Screening automation for systematic reviews has largely been evaluated based on recall and workload reduction. Recall is however a crude measure of coverage, and a systematic review may achieve high recall but still fail to identify sufficient evidence. In this presentation we describe work on estimating when meta-analyses in systematic reviews on diagnostic accuracy are sufficiently stable that screening can be interrupted without impacting the findings of the systematic review.
Learning Objective: Understand the trade-off between screening workload and estimation accuracy in systematic reviews of diagnostic test accuracy.
Christopher Norman, Limsi-CNRS
Mariska Leeflang, Academic Medical Center
Raphaël Porcher, Université Paris Descartes
Aurelie Neveol (Presenter)