Understanding delayed access to antenatal care: a qualitative study Rosalind Haddrill

4.3.7 Analysis of the data

The data was analysed using a thematic approach, as discussed in chapter three. Braun and Clarke (2006) propose a step-by-step guide to undertaking thematic analysis which attempts to articulate the balance between analysis and interpretation suggested by Wolcott (2009). I have attempted to elucidate the process below using (loosely) the six stages they propose, though the reality was less linear, and illustrate some of the tensions and complexities inherent in the process. Critical to all stages of the analytical process was the interaction between myself and the research team, through research meetings and academic supervision, providing independent verification of the data and critical challenge.

4.3.7.1 Stage 1: becoming familiar with the data

As discussed previously, the analytical process began with the transcription of the interviews and the checking of the audio files to ensure accuracy. This was done repeatedly to examine all relevant data and get a clear picture of the participants’ language and views (Ziebland and Wheeler 1997). Transcripts were read and re-read and initial notes were made summarising each woman’s story (table 5.1.2, chapter five). These were compared with notes made after the interviews, to start to build an overall picture of the women’s views and to begin the process of coding the data. Members of the supervisory panel were involved in verification of the data during this initial process.

4.3.7.2 Stage 2: generating initial codes

Called the heart of thematic analysis (Liamputtong and Ezzy 2005), initial coding of the data involved identifying, sorting and organising data, labelling interesting features of the interview data in a systematic fashion, line by line. This inductive process was completed across all the data, collating data relevant to each code, using only codes originating from the interviews themselves (‘in vivo’) rather than imposing predetermined categories. The process of initial coding of the data involved the use of a computerised analysis package for qualitative data: NVIVO 8, which enabled the creation of hierarchies of codes and the ‘attachment’ of interview data to these. An illustration of the coding of the data in NVIVO is presented in appendix 4.13. NVIVO offered an efficient way of handling and organising the data. Its use improved the management of the large amount of interview data generated, and improved the transparency and rigour of the analysis. However, NVIVO acts to facilitate the process rather than carrying it out and does not help with complex skills like coding decisions or the interpretation of findings (Murphy et al 1998, Bryman 2008). Lofland et al (2006) comment about such software that

“at best they expedite and expand data organisation, storage, and retrieval possibilities, but they cannot do the hard work of data analysis, which requires certain intellectual and creative skills that, to date, only the analyst can bring to the enterprise” (p204).

It has also been argued that the use of such software may distance the analyst from the data, blunting the sensitivity which is central to the data analysis process (Murphy et al 1998). This was evident in my experience. NVIVO was a useful organisational tool, however dividing the data into coded pieces led to some fragmentation and loss of context and narrative flow.
The coding process considered not only what was said but also how it was said, some of which was lost during transcription. The coding of long narrative and discursive accounts was particularly challenging, as they included several themes in a coherent sequence, which required multiple, overlapping codes. As a result of these factors the analytical process required a combination of methods, including repeated comparison of the complete transcripts, the coded data in NVIVO and in some cases the audio files. This was supported by the production of memos and analytical diagrams describing the data, initial codes and theories and ideas for further sampling, reflecting the influence of grounded theory.

4.3.7.3 Stage 3: searching for themes

Refinement of the initial coding process gradually led to the creation of hierarchies or patterns of codes. Many codes were generated initially, which were then combined and recombined into recurrent themes. This reflects grounded theory’s ‘axial coding’: looking for relationships and putting the fractured data back together. The original research question was central to this process, which tried to identify, using the language of the participants themselves, the essence of women’s reasons for late booking. The themes, ‘meaningful sequential units’ (Leininger 1985), reflected differing levels of complexity and abstraction (DeSantis and Ugarizza 2000), as Boyatzis (1988) notes:

“A theme is a pattern found in the information that at minimum describes and organises the possible observations, and at maximum interprets aspects of the phenomenon” (p4)

The process was undertaken systematically and repeatedly for all the interviews, as I became more familiar with the data and the software. Again, the aim was to allow the themes to be found in the original data itself, using a careful and rigorous process, rather than superimposing pre-existing ideas onto the data and using ‘cherry picking’ methods to find supporting evidence of their existence (DeSantis and Ugarriza 2000, Green and Thorogood 2014). Although a priori considerations of the women who might be recruited for the study had been made, based on the surveys discussed in chapter one, there was no such influence in relation to the thematic analysis. As discussed in chapter 2, the literature synthesis was undertaken following data collection, and separate to the thematic analysis for the Sheffield study. This contributed to maintaining the originality of the findings.

