Why English? | Appraising ROBINS-E Exposure Results Through Confounding, Measurement, Missing Data and Reporting

ROBINS-E exposure study appraisal helps readers examine a specific result from a non-randomised study of environmental, occupational, behavioural or other exposures. People searching for a ROBINS-E tool, observational exposure study risk of bias or confounding assessment need more than a general list of study weaknesses. They need a defined causal effect, exposure contrast, outcome, time point and mechanism connecting each design feature to possible bias.

The official ROBINS-E page, checked on 5 November 2026, identifies the current full version for follow-up studies as 24 March 2024. It says the tool is designed primarily for systematic reviews, assesses a specific exposure-effect result, uses seven bias domains and makes three kinds of judgement after each domain: risk of bias, predicted direction and whether the bias could threaten conclusions about an important effect. The official page also links the full tool and a 2024 development paper.

Clear English prevents several category mistakes. Exposure is not the same as intervention; association is not automatically causation; no information is not low risk; and a study-wide reputation cannot replace a result-specific assessment. The language of direction also matters. Bias may move an estimate towards or away from the null, towards or away from harm, or in an unpredictable direction depending on the mechanism.

ROBINS-E has a separate owner from the existing ROBINS-I intervention-result guide, STROBE reporting guide and RECORD routinely collected data guide.

---

Use the Current ROBINS-E Version and Scope

Record the 24 March 2024 full version for follow-up studies and do not assume unsupported variants for other designs.

Bias mechanism

Define the causal contrast, exposure window, outcome and follow-up before answering any signalling question. In use the current robins-e version and scope, the question is not whether the paper sounds cautious. It is whether the specified result could differ systematically from the causal effect of interest because of study design, conduct, measurement, missingness, analysis or selective reporting.

Evidence needed for judgement

A study of air pollution and asthma controls for age and smoking but not neighbourhood deprivation, which may affect both exposure and outcome; the remaining confounding requires an evidence-linked judgement. For use the current robins-e version and scope, link each answer to a protocol, methods section, flow account, table, appendix or analysis plan. ROBINS-E signalling questions gather evidence, but the assessor still has to explain how the observed feature could bias the result, in which direction and whether the risk is high enough to threaten a conclusion about an important exposure effect.

Direction, threat and next action

The official ROBINS-E page identifies the current full version for follow-up studies as 24 March 2024; variants for other designs should not be assumed complete. This boundary is decisive for use the current robins-e version and scope. Keep three official judgement ideas separate: risk of bias, predicted direction of bias and whether the bias could threaten the conclusion about an important effect. Then state the practical next action—retain with a clear caveat, explore in sensitivity analysis, seek missing information or reconsider how the result contributes to synthesis.

---

Choose One Specific Result to Assess

Tie the assessment to one exposure contrast, outcome, time point and numerical effect estimate.

Bias mechanism

Compare the observed study with a hypothetical well-conducted target experiment that estimates the same causal effect. In choose one specific result to assess, the question is not whether the paper sounds cautious. It is whether the specified result could differ systematically from the causal effect of interest because of study design, conduct, measurement, missingness, analysis or selective reporting.

Evidence needed for judgement

A single short-term biomarker is used to represent years of exposure; the assessor asks whether the measure captures the intended exposure metric and how misclassification could shift the estimate. For choose one specific result to assess, link each answer to a protocol, methods section, flow account, table, appendix or analysis plan. ROBINS-E signalling questions gather evidence, but the assessor still has to explain how the observed feature could bias the result, in which direction and whether the risk is high enough to threaten a conclusion about an important exposure effect.

Direction, threat and next action

ROBINS-E appraises a specific result from a non-randomised exposure study, not the reputation of authors or the truth of an entire field. This boundary is decisive for choose one specific result to assess. Keep three official judgement ideas separate: risk of bias, predicted direction of bias and whether the bias could threaten the conclusion about an important effect. Then state the practical next action—retain with a clear caveat, explore in sensitivity analysis, seek missing information or reconsider how the result contributes to synthesis.

---

Specify the Causal Effect of Interest

Write the effect the study intends to estimate before judging whether its design can recover it.

Bias mechanism

List important confounding factors and explain how they were measured and controlled without adjusting for variables caused by the exposure. In specify the causal effect of interest, the question is not whether the paper sounds cautious. It is whether the specified result could differ systematically from the causal effect of interest because of study design, conduct, measurement, missingness, analysis or selective reporting.

