The Swedish six-community alcohol and drug prevention trial: Effects on youth drinking


  • Mats Hallgren PhD, Researcher, Sven Andréasson MD, PhD, Professor.

Correspondence to Dr Mats Hallgren, Department of Public Health Sciences, Division of Social Medicine, Karolinska Institutet, c/o STAD, Teknologatan 8E, 113 60 Stockholm, Sweden. Tel: +46 08 12345 501; Fax: +46 08 12345509; E-mail:


Introduction and Aims

Local communities are increasingly targeted for alcohol and drug prevention campaigns. This study describes some of the key findings from the Swedish six-community alcohol and drug prevention trial (2003–2007) and lessons learned following an evaluation of the trial's effectiveness. The paper focuses mainly on changes in youth drinking and related harms.

Design and Methods

This was a pre- to post-intervention effect study comparing six trial communities that received added training and technical support with six control communities where regular prevention efforts were supported by national alcohol and drug action plans. A repeated, cross-sectional survey of 8092 youths aged 15–19 years assessed changes in alcohol consumption, binge drinking, perceived alcohol availability, access to alcohol via parents and adult attitudes towards the supply of alcohol to youths. National registry data were used to assess changes in hospital admissions due to alcohol intoxication.


Overall, there were few significant improvements in the six trial communities compared with the control communities.

Discussion and Conclusions

The absence of program effects was largely attributable to the selection of strategies (in particular, school and parental programs) lacking evidence of effectiveness in reducing alcohol consumption at the aggregate level. Prevention programs based on efficacy studies need to be tested in community-based effectiveness trials before being disseminated. [Hallgren M & Andréasson S. The Swedish six-community alcohol and drug prevention trial: Effects on youth drinking. Drug Alcohol Rev 2013;32:504–511]


To date, only a small number of multi-component trials to reduce alcohol problems have been reported in the scientific literature [1-9]. Importantly, the focus of recent community interventions has shifted from changing the behaviour of ‘high risk’ individuals to modifying identifiable social and physical environments associated with alcohol and drug use—the ‘systems’ approach to prevention [10]. In Sweden, alcohol and drug availability has increased over the past 15 years, and serious alcohol-related harms among adolescents have increased sharply [11]. Total alcohol consumption rose by around 19% between 1996 and 2010, and reports of drug-related activity have also risen. These increases prompted the adoption of new national alcohol and drug action plans with a greater emphasis on local community prevention.

In 2002, a decision was taken by key alcohol and drug agencies in Sweden to implement and evaluate a comprehensive community-based intervention that aimed to mobilise and enhance local alcohol and drug prevention efforts in six Swedish communities (Umeå, Kramfors, Solna, Kalmar, Laholm and Lund). The aim of this six community project was to reduce alcohol and drug use and related harms among adults and youth and serve as a model to other municipalities in Sweden. A description of the planning phase of the trial has been reported previously [12].

The research question was whether support for the implementation of evidence-based prevention would produce better outcomes in the intervention communities collectively compared with control communities where regular prevention efforts were supported by the National Alcohol and Drug Action Plans. Importantly, the additional programs in the trial communities all had evidence of effects from efficacy trials; few had evidence from effectiveness trials demonstrating outcomes at the community level. The focus was neither on individual communities nor on specific prevention methods. Furthermore, in this first paper from the evaluation, our focus is on youth drinking.


The intervention

Six intervention communities in Sweden were selected on the basis of their willingness and capacity to participate in the trial by increasing alcohol and drug prevention efforts in their local communities. Six demographically matched control communities were also selected for comparison purposes. Implementation of the project was guided by a national steering committee, a project manager with responsibility for overall project coordination, and action groups in each trial community responsible for implementing the chosen strategies locally. Each trial community was provided with a menu of evidence-based prevention methods and offered training and technical support for the implementation of these. However, the final selection was made by local community stakeholders and was seen as a compromise between evidence from the scientific literature and community priorities. Most of the prevention strategies chosen involved youth-related activities, such as parental education and training, and school-based interventions. Other prevention strategies involving adults, such as responsible beverage service (RBS), prevention of drunk driving, and screening and brief intervention in primary health, was taken up by some communities but to a lesser degree. Underpinning all prevention programs were sustained efforts to organise and mobilise community resources in ways that encouraged work towards the trial objectives. It was anticipated that the interventions would result in a number of intermediate and long-term changes, including decreased heavy drinking, especially among minors. An intermediate goal was to initiate a shift in community attitudes towards the regulation and supply of alcohol to young people. The strategies were implemented over a 4-year period, commencing in 2003, with the first 2 years taken up primarily with planning and implementation activities.

