RoB 2 Domain 3: Bias due to missing outcome data


These videos, originally part of the RoB 2: Learning Live webinar series, explain how to use the third domain of the RoB 2: bias due to missing outcome data. The presenters explain when bias arises and introduce the signalling questions for the domain. They then cover how to determine when sensitivity analyses provide evidence that a result was not biased by missing outcome data, and they also discuss how to assess whether missingness in the outcome depends on its true value, which is the key consideration in the domain. Finally, the presenters show the algorithm for how answers to signalling questions are mapped to judgements about risk of bias for the domains, along with worked examples.

This session was intended for people who are interested in using RoB 2 to assess risk of bias in their review. In addition to review author teams, CRG editors can learn about the risk of bias from missing outcome data so that they are able to assist authors with any queries they may have and also ensure information included in the review for this domain are relevant.

The webinar was delivered in August 2020 and below you will find the videos from the webinar, together with accompanying slides to download [PDF].

Part 1: Missing outcome data and when they lead to bias
Part 2: How do we know whether there is bias?
Part 3: Assessing the RoB due to missing outcome data


Presenter Bios

Julian Higgins is Professor of Evidence Synthesis in the University of Bristol’s Department of Population Health Sciences where he is head of the Bristol Appraisal and Review of Research (BARR) group. His research interests span all areas of systematic review and meta-analysis. Among his methods contributions are: a Bayesian approach to network meta-analysis; the I-squared statistic to quantify inconsistency across studies in a meta-analysis; simple prediction intervals for random-effects meta-analysis; a general framework for individual participant data meta-analysis; a library of prior distributions for between-study variation in a meta-analysis; and risk-of-bias assessment tools for clinical trials and other study designs. He has long been an active contributor to Cochrane, is a former member of the Cochrane Collaboration Steering Group, the Cochrane Editorial Board and the Cochrane Scientific Committee, and is currently co-convenor of the Cochrane Bias Methods Group. He has co-edited the Cochrane Handbook for Systematic Reviews of Interventions since 2003.

Jonathan Sterne is Professor of Medical Statistics and Epidemiology in the University of Bristol’s Department of Population Health Sciences, and Deputy Director of the NIHR Bristol Biomedical Research Centre. He has a longstanding interest in methodology for systematic reviews and meta-analysis, led development of the ROBINS-I tool for assessing risk of bias in non-randomised studies of interventions, and co-leads development of version 2 of the Cochrane risk of bias tool for randomised trials. He leads a large-scale collaboration of HIV cohort studies that led to advances in our understanding of prognosis of HIV positive people in the era of effective antiretroviral therapy. Jonathan is a former co-convenor of the Cochrane Bias Methods Group. He has published influential papers on reporting bias in meta-analysis, meta-epidemiology, causal inference and statistical methodology.


Part 1: Missing outcome data and when they lead to bias


Part 2: How do we know whether there is bias?


Part 3: Assessing the RoB due to missing outcome data


Additional materials

Download the slides from the webinar [PDF]