Use of theory in computer-based interventions to reduce alcohol use among adolescents and young adults: a systematic review

PsychINFO and PubMed (electronic databases) were searched to identify peer-reviewed journal articles on computer-based interventions aimed at preventing or reducing alcohol. The search included previous reviews of CBIs. In addition, Google Scholar was searched to identify additional articles/abstracts that may have been published. The reference lists of all the identified articles were also reviewed. The search, which used both Medical Subject Headings [MeSH] and non-MeSH terms, used the search terms: “alcohol abuse prevention,” “alcohol,” “alcohol drinking/prevention and control,” “computer,” “internet,” “web,” “computer software,” “computer games,” and “intervention.” The search was conducted between November and December of 2014.

To be included in this review, the main component of the intervention was required to be delivered via computer, tablet or smartphone. Interventions could include a video game, computer program, or online module. In addition, the intervention needed to target alcohol use among adolescents and young adults between the ages of 12 to 21 years. While adolescence covers a wide range, we chose this age range because there is general consensus that it has begun by age 12, and we included youth up to age 21 since that is the legal drinking age in the U.S. Studies whose participants’ had a mean age between 12 and 21 years were included even when individual study’s participants’ ages extended outside this age range. Interventions intended to treat a substance use disorder were excluded. Non-English language articles, research protocols, and intervention studies that did not report outcomes were also excluded from analyses.

Once eligible studies were identified, the characteristics of the intervention, the context of the intervention, the population targeted, intervention dosage, study author, year and outcomes were entered into a spreadsheet for analyses. Duplicate articles were deleted and journal articles which discussed the same intervention were grouped together. When there was more than one unique article for any given CBI, the CBI was counted only once. In some cases, a given CBI existed in several editions, was modified, or was applied to a different study population. These variations of the CBI were grouped together.

Painter et al.’s classification system was used to categorize the use of theory in each of the CBIs [22]. Consistent with this system, first a CBI was examined to see if an established, broad theory was mentioned in any of the corresponding articles for a given CBI. If so, the CBI was classified as “mentioned”. Second, articles were reviewed to see if they provided any information about how the CBI used theory to inform the intervention. If any of the articles associated with a given CBI provided any information about the use of theory, the CBI was classified as “applied. For our third category, we used “measured” to classify CBIs if any associated article included at least one specific measure of a construct within the theoretical framework. This third category is a slight departure from Painter’s typology which classifies interventions as “tested” if over half of the constructs in the theory are measured in the evaluation of the intervention. We opted for “measured” because testing theories is a complex process and not a common practice of CBIs. We did not use Painter’s 4th category, “building or creating theory” because this was not applicable for any of these interventions.

For all articles reporting on effects of the intervention on alcohol use, attitudes, or knowledge on an included CBI, the effectiveness of the CBI on these outcomes was also examined.

Two senior health research scientists (a counseling/health psychologist and developmental psychologist), with advanced training in theories of behavior change, oversaw the classifications system and addressed questions about the application of a theory/theoretical constructs. The review was conducted by a trained research associate with a master’s degree in public health. A spread sheet was created that included each classification, a description of how the theory was applied, and a list of relevant constructs that were measured.

Tables 1 and 2 summarize the included studies. The interventions were largely studied exclusively in the United States (30/42). The remaining interventions were studied in Australia (n = 3), New Zealand (n = 3), the Netherlands (n = 2), the United Kingdom (n = 2), Sweden (n = 1) and in both the United States and Canada (n = 1). Study sample sizes ranged widely. Included studies had between 59 and 20,150 participants. The number of study participants was less than 200 in 26 % of included studies, from 200 to 1,000 in 53 % of studies, and over 1,000 in 21 % of studies. Nearly all interventions had at least one study that measured alcohol use as a primary outcome (n = 37). Other common primary outcomes included binge drinking (n = 17), perceived alcohol norms (n = 14), consequences of alcohol use (n = 14), alcohol-related attitudes (n = 8), and alcohol-related knowledge (n = 6).

Table 1 provides a list of the CBIs (and corresponding articles) and how theory was used according to the classifications of “mentioned”, “applied” or “measured”. In addition, if the theory was applied to the intervention, a brief description of its application is provided. Similarly if it was classified as “measured” the measure of the theoretical construct was also listed. The CBIs in Table 1 all indicated use of a broad theoretical framework. Broad theories specify the relationship between a number of constructs and associated variables that explain or predict behaviors. Broad theories of behavior change take into account a number of complex contextual factors (e.g. social, cultural, economic, etc.) and inter-related sets of constructs that influence behaviors. CBIs that did not mention use of a broad theoretical framework are listed in Table 2. These CBIs typically mentioned use of a specific theoretical construct without reference to a broader theory, or intervention technique. In addition, sometimes a specific construct or intervention technique can be associated with more than one theory. For example, several of these CBIs mentioned that the goal of the intervention was to improve “self-efficacy”, a specific construct that is most often associated with Social Cognitive Theory [26], but is also incorporated within other theories such as the Theory of Reasoned Action [27]. We applied the same classification system to these CBIs with regard to mention, application and measure for the construct and/or techniques. For each CBI listed in Tables 1 and 2, the use of the theory or construct/technique are classified as (1) mentioned, (2) applied, or (3) measured (using at least one of the theoretical constructs).

The effectiveness of the CBI was also examined. Tables 3 and 4 provides information about the 83 articles associated with an includedCBI that reported study outcomes: the setting, participants, a brief description of the intervention, comparators and the primary outcome measures that were used to evaluate the effectiveness of the CBI. The measures listed in Table 3 and 4 are primary outcome measures and, in many cases, are different from those listed in Tables 1 and 2 which lists the measures of theoretical constructs which were often secondary rather than primary outcomes. For the outcomes listed in Tables 3 and 4, an asterisk denotes statistical significance (at the level of p ≤ 0.05) indicating that the intervention showed more favorable results than the comparator (e.g., lower alcohol use or frequency of binge drinking, greater negative expectancies related to alcohol use, etc.) Of the 42 CBIs, all but one [48] demonstrated improvements in alcohol knowledge and/or attitudes. In addition to these knowledge or attitude outcomes, the majority (62 %) of the CBIs showed significant reductions in alcohol related behaviors. The proportion of CBIs reporting significant behavioral outcomes was greater among those that used a broad theoretical framework (71 %) compared to those that targeted a specific theoretical construct and/or intervention technique (51 %).