revisiaLog inSign up

Risk of bias figure › ROBINS-I

ROBINS-I traffic light plot

Risk of bias in non-randomized studies of interventions, domain by domain. With the 2016 version and the 2025 V2 draft.

What ROBINS-I is

ROBINS-I assesses the risk of bias in a result from a non-randomized study of an intervention, compared with a hypothetical target trial (Sterne JAC et al. BMJ 2016;355:i4919). It has seven domains: two before the intervention (confounding and selection), one at the start of the intervention (classification) and four after it.

Version 2 was posted as a draft on 20 November 2025 and may still change. It has six domains: it drops deviations from intended interventions, which are now handled within confounding when the analysis estimates a per-protocol effect, and moves classification ahead of selection. To draw it, pick “ROBINS-I V2” in the instrument list.

The ROBINS-I domains

2016 version, seven domains:

  • D1 Bias due to confounding
  • D2 Bias in selection of participants into the study
  • D3 Bias in classification of interventions
  • D4 Bias due to deviations from intended interventions
  • D5 Bias due to missing data
  • D6 Bias in measurement of outcomes
  • D7 Bias in selection of the reported result

Low · Moderate · Serious · Critical. Grey: cells not filled in yet.

V2 draft of 20 November 2025, six domains:

  • D1 Bias due to confounding
  • D2 Bias in classification of interventions
  • D3 Bias in selection of participants into the study (or into the analysis)
  • D4 Bias due to missing data
  • D5 Bias in measurement of the outcome
  • D6 Bias in selection of the reported result

Low · Moderate · Serious · Critical · Low except for concerns about uncontrolled confounding. Grey: cells not filled in yet.

Judgements and the overall judgement

Each domain is low, moderate, serious or critical risk; the 2016 version also allows “no information” (V2 does not). Critical means the study is too problematic to provide useful evidence, and such a result is generally left out of the synthesis (in the 2016 version and in V2).

The overall judgement starts from the worst domain: critical if any is critical, serious if any is serious, moderate if any is moderate, low only if all are low. It can go up a level when several domains add up: several moderate can make serious, several serious can make critical. In V2 the best overall level is called “low risk of bias except for concerns about uncontrolled confounding”.

How to make the ROBINS-I figure

  1. The table above is already set to ROBINS-I. Replace the example studies with yours, or paste the table from Excel or Google Sheets (the robvis template works).
  2. Pick the judgement for each domain and, if the instrument has one, press “Fill in the overall judgement” and correct what needs correcting.
  3. Download the traffic light plot, the summary plot or both: PDF or SVG when the journal wants vector artwork; PNG or TIFF at the column width and 300 or 600 dpi when it wants an image.

Frequently asked questions

What is the difference between ROBINS-I and ROBINS-E?

ROBINS-I is for interventions: treatments, programmes or policies someone decides to apply. ROBINS-E is for exposures nobody assigns, such as pollution, diet or occupation (Higgins JPT et al. Environ Int 2024;186:108602). Both are in the instrument list.

Should I use ROBINS-I or ROBINS-I V2?

V2 is a draft dated 20 November 2025 and its authors say it may change. If your protocol just says ROBINS-I, use the 2016 version. If you use V2, cite it as a draft with its date.

What does critical risk of bias mean?

That the study is so problematic in that domain that it provides no useful evidence on the effect. A single critical domain makes the overall judgement critical, and such a result is generally left out of the synthesis.

Can I use the figure in my paper?

Yes. Unless you are signed in to a revisia account with a verified email, what you download carries a small revisia mark under the figure, outside the drawing. With a verified email the figure comes out clean, in the preview and in every format. The account is free.

The figure is the last step of the assessment. revisia does the ones before it.

revisia assesses the risk of bias of every included study, domain by domain, and points to the sentence in the paper that supports each judgement so you can confirm or change it. The figure draws itself from what you decide. The account is free; the AI assessment is part of the paid plans.

Create a free account

Domains and judgement levels are those published by each instrument’s authors: RoB 2, Sterne JAC et al. BMJ 2019;366:l4898; RoB 1, Higgins JPT et al. BMJ 2011;343:d5928; ROBINS-I, Sterne JAC et al. BMJ 2016;355:i4919; ROBINS-I V2, draft of 20 November 2025 (riskofbias.info); ROBINS-E, Higgins JPT et al. Environ Int 2024;186:108602; QUADAS-2, Whiting PF et al. Ann Intern Med 2011;155:529-36; PROBAST, Wolff RF et al. Ann Intern Med 2019;170:51-8; Newcastle-Ottawa Scale, Wells GA et al., Ottawa Hospital Research Institute; JBI checklist for analytical cross-sectional studies, JBI. The traffic light and summary plots follow the format of the robvis package: McGuinness LA, Higgins JPT. Risk-of-bias VISualization (robvis): an R package and Shiny web app for visualizing risk-of-bias assessments. Res Synth Methods 2021;12:55-61. This tool is independent: it is not affiliated with, endorsed by or maintained by Cochrane, the instruments’ authors or robvis. Your data never leaves your browser. Other free tools: all tools · PRISMA flow diagram.