RESEARCH METHODS AND REPORTING A guide to systematic review and meta-analysis of prediction model performance Thomas P A Debray,1,2 Johanna A A G Damen,1,2 Kym I E Snell,3 Joie Ensor,3 Lotty Hooft,1,2 Johannes B Reitsma,1,2 Richard D Riley,3 Karel G M Moons1,2 1Cochrane Netherlands, Validation of prediction models is diagnostic test accuracy studies. Compared to therapeu- University Medical Center tic intervention and diagnostic test accuracy studies, Utrecht, PO Box 85500 Str highly recommended and increasingly there is limited guidance on the conduct of systematic 6.131, 3508 GA Utrecht, Netherlands common in the literature. A systematic reviews and meta-analysis of primary prognosis studies. 2Julius Center for Health review of validation studies is therefore A common aim of primary prognostic studies con- Sciences and Primary Care, cerns the development of so-called prognostic predic- University Medical Center helpful, with meta-analysis needed to tion models or indices. These models estimate the Utrecht, PO Box 85500 Str 6.131, 3508 GA Utrecht, summarise the predictive performance individualised probability or risk that a certain condi- Netherlands of the model being validated across tion will occur in the future by combining information 3Research Institute for Primary from multiple prognostic factors from an individual. Care and Health Sciences, Keele different settings and populations. This Unfortunately, there is often conflicting evidence about University, Staffordshire, UK article provides guidance for the predictive performance of developed prognostic Correspondence to: T P A Debray
[email protected] researchers systematically reviewing prediction models. For this reason, there is a growing Additional material is published demand for evidence synthesis of (external validation) online only.