Evidence needed for judgement

Participants enter analysis only after surviving long enough to receive an exposure classification, creating a selection route that may favour one group. For specify the causal effect of interest, link each answer to a protocol, methods section, flow account, table, appendix or analysis plan. ROBINS-E signalling questions gather evidence, but the assessor still has to explain how the observed feature could bias the result, in which direction and whether the risk is high enough to threaten a conclusion about an important exposure effect.

Direction, threat and next action

The seven domains address confounding, exposure measurement, participant selection, post-exposure interventions, missing data, outcome measurement and selective reporting. This boundary is decisive for specify the causal effect of interest. Keep three official judgement ideas separate: risk of bias, predicted direction of bias and whether the bias could threaten the conclusion about an important effect. Then state the practical next action—retain with a clear caveat, explore in sensitivity analysis, seek missing information or reconsider how the result contributes to synthesis.

---

Describe the Target Experiment

Imagine a well-conducted experiment addressing the same causal effect and compare the observed study with it.

Bias mechanism

Examine whether exposure measurement captures the metric of interest and whether error differs by outcome status or participant characteristics. In describe the target experiment, the question is not whether the paper sounds cautious. It is whether the specified result could differ systematically from the causal effect of interest because of study design, conduct, measurement, missingness, analysis or selective reporting.

Evidence needed for judgement

Highly exposed workers receive extra medical surveillance that changes detection or treatment; this post-exposure pathway can affect the outcome independently of the exposure effect of interest. For describe the target experiment, link each answer to a protocol, methods section, flow account, table, appendix or analysis plan. ROBINS-E signalling questions gather evidence, but the assessor still has to explain how the observed feature could bias the result, in which direction and whether the risk is high enough to threaten a conclusion about an important exposure effect.

Direction, threat and next action

A low-risk judgement does not prove that the exposure causes the outcome or that all confounders were measured perfectly. This boundary is decisive for describe the target experiment. Keep three official judgement ideas separate: risk of bias, predicted direction of bias and whether the bias could threaten the conclusion about an important effect. Then state the practical next action—retain with a clear caveat, explore in sensitivity analysis, seek missing information or reconsider how the result contributes to synthesis.

---

Define the Exposure Window and Contrast

State when exposure begins, how it accumulates and which levels or groups are compared.

Bias mechanism

Trace entry into the study and analysis from the start of the exposure window to avoid selection conditioned on common effects of exposure and outcome. In define the exposure window and contrast, the question is not whether the paper sounds cautious. It is whether the specified result could differ systematically from the causal effect of interest because of study design, conduct, measurement, missingness, analysis or selective reporting.

Evidence needed for judgement

Outcome loss is greater in one exposure group and related to early symptoms; a complete-case analysis may not recover the intended effect without strong assumptions. For define the exposure window and contrast, link each answer to a protocol, methods section, flow account, table, appendix or analysis plan. ROBINS-E signalling questions gather evidence, but the assessor still has to explain how the observed feature could bias the result, in which direction and whether the risk is high enough to threaten a conclusion about an important exposure effect.

Direction, threat and next action

Predicted direction may be towards or away from the null, towards or away from harm, or unpredictable; it should follow the mechanism rather than a preferred conclusion. This boundary is decisive for define the exposure window and contrast. Keep three official judgement ideas separate: risk of bias, predicted direction of bias and whether the bias could threaten the conclusion about an important effect. Then state the practical next action—retain with a clear caveat, explore in sensitivity analysis, seek missing information or reconsider how the result contributes to synthesis.

---

Define the Outcome and Follow-Up

Preserve the outcome measurement, start of follow-up and time horizon attached to the result.

Bias mechanism

Separate post-exposure interventions from the exposure itself and ask whether they create another path to the outcome. In define the outcome and follow-up, the question is not whether the paper sounds cautious. It is whether the specified result could differ systematically from the causal effect of interest because of study design, conduct, measurement, missingness, analysis or selective reporting.

Evidence needed for judgement

Assessors know exposure history when judging a subjective outcome; the possible direction of measurement bias depends on how that knowledge influences classification. For define the outcome and follow-up, link each answer to a protocol, methods section, flow account, table, appendix or analysis plan. ROBINS-E signalling questions gather evidence, but the assessor still has to explain how the observed feature could bias the result, in which direction and whether the risk is high enough to threaten a conclusion about an important exposure effect.

