Evidence Synthesis

Keeping track of your search is an essential part of the evidence synthesis process. You must report your search strategy in the methods section of your manuscript and include the full final search in each database (typically as an appendix) to make your search reproducible. The goal of evidence synthesis searching is to, insofar as possible, capture every single piece of available evidence on a topic. In order to do this, you must write a detailed and comprehensive search string and translate it for multiple databases. 

Search steps:

1. Brainstorm keywords 
2. Choose your databases + journals or grey literature if using
3. Search the database thesaurus / subject headings to find controlled vocabulary 
4. Construct search using keywords AND controlled vocabulary 
5. Revise search as needed to retrieve more results 
6. Translate search for all databases 

Use the Topic & Terms tab of the Search Log Template spreadsheet.

Keywords are nouns and short noun phrases from your research topic and question. Brainstorm some initial terms and discover more as you search by reading the title, abstract, author-assigned keywords, and subject terms of relevant articles. 

Example: 

Topic: statins vs. red yeast rice for lowering cholesterol 

Question: How effective is red yeast rice in lowering LDL cholesterol compared to statins?

Keywords: "LDL cholesterol", "high cholesterol", "bad cholesterol", "red yeast rice", statins, atorvastatin, lipitor 

Keep track of the databases you're using on the Search Translation Starter tab of the Search Log Template spreadsheet. 

As a general rule, an evidence synthesis project should search at least three databases: two subject databases and one general database to pull in anything the subject databases missed. You can search more databases, and you may also want to include individual journals or grey literature sources in your search. 

If you don't know where to get started with database searching, you can:

• Use the Evidence Synthesis Services link on the homepage of this guide to request an appointment with a librarian
• Search for reviews on similar topics and check the methods section for the databases used 

In addition to databases, you may want to search individual journals and grey literature. You may identify a relevant journal that is not indexed in the databases you've chosen to search. Grey literature includes anything that is not traditionally published, e.g., anything outside of journal articles and books. This includes article preprints, clinical trials, conference proceedings, theses and dissertations, laws, government documents, and more. 

Many databases already include grey literature. For example: PubMed has preprints, Web of Science has conference proceedings, and Cochrane links to clinical trials. FSU also has a ProQuest Dissertations & Theses database. So, you may not need to run a lot of additional searches for grey literature, but it's a good thing to keep in mind. 

Where NOT to search:

• FSU Libraries OneSearch (the big search bar on the FSU Libraries website)
• Google Scholar 

Searching these will give you a lot of results at once, but your search will not be reproducible, and neither has the search tools available in the individual databases. 

Use the Search Explore tab in the Search Log Template spreadsheet to keep track of your search. 

It's important to include both controlled vocabulary from each database AND keywords that you come up with in your search. Combining them will yield more results that searching only with one or the other, and the goal of evidence synthesis searching is to discover all available evidence on a topic. 

To get started, pick one database. Look up the controlled vocabulary in the database's thesaurus and combine them with your related keywords by building search filters. A search filter is a string of search terms related to one of the concepts you're searching. Typically, the terms within a search filter are connected with OR. Multiple filters are connected to each other with AND. 

Here are the keywords I brainstormed from the example topic:

Keywords: "LDL cholesterol", "high cholesterol", "bad cholesterol", "red yeast rice", statins, atorvastatin, lipitor 

I'm looking for (high cholesterol) AND (statins) AND (red yeast rice). Those are my three filters for this search. 

Let's add more keyword terms to each filter. 

("LDL cholesterol" OR "high cholesterol" OR "bad cholesterol") AND (statins OR atorvastatin OR lipitor) AND ("red yeast rice")

I can use the above keyword terms in all of my databases. However, I'd like to start with PubMed. So, I'm going to go to the MeSH search in PubMed (refer to PubMed - MeSH video in the last box) and find some Medical Subject Headings. 

("Cholesterol, LDL"[Mesh] OR "Hypercholesterolemia"[Mesh] OR "LDL cholesterol" OR "high cholesterol" OR "bad cholesterol" OR "elevated cholesterol" OR "hypercholesterolemia") AND ("Hydroxymethylglutaryl-CoA Reductase Inhibitors"[Mesh] OR "hydroxymethylglutaryl CoA reductase inhibitors" OR "HMG CoA reductase inhibitor" OR statins OR atorvastatin OR lipitor) AND ("red yeast rice" OR cholestin)

Now I have a beginning search for my topic that I can use to search PubMed. Note that there are a different number of MeSH terms for each filter. High cholesterol has two, statins has one, and red yeast rice doesn't have a MeSH term associated, so the terms in that filter are all keywords. 

Tip: try searching PubMed by copying the first search string that only contains keywords and then the second one that contains both keywords and MeSH terms. How many more results do you get?

Use the Search Explore tab in the Search Log Template spreadsheet to keep track of your search. 

Remember that the goal of evidence synthesis searching is to discover all available evidence on a topic. You will add more controlled vocabulary and keyword terms as you search and may end up removing or changing some as well. Give yourself time to think about your search, and pay attention to new terms coming up in relevant articles you discover as you start searching. 

