Tuesday, August 22, 2017

Translating search strategies (updated)

This is an update of a post from November 2015.

If you are doing a systematic review, you will need to run the same search in more than one database.


I use the term "translation" for the process of modifying a search so that it works in another database.


It is important to run the same search in all databases.


So, if you have a search that works in Ovid Medline, what do you need to change to make it work in another database like PsycINFO or Embase?  Those other databases have different features or functionality.


There are three areas where translation is necessary, once you have finalised your strategy:


Search syntax – phrase searching, proximity, truncation, field searching
Subject headings
Limits



Here are some pointers.    For detailed help, consult the help files of the database itself.


Search syntax


Phrase searching


Ovid will assume that the keyword heart disease is the phrase heart disease. 
NHS databases, Cochrane, Web of Science and PsycINFO via Ebsco will not assume this and will treat the keyword as if it was heart AND disease.  To force a phrase search, put the phrase in double quotes (speech marks):


"heart disease".


Proximity operators
A proximity operator finds one keyword next to another, or one within a specified number of words of the other.


So, in Ovid Medline:


heart adj disease


finds the word heart next to the word disease, in that order.   (This is the same as searching for the phrase, of course)


heart adj2 disease


finds the word heart within two words of the word disease, so will find heart disease, disease of the heart.


You can use proximity operators with the truncation symbol (see below), so


heart adj3 disease*


finds heart disease, disease of the heart, diseases of the heart, diseases of the human heart.


Proximity operators in other databases are:


Ovid Embase and NHS databases - the same as Ovid Medline


PsycINFO via Ebsco - Nn finds one word within n words of the other, in either order.  Wn finds one word within n words of the other, in that order.


Cochrane - NEAR/n finds one word within n words of the other, in either order.  NEXT finds two words adjacent to each other (you cannot use truncation and phrase searching together in Cochrane, so this is an alternative).


Web of Science  - NEAR/n works the same as in Cochrane.  


Truncation and wildcards


Using a truncation symbol will find words that begin with that string of letters, so disease* finds disease, diseases, diseased...


I think * is the truncation symbol in most databases but check the help files.  Also check the help files to see how many characters the * can stand for.


Field searching


In Ovid you will see:


Heart.mp.


.mp. stands for the (article) title, abstract and some other fields and Ovid has searched for the word heart in any of these fields.   .mp. is the default for free text searching in Ovid.   To search just in titles and abstracts in Ovid, you can use .ti,ab.:


Heart.ti,ab.


Ovid Embase also uses .mp.   In other databases there is probably no exact equivalent of .mp.   To search for words in titles and abstracts, do this:


NHS databases: .ti,ab   (no second full stop)


PsycInfo via Ebsco - you can use the drop down box to select title, or abstract.   Not making a choice searches all fields, and this is the closest equivalent to .mp.


Web of Science - select Topic in the drop down box.   This is the closest match to Ovid's .mp.


Cochrane Library - (heart):ti,ab    You must put the search term(s) in brackets.  Note the colon and that there is no final full stop.  


Subject headings


Medline has subject headings, called MeSH, and you should use these in your search.   These subject headings are listed in a “thesaurus”.   In your search history they have a final /, for example Diabetes Mellitus, Type 2/.    Exp before the term means the term is exploded.  


Some databases do not have a thesaurus.  Some databases have their own, and so the subject headings may not be the same as those in Medline.


You need to take this into account when modifying your search.


Cochrane uses MeSH (but not in all databases).  Use the MeSH terms you used in Medline.  In the Search Manager, you can enter them like this:


[mh heart]


This will explode the term heart.  If you do not want to explode it, use ^, like this:


[mh ^heart].


If the MeSH term is a phrase, put it in double quotes:


[mh "diabetes mellitus"]


Web of Science has no thesaurus at all.  In a database without a thesaurus, I would ensure
I have searched for the MeSH term as free text.    You have probably done this anyway.


PsycINFO, Cinahl and Embase have their own thesauri.


Identify the term that is used for the concept you are searching, by entering the MeSH term or another keyword and exploring the thesaurus.    


If you have used explode in Medline, a good rule is to use it in other databases.


PsycInfo has more detailed terms for psychological concepts, and Embase for drugs and drug administration, so there may be thesaurus terms available that Medline does not have.  


Any thesaurus may have different terms from Medline, for example, the subject heading in Medline for type 2 diabetes is Diabetes Mellitus, Type 2/, and in Embase it is non insulin dependent diabetes mellitus/.


It may be too late to tell you now, but translating a Medline strategy is easier if you have MeSH and freetext on separate lines of the Medline strategy.


Limits


You may have used limits in Medline to find particular publication types, or work relating to a particular age group.


There are issues with limits and you might have used search filters instead (see this separate post).  If you have used a search filter, look for a version for the database you are
translating into.  


If you have used limits, check if those same limits are available in the database you are translating into.


Here are some things you should know:


Date and language limits are available in most databases.


There are detailed publication type limits in Medline, and also in Embase and Cinahl.   The definition of, say "randomised controlled trial" should be the same across the three databases but you might be wise to check.


PsycInfo has detailed publication type limits, and you should check that they match what you used in Medline.


Web of Science has no publication type limits beyond "journal article" and "review".    You would need to screen your Web of Science search results for references that match your chosen publication type.


The Cochrane Database of Systematic Reviews is of course all systematic reviews, so no limit needed.   The Trials database in Cochrane is all clinical trials (randomised and otherwise) so again no limit is needed.


Embase, Cinahl and PsycInfo have age group limits, but they might not be the same as Medline's (or each other's).  


Web of Science and Cochrane have no age group limits.