Direction, threat and next action

No information is not the same as low risk and should prompt a visible information-gap note. This boundary is decisive for define the outcome and follow-up. Keep three official judgement ideas separate: risk of bias, predicted direction of bias and whether the bias could threaten the conclusion about an important effect. Then state the practical next action—retain with a clear caveat, explore in sensitivity analysis, seek missing information or reconsider how the result contributes to synthesis.

---

Complete the Preliminary Considerations

Check study type, result identity, causal question and whether a full assessment is appropriate.

Bias mechanism

Account for missing exposure, confounder and outcome data and evaluate the assumptions behind complete-case or imputed analyses. In complete the preliminary considerations, the question is not whether the paper sounds cautious. It is whether the specified result could differ systematically from the causal effect of interest because of study design, conduct, measurement, missingness, analysis or selective reporting.

Evidence needed for judgement

Many exposure definitions, outcome time points and adjustment sets were possible, yet no protocol is available; selective reporting cannot be ruled out simply because one model is plausible. For complete the preliminary considerations, link each answer to a protocol, methods section, flow account, table, appendix or analysis plan. ROBINS-E signalling questions gather evidence, but the assessor still has to explain how the observed feature could bias the result, in which direction and whether the risk is high enough to threaten a conclusion about an important exposure effect.

Direction, threat and next action

Observational exposure evidence is often necessary because randomisation may be infeasible or unethical; careful appraisal should not devalue it automatically. This boundary is decisive for complete the preliminary considerations. Keep three official judgement ideas separate: risk of bias, predicted direction of bias and whether the bias could threaten the conclusion about an important effect. Then state the practical next action—retain with a clear caveat, explore in sensitivity analysis, seek missing information or reconsider how the result contributes to synthesis.

---

Domain 1: Assess Bias Due to Confounding

Identify factors that affect both exposure and outcome and evaluate whether control was adequate.

Bias mechanism

Look for prespecified analysis plans and the space of possible exposures, outcomes, subgroups and models from which the reported result could have been selected. In domain 1: assess bias due to confounding, the question is not whether the paper sounds cautious. It is whether the specified result could differ systematically from the causal effect of interest because of study design, conduct, measurement, missingness, analysis or selective reporting.

Evidence needed for judgement

A student example follows screen time and sleep but distinguishes association, confounding by household routines, measurement error and the causal effect the study intended to estimate. For domain 1: assess bias due to confounding, link each answer to a protocol, methods section, flow account, table, appendix or analysis plan. ROBINS-E signalling questions gather evidence, but the assessor still has to explain how the observed feature could bias the result, in which direction and whether the risk is high enough to threaten a conclusion about an important exposure effect.

Direction, threat and next action

Policy and personal decisions require the wider body of evidence, absolute risks, feasibility, values and local context in addition to one ROBINS-E assessment. This boundary is decisive for domain 1: assess bias due to confounding. Keep three official judgement ideas separate: risk of bias, predicted direction of bias and whether the bias could threaten the conclusion about an important effect. Then state the practical next action—retain with a clear caveat, explore in sensitivity analysis, seek missing information or reconsider how the result contributes to synthesis.

---

Measure Confounders Validly and Reliably

Poorly measured confounders can leave residual bias even when they appear in a model.

Bias mechanism

Define the causal contrast, exposure window, outcome and follow-up before answering any signalling question. In measure confounders validly and reliably, the question is not whether the paper sounds cautious. It is whether the specified result could differ systematically from the causal effect of interest because of study design, conduct, measurement, missingness, analysis or selective reporting.

Evidence needed for judgement

A study of air pollution and asthma controls for age and smoking but not neighbourhood deprivation, which may affect both exposure and outcome; the remaining confounding requires an evidence-linked judgement. For measure confounders validly and reliably, link each answer to a protocol, methods section, flow account, table, appendix or analysis plan. ROBINS-E signalling questions gather evidence, but the assessor still has to explain how the observed feature could bias the result, in which direction and whether the risk is high enough to threaten a conclusion about an important exposure effect.

Direction, threat and next action

The official ROBINS-E page identifies the current full version for follow-up studies as 24 March 2024; variants for other designs should not be assumed complete. This boundary is decisive for measure confounders validly and reliably. Keep three official judgement ideas separate: risk of bias, predicted direction of bias and whether the bias could threaten the conclusion about an important effect. Then state the practical next action—retain with a clear caveat, explore in sensitivity analysis, seek missing information or reconsider how the result contributes to synthesis.

