Chapter 5. Research design and methods
5.3 Research setting
5.4.4 Data analysis
In order to analyse the data to answer my research questions, I used the principles of thematic analysis described by Braun and Clarke (2006) (Figure 3). This approach fitted well in the context of the generic qualitative methodology I adopted for this study because of its theoretical flexibility (Braun and Clarke, 2006).
In the sections below I describe in more detail the data I drew on for the analysis and the process I followed. In reporting this process it is important to point out that although it is presented as a series of stages, in reality it was much more iterative.
During the analysis I moved back and forth between strategies to explore the data, developing and refining codes, and abstracting higher level themes. This continued
89 as I wrote up my findings, reflecting the messy and complex nature of qualitative analysis (Barbour, 2014).
Figure 3: Phases of thematic analysis (adapted from (Braun and Clarke, 2006))
Exploring the data
In the first phase of analysis I explored the data by reading through my entire dataset, and adopting a number of different strategies to familiarise myself with the data (Table 7). Using these approaches while coding prevented the process from becoming too mechanical (Bazeley, 2013).
90 Table 7: Strategies used to explore the data
Technique Description Examples
Shifting sequence Reading through transcripts in a non-linear way (Dey, 2005)
Reading observation transcripts
‘backwards’ – starting at the bottom and working up, rather than always starting with the cases discussed first by the team
Writing memos: scribbling in the margins of paper copy
transcripts and field notes to record thoughts and reflections Questioning Asking questions of the
data (Dey, 2005, Bazeley, 2013)
Asking: what’s interesting? Why is it interesting?
Counting Simple counting in order to avoid the risk of anecdotalism (Barbour, 2014)
Carrying out simple counts of who said what and exploring the context using the Nvivo matrix coding query function
Reflexive coding Interrogating and challenging each code created (Bazeley, 2013, Barbour, 2014)
Using memos to record definitions of codes and how used
Conducting a trial run to see if I could answer my research questions from the codes created. Are there any gaps that need filling?
91 Following this, I conducted two separate thematic analyses of the data I had
collected. The first thematic analysis (Thematic Analysis I) addressed research questions one and two by exploring the role of status hierarchies and the factors that facilitated or inhibited multidisciplinary discussion. This analysis was based on the observation and interview data. The second thematic analysis (Thematic Analysis II) addressed research question three, and was initially based only on the selective transcripts from the observation audio files. Conducting this analysis separately enabled me to focus solely on the approaches used by lower status groups to contribute to the MDT meeting using relevant extracts from across the whole dataset.
Thematic Analysis I
Generating initial codes
The first phase of coding I undertook for Thematic Analysis I was both inductive and deductive (Langley, 1999, Dey, 2005). This meant that some codes were grounded in the data, while others were drawn from relevant literature and my theoretical framework. Using this inductive and deductive approach I continually expanded and refined the coding framework (Dey, 2005, Barbour, 2014).
My coding framework covered the four full meeting transcripts, observations from all 122 sets of field notes, the observation audio files, the professional interview data, and the patient interview data (i.e. all data, except for the selective transcripts of lower status contributions). Having a single coding framework for these different datasets meant that I was able to consider key issues from different perspectives throughout the analysis process. This enabled me to undertake a process of
complementary triangulation, using different types of data to gain a more nuanced understanding (Moran-Ellis et al., 2006). For example, I could explore features of the initial case presentation based on my observations and the views of
professionals within the team, as illustrated in Table 8.
92 Table 8: Integration of datasets during the analysis process
Code Illustrative quotations
Initial case presentation
Observation transcript:
Consultant Haematologist: So the next one is [name of patient] who is an elderly gentleman that I saw with an isolated significant large sub mandibular lymph node, no other systemic symptoms and actually, remarkably fit and well for his age past history of prostatic carcinoma which he had major surgery on er in 1993 and he now has zoladex but his PSA is normal (Haematology 1)
Observation field note:
Consultant Haematologist 1 presented most of the patients, although StR1 presented a couple, as did the Consultant Pathologist towards the end (Haematology 1) Interview transcript:
Chemotherapy CNS: it is usually the doctor who has seen them in the clinic that presents that patient to the rest of the team (Haematology 1)
During this phase of the analysis, I reviewed my initial coding framework alongside excerpts of data and my research questions with my supervisors. This provided an opportunity for me to reflect, and to be challenged on, my approach to coding.
