coreSR.AI
Learn

Risk of Bias Tools

ROBINS-I, ROBINS-E, and ROBUST-RCT with structured AI review

CoreSR provides structured risk of bias assessment for non-randomized studies and RCTs. GLASS evaluates each domain with structured prompts and provides transparent rationale.

AI Assessment Is a Starting Point

AI-generated assessments should always be reviewed and verified by a human assessor with domain expertise. For published systematic reviews, all risk of bias assessments should be conducted by at least two independent human reviewers.

Available Tools

Choose the appropriate tool based on your study design.

ROBINS-I

Risk Of Bias In Non-randomized Studies — of Interventions

Assesses risk of bias in non-randomized studies that compare the health effects of two or more interventions. Evaluates seven bias domains: confounding, selection of participants, classification of interventions, deviations from intended interventions, missing data, measurement of outcomes, and selection of the reported result.

ConfoundingSelection of participantsClassification of interventionsDeviations from intended interventionsMissing dataMeasurement of outcomesSelection of reported result
Judgement scale:Low /Moderate /Serious /Critical /No information

ROBINS-E

Risk Of Bias In Non-randomized Studies — of Exposures

Designed for non-randomized studies of exposures (e.g., environmental, occupational, dietary). Evaluates bias across seven domains adapted for exposure studies where the exposure is not assigned by researchers. Note that ROBINS-E does not share ROBINS-I’s judgement scale — its worst tier is "Very high", not "Critical".

ConfoundingMeasurement of the exposureSelection of participantsPost-exposure interventionsMissing dataMeasurement of the outcomeSelection of the reported result
Judgement scale:Low /Some concerns /High /Very high /No information

ROBUST-RCT

Risk Of Bias for Randomized Controlled Trials

A structured two-step approach for assessing risk of bias in RCTs. Step 1 evaluates study-level characteristics; Step 2 evaluates outcome-level characteristics. Each item produces a judgement that contributes to an overall risk of bias assessment.

RandomizationAllocation concealmentBlindingIncomplete outcome dataSelective reportingOther bias
Judgement scale:Low /Unclear /High

Step-by-Step Guide

The same workflow applies across all three tools.

Step 1

Upload Study Documents

Upload full-text PDFs for the study. The AI extracts text to inform domain-level assessments. You can upload multiple files per study (main paper + supplements).

Publisher PDFs with embedded text work best. Scanned PDFs may have reduced accuracy.
Step 2

Enter Study Details

Provide study name, reference, and key details. These help the AI contextualise the assessment and apply the correct evaluation criteria.

If you used bulk upload, the study name is pre-filled from the PDF filename.
Step 3

Run AI Assessment

GLASS evaluates each domain and its signaling questions with structured prompts. Framework-specific logic computes or supports the final judgement.

Assessment typically takes 30-60 seconds per study.
Step 4

Review and Override

Review each domain’s judgements and AI justifications. Override any answer. Low-confidence or incomplete evidence should get extra attention.

Always verify AI judgements — the final assessment is your responsibility.

Dual AI Review

GLASS assesses each domain with structured prompts and evidence-linked rationale.

Agreement

When both reviewers reach the same judgement, it is used directly with a green "AI Agreed" badge.

Disagreement

When reviewers disagree, the more conservative judgement is applied. An orange badge flags it for your attention, with each reviewer's reasoning shown.

Human Override

You can override any judgement. Your decision always takes precedence and is clearly tracked for audit trails.

Frequently Asked Questions

Common questions about risk of bias tools in CoreSR.

What is the difference between ROBINS-I and ROBINS-E?
ROBINS-I assesses bias in studies of interventions (treatments, procedures), while ROBINS-E assesses bias in studies of exposures (environmental, dietary, occupational). The domain structures are similar but the signaling questions are adapted to the study type.
When should I use ROBUST-RCT vs RoB 2.0?
RoB 2.0 (available as a separate module) is the revised Cochrane tool specifically for individually-randomized parallel-group trials. ROBUST-RCT is an alternative structured approach. Both assess RCTs but use different frameworks and signaling questions.
How does the AI review work?
GLASS evaluates the structured framework questions for each domain. The app stores the answer, rationale, evidence quote, and computed or suggested judgement for human review.
Can I override the AI’s judgements?
Yes. Every signaling question and domain judgement can be overridden. Your decision always takes precedence. Overrides are tracked and clearly marked in exports.
Does the AI assessment replace human review?
No. AI assessment is a starting point to accelerate the process. For published systematic reviews, risk of bias should be assessed by at least two independent human reviewers. The AI helps by providing a first pass and flagging areas of disagreement.
Can I export the results?
Yes. You can export individual study assessments or the full project summary. The export includes domain judgements, AI justifications, and any user overrides.
What PDF formats are supported?
Standard PDF files with embedded text are fully supported. Scanned (image-only) PDFs may have reduced accuracy. Use publisher PDFs when possible.
Can I use bulk upload?
Yes. In RoB 2.0 projects, you can upload multiple PDFs at once and each will create a new assessment automatically. For ROBINS-I/E and ROBUST-RCT, you add studies individually and then upload PDFs to each.

Ready to assess?

Create a project and start assessing risk of bias with AI assistance.

Go to Risk of Bias

References

1. Sterne JAC, Hernán MA, Reeves BC, et al. ROBINS-I: a tool for assessing risk of bias in non-randomised studies of interventions. BMJ. 2016;355:i4919. doi:10.1136/bmj.i4919

2. ROBINS-E development group. Risk Of Bias In Non-randomized Studies — of Exposures (ROBINS-E). Environ Int. 2022;154:107108. doi:10.1016/j.envint.2022.107108

3. Wang Y, Keitz S, Briel M, et al. Development of ROBUST-RCT. BMJ. 2025;388:e081199. doi:10.1136/bmj-2024-081199

4. Schünemann HJ, Cuello C, Akl EA, et al. GRADE guidelines 18: How ROBINS-I and other tools to assess risk of bias in nonrandomized studies should be used. J Clin Epidemiol. 2019;111:105-114.

CoreSR · Evidence Synthesis Platform