Searching the literature so that it can be checked
A search you cannot describe is a search nobody can evaluate, including you in six months.
Where to look
| Source | Covers | Note |
|---|---|---|
| PsycINFO | Psychology and psychotherapy, comprehensively | The primary database for this field; APA thesaurus terms |
| MEDLINE / PubMed | Biomedical, including psychiatry | MeSH terms; free via PubMed |
| Embase | Biomedical, stronger European and pharmaceutical coverage | Subscription; important for systematic reviews |
| CENTRAL (Cochrane) | Randomised trials | Essential for any review of trials |
| Web of Science / Scopus | Cross-disciplinary; citation tracking | Use for forward citation searching |
| ClinicalTrials.gov / ISRCTN | Registered trials, including unpublished | The main tool against publication bias |
| PROSPERO | Registered systematic reviews | Check before you start; someone may be doing it |
| Google Scholar | Everything, imprecisely | Good for finding a known item and for grey literature; not reproducible as a primary search |
Building a search you can report
- 1Frame the question. PICO for quantitative — Population, Intervention, Comparator, Outcome. SPIDER or SPICE for qualitative.
- 2Identify concepts, then list synonyms for each: free-text terms plus controlled vocabulary (MeSH, APA thesaurus).
- 3Combine synonyms with OR inside a concept, and concepts with AND between them.
- 4Use truncation and proximity operators; check each database’s syntax because they differ.
- 5Apply limits deliberately and justify them — language limits in particular introduce bias and should be defended, not assumed.
- 6Record the exact search string, the database, the interface and the date run. Every one of them.
- 7Supplement: reference lists of included studies, forward citation searching, trial registries, and contacting authors.
The flow diagram
PRISMA’s flow diagram is not decoration. It forces you to account for every record: identified, duplicates removed, screened, excluded with reasons, full text assessed, excluded with reasons, included. If you cannot fill it in, you did not keep adequate records, and the review cannot be reported honestly.
Reference management
Use Zotero, Mendeley or EndNote from the first paper you read, not from the first draft. Retrospective reference construction is where citation errors are introduced, and errors introduced this way propagate: a substantial proportion of citations in published papers contain errors, and a meaningful fraction misrepresent what the cited source said.
Sources
Higgins, J. P. T., et al. (Eds.) (2023). Cochrane Handbook for Systematic Reviews of Interventions (v6.4). Cochrane
manualRethlefsen, M. L., et al. (2021). PRISMA-S: An extension to the PRISMA statement for reporting literature searches in systematic reviews. Systematic Reviews, 10, 39
guideline