How to screen titles and abstracts for a systematic review
Short answer
Two reviewers independently screen every title and abstract against pre-specified eligibility criteria, after a short pilot to calibrate. At this stage you exclude only records that clearly do not meet the criteria; anything uncertain goes on to full-text review. Disagreements are resolved by discussion or by a third reviewer, and the numbers are reported in the PRISMA 2020 flow diagram.
Step by step
- 1
Turn the protocol into explicit eligibility criteria
Write inclusion and exclusion criteria for population, intervention, comparator, outcomes and study design, exactly as in your protocol. Put them in the order you will check them, so that the most frequent reason for exclusion is checked first.
- 2
Pilot the criteria on a sample
Have both reviewers screen the same small sample of records, compare decisions and discuss every disagreement. Clarify the criteria wherever disagreements show they are ambiguous. Some teams report an agreement statistic such as Cohen's kappa from the pilot.
- 3
Screen independently and in duplicate
The Cochrane Handbook recommends that at least two people independently assess each record. Reviewers should not see each other's decisions until both have finished, so the second opinion is truly independent.
- 4
Be inclusive at this stage
Titles and abstracts are short and often incomplete. Exclude a record only when it clearly fails a criterion; if the abstract is missing or the information is unclear, keep it for full-text review. A record wrongly excluded here is lost for good, while one wrongly included costs only a full-text check.
- 5
Resolve conflicts
When the two reviewers disagree, they discuss the record and refer to the criteria. If they still disagree, a third reviewer decides. Keep a record of how conflicts were resolved.
- 6
Move to full-text screening and record reasons
Retrieve the full text of every record that passed. At full-text stage, record the reason for each exclusion: PRISMA 2020 asks for the number of reports excluded at full text together with the reasons.
Can automation or AI do the screening?
Tools can prioritize records or suggest decisions, and the Cochrane Handbook discusses machine-learning support for screening. Whatever tool you use, report how it was used, and keep a human responsible for the final decision on every record. PRISMA 2020 asks reviewers to state whether automation tools were used and how.
How Prolytiq does this step
In Prolytiq, screening follows a two-reviewer PRISMA workflow: two independent reviewers, optional blinding of each other's decisions, and a resolver for conflicts. An AI step can pre-screen titles and abstracts against your criteria and propose a decision with a written rationale; the proposal is a suggestion, and only a human reviewer can exclude a record.
See the workflow →Frequently asked questions
Is single-reviewer screening acceptable?
The Cochrane Handbook recommends duplicate independent screening. If you use a single reviewer for part of the process, for example with a second reviewer checking a sample, state it clearly in the methods and discuss it as a limitation.
What do I do with records without an abstract?
Screen the title. If it does not let you rule the record out clearly, carry it forward to full-text review rather than excluding it.
Do I need to give reasons for exclusions at title and abstract stage?
No. PRISMA 2020 asks for the number of records excluded at this stage, and for reasons only for reports excluded at full-text stage.
Methodological sources
- Cochrane Handbook for Systematic Reviews of Interventions, chapter 4: Searching for and selecting studies
- PRISMA 2020 statement and explanation and elaboration (Page et al., BMJ 2021)
Run your next systematic review in Prolytiq
From research question to publishable chart, with every decision logged and exportable for peer review.