Descriptive qualitative analysis is an iterative process, with the aim of meaningfully re-classifying codes into categories and themes. Published examples of qualitative research conducted in local languages sometimes do not describe the analysis process adequately; it is often difficult to discern in what language data were analysed, how coding frames were developed and codes derived, and how consensus was reached on analysis and interpretation. Some researchers devise their own ways of assuring data quality, analysing data in the local language using translated summaries of relevant text extracts [26
], and we believe these should be made explicit.
We have described methods we used that allowed us to collect detailed qualitative data in Mandarin, manage that data effectively, produce plausible data categories and arrive at a meaningful interpretation of our data. This level of teamwork across languages was made possible by using a matrix based approach that allowed each stage of the analysis to remain visible to all researchers, and software that facilitated browsing and retrieval of relevant data. We are aware that these may not be the only solutions to the problems highlighted in this paper, and obstacles remain even within our proposed methods.
We found qualitative analysis software useful for keeping track of decisions made during analysis and for sharing the project coding and categorisation between the whole team. We are aware that there are limitations with most software packages, for example, the language restrictions in MAXqda (version 2) meant we underutilised the text search functions available. However, we are aware that the most recent version (MAXqda2007) and other software programmes such as NVivo 7 now support coding and searching in any language including Unicode [27
] character languages such as Chinese. We developed data matrices and charts using Microsoft Word, but MAXqda2007 now includes a facility to construct data matrices, tables and other visual models [28
]. In addition, the founders of the Framework approach, the National centre for Social Research, have just launched their own software with matrix capabilities, specifically for use alongside the Framework approach [29
We accept that cultural issues may only be partly responsible for the problems we encountered in obtaining depth and detail in qualitative interviews. Another important consideration is capacity in interviewing technique. Although the use of qualitative data collection methods in public health research is growing, medical and public health training in universities in mainland China emphasises epidemiological methods, and qualitative methods receive limited consideration. We believe capacity in qualitative health research will develop as researchers begin to recognise the contribution of this approach. Better integration of basic and social science research is needed in health systems and health services research, and better collaboration between researchers across cultures is important in achieving this, particularly in countries like China. International collaborative programmes present the opportunity to acquire research methods expertise and disease specific knowledge that can be applied to public health priorities in mainland China, and this will encourage greater participation of Chinese researchers [30
]. However, due to the language barrier, cultural differences, and difficulties in applying the research methods across languages, Chinese researchers may face obstacles to participating in global research programmes, and particularly in qualitative health research. We think the methodological principles outlined in this paper contribute to the growing consensus on acceptable methods for conducting qualitative health research across languages and cultures.
We are aware that in our example we relied on English summaries to form the basis of team discussions about patterns in the data, explanations of themes and interpretation. We emphasise that these translated extracts are to aid discussion; often interpreting the data requires the research team returning to individual transcripts to clarify meaning and concepts and make sure that data categories are reasonable and emerging themes are meaningful. In various research projects we have found that this stage in the analysis is the most time consuming.
When qualitative research conducted in one language is written up in another (for example in English for publication in international journals) in our experience it makes sense to provide at least some of the data as illustrative quotes in the local language. Depending on the journal, this may or may not be straightforward. Public health journals whose editors are used to publishing epidemiological research sometimes have restricted word limits. On the other hand, this can be easier in journals operating an open access policy, or who publish primarily online, as they often allow for additional tables or files.