---

Avoid Inappropriate Adjustment

Do not control mechanically for variables caused by exposure or for colliders that can open new bias paths.

Bias mechanism

Compare the observed study with a hypothetical well-conducted target experiment that estimates the same causal effect. In avoid inappropriate adjustment, the question is not whether the paper sounds cautious. It is whether the specified result could differ systematically from the causal effect of interest because of study design, conduct, measurement, missingness, analysis or selective reporting.

Evidence needed for judgement

A single short-term biomarker is used to represent years of exposure; the assessor asks whether the measure captures the intended exposure metric and how misclassification could shift the estimate. For avoid inappropriate adjustment, link each answer to a protocol, methods section, flow account, table, appendix or analysis plan. ROBINS-E signalling questions gather evidence, but the assessor still has to explain how the observed feature could bias the result, in which direction and whether the risk is high enough to threaten a conclusion about an important exposure effect.

Direction, threat and next action

ROBINS-E appraises a specific result from a non-randomised exposure study, not the reputation of authors or the truth of an entire field. This boundary is decisive for avoid inappropriate adjustment. Keep three official judgement ideas separate: risk of bias, predicted direction of bias and whether the bias could threaten the conclusion about an important effect. Then state the practical next action—retain with a clear caveat, explore in sensitivity analysis, seek missing information or reconsider how the result contributes to synthesis.

---

Domain 2: Assess Exposure Measurement

Judge whether the observed measure captures the exposure metric of interest.

Bias mechanism

List important confounding factors and explain how they were measured and controlled without adjusting for variables caused by the exposure. In domain 2: assess exposure measurement, the question is not whether the paper sounds cautious. It is whether the specified result could differ systematically from the causal effect of interest because of study design, conduct, measurement, missingness, analysis or selective reporting.

Evidence needed for judgement

Participants enter analysis only after surviving long enough to receive an exposure classification, creating a selection route that may favour one group. For domain 2: assess exposure measurement, link each answer to a protocol, methods section, flow account, table, appendix or analysis plan. ROBINS-E signalling questions gather evidence, but the assessor still has to explain how the observed feature could bias the result, in which direction and whether the risk is high enough to threaten a conclusion about an important exposure effect.

Direction, threat and next action

The seven domains address confounding, exposure measurement, participant selection, post-exposure interventions, missing data, outcome measurement and selective reporting. This boundary is decisive for domain 2: assess exposure measurement. Keep three official judgement ideas separate: risk of bias, predicted direction of bias and whether the bias could threaten the conclusion about an important effect. Then state the practical next action—retain with a clear caveat, explore in sensitivity analysis, seek missing information or reconsider how the result contributes to synthesis.

---

Predict the Direction of Exposure Misclassification

Distinguish differential from non-differential error and avoid assuming all error moves towards the null.

Bias mechanism

Examine whether exposure measurement captures the metric of interest and whether error differs by outcome status or participant characteristics. In predict the direction of exposure misclassification, the question is not whether the paper sounds cautious. It is whether the specified result could differ systematically from the causal effect of interest because of study design, conduct, measurement, missingness, analysis or selective reporting.

Evidence needed for judgement

Highly exposed workers receive extra medical surveillance that changes detection or treatment; this post-exposure pathway can affect the outcome independently of the exposure effect of interest. For predict the direction of exposure misclassification, link each answer to a protocol, methods section, flow account, table, appendix or analysis plan. ROBINS-E signalling questions gather evidence, but the assessor still has to explain how the observed feature could bias the result, in which direction and whether the risk is high enough to threaten a conclusion about an important exposure effect.

Direction, threat and next action

A low-risk judgement does not prove that the exposure causes the outcome or that all confounders were measured perfectly. This boundary is decisive for predict the direction of exposure misclassification. Keep three official judgement ideas separate: risk of bias, predicted direction of bias and whether the bias could threaten the conclusion about an important effect. Then state the practical next action—retain with a clear caveat, explore in sensitivity analysis, seek missing information or reconsider how the result contributes to synthesis.

---

Domain 3: Assess Selection Into Study or Analysis

Ask whether inclusion depended on factors influenced by exposure and related to outcome.

