Commentary on Kelly et al. (2012): Multiple mediation-many mechanisms signal more questions

Abstract

Using very sophisticated multiple mediator analytical techniques, Kelly et al. [1] have shown empirically that the effects of Alcoholics Anonymous (AA) meeting attendance on abstinence are explained largely by changes that occur in members' social networks, namely: more abstinence influences and fewer heavy drinking influences (which together explain 40% of AA's effect in Project MATCH's aftercare sample and 64% of AA's effect in the less severe out-patient sample). A third mediator, a strengthening in ones' ability to say ‘no’ to drinking in social situations, further fleshes out a story of very effective peer learning in the AA context. The interplay between both negative and positive social networks (i.e. heavy drinkers and abstainers, respectively) is especially important, as it shows that AA works both by reducing the ‘wet’ influences and by increasing the ‘dry’ influences, with both phenomena occurring as the number of meetings increases. Treatment clients often ask whether they will have to jettison their existing friends and family who are heavy drinkers. Kelly et al. show that gains in abstinence can occur without such a seemingly drastic move, by adding only individuals to the network who are supportive of abstinence; but that any reduction in the number of heavy drinkers will yield even more abstinence. Complementing those findings, work on AA mediators from another study has found that having support for less drinking from non-AA members is not associated with abstinence, whereas such support from AA-based social network influences is [2]. This adds still more specificity regarding the type of network changes needed to sustain recovery, and suggests that treatment clients should cultivate a relationship with at least one AA member. One question that arises from my reading of their paper is their rationale for exploring both abstinence and heavy drinking outcomes in the context of AA effects, given that AA's goal is to help members stop drinking (and not to help them simply drink less heavily). Perhaps future analyses could be conducted among the out-patient sample (which was followed for a longer period) to determine whether reductions in heavy drinking lead ultimately to eventual abstinence (and whether that was mediated by subsequent AA attendance). This would be consistent with the now dominant view of a cyclic nature of recovery. Seeing the Kelly et al. results, one cannot help but wonder whether treatments can be developed that can yield similar gains in abstinence for those clients who are resistant to AA and other 12-Step groups. One approach to consider is 12-Step facilitation, such as Project MATCH's 12-Step facilitation [3], or the author's Making AA Easier/MAAEZ [4]. Another is network support facilitation, such as Litt's Network Support intervention [5] that does not rely exclusively on the 12-Step fellowship for the social network mechanism. Interesting mediational effects have been found for these approaches, and have included pro-abstinence influences (as was found here) as well as increases in having a 12-Step sponsor, doing service and feeling comfortable being around 12-Step groups [6,7]. I also encourage moderation analyses that can isolate mechanisms unique to particular subgroups known to be especially challenging to treat, such as those with high levels of prior treatment, prior AA or comorbid psychiatric severity. None.