A process evaluation, with mediation analysis, of a web-based intervention to augment primary care exercise referral schemes: the e-coachER randomised controlled trial

Background: The e-coachER trial aimed to determine whether adding web-based behavioural support to exercise referral schemes (ERS) increased long-term device-measured physical activity (PA) for patients with chronic conditions, compared to ERS alone, within a randomised controlled trial. This study explores the mechanisms of action of the e-coachER intervention using measures of the behaviour change processes integral to the intervention’s logic model. Methods: Four hundred ffty adults with obesity, diabetes, hypertension, osteoarthritis or history of depression referred to an ERS were recruited in Plymouth, Birmingham and Glasgow. The e-coachER intervention comprising 7-Steps to Health was aligned with Self-Determination Theory and mapped against evidence-based behaviour change techniques (BCTs). Participants completed questionnaires at 0, 4, and 12 months to assess PA and selfreported ofine engagement with core BCTs in day-to-day life (including action planning and self-monitoring) and beliefs relating to PA (including perceived importance, confdence, competence, autonomy and support). We compared groups at 4 and 12 months, controlling for baseline measures and other covariates. Mediation analysis using the product of coefcients method was used to determine if changes in process variables mediated intervention efects on moderate to vigorous physical activity (MVPA) recorded by accelerometer and self-report at 4- and 12-months. Results: The internal reliability (Cronbach’s alpha) for all multi-item scales was>0.77. At 4-months, those randomised to e-coachER reported higher levels of PA beliefs relating to importance (1.01, 95% confdence interval (CI): 0.42 to 1.61, p=0.001), confdence (1.28, 95% CI: 0.57 to 1.98, p<0.001), competence (1.61, 95% CI: .68 to 2.54, p=0.001), availability of support (0.77, 95% CI: 0.07 to 1.48, p=0.031), use of action planning (1.54, 95% CI: 0.23 to 2.85, p=0.021) and use of self-monitoring (0.76, 95% CI: 0.19 to 1.32, p=0.009) compared to ERS alone. There were no intervention efects on autonomous beliefs or perceived frequency of support, compared to ERS alone. At the 12-month follow-up, participants belief in the importance of PA was the only process measure to remain signifcantly higher in the e-coachER group when compared to ERS alone (0.75, 95% CI: 0.05 to 1.45). Intervention efects on perceived importance (2.52, 95% CI: 0.45 to 5.39), action planning (1.56, 95% CI: 0.10 to 3.54) and self-monitoring (1.92, 95% CI: 0.21 to 4.33) at 4-months signifcantly mediated change in accelerometer measured MVPA at 12-months (recorded in≥10-min bouts). Conclusions: e-coachER led to some short-term changes in most process outcomes. Some of these processes also appeared to mediate e-coachER efects on changes in accelerometer measured MVPA. Further work should be carried out to understand how best to design and implement theoretically underpinned web-based physical activity promotion interventions within ERS. Trial registration: ISRCTN, ISRCTN15644451