Xuxi Zhang

and anxiety. Each dimension has five levels, ranging from no problems (level 1) to serious problems (level 5). Hence, the EQ ‐ 5D ‐ 5L has 3,125 possible health states. Utility values for these health states are available for the study sites of each participating country. 34 As part of the EQ ‐ 5D ‐ 5L, participants are also asked to indicate their experienced current health state on a visual analog scale, 0 being the worst imaginable health and 100 being the best imaginable health. Additionally, we will evaluate healthcare utilization and productivity losses. Healthcare utilization is measured with four questions from the Self ‐ Management Resource Center (SMRC) Health Care Utilization questionnaire regarding doctor appointments, the use of hospital emergency rooms and hospital admissions. 35, 36 Productivity losses are measured with two domains from the Productivity Costs Questionnaire (PCQ) 37 : lost productivity at paid work due to absenteeism (6 items) and lost productivity at unpaid work (3 items). Socio ‐ demographic characteristics include age, gender, country of birth, marital status, household composition, education level, employment situation and health conditions. There is an open box at the end of the questionnaire for any additional remarks. The follow ‐ up questionnaire at 6 months (T1) will be identical to the baseline questions except for the addition of questions on the satisfaction of the target population with the intervention. In the T1 ‐ questionnaire, we will add 6 items to rate the satisfaction with the whole SEFAC intervention as well as specific concepts (mindfulness, social engagement and ICT support) on a scale from 1 to 10. Power considerations The power considerations are conducted according to the methods of a previous study. 38 We will include net 113 participants at T0 in each study site (4 study sites * 113 = 452 study participants). When the loss to follow ‐ up between T0 and T1 will be 20%, we will have complete data of 360 participants at T1. Assuming equal standard deviations (SD) at T0 and T1, an alpha of 0.05 and power of 0.80, and taking into account the cluster design (4 participating study sites) with an average cluster size of 90 participants (360/4) and an intra ‐ class correlation coefficient of 0.02, a difference of 0.24 SD between T0 and T1 can be established regarding the continuous outcome measures for this expected sample size and under these conditions. For instance, regarding HRQoL as measured by the SF ‐ 12, a difference of 2.74 points can be established between T0 and T1 for the PCS (SD = 11.4) and 2.86 points for the MCS (SD = 11.9). 39 Statistical analysis Descriptive statistics will describe characteristics of participants in the total study population and in each study site. In order to evaluate differences between T0 and T1 measurements, multiple linear regression analyses (for continuous outcome variables) and multiple logistic regression analyses (for dichotomous variables) will be adopted in the total study population. In addition, the analyses will be done for each study site separately, and possibly 212 Chapter 8