Expert Elicitation of Study Weights for Bayesian Analysis and Meta-Analysis

Abstract

Meta-analysis and Bayesian informative prior distributions are used for updating knowledge about treatment effects with new data. When the available data come from slightly different study populations, or from slightly different trials as compared with the new data, the researcher has to specify study weights to control the influence of the historical data on the current data. This research evaluates whether an internet Delphi technique is useful to elicit valid and reliable study weights from an expert panel. Weights were elicited for four historical studies by four experts. Despite some difficulties regarding the panel, such as respondent burden, the method seemed useful within this context. The authors advise to include sensitivity analyses to assess the required number of Delphi rounds.