Figure provides a summary of the study; an exploratory cluster randomised controlled trial design with nested mixed methods process evaluation. Ethical approval was provided by Cardiff University, School of Social Sciences Research Ethics Committee, with separate applications approved for the process evaluation (SREC/752) and the cluster randomised trial (SREC/857)
The intervention is a social norm marketing campaign, which aims to correct misperceptions regarding the behaviours and social expectations of peers and in so doing influence alcohol consumption. The campaign will be delivered in two phases between October 2011 and May 2012 in intervention halls of residence in four universities and will use a variety of materials encompassing posters, beer mats/coasters, leaflets, meal planners and drinking glasses. The campaign will be implemented by university accommodation staff. Social norm messages were based on the results of a survey of first year university students conducted in study universities in late April/May 2011 which identified discrepancies between norms and behaviours. Table highlights the intervention materials and main social norm messages within them.
Social norms intervention components and examples of core messages communicated within them
Universities in the study also receive a toolkit to promote institutional responsibility for prevention, audit current alcohol misuse policies and practices and which provided advice and guidance on prevention. The toolkit was developed by a National Union of Students (NUS) intervention project officer in consultation with the universities in the study. It was distributed to key university stakeholders in October 2011 with the intention of developing a supportive environment for the intervention. Control halls will be exposed to the toolkit only. The toolkit and social norms intervention were developed collaboratively by Drinkaware, NUS Wales and the Welsh Government following a review of previous interventions and support from an academic supervisor. Their implementation was facilitated by a dedicated NUS project officer. Given the nature of the intervention, it was not possible to blind participants to condition.
Six universities who had collaborated on the development of the intervention were approached by the evaluation team to participate in the study, with four agreeing to implement the intervention in the study period. Reasons for non-participation were related to difficulties in implementing the intervention and obtaining university consent within the evaluation timeframe. Informed consent for the study was obtained from directors of student services and halls of residence managers. The universities varied in terms of the number of full-time first year students (from 1100 to 3327), and location, with a mixture of urban and rural locations.
The four universities had a total of 51 on campus university owned halls of residence. One female only hall was excluded due to lack of trial arm balance. All remaining halls (n = 50) were eligible for inclusion and consented to randomisation, although 5 halls in one site were empty during the first phase of the campaign due to renovation.
Blind remote randomisation was used to allocate halls of residence to receive the social norms plus toolkit intervention or toolkit only. Halls were stratified by institution and halls allocated alternately in a list ordered by size, with the group allocation determined by one random number within each stratum.
Units consumed per week - daily drinking questionnaire
The primary outcome is alcohol consumption in units per week assessed via the Daily Drinking Questionnaire [13
]. The measure asks students for details of a typical week rather than exact quantities for the last 7 days, in order to ensure that it reflects habitual drinking. The DDQ has emerged as a favoured measure within RCTs with students due to its brevity, its convergent validity with more laborious drinking measures [13
], acceptable internal consistency (Neighbors et al. 2002), good 2-month test-retest reliability for volume and adequate test-retest reliability for frequency [14
]) and established ability to detect post-intervention changes. Importantly, the measure also provides comparable estimates regardless of whether administered via the internet or as a pen and paper exercise [15
Weekly alcohol consumption behaviours - daily drinking questionnaire
Responses can also provide a measure of i) number of days per week drinking in a typical week, ii) number of units per sitting and iv) number of heavy drinking episodes per week.
[Prevalence of higher risk drinking - AUDIT
The consumption subscale of the Alcohol Use Disorders Identification Tool (AUDIT p15;[16
] provides an additional measure of alcohol consumption, allowing estimation of the prevalence of potentially hazardous drinking in control/intervention halls. The scale includes items on frequency of drinking, volume per drinking occasion and frequency of 'binge drinking' (e.g. 8+/6+ units on one occasion for men/women), each scored on a scale of 0-4. In primary care studies, a total summed score of 4 or above for men, or 3 or above for women, has been shown to optimally identify potentially hazardous drinkers [17
Alcohol related consequences - Rutgers alcohol problem index
Secondary outcomes include the 18-item version of the Rutgers Alcohol Problem Index (RAPI) [18
]. The index is a well validated measure of alcohol problems with well established psychometric properties among clinical and general population samples ranging from 12 to 21 years. It is commonly used among general university populations in evaluations of alcohol based interventions. All items are typically summed to provide a single continuous variable for alcohol problems, although the factor structure in the current population will be carefully checked.