Prevention activities

Seven key prevention programs implemented by at least four of the trial communities are described below with an indication of how widely each program was implemented during the project. Table 1 summarises the main features of the prevention programs. As shown, most of the programs targeted individual level behaviour and the demand for alcohol. The six control communities were also engaged in alcohol prevention during the project period but with considerably less support and fewer programs overall [13].

Table 1. Summary of the programs implemented by the trial communities
Program nameType of programTarget groupNumber of Trial Communities
  1. RBS, responsible beverage service.
Information and media advocacyCommunity basedAdults and youths6
AvailabilityPolice trainingPrimarily youths6
Social and emotional trainingSchool programPrimary school children6
Responsible beverage serviceAvailability/accessBars and clubs serving alcohol5
Örebro Prevention ProgramParent programParents with secondary school aged children5
Motivational interviewing for student healthSchool programPrimary and secondary school students4
KOMET for parentsParent programParents with children aged 3–12 years4
Parental PowerParent programParents of children 13–17 years3
RBS for university studentsAvailability/accessStudent pubs3
Strengthening Families Program ‘Steg för Steg’Parent programParents with children aged 10–14 years2
SMADIT'Treatment strategy for drunk driversAnyone with a driving license3
Alcohol screening and counselling during pregnancyAlcohol screening and counsellingPregnant women3
Early detection and brief counsellingBrief intervention in primary health careAdults2
Parental stepsParental behaviour programParents of children 13–172
‘Don't drink and drive’ campaignDrink-driving information/educationYoung adults aged 15–25 years1


Participation in the program meant that venue staff completed a two-day RBS training program at least once during the project period. In 2007, the participation rates were as follows: Laholm and Umeå (20%), Solna (71%) and Kramfors (55%). No data were available for Kalmar or Lund. The number of compliance checks in the trial communities increased from 277 in 2004 to 367 in 2006. At the same time, the number of checks decreased in the control communities from 275 in 2003 to 125 in 2005 (data were not reported in 2006/2007).

Social and emotional training (SET)

SET aims to develop adolescents' social skills so they are better equipped to make choices that reduce their exposure to alcohol and drugs. A recent Swedish study using a quasi-experimental, longitudinal design found that the program had generally favourable effects on mental health [14]. By 2007, SET had been implemented in 10% of schools in Umeå, 29% in Lund, 60% in Laholm, 64% in Solna and 51% of schools in Kalmar, with no data available from Kramfors.

Motivational interviewing (MI)

In the trial communities, MI was integrated into school health services work, when students attend health examinations or request counselling. By 2007, the proportion of student health personnel (nurses, psychologists and counsellors) who were trained in MI ranged from 2% of all schools in Lund to 30% in Solna, 85% in Kalmar and 100% in Laholm and Umeå, with no data available from Kramfors.

Örebro Prevention Project (ÖPP)

The ÖPP program aims to positively influence parents' attitudes to young people's drinking and teaches parents how they can act to prevent alcohol misuse in young people [15]. ÖPP was one of the more widely implemented strategies among the trial communities. In 2007, all schools in Kalmar and Laholm were actively working with the program. In Umeå, 47% used the program, and in Solna and Lund, 64% and 69% of schools, respectively, had implemented ÖPP.

Komet for parents

This program is intended for parents with children aged three to 12 years with disruptive behaviour or who have consistent problems establishing peer relationships or difficulty concentrating at school [16]. The program was implemented in three communities between 2004 and 2007: Kalmar, Solna and Lund. The number of parents receiving the program were few during the trial years and varied between five (Kalmar, 2004) and 67 (Lund, 2006).


All six trial municipalities were encouraged to work with availability measures. This primarily involved offering the police additional training in alcohol availability and enforcement measures. The percentage of police officers trained in availability regulations in 2006–2007 varied widely, from 10% in Solna, 35% in Umeå and 75% in Laholm. The application of this knowledge by police, in terms of venue inspections at the local level, was not assessed.