I reached a stage where no new codes were generated by review of the additional material following analysis of the four full meeting transcripts, and a selection of interview transcripts from both patients and professionals. The field notes and audio files therefore provided me with a wider pool of examples to draw on for my analysis, but the key codes emerged from the meeting and interview transcripts.
Developing and reviewing themes
Building on my coding framework, I grouped my codes into broader categories to develop an initial set of themes. For example, codes such as ‘provides information’,
‘interruption’ and ‘introduces new topic’ all seemed to represent attempts to participate in the meeting, so I grouped them under a theme titled ‘initiating participation’.
93 Defining and naming themes and producing a written report
Braun and Clarke (2006) describe these as two separate stages. The first (defining and naming themes) is presented as the stage at which the essence of each theme is defined. The second is described as beginning once a set of themes are “fully worked out” (Braun and Clarke, 2006) (p. 23). Instead, I found that these stages were heavily intertwined. It was through the process of writing that I was able to achieve clarity within each theme and to determine the most appropriate names for these. In practice, this meant that my themes continued to develop as I wrote, and my writing developed as my themes became clearer and more well-defined.
The final overarching themes from Thematic Analysis I are presented in Box 3 below. A full list of the sub-themes and codes is presented in Appendix 11. Themes I and II are discussed in detail in Chapter 6, and Theme III is discussed in detail in Chapter 7.
Box 3: Themes from Thematic Analysis I
I. The decision making process in MDT meetings II. The effects of status hierarchies
III. Multidisciplinary discussion
Thematic Analysis II
Generating initial codes
I created a second thematic framework based only on the 88 selective transcripts from discussions that involved lower status contributions in the 122 MDT meetings observed. This formed the basis of the analysis for research question three. I used a hybrid inductive and deductive approach to generate two sets of codes: ‘outcome codes’ and ‘strategy codes’.
94 Outcome codes
I initially reviewed each of the 88 discussions to deductively identify successful contributions. I made a decision about whether a contribution was successful or not based on the content of the selective transcripts themselves. My focus was on the effect of a contribution on the decision making process, rather than the quality of the decision itself. For example, contributions made by a lower status individual that prompted further discussion or influenced the clinical decision made by the team were deemed to be ‘successful’. Where the outcome of a contribution was unclear I used the quantitative dataset from the MDT Study to ascertain what decision had ultimately been implemented as a way of determining if a contribution had been incorporated into the final treatment plan or not.
Following this I generated inductive codes to explore the way in which a
contribution was successful. I grouped these codes into three themes linked to the decision making process, as illustrated in Table 9.
Table 9: Final themes and codes for successful contributions (outcomes)
Theme Codes
Prompts discussion Prompts others to share expertise
Raises awareness of an issue with the team
Determines logistics of treatment
Influences decision Acts as patient’s advocate
Ensures a decision is made
Influences a clinical decision
Influences practicalities of treatment delivery Facilitates team work Promotes social cohesion
Supports higher status decision
Strategy codes
I then generated another series of inductive codes based on the approaches being used by lower status groups to contribute in successful cases (Table 10).
95 Table 10: Final themes and codes for successful contributions (strategies)
Theme Codes
Asking questions Asks questions
Framing Psychosocial frames from medical perspective
StR provides clinical rationale for discussion Providing a practical
alternative or solution
Proposes an alternative solution (specifically challenging another suggestion)
Proposes solution to a problem facing the team Sharing information Provides clinical information
Provides information about patient as a person Using humour Uses humour
I discussed the development of this coding framework with my Departmental Qualitative Research Group. I presented a selection of data extracts to the group, before facilitating a discussion about different interpretations of the data. I found this process helpful in stimulating my thoughts at an early stage of coding.
Developing and reviewing themes
At the end of this stage my coding framework was composed of a series of ‘strategy’
and ‘outcome’ codes, which I explored in Nvivo using the ‘Matrix coding query’. This enabled me to explore each of the approaches used by lower status groups to contribute in relation to the different outcomes, as illustrated in Figure 4.
96 Figure 4: Matrix coding query: strategy and outcome codes Thematic Analysis II
97 Defining and naming themes and producing a written report
In line with my approach for Thematic Analysis One, I continued to develop themes through the process of writing. I discussed these themes with my PhD supervisors as my analysis progressed to further clarify and refine them. The final thematic framework incorporating the three outcome and five strategy themes is provided in Appendix 12.