Systematic reviews - for researchers

In many databases, articles are indexed with subject headings or controlled vocabulary search terms meant to make it easier to find articles about a certain subject. PubMed, for instance, uses the controlled vocabulary list MeSH (Medical Subject Headings).

You can find controlled search terms in many ways. One way is to search directly in a given database’s thesaurus. You can also find out which search terms were used to index your key articles. 

Controlled search terms are usually hierarchically organized, with a superior – broader – term, and a subordinate – narrower – term. In the thesaurus you may also find referrals to related terms. One tip is to use both broader, narrower and related terms, in order to ensure that you’re getting all the relevant subject headings into your search. 

The default setting in most databases is the automatic explosion of subject headings, that is, the search will include subordinate terms. This is often useful, but sometimes it can be more helpful to turn this setting off and search for a term without explosion.

To find new articles that have not yet been indexed, or articles that have been indexed with other terms than those you’re using yourself, you should also search in the title, abstract and author’s keywords. This is usually referred to as a free-text search. It is recommended that you use the same search terms, both as controlled terms and as free-text terms. 

As free-text terms, you should include all the relevant synonyms and spelling variations that you can find in the key articles’ titles and abstracts. 

The databases’ thesauruses (the lists of the controlled terms) can be a helpful tool to use to find free-text terms, since these often list synonymous concepts. In PubMed, such synonyms are called Entry terms. Narrower controlled terms can also be included as free- text terms. 

Truncation

Use truncation to find different variants of a word. Therap* retrieves, for example:
•    therapy
•    therapies
•    therapeutic

Spelling variations in British and American English

Quite a few medical terms are spelled differently in British and American English. Consider whether or not your search terms include a word with a different spelling. If so, include both spellings. A few examples of different spellings in British vis-à-vis American English:
•    Tumour / Tumor
•    Gynaecology / Gynecology
•    Coeliac / Celiac
•    Ageing / Aging
•    Behaviour / Behavior

Searching for phrases 

Quotation marks are useful for keeping words in phrases together, but be careful when doing so during a systematic search. When you search for literature using quotation marks, the search becomes more precise; you will only find articles where that exact phrase is used and miss articles that contain variations on that phrase or a similar combination of search terms.  

Proximity operators

An alternative might be using proximity operators. These look a little different depending on which database you’re using:

NEAR in Web of Science and adj in Medline Ovid. 

Adding a number after NEAR/adj will determine how many words you want to “allow” between your two terms. This allows you to retrieve any variations on a phrase, for instance phrases that contain a different word order. 

For instance, if you did a search for “oxygen treatment” as a phrase in citation marks, you might miss any references containing phrases such as:

• "oxygen (HBO) treatment"
• "treatment with oxygen"
• "oxygen in the treatment"

If you instead use a proximity operator (oxygen NEAR/3 treatment), your search will retrieve the variations.  

One way of testing your search strategy is to see if it succeeds in retrieving your key articles. Test this by doing the following: 

• Conduct your search (A).
• Do a new search (B), this time for your key articles only. In PubMed, you can do so using PMID.
• Which key articles were not retrieved? Search B NOT A. Here you will want the result to be zero. If that’s your result, then you’ve retrieved all your key articles. 

If a key article was not retrieved, analyze why. Can a search term be added that might include this key article? Should you remove an entire search block? Keep in mind that it may not always be possible to alter your strategy to such an extent that all your key articles are found without your results becoming too comprehensive. 

There are many different tools and databases available that can help you retrieve citation data. Here we’ll review how you can do so in Web of Science and in SpiderCite, which builds on data from Lens.org.

Web of Science

Start by conducting a search in Web of Science where you retrieve all the articles you want to use (ex., the articles you intend to include in your review). For instance, you can search for the articles using titles, but you can also do so using PMID-nr or DOI-nr. Select “OR” between each row and add more rows if needed.

You will receive a list of hits that only contain the publications you want to use as your starting point (in the example, four publications). To see which publications that have cited these articles, click on Citation Report, which you’ll find on the right side of the search box:

In this example, 73 publications have cited the four publications we used as our starting point. Click on the number under the heading Citing Articles, you’ll receive a list of these 73 publications. You can then export the references from the list to EndNote or another reference management program.

