Closing the gap between ‘mean effect size’ and data desired by decision makers—exploring heterogeneity in complex interventions

Article type
Authors
Ivers N1, Grimshaw JM2, Tricco A3, Trikalinos T4, Straus S3, Dahabreh I5, Yu C6, Lavis JN7
1Women’s College Hospital, University of Toronto, Canada
2Director of the Canadian Cochrane Centre, Canada
3Li Ka Shing Knowledge Institute of St Michael’s Hospital, Canada
4Center for Evidence-based Medicine, Brown University, USA
5Brown University, USA
6Li Ka Shing Knowledge Institute, Canada
7McMaster University, Canada
Abstract
Objectives:

To discuss the role of engaging decision makers to better understand their informational needs when reviewing complex interventions. To introduce novel statistical approaches for exploring heterogeneity of complex interventions.

Description:

Limitations in the reporting of primary studies of complex interventions and in the commonly used meta-analytical methods (including approaches for exploring heterogeneity) restrict the utility of existing systematic reviews for decision makers who wish to identify and optimize the design of new initiatives for their own context. Our recent systematic review of complex QI interventions for diabetes care, will serve as the foundation for these topics. We will introduce the role of and discuss methods for engaging stakeholders. Participants will briefly play the role of various stakeholders, including patients, providers, and policy makers, as an exercise in prioritizing key questions beyond ‘mean effect size’. We will demonstrate analytical approaches, including combinatorial meta-analysis techniques and hierarchical meta-regression that can be used to pursue answers of interest to decision makers. The pros and cons of agnostic/inductive and deductive approaches to explore heterogeneity will be examined, and the role of data enrichment through author surveys will be discussed. Broad application of these techniques in reviews of heterogeneous, complex interventions will be considered.