How to Organise Evidence for a Literature Review

A literature review becomes difficult to defend when the research team remembers conclusions but cannot easily recover which source supported them.

Organisation should therefore preserve more than bibliographic references. It should connect each source to the evidence extracted from it and the synthesis decisions made later.

Start with one stable record for each source. Record enough bibliographic information to identify it reliably, then add fields relevant to the review such as study setting, population, method, sample, intervention or theoretical approach.

Extract findings with context. A sentence removed from the results section may mean something different once the study design, subgroup or limitation is considered.

Avoid creating one giant summary per article. Structured extraction makes comparison easier. Depending on the review, fields may include:

• research question;
• design;
• setting;
• key finding;
• evidence or quotation;
• limitation;
• relevance to the synthesis question.

When several reviewers work together, agree on the extraction rules before the dataset becomes large. Pilot the form on a small number of contrasting studies.

Keep synthesis separate from extraction. The source says one thing; the reviewer interprets its relevance to the wider question. Both are important, but they should not be confused.

Literature reviews also benefit from negative evidence. Record studies that contradict the emerging narrative rather than allowing them to disappear because they are inconvenient to summarise.

AI can assist retrieval or preliminary organisation, but reviewers should verify citations and source claims directly. A generated synthesis should never become the only record of what the literature actually said.

The basic standard is simple: a reviewer should be able to move from a synthesis claim back to the relevant sources and, where necessary, the exact context in which those sources made the underlying claim.