Table 4 Information on methods, results, and possible bias of the six studies included in the present review[32–37]

Incerto, et. al., 2011 [32], “Fourth Year Fifth”event, determine effectiveness of social norms marketing campaign to prevent participating in the “Fourth Year Fifth” event ¹.

Observational cross-sectional. Web-based survey to random sample. Wilcoxon Signed-Rank tests. Descriptive statistics.

n = 536/1,000, response rate = 53.6%.

Participation in the “Fourth Year Fifth” event: 19.6% participated. Application of protective behaviours: 86.3% diluted alcohol; 80.6% had sufficient sleep; 78.6% ate large breakfast. Relation participation and exposure to campaign elements : χ² = 34.81, d.f. = 6, p ≤ 0.001.

No control group. No pretest. Different response prompts to questions: no comparison possible. Short-term effect only.

Glassman, et. al., 2010 [33], “Less is more”, determine effectiveness of the social marketing campaign “Less is more”.

Observational longitudinal. Standardized quantitative survey with random sample. Data collection was done six times, from fall 2004 until spring 2008. Descriptive statistics.

n = 473/2,400 in fall 2004, 19.7%

Impact on high-risk drinking: significant decrease from 56.5% to 37.8%. Impact on drinking and driving: significant decrease from 37.5% to 20.6%. Impact on the perception that alcohol use increases sexual opportunities: significant decrease from 64.0% to 50.7%.

No control group. Low response rates. Other prevention efforts may have caused the effect. Short-term effect only.

n = 1,006/4,000 in fall 2005, 25%

n = 785/4,000 in fall 2006, 19.6%

n = 835/4,000 in spring 2007, 20.9%

n = 745/4,000 in fall 2007, 18.6%

n = 546/4,000 in spring 2008, 13.7%.

Slater, et. al., 2006 [34], “Be under your own influence”, determine effectiveness of an in-school media campaign “Be under your own influence” reinforced by community-based media efforts, on the reduction of increase of substance uptake. For this review study, only results of “Be under your own influence” are discussed.

Experimental longitudinal. Randomised community crossed design: 8 communities received social marketing in-school media intervention and 8 communities did not. Four waves of data collection, during two years. Generalised linear mixed models (four-level random-intercept model).

n = 4,216 Response rates: 68.6% provided data at 4 measurements, 16.8% at 3, 10.9% at 2, and 3.7% at 1.

Alcohol use: odds ratio (OR) = 0.40, p ≤ 0.01. Effect on rate of change in alcohol use: OR = 0.82, p > 0.05. Recognition of campaign messages: Time 2, OR = 4.70, p ≤ 0.01; time 3, OR = 6.80, p ≤ 0.01; time 4, OR = 10.13, p ≤ 0.01.

Short-term effect only. Other prevention efforts may have caused the effect.

Rothshild, et. al., 2006 [35], “Road Crew”, determine effectiveness of social marketing intervention “Road Crew”.

Experimental longitudinal. Treatment for 1 year, with pre- and post-test. Three treatment communities and five control communities. Generalized linear models.

n = 710 and n = 693 at pre-test in treatment and control groups. n = 573 and n = 371 at post-test in treatment and control groups.

Count of all rides taken in treatment communities: 10,097 rides taken by 21-34-year-olds. Self-report of drinking and driving behaviour: less likely to drive themselves or ride with someone else (OR = 0.40, p ≤ 0.05); no significant changes in alcohol-impaired driving (χ² = 0.82, p > 0.05); decrease in reported number of alcohol-impaired driving (χ² = 4.85, p ≤ 0.05).

Observing changes in the number of actual crashes was not possible. Possible differences between communities of treatment and control groups. Self-reported data of bar patrons possibly underestimated.

Gomberg, et. al., 2001 [37], “Just the facts”, analyse the results of the “Just the Facts” (JTF) campaign.

Observational longitudinal. Survey with random sample. Three times of data collection (1 pretest and 2 posttests, just after the two campaign phases). Two-sample independent t tests. Chi-square analyses. Linear regression analyses. Logistic regression analyses.

n = 785 for pretest, n = 698 for first posttest n = 583 for second posttest.

Recognition of campaign logo: 6.2% at pretest, 55.4% at first posttest, 78.5% at second posttest. Alcohol use: decrease of mean number of drinks from 15.80 at pretest to 12.61 at second posttest; decrease in mean number of days from 2.96 at pretest to 2.65 at second posttest. High risk drinking: decrease for male students from 65.6% at pretest to 58.4% at posttest and for female students from 40.5% at pretest to 34.7% at second posttest. Perceived drinking norms: significant increases in correctly answered questions about the drinking norms.

Shortcomings in research design. No control group. Decreasing response rates for three surveys. Not asked to recall campaign messages, only logo and advertisements. Measurement for high-risk drinking is not comparable.

Caverson, et. al., 1990 [36], “Thanks for being a sober driver”, determine how the “Thanks for being a sober driver” intervention was received by the community.

Observational cross-sectional. Field experiment of 1 year. Telephone interview, conducted several months after end of pilot. Other measures: number of cars stopped, number of offences and number of folders handed out at spot-checks. Further, interviews with key informants from police department and senior officers. Descriptive statistics.

n = 445/667, response rate = 67%.

Awareness of intervention: 76%. Knowledge of slogan: 13%. Stopped by the police: 79% not been drinking prior to driving. Reaction to this and other equivalent interventions: 93% good idea to reward sober drivers.

No control group. Short-term effect only. Other drinking-driving countermeasure programs were run simultaneously: not clear whether results can be attributed to “Thanks for being a sober driver”.