PROBAST: introduction to a new risk of bias tool for prediction modelling studies

Article type
Authors
Wolff R1, Whiting P2, Moons K3, Mallett S4, Riley R5, Westwood M6, Kleijnen J6
1Kleijnen Systematic Reviews Ltd
2Kleijnen Systematic Reviews
3University Medical Center Utrecht
4University of Oxford, United Kingdom
5University of Birmingham, United Kingdom
6Kleijnen Systematic Reviews Ltd, United Kingdom
Abstract
Objectives:
Quality assessment of included studies is a crucial step in any systematic review. Review and synthesis of prediction modelling studies is a relatively new and evolving area. The QUIPS tool for prognostic factor studies has been recently updated. However, a tool facilitating quality assessment for prediction modelling studies, both for diagnostic and prognostic prediction models, is needed. We have developed PROBAST, a tool for assessing the risk of bias and applicability of all types of prediction modelling studies. PROBAST assesses risk of bias and applicability of prediction modelling studies. It consists of five domains (participant selection, outcome, predictors, sample size and flow, analysis) and 27 signalling questions across these domains.

Description:
The workshop will be split into two sessions. The first session will give an overview of the development and structure of PROBAST. We have used a Delphi process, including 42 experts in the field of prediction research. The presentation will give an overview of the process, the current version of the tool (including the domains covered and signalling questions) as well as an insight into underlying discussions. In the second half of the workshop, participants will have the opportunity to experience this new tool first hand and to discuss issues with the creators.