the intervention and control groups on outcomes at baseline. Although there is some evidence of converging rates of cannabis use between young males and females [31], considerable evidence suggests that males are more likely to use drugs than females. For this reason, in all analyses directly examining drug use behaviour, gender was also taken into account.

A total of 1734 Year 10 students from 21 Independent and Catholic High Schools completed the survey at baseline (mean age = 15.44 years, SD = 0.41; 66.2% male). Of the baseline sample, 69.7% completed the survey immediately post-test, 61.4% provided data at five month follow-up and 56.1% were present at the 10-month follow-up. A total of 651 students (37.5%) had data for all four survey occasions. The number and percentage of students over time is presented in Figure 1.

Despite randomisation, at baseline, students in the intervention and the control groups differed on a number of outcome variables. Compared to the intervention group, the control group had greater pro-drug attitudes about psychostimulants (F(1, 1727) = 8.20, p < 0.01), and cannabis (F(1,1730) = 22.45, p < 0.001), greater lifetime cannabis use (χ2(1) = 18.47, p < 0.001), greater frequency of cannabis use (F(1, 1720) = 9.57, p = 0.002) and had greater intention to use cannabis (F(1,1729) = 21.77, p < 0.0001), meth/amphetamine F(1, 1728) = 9.48, p = 0.002 and ecstasy F(1,1727) = 5.77, p = 0.02 in the next 12 months.

Attrition analyses were conducted to assess the comparability of those students who were present at baseline only (single), to those who were present at baseline and at least one further survey occasion (repeat) on outcomes measures. Table 3 indicates that students present only at baseline reported significantly higher levels of cannabis and psychostimulant use, greater intentions to use these substances and greater pro-drug attitudes. Attrition resulted from a number of different factors including one school withdrawing after the baseline survey due to changes in staff, time constraints prohibiting certain classes from completing the post-intervention survey, students being absent on the day of the survey, students no longer attending the school at the time of the survey, students failing to maintain the same identifier code, and students whose surveys were excluded as they had failed to include unique identifier codes or had not answered the questions.

Differential attrition analyses were conducted to assess if there were any differences between the intervention and control condition in the students who were retained beyond baseline (repeat) and those that were lost to follow-up (single). On all outcome variables, there was no evidence of differential attrition.

Tables 4, 5, 6 and 7 provide the descriptive statistics for all outcome variables over the four survey occasions for the intervention and the control groups.

A total of 749 students and 34 teachers completed an evaluation survey about the Climate Schools course. Feedback indicated that the program was well received and enjoyed by both teachers and students. The majority of students (71%) indicated that the cartoon story was an enjoyable way of learning and 63% said that they would like to learn other health theory topics in this way. Eighty-seven per cent of teachers reported that students could recall the information taught in the computer component ‘very well’ or ‘well’. In terms of implementing a computer-based program, the vast majority of teachers reported that it was ‘very easy’ (45%) or ‘easy’ (42%) to implement. Although most teachers did not have problems accessing computer resources in their schools, some teachers (23%) did report difficulties. Ninety-seven per cent of teachers said that they would be likely to use the program in the future and 94% reported that they would be likely to recommend it to other teachers. Teachers varied widely in which classroom activities, and how many activities, they completed with their class. Over eighty per cent of teachers rated the educational quality of the activities positively and thought the activities were helpful in reinforcing information about psychostimulants and cannabis to students.

Eight schools in the control condition provided details of the content and timing of the cannabis and psychostimulant education that was provided to their students. Three schools delivered a social influence program based on a harm-minimisation approach, one school delivered a harm-minimisation program and four schools did not cover psychostimulants as a topic at all. The number of lessons spent on cannabis and psychostimulant education varied, ranging from eight to 13.5 lessons.

