Meta-analysis of diagnostic test accuracy studies for healthcare policy and decision making

Jones H1, OWEN R2, Ades AE3, Sutton AJ4, Cooper NJ4
1University of Bristol, 2 NIHR CRSU/ UNIVERSITY OF LEICESTER, 3School of Social and Community Medicine, University of Bristol, 4Department of Health Sciences, University of Leicester

Objectives: We will discuss whether and how a meta-analysis of diagnostic test accuracy can inform key decisions about testing strategies. This workshop will focus on conceptual issues rather than statistical methods, but we will signpost to the relevant methods literature where appropriate.

Description: Meta-analysis is frequently used to aid healthcare policy decision making for medical technologies. Key decisions to be made include: Should we test? Which test should we use? At which threshold should we call a continuous test result ‘positive’? How should the testing strategy depend on patient characteristics?

Standard methods for meta-analysis of diagnostic test accuracy produce a ‘summary’ estimate of sensitivity and specificity and/or a summary Receiver Operating Characteristic (ROC) curve. In this workshop we will explore (i) the extent to which these results can inform decision making; and, (ii) what information a meta-analysis model would ideally provide, to answer the questions of interest more effectively. We will discuss applied case studies including use of natriuretic peptides in the diagnosis of acute heart failure and tests for dementia and multidomain cognitive impairment following a stroke.

This will be an interactive workshop, involving group discussions. If you have a smart phone, tablet computer or laptop then we ask that you bring it, as we may use these for multiple choice voting. This is not essential, however.