Bias mechanism

Trace entry into the study and analysis from the start of the exposure window to avoid selection conditioned on common effects of exposure and outcome. In domain 3: assess selection into study or analysis, the question is not whether the paper sounds cautious. It is whether the specified result could differ systematically from the causal effect of interest because of study design, conduct, measurement, missingness, analysis or selective reporting.

Evidence needed for judgement

Outcome loss is greater in one exposure group and related to early symptoms; a complete-case analysis may not recover the intended effect without strong assumptions. For domain 3: assess selection into study or analysis, link each answer to a protocol, methods section, flow account, table, appendix or analysis plan. ROBINS-E signalling questions gather evidence, but the assessor still has to explain how the observed feature could bias the result, in which direction and whether the risk is high enough to threaten a conclusion about an important exposure effect.

Direction, threat and next action

Predicted direction may be towards or away from the null, towards or away from harm, or unpredictable; it should follow the mechanism rather than a preferred conclusion. This boundary is decisive for domain 3: assess selection into study or analysis. Keep three official judgement ideas separate: risk of bias, predicted direction of bias and whether the bias could threaten the conclusion about an important effect. Then state the practical next action—retain with a clear caveat, explore in sensitivity analysis, seek missing information or reconsider how the result contributes to synthesis.

---

Align Time Zero With the Exposure Window

Prevent selection or survival requirements that arise after the causal comparison should have begun.

Bias mechanism

Separate post-exposure interventions from the exposure itself and ask whether they create another path to the outcome. In align time zero with the exposure window, the question is not whether the paper sounds cautious. It is whether the specified result could differ systematically from the causal effect of interest because of study design, conduct, measurement, missingness, analysis or selective reporting.

Evidence needed for judgement

Assessors know exposure history when judging a subjective outcome; the possible direction of measurement bias depends on how that knowledge influences classification. For align time zero with the exposure window, link each answer to a protocol, methods section, flow account, table, appendix or analysis plan. ROBINS-E signalling questions gather evidence, but the assessor still has to explain how the observed feature could bias the result, in which direction and whether the risk is high enough to threaten a conclusion about an important exposure effect.

Direction, threat and next action

No information is not the same as low risk and should prompt a visible information-gap note. This boundary is decisive for align time zero with the exposure window. Keep three official judgement ideas separate: risk of bias, predicted direction of bias and whether the bias could threaten the conclusion about an important effect. Then state the practical next action—retain with a clear caveat, explore in sensitivity analysis, seek missing information or reconsider how the result contributes to synthesis.

---

Domain 4: Assess Post-Exposure Interventions

Identify interventions triggered by exposure that also change outcome risk.

Bias mechanism

Account for missing exposure, confounder and outcome data and evaluate the assumptions behind complete-case or imputed analyses. In domain 4: assess post-exposure interventions, the question is not whether the paper sounds cautious. It is whether the specified result could differ systematically from the causal effect of interest because of study design, conduct, measurement, missingness, analysis or selective reporting.

Evidence needed for judgement

Many exposure definitions, outcome time points and adjustment sets were possible, yet no protocol is available; selective reporting cannot be ruled out simply because one model is plausible. For domain 4: assess post-exposure interventions, link each answer to a protocol, methods section, flow account, table, appendix or analysis plan. ROBINS-E signalling questions gather evidence, but the assessor still has to explain how the observed feature could bias the result, in which direction and whether the risk is high enough to threaten a conclusion about an important exposure effect.

Direction, threat and next action

Observational exposure evidence is often necessary because randomisation may be infeasible or unethical; careful appraisal should not devalue it automatically. This boundary is decisive for domain 4: assess post-exposure interventions. Keep three official judgement ideas separate: risk of bias, predicted direction of bias and whether the bias could threaten the conclusion about an important effect. Then state the practical next action—retain with a clear caveat, explore in sensitivity analysis, seek missing information or reconsider how the result contributes to synthesis.

---

Domain 5: Assess Missing Data

Examine missing exposure, confounder and outcome information and who remains in analysis.

Bias mechanism

Look for prespecified analysis plans and the space of possible exposures, outcomes, subgroups and models from which the reported result could have been selected. In domain 5: assess missing data, the question is not whether the paper sounds cautious. It is whether the specified result could differ systematically from the causal effect of interest because of study design, conduct, measurement, missingness, analysis or selective reporting.

