How to screen titles and abstracts

Title and abstract screening asks one question of each record: could the full text meet your criteria? Most records leave a review at this stage, so the way you decide here shapes everything after it.

This guide covers the pilot, the three decisions, reasons, records with no abstract, notes, pace, and screening in a spreadsheet when you have no screening tool. Four one-page sheets go with it.

Before you start

• Your eligibility criteria are written down, numbered in order of importance, with a version number and a date.
• Your records are in one list, with duplicates removed.
• You have filled in the criteria card that goes with this guide, and put it beside your keyboard.

Ask one question of each record

At this stage you are not deciding whether a study belongs in your review. You are deciding whether its full text might. The Cochrane Handbook advises being over-inclusive at this stage (section 4.6.3). A record you cannot rule out goes forward, and the full text settles it.

On the criteria card, list each criterion with its number. Beside each, write the words that usually settle a record: population terms, the phenomenon or intervention, and design words such as “interview” or “trial”. Your eye can then look for them first.

Pilot your criteria

Before screening in earnest, screen a small set with the people who will screen, or with your supervisor if you are working alone, and compare. The sizes differ by source:

• Cochrane Handbook: pilot the eligibility criteria on about six to eight reports, including some you think clearly eligible, some clearly not, and some doubtful (section 4.6.4).
• JBI, for scoping reviews: a random 25 titles and abstracts, screened independently by the whole team, then discussed, with the criteria refined as needed. Screening starts once agreement is 75% or more (section 10.2.7.1).
• Cochrane Rapid Reviews Methods Group: pilot the screening process on “e.g., 20–50 citations” (Garritty et al. 2026).

Talk through every difference, not only the total. When a difference comes from a criterion’s wording, change the wording, give the criteria a new version number and date, and note what changed. The pilot sheet has a line for each record, both decisions and what you agreed.

Decide Include, Maybe or Exclude

Three decisions cover every record.

• Include when the title and abstract suggest the study meets every criterion you can judge from them.
• Exclude when the record clearly does not meet at least one criterion. Not meeting one criterion is enough to exclude a study (Cochrane Handbook, section 4.6.4).
• Maybe when you cannot tell. Under the over-inclusive rule, a Maybe goes forward to full text.

Maybe is a decision in its own right. It records that the title and abstract do not settle the question. A short note saying why helps at full text and makes sense to a colleague: which criterion is unclear, and what would settle it. The uncertain-note sheet gives the three lines.

Keep reasons for full text

At title and abstract, numbers are enough. Cochrane’s standard for documenting selection says counts of records are sufficient for exclusions at this stage; reasons are needed for studies excluded at full text (MECIR C41).

If you keep a reason anyway, use the number of the first criterion the record does not meet, taking the criteria in order of importance. That is how the Handbook suggests choosing the primary reason for exclusion (section 4.6.4).

Decide who screens

In Cochrane intervention reviews, two people screening titles and abstracts independently is “desirable, but not mandatory”; two working independently are mandatory for the final decision on each full text (MECIR C39). JBI asks for two independent reviewers at both stages of a qualitative review (section 3.3.10). A second screener need not know the topic: the Cochrane Handbook notes that one who is not a content expert can be an advantage (section 4.6.4).

The evidence on screening alone is small. A methodological review of 4 evaluations (23 single screenings) found a median of 5% of relevant studies missed (range 0–58%); 3% for 6 experienced reviewers and 13% for 3 less experienced ones (Waffenschmidt et al. 2019). For 7 of those screenings the effect on meta-analyses was reported, and the 3 that would have changed the findings substantially were all by the same less experienced reviewer. The authors conclude that single screening “is not equivalent to double screening, as substantially more studies are missed”. They call the evidence on the impact of missed studies insufficient.

Screen records with no abstract

Some records have only a title. Look for the abstract first: the publisher’s page or another database may have it. If there is none and the title leaves the question open, choose Maybe and let the full text decide. If the title alone rules the record out, exclude it as you would any other.

Write notes someone else can read

Write each note for a reader who has not seen the record: your colleague, your supervisor, or you in three months.

• Name the criterion by its number.
• Say what the abstract says, or leaves out.
• Say what would settle it.

Keep notes about the paper, in words you would be content for an examiner to read.

Time your first 100 records

This guide gives no figure for how long screening takes. It depends on your topic, your criteria and you, so time your own.

1. Screen 100 records and note the minutes they took.
2. Minutes per record = minutes ÷ 100.
3. Hours left = records left × minutes per record ÷ 60.

For example (illustrative numbers), 100 records in 50 minutes is 0.5 minutes a record. With 1,800 records left, that is 1,800 × 0.5 ÷ 60 = 15 hours. Time another 100 later on and update the estimate. The pace log has a line for each session.

Stop for the day

Stop while you are still reading each abstract to the end. Before you stop, note where you are, the time, and anything to raise at your next meeting. If you screen in a spreadsheet, save a copy with the date in its name.

Screen in a spreadsheet when you have no tool

Put one record on each row of a sheet named Records, with a header in row 1.

Make column F a drop-down list of Include, Maybe and Exclude, so every decision is spelt the same way. In Excel, choose “Data” › “Data Validation”, set “Allow” to “List”, and type Include,Maybe,Exclude in “Source”. In Google Sheets, choose “Insert” › “Dropdown” and enter the three options. Then keep your counts on a second sheet:

If two of you screen, each works in their own copy of the file. The guide “Screening alone or with a colleague” has the formulas that compare the two copies.

If your records are in EndNote, Bramer, Milic and Mast (2017) describe a way to screen there: a group set for includes and one for excludes, each screener dragging records into them in their own copy of the library, with a display style that shows the abstract.

Check your work

• Include + Maybe + Exclude + Undecided equals the number of records in your list.
• Every Maybe has a note saying what is uncertain.
• Every sheet and file names the version of the criteria it was screened under.
• Each change to the criteria has a new version number, a date and a line saying what changed.

Writing it up

PRISMA 2020 item 8 asks how many reviewers screened each record, whether they worked independently, and what automation tools were used, if any (Page et al. 2021). A paragraph to adapt:

Titles and abstracts were screened against the eligibility criteria (version [n], [date]) by [one reviewer / two reviewers working independently, who resolved disagreements by discussion]. The criteria were piloted on [n] records by [initials] and [revised once / not changed] before screening began. Records that could not be excluded on their title and abstract were carried forward to full-text assessment. No automation tools were used to exclude records.

Reading on

• Agreement between screeners: percent, Cohen’s kappa and PABAK
• Screening alone or with a colleague
• Cochrane Handbook, chapter 4: Searching for and selecting studies
• Waffenschmidt et al. (2019), single versus double screening