Mariken Stegmann

invited from different fields, ensuring variety in characteristics such as age, gender,  academic qualification and place of employment. Patient associations in the Netherlands  were approached to identify patients. Given the number of sample characteristics, we aimed  to include 20 GPs, 20 oncology specialists and 15 patients after the final round.  Data collection and analysis  The Delphi procedure was performed over at least two rounds, with participants using the  online Qualtrics tool, version 2005, to complete questionnaires (Qualitrics, Provo, Utah,  USA). The patient panel received a list of definitions for the items used in the questionnaire.  During Round 1, we also sent a short questionnaire about demographic characteristics and  preferences for the layout and writing style of the respective letters. Descriptive statistics  were applied to the answers of these questions. All data were analysed using IBM SPSS for  Windows, version 22.0 (IBM Corp., Armonk, NY, USA). Delphi procedure  Round 1  Participants were asked to rate each list item on four‐point Likert‐type scales (not relevant =  1 point; moderately relevant = 2 points; relevant = 3 points; and highly relevant = 4 points),  with a fifth “no opinion” option (0 points). Participants were provided with a free text area  at the end of the questionnaire in which they could propose additional items. The median  scores for each item were calculated for the panels combined and for each panel  individually. If the difference between the median scores of the GPs and the specialists was  ≥2 points, the item score of the receiving party was weighed more heavily (weight factor  1:2). Items with median scores <2.5 were considered  not relevant. Items with median scores  ≥2.5 were included in the concept consensus list: those with scores ≥2.5 but <3.5 were  considered less relevant and those with scores ≥3.5 were defined as highly relevant.  Additional items were added to the concept consensus list if more than one participant  suggested the same item.  Round 2  Participants were asked to indicate whether they agreed (yes/no) with inclusion of the highly  relevant items on the consensus list. Based on the content validity index (CVI, which is the  proportion of respondents who agree with the proposed relevance of the item), if ≥70% of  respondents disagreed with a proposed item, it was excluded from the final list [13,15]. For  all items scored as less relevant in the first round, participants were asked to indicate  whether they agreed (yes/no) with their exclusion from the consensus list; if ≥70% of  respondents disagreed with this proposal, the item was included in the final consensus list  [12,13]. Participants were not asked to review items with low relevance a second time. If  ≥70% of the items included in round two reached a CVI of ≥0.70, an additional (third) round  was not deemed necessary. Agreement (CVI) between the panels was calculated as the  percentage of items in which the different panels separately reached the same decision.  Results  Composition of the expert panels  In both rounds, GPs (n = 39; 42%), medical specialists (n = 42; 41%) and patient  84 Chapter 7