In my PhD I am investigating ethical issues for health and demographic surveillance systems (HDSS) in sub-Saharan Africa (click here for more information on the study).  HDSS occupy a grey area between research, health care and public health, and have received little attention in the ethics literature and guidelines.  Together with my supervisors, I recently developed a coding framework to analyse qualitative individual interview and focus group discussion data that I collected from two HDSS sites in Kenya.  Data include stakeholders’ perceptions of the Verbal Autopsy component of HDSS, which are interviews with final caregivers of people in the HDSS who have died recently to determine probable cause of death.  Here I share some of my learning in developing the coding framework, and beginning to code my transcribed and translated data.  

Developing a coding framework as part of a wider iterative process.  During the workshop, Prof Sally gave an overview of the common approaches to qualitative data analysis, including content analysis, grounded theory, and thematic analysis. I am using a framework analysis approach, a widely used form of thematic analysis which involves familiarisation of the researcher(s) with the data, developing a coding framework, indexing, charting, and mapping and interpretation.  I chose this because framework analysis was first used in problem-based research to inform policy, and our study seeks to contribute to understanding of ethics in HDSS and to inform ethics policy and practice.  I developed my initial coding framework using the research guides (in turn based on a review of the literature and inputs from those familiar with the context), as well as familiarisation with the data collected by then which I’d collated in detailed summaries, and my own learning about the context. Questions in the research guides had been organised under four broad themes; benefits, burdens, consent, community engagement and governance, with the research guides regularly revised during data collection.

Involving others in developing the coding framework.  In developing the coding framework I selected three interview transcripts that I felt had the range of information given across many of the interviews (data from a HDSS manager, a researcher and a verbal autopsy field worker).  I also shared the interview summaries with my supervisors, both of whom have more qualitative research experience than I do, but who have different backgrounds from each other and from me.  We independently read each of the three transcripts and developed coding frameworks based on the a priori themes and other issues emerging from the data. Next, we compared the similarities and differences in our frameworks. Our frameworks included all the main a priori themes but differed in terms of level of detail and perspective taken. In one framework, the codes had been organised under several high level topics representing important entities e.g verbal autopsy respondents, staff, community, institutions (with ethics themes as sub-codes under each), but in another, they had been organised under the main ethical themes (such as benefit-sharing, with sub-nodes such as ‘staff’).   Some frameworks were quite broad (e.g individual-level burdens) while others became more specific e.g “factors influencing burdens to individuals” (as a sub-code).   We compared the different codes, discussed and revised them to develop an initial framework which mainly aimed to move us closer to the data, ensuring that there was less abstraction at this stage.  We are now applying this initial coding framework to analyse the rest of the data.  The image is a sample of our coding framework. 

There’s no one way to develop a coding framework, but there are some tips.  Embracing uncertainty and iteration were highlighted as important aspects of qualitative research, and that’s been the case even in developing the coding framework.  In developing and applying my framework, I’ve found it helpful to think about the following:

  • Be really clear on what each code and sub-code means, and the kind of information to put in there. Keep a note ideally in the software system you’re using of that so that I don’t forget over time.
  • Try to make sure that if I’m coding text under two different codes, it’s clear how those two codes are showing different things to each other; that I’m not just unnecessarily double coding
  • How to balance between making sure that codes are not so broad they are not very meaningful, and not so narrow that there’s lots of double coding. 
  • Recognise that although it’s worth getting my coding structure as clear and useful as I can, it’s not necessarily the end of the coding process… I can always go back to a specific code and ‘code-on’ (where I add on sub-codes later on), or merge sub-codes
  • Where I’m not sure whether to code something or leave it uncoded, just code it including as a new code or free node.

Prof Sally shared examples of “things” that can be coded in framework analysis. These include behaviours, events, meanings, interactions, barriers, facilitators among others.  We think this could be helpful for others.  In our study, we sought from early on to draw a distinction between practical and ethical issues, the empirical and normative, in order to focus the research.  These distinctions informed the interview tools and will inform the next stages of analysis.  We had already developed codes on states, such as feelings of helplessness among stakeholders and events, for example, deaths in the community.  As we continue with data analysis we hope to draw on some of the insights gained during the workshop

Categories

Reply