68 Chapter 3 MCD sessions, with two versions: 1) for participants (Appendix 3), including 19 closed and 5 open questions (of which 16 and 3 where focuses out MCD outcomes and are used in this study), and 2) for facilitators (in Appendix 4), including 10 closed and 7 open questions (of which 9 and 5 are outcome-related and are used). Items from these forms refer to the different steps of the used dilemma method (see Appendix 2). The forms are not validated but were based on earlier developed forms, e.g., as used by Janssens et al. (2015) and Seekles et al. (2016). The closed questions consisted of a 5-point Likert scale or rated scores between 0-10. Every MCD session per team was given a series of ascending numbers, from 1 to 10, that we used to compare the influence of the number of sessions with evaluation scores. Our second type of questionnaire, i.e., the survey-after-the-series (forthcoming), was directed at participants and conducted after each MCD series. This questionnaire asked whether respondents participated in MCD at their location during this research, if so, 1 open and 6 closed evaluative questions were asked. Furthermore, we conducted mid-term interviews, between 30-60 minutes, with prison staff who were MCD participants (n=19), local managers (n=12), and 3 focus group meetings with MCD facilitators. All were audio-taped and transcribed. Transcripts were mainly used as a monitoring instrument during the implementation process, generating input for the steering committees and facilitators’ peer meetings. Inclusion and exclusion of data Before the analyses of the initial 148, sessions were excluded that were clearly understood to have used other conversation methods (n=3). One team quit participation; therefore, we excluded their MCD sessions (n=4). When DCIA announced the closure of one of our research locations, most sessions that were still planned at that location got canceled. However, the few MCD sessions that continued, in that location before the closure, showed to be dominated by frustrations instead of focusing on a moral dilemma; hence we excluded these sessions (n=3). After the first analysis phase, we again excluded sessions (n=7) that used a conversation method too deviant from the dilemma method. Required steps of the dilemma method were a concrete dilemma, discussed perspectives, and related values. Data analyses We used an embedded mixed methods design (Creswell, 2014, p. 44), which is needed for a responsive evaluation to provide an insightful understanding of data (Abma, 2005). The qualitative data enhances the insights gathered by quantitative data. We conducted all qualitative analyses via MAXQDA® software, version 2020. The written answers of