Data synthesis

The data were analysed using thematic synthesis outlined by Thomas and Harden [30]. Accordingly, all text under the ‘results/findings’ or ‘discussion/conclusion’ section of each article were extracted. Starting with one article, SK performed line-by-line coding of the findings. At least one code was given to all statements relating to pain; example codes included ‘cause of pain unknown’ ‘no sign of swelling’ and ‘doctors do not understand’. After coding the first paper, a ‘bank’ of codes was generated which identified aspects of the pain experience and discussed with all authors. Subsequent articles were analysed similarly with new codes being added to the existing bank of codes. Codes were then organised into descriptive themes by grouping the codes based on their similarities and differences. This process involved repeated reference back to the original papers to ensure interpretations were grounded in the words of participants. Examples themes included ‘behaviour of health professionals’ and ‘feeling disbelieved’. Descriptive themes were then further interpreted to develop analytical themes through discussion with all authors. This involved a process of examining and discussing patterns and relationships across descriptive themes, which led to the emergence of more abstract messages and themes that go beyond the content in the original studies. The final themes are discussed below.