Chapter II: Method
2.4 Data analysis
2.4.2 Coding
Interpretive Phenomenological Analysis (IPA; Braun & Clarke, 2006; Giorgi & Giorgi, 2003; Willig, 2001) was adopted as the method of choice for coding and analysis. The following stages of IPA were used:
1. Familiarisation with the text. Each individual transcript was entered into a Microsoft Word file, printed in hard copy and read over twice, to obtain an overall sense of the data. Broad notes were made in the right hand margin of the text while the researcher reflected on the content of the text.
2. Identification of themes. Inductive and deductive coding of raw data transcripts were conducted in an iterative manner after each interview was completed. More specifically, using Word file documents which were all merged into one Word file, transcripts were read line by line, section by section. During this process, the identification and organisation of information into categories, and broader subthemes and major master themes was carried out using descriptive labels. These short thematic labels and descriptions were entered in the right hand margin of the word file as a „Comment‟. These labels often corresponded to the topics outlined in the interview guide or psychological concepts such as support, emotional functioning and loss that reflect the
„psychological meanings‟ that are explicitly or implicitly articulated by caregivers and stroke survivors (Giorgi & Giorgi, 2003).
To ensure the reduction of subjective biases in analyses, coding was carried out by the investigator and an independent experienced qualitative coder, and occurred at different phases across data collection. The independent researcher coded ten interviews all together, with the first 6 interviews of the study coded initially, developing a list of broad themes. The investigator (who also coded the first six interviews) and independent coder then met to discuss and compare their independent findings and resolve any disparities in themes/definitions generated. Ninety percent agreement in themes was initially found (although sometimes different labels were given by the two coders) before the remaining discrepancies were then resolved until a common theme list was constructed. Coding was also conducted by the independent researcher during the middle (two interviews) and final phases of data collection (two interviews), and compared with any new themes found by the chief investigator to ensure consistency and that no important information was overlooked (see Appendix D for the original two theme lists generated by the qualitative coders).
Master themes, subthemes and categories were then recorded in a working Microsoft Excel spreadsheet summary table. This summary table also contained key descriptive labels, brief quotes and the corresponding source (see Appendix E for a copy of the final excel spreadsheet used by the chief investigator and the final theme list generated by the independent coder). In conjunction with this summary table, constant organising/re-assignment, comparison and refining of coded data were carried out in a cyclical manner. This was done by comparing emerging master themes, subthemes and
categories that were being coded in new interviews with existing themes in a back and forth movement between raw data and the thematic summary table. This ensured only new independent master themes were added to the overall summary theme list, and that emerging subordinate sub-themes and categories were systematically integrated under these master themes.
3. Data Saturation. As previously discussed, qualitative data collection is generally said to end when no new theoretical insights emerge that would further develop a conceptual theme (Bowling, 2002; Strauss & Corbin, 1990). Unfortunately, most if not all of the qualitative literature in this area does not actually specify the process and criteria by which saturation of data are reached (Francis et al., 2009). In response to this gap, Francis and colleagues recommended procedures by which saturation of themes could indeed be systematically conducted and reported, and hence sample size more adequately justified. In accordance with these procedures, once an initial sample size of 10 was reached, analysis of saturation using the coding summary/paradigm outlined above and in Appendix E was carried out. A stopping criterion of 3 was initially set; that is, data collection was planned to stop when no master themes emerged from the data after three consecutive interviews. However, because the recruitment and data collection was slow in the earlier stages and fast at the later stages, the coding of an interview was not always completed before the next consecutive interview took place. Therefore, the stopping criterion of three was sometimes exceeded (see Appendix E for the saturation of themes table which illustrates this process). Although the stopping criterion could not strictly be adhered to, what is clear and of greater importance is that the sample size is sufficient for the saturation of themes.
4. Clustering of themes. Master themes, subthemes and where applicable, categories, were then clustered and refined into five broad master themes with respective subthemes:
Relationships and Support
-Caregiver-Stroke Survivor Relationship -Informal Support
-Social Functioning
-Formal Support and Access to Services -Evaluation of Services and Unmet Needs -Messages from Health Professionals
Caregiver Factors -Emotional Functioning
-Social Functioning and Independence -Resilience
-Caregiver‟s Physical Health -Caregiver‟s Characteristics
Stroke Survivor Factors
-Cognitive and Behavioural Changes -Stroke Survivor Emotional Functioning -Stroke Severity and General Physical Health
External Stressors
-Employment and Financial Strain
Positive Outcomes
5. Integration of themes. Integration of themes was then carried out within and between each individual interview transcript. This was done by copying and pasting the direct quotation extracts of all common master, subtheme and category labels („Comments‟) into individual word files that corresponded to each of the five master themes. See Table 3 below for the layout of the first master theme file: Relationships and Support. This was completed for all interviews until there were five major Microsoft Word documents for each of the five main master themes.
Table 3.
Layout of Master Theme File for Relationships and Support Master Theme: Relationships and Support
Interview 1
Subtheme 1: Caregiver-Stroke Survivor Relationship Extracts
E.g. “Yea…sex has gone out the window…” Subtheme 2: Informal Support
Extracts
E.g. “Yes...but it’s something I can handle…It’s not a problem. And my kids are great, they say mum if you want to go out or something we’ll look after dad while you’re gone…cause’ two of them are still living at home…yea”
Cont. for all subthemes and categories of this master theme Interview 2