To the right of every hit in the list in Web of Science you can also see how many references each publication has. Unfortunately, it is not possible to easily retrieve a list containing all the references from several different publications.

SpiderCite

SpiderCite is a tool where you can see both the citations and references from several publications. SpiderCite builds on citation data from Lens.org. To use SpiderCite you need to save the publications you want to use as your starting point in an XML, RIS or BibText-file. If you want to include your references in some other way, for instance by using PMID or the DOI-nr, we recommend CitationChaser.

Start by going to https://sr-accelerator.com/#/spidercite and upload the files with the publication list you want to use. To export references in an XML-file from EndNote, do the following: in EndNote, mark (CTRL + A) those references you want to export, then click File > Export. In the roll-down menu Save as type, choose XML (*.xml). What kind of Output style you choose doesn’t matter.

When you’ve uploaded your file, you will see two different tabs to the left of the page. Input contains the publications you’ve uploaded. Output contains the citations and references from the publications you’ve uploaded.

When you click on Input you will see two different alternatives, "Loaded with Lens.org" and "Citations to be searched manually". The publications that end up under "Loaded with Lens.org" have been retrieved and searched for citations and references. The publications that have eventually ended up under "Citations to be searched manually" haven’t been found and will need to be retrieved manually, for instance in Web of Science, using the method detailed above.

When you click on Output, you should see both the reference list and the the citations from the publications you’ve uploaded.

To export references and citations, go to File, then Export.

You are probably familiar with PubMed, and it is an easy database to get started with. However, there are a few things to consider: 

• When you search in "All fields" PubMed maps your search terms by adding MeSH terms and alternate endings. If you want to see how this automatic mapping works you can do a search without choosing a specific field, and then go to History and Search Details (on the Advanced page). Click on the arrow in the column Search Details. If you use truncation, the automatic mapping won’t work. 
• In systematic searches, it is recommended that you not trust automatic mapping and instead specify which fields you want to search. Usually you will search for MeSH terms in the field "MeSH Terms", and free-text terms in the field "Title/Abstract". 
• MeSH terms will be exploded by default. To search for a MeSH term without exploding it, write [Mesh:NoExp] after the search term. 
• Not all phrases are searchable in PubMed, only the phrases included in the phrase index.
• If you’re searching in the MeSH-terms-field and Title/Abstract-field, the words automatically will be searched as phrases.
• You can use truncation, but you can only truncate the last word in a phrase.
• You cannot use proximity operators, in contrast to almost all other databases.

This is how a search using two different search blocks might look in PubMed. You can find PubMed Advanced Search Builder by clicking on "Advanced" in PubMed’s homepage. 

In MEDLINE Ovid, the indexed part of PubMed is included, but also newly added references (Epub Ahead of Print and In-Process) and other non-indexed material. Ovid is a database platform used for many different databases. 

• There are several advantages of searching MEDLINE via Ovid instead of PubMed. In Ovid it is possible to use proximity operators. It is also easier to structure comprehensive and advanced search strategies, and to edit search strategies.
• To search in the MeSH field, type a forward slash symbol after the MeSH term. If you write "exp" before the MeSH term, it means that it is exploded, i.e. narrower terms are also included in the search.
• The free-text terms are usually searched for in the Title, Abstract and Author Keywords fields. Put the free-text within a parentheses and write ".ti,ab,kf." immediately after the parenthesis.
• The search terms are searched as phrases when you search in Title, Abstract and Author Keywords search fields.
• In Ovid, the proximity operator adj is used. Adding a number after adj will determine how many words you want to “allow” between your two terms. 
• To edit a row, click the arrow next to "More", and then "Edit".

This is how a search using two different search blocks might look in MEDLINE Ovid. The search terms within each block are combined with OR, then the blocks are combined with AND:

Embase is a database that contains references in biomedicine, with focus on pharmacology and toxicology. There are several different options for searching in Embase (via embase.com). In the search example below, we show how you can structure the search in a similar way as in Medline Ovid.

