Action and inaction in multi-behaviour recommendations: a meta-analysis of lifestyle interventions

ABSTRACT

This meta-analysis examined theoretical predictions about the effects of different combinations of action (e.g., start an exercise regime) and of inaction (e.g., reduce screen time, rest in between weight lifting series) recommendations in smoking, diet, and physical activity multiple-domain interventions. The synthesis included 150 research reports of interventions promoting multiple behaviour domain change and measuring change at the most immediate follow-up. The main outcome measure was an indicator of overall change that combined behavioural and clinical effects. There were two main findings. First, as predicted, interventions produced the highest level of change when they included a predominance of recommendations along one behavioural dimension (i.e., predominantly inaction or predominantly action). Unexpectedly, within interventions with predominant action or inaction recommendations, those including predominantly inaction recommendations had greater efficacy than those including predominantly action recommendations. This effect, however, was limited to interventions in the diet and exercise domains, but reversed (greater efficacy for interventions with predominant action vs. inaction recommendations) in the smoking domain. These findings provide important insights on how to best combine recommendations when interventions target clusters of health behaviours.

KEYWORDS:

• Multiple behaviour domain change
• lifestyle intervention
• multi-domain intervention
• multi-behaviour intervention
• action
• inaction

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1 When ethnicity data were not reported and countries were highly ethnically homogeneous (e.g., the Netherlands, Italy), we obtained the information from population reports from those countries.

2 When the N at the pretest differed from the N at the post-test, the smaller N was used.

3 The distribution of weights was skewed due to 16 cases with large weights. To correct for this, we curved weights over 999 to fit between the range of 1000–2000.

4 We controlled for whether samples were self-selected, as self-selected samples were associated with stronger improvements in overall change than samples that were not self-selected, fixed-effects QB = 91.99, p < .001, k = 216.

5 The weighted-mean effect sizes for the multiple behaviour intervention, single-behaviour intervention and no-intervention control groups have been reported elsewhere (Wilson et al., 2015). However, the main analyses reported in this article have not been included in any other article.

6 Attempts were made to compute a more precise measure of the balance of action and inaction recommendations. Because little was gained due to empty cells in our analyses, we present only results from this measure of the balance of action and inaction recommendations.

Additional information

Funding

This research was supported by Grant R01 MH094241 (Dolores Albarracín) from the National Institutes of Health.