3.6. Data collection

A semi‐structured interview guide (Polit & Beck, 2018) with open‐ended questions was informed by previous research, the self‐determination theory and through discussions in the research group. The interview guide was tested on a daughter who met the inclusion criteria and she contributed to the interview guide and the interview itself. For example, she found it hard to answer the vague question of why she had put so much effort in helping her father. She recommended breaking the question into smaller themes to encouraging participants to start talking about their motivation. She requested a more precise question about her relationship to her father before and after the dementia diagnosis; her knowledge of dementia and of her father's symptoms; and the importance of being able to decide for herself what she should and should not do as a caregiver. Her feedback led to a more detailed and expanded interview guide [see Appendix S1]. This pilot interview contributed with rich data and was therefore included in the sample.

Data were obtained from individual face‐to‐face interviews (Polit & Beck, 2018) conducted by HD in 2017. HD is an experienced nurse trained in interviewing and consulting elderly patients and their caregivers. All participants were informed that she was a PhD student writing her PhD thesis on caregivers. The interviewer was unknown to all the participants except for the pilot interview; that participant was an acquaintance. The same interview guide was used in all interviews except for the pilot interview. All participants were interviewed once and each interview lasted from 56 min to 1 hr and 47 min; the median length was 1 hr and 11 min. The data material was digitally audio‐recorded and later transcribed, mainly by HD. A professional transcriptionist was hired to transcribe parts of the last seven interviews. The transcripts were not returned to participants for comments as their intuitive experiences were considered essential for the research question. Permitting elaboration and corrections could have resulted in a data set that did not represent spontaneous answers to the interview questions.

The recruitment process lasted for several months. The participants were interviewed until no new relevant knowledge was generated (i.e. after 21 interviews). The research group discussed the point of data saturation (Polit & Beck, 2018). A bias in retrospective interviews might be that some participants easily could switch from the past to the present time and talk about the parent in the nursing home. Therefore, there was a risk of obtaining irrelevant information. This was handled in the interview situation by asking the participants to recall examples from the home‐dwelling period. Data on caregiving after the parent had moved to a nursing home were excluded from the analysis because it was not relevant to the research question.