207 Training for supervisors to guide employees with a work disability efficacy, intention to adopt and actually applied an open and involved attitude towards employees with a disability. An item to measure skills was, for example, that we asked supervisors whether they have self-efficacy, intention to adopt and actually applied a supporting environment at the workplace for employees with a work disability. Reliability and validity were not tested, but items for each outcome measure were based on the ‘Mentorwijs’ theoretical handbook (2). The items in the questionnaire were aligned to the defined objectives and expected results in this theoretical handbook. Process measures (only measured after the training – T1) focused on factors that could affect the implementation of the training in practice: 1) dose delivered – i.e. to what extent was the intervention implemented as planned, 2) dose received – i.e. number of meetings followed, 3) satisfaction towards the training, 4) extra time spend on the guidance of employees with a disability, 5) and contextual factors on the level of the supervisor and organization, which were based on an existing instrument to measure determinants of innovations (17). Statistical analysis For aim 1 we applied a difference-in-difference estimation to the matched sample in Stata 14, which allowed us to estimate the causal effect of ‘Mentorwijs’. The difference-in-difference estimation together with matching corrects for potential pre-treatment differences between the ‘Mentorwijs’ and control group. A similar approach has been followed by De Graaf-Zijl et al (2020) (18). In the analysis, every person in the control group is weighted according to their propensity score. The use of difference-in-difference techniques is only allowed if there is a common trend between Mentorwijs and the control group prior to the intervention. Tests showed that a placebo effect of Mentorwijs 6 months before the actual start of the intervention was not statistically significant for any of the outcome measures. This implies that the common trend hypothesis for using the difference-indifference design has not been violated. The model specification is: Where i is the individual employee and t calendar time. Yit is the outcome of interest (employment status) for individual i in month t. Individuals have to be employed in month 0. Month 1 is the month of the end of the intervention or 7
RkJQdWJsaXNoZXIy MTk4NDMw