ROBINS-E
Risk Of Bias In Non-randomized Studies — of Exposures
CoreSR implements the ROBINS-E tool with structured GLASS assessment. The model answers signaling questions across seven bias domains for non-randomized studies of exposures, and the tool keeps transparent rationale for every judgement.
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 output highlights uncertainty and source-evidence gaps, 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 for non-randomized exposure studies.
Step-by-Step Guide
Each step builds on the previous one. AI assessment runs automatically after you provide the study details.
Upload Study PDFs
Upload one or more full-text PDF files for the non-randomized exposure study you want to assess. The AI extracts text from the PDFs to inform its domain-level judgements.
Enter Study Details
Provide the study title, authors, publication year, exposure of interest, and comparator. Specify the target trial design that the observational study is emulating.
Run AI Assessment
Click "Evaluate with AI" to run a structured GLASS assessment. The model answers ROBINS-E signaling questions for each domain, and the tool derives domain and overall judgements from those structured answers.
Review and Override
Review each domain's signaling questions, AI justifications, and judgements. You can override any answer or domain judgement. Pay special attention to confounding and exposure classification, as these often require expert knowledge.
The Seven Domains
ROBINS-E evaluates risk of bias across seven domains, each assessed through a series of signaling questions tailored to exposure studies.
Bias due to confounding
Did the authors control for all the important confounding factors for which this was necessary, and were those factors measured validly and reliably? Did they control for variables after the start of the exposure period that could have been affected by the exposure? Did negative controls, or other considerations, suggest serious uncontrolled confounding? When exposure changes over follow-up and with prognostic factors, a second variant asks whether the analysis method could control for time-varying as well as baseline confounding.
Bias arising from measurement of the exposure
Does the measured exposure well-characterize the exposure metric of interest? Was the exposure likely to be measured with error, or misclassified? Could the mismeasurement or misclassification have been differential (related to the outcome or the risk of the outcome)? Is non-differential measurement error likely to bias the estimated effect of exposure on outcome?
Bias in selection of participants into the study (or into the analysis)
Did follow-up begin at (or close to) the start of the exposure window for most participants, and if not, is the effect of exposure likely to be constant over the follow-up analysed? Was selection into the study (or into the analysis) based on characteristics observed after the start of the exposure window that were influenced by the exposure and by the outcome? Did the analysis, or sensitivity analyses, deal with these selection biases?
Bias due to post-exposure interventions
Were there post-exposure interventions that were influenced by prior exposure during the follow-up period? If so, is it likely that the analysis corrected for their effect?
Bias due to missing data
Were complete data on exposure status, the outcome and confounding variables available for all, or nearly all, participants? If not: was the result based on a complete case analysis, on imputation or on another correction method, was that done appropriately, and is there evidence that the result was not biased by missing data?
Bias arising from measurement of the outcome
Could measurement or ascertainment of the outcome have differed between exposure groups or levels of exposure? Were outcome assessors aware of study participants' exposure history, and could that knowledge have influenced the assessment of the outcome?
Bias in selection of the reported result
Was the result reported in accordance with an available, pre-determined analysis plan? Is the reported effect estimate likely to have been selected, on the basis of the results, from multiple exposure measurements, multiple outcome measurements, multiple analyses or different subgroups?
Structured AI Review Process
GLASS answers structured signaling questions, then the tool stores domain judgements with source-grounded rationale for human review.
Structured Answers
The assessment is built from signaling-question answers rather than an unstructured overall guess. Each answer includes rationale and confidence so the basis for the domain judgement is inspectable.
Uncertainty Flags
Low-confidence answers, no-information answers, and sparse source evidence remain visible in the domain details so reviewers can focus attention where the automated assessment is least certain.
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
There is little or no concern about bias with regard to this domain. For Domain 1 (and overall) this level is "Low risk of bias except for concerns about uncontrolled confounding": confounding is well addressed but cannot be ruled out in an observational study.
There is some concern about bias with regard to this domain, although it is not clear that there is an important risk of bias.
The study has some important problems in this domain: its characteristics give rise to a high risk of bias.
The study is very problematic in this domain: its characteristics give rise to a very high risk of bias.
Not a level of the official tool; CoreSR keeps it for a domain the report gives no basis to judge.
Frequently Asked Questions
Common questions about ROBINS-E assessments in CoreSR.
What is ROBINS-E?
How does ROBINS-E differ from ROBINS-I?
When should I use ROBINS-E?
How does confounding work in exposure studies?
What is the "target trial" in ROBINS-E?
How does the AI review work?
How should I handle time-varying exposures?
Can I export the results?
Ready to assess?
Upload a study PDF and let AI provide a first-pass ROBINS-E risk of bias assessment.
Go to ROBINS-EReferences
1. 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
2. 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
3. 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. doi:10.1016/j.jclinepi.2018.01.012