The Climate Schools program was shown to increase cannabis and psychostimulant-related knowledge and decrease pro-drug attitudes. Immediately after the intervention, students who received the Climate Schools program had significantly higher levels of knowledge than the control group. The absolute difference in knowledge between groups did diminish over time, but was still significantly different 10 months post-intervention. The capacity to positively impact on knowledge and attitudes is consistent with a large body of previous research on school-based drug prevention [35‐38].

These results directly challenge the notion that harm-minimisation information is too complex for young people to learn, a justification used to support the teaching of simpler abstinence based messages alone [39]. Specifically, consistent with a harm-minimisation approach, the knowledge taught in the current module was to strongly encourage young people to refrain from drug use, but it also provided the knowledge required to practice harm-minimisation skills to prevent harms as a result of one’s own or other people’s drug use. The demonstrated increase in knowledge in the intervention group provides evidence that young people can learn harm-minimisation information. These findings are consistent with previous Climate Schools research [18‐21] and the SHAHRP program which was also conducted in Australia [40]. The heartening aspect of the results from the current study is that students in both the intervention and control groups predominantly had very negative attitudes towards drug use from the outset of this study, attitudes which became even more negative in the intervention group after they received the intervention.

Although it is important to change young people’s attitudes to drug use, the research has clearly demonstrated that this is not sufficient to achieve behavioural change [41] and the demonstration of behavioural change is what is considered to be essential [36, 42, 43]. The behavioural change of interest in this study is to alter the otherwise predicted course of development, which is either to suppress drug use behaviour or keep it from occurring [44].

The Climate Schools program was effective in delaying initiation to ecstasy use for the intervention group, compared to the control group. Even though this difference between groups diminished by the five-month follow-up, it is an important result as a substantial body of evidence has shown that early onset of drug use is a risk factor for developing a substance use disorder in later life [45, 46]. The Climate Schools program was also related to a plateau in the frequency of ecstasy use in the intervention group, with this difference gradually diminishing over time. Halting the use and frequency of ecstasy use is vital, because even though serious acute adverse events related to ecstasy may be relatively rare, when they do occur they are unpredictable and are associated with considerable morbidity and mortality [23]. Although a plateauing of frequency of ecstasy use was demonstrated in the current study, it is essential to also bear in mind that the overall prevalence of use in the current sample was very low and could at most be described as experimental use.

The Climate Schools program did not delay initiation to cannabis use or impact on cannabis related harms, but it was related to decreasing the frequency of cannabis use for females, although not for males. Although the frequency of cannabis use in the current sample is typical of occasional and experimental cannabis use, even the smallest reductions in frequency of use may assist in subduing the growth in trajectories to riskier levels of use in the future. The fact that the Climate Schools program was not successful in delaying initiation to cannabis use can likely be explained by research that indicates that for prevention to be effective in delaying initiation, it needs to be delivered early. That is, prevention for cannabis needs to be delivered before experimentation occurs and that by 15 years of age, it is too late to deliver a standalone intervention for the prevention of cannabis use, as has been done in the current trial. Research does, however, suggest that ongoing interventions are also important, as young people who cease cannabis use or decrease the frequency of use have better prognoses for the future [47]. This supports the need for sequential and developmentally appropriate interventions, which have now been catered for within the Climate Schools drug prevention resource.

The lack of preventive effects for males in changing the frequency of cannabis use is consistent with a growing body of evidence which shows that school-based prevention programs may be less effective for males than females [19, 48]. Although there is some evidence of converging rates of cannabis use between young males and females [31, 49], considerable evidence suggests that males are not only more likely to have tried cannabis, but are also more likely to become long-term problematic users [50]. The cannabis use of male students in this sample was also elevated in comparison with females, with males being significantly more likely to have ever used cannabis and use cannabis more frequently post-test. Males were significantly more likely to have ever used meth/amphetamine at post-test, in the last three months, and used more frequently in the last 12 months. The majority of these differences between males and females remained consistent over the duration of the study. Likewise for ecstasy, with the exception of lifetime use, males were more likely to have used ecstasy in the last three months and used more frequently in the last 12 months.