Evidence needed for judgement

A student example follows screen time and sleep but distinguishes association, confounding by household routines, measurement error and the causal effect the study intended to estimate. For domain 5: assess missing data, link each answer to a protocol, methods section, flow account, table, appendix or analysis plan. ROBINS-E signalling questions gather evidence, but the assessor still has to explain how the observed feature could bias the result, in which direction and whether the risk is high enough to threaten a conclusion about an important exposure effect.

Direction, threat and next action

Policy and personal decisions require the wider body of evidence, absolute risks, feasibility, values and local context in addition to one ROBINS-E assessment. This boundary is decisive for domain 5: assess missing data. Keep three official judgement ideas separate: risk of bias, predicted direction of bias and whether the bias could threaten the conclusion about an important effect. Then state the practical next action—retain with a clear caveat, explore in sensitivity analysis, seek missing information or reconsider how the result contributes to synthesis.

---

Evaluate Complete-Case and Imputed Analyses

Check assumptions, predictors of missingness and alignment between imputation and analysis models.

Bias mechanism

Define the causal contrast, exposure window, outcome and follow-up before answering any signalling question. In evaluate complete-case and imputed analyses, the question is not whether the paper sounds cautious. It is whether the specified result could differ systematically from the causal effect of interest because of study design, conduct, measurement, missingness, analysis or selective reporting.

Evidence needed for judgement

A study of air pollution and asthma controls for age and smoking but not neighbourhood deprivation, which may affect both exposure and outcome; the remaining confounding requires an evidence-linked judgement. For evaluate complete-case and imputed analyses, link each answer to a protocol, methods section, flow account, table, appendix or analysis plan. ROBINS-E signalling questions gather evidence, but the assessor still has to explain how the observed feature could bias the result, in which direction and whether the risk is high enough to threaten a conclusion about an important exposure effect.

Direction, threat and next action

The official ROBINS-E page identifies the current full version for follow-up studies as 24 March 2024; variants for other designs should not be assumed complete. This boundary is decisive for evaluate complete-case and imputed analyses. Keep three official judgement ideas separate: risk of bias, predicted direction of bias and whether the bias could threaten the conclusion about an important effect. Then state the practical next action—retain with a clear caveat, explore in sensitivity analysis, seek missing information or reconsider how the result contributes to synthesis.

---

Domain 6: Assess Outcome Measurement

Judge whether outcome ascertainment differs by exposure level or knowledge of exposure history.

Bias mechanism

Compare the observed study with a hypothetical well-conducted target experiment that estimates the same causal effect. In domain 6: assess outcome measurement, the question is not whether the paper sounds cautious. It is whether the specified result could differ systematically from the causal effect of interest because of study design, conduct, measurement, missingness, analysis or selective reporting.

Evidence needed for judgement

A single short-term biomarker is used to represent years of exposure; the assessor asks whether the measure captures the intended exposure metric and how misclassification could shift the estimate. For domain 6: assess outcome measurement, link each answer to a protocol, methods section, flow account, table, appendix or analysis plan. ROBINS-E signalling questions gather evidence, but the assessor still has to explain how the observed feature could bias the result, in which direction and whether the risk is high enough to threaten a conclusion about an important exposure effect.

Direction, threat and next action

ROBINS-E appraises a specific result from a non-randomised exposure study, not the reputation of authors or the truth of an entire field. This boundary is decisive for domain 6: assess outcome measurement. Keep three official judgement ideas separate: risk of bias, predicted direction of bias and whether the bias could threaten the conclusion about an important effect. Then state the practical next action—retain with a clear caveat, explore in sensitivity analysis, seek missing information or reconsider how the result contributes to synthesis.

---

Keep Assessor Knowledge and Objectivity Visible

Subjective outcomes may be especially vulnerable when assessors know exposure status.

Bias mechanism

List important confounding factors and explain how they were measured and controlled without adjusting for variables caused by the exposure. In keep assessor knowledge and objectivity visible, the question is not whether the paper sounds cautious. It is whether the specified result could differ systematically from the causal effect of interest because of study design, conduct, measurement, missingness, analysis or selective reporting.

Evidence needed for judgement

Participants enter analysis only after surviving long enough to receive an exposure classification, creating a selection route that may favour one group. For keep assessor knowledge and objectivity visible, link each answer to a protocol, methods section, flow account, table, appendix or analysis plan. ROBINS-E signalling questions gather evidence, but the assessor still has to explain how the observed feature could bias the result, in which direction and whether the risk is high enough to threaten a conclusion about an important exposure effect.

