Text Messaging Interventions for Adolescent and Young Adult Substance Use: A Meta- Analysis

Introduction

Tobacco and alcohol use continue to be associated with negative health outcomes among adolescents and young adults. New technologies such as text messaging can increase access to substance use interventions and have now been established as an evidence-based, recommended approach towards substance use prevention.  In a recent survey the Pew Research Center found that 83% of adults own mobile phones and nearly three-quarters (73%) of these send and receive text messages. Young adults are the most active text message users across all age groups. However, little is known about the effect of these interventions on this age group. The present meta-analytic study sought to expand on this literature by addressing the following research question: What is the comparative effectiveness of text message programs to no treatment or treatment as usual among adolescents and young adults using tobacco or alcohol?

 Methods

Eleven studies were included that used text message intervention as a stand-alone intervention or in combination with another program as long as sufficient detail was reported to calculate effect sizes. Eligible studies were included only if they utilized a sample that was inclusive of adolescents and/or young adults ages 12 to 29.  All included studies were required to be randomized controlled trials. Eligible studies had to include use of at least once substance as a primary outcome measure, and were required to be published in 2000 or later.

 Results

We obtained Cohen’s d by computing the standard mean difference and by converting odds ratios provided in articles’ results to standard mean differences.  Then we converted Cohen’s d to Hedges’ g to remove the estimation bias of effect sizes ascribed to small sample sizes.  We computed the variance for each study. The inverse variance was assigned as a weight for each study, with larger weights indicating better precision. We then combined the effect size across studies to create a summary effect size.

Nine of the eleven studies had positive effect sizes (Hedges’ g), ranging from -0.04 to 1.38.  One study had a borderline medium effective size 0.46 for tobacco (Free et al., 2011) and one study had a large effect size 1.38 for alcohol (Mason et al., In press); the rest had small effect sizes. Effect sizes were combined across studies to produce a summary effect size of 0.20, indicating that the combined treatment effect is small. Nonetheless, participants’ reduction of tobacco or alcohol use would be 20% greater in the treatment conditions than in the control conditions.    

 Conclusions

Results suggest that there is sufficient evidence to warrant further research into the effects of a low-cost intervention delivery model. Effect sizes varied, but in general those more participants in the treatment conditions experienced reduction of substance as compared to control participants. This line of research is no doubt poised to rapidly expand as low cost, large studies can continue to refine and improve their design. The potential for highly flexible, adaptive interactive interventions that respond in real time has promise to support current evidence based prevention interventions and to expand into novel areas of health promotion and risk reduction.