In order to assess whether the campaign achieves the hypothesised mechanism of changing perceived descriptive norms for drinking, the evaluation requires a measure of perceived descriptive norms. The drinking norms rating form has been widely used in RCTs and cross sectional studies (Baer et al. 1991) and involves rewording the DDQ to reference others rather than self, therefore having the advantage that perceived normative behaviour is measured in exactly the same way as own behaviour.
Whilst most previous studies have focused on descriptive norms, many psychological models argue that injunctive norms (i.e. perceived social pressure or social approval) are equally important in shaping behaviour. A scale previously used by Neighbors et al. (2008) was therefore included.
Measures of gender, age, ethnicity, international/home student status, course studied and place of residence will facilitate an examination of the representativeness of the sample, to assess comparability between groups of students assigned to receive/not receive the social norms intervention and assess potential contamination between trial arms.
Acceptability of objective measures
Students will also be asked to indicate whether they would be willing to provide hair samples as an objective measure of alcohol consumption, although it will be made clear that this is a hypothetical question, and that we will not be attempting this at any point in the present study. The question is simply included to evaluate the acceptability of this method among university students if we were to seek funding for a larger definitive trial using more objective measurement approaches in the future.
At four months after initial implementation of the intervention, measures will be collected via a survey to all 1st year university students, offered in web and paper format. They will be recruited via nominated university distribution contacts, who will circulate the link to first year students via email and electronic notice boards between mid-February and the end of March 2012. At least one reminder will be emailed to students during the data collection period. On completion of the questionnaire, data will be captured and processed by a market research company, who will prepare a complete anonymised dataset for analysis. Heads of student services provided consent for the conduct of the survey and students will not be able to complete the survey without completing informed consent tick-boxes. Students will not be asked to provide any identifiable information, other than email addresses, which will be used purely for the purpose of selecting a winner for the £100 prize draw in each university, offered as an incentive for participation. Email addresses will be separated from responses to the web survey and destroyed after the prize draw.
In an attempt to boost student responses, residence hall managers will be asked to promote the survey and the prize draw to residents. To compare the efficacy of two data collection approaches, a paper copy of the questionnaire will be distributed to student halls of residence via accommodation managers, inviting students either to complete the paper copy and return it to the research team in a freepost envelope, or to go to the web-page to complete the survey online. Questionnaires completed in paper form will be returned to the research team in freepost envelopes. These will be stored in a locked cabinet until the web survey data-file is received from the survey company, at which point, questionnaires will be retrieved, data entered into the data-file, and questionnaires returned to the cabinet. Participants will be offered the opportunity to enter a prize draw, with £100 offered to one winner in each participating university by supplying a university email to be kept separately to questionnaire responses. Email addresses will be recorded on a detachable sheet at the start of the questionnaire, which will be separated from the questionnaire once received, with the email address entered into a separate spread sheet and the paper copy destroyed.
Assuming a student response rate of 40%, 1600 completed questionnaires will be available for analysis, an average of 32 students per hall. Assuming an intra-cluster correlation of 0.03, fifty halls of residence will therefore provide 80% power to detect a 0.2 standard deviation difference in units of alcohol consumed using a two-tailed alpha of 0.05.
Assuming a student response rate of 25%, 1000 completed questionnaires will be available for analysis, an average of 20 students per hall. Assuming an intra-cluster correlation of 0.03, fifty halls of residence will therefore provide 80% power to detect a 0.23 standard deviation difference in units of alcohol consumed using a two-tailed alpha of 0.05.
