- Published
- Updated
Two reviewers may extract different results from the same publication. This is not always carelessness: one may count people while the other counts procedures or uses another follow-up period. Extraction should make clear which question each number answers and where it comes from.
Scope of the decision
Connect extraction-field definitions to the analysis plan. Record outcome unit, population, time point and handling of missing information separately. A number appearing in a table is not automatically suitable for the intended analysis. Preserve uncertainty or seek appropriate clarification when necessary detail is unavailable.
Worked review
For example, a paper reports seventy procedures in fifty people. One reviewer uses fifty and the other seventy as the complication denominator. The team first examines whether the outcome is defined per person or per procedure. It chooses the field appropriate to the question and explains why the alternative was not used.
Practical checks
- Link every extracted number to its source table, population and time point.
- Verify meaning before converting between participant, device and procedure units.
- Track the resolution and any related records that need revision.
Acceptance and limits
Acceptance should include the ability to relocate the result and its denominator in the source. Averaging two conflicting extracted values is not a scientific solution. When a rule is clarified, revisit comparable earlier extraction decisions where appropriate.
Frequently asked questions
- Is the table total always the right denominator?
- No. The evaluable population for a particular outcome and time point may differ.
- Does second review only check arithmetic?
- Definitions, context and connection to the analysis question also need assessment.