Repair a double-counted certainty downgrade by tracing each limitation to its GRADE domain, removing the duplicate application, and rechecking the revised rating against the evidence profile. GRADE rates certainty with explicit criteria and outcome-level judgements, so the repair works at the outcome level where the downgrade was applied.
- List every downgrade recorded for the outcome and name the domain each one cites: risk of bias, inconsistency, indirectness, imprecision, or publication bias. Expected result: each limitation appears once, attached to the domain it actually concerns.
- Compare entries that describe the same underlying limitation. If one limitation was entered under two domains, or the same concern was counted in two separate judgements, keep the single entry that matches the limitation and remove the duplicate. Expected result: the number of downgrades reflects distinct limitations.
- Recalculate the certainty rating from the starting level using the remaining downgrades. Expected result: a revised rating that follows from the deduplicated set.
- Verify the revised rating against the evidence profile for that outcome, checking that each remaining downgrade still has supporting evidence and that no distinct limitation was lost in the cleanup. Expected result: a rating consistent with the documented limitations.
- Document the correction: record which limitation was double-counted, which domain it belongs to, the original and revised ratings, and the reasoning. Expected result: an audit trail that lets reviewers follow the change.
GRADE's transparent criteria and outcome-level judgements give the structure for this trace, and Cochrane's GRADE resources cover rating certainty in systematic reviews and Summary of Findings tables, which is where the corrected rating is reported. Keep the correction at the outcome level so the Summary of Findings entry matches the reasoning recorded for that outcome.