It is not anticipated that the effect size will be of this magnitude, with a much larger trial likely to be necessary to detect realistic effect sizes below 0.1 standard deviation. This study is therefore designed as an exploratory trial to assess the value of the intervention and plan a larger scale study if warranted.
Universities are complex systems, whose ethos, policies and practices may provide a context supportive of change, or of maintaining the status quo [10
]. Within evaluations of complex interventions, process evaluation is crucial in order to understand what was implemented, how it was received and ultimately, how outcomes were produced. A process evaluation will run alongside the implementation of the programme, throughout the 2011/12 academic year. The process evaluation is concerned with 5 core research questions:
i. What role does alcohol play in students' social life during the transition to university and throughout university life?
ii. How are the toolkit and social norms activities developed and what are their underlying logic models?
iii. How are the toolkit and social norms activities implemented?
iv. How, for whom and in what circumstances, does the toolkit brings about change in university practices
v. How, for whom and in what circumstances does the social norms intervention influence alcohol related beliefs and behaviour?
The process evaluation will encompass 1) group interviews with up to twenty 2nd and 3rd year students in each university focusing upon experiences of alcohol throughout student life, 2) visits by a researcher to each intervention residence hall in order to monitor the distribution and placement of materials, 3) group interviews with up to 6 students in 2 case study halls in each university (one receiving and one not receiving the social norms intervention) exploring awareness and responses to the intervention 4) interviews with stakeholders in each university involved in delivering the intervention. In addition, all residence hall wardens will be asked to complete a brief questionnaire to assess changes in practice over time. Permission will also be requested from university representatives to use routine public data gathered during audits forming part of the toolkit.
Finally, within the survey described above, to assess intervention reach, students will be asked to indicate whether they had seen the intervention materials in their own hall of residence, or in another students' hall of residence. To assess recall, students who recalled seeing any of the norms materials will be asked to identify core messages from a list. Students will also be asked whether messages within the materials were credible and relevant, and whether they felt that exposure to the materials had influenced their normative perceptions or behaviour. These questions will be identical for students in control and intervention halls, allowing assessments of contamination between trial arms. The survey also includes a number of bespoke items from the intervention survey, which informed the social norm intervention, but only where these are linked to specific intervention communications (e.g. some materials focused on round buying behaviour and alternating alcoholic and soft drinks, hence items assessing the prevalence of these behaviours are retained).
Hall of residence managers will be asked for their consent for researchers to visit halls to monitor the placement of campaign materials. Prior to group interviews, an information sheet would be provided, with participants offered the opportunity to ask questions prior to obtaining informed consent. Since part of the process evaluation requires asking different questions of intervention and control premises representatives, the research team members who conduct the process evaluation will be unblinded.
In order to assess exposure to intervention materials and contamination between trial arms, percentages of students within the intervention and control groups reporting having seen each of the intervention materials i) in their own hall of residence and ii) in another students' hall of residence will be examined. Among those students reporting exposure to intervention materials, percentages correctly identifying the messages within them will be calculated for each trial arm. Percentages of students reporting each level of agreement with statements regarding the credibility, relevance and perceived impacts of intervention materials will also be examined for each trial arm.
Whilst the study is likely not sufficiently powered to detect impacts on behaviour, it is likely that relatively large changes in perceived norms will be necessary to produce small changes in behaviour. Hence, regression analyses, with random terms to adjust for clustering at the hall level, and fixed terms to adjust for stratification variables, will examine differences between intervention and control participants in terms of normative perceptions for alcohol consumption and alcohol related consequences. Comparisons between trial arms will be conducted on an intention-to-treat basis. Secondary analyses would compare halls on the basis of researcher observations of whether or not materials were placed.
To inform the design of a potential large scale definitive trial with sufficient power to detect changes in behaviour, intra-cluster correlations and standard deviations will be calculated for total number of units per week. Response rates will be calculated in each trial arm. The percentage of students reporting willingness to provide hair samples will also be presented, whilst among those students reporting that they would only do so if paid, percentages reporting that each level of payment would be required would be presented.