Chapter 4: Research Design and Methods
4.6 Design Considerations: Data Analysis
It is suggested that IPA data analysis should not be viewed as a ‘prescriptive methodology’, as there is no ‘right or wrong way’ to perform data analysis utilising this approach (Smith et al., 2009, p.80). However, a broad guidance in conducting data analysis is proposed (Smith et al., 2009), incorporating the flexibility of being able to return to the data to focus on meanings throughout the process of analysis (Smith & Osborn, 2003). Data analysis utilising this approach goes beyond capturing just a description of experiences, but initiates an understanding of those
experiences (Larkin et al., 2006), by engaging in a detailed analysis of each
interview to determine shared meanings (Shinebourne, 2011). Any interpretation of the data is based solely on what the participants have expressed during their interviews, and the role of the researcher is to endeavour ‘to make sense of the participants trying to make sense of what is happening to them’, therefore following the process of ‘double hermeneutics’ (Smith et al., 2009, p.3). This hermeneutic endeavour by the researcher is ultimately aiming to realise the meanings attached to these experiences. The researcher therefore engaged with the interview transcripts to identify themes and achieve an interpretation of the meaning of the participants’ experiences (Quinn & Clare, 2008) within the context of their world, their working
environment (Larkin et al., 2006), providing maternity care to women with raised BMIs.
Data analysis utilising IPA commenced during the interview process, whereby the intention of the researcher was to initiate and affix meaning to what was being said by the participants (Smith et al., 2009), as opposed to attempting to ‘fit a pre- existing theoretical viewpoint’ (Smith, 1999, p.285) on to the participants’ accounts. This initial assessment and impression of what the participants had expressed during their interview was handwritten in the researcher’s notebook as soon as the interview had been completed. Following interviews held in the participants’ own homes these notes were made in the researcher’s car.
The flexible process of data analysis as advocated by Smith et al. (2009) for larger sample sizes was followed. This meant that a thematic analysis collectively representing the two different groups for part 1 and part 2 of the study was
performed, rather than a case study approach, which would be the idiographic representation of what it meant to each individual to care for women with BMIs ≥30kg/m2 during the childbirth continuum (Larkin et al., 2006). All the interviews in
both parts 1 and 2 of the study had been digitally recorded with consent, and were transcribed verbatim (Blaxter et al., 2006). The stages of data analysis required the researcher to engage with the interview transcripts, ‘reading and re-reading’ them in step 1 of the data analysis process as advocated by Smith et al. (2009, p.82).
Step 2 involved ‘initial noting’, a manual task which involved ‘exploratory commenting’ to determine similarities, differences or connections within the texts and annotating them as such (Smith et al., 2009, p.83). Essentially, the data was systematically interrogated so that rigorous explanations (Barbour, 2008) of what it meant to the participants to care for this client group could be achieved. This was a lengthy and complex process, and the researcher worked with both the transcribed texts and verbal recordings of the interviews to become immersed in attaching meaning to what the participants had verbalised (Boeiji, 2010).
Appendix 18 provides an example of a transcript with manual initial noting and annotations.
This immersion with both the texts and recordings of the interviews produced meaningful information, which was represented by the exemplar quotes (Black, 2006) realised from this stage of analysis (found in chapters 5 and 6), therefore initiating a construct for assigning meaning for both sets of participants.
Step 3 concerned ‘developing emergent themes’, essentially fragmenting the transcripts into a thematic analysis by corralling the ‘initial noting’, which had
(Smith et al., 2009, p.91). Essentially, the researcher was considering the smallest units of data in terms of the larger data sets (Cohen et al., 2000). This was
commenced manually. However, as the researcher became more familiar and confident with the use of NVivo 10, this was utilised to manage the collected data (Bazeley & Jackson, 2013) and to confirm the emergent themes. In-depth reading of the transcripts by the researcher, corroborating initial themes and performing data reduction with the aid of NVivo 10, resulted in 21 themes being identified by 19th
July, 2012 (appendix 20).
As previously discussed under recruitment of the participants, the concept of data saturation was followed to justify ceasing sampling (Mason, 2010). In these circumstances it provided a rationale during the process of data analysis that when no new themes were produced (Kumar, 2014), the conclusion was that saturation of the data had been achieved and therefore no further data collection transpired (Curry & Nunez, 2015).
Like part 1, part 2 followed the same processes for data analysis and by 20th
December 2013, 21 themes had resulted; however, further reduction of the themes did occur during the writing-up stage of the analysis. This resulted in 20 themes for part 2 being realised (appendix 21). The theme of ‘Students are surprised and shocked at the size of the women’ was amalgamated with the theme ‘Size of woman’ which strengthens this theme.
The emergent themes for the whole groups of participants were allocated to ‘nodes’ as this is how themes are represented by NVivo 10 (Bazeley & Jackson, 2013). Smith et al. (2009) suggest that emergent themes can be created in word- processed documents by compiling the extracts from the transcripts; the researcher, however, found the NVivo 10 software package user-friendly in enabling
straightforward management of the data sets.
Step 4 entailed ‘searching for connections across the emergent themes’. Smith
et al. (2009, p.92) suggest that the themes be listed chronologically as they occur in
the transcribed texts, but also state that researchers can be innovative and not too prescriptive. NVivo 10 produced the ‘nodes’ alphabetically, and the researcher printed off all the themes which had been created. Next, the researcher read the abstracted comments from the participants, which had been created into themes (nodes), endeavouring to search for connections between the themes to create super-ordinate themes. Super-ordinate themes represent the next step in analysis by contextualising and providing an overarching concept to encapsulate the emergent themes.
