We use the basic science of behavior change1,2 experimental medicine approach to efficiently test the effects
of four distinct behavior change techniques (BCTs). These BCTs are selected to engage one of the key
behavioral mechanisms of action (MoAs)--self-efficacy--for improving daily walking by at least 1000 steps in
those who are very sedentary and at elevated cardiovascular disease risk.3,4 We propose a randomized 25
factorial experiment to do so. Whereas we have hundreds of behavior change interventions aimed at physical
activity initiation, most test multi-BCT interventions, averaging six BCTs in each randomized controlled trial.6
Systematic reviews of these trials concluded that self-efficacy was in many cases the key proposed
mechanism of action that improved initiation.7-11 They also concluded that self-efficacy was significantly
improved by these multi-BCT interventions, as was physical activity.7,12 However, when they attempt to
conclude that one BCT uniquely engaged or changed self-efficacy, and that the change in self-efficacy in turn
increased the physical activity, they cannot do so.13 They cannot, because currently published physical activity
interventions--with one single exception--always have multiple BCTs involved. These interventions don't
quantitatively assess whether the BCT was used by the participant. These interventions don't assess the key
MoA and behavior repeatedly across time, so that a full mediation model can be appropriately tested and
analyzed. And, they don't estimate statistical power for conducting such a mediation model test, and so are
likely under-powered to detect mediation.
The Science of Behavior Change experimental method proposes that researchers identify the putative MoA,
engage this MoA through experimentation and/or intervention, and then assess the degree to which MoA
engagement produces behavior change. We used this approach to design an experiment in which eligible
sedentary adults will be randomized to one of 16 cells for five months of intervention. Innovations of this project
include: 1) identifying the individual effects of each of four BCTs on the hypothesized MoA, with a full factorial
design; 2) using sequential mediation analysis to address longitudinal mediators; 3) including sufficient
participants (N=480) to have adequate statistical power; and 4) quantitatively assessing whether the BCT was
used.
Sedentary behavior among participants at risk for CVD is highly prevalent, and is associated with increased
risk for morbidity and mortality. Identifying which unique behavior change techniques engage self-efficacy and
improve walking, and which ones don't, may allow future interventions to be more efficient and scientifically
informed to precisely target this important public health problem.