Scottish Cancer Patient Experience Survey 2015/16: Analysis of Free-text Comments

Appendix B Data analysis process

Free-text comments were analysed by a team of three researchers with significant experience in qualitative analysis. The data were analysed using Thematic Analysis (Braun and Clarke, 2006), employing an inductive approach – coding and theme development were driven by the content of the free-text comments. The process involved identifying commonalities in the data, and searching and comparing the data to identify and record relationships and themes. In order to apply some structure to the large volume of data, the following steps were undertaken during analysis:

1. The complete data set was split into responses by comment box, creating seven data-sets.

2. For each of the individual comment box data-sets, one researcher familiarised themselves with the data by reading all the responses. During familiarisation, the researcher made a note of potential codes for that individual data-set by identifying recurring words or units of meaning (positive, negative or neutral observations). A second researcher familiarised themselves with a sample of the responses and the two researchers discussed and agreed the coding decisions.

3. The researcher sorted all responses in each individual data-set by allocating responses into the following top level categories (i) positive comment; (ii) negative comment; (iii) factual/neutral comment; (iv) irrelevant/miscellaneous comment; (v) both positive and negative comment. This process enabled the team to gain an overview of the nature and emphasis of comments made.

4. The same researcher then applied detailed codes to all responses in each individual data-set. Because the codes had been derived inductively from the responses to comment boxes, the coding sheet was different for each individual data-set. Comments were given as many codes as were appropriate to cover the content of the comment. In total, there were 174 codes across the seven comment boxes, for example speedy action; uncomfortable environment; rude or insensitive communication.

5. Each individual dataset was then split into separate sheets containing all comments for every code. All comments assigned to every code were then re-read to check for consistency of meaning within the code. During this process of constant comparison, where comments were not seen as a good 'fit' with the code, either the code was refined to reflect the comments within that code more meaningfully, or the comment was moved to an alternative code. At this stage, it emerged that an additional code for 'family history not taken seriously' was needed. The content for all comment boxes was then searched again, using the key words 'family' and 'history' and an additional coding sheet was created for relevant comments.

6. A second researcher then checked 5% of all comments for consistency in terms of splitting the data into top level categories, and coding decisions. Any discrepancies or disagreements (of which there were only a small number) were discussed and codes adjusted as necessary.

7. Two researchers then worked together to identify similar codes across the seven comment boxes, in order to look for common themes across the whole data-set ( i.e. across all stages of the cancer journey represented in the SCPES). The researchers amalgamated codes which shared similar meaning into sub-themes.

8. The research team then mapped subthemes into overarching themes which described the main issues highlighted in the data.