A graphical meta-analysis module for facilitating transparent healthcare decision-making

Article type
Authors
Bujkiewicz S1, Lai M1, Jones H2, Turner R3, Cooper N1, Hawkins N4, Abrams K1, Spiegelhalter D5, Sutton A1
1Department of Health Sciences, University of Leicester, Leicester, UK
2Department of Social Medicine, University of Bristol, Bristol, UK
3Institute of Public Health, MRC Biostatistics Unit, Cambridge, UK
4Oxford Outcomes, Oxford, UK
5Centre for Mathematical Sciences, University of Cambridge, Cambridge, UK
Abstract
Background: We have developed an Excel based graphical user interface for evidence syntheses developed in R or WinBUGS. This interactive tool aims to facilitate transparent healthcare decisionmaking by allowing in-depth access to meta-analysis to a wider community of stakeholders. Objectives: This graphical interface allows evidence synthesis to make use of the advanced statistical and graphical analysis provided by R and WinBUGS and at the same time it enables users not only to run the meta-analysis but also to have control over its structure. It aims to facilitate interactive adjustment of meta-analysis for relevance and potential biases. Originally it has been developed as one of the modules of the Transparent Interactive Decision Interrogator (TIDI) that aims to facilitate transparent and efficient decision making in HTA. This module can be used as part of TIDI where meta-analysis is used to inform health economic decision models or as stand-alone interface to help in any healthcare decisions that are made based on evidence synthesis. We suggest this concept, could be taken further to be accessible online, and ultimately integrated into the systematic reviews within the Cochrane library. This would allow the reader to incorporate their beliefs, carry out sensitivity analysis or tailor the analysis for a particular patient population. Design: The interface is programmed in Visual Basic. It allows interactive selection of studies (to be included in the meta-analysis) from among those whose data is stored in the Excel spreadsheet of the interface. An extension to Excel, called RExcel, passes the data to the meta-analysis running in R or WinBUGS (via R2WinBUGS). The results are then returned to the Excel spreadsheet in a tabular format as well as using interactive forest plots. The interface gives also the possibility to explore the impact of adjusting the primary studies in the meta-analysis for potential biases.