To assess interaction, we performed three analyses to determine: (i) what themes were addressed by multiple stakeholder groups, (ii) which participants interact with each other, and (iii) how participants interact. All these analyses were performed after the data were collected, and supported by Atlas.ti software (V.9). For the first analysis, we calculated the percentage of the themes that was addressed by two or three stakeholder groups in each data-collection method. For example, if both HCPs and policy experts addressed theme X, this theme was marked as ‘shared’, and consequently the percentage of shared themes of the total amount of themes could be calculated. For this comparison, all methods were taken into account, even though participants did not directly interact in surveys. For the other two analyses, we examined how participants interact. To this aim, we adapted coding schemes by Morgan & Hoffman [
28] and Keyton [
29], which can be applied on verbatim transcripts to code
how people react. Our main aims for this analysis were to identify how extensive topics were addressed, and to identify whether stakeholders could reach forms of agreement while interacting. The codes as described by Morgan & Hoffman [
28] were most applicable for these aims so we have used their code book, whereas Keyton [
29] more concisely describes how different turns in an interaction can be coded, so we have adapted her coding process. The main change in the coding book that we made, is that we distinguished who interact from the ways how people interact, which was not clearly incorporated in Morgan & Hofmann’s code book (see Appendix
3). We coded who the speaker is and to whom this speaker addresses the message, which will be referred to as
turn. When a moderator asked a question to participants, or vice versa, this is coded as a ‘moderator-participant’ turn, and when participants ask and answer questions to each other, these are coded as ‘turn between participants’. When a moderator presented views of other participants, and when participants reacted, these were separately coded as a ‘turn between moderator and participant where moderator presents view of others’. Besides
turns we also coded how participants interact in the conversation, using the codes ‘question’, ‘answer’, ‘expansion: sharing new aspect of previous topic’, and ‘agreement’, which will be referred to as
acts. Acts are mutually exclusive, but multiple acts could be assigned to one turn in a conversation and vice versa, for example when a participant gives an answer to a moderator and directly poses a question the moderator. KW performed the coding process. These analyses were discussed in-depth with MT and RR, to ensure consistency and minimise a possible bias.