Darcy Ummels

Patients’ experiences with commercially available activity trackers embedded in physiotherapy treatment | 49 3 Data analysis The data were analyzed according to the Framework Method. 40 In stage 1, all audio recordings were transcribed verbatim by two researchers (DU, KT) and primary identifiers were de ‐ identified. In stage 2, the entire transcripts were read and, if needed for the context of the text passage, the audio recordings were reviewed. In stage 3, paraphrases or labels (a code) were applied to relevant text fragments. Deductive and inductive content analysis was used. 41 Deductive content analysis was used for the majority of codes predefined according to empirical information. 16 ‐ 19,22 ‐ 24,29,30,42 ‐ 47 Inductive coding was applied when the text passages did not fit a predefined code but were considered to be relevant. An ‘ other ’ code was defined to include such data. During the analysis, two subcategories were added to the framework: choice of activity tracker and discussing results with a physiotherapist. In stage 4, the first two interviews were coded by both researchers (DU, KT). An alignment session was held with the two researchers and one other independent researcher (EB) to fine ‐ tune the coding. Differences in interpretation were solved by a dialog between DU and KT to reach consensus, because the aim was the find a suitable interpretation grounded in the original quote. The codes were grouped together into categories and subcategories by the research team, using a tree diagram. Expert assessment of this working analytical framework took place in four iterative expert meetings with two participant representatives, the research team, and two mHealth experts. In stage 5, the analytical framework was applied by indexing subsequent transcripts using the existing categories and codes (see Supplement File S3.2). In stage 6, the data were entered into the framework matrix from a spreadsheet including a summary and a reference to illustrative quotations. In addition, an analytical tabulation was performed. Finally, in stage 7, the data were interpreted and presented in a descriptive way using NVivo (version10). Descriptive statistics of the participants ’ characteristics were presented. Means (± standard deviation) or medians (range) (depending on the data distribution) were given. If data were missing, pairwise deletion was applied. Trustworthiness To ensure the quality and trustworthiness of this study, credibility and transferability were checked in several ways. 31 Credibility was examined by method, investigator and data triangulation. Method triangulation involved multiple methods of data collection (interviews and focus group discussions); investigator triangulation was achieved by having all authors reflect on the design, data collection and analysis process during this study; and data triangulation used different sources of the same information (multiple