Mariken Stegmann

Codes concerning the same topic were grouped into a theme. For each tumour site (lung,  breast or colorectal), a second researcher independently analysed random samples of the  documents. Agreement between the researchers was calculated as the percentage of  exactly corresponding codes. Any discrepancies were discussed until consensus was  reached, and if necessary, the coding list was refined in an iterative process. In the  interviews, two researchers independently identified relevant text fragments.  Sampling was continued until saturation, which was defined as the moment that no new  codes were identified. When saturation was reached for one tumour type in one hospital,  we analysed two more medical files to ensure saturation. All documents and notes were  anonymously uploaded in Atlas.ti (version 7.15.12). Any discrepancies were discussed with  a third member of the research team until consensus was reached. Finally, the results were  discussed in a multidisciplinary meeting with all co‐authors.  Results Sample characteristics  All six departments that we approached agreed to collaborate. In total, 50 patient files  containing 419 documents were analysed and nine healthcare professionals were  interviewed. The purposive sample of interviewees consisted of four GPs, four oncology  specialists and one specialist oncology nurse. The characteristics of the included patients and  the interviewed healthcare providers are  detailed in Table 1.  Table 1 . Sample Characteristics  Patient files (%)  n = 50 Interviews (%)  n = 9  Patients  Gender   Male  19 (38%)  Tumour site   Lung   Mamma   Colorectal  20 (40%)  16 (32%)  14 (28%)  4 (44%)  3 (33%)  2 (22%)  Age (mean, SD)  65 (10)  Tumour stage   Stage I   Stage II   Stage III   Stage IV   9 (18%)  11 (22%)  16 (32%)  14 (28%)  Oncological healthcare providers  Gender   Male  3 (60%)  General practitioners  Gender   Male  3 (75%)  6 69 Qualitative mixed-methods analysis of correspondence