Discrepancies between qualitative and quantitative evaluation of randomised controlled trial results: Achieving clarity through mixed methods triangulation
Tonkin-Crine S., Anthierens S., Hood K., Yardley L., Cals JWL., Francis NA., Coenen S., van der Velden AW., Godycki-Cwirko M., Llor C., Butler CC., Verheij TJM., Goossens H., Little P.
© 2016 Tonkin-Crine et al. Background: Mixed methods are commonly used in health services research; however, data are not often integrated to explore complementarity of findings. A triangulation protocol is one approach to integrating such data. A retrospective triangulation protocol was carried out on mixed methods data collected as part of a process evaluation of a trial. The multi-country randomised controlled trial found that a web-based training in communication skills (including use of a patient booklet) and the use of a C-reactive protein (CRP) point-of-care test decreased antibiotic prescribing by general practitioners (GPs) for acute cough. The process evaluation investigated GPs' and patients' experiences of taking part in the trial. Methods: Three analysts independently compared findings across four data sets: qualitative data collected view semi-structured interviews with (1) 62 patients and (2) 66 GPs and quantitative data collected via questionnaires with (3) 2886 patients and (4) 346 GPs. Pairwise comparisons were made between data sets and were categorised as agreement, partial agreement, dissonance or silence. Results: Three instances of dissonance occurred in 39 independent findings. GPs and patients reported different views on the use of a CRP test. GPs felt that the test was useful in convincing patients to accept a no-antibiotic decision, but patient data suggested that this was unnecessary if a full explanation was given. Whilst qualitative data indicated all patients were generally satisfied with their consultation, quantitative data indicated highest levels of satisfaction for those receiving a detailed explanation from their GP with a booklet giving advice on self-care. Both qualitative and quantitative data sets indicated higher patient enablement for those in the communication groups who had received a booklet. Conclusions: Use of CRP tests does not appear to engage patients or influence illness perceptions and its effect is more centred on changing clinician behaviour. Communication skills and the patient booklet were relevant and useful for all patients and associated with increased patient satisfaction. A triangulation protocol to integrate qualitative and quantitative data can reveal findings that need further interpretation and also highlight areas of dissonance that lead to a deeper insight than separate analyses.