Informing the local community

Considerable efforts were made during the trial period to increase the number of media reports (mainly print media) describing alcohol-related problems and harms occurring at the local level in the trial municipalities. In practice, this frequently involved local communities working with journalists to ensure that particular issues were highlighted in the media. Every three months, the total number of relevant print media articles was counted, and in January to March 2003, there were approximately 10 articles in the control communities compared with 65 articles in the six trial communities. Four years later in 2007, this number had increased to 75 (control) and 135 (trial) articles, respectively.

Measuring the overall prevention ‘dose’

Given the comparative nature of the trial, the prevention activities undertaken by the six control communities are also relevant. In an effort to quantify the overall dose of structured prevention work in the 12 communities, we developed a measure consisting of two indices. First, a 16-item ‘Organisation’ index to assess how well the municipality's prevention work was structured. This index included questions about alcohol and drug policy, personnel and compliance with regulations. For example, ‘the municipality has a structured alcohol policy in place’, ‘the municipality works with (at least three) other organisations to reduce alcohol problems’ and ‘the percentage of liquor licence compliance checks conducted in venues’. Second, an 11-item ‘Activity’ index that assessed the total number of prevention activities initiated during the last year. The maximum possible score is 27 points. The indices are based primarily on data collected through the Swedish National Institute of Public Health's work development questionnaire, which is posted to all municipalities annually. In 2006, three years in to the project, the six trial communities scored higher on both the organisation and activity index compared with the control communities and the remaining 198 communities not involved in the trial. The total scores (out of 27) for trial, control and other municipalities were 21 points, 17 points and 16 points, respectively. Indices for 2007 were not calculated. A forthcoming paper will describe the development and validation of the ‘Prevention Index’ in detail.

Effect evaluation

An effect evaluation was undertaken to assess the overall impact of increased prevention support and activity within six trial communities over five years, compared with the regular prevention work in six matched control communities. The main outcomes of interest were changes in youth drinking (total volume and heavy episodic drinking) and alcohol-related hospitalisations within the six trial and six control communities. We were also interested to know what impact the trial had on three additional factors frequently associated with hazardous adolescent drinking, parental provision of alcohol to children [17, 18], adult attitudes towards youth drinking [19] and youth's perception of alcohol availability in the community. We regard the last factor as a pseudo-indicator of actual alcohol availability—a well-established risk factors for alcohol misuse [20, 21]. Given that some of the strategies implemented by the trial communities aimed to reduce youth access to alcohol (e.g. RBS training) and increase community awareness of alcohol problems (media advocacy), it is conceivable that the trial could influence these outcomes. Importantly, we did not set out to compare individual communities with each other or assess the impact of specific program components.

Changes in alcohol use were assessed by a self-report questionnaire mailed to residents within the 12 respective communities in May of each year (except 2005). Two stratified, unbound, random samples were drawn on the four measurement occasions; one for adults consisting of 7200 individuals aged 19–70 years and one for adolescents consisting of approximately 4800 individuals aged 15–19 years. This meant that 1000 people (400 adolescents and 600 adults) were surveyed from each of the 12 communities. In total, 12 000 questionnaires per year or 48 000 across the four-year study period were posted to residents. Lastly, changes over time in alcohol-related harms were assessed with hospitalisation data from the Swedish National Board of Health and Welfare. A separate process evaluation was conducted during the trial with the aim of understanding what factors helped or hindered the implementation of prevention work at the local level [22].

Statistical analyses

Descriptive statistics are presented based on cross-sectional analyses of the survey data and national registry data for hospitalisations. Two-way analysis of variance (anova) was used to explore differences over time between trial and control communities (interaction effects). The Kruskal–Wallis H non-parametric test assessed changes in binge drinking frequencies and the proportions of parents offering alcohol to their children. Changes in the proportion of alcohol-related hospitalisations were also examined using the Kruskal–Wallis H-test. As the prevention activities were implemented progressively over five years, data for each year (except 2005) are presented to enable cumulative effects of the prevention work to be seen. All analyses were completed with spss version 20 (IBM SPSS Inc. SPSS Base 20.0, Chicago, USA).


In total, 8092 questionnaires (42%) were returned over the 4-year study period. When the four surveys were combined, the response rates for year 9 and 11 students were 30.5% and 53%, respectively. The lower response from younger adolescents may be attributable to the fact that parental consent was required before the surveys could be returned by regular mail (parental consent was not required for year 11 students). Response rates by school year are shown in a separate table available in the online supplement to this journal.

