Fix an inconsistent data-extraction form by piloting it on two studies, rewriting every ambiguous field definition, adding derivation flags, and re-extracting the studies the old form affected.
- Pilot the form on two studies and compare the completed forms field by field. Mark each field where two extractors recorded different values or interpreted the field differently; those marks identify the ambiguous fields to rewrite.
- Rewrite each flagged definition to state what belongs in the field, what stays out, and the unit or format to record. For example, specify whether a reported effect measure is recorded as published or recalculated, and which time point counts when a study reports several.
- Add a derivation flag to any field whose value is calculated or inferred from another source rather than copied from the study report. The flag lets a reader see which values came directly from the paper and which were derived.
- Document the revised form, including the changed definitions and the flag convention, so extractors work from one version.
- Re-extract the studies completed under the earlier form and compare old and new values. Resolve each difference against the study report before the data enter analysis.
Cochrane Handbook Chapter 4 covers documenting searches and managing records for systematic reviews, which supports keeping the form version and its changes on file. Cochrane Handbook Chapter 10 covers effect measures, missing data, and sensitivity analyses; use those sections when a rewritten field concerns an effect measure or a missing value, since the recorded format affects how the value is later pooled or tested.