Nature and Predictors of Response Changes in Modified-Delphi Panels
Published in: Value in Health, Volume 23, Issue 12, pages 1630–1638 (December 2020). doi: 10.1016/j.jval.2020.08.2093
Posted on RAND.org on January 07, 2021
To describe the extent and nature of changes in participants' responses after exposure to group feedback and discussion in modified-Delphi panels and to identify factors affecting those changes.
We analyzed data from 2 online modified-Delphi panels, each consisting of 2 rating rounds and an online discussion round. We included responses from 55 participants who answered 38 questions in both rating rounds. Because not all participants answered each question twice, our sample consisted of 1846 cases (response changes). We used mixed-effect logistic and multinomial logistic regression to identify factors predicting response changes and their direction relative to group median—our consensus measure.
Participants changed, on average, 49% of their responses. A response was changed in 47% of the 1846 cases: 28% of responses were changed toward consensus and 19% away from it. Although some measures of subjective participation experiences had a marginally significant impact on the propensity and direction of response changes, several objective measures of discussion engagement were statistically significant predictors of both the presence and direction of response changes.
Our results illustrate the nature of response changes and highlight the importance of exposing participants to alternative perspectives and encouraging them to explain their perspectives.