Exporting search results for screening, database by database

Every database lets you take your results away as a file, and each does it a little differently. By the end of this guide you will have every record from each database in RIS or PubMed files, abstracts included, and a note of the count each search showed.

That count matters twice. It is the number you report for each database, and it is how you know a file is complete. An export can stop at a limit, or leave the abstracts behind, and neither is easy to see later.

Before you start

• Run your final search in each database and keep it open.
• Start a search log with one row per database, as in Table 1. A blank one to print follows “Check your work”.
• Name each database with its platform. PRISMA-S item 1 asks you to “Name each individual database searched, stating the platform for each”. APA PsycInfo on Ovid and APA PsycInfo on EBSCOhost are two different searches.
• If the platform offers a personal account, sign in to it. Several of the larger exports below need one.

Write down the count before you export

The number a database shows for your final search is the number of records identified from that database. It goes into the flow diagram’s box “Records identified from*: Databases (n = )”, whose footnote suggests one number for each database rather than a total. PRISMA-S item 15 asks you to “Document the total number of records identified from each database and other information sources”, and item 13 asks for “the date when the last search occurred”.

Keep the strategy itself as well. PRISMA-S item 8 asks for the search strategies “copied and pasted exactly as run”. Ovid can put the search history inside the export file. For PubMed and the Cochrane Library, save it separately, as their sections below show.

Choose RIS, with the abstract

Choose RIS wherever it is offered. It is a plain-text format made for moving references between programs, and every platform in this guide exports it except PubMed. PubMed has its own tagged format, the PubMed format, which comes as an .nbib or .txt file.

Screening needs the abstract, and several platforms let you leave it out. Look for the setting as you export:

• Ovid: choose a set of fields that includes the abstract.
• EBSCOhost: tick “Include abstract”, or choose fields that include it.
• Scopus: tick “Abstract and Keywords”.
• Web of Science: choose “Full Record” rather than “Fast 5000”, which carries only author, title and source.
• Embase.com: choose “Citations and Abstracts” or a fuller option.
• Cochrane Library: leave “Include abstract” ticked.
• ProQuest: choose to include the available abstracts.

Then open one file in a text editor and find an abstract in it.

Export in batches under each limit

Each platform caps the number of records in one export. Table 2 gives each cap with the date of the page it comes from.

When a search is larger than the cap, split it and export each part:

• By date. PubMed, the Cochrane Library and ProQuest each suggest splitting a large result by date. Run the same search for, say, up to 2015 and from 2016 on.
• In blocks. Otherwise export the records in blocks no larger than the cap, one after another, and number the files.

Write each part’s count in the log. The parts should add up to the count the database showed.

Export from each database

PubMed

1. Note the number of results and the date.
2. Choose “Save”, then “All results” and the format “PubMed”, then “Create file”. Or choose “Send to”, then “Citation manager” and “All results”, then “Create file”, which gives an .nbib file. Both stop at 10,000.
3. Over 10,000, the PubMed User Guide says to “Divide the result set into smaller chunks using the results timeline or custom date range filter”, then export each part.
4. On the Advanced Search page, “Download” saves your search history as a CSV file. PubMed keeps only “the last 100 searches” and the history “expires after 8 hours of inactivity”, so download it the day you search.

Ovid

1. Note the count on your final line and the date.
2. Select the results and export them as RIS, with a set of fields that includes the abstract. Ovid can include the search history in the export, so ask for it: the strategy then travels in the same file as the records, which is what PRISMA-S item 8 needs.
3. Ovid’s help page gives no maximum per export. Export in blocks of 1,000 records (1–1000, then 1001–2000) and check each file’s count.

EBSCOhost

This covers APA PsycInfo, CINAHL and the other databases on EBSCOhost.

