The skills you learn when designing a basic search strategy can be applied to virtually all academic databases and are helpful when searching other places too. However, if you plan on crafting a fully comprehensive search or want to make your results even more precise there are some advanced search techniques you can use. This is especially helpful if you want to publish in a peer-reviewed journal and want to be assured you have found ALL relevant results.
(elderly[tiab] OR senior[tiab] OR aged[tiab] OR "Aged"[Mesh]) AND (chemotherapy[tiab] OR "Chemotherapy, Adjuvant/adverse effects"[Mesh] OR "Consolidation Chemotherapy"[Mesh] OR "Induction Chemotherapy"[Mesh]) AND (diarrhea[tiab] OR diarrhoea[tiab] OR "Diarrhea"[Majr])
You can see that this is at least 10,000 fewer results than the initial search on the 'basic search' page.
elderly OR senior OR aged OR 'aged'/exp/mj AND ((chemotherapy OR 'chemotherapy') NEXT/1 (related OR associated OR induced) NEXT/2 (diarrh*ea OR 'diarrhea')):ti,ab
Note that we went from 15000 results to 28 using the advanced search skills and features.
Proximity searching is searching for keywords within X number of words from each other.
In Pubmed, you can only do proximity searching within one pair of words. For example, "surgical repair"[Title/Abstract:~3] gets the words surgical AND repair within 3 words of each other which would find phrases like surgical hernia repair.
In Embase,
NEAR/n requests terms that are within 'n' words of each other in either direction.
NEXT/n requests terms that are within 'n' words of each other in the order specified.
For example: (chemotherapy) NEAR/2 (related OR associated) NEXT/2 (diarrh*)
In basic search design we discussed using truncation at the end of a word to find variations. In Embase you can use truncation (also refererred to as "wildcards") to find alternative spellings within words. You can also use them while phrase searching in Embase - something you cannot do in PubMed.
This can be extremely helpful when searching for words that have slight spelling variations between countries or pharmaceutical names.
For one or more letters use *
For a single letter use ?
Example:
"heart attack*" finds heart attack, heart attacks
sul*r finds sulfur and sulphur
f*cal finds fecal and faecal
sulf?nyl finds 'sulfonyl' and 'sulfinyl'
Every database uses unique field syntax to tell the search algorithm WHERE to search in an article's record. These can be powerful, when used correctly, and help narrow and make your search more precise. They often include a container element like square brackets or colons and abbreviations.
PubMed puts search field tags in square brackets [ ].
The major syntax we learned above was for Medical Subject headings, [MeSH]. Here are a few more:
[Majr] - use this if you want your MeSH subject heading to be the major focus of the article. Either click the check box in the MeSH term record that says "restrict to major topic" or replace the [Mesh] with [Majr] at the end of your subject heading.
[tiab] - put this after a key term, either singular or a phrase in quotations, to only search for it in the title and abstract fields. This prevents PubMed from "automatic term mapping" which can artificially broaden your results and introduce irrelevant results.
Embase has similar syntax to PubMed. Instead of square brackets it uses a colon or / and an abbreviation.
Some commonly used field codes are:
mj for major term /mj
ab to search only in the abstract :ab
ti to search only in the title field :ti
Controlled vocabulary terms are a way to ensure your searches are comprehensive by not only relying on authors to use the exact same free text keywords that you have in your search. These are terms in databases that are applied to the article and describe the "aboutness" of the article.
Sometimes subject headings and keywords are the same, but often they aren't. The MeSH heading in MEDLINE for support groups is Self-Help Groups. Using only the keyword "support groups" will leave out some articles tagged with the subject heading, but using only the subject heading will leave out some articles that use the keyword. For a comprehensive search, always use both subject headings and keywords.
From the University of Toronto Gerstein Health Information Centre
Tips on using controlled vocabulary terms:
In PubMed each article is manually labeled with a set of subject headings called Medical Subject Headings or MeSH. If you use a Medical Subject Heading (MeSH) for each concept of your search and each MeSH is assigned to a given article that article should be extremely relevant to your topic. MeSH help narrow and define your searching further.
Like PubMed, Embase has its own subject headings called Emtree. You'll know a term is an Emtree term by the /exp that is on the end of the word. They are not always the exact same wording as a MeSH term, although there is broad overlap.
For example, in Emtree "support group" is it's own controlled vocabulary term 'support group'/exp. In PubMed the MeSH term for "support group" is "Self-Help Groups"[Mesh].
CINAHL uses "CINAHL Headings" as their controlled vocabulary. You find them in the row of tabs below the main search bar.
PsycInfo uses thesaurus terms as well, these can be very helpful when you want to limit to a specific kind of therapeutic technique or mental disorder. They are often much more granular in these areas than other biomedical databases because of the scope of the included articles.
Subheadings are a useful but often misused component to advanced searching in PubMed.
When you choose a MeSH term there is an option to select a subheading to go along with it. This further narrows the potential results because the results will only include that specific subset of articles.
For example, if your research questions was about (strictly) the surgical treatment or outcomes for heart attacks you would search for:
"Myocardial Infarction/surgery"[Mesh]
The word after the / is the subheading. A list of all the potential subheadings can be found below but will not be available for all MeSH because of indexing choices.
Instead of using the limits in PubMed to select only humans instead you will want to add in a "hedge" or search string. If you select the "humans" filter this will remove studies that include humans AND animals. These could still be relevant to your topic.
Use this instead: NOT (("Animals"[Mesh] NOT ("Animals"[Mesh] AND "Humans"[Mesh]))
The hedge/filter for humans in PubMed comes from the Cochrane Handbook for Systematic Reviews of Interventions http://www.cochrane-handbook.org/ Chapter 2, section 6.4.11 in Box 6.4.a.
In Embase use this: NOT ((exp animal/ or nonhuman/) NOT exp human/)
we recommend building and refining your search in one database first. Then when you are satisfied with the results translate it to the other databases you are wanting to search. Consistency is key here, the searches won't be exactly the same because controlled vocabularies differ and sometimes you need field syntax to enhance precision but they should be comparable.
Translating searches between databases can be complicated because each databases uses different syntax and field codes, as discussed above. There are some tools that can be used to automate this process but you must check the query before running it. We find that the places that can be the most inaccurate during translation are database field syntax and controlled vocabulary terms. Always check to see if the term generated by the tool actually exists in the database.
Below are some tools that can help.
If you find an article that is exactly what you are looking for and you want to find more like it one way to do this is to use citation chaining. This is an additional way to augment searching databases and ensure higher recall of relevant research.
Backward citation chaining: look at the article's bibliography and see what they cited, odds are many are quite relevant to the topic you are researching too.
Forward citation chaining: sometimes called 'snowballing', you can use search engines to find who cited the article you like. This is a great way to find new research including conference proceedings and reports your search may have missed.
There are three databases that do this: Scopus, Web of Science, and GoogleScholar.
These databases will generate a list of results of articles that have cited your chosen article and which, in theory, should be highly relevant to the topic you are currently searching. You may want to create a free profile on Scopus and Web of Science to export, you can make a list of citations and add the ones you want to export as you look up each article. This will save you from multiple downloads.
