---
name: evidence-extraction-tables
description: Design or populate traceable evidence extraction tables from research reports, with source locations and explicit missing data.
---

# Evidence extraction tables

Use when a user needs structured information from studies or wants an extraction form. Define the research question, unit of analysis, and eligible report types before selecting fields. Separate a study from multiple publications, cohorts, arms, outcome measurements, and follow-up times so that results are not counted twice.

## Workflow

Create fields that serve the question: citation and study identifiers; design and setting; participant eligibility and counts; intervention or exposure and comparator; outcome definition, measurement and time point; numerical result and uncertainty; funding and relevant limitations. Add fields for exact source location and extraction status. Choose units, coding rules, allowed values, and how to represent missing, unclear, not applicable, and not reported data before populating rows.

Read the available source at the level required for each value. Capture the authors' original value and units alongside any conversion or calculation; show the formula and inputs for derived quantities. Do not treat an abstract-only report as a full-text extraction. Note conflicting values within or across reports and preserve the versions and locations instead of silently choosing one.

Pilot the form on a few varied studies when material is available, revise ambiguous fields, and record changes. For a multi-reviewer process, identify who extracted and checked each record, log disagreements and resolution, and mark entries awaiting verification. Do not claim independent checks that were not performed.

## Deliverable and dependencies

Return a table or editable file plus a short data dictionary, source list, and unresolved issues. Full-text access, source quality, and the chosen output format constrain completeness. Report unlocated values as missing and avoid extrapolating results beyond the reports. An extraction table is descriptive evidence, not itself a conclusion about causality or quality.