Search Fields

When you enter a search into a database, which fields does that database search automatically? What fields are available to search? Can you search multiple fields at once? How do you do this without excluding relevant results?

PubMed defaults to searching [all fields]. You can also specify a narrower search in the title/abstract field [tiab] or the text word field [tw]. As a general rule, you probably want to start out with a narrower field like [tw] when searching for an evidence synthesis project and only expanding to all fields if your topic does not have a lot of literature and is yielding very few results. 

Consider how many results you're getting, if they seem relevant, and which search fields would be best for your search in each database.

Truncation, Wildcards, Phrase Searching, and Proximity Searching

Truncation allows you to use one term to search for multiple related words with the same ending: nurs* searches for nurse, nurses, nursing 

Wildcards allow you to search for terms with variations in the middle of the word and can use the symbols * or ? depending on the database.

"Pregnant wom*n"[tiab] in PubMed yields results for "pregnant woman" or "pregnant women" in the title or abstract; "general an*esthesia"[tiab] searches for the British or American spelling of anesthesia / anaesthesia.

Phrase searching uses "quotation marks" to search for two or more words in a particular order.

"High LDL cholesterol"[tiab] searches for these words in this order. Removing the quotation marks may yield results that include all of those words separately somewhere in the title or abstract.

Proximity searching can be used when terms appear near each other, but maybe not right next to each other as in phrase searching.

"Radical prostatectomy"[tiab:~1] searches for the phrase "radical prostatectomy" with 0-1 words in between, and so also yields results for "radical cytoreductive prostatectomy", "radical retropubic prostatectomy", etc.

Important note! Some databases look for variants such as plurals and alternative spellings automatically. Some do not allow for truncation, wildcard, and/or proximity searching or only use them with certain fields (PubMed allows for proximity searching with [tiab] but not with [tw]). Each database should have its own help or documentation page describing what terms it allows, so look for these, especially if there are issues with your search!

Database Limiters

Many databases provide tools to limit your search results on the advanced search page and/or on a sidebar after you search. These may include date, language, publication type, population, and more. It is generally not recommended to use limiters for evidence synthesis, as this may exclude relevant results. If you do use a limiter, it is expected that you describe doing so and provide a justification in your manuscript. 

Using NOT 

AND and OR are called Boolean operators. A third one is NOT, which takes results out of your search. As with database limiters, NOT is rarely used for evidence synthesis because it can exclude relevant results. 

Here is my revised search. I've applied search field tags to my keyword terms so I'm no longer searching in [all fields] and truncated some terms. Try copying and pasting this into PubMed. How many results do you get compared to the previous search? Do the results seem more relevant? 

("Cholesterol, LDL"[Mesh] OR "Hypercholesterolemia"[Mesh] OR "LDL cholesterol"[tw] OR "high cholesterol"[tw] OR "bad cholesterol"[tw] OR "elevated cholesterol"[tw] OR "hypercholesterolemia"[tw]) AND ("Hydroxymethylglutaryl-CoA Reductase Inhibitors"[Mesh] OR "hydroxymethylglutaryl CoA reductase inhibitor*"[tw] OR "HMG CoA reductase inhibitor*"[tw] OR statin*[tw] OR atorvastatin[tw] OR lipitor[tw]) AND ("red yeast rice"[tw] OR cholestin[tw])

Important note! This is my search after one round of revisions, kept relatively simple for readability. It will take many more revisions before I'm ready to export my results from the database. What other related terms could be added to this search?

Database-Specific Resources:

Use the Search Explore tab in the Search Log Template spreadsheet to keep track of your search. 

Once you have finalized your search in the first database, you must translate it for all other databases. This will involve reformatting the search to work your other databases and looking up the controlled vocabulary in each. Remember that some general databases like Scopus and Web of Science do not have controlled vocabulary, so in these databases, you would only use keywords. Some databases will also have a larger thesaurus than other databases. If one of your key concepts does not show up in PubMed's MeSH, for example, try searching it the next database's thesaurus when translating the search. 

Here is the revised PubMed search from the box above translated for two different databases: 

The Embase thesaurus is called Emtree. Emtree is similar to MeSH, but uses some different terms. Here the MeSH term "Cholesterol, LDL"[Mesh] translates to 'low density lipoprotein cholesterol'/exp in Emtree while the term "Hypercholesterolemia" is the same in both. MeSH doesn't have a term for "red yeast rice"/cholestin, but Emtree does have 'cholestin'/exp that can be added to the search--this is why it's important to do a new search of each database's thesaurus. The PubMed search looks for keywords in the text word [tw] field. The closest equivalent in Embase is searching in title, abstract, and keywords :ti,ab,kw. 

In CINAHL, the thesaurus is called CINAHL Subject Headings. It again has some terms that are the same as MeSH and some that are not. Keywords entered into this database are automatically searched in the title, abstract, and subject terms fields, so if those are the fields you'd like to search, you don't have to specify as with PubMed and Embase.

Search translators:

You may be able to find and use existing search filters for your evidence synthesis project in an already published article or in one of the search filter resources listed below. If you incorporate an existing search filter into your search, remember that you must cite it accordingly!