Analysis of qualitative data

In week 1, we learned about collecting data through qualitative methods such as in-depth interviews or focus group discussions. With the participants’ consent, these discussions are typically audio recorded. This audio recording is your data. Now, we need to organise and interpret this data into a cohesive whole to gain some form of understanding of what we are investigating. This is the purpose of analysing qualitative data. First, you need your data transcribed. This is writing down everything recorded in the audio. For most types of analysis, it is sufficient to transcribe just the words that were said exactly as they were said, without summarising or attempting to turn them into neat sentences.

If the interview or focus group discussion was conducted in a language you don’t speak, you would need to have it translated. Remember to allow sufficient time and budget for transcription and translation to avoid rushing in these steps and losing valuable content in your data. If you speak the language of the interview or focus group, you don’t need to translate it. You can analyse the data in the original language and avoid losses in translation. Then you would need to anonymize the transcripts. This means removing all personal information, replacing these with fake names or aliases so that participants’ identity and private information are protected during reporting.

Now, start familiarising yourself with the data by reading through the transcript freely several times while keeping in mind the purpose of the study what you’re trying to explore. What are the ideas emerging from the transcripts? Start noting these reflections in the margin or start a summary document. This will help you start coding. Coding is not as intimidating as it sounds. It is essentially organising the data into a structure so that you can see patterns. Let’s consider an example. We conducted a study to explore barriers faced by people with disabilities in accessing water sanitation and hygiene facilities. Going through the transcripts, we highlighted phrases that describe different barriers. We mark these as barriers. We’ve started coding.

After going through all the transcripts, we pull together all these phrases coded as barriers. Here are the barriers we identified. Now, we need to organise them. One way to do this is to group them into categories of similar types. There are barriers to do with the physical environment. These might be barriers in the natural environment, like distance to the facility, or it could be barriers to do with the built infrastructure, like narrow doors. Then there are barriers to do with the way services are delivered. For example, lack of information. Then, there are social barriers such as attitudes, cultural beliefs, and traditional practises. An example of this is stigma. Here are these barriers again. Now divided into categories we just discussed.

We have organised the codes into a coding scheme. You can revise the codes or the scheme as you go on. You can add more detail or you can combine some codes to make it broader. NVivo is a commonly used software to help with coding and organising your coding scheme. Note that it doesn’t help you analyse. It just helps you to manage your data. It is the researcher who actively searches for patterns and interprets them. There are several resources for using NVivo. Alternatively, you can code manually by highlighting transcript printouts, marking different codes with different colours.

There are various approaches to qualitative analysis. The most common two are thematic analysis and content analysis. Content analysis focuses on characteristics of the data, what was said, and who said it. Thematic analysis focuses more on identifying and analysing the ideas and patterns, called themes, that are emerging from the data. We’ve provided an article that illustrates the difference between these two approaches in more detail.

In our example, we took a thematic approach and started noticing that many of the social and cultural barriers affected women with disabilities more than men. As we progressed, we kept coming across menstrual hygiene management as a key element that related to gender differences in access to wash facilities. This became a theme. It is important to synthesise. This is where you explicitly state the links between codes, identifying themes, and patterns across the different transcripts. It is not enough to simply describe patterns such as women with disabilities experience more barriers to watch than men. You must interpret the data to try and figure out why there is this difference.

Through this interpretation, you can explain how your findings relate to the research question, suggest wider relevance, and highlight areas where an intervention can make most difference. Based on our data, we can see that merely addressing physical barriers by building ramps would not lead to better access to wash facilities. There is a need to also address social taboos and gaps in policy to strengthen institutional practise. Let’s look at some features of a good analysis. First, always anchor your analysis in what people actually said. Provide quotes in your write-up as evidence to back up your findings and analysis. A good analysis is comprehensive and comparative. Avoid looking for data that just backs up your theory or hunch.

And don’t ignore deviant cases, which are data that doesn’t seem to fit the general pattern. You can strengthen your analysis by comparing findings across different cases or transcripts and other data. It is also important to consider the impact of context, methods, process, and researchers’ characteristics. This is called reflexivity. It is not about removing interviewer or context effects, but about accounting for and explaining them. We discussed this among things to consider when you collect qualitative data. Reflect on this when you write up your analysis, too. Similarly, good analyses pay attention to subjectivity. That is the way the data or analysis is shaped by our feelings or opinions. We are all biassed.

But it’s important to acknowledge it and challenge what we think and question our assumptions. We must also check the trustworthiness or reliability of the data we collected. You can do this by having your data coded by another researcher and then examine differences between your coding scheme and findings with theirs. Some researchers also test emerging themes and theories in subsequent data collection. Another way to strengthen analysis is triangulation by cross-checking information from different sources. We can check between methodologies, quantitative and qualitative, between methods– for example, interviews, focus groups. Or observations between respondents– for example, compare narratives from different groups of participants like caregivers and health workers, and between investigators by using different researchers to collect and analyse.

Where possible, it’s important to look at how people with disabilities can also participate and contribute to analysis of your data, as stated in the principle nothing about us without us. It’s their right to be involved, not just in conducting the research, but also in the analysis. It can aid interpretation, help to validate and contextualise some of the findings. A key point to remember is that their participation must be meaningful, not tokenistic. In one of five studies in Nepal, called strengthening the voices of adolescents with disabilities, we worked with youth research associates who are young people with a range of impairments. During data collection, they discussed emergent themes at the end of each day.

After fieldwork, we had a two-day workshop with the youth researchers to analyse the key themes. Each pair of researchers worked on an analysis of their interviews. There was then an opportunity to discuss key themes and to look at differences across regions and experiences across impairments. At the end of the workshop, they also presented key findings to the donor organisations. We’ve provided the report for more details on this study. In summary, you should now be able to recognise how data from qualitative studies can be analysed, identify features of good quantitative analysis, and appreciate how people with disabilities can be involved in data analysis.