Stefan Elbers

191 AGRIPPA: protocol for an RCT repeated observations and treatment locations, we will perform a multilevel analysis. In our hierarchical model, time points (level 1) will be nested in patients (level 2) and patients nested in treatment locations (level 3). Our main analysis will include the effects of time, treatment condition, treatment location, and the interaction between time and treatment condition, with a random intercept for patients. In addition, the model will be adjusted for sex, age, pain intensity, and pain duration. In the case of a significant 2-way interaction between time and treatment condition, posthoc contrasts between the treatment conditions at 3, 6, and 12 months will be calculated. We will also perform a subgroup analysis for patients that report a positive treatment effect at t 1 (i.e. a GPE score of 1 or 2). Analysis of Secondary Outcomes Perceived Usability and App Engagement. Based on the results of the feasibility study and the nature of the behaviour regulation strategies, we expect an engagement pattern in frequency, type, and depth of engagement that differs for each behaviour change strategy and changes over time. Specifically, for Insight Cards, we believe that patients will actively engage with this component during treatment (i.e. creating Insight Cards during use), but shift to more passive engagement (i.e. reading the input, but only creating new content at limited occasions) during follow up. For Value-Based Goals, we expect that the formulation of goals and steps will increase during the final part of treatment, together with a growing emphasis within the treatment progamme on integrating newly learned strategies into daily life routines. After treatment, we anticipate that patients will engage in a reflective (e.g., documenting progress) and active (i.e. formulating new goals) manner with this strategy. In general, we expect a decreasing trend of the number of logins over time, but an increase in the “depth” of use (i.e. the average number of features accessed per login) as well as an increase in the duration of a login. We will calculate descriptive statistics to explore patterns of engagement. Furthermore, to examine the extent to which user engagement and usability are negatively associated with the change of pain disability during follow up, we will perform a multiple regression analysis, with the change score of pain disability (t 4 – t 1 ) as the outcome variable and engagement and usability measures as predictors. We will adjust for age, sex, pain intensity, and baseline PDI. Cost Effectiveness. To investigate the efficiency of the intervention, we will perform a cost-effectiveness analysis at 3 and 6 months posttreatment according to the intention-to- treat principle.We hypothesize that patients in the enhanced treatment condition will have more quality-adjusted life years (QALYs) relative to the health care expenses compared to the regular treatment condition within the 6-month study period. Expected health gain will be expressed in QALYs and calculated using the procedure of Brazier and colleagues (2004) to estimate the 6-dimensional health state form (SF-6D) using the SF-12 assessment at 3 and 6 months posttreatment (Brazier & Roberts, 2004; Cunillera et al., 2010).