PRESS Checklist Explained: How to Peer Review a Literature Search Strategy, Element by Element (PRESS 2015)

The PRESS checklist comes from the PRESS 2015 guideline statement (Peer Review of Electronic Search Strategies), an evidence-based tool for checking a literature search strategy before it is run [1]. A second searcher works through six elements to find errors and omissions in the electronic search strategies for each database [1].

PRESS covers the search itself, not the eligibility criteria or synthesis. PRESS recommends reviewing the primary strategy, usually MEDLINE, once it is built and before it is translated to other databases [4], when changes cost little. It applies to any literature review with a structured search, including a systematic review, and to the Boolean search strings you build for it.

Table 1 summarises each element of the PRESS 2015 guideline; read the published checklist for its full wording.

Elements 1 and 2 deserve the most time. In the expert survey behind PRESS 2015, 81% rated translation of the research question as the most important element, and errors in Boolean and proximity operators received the highest “most severe” ratings for their potential effect on recall (77%) and precision (67%) [4].

Elements 1 and 2: question and operators

Lay the question’s concepts next to the search lines. Each concept should map to one block of terms, and each block should be there on purpose. Then read the Boolean operators: synonyms joined with OR and bracketed, blocks joined with AND, and NOT used only where a test shows it removes nothing relevant.

Element 3: subject headings

• Is the heading the one this database uses? MEDLINE indexes with MeSH [3]; Embase uses Emtree, and CINAHL uses CINAHL Headings.
• Is each heading relevant, and is explosion to narrower subject headings intended?
• Is there a free-text term for every heading, to catch records not yet indexed [5]?

Element 4: text word searching

• Are there synonyms, acronyms and spelling variants, including German terms where the field uses them?
• Does truncation catch the endings you want without pulling in unrelated words?
• Is text word searching done in suitable fields, such as title and abstract?

Elements 5 and 6: syntax and limits

• Does each line number refer to the right earlier line, and does the syntax suit the platform?
• Are search limits by date, language or study design justified in the protocol?
• Could a design filter or language limit remove studies you would want to read?

1. The searcher shares the research question and the draft strategy for each database, with hits per line.
2. The reviewer checks the six elements in order and comments against each line.
3. The searcher accepts or rejects each comment and records why.
4. The strategy is revised and run; the final version goes into the search report.
5. The review is described in the methods.

If a comment is disputed, test it: run the line both ways, compare the counts and look at what changes. Record each comment and decision in the search log so the report can show what the search strategy peer review changed.

PRISMA-S item 8 asks authors to describe any search peer review process [2]. A short methods statement is enough, for example: “The MEDLINE strategy was peer reviewed by [name, role] using the PRESS 2015 guideline before the final search on [date]; [n] comments were made and [n] accepted.” Keep the comment log with the search log, and report grey literature sources in the same way (see grey literature and PRISMA-S).

• Reviewing too late. A strategy reviewed after the searches are run can only be re-run, not improved.
• Using NOT casually. NOT can silently remove relevant records, which is why PRESS flags unintended exclusions.
• Copying syntax between databases. Subject headings and operators differ by platform.
• Not recording the review. If it is not reported, a reader cannot tell it happened.