4.3.7.4 Stage 4: reviewing and refining themes

Pope, Ziebland and Mays (2006) discuss how a stronger analysis moves beyond simple description to examine the interconnection of themes. A process of reviewing, reorganising and refining themes was undertaken for all the interviews in the study. This involved drawing thematic maps and tables of the data, exploring relationships between the themes, subthemes and the codes within them, and checking the coded extracts; looking for repetition, omissions, similarities and differences (Ryan and Bernard 2003) (Appendix 4.14). This reflects Wolcott’s (2009) description of the more intuitive process of ‘interpretation’ which started to take place at this point. Leininger (1985) calls this interpretive process a ‘synthesis’ of themes, to obtain a comprehensive and holistic view of the data.

4.3.7.5 Stage 5: defining themes, developing theories

As the analytical process progressed I started to focus on developing more theoretical ideas in relation to the data. The themes, their boundaries, hierarchy and the language used to define them, were refined further. As the pattern of themes became more defined visual mapping methods were used, as illustrated in appendix 4.15, to give an overall graphic picture of the data and aid the analytical process. Lofland and Lofland (1995) suggest that at this stage in the analytical process the themes or ‘elaborated codes’ as they call them ‘begin to assume the status of overarching ideas or propositions that will occupy a prominent place in the analysis’ (p192). A taxonomy of themes and subthemes relating to late booking was developed and went through many transitions, with interpretive challenge provided during analysis meetings from the supervisory panel. As this developed I started to consider the overall story that the analysis was telling, whilst trying to ensure that it remained honest to its component parts: the individual accounts from the women. This was a challenging and complex task, as Leininger (1985) discusses:

“Much creative thought and analytical ability is needed to literally “put the pieces together” so that a theme or pattern of behavior is formulated that is congruent to the people being studied” (p60)

Differing depths of analysis are appropriate for different studies (DeSantis and Ugarizza 2000). The study’s aim was to gain a broad understanding of the attitudes and beliefs of late booking women, to improve access and health outcomes, rather than exploring language in detail, such as in discourse analysis. The process of thematic analysis provided an analytical method with breadth but also an attention to detail, variation and ‘deviant’ cases, to provide a credible and thorough account of the data, beyond the taxonomy (Green and Thorogood 2014).

4.3.7.6 Stage 6: presenting the findings

Braun and Clarke (2006) identify the writing up stage as the final opportunity for analysis, the creation of a ‘big picture’, whose parts fit logically and meaningfully together. This final stage reflects Sandelowski’s (1995b) ‘qualitative interpretation’, a creative process that produces something new that is different from, whilst faithful to, the data in its original form. The end point of the analysis aimed to be both descriptive and theoretical about late booking; to create something that could have the potential to influence policy or practice interventions. My aim was to present a taxonomy of themes relating to late booking (a common method of ‘measurement’ in qualitative research (Pope and Mays 2006)). However, I also wanted to create a story (Giorgi’s (1985) overall ‘structure of experience’) around the voices of the women, a narrative representation of the accounts but one which was anchored (grounded) in the words of the participants.
This narrative attempts to reflect common but also unique and striking themes in the data, alternating description and analysis with illustrative quotes (Holloway and Wheeler 2010). As it was recognised that the sample was not representative of the whole population, and because of the flexibility in the interview process, I felt it was misleading to report relative frequencies of responses. However, simple numeric indicators were used in some cases to give an overall flavour of the data (Ziebland and Wright 1997). The language of the themes aims to be consistent with the interpretivist approach and feminist influences on the research itself, particularly in terms of being respectful of the participants’ perspectives, not labelling or blaming. The narrative attempts to develop explanatory theories, in order to provide a comprehensive view of late booking that has depth and breadth, and to answer the research question.