Appendix 19 provides an example of how the data was managed in terms of collating information for an emergent theme (node) using NVivo 10. The emergent theme of ‘promotes normality’ from part 2 of the study was chosen to illustrate the process due to word count considerations, as it had the least amount of text for the quotes within the node. The researcher, having read all the emergent themes, then sought connections between them. Each theme was then grouped and categorised with other themes to create an overall intended meaning for the cluster. In this example, the emergent theme of ‘promotes normality’, together with the emergent themes of ‘non-judgemental, aims not to discriminate, treats everyone the same’ and ‘medicalised and high risk’, were determined to create the super-ordinate theme of ‘normalising the risk’. This illustrated to the researcher the midwifery students’ desire to provide non-judgemental care in promoting normality to a defined high risk and potentially medicalised group of women (chapter 6, section 6.5).
The creation of the super-ordinate themes should also be representative of the participants to ensure credibility (Smith et al., 2009). The researcher chose in parts 1 and 2 of this study to create five super-ordinate themes for part 1 and five super- ordinate themes for part 2, by linking the emergent themes to an over-arching premise in support of generating meaningful subsets of the participants’ expressed perceptions of caring for this client group. Two tables incorporating the super- ordinate themes and their relevant emergent themes can be found in chapter 5 to represent the process of data analysis and findings for part 1, and in chapter 6 for part 2 of the study (tables 5.1 and 6.1).
Metaphors were chosen to name the super-ordinate themes, representing the emergent themes within the overarching themes and encapsulating what they represented to the researcher in terms of meaning (Smith et al., 2009). The use of metaphors is supported by Carpenter (2008), who believes that they can illuminate the intended research message, and by Charlicke et al. (2016) who contend they are integral to producing the findings for an IPA study. An explanation for the application of the chosen metaphors is given under the introduction of each super- ordinate theme in chapters 5 (tables 5.2. 5.3, 5.4, 5.5, 5.6) and 6 (tables 6.2, 6.3, 6.4, 6.5, 6.6). Chapters 5 and 6 also provide the findings of the study in narrative accounts under each emergent theme.
The term ‘numeration’ is used by Smith et al. (2009, p.98) to determine how important the theme is to the individual participant by using the frequency with which it is mentioned during the interview. Appendices 20 and 21 demonstrate this concept for both groups of participants (Smith et al., 2009). The researcher in this instance was also seeking to interpret the information to establish how important it is to the
research by realising how many participants have expressed the theme. Smith et al. (2009) recommend that it is up to the researcher to make the decision about how many participants’ comments would credibly make up a theme; they suggest anything from a third to half of all participants. A table was created and a cut-off point of at least five out of sixteen participants in part 1 and three out of eight participants in part 2, expressing each respective theme, was chosen by the researcher as a means of realising at least a third of contributions made by the subjects. As previously stated, the themes most frequently mentioned by the participants and how often these emergent themes have been referenced by them are referred to as ‘numeration’, and to support the creation of the super-ordinate themes each one had a specific table devised relating to the emergent themes. The tables (tables 5.2–5.6, 6.2–6.6) identify the emergent themes that the researcher felt best encapsulated individual super-ordinate themes, and provide demonstrable ‘numeration’ as recommended by Smith et al. (2009).
If the study was utilising case studies with small sample sizes of three to six participants, the next stage would be step 5 in the analysis; Smith et al. (2009, p.100) would suggest this is ‘moving to the next case’. This would involve examining each subject in depth, therefore taking an idiographic stance (Coyle, 2014).
However, for larger samples it is not always possible to examine each case in depth to determine what the key themes are that represent the whole group. It is therefore suggested that the ensuing step in the process of analysis is to write up the findings to produce a narrative account of what meaning the researcher has attached to the findings (Smith et al., 2009). The following step, therefore, that the researcher undertook was to read and re-read the themes making up each super-ordinate theme, to write a generic account relating to each of the themes with supporting extracts, and to consider what it meant to the participants as a whole to care for women with raised BMIs during the childbirth continuum, consequently adhering to the ‘hermeneutic circle’ (Smith et al., 2009. p.27). In doing so, an analytical
interpretative narrative was produced by the researcher and can be found in chapters 5 and 6 for parts 1 and 2 of the study.
To achieve this interpretation, however, the researcher did not just accept what was expressed by the participants, but also endeavoured to critically question these accounts (Shinebourne, 2011) by returning to the data to focus on meanings
throughout the process of analysis (Smith & Osborn, 2003).
The data analysis approach for IPA provided a rigorous systematic process by which the data could be interrogated (Barbour, 2008), in the first instance to ensure that the researcher focused on both groups of participants’ experiences of caring for
this client group to determine the participants’ ‘world’ (Larkin et al., 2006); and secondly, to generate an analytical interpretation of what the experience of caring for women with raised BMIs meant to them (Smith & Osborn, 2003). The process of performing data analysis using IPA was an iterative, dynamic, empathic
(Shinebourne, 2011), interesting, motivating and informative experience for the researcher. The concept of the ‘double hermeneutic’ approach was applied in that the researcher endeavoured to make sense of the participants making sense of their experiences of caring for this client group (Smith et al., 2009). This was a constant process of returning to the data to establish meaning from the participants’
perspectives within the context of midwifery practice and then questioning the truth of that derived meaning. The use of IPA also guided the researcher to attempt a deeper level of analysis, referred to as second order by Larkin et al. (2006), helping to provide unique insights on what it means to care for this client group, which may in turn offer broader implications for care delivery.