Objectives: To develop methods of meta-analysis which will permit the inclusion of qualitative data as well as randomised and non-randomised quantitative study evidence. This should address deficiencies in current methods of meta-analysis by allowing for the systematic inclusion of all of the available research evidence, which is particularly important when meta-analyses form the basis for policy decision making in health-related contexts.

Methods: Several alternatives have been explored, one of which extends existing methods of Bayesian hierarchical meta-analysis. Using this approach, the qualitative data is synthesised within an elicited prior, and the quantitative data is combined in the likelihood, yielding posterior results based on both sources of evidence.

Results: The feasibility of this approach is demonstrated using an example concerning what the determinants of uptake of childhood immunisation in developed countries might be. Syntheses of the available qualitative and quantitative data yield quite dissimilar pictures regarding determinants of uptake, and differences in the results of the meta-analysis based on the inclusion and exclusion of the qualitative data are explored.

Discussion: The value of this approach is critically appraised, extensions are offered, and implications for the Cochrane Collaboration are highlighted.