Direction, threat and next action

The seven domains address confounding, exposure measurement, participant selection, post-exposure interventions, missing data, outcome measurement and selective reporting. This boundary is decisive for keep assessor knowledge and objectivity visible. Keep three official judgement ideas separate: risk of bias, predicted direction of bias and whether the bias could threaten the conclusion about an important effect. Then state the practical next action—retain with a clear caveat, explore in sensitivity analysis, seek missing information or reconsider how the result contributes to synthesis.

---

Domain 7: Assess Selection of the Reported Result

Look for choice among multiple exposure measures, outcomes, analyses and subgroups based on results.

Bias mechanism

Examine whether exposure measurement captures the metric of interest and whether error differs by outcome status or participant characteristics. In domain 7: assess selection of the reported result, the question is not whether the paper sounds cautious. It is whether the specified result could differ systematically from the causal effect of interest because of study design, conduct, measurement, missingness, analysis or selective reporting.

Evidence needed for judgement

Highly exposed workers receive extra medical surveillance that changes detection or treatment; this post-exposure pathway can affect the outcome independently of the exposure effect of interest. For domain 7: assess selection of the reported result, link each answer to a protocol, methods section, flow account, table, appendix or analysis plan. ROBINS-E signalling questions gather evidence, but the assessor still has to explain how the observed feature could bias the result, in which direction and whether the risk is high enough to threaten a conclusion about an important exposure effect.

Direction, threat and next action

A low-risk judgement does not prove that the exposure causes the outcome or that all confounders were measured perfectly. This boundary is decisive for domain 7: assess selection of the reported result. Keep three official judgement ideas separate: risk of bias, predicted direction of bias and whether the bias could threaten the conclusion about an important effect. Then state the practical next action—retain with a clear caveat, explore in sensitivity analysis, seek missing information or reconsider how the result contributes to synthesis.

---

Answer Signalling Questions With Evidence

Use yes, probably yes, probably no, no or no information as prompts for a documented mechanism.

Bias mechanism

Trace entry into the study and analysis from the start of the exposure window to avoid selection conditioned on common effects of exposure and outcome. In answer signalling questions with evidence, the question is not whether the paper sounds cautious. It is whether the specified result could differ systematically from the causal effect of interest because of study design, conduct, measurement, missingness, analysis or selective reporting.

Evidence needed for judgement

Outcome loss is greater in one exposure group and related to early symptoms; a complete-case analysis may not recover the intended effect without strong assumptions. For answer signalling questions with evidence, link each answer to a protocol, methods section, flow account, table, appendix or analysis plan. ROBINS-E signalling questions gather evidence, but the assessor still has to explain how the observed feature could bias the result, in which direction and whether the risk is high enough to threaten a conclusion about an important exposure effect.

Direction, threat and next action

Predicted direction may be towards or away from the null, towards or away from harm, or unpredictable; it should follow the mechanism rather than a preferred conclusion. This boundary is decisive for answer signalling questions with evidence. Keep three official judgement ideas separate: risk of bias, predicted direction of bias and whether the bias could threaten the conclusion about an important effect. Then state the practical next action—retain with a clear caveat, explore in sensitivity analysis, seek missing information or reconsider how the result contributes to synthesis.

---

Make the Three Domain Judgements Separately

State risk of bias, predicted direction and whether the threat could change an important-effect conclusion.

Bias mechanism

Separate post-exposure interventions from the exposure itself and ask whether they create another path to the outcome. In make the three domain judgements separately, the question is not whether the paper sounds cautious. It is whether the specified result could differ systematically from the causal effect of interest because of study design, conduct, measurement, missingness, analysis or selective reporting.

Evidence needed for judgement

Assessors know exposure history when judging a subjective outcome; the possible direction of measurement bias depends on how that knowledge influences classification. For make the three domain judgements separately, link each answer to a protocol, methods section, flow account, table, appendix or analysis plan. ROBINS-E signalling questions gather evidence, but the assessor still has to explain how the observed feature could bias the result, in which direction and whether the risk is high enough to threaten a conclusion about an important exposure effect.

Direction, threat and next action

No information is not the same as low risk and should prompt a visible information-gap note. This boundary is decisive for make the three domain judgements separately. Keep three official judgement ideas separate: risk of bias, predicted direction of bias and whether the bias could threaten the conclusion about an important effect. Then state the practical next action—retain with a clear caveat, explore in sensitivity analysis, seek missing information or reconsider how the result contributes to synthesis.

