Chapter 4: Research Design and Methods
4.8 Data analysis
Analytical strategies within qualitative data analysis are diverse with different purposes and epistemological underpinnings (Morse and Richards, 2002). The transparency of the data findings and coding are presented throughout the data analysis. The epistemological
underpinnings of this study was to present a description of the child and parent perspectives of decision-making for asthma management, staying close to the data and presenting the findings in the participants’ own words, unlike researchers conducting ethnographic, grounded theory or phenomenological studies. Although the framework approach
underpinned data analysis, in order to make an informed choice a range of approaches and their purpose were considered including: content analysis; thematic analysis;
phenomenological analysis; and grounded theory. Although these approaches are inductive and focus on meaning, they differ in their ontological and epistemological underpinnings and purpose (Morse and Richards, 2002). Content and thematic analysis focus on systematically exploring data to find common meaning; phenomenological analysis aims to capture the uniqueness of participants’ stories as co-constructed with the researcher; and constant comparative analysis used within grounded theory, promotes theory generation (Patton, 2002; Crabtree and Miller, 1992). Other approaches such as discourse or conversation analysis are socio-linguistic in that they focus on the use and meaning of language and are rooted in the disciple of ethnography, which was not congruent with the aims of this study (Smith and Firth, 2011).
Thematic analysis is the most widely used approach to qualitative data analysis (Braun and Clark, 2006). Thematic analysis, like the framework approach, is gaining popularity within nursing and health psychology research settings, because of the increase in adopting descriptive designs to underpin qualitative research (Swallow et al., 2011; Tierney et al., 2011). Both approaches are flexible because they can be adopted across qualitative research paradigms and are suitable for analysis of different types of data and guide the analysis through a series of interconnected stages adding to the transparency of the analytical process (Patton, 2002; Pope et al, 2000). These methods can provide rich insights into complex phenomena and be applied across a range of theoretical and epistemological approaches (Braun and Clark 2006; Spencer et al, 2003b). However, thematic analysis has been criticised for lacking depth, fragmenting data, being subjective and lacking
transparency in relation to the development of themes, which can result in difficulties when judging the rigour of the findings (Smith and Firth, 2011; Attride-Stirling 2001).
The framework approach has similar underpinning assumptions as grounded theory, in that data analysis may generate new theories, but the focus of framework approach is to
describe and accurately present participants accounts toexplore a topic that has received little attention (Ritchie and Spencer, 1994). The framework approach offers a systematic approach to data analysis through distinct but interrelated stages, allowing the researcher to move iteratively through the data until a coherent account emerges (Johnston et al., 2011; Rabiee, 2004). This method was adopted to underpin data analysis because it offered clarity, guided a novice researcher and assisted in ensuring a clear audit trail was maintained and consequently ensured greater transparency in the development of the findings (Ward et al., 2013; Smith and Firth, 2011; Srivastava and Thomason, 2009; Pope et al., 2000). The
framework approachfacilitates cross-sectional data analysis (Smith and Firth, 2011); allows for between-case and within-case analysis comparisons and associations (Swallow et al., 2011; Ritchie and Lewis, 2003; Snape and Spencer, 2003; Spencer et al., 2003); and in-vivo codes promote the authenticity of the data (Ritchie and Lewis, 2003). Indeed within this study such a systematic approach allowed the child and parent data to be analysed separately then child-parent dyad, where relevant, to becombined in the final analysis to
identify similarities and differences between emerging themes (Furber and McGowan, 2011).
The interconnected stages of the framework approach guided data analysis as a systematic and transparent process (Swallow et al., 2013; Ward et al., 2013; Ritchie and Lewis, 2003; Popes et al., 2000). Consistent with the framework approach data collection and data analysis were undertaken simultaneously facilitating the recognition of data saturation (Smith and Firth, 2011). The stages of framework are located with three discrete but interconnected stages: data management (coding the transcribed data to form an initial category); descriptive account (refining the categories and themes to develop more abstract concepts); explanatory accounts (reflecting on the associations and patterns within the core concepts). The three inter related stages of the framework approach are now presented:
Data Management
Although data management is time consuming it is an essential stage of the analysis and commences with transcribing the interviews verbatim. Data management involves
condensing the data into manageable units, initially using the participant’s own words (in- vivo code). Each interview was transcribed to facilitate familiarity with the data which also facilitated preliminary analysis by annotating transcripts with memos recording initial thoughts (Ward et al., 2013). Codes were summaries of a phrase or paragraph from the transcript, which are linked together to form initial categories. Categories in turn form a coding index that is used as a means of organising the whole data set.
Descriptive accounts
Descriptive accounts involved reviewing the data within each category in the coding index, summarising and synthesising the categories, identifying links between the categories and ascertaining if categories could be grouped into an overall theme. The final stage of the data analysis makes associations across the data sets. Refining the categories and themes was an ongoing process throughout data analysis, which as the process evolved led to develop more abstract concepts leading into the conceptual framework. Within the study findings
typologies of parenting styles emerged from the descriptive accounts of data analysis, ‘Typologies are specific forms of classification that help to describe and explain the
segmentation of the social world or the way that phenomena can be characterised or
differentiated’ (Ritchie and Lewis, 2003:p214). Family specific maps are presented in Chapter 6, Section 6.1.1 to illustrate cases representing three of the four different parenting style groups (Maccoby and Martin, 1983; Baumrind, 1967). All of the parents fit within one discrete parenting style.
