SOAD2021 Chap.2 Desk-based Review and Evidence Searching
Desk-based Review and Evidence Searching
A desk-based review turns a broad concern into a transparent evidence search. Begin by defining purpose, population or setting, focal concept and intended decision. Translate each into alternative search terms, combine concepts deliberately, and record where and when you searched.
Searching is iterative rather than a single perfect query.
A broad pass maps language; a focused pass tests combinations and subject headings; citation tracking follows influential sources. Academic databases, government material, organisational reports and other grey literature answer different evidentiary needs. Their authority must be evaluated rather than assumed.
Selection decisions belong in a search ledger.
Record inclusion and exclusion criteria, duplicate handling, date and jurisdiction limits, and reasons for rejecting apparently relevant material. The final synthesis should distinguish consistent findings, contested findings, evidence gaps and transfer limits. A fluent summary is not enough if the reader cannot see how the evidence set was produced.
Synthesis assigns evidence a job.
Descriptive studies may establish scale, qualitative work may explain mechanisms or acceptability, and implementation reports may reveal staffing and system conditions. Compare sources within those jobs before drawing an overall conclusion. When studies disagree, inspect their populations, settings, measures and follow-up periods rather than averaging away the difference.
The review should end with a bounded answer to the decision question, an account of confidence, and a next evidence-producing action. Record whether a recommendation depends on a local assumption, whether a missing group may change the conclusion, and what new information would reverse it. This converts a bibliography into a revisable practice argument.
It also prevents a polished search process from disguising weak transfer from the studied context to the program being designed.
What this chapter covers
- 01
Review purpose
- 02
Search concepts
- 03
Boolean logic
- 04
Database selection
- 05
Grey literature
- 06
Search ledger
- 07
Evidence synthesis
Worked example · free
Worked example: construct a search for outreach engagement
- 1Separate the question into population, intervention or setting, and outcome concepts.
- 1Add synonyms within each concept using OR, then connect concept groups using AND.
- 1Choose at least one scholarly database and one relevant grey-literature source, recording the rationale.
- 1State inclusion criteria and record the exact query, date, filters, result count and screening decision.
Key terms
- Search string
- A search string is a documented combination of terms, operators and field limits used in a particular source. It should be reproducible and adapted to the source's syntax.
- Inclusion criteria
- Inclusion criteria are pre-stated features that evidence must meet to enter a review. They should follow the review purpose rather than a desired conclusion.
- Citation tracking
- Citation tracking is the use of reference lists and later citations to locate related evidence. It complements rather than replaces a structured database search.
Desk-based Review and Evidence Searching FAQ
How narrow should the first search be?
Start broad enough to learn the language used in the field, then narrow through concept combinations and justified limits. A very narrow first query can hide relevant terminology, while an unbounded final query produces an evidence set that cannot be screened transparently.
Can grey literature be credible evidence?
Yes, especially for policy, implementation and local service context. Appraise who produced it, for what purpose, with which methods and with what access to data. Publication outside a journal is neither an automatic weakness nor a guarantee of practical relevance.
Should I use artificial intelligence to generate search terms?
It can suggest vocabulary, but every term and citation requires verification. Test suggested terms in actual databases, inspect the records they retrieve and retain a human-readable audit trail. Never treat a generated reference as real until it is confirmed in a reliable source.
What belongs in the synthesis?
Organise evidence by the review question, not source order. Compare patterns, methods and settings; identify disagreement and gaps; and state how transfer to the target service may be limited. Separate what studies found from your interpretation of the combined evidence.
Assessment move
Practise with one question across two databases and one government site. Before searching, predict useful terms. After each pass, annotate which term created noise, which subject heading improved precision and which important source would have been missed. Reconstruct the final search without notes, then compare it with the ledger.
For revision, explain aloud why each inclusion criterion is necessary and what bias it might introduce. Finish by assigning every retained source one evidentiary job and removing a source whose presence does not change the synthesis or recommendation. Write the final search limit as a transparent boundary and explain how it might shape what the review concludes.
