skills/evidence-comparison/references/evidence-matrix.md

# Cross-study evidence matrix

Choose a compact subset of columns relevant to the user's question. Preferred structure:

| Paper (DOI/link) | Design and setting | N / population | Comparison/exposure | Outcome and timepoint | Reported result (units and uncertainty) | Main limitation | Source / verification |
|---|---|---|---|---|---|---|---|

## Evidence provenance

- **FULL TEXT / page N:** extracted passage or exact paper section supports entry.
- **ABSTRACT ONLY:** no full-text confirmation; do not attach invented page numbers.
- **METADATA ONLY:** bibliographic fields, not findings.
- **NOT REPORTED:** source checked, value not found there; does not imply universally absent.
- **NOT RETRIEVED:** needed full text or part unavailable; do not extrapolate.
- **NOT COMPARABLE:** outcome, timepoint or population differs; explain why.

## Synthesis checks

- Do different groups measure change vs final values, use different units, or report different follow-ups?
- Are directions/effect magnitudes and uncertainty truly comparable? Do not convert an odds ratio into a risk ratio without sufficient data.
- Distinguish within-group changes from between-group effects.
- Do not interpret non-significant results as demonstrated equivalence.
- Keep contradictory findings visible rather than voting by study count.