Chapter 5 Methods
5.3 Data Analysis
5.3.1 Thematic content analysis
I first conducted a thematic analysis of the documents collected. Each document was scanned for main themes as identified from the literature review in Chapter 3. The analysis of themes was used to identify elements of the policy process and the official definition of poor-quality medicines. I extracted key information on how poor-quality medicines were defined, on the main policy outcomes of actors’ interactions as well, as information about key focusing events leading to policy changes and reforms. I note here the potential discrepancies when analysing the framings and definitions of poor-quality medicines from translated English version rather than original documents. To do so, I read each document in full twice, highlighting relevant clauses or sections of these documents that detailed the structure of the regulatory framework to improve access to quality medicines for each country. Then, I coded the content of the available legal and policy documents using an excel sheet in which I recorded recurring themes iteratively, re-arranging those throughout the process of thematic content analysis. I took note of policy developments and focusing events leading to policy change. At this stage, I was able
to cross-reference the findings from this thematic analysis with the map of stakeholders, highlighting where and how actors and networks influenced policy developments. The findings gathered from this phase of the analysis are represented in each results chapter as a chronological narrative of the policy process, and illustrated through a timeline. I also paid specific attention to the use of terminology in each document, making note of the definitions of poor-quality medicines. This process provided an indication of how the problem for poor- quality medicines was conceptualized in official documents, as material for reflection on how the official definitions might influence policy developments. The outcome of this analysis is represented in the form of a table of official definitions for each country.
I then conducted a thematic content analysis of the data from the semi-structured interviews against the themes emerging from Chapter 3. The analysis of themes here served to identify key determinants and challenges of the issue of poor-quality medicines as perceived by policy actors, and to discern contextual factors that might influence policy developments. I followed five main steps in the thematic content analysis process, from data familiarization, coding, indexing, charting and mapping to data interpretation (Ritchie & Spencer 1994). Recordings of interviews were first converted into time-stamped transcripts that were uploaded into a Computer-Assisted Qualitative Data Analysis software (Nvivo 10), along with word- processed field notes. After the transcription process, which served as a phase of data familiarization, the main themes were coded along a pre-determined coding framework, created from the list of themes and the conceptual framework emanating from the literature review (See Appendix 7 on page 328 for coding framework). This coding stage involved a tagging process using highlighting tools to identify key themes from the data (Pierce 2008). While the initial coding framework was based on the themes identified in the literature review, the analysis phase gave special attention to themes that emerged from the data as ‘free nodes’ (Glaser & Strauss 2008).
To identify key themes and causality references in this coding stage, I paid specific attention to key terms such as ‘because’, ‘after’, ‘then’, ‘same as’, where participants shared their perceptions and experiences of the policy issue (Green & Thorogood 2014). This information also contributed to my understanding of the policy process and the policy context. Nvivo 10 software facilitated the management of data permitting easier navigation between cluster of themes (Spencer et al. 2014; Green & Thorogood 2014). I conducted a comprehensive analysis of the data until no new themes emerged that had not already been taken into consideration (Silverman 2011). This stage of the thematic content analysis process served to highlight what determinants and challenges policy actors attribute to the policy issue of poor-quality
antimalarial medicines. These findings were then coded as a final list of themes and sub- themes for each country.
The indexing phase involved a deeper theoretical reading of these emerging themes (Kvale & Brinkmann 2008), during which I was able to draw key observations on stakeholders’ perceptions of the policy issue. I matched the emerging themes to the pre-determined list of frames. To do so, I used the pre-set list of six ‘master’ frames (as identified in Chapter 4), which allowed for a more systematic and meaningful comparison of framings of the problem of poor-quality medicines across countries. This process helped to refine the list of frames iteratively and in accordance with the themes emerging from the data. During this indexing phase, I applied a method of ‘constant comparison’ in order to spot patterns in the data (Glaser 1999; Pierce 2008; Green & Thorogood 2014). As a result, I identified salient themes among all three categories of respondents and paid specific attention to those that bore the most references. From the data, I depicted both rhetorical frames (the debates and stories in relation to the policy issue) and action frames (evidence of practice, policy solutions, laws or policy) (Schön & Rein 1994). I referred to both the content of the transcripts and my field notes in order to depict the specific attitudes of local stakeholders towards the policy problem (Chong & Druckman 2007). This helped to capture the tone of the discussion and elements of the social context that may influence the data. I paid particular attention to the terminology used by respondents, their specific definition of the problem and the rationales used repeatedly by local respondents from each country to explain this problem. I created a separate coding scheme in Nvivo for these dominant or ‘master frames’ used across and within all three categories of respondents.
The fourth stage (charting) involved drawing linkages between the master frames, and the dominant themes emerging from the data that describe the substance of each dominant frame. I charted different interpretations of the policy issue across institutional settings (across the three categories of respondents) within the same frames. By juxtaposing the framings that emerged from the document analysis and the interview data, I was able to explore the implications of perceptions of the problem for policy developments to improve access to quality medicines within each country. The results from the thematic analysis of the interview data are presented separately for each case country, as a thematic narrative describing the six frames in detail and acknowledging variations in interpretations of the policy issue across institutional setting, while reflecting on analogies between actors’ perceptions and recent policy developments.
The limitation of thematic content analysis for the study of frames is that the classification of frames by the researcher is itself a subjective process – a limitation which I have sought to address through a reflexive data collection and analysis process. More specifically, I followed an iterative process of data collection and analysis based on the Spiralling Research Approach by Berg & Lune (2012), as illustrated in Figure 7 on page 92. This approach is particularly adapted to the nature of this study which required a strong awareness of the socio-political context, as it provided an opportunity to adapt data collection tools throughout the data collection process (Green & Thorogood 2014; Berg & Lune 2012). The Spiralling Research Approach is inherently reflexive and allowed for Methods A, B and C to feed into one another throughout the data analysis process. For example, while the stakeholder mapping (Method A) supported the data collection process for Methods B and C, both the data from the analysis of documents (Method B) and from the interviews (Method C) helped to refine the maps of stakeholders.
Figure 7: Research methods in a reflexive research process
Cross-feeding between the findings of Methods B and C was useful to triangulate findings and develop my observations of the level of enforcement of written policies and laws. The data from the interviews helped to verify the findings from the analysis of official documents – enhancing my understanding of the policy process. I ensured that the research design per se remained unchanged through maintaining careful awareness of the data collection and a systematic process of data analysis. In practice, therefore, the spiralling research approach allowed for the refinement of the criteria for the document analysis and the topic guides for subsequent semi-structured interviews, with the aim of delving deeper into emerging themes without drastically altering the systematic application of each method.