The propensity for males to be a higher risk group may be a result of the differences in the level of risk and protective factors for drug use experienced by males in comparison with females [51]. In particular, research suggests that the biological and social consequences of drug use may be stronger protective factors for females in comparison with males; in particular, greater social sanction against drug use by females in comparison with males [51, 52]. Whereas for males, it may be higher levels of impulsivity or the need to use drugs with the motivation of enhancing their capacity to cope that explain some of the increased levels of drug use, more so than for females [53]. Gaining a greater understanding of the risk and protective factors which lead to gender differences in drug use is essential because a more comprehensive understanding of these factors could assist in making prevention program equally and more effective for both males and females.

The Climate Schools program was not effective in changing meth/amphetamine use. This result may explained by the low levels of use in the sample (4-5%) and the lack of variation in use both within and between schools over the study duration, which were evident in the HLM analyses. Students’ intentions to use cannabis in the next 12 months did not significantly differ among the intervention and control groups at any of the survey occasions. Although students in the control group reported significantly greater intent to use meth/amphetamine than the intervention group immediately post-intervention, this effect did not persist at the later follow-up occasions. Similarly, by the 10-month follow-up the intervention and control groups did not differ in terms of intent to use ecstasy in the next 12 months. These findings are likely to reflect the very low levels of intention reported by students in both groups to use cannabis, meth/amphetamine and ecstasy throughout the study period.

Finally, it is important to note that the Climate Schools program was well received by teachers and students. Feedback from the current study indicates that it was feasible to implement a six-lesson computer-based program about illicit substances to Year 10 students. By delivering the program via computers, traditional obstacles to successful implementation, such as the need for teacher training and high costs, appear to have been overcome. Although some teachers reported finding it difficult to access computers to deliver the program, there have been improvements in access to computer technology at Australian schools in the past six years since this trial was conducted and several recent initiatives. For example, the Bring Your Own Device (BYOD) scheme, in which students are encouraged to bring their own laptop or tablet to school, is likely to increase access to computers and improve the feasibility of using computer- and Internet-based programs.

Despite randomisation, the control group was typically a higher risk group than the intervention group at baseline on a number of drug use measures. Although all the analyses took baseline scores into account to control for differences between the intervention and control group, this does not necessarily control for the possibility that the control group may have a higher-risk trajectory over time.

In terms of the analyses employed in the current study, it should be noted that ANCOVA and hierarchical regression procedures are not necessarily the most powerful alternatives when assumptions violated. In recent years, Generalized Estimating Equations (GEEs) have emerged as a superior approach for analysing longitudinal repeated measures data [54, 55] and as such, it would be beneficial to conduct future analyses using a GEE approach.

The attrition of high-risk students, which is a common occurrence in school-based prevention research [56‐59], is also potentially problematic. Attrition of high-risk students has the potential to limit the external validity of findings and may also result in an overestimation of program effects. In the current study there was no evidence of differential attrition, which suggests that the prevention effects which have been detected are not spurious positive findings which have resulted from higher-risk subjects dropping out of the intervention group rather than the control condition.

Another limitation pertains to the accuracy of collecting self-report data on substance use. However, there is evidence to support the validity of self-report of alcohol and other drug use with adolescents [60‐62]. In the current study, assurances of confidentiality were repeatedly provided during survey administration [63] and clear information was provided about the anonymity of individuals and schools in the provision of results [44], factors which have both been shown to enhance self-report accuracy.

Finally, the current study was powered for a standardized difference of 0.15, which is consistent with effects reported for other universal school-based interventions for alcohol and other drugs [33, 64]. Such a modest size of effect is most impactful when applied across large populations and the Internet-based delivery of the Climate Schools program makes such large-scale implementation possible. However, a limitation of universal approaches is their limited impact on adolescents who are most highly at risk of developing problems or who have already developed problems. Targeted prevention programs specifically tailored to the needs of these individuals, such as the selective personality-targeted Preventure program [65, 66], may be more suitable for high-risk students and produce greater effects.