When prioritizing clinical questions in the development of the clinical practice guidelines, clinical questions with high recognition and low variability, or high score and less disagreement among experts were often prioritized, while questions with high recognition but high variability were excluded. By this approach, clinical questions with practical value but also showed high variability due to different causes were not accepted as priorities. There were some methodological and clinical limitations by doing so. By summarizing the causes and connotations of expert opinion variability in terms of clinical experience, expertise and values, this paper analyzed the advantages of the variability quantification application, and proposed corresponding methodological recommendations, so as to provide references for guideline developers in the priority selection of clinical questions.
ObjectiveThe application of the coefficient of variation (CV) in the development of clinical practice guidelines is limited to evaluating the consistency of the consensus panel in clinical questions rating, and the application of variability was limited. This study presents the application and results of variability evaluation in the development of guidelines. MethodsWe conducted a large-scale clinical survey through questionnaire survey, and conducted two rounds of questionnaire survey and face-to-face consensus meeting for the consensus group. Means and CV were calculated for clinical questions and outcome importance ratings. We performed the summary and analysis by SPSS and Microsoft Excel. ResultsA total of 356 clinical survey questionnaires and two rounds survey in consensus panel were collected. We found that in the clinical survey and the first-round of the consensus panel, the CV was greater than 25% for all clinical questions regardless of the overall importance score. In the consensus panel, the results of the second-round were greatly changed. On the one hand, compared with the first-round, the CV of almost all clinical questions was smaller in the second-round, and the CV of high-priority clinical questions was less than 25%, while the clinical questions with a CV greater than 25% were of low-priority. In view of the CV of outcome importance, the clinical survey was similar to the results of the first-round of consensus panel. The CV of very important outcomes was less than 30%. In the second-round of consensus panel, the variability of very important outcomes was less than 20%. The higher the importance level of the outcome was, the smaller the CV was. ConclusionThe study of variability evaluation has practical methodological value, which can assist clinical questions and outcomes priority selection, and help to fully consider the influence of different factors and values, and develop high-quality guidelines.