Alcohol consumption

Changes in per capita alcohol consumption are shown in Table 2. Generally, there was a trend towards less consumption in both the trial and control communities, except among year 11 females and to some degree year 11 males in the control areas only. There were no statistically significant group by time interactions, indicating that the trial interventions had no measurable impact on youth drinking. Table 3 shows changes in the percentage of young people who reported binge drinking once or more per year by age and school year. All young people reported reductions in binge drinking over time, with no differences overall between the trial and control communities. However, the proportion of year 11 males who binge drink reduced somewhat more in the trial communities. Among year 9 females, binge drinking reduced by 20% (P < 0.001) in the trial communities compared to 10% in the control communities (P < 0.05).

Table 2. Changes in youth alcohol consumption over time*
 20032004200620072 group × 4 year anova (interaction effect)
  1. *Centiliters of 100% pure alcohol consumed per year.
  2. anova, analysis of variance.
Year 9 Males     
Trial (n = 621)349320274237F3,1267 = 0.065, P = 0.97
Control (n = 645)329253231196
Year 9 Females     
Trial (n = 814)236185242180F3,1692 = 0.033, P = 0.80
Control (n = 887)232182197132
Year 11 Males     
Trial (n = 1094)513727545553F3,2207 = 0.988, P = 0.39
Control (n = 1122)643694584699
Year 11 Females     
Trial (n = 1486)400378402330F3,29787 = 0.300, P = 0.82
Control (n = 1501)392328370526
Table 3. Changes in the proportion of youth self-reported binge drinking over time
  1. P-values for Kruskal–Wallis H non-parametric test of statistical significance, **P < 0.001, *P < 0.05.
Year 9 Males      
Trial (n = 623)242020173.01 (3)0.389
Control (n = 632)221816171.76 (3)0.624
Year 9 Females      
Trial (n = 797)3119171135.21 (3)0.001**
Control (n = 868)211415118.90 (3)0.031*
Year 11 Males      
Trial (n = 1079)4756434112.79 (3)0.005*
Control (n = 1105)475043443.40 (3)0.333
Year 11 Females      
Trial (n = 1471)424035346.77 (3)0.079
Control (n = 1484)423733348.70 (3)0.033*

Parents offering alcohol to their children

The self-report survey included one item to assess whether or not parents had offered alcohol to their children (Table 4). The 5-point scale ranged from 1 (my parents do not drink) to 5 (yes, I am often offered alcohol by my parents). Differences over time (2003–2007) were assessed with the Kruskal–Wallis H non-parametric test. There were clear improvements over time, with a lower proportion of parents offering alcohol to their children by 2007. The largest improvements were seen in the trial communities.

Table 4. Change over time in the proportion of parents who offer alcohol to their children
  1. This item is based on an affirmative response to the question: ‘Have you been offered alcohol by your parents? Yes, a taste from my parents' glass, or Yes, from my own glass, or Yes, I am often offered alcohol by my parents'. P-values for Kruskal–Wallis H non-parametric test of statistical significance **P < 0.001, *P < 0.05.
Year 9 Males      
Trial (n = 618)6262484120.50 (3)0.000**
Control (n = 633)626154566.57 (3)0.087
Year 9 Females      
Trial (n = 799)6769534435.21 (3)0.000**
Control (n = 876)7261545223.61 (3)0.000**
Year 11 Males      
Trial (n = 1079)7567575923.49 (3)0.000**
Control (n = 1097)7477696510.52 (3)0.015*
Year 11 Females      
Trial (n = 1471)7779706231.14 (3)0.000**
Control (n = 1480)8175706819.17 (3)0.000**

Adult attitudes towards alcohol availability

Young people's access to alcohol is influenced by the behaviour and attitudes of adults in their wider community. We therefore assessed changes in adults' attitudes towards the supply of alcohol to youth. Results indicated that adults in both the trial and control communities developed significantly more restrictive attitudes towards the supply of alcohol to youth, but there were no group by time interactions.

Alcohol-related hospitalisations

Table 5 shows changes in the proportion of adolescents hospitalised with a primary and/or secondary alcohol related diagnosis [International Classification of Diseases (version 10) (ICD-10) codes f10, acute intoxication and T51, toxic effect of alcohol]. Between 2003 and 2007, there was an increase in alcohol-related hospital admissions nationally and in the trial communities but with variability between years. There were no significant group differences over time.