• Embase has a controlled vocabulary, Emtree. When you translate a search strategy from PubMed/Medline, you need to look up your MeSH-terms in the list of Emtree-terms to see what the corresponding controlled term is in Embase.
• To search the Emtree field, type single quotation marks around the term, and a forward slash symbol immediately after. If you want the term to be exploded, type "exp" after the slash. If you don't want the term to be exploded, write "de" after the slash.
• The free text search terms are usually searched for in the Title, Abstract and Author Keywords fields. Put the free text within a parentheses and write ":ti,ab,kw" directly after the parenthesis.
• You need to put single quotation marks around phrases. If you search without quotation marks, the terms are not kept together as a phrase.
• In Embase, the proximity operator NEAR/ is used. Adding a number after NEAR/ will determine how many words you want to “allow” between your two terms. 

This is how a search using two different search blocks might look in Embase. The search terms within each block are combined with OR, then the blocks are combined with AND:

Web of Science Core Collection is a multi-disciplinary database that includes research areas other than biomedicine and health.

• There is no controlled vocabulary in Web of Science and only free-text searches are possible. For systematic searches, it is recommended to search in "Topic", which means that you search the Title, Abstract, Author's Keywords, and Keywords Plus.
• Use quotation marks around phrases for more precision. If you search without quotation marks, the terms are not kept together as a phrase.
• Truncation can be done with an asterisk*.
• In Web of Science the proximity operator NEAR/ is used. Adding a number after NEAR/ will determine how many words you want to “allow” between your two terms. 
• In Basic Search, you can enter a search block on each line, with OR between the words. The different rows are combined with AND.
• If you have a long search string with several different lines within each block, it may be easier to build up the search in Advanced

This is how a search with two different search blocks can be done in Web of Science, in Basic Search:

This is how a search with three different search blocks, where each block contain several different lines, can look like in Advanced Search. The different rows with search terms within each block are combined with OR, then the blocks are combined with AND. TS is the same as TOPIC.

CINAHL is a database that contains references to articles in nursing, physiotherapy and occupational therapy.

• CINAHL has a controlled vocabulary, CINAHL Subject Headings. When you translate a search strategy from PubMed/Medline, you need to look up your MeSH terms in the CINAHL Subject Headings-list to see what the corresponding term is called there.
• To search in the Subject Headings field, type MH before the term. If you write a + after, it means that the term is exploded, i.e. underlying terms are also included in the search.
• The free text search terms are usually searched for in the TI (Title) and AB (Abstract) fields. Put the free text words within a parenthesis and write TI in front of the parenthesis. Rewrite the words within a parenthesis and write AB in front of the parenthesis. Combine the two rows with OR.
• You need to put quotation marks around phrases. If you search without quotation marks, the terms are not kept together as a phrase.
• In CINAHL the proximity operator N is used. Adding a number after N will determine how many words you want to “allow” between your two terms. 

This is how a search in CINAHL can look like. The search terms within each block are combined with OR, then the blocks are combined with AND:

Wichor Bramer, information specialist at Erasmus MC, has developed an method for removing dublicates from different databases in Endnote. The method exists in an unpublished and relative simple version presented here. However, Wichor Bramer and co-authors have also later published a modified and more advanced version (Bramer et al., 2016).

Step 1

Go to Duplicates and Preferences in Endnote. Select fields for each comparison below and click on Find Duplicates under References and remove all duplicates. Repeat the procedure for each comparison.

1. Author; Year; Title; Secondary title
2. Author; Year; Title; Pages
3. Author; Year; Secondary title; Pages
4. Title; Secondary title; Pages
5. Author; Volume; Issue; Pages
6. Title; Volume; Issue; Pages

Step 2

Follow the procedure in step 1, but also check the fields in bold.

1. Title; Pages – compare Author
2. Author; Pages – compare Title
3. Author; Title – compare Journal Title
4. Title; Secondary title – compare Author

Step 3

Follow the procedure in step 2, but de-select all duplicates and select the real duplicates instead.

1. Author; Secondary title – compare Pages
2. Volume; Issue; Pages – compare Author (sorted by page number)
3. Title – compare Abstract
4. Author

You're done!