---

Build the Overall Judgement Across Seven Domains

Combine domain evidence transparently rather than hiding decisive concerns inside an average.

Bias mechanism

Account for missing exposure, confounder and outcome data and evaluate the assumptions behind complete-case or imputed analyses. In build the overall judgement across seven domains, the question is not whether the paper sounds cautious. It is whether the specified result could differ systematically from the causal effect of interest because of study design, conduct, measurement, missingness, analysis or selective reporting.

Evidence needed for judgement

Many exposure definitions, outcome time points and adjustment sets were possible, yet no protocol is available; selective reporting cannot be ruled out simply because one model is plausible. For build the overall judgement across seven domains, link each answer to a protocol, methods section, flow account, table, appendix or analysis plan. ROBINS-E signalling questions gather evidence, but the assessor still has to explain how the observed feature could bias the result, in which direction and whether the risk is high enough to threaten a conclusion about an important exposure effect.

Direction, threat and next action

Observational exposure evidence is often necessary because randomisation may be infeasible or unethical; careful appraisal should not devalue it automatically. This boundary is decisive for build the overall judgement across seven domains. Keep three official judgement ideas separate: risk of bias, predicted direction of bias and whether the bias could threaten the conclusion about an important effect. Then state the practical next action—retain with a clear caveat, explore in sensitivity analysis, seek missing information or reconsider how the result contributes to synthesis.

---

Communicate Exposure Evidence Responsibly

Explain what the result contributes to synthesis, policy or personal understanding without turning association into certainty.

Bias mechanism

Look for prespecified analysis plans and the space of possible exposures, outcomes, subgroups and models from which the reported result could have been selected. In communicate exposure evidence responsibly, the question is not whether the paper sounds cautious. It is whether the specified result could differ systematically from the causal effect of interest because of study design, conduct, measurement, missingness, analysis or selective reporting.

Evidence needed for judgement

A student example follows screen time and sleep but distinguishes association, confounding by household routines, measurement error and the causal effect the study intended to estimate. For communicate exposure evidence responsibly, link each answer to a protocol, methods section, flow account, table, appendix or analysis plan. ROBINS-E signalling questions gather evidence, but the assessor still has to explain how the observed feature could bias the result, in which direction and whether the risk is high enough to threaten a conclusion about an important exposure effect.

Direction, threat and next action

Policy and personal decisions require the wider body of evidence, absolute risks, feasibility, values and local context in addition to one ROBINS-E assessment. This boundary is decisive for communicate exposure evidence responsibly. Keep three official judgement ideas separate: risk of bias, predicted direction of bias and whether the bias could threaten the conclusion about an important effect. Then state the practical next action—retain with a clear caveat, explore in sensitivity analysis, seek missing information or reconsider how the result contributes to synthesis.

---

Frequently Asked Questions

What does ROBINS-E stand for?

Risk Of Bias In Non-randomized Studies—of Exposures.

Which version is current here?

The official site identifies the full version for follow-up studies as 24 March 2024.

What does ROBINS-E assess?

A specific exposure-effect result from an individual non-randomised observational study.

How many bias domains are there?

Seven.

What are the three judgement ideas?

Risk of bias, predicted direction of bias and whether the bias could threaten a conclusion about an important effect.

How is ROBINS-E different from ROBINS-I?

ROBINS-E addresses exposures; ROBINS-I addresses non-randomised studies of interventions.

Does low risk prove causation?

No. It indicates little concern from the assessed bias domains for that result, not proof of a causal effect.

Is no information the same as low risk?

No. Missing information should remain visible and may prevent reassurance.

Can ROBINS-E be used outside a systematic review?

Its official design is primarily for systematic reviews, though its result-specific reasoning can inform careful reading of individual studies.

Why predict direction of bias?

Direction helps readers understand whether a mechanism may inflate, reduce or unpredictably change an exposure-effect estimate.

---

Useful Next Reading

• Read the official ROBINS-E page and current tool.
• Read the open 2024 ROBINS-E development paper.
• Compare exposure appraisal with ROBINS-I intervention appraisal.
• Practise critical reading across claims, evidence, assumptions and bias.

Closing thought: ROBINS-E turns a broad suspicion of observational bias into a result-specific explanation of mechanism, direction, threat and responsible use.