Rate certainty by testing whether population, dose, or risk of bias explains the conflicting directions before you assign an inconsistency downgrade, then record the explanation you tested and the level you chose. GRADE rates certainty at the outcome level using explicit criteria, so the judgement applies to the specific winter-health outcome under review.
- Group the trials by direction and list the population, dose, and risk-of-bias profile of each group. A pattern that tracks one of these factors gives you a candidate explanation to test.
- Run a sensitivity analysis that removes the trials carrying the suspected factor, and compare the pooled result with the full analysis. Use the Cochrane Handbook guidance on sensitivity analyses and heterogeneity to judge whether the direction of the pooled effect changes.
- Assign the inconsistency level from what the comparison shows. If the explanation accounts for the conflict, state it and set the downgrade accordingly; if the conflict persists after the test, record that result and apply the downgrade your review protocol specifies.
- Write the justification as two sentences: the explanation tested with the comparison that tested it, and the downgrade level chosen with its reason.
A worked example: if trials in older adults point one way and trials in younger adults point the other, group by age, remove the older-adult trials in a sensitivity analysis, and check whether the pooled direction shifts. Report the outcome of that check rather than the presence of conflict alone.
Keep the judgement at the outcome level and tie each downgrade to the specific comparison that produced it, so reviewers and guideline panels can trace the reasoning.