Fix a missing-outcome-data problem by identifying which studies have missing outcome data and whether they report it, then applying a prespecified imputation or sensitivity analysis and documenting how the choice changes the pooled estimate.
Work through the problem in this order
- Map the missingness study by study. For each included study, record which outcome is missing, how many participants are affected, and whether the report states the reason or the handling method. This tells you whether the gap sits in one trial or across several.
- Decide on a handling approach before you look at the pooled result. Cochrane Handbook Chapter 10 covers missing data alongside sensitivity analyses, so set the imputation or sensitivity rule in advance and apply it consistently across studies.
- Run the analysis with and without the chosen handling. Compare the pooled estimate and its confidence interval across the two runs. A shift in the estimate or a widened interval shows how much the missing data influences the conclusion.
- Report the choice and its effect. State which studies contributed missing data, the handling method used, and the change in the pooled estimate. Cochrane Handbook Chapter 15 supports interpreting results through confidence intervals and certainty, so frame the missing-data handling as part of that interpretation.
What changes the decision
If the missingness is concentrated in one study, a sensitivity analysis that excludes or reweights that study shows whether the pooled result rests on it. If several studies are affected, the imputation rule itself becomes the main driver, and the comparison across runs matters more than any single pooled number.
Documentation to include
Record the affected studies, the number of participants with missing outcomes, the prespecified handling method, and the pooled estimate from each analysis. This record lets readers and reviewers see how the missing data shaped the result.