Methods
Design and Setting
A qualitative study, comprising semi-structured interviews and focus groups, in general practices across Leeds and Bradford, cities in the North of England.
Sampling
Thirty-seven (23 %) of 158 general practices approached expressed an interest. To ensure diversity in their experiences, we sampled patients from practices with high and low opioid prescribing levels. Electronic searches identified patients aged 18 and over, with a current repeat prescription for either a strong opioid (e.g. morphine) or a weak opioid (e.g. codeine). We excluded patients with cancer or cognitive impairment. Each practice invited up to 60 patients by letter to contact the study team. As recruitment progressed, we purposively sampled by sex, age, recent significant changes in opioid prescriptions, opioid strength, the presence of coded mental health problems and ethnicity in order to maximise the range of relevant attributes.
We invited individual GPs from the 37 practices to take part in a focus group held in their locality, seeking one GP from each practice.
Data Collection
Semi-structured interviews (by CM and LZ) with individual patients covered accounts of their prescribing histories, experiences of treatment, and hopes and expectations (Appendices 1 and 2).
Early patient interviews informed the development of patient vignettes to prompt the subsequent focus group discussions with GPs (facilitated by RF and CM). The two vignettes illustrated patients experiencing suboptimal pain control on opioids. We prompted discussion on experiences and expectations of the initiation and maintenance of opioid prescribing, management options, perceived 'red flags' for long-term opioid use and experiences of trying to reduce prescribing. All interviews and focus groups were audio taped and transcribed verbatim following written consent.
Analysis
Constant comparison was used to analyse transcriptions. Analysis involved deconstructing each transcript to identify primary categories (open coding). These categories were compared with others within and across transcripts, followed by cross-linking of categories to generate new meanings and concepts (axial coding) then cross-linking of concepts to generate themes (selective coding). Interviews and focus groups were initially coded separately. Coding reliability was checked by two further researchers versed in qualitative analysis (SJC and LZ).
Data collection and analysis proceeded iteratively, with focus group findings informing later patient interviews, allowing progressive focusing on key themes to explain findings. The wider study team also read selected transcripts before discussing emerging findings and analyses. From the two perspectives (patient and GP) we identified, we derived an account of the main features of transactions between patients and GPs and subsequent prescribing decisions.
Ethical Approval
The study was approved by Yorkshire and the Humber (Humber Bridge) NRES Committee 12/YH/0109.