Explanatory accounts
Explanatory accounts involve reviewing and constantly refining the original themes and categories, reflecting on the original data and the analytical stages, reflecting on associations and patterns within the core concepts to describe the decision-making between the child and the parent in relation to the management of childhood asthma and to seek wider application of the core concepts (Ritchie and Lewis, 2003). Explanatory accounts are constructed by the researcher by reflecting on the original data, identifying the emergent patterns within that data in order to produce a coherent account, interpretation and explanation of participant accounts (Srivastava and Hopwood, 2009).
4.8.1 Application of the framework approach to child and parent findings
Codes were developed by summarising units of data (key words, phrases or sentences) from the transcript using children and parents’ own words. Each unit of data was summarised to form an initial category. A list of categories formed the coding index that was used as a means of organising the whole data set. Ongoing records of preliminary thoughts and development of initial categories were maintained. An example of the coding matrix, of a child and parent individual interview transcription, highlighting the processes involved in identifying codes and initial categories is presented in Appendix 12 and 13 respectively. As the coding progressed similar categories were grouped together forming initial themes. The initial categories and initial themes formed the coding index. The process was undertaken separately for the child and parent data, Appendix 14 and 15 respectively.
Data management could have been managed using computer software such as NVivo® Version 10, Microsoft Excel (Swallow et al., 2013) or Microsoft Word (Furber, 2010). Qualitative data software packages are not a pre-requisite for undertaking qualitative
analysis and have been criticised for extensive coding and categorising, resulting in data becoming unmanageable with the analysis becoming a reductionalist rather than
interpretive process (Morrison, 1998). An initial attempt at using NVivo® Version 10,proved unproductive due to lack of familiarity with the software and learning the system hindered the ability to make links when developing codes. Consequently, a database spread sheet was used, enabling related codes to be viewed simultaneously. Using printed version of the coding matrix and coding index facilitated by the use of post-it notes to write initial
categories and themes assisting with the iterative process of the analysis by moving post-it notes around, as the categories and themes were refined, combined and developed (Ward et al., 2013). Using the spread sheet enabled the supervisors to be provided with
anonymised transcripts for independent coding, which enabled decisions to be challenged and added to the depth of data analysis (Ward et al., 2013). In addition coding and
annotations were included alongside interview extracts, which facilitated discussion during supervision meetings. As data coding of the transcribed interviews progressed the coding index was refined. Similar codes formed initial categories. The categories were refined and similar categories formed initial themes. The refined categories and themes were used to sort and organise the whole data set and is presented in Appendix 16 and 17 respectively. Findings were contextualised by exploring the relationship between the final concepts and the established literature and theoretical perspectives related to child-parent shared decision-making of asthma management.
4.8.2 Data reduction
The data coding was a lengthy iterative process and was particularly challenging during the data management phase; progress was eventually achieved by moving away from a ‘nursing lens’ to a ‘researcher lens’. Refining the categories and the themes as an iterative process is important to establish the credibility of the findings (Ritchie and Lewis, 2003). The four interview transcripts were used to develop the coding index and were independently reviewed by both research supervisors. An initial list of the child and parent data generated 69 and 55 categories, respectively. The child and parent categories were grouped
thematically into 40 then 28, initial categories, respectively. Links between categories were identified and refined into 10 and 13 child and parent themes. Three child and three parent
core concepts emerged from the 10 key child and 13 key parent embedded themes. The coding index was reviewed on several occasions, reducing the biases through the exploration and discussion of different interpretations within the data. The study objectives were
constantly referred to during this stage of data analysis, to remain focussed on the study aims. Two data sets were created: child and parent. As there was only one joint parent interview, this was included as one parent interview, although quotes are presented in Section 5.3. from both. Transcribing data verbatim enabled the analysis to remain close and true to the participant’s own words, (Sandelowski, 2000), increasing the credibility of the research findings (Speziale and Carpenter, 2005). To capture the diversity of participant perspectives a range of quotes are presented (Spencer et al., 2003). A particular strength of the framework approach is to enable transparency between the participant perspectives and the researcher interpretations (Morse and Richards, 2002), promoting the credibility of the findings. Regular discussions and reviews with both supervisors ensured the data analysis remained focussed on the child and parent perspective of shared decision-making for the asthma management. The application of critical thinking to how the child and parent description were coded, the links between the codes and categories and between the categories and themes, required nurturing and developed as the data analysis progressed. The explanatory accounts were contextualised by exploring the final themes with the theoretical perspectives (Chapter 2) and existing literature (Chapter 3). A service user day is planned to confirm concepts and themes are an accurate reflection of child-parent shared decision-making for asthma management. Dissemination of findings has primarily been through local, national and international conferences, outlined in Chapter 9, Section 4.