43 Moral dilemmas of Dutch prison staff based on explicitly mentioned words. To reach consensus on the formulation and coding style, we used a pilot of 30 cases where four researchers independently analyzed 10 cases before discussing the final coding. Subsequently, two junior researchers independently analyzed all cases, discussed their viewpoints, and together formulated final themes. In case of disagreement or doubt a senior researcher was consulted. We created reliability and credibility of the analyses by this triangulation of researchers (Green & Thorogood, 2013). In between analyses, to prevent bias, we were constantly reflective about the data. We were intent on increasing the reliability by being transparent about our methods and its challenges. In case of doubts about the correct interpretation, information from the facilitator-forms was often used to lead the interpretation, as these usually offered more detailed information than forms of the MCD participants. In addition to collecting themes and subthemes, we wrote down observations regarding the formulation and context of the dilemma. After consulting such additional notes with the three researchers (including one senior), we worked toward consensus in order to come to final categorizations. This triangulation of data and researchers – complementing each other with additional information about the cases – led to a stronger credibility of this data analyses. Our coding method asked for intermediate and ongoing consultations by several researchers, therefore it was not possible to calculate inter-rater reliability scores. During the second step, we used the MindMeister program to create a mind map in which we further categorized main themes, by means of a more deductive process. We included abstract concepts that could accommodate multiple coded themes, e.g., ‘professional responsibility’ or ‘deviation from protocol’. We merged themes that clearly overlapped. Newly created main themes were expanded with layers of subthemes, based on step one of our analyses. We verified that every case was correctly covered by its main theme, as well as the names of themes. The search for meaningful and context-relevant categories resulted in 9 main themes, and 2 to 5 subthemes per main theme. Furthermore, we included an overview of the ‘organizational levels’ (Table 1) which provided insight into the organizational level at or between which the dilemma occurred, e.g., between a team and their supervisor. Each subtheme showed to have multiple ramifications based on this organizational structure. We checked our additional notes about observations regarding the formulations of dilemmas, to reach consensus on conclusions. During previous steps we were unaware of which team mentioned which dilemma. For our last step of analysis, we added team names, and therefore the professions of participating staff, to examine whether moral dilemmas and themes were associated with 2