BackgroundDelphi studies are primarily used in the health sciences to find consensus. They inform clinical practice and influence structures, processes, and framework conditions of healthcare. The practical research—how Delphi studies are conducted—has seldom been discussed methodologically or documented systematically. The aim of this scoping review is to fill this research gap and to identify shortcomings in the methodological presentation in the literature. On the basis of the analysis, we derive recommendations for the quality-assured implementation of Delphi studies.MethodsForming the basis of this scoping review are publications on consensus Delphi studies in the health sciences between January 1, 2018, and April 21, 2021, in the databases Scopus, MEDLINE via PubMed, CINAHL, and Epistemonikos. Included were publications in German and English containing the words “Delphi” in the title and “health” and “consensus” in the title or abstract. The practical research was analyzed for the qualitative content of the publications according to three deductive main categories, to which an influence on the result of Delphi studies can be imputed (expert panel, questionnaire design, process and feedback design).ResultsA total of 287 consensus Delphi studies were included in the review, whereby 43% reported having carried out a modified Delphi. In most cases, heterogeneous expert groups from research, clinical practice, health economics, and health policy were surveyed. In about a quarter of the Delphi studies, affected parties, such as patients, were part of the expert panel. In the Delphi questionnaires it was most common for standardized Likert scales to be combined with open-ended questions. Which method was used to analyze the open-ended responses was not reported in 62% of the Delphi studies. Consensus is largely (81%) defined as percentage agreement.ConclusionsThe results show considerable differences in how Delphi studies are carried out, making assessments and comparisons between them difficult. Sometimes an approach points to unintended effects, or biases in the individual judgments of the respondents and, thus, in the overall results of Delphi studies. For this reason, we extrapolate suggestions for how certain comparability and quality assurance can be achieved for Delphi studies.
Read full abstract