EPIB 629 Knowledge Synthesis

Keywords

Keyword searching (also called textword, natural language, or free-text) is when you search for words you expect to find in the title, abstract, or author-assigned keywords. Remember, most article databases do NOT search the full text. Create a list of keywords or phrases based on the main concepts in your research question. You can organize this using the search strategy worksheet above, or any layout that works for you. When using this technique, be aware of synonyms and spelling variations.

While subject headings are usually the foundation of search strategies because they can capture a vast amount of information, there are various reasons why you should add keywords to a search:

• For an emerging research topic that may not yet have a subject heading to describe it.

• To retrieve the most recent articles (which have not yet been indexed with subject headings).

• Occasionally, articles can be incorrectly indexed (e.g., not assigned a relevant subject heading) and may be missed by a search using only subject headings.

• To develop a thorough search strategy for a systematic review or other knowledge synthesis, we always supplement the search with keywords.

***You should keep in mind that keywords do not generally account for:

• Spelling variations
• Synonyms
• Plural forms

        ⇒ Truncation symbols and operators can help with this.

        ⇒ However some databases do include lemmatization.

Example of Using Subject Headings and Keywords

Using a concept map we will demonstrate how to use subject headings and keywords with the following research question:

⇒ What specific occupational risks are associated with an increased cancer risk in firefighters?

***Subject headings were chosen according to MeSH terms in MEDLINE

► in Ovid MEDLINE, occupational risks mapped to several subject headings, and based on the scope note these are the most relevant. See the scope note for Occupational Exposure below:

Proximity Operators

Many bibliographic databases also support the use of proximity operators, also known as adjacency operators. These operators allow you to refine your search by specifying the proximity of your search terms, identifying the number of words near or between them. 

• The availability of proximity operators depends on the database and platform you are using. For example, SCOPUS uses W/n, EBSCO uses Nn, and OVID uses adjn (where "n" represents the maximum number of words).
• For more details, refer to the Database Specific Operators and Fields section.

For example:

• on the Ovid platform, a search for primary adj3 care could retrieve records containing, primary health care, or  primary and secondary health care.

Parentheses/Brackets

Parentheses (or brackets) are used to set the order of the execution of Boolean logic in a search. When combining different search terms with Boolean operators, the database doesn't automatically recognize synonyms. By using parentheses to group similar terms, you can ensure they are treated as a single unit in your search.

• Parentheses work in most but not all systems (e.g., they work on the Ovid, PubMed, EBSCOhost, ProQuest, Scopus, and Web of Science platforms, but they are ignored in Google and Google Scholar)

For example: 

(chest OR thorax OR thoracic) AND (imaging OR radiographs OR radiography) 

Searching with vs without parentheses 

To demonstrate the impact of searching using parentheses see the image below: 

• S1 - S3: These search lines demonstrate line searching, where each concept is searched separately. In S1, "Chest" and its synonyms are searched using the OR operator. In S2, "Imaging" and its synonyms are searched using the OR operator. These two searches are then combined in S3 using the AND operator, resulting in 35,378 results.
• S4: This search line includes both "Chest" and "Imaging" concepts without using brackets, but combined using the same Boolean operators, resulting in 232,120 results. This search has a surplus of results compared to the S1 - S3 search, indicating an increase in unintended results.
• S5: This search line includes both "Chest" and "Imaging" concepts with brackets and combined using the same Boolean operators, resulting in 35,378 results, replicating the S1 - S3 search and targeting the intended articles.

Truncation and Wildcards for Keywords 

Truncation symbols are shortcut characters which can help to include variations of your text word without having to type each variation into the search separately. This feature is also known as stemming. 

The asterisk, *, is the most common truncation symbol. Other truncation symbols and wildcards include, #, ?, $, !. Depending on the platform you are using to search a given database (e.g., EBSCOhost, Ovid, ProQuest), other truncation and wildcard symbols may be available to use. Please see the database-specific operators and fields for more information.

For example: 

Searching for computer*, searches for computer, computers, computerised, computerized, etc.

Advanced Searching Tips

Advanced search techniques can include quotation marks (" ") to force phrase searching or to turn off lemmatization (this is database-dependent) and specifying search fields (such as title, subject, etc.).

Example of using quotation marks:

Imagine you're researching the phrase adverse drug reaction:

• Without quotation marks: Searching adverse drug reaction will find articles that include the words adverse, drug, or reaction—but these words can appear anywhere in the text and in any order.

• With quotation marks: Searching "adverse drug reaction" will find articles where these three words appear together in exactly this order.

Example of using search fields:

When you're researching the phrase adverse drug reaction, you can use a search field to focus your results

• Searching without a field: Searching "adverse drug reaction" will give you results from anywhere in the article—titles, abstracts, body text, etc.

• Searching with a field: Searching TI: "adverse drug reaction" tells the database to find articles where the phrase adverse drug reaction appears specifically in the title.

***Each database is different, so be sure to consult our database-specific tables to check for exact field codes and additional functions.

Recommended Search Structures

When searching in databases, there are several ways to structure and organize your searches to make them more efficient and concise. The following simplified examples of searching in PubMed illustrate different options you could use if you were looking for literature on the use of ultrasound to diagnose pulmonary tuberculosis.

Please note: Depending on the database you're using, you may see different ways to combine search line numbers (for example, #1 OR #2 in PubMed, or S1 OR S2 in CINAHL).

"Line by line" -- Easier to see if you have made spelling mistakes when you look at the number of records generated by each line; not all databases allow you to enter searches this way as it requires access to a search history (Example screenshot from PubMed):

"Block by block" -- One key concept or set of terms per line: More succinct but you must check even more carefully for spelling mistakes; once again, not possible in all databases (requires access to a search history) (Example screenshot from PubMed):

"Single line" -- Many databases will allow for this type of entry, but you must check that parentheses (and intended application of Boolean operators, in some cases) are supported (Example screenshot from PubMed):

• Complex searches will often mix these approaches, and that's okay.
• It is important to note that this is an oversimplified example to illustrate the basic principles of structuring a search. Normally, a fully developed search strategy should include subject headings and text words (with proximity operators, truncation, field codes, etc. as appropriate).