1. Note the count and the date.
2. If EBSCO has enabled it for your institution, the menu at the top of the result list offers “Export results (Up to 25,000)”. You must be signed in to MyEBSCO. Choose the export format for your screening tool, and EBSCO emails you a link to the file. The link works for 7 days.
3. Records from APA PsycTests are not included in these bulk exports. EBSCO also says the number “may be lower depending on your institution’s permissions, duplicates, and database restrictions”, so compare the file with your count.
4. Without the bulk option, select up to 50 records on the result list, or 500 when signed in. “Download” offers a “Range” of 500 and an “Include abstract” box. “Cite”, then “Export citation”, then “Export in RIS format” gives a RIS file.

Scopus

1. Note the count and the date.
2. Select the documents and choose “Export”. Scopus asks you to sign in. Choose RIS, and tick “Abstract and Keywords” as well as “Citation information”.
3. One export holds up to 20,000 items. Over that, split the search by date.

Web of Science

1. Note the count, the date and which indexes you searched. PRISMA-S item 2 asks you to list “all of the databases searched” on a single platform.
2. Choose “Export”, then “RIS (other reference software)”, with “Full Record” as the content.
3. Up to 1,000 records go out at a time, or 500 when cited references are included. “Fast 5000” exports 5,000 at a time “with only Author, Title and Source fields”, so it has no abstracts.

Embase

1. Note the count and the date.
2. On Embase.com, select the results and choose “Export”. Choose RIS and “Citations and Abstracts”.
3. Registered users can export 10,000 records at a time. Anonymous users can export 500, so sign in first.

Cochrane Library

1. Note the count and the date.
2. Choose “Select all”, then “Export” and “RIS (EndNote)”. Leave “Include abstract” ticked, then “Download”.
3. One search exports up to 20,000 records. Above that, the help page says to “limit your results by date”.
4. Exporting CENTRAL (trials) records needs a subscription: “citation export options display only when a user has a subscription to the Cochrane Library”.
5. For your strategy with its counts, use “Print” in Search Manager. The Strategy Library’s “Export” gives “a text version of your saved strategy without counts”.

ProQuest

1. Note the count and the date. When several databases are searched together, ProQuest includes duplicates by default, unless you or your library have changed that setting. Write in your log whether “Exclude duplicate documents” was ticked.
2. If your library has enabled it, “Export Results” appears at the top left of the results when you are signed in to My Research and the search holds 20,000 results or fewer. Choose RIS and include the available abstracts. ProQuest asks you to agree to its terms and conditions, then places the file in “Saved Searches” in My Research.
3. It exports the whole result set only, and you can use it ten times a day. Over 20,000, ProQuest suggests you “break your query into separate date ranges”.
4. Some databases do not support it, among them APA PsycARTICLES, PsycBOOKS, PsycEXTRA and PsycTESTS. Deselect them on the “Select Databases” page and run the search again.
5. Without the feature, add results to “Selected items” and use “Export/Save as file” in smaller batches. ProQuest’s help pages give no number for this route.

Check your work

• Each file holds the count the database showed, or its parts add up to it.
• One record in each file has its abstract.
• Each file name says the database, the platform and the date.
• The log has a row for every database, with its date and the search as run.

To count the records in a file, search it in a text editor. In RIS, each record begins with a TY line; in the PubMed format, each begins with a PMID line.

When a file holds fewer records than the search, the vendors’ pages give these reasons: a cap that cut the export short; records a platform leaves out, such as APA PsycTests records in EBSCOhost’s bulk exports; and, for those bulk exports, duplicates and your institution’s permissions. Export again in smaller parts, and note what you did.

Writing it up

Searches were run in [database (platform)], [database (platform)] and [database (platform)] on [date]. Results were exported with their abstracts as RIS files (PubMed format for PubMed), in parts where a platform’s export limit required it, and the number of records in each file was checked against the number of results each database showed. The full strategies, as run, and the number of records from each database are in Supplement [n].

Reading on

• The PRISMA 2020 flow diagram: counting every box
• Removing duplicates, with a record of each one
• PRISMA-S: reporting literature searches in systematic reviews