Table 5. Number of adolescents (per 10 000) aged 15–19 years hospitalised with a primary or secondary alcohol-related diagnosis


This community trial resulted in almost no positive effects at the aggregate level with two partial exceptions: the proportion of parents who offer alcohol to children reduced more in the trial communities and the proportion of binge drinking year 11 males reduced slightly more in the trial communities. A similar trend was seen among year 9 females.

Reasons for the absence of program effects

The trial communities ultimately chose to implement programs with weak evidence of population level effects. Only four of the 15 programs listed in Table 1 have demonstrated evidence of effectiveness in reducing alcohol consumption or related harms. These include availability measures (RBS programs and police training in regulation enforcement) and brief interventions in primary health care [21, 23, 24]. Such programs received relatively little attention compared with the demand-reducing programs prioritised by the trial communities. One reason for these choices was that several of the programs recommended by the steering group had evidence based on efficacy trials, with little evidence from effectiveness trials. Failure to observe this distinction was combined with traditional views of prevention in the communities. Many stakeholders in this trial regarded the youth- and school-based activities as effective in a self-evident way. Local decision makers often regarded individual behaviour change as the goal of alcohol and drug prevention, rather than aiming for legislative or systemic change. Another relevant factor was that some of the programs chosen were already funded. and the personnel running the programs were known to stakeholders in the community.

Furthermore, it was difficult to assess whether or not the strategies were uniformly implemented as intended. Process data from the local coordinators showed that even where the trial communities had decided to implement certain programs (e.g. school-based SET), on average no more than 40% of all schools received the intervention. This illustrates that the programs were complex, requiring extensive training and technical support to an extent that average communities were not prepared to pay for.

Another explanation for the generally poor program effects concerns the prevention work in the control communities. During the 4-year study period, the control communities were also engaged in prevention activities as part of the National Alcohol and Drug Action plans in Sweden these years. The possibility of spill over effects, where control communities are influenced by the activities of trial communities cannot be excluded entirely, even though they were separated geographically. One could also legitimately question whether strategies targeting adults (e.g. RBS and screening and brief intervention) could have had an impact on youth drinking at the population level,

Our results show a divergence between youth alcohol consumption, which reduced during the trial, and alcohol-related hospitalisations, which increased overall. The most likely explanation for these changes is a polarisation in youth drinking. Although most young people are drinking less alcohol over time, a minority continue drinking at hazardous levels resulting in increased harm [25].

Limitations and methodological issues

Response rates from the community survey were somewhat low, ranging between 25% and 55%. However, the survey data appear to be representative of current alcohol consumption trends reported by Swedish youth nationally. Differences between trial and control communities were assessed with anova, which assumes independent selection of cases. Multi-level modelling is an extension of multiple regressions and is appropriate for analysing hierarchically structured or nested data and may be considered the preferred analytical method. However, the general absence of significant interaction effects using a less statistically rigorous approach (anova) brings into question the necessity of a multi-level analysis; a methodology that makes it substantially more difficult to obtain significant pre–post program effects.

Implications for future prevention trials

What can be learned from this community trial? Even where prevention programs have shown promise in efficacy trials, it remains a challenge to demonstrate effectiveness in real-world settings. However, our data show that local communities can be mobilised to develop and implement alcohol prevention initiatives. This process takes time: despite the favourable circumstances that prevailed in the test municipalities, it took 2–3 years for the concrete work to get under way. Gaining acceptance for these measures requires considerable advocacy and efforts to bring together researchers and practitioners, a process that demands a respectful exchange and an appreciation for what both groups can bring to a joint prevention effort [26]. Presenting the scientific evidence is necessary but, in this instance, was not sufficient to shift entrenched ideas about ‘high risk youth’ and individually focused programs, some of which were already in place with ongoing funding. Programs that address availability issues and locally identified risks are more likely to succeed [23, 27].


We wish to thank all our participants. Financial support was provided jointly by the Karolinska Institutet (KID funding for doctoral education), the Swedish Council for Working Life and Social Research (grant nr: 2005-0064) and the Swedish National Institute for Public Health (grant nr: 2008/167). The authors would also like to thank Associate Professor Richard Bränström for his comments on an earlier draft of this paper.