Stefan Elbers

192 Chapter 7 Intervention costs will be determined by the standardized cost prices for rehabilitation treatment (Hakkaart-van Roijen et al., 2015). Other health expenses will be obtained using the iMTQ at 3 and 6 months posttreatment. This questionnaire includes visits to health care providers, prescribed and over-the-counter medication, and alternative health care. We will also calculate the productivity loss due to pain-related absence from work, adopting the gross human capital approach (Pike & Grosse, 2018). Productivity loss will be obtained with two questions of the iPCQ at 3 and 6 months posttreatment. Total costs are calculated using the Dutch manual for cost analysis in health care research (Hakkaart-van Roijen et al., 2015; Kanters et al., 2017). Following the procedure of Den Hollander and colleagues (Den Hollander et al., 2018), a standardized cost price will be used for each hour of productivity loss. Total costs and total health gains for each condition will be used to calculate the incremental cost-effectiveness ratio (ICER). Furthermore, we will construct cost-effectiveness acceptability curves based on mean costs and using incremental costs and incremental effects, employing nonparametric bootstrapping with 5000 replications. This will result in a scatter plot over four quadrants, where each quadrant indicates a different implication for economic evaluation (i.e. a combination of positive or negative costs and effects) (Fenwick & Byford, 2005). According to the National Institute for Health Care and Excellence guidelines, all intervention costs in cost-effectiveness analyses should relate to health care or social services funding ( NICE , 2013; Li et al., 2018). This excludes the development costs of the AGRIPPA app because this project has been funded with research grants. However, to account for future development and maintenance costs, we will perform a sensitivity analysis and explore various cost scenarios. We will calculate multiple ICERs, each with a different cost input value that corresponds to a possible future pricing scenario (e.g., subscription, pay to download). Missing Data. Following recommendations of Twisk and colleagues (Twisk et al., 2013), we will perform the multilevel analysis on incomplete data, rather than using multiple imputation procedures. However, we will use the R MICE package to search for patterns of missing data across the included variables and to perform t tests to explore the relationships between the amount of missingness of each variable and all other variables (Tabachnik & Fidell, 2013). Participant Timeline Patient eligibility screening, informed consent, and treatment allocation procedures, as well as the baseline assessment, will be completed prior to the start of the treatment progamme. Patients in the enhanced treatment condition will receive instruction on how to download and use the app, and both strategies will be explained by a member of the treatment staff.