208 Chapter 7 fictional intervention in case i belongs to the control group. are quarterly calendar time dummies for each quarter and can capture business cycle and other time calendar time effects. is an indicator taking the value 1 if the individual is in the ‘Mentorwijs’ group. are time dummies representing the month compared to the start of (fictive) treatment. is the error term. and are parameters and is the effect of analysis time. Note that controls do not necessarily have to start in the same month as ‘Mentorwijs’ cases, meaning that calendar time and analysis time can differ. is the parameter of interest, the estimate of the effect of ‘Mentorwijs’. The beta is the difference in the change of the outcome between the intervention and control group in month t, with respect to the baseline measurement. For aim 2 we used the same main model but with interaction effects for subgroups. For aim 3 we used mixed modeling in SPSS statistics 26 to estimate the change after ‘Mentorwijs’ on all outcome measures among supervisors, wherein time was used as a categorical independent variable and T0 was used as the reference category (model 1). This technique deals better with missing data than generalized estimation equations (GEE) and considers that repeated measurements are correlated (19). In a second model we tested for the following possible confounders 1) demographics (i.e. age, educational level, and sex), 2) number of years of experience with the guidance of employees with a work disability, 3) number of years employed at current employer, 4) company size, 5) number of employees they guide at the workplace and 6) number of employees with a work disability they guide at the workplace. Only confounders that changed the beta of the independent variable (i.e. time) with more than 10% were added to the model (model 2). In both models we estimated Beta coefficients (B) and 95% confidence intervals (CI). For aim 4 process evaluation data were analyzed using descriptive statistics (i.e. mean (SD) and percentage). Results Participants We included 127 supervisors that followed ‘Mentorwijs’ and 118 employees with a work disability who were guided by these supervisors. Not every employee gave consent to be identified in the register data and not every employee could be identified in the register data. Therefore, register data were collected from 78 employees. Four employees were excluded from matching with controls, as they were not registered as having a job at baseline. One employee could not be matched with controls. In the end, 73 employees were matched with 1.526 controls. Figure 1 shows the flow diagram of the selection process of supervisors and employees in this study.
RkJQdWJsaXNoZXIy MTk4NDMw