coreSR.AI
Learn

Risk of Bias 2.0

AI-assisted risk of bias assessment for randomized controlled trials

CoreSR implements the revised Cochrane RoB 2.0 tool with GLASS structured assessment. The model evaluate each domain, flag disagreements, and provide transparent rationale for every judgement.

Watch the 3-minute tutorial

AI Assessment Is a Starting Point

AI-generated assessments should always be reviewed and verified by a human assessor with domain expertise. The structured AI assessment highlights the evidence used, but the final judgement is yours. For published systematic reviews, all RoB assessments should be conducted by at least two independent human reviewers.

How It Works

Five steps from PDF upload to a complete domain-level risk of bias assessment.

1UploadPDFsFull-text papersSupplements2StudyDetailsTitle & authorsIntervention3Dual AIReviewAI Reviewer 1AI Reviewer 24DomainJudgements5 domainsSignaling Qs5Review &OverrideUser overridesExport

Step-by-Step Guide

Each step builds on the previous one. AI assessment runs automatically after you provide the study details.

Step 1

Upload Study PDFs

Upload one or more full-text PDF files for the study you want to assess. The AI extracts text from the PDFs to inform its domain-level judgements.

You can upload multiple PDFs per study (e.g., main paper + supplementary appendix) for more complete assessment.
Step 2

Enter Study Details

Provide the study title, authors, publication year, experimental intervention, and comparator. These details help the AI contextualise the assessment.

If you created the study from the Reference Screener, some fields may be pre-filled.
Step 3

Run AI Assessment

Click "Evaluate with AI" to run the GLASS structured assessment. GLASS answers each domain signaling question, and RoB 2 judgements are computed from those answers.

Assessment typically takes 30-60 seconds per study depending on PDF length.
Step 4

Review and Override

Review each domain’s signaling questions, quotes, and judgements. Switch on "Show AI reasoning" to see the one-line rationale behind each answer, which answers decided each domain rating, and whether each quote is about the outcome being assessed. You can override any answer or domain judgement.

Focus extra attention on low-confidence answers, quotes flagged as not about the assessed outcome, and sparse evidence.

The Five Domains

RoB 2.0 evaluates risk of bias across five domains, each assessed through a series of signaling questions.

D1

Randomisation process

Was the allocation sequence random? Was the allocation sequence concealed? Were there baseline imbalances suggesting problems with randomisation?

D2

Deviations from intended interventions

Were participants aware of their assigned intervention? Were there deviations from the intended intervention beyond what would be expected in usual practice?

D3

Missing outcome data

Were data for this outcome available for all, or nearly all, participants randomised? Is there evidence that the result was not biased by missing outcome data?

D4

Measurement of the outcome

Was the method of measuring the outcome inappropriate? Could measurement or ascertainment of the outcome have differed between intervention groups?

D5

Selection of the reported result

Were the data that produced this result analysed in accordance with a pre-specified analysis plan? Is the numerical result being assessed likely to have been selected from multiple eligible outcome measurements or analyses?

Dual AI Review Process

GLASS answers each domain's signaling questions, then the RoB 2 algorithm computes judgements.

When Reviewers Agree

If the signaling-question answers support a clear domain judgement, the computed judgement is used directly. A green "AI Agreed" badge is shown. The combined rationale is displayed for transparency.

When Reviewers Disagree

If the reviewers reach different judgements, the more conservative (higher risk) judgement is applied as a default. An orange "Disagreed" badge flags the domain. Expanding the domain shows each reviewer's individual judgement and a resolution rationale explaining why they differed.

Human Override

You can override any signaling question answer or domain judgement. Your decision always takes precedence over the AI. Overrides are tracked and clearly marked in exports for audit trails.

Judgement Scale

Low risk of bias

The study is judged to be at low risk of bias for this domain.

Some concerns

The study raises some concerns about risk of bias but is not clearly high risk.

High risk of bias

The study is judged to be at high risk of bias for this domain.

Frequently Asked Questions

Common questions about RoB 2.0 assessments in CoreSR.

What is RoB 2.0?
RoB 2.0 (revised Cochrane risk-of-bias tool for randomized trials) is a structured framework for assessing risk of bias in randomized controlled trials. It evaluates five domains through a series of signaling questions, producing a domain-level judgement of "Low", "Some concerns", or "High" risk of bias.
How does the AI review work?
GLASS answers each RoB 2 signaling question with rationale and evidence. The app then computes domain and overall judgements from those structured answers.
Can I override the AI’s judgements?
Yes. Every signaling question answer and domain-level judgement can be overridden. Your override takes precedence over the AI’s assessment. Overrides are clearly marked in the interface and exports.
What PDF formats are supported?
Standard PDF files with embedded text are fully supported. Scanned PDFs (image-only) may have reduced accuracy as they require OCR. For best results, use publisher PDFs rather than scanned copies.
Can I assess multiple outcomes per study?
Yes. You can define multiple outcomes and run parallel assessments. Each outcome receives its own set of domain-level judgements, as risk of bias can differ across outcomes within the same study.
Does my PDF data leave the server?
GLASS performs the primary RoB 2 assessment. Configured independent review and recovery paths can send study text or evidence excerpts to Gemini or DeepSeek. Your PDFs and assessment results are stored on the CoreSR server.
How should I interpret "Some concerns"?
A "Some concerns" judgement means the domain raises some risk of bias but it is not clearly high. Per Cochrane guidance, if multiple domains have "Some concerns", the overall study judgement may warrant "High" risk of bias due to cumulative concerns.
Can I export the results?
Yes. You can export individual study assessments or the full project summary as CSV. The export includes all domain judgements, signaling question answers, AI justifications, and any user overrides.
What is the traffic light visualisation?
The traffic light shows each domain’s judgement as a coloured circle: green (Low), yellow (Some concerns), or red (High). This provides a quick visual summary following Cochrane’s standard RoB 2.0 presentation format.
How does this compare to manual assessment?
AI assessment serves as a first pass to accelerate the process. All judgements should be reviewed by a human assessor. The structured signaling-question output helps focus human attention on domains with sparse evidence or low confidence.

Ready to assess?

Upload a study PDF and let AI provide a first-pass risk of bias assessment.

Go to RoB 2.0

References

1. Sterne JAC, Savović J, Page MJ, et al. RoB 2: a revised tool for assessing risk of bias in randomised trials. BMJ. 2019;366:l4898. doi:10.1136/bmj.l4898

2. Higgins JPT, Altman DG, Gøtzsche PC, et al. The Cochrane Collaboration's tool for assessing risk of bias in randomised trials. BMJ. 2011;343:d5928. doi:10.1136/bmj.d5928

3. Guyatt GH, Oxman AD, Vist G, et al. GRADE guidelines: 4. Rating the quality of evidence — study limitations (risk of bias). J Clin Epidemiol. 2011;64(4):407-415. doi:10.1016/j.jclinepi.2010.07.017

CoreSR · Evidence Synthesis Platform