Search filters/hedges

What is a search filter or a search hedge?

Very simply, a search filter (also known as a "search hedge") is a comprehensive canned search that you can add to your database search. [Note: The term "search filter" will be used throughout this guide.] Each database and search interface has a specific controlled vocabulary and syntax that can be used to harness the power of the database interface and content. (Yes, this is "librarian-speak" but it is important to the process!)

Who develops search filters?

Most search filters have been developed by experienced librarians who understand controlled vocabulary and database syntax as well as general search strategy rules. Experienced librarians understand the way to develop good searches through years of practice and, yes, through trial, error, and testing! 

What should be expected from a search filter?

A well-designed search filter will very deliberately return extremely (and possibly overly) comprehensive results.

High specificity: If there is any possibility an article is describing some kind of whatever it is that filter is trying to find, that citation should be returned in the search. Ideally, a search filter should approach 100% sensitivity. Many true negatives will be kicked out with a search filter but false positives will abound in the returned results, resulting in....

Low specificity: You will notice many false positives, i.e. items that are returned in the filter search that do not belong. Fortunately, many of those false positives will be eliminated when combined with the other concepts of your search.

How current is the search filter?

Language changes and terms get added. You don't want to miss out on the most current studies simply because the search filter was out of date! Each search filter on this site will indicate when it was last updated.

Is the search filter designed for the database and interface that I am using?

Search filters are not interchangeable across databases! There are two reasons for this.

1. The controlled vocabulary is different for each database. "Controlled vocabulary" refers to the indexing terms used by the database creator. A very simple example is a comparison between PubMed (MeSH) and Embase (Embase thesaurus). MeSH terms tend to be written as plural terms: neoplasms. Embase terms tend to be written as singular terms: neoplasm. It seems minor, but it isn't! 
2. Interface syntax varies across databases.
   • PubMed search for cancer would look like this: neoplasms[mesh] OR cancer[tiab] OR neoplasms[tiab]
   • A comparable Ovid Medline (Medline is the content database of PubMed) search would look like this: exp neoplasms/ or (cancer or neoplasms).ti,ab,kf. 

You cannot grab any search filter and add it to your search. You must make certain you have a search filter designed specifically for your interface and database.

Was the search filter tested?

Ideally, some guidance should be given as to how the search filter was tested. As librarians, we have all thrown together searches that are reasonably comprehensive/thorough and felt comfortable that, at the end of the day, the search met the needs of our patron. Search strategies are definitely a different animal!

How are the search filters tested on this site?

The search filters found on this site have undergone multiple tests by the developer of the filters. Revisions are performed several ways. One is by comparing against other filters to determine if there are additional terms that should be added. Another way revisions are performed is by running a filter, then reviewing items returned to determine if there are term variations to consider. You might think a trial is a trial is a trial but authors have funny ways of describing their research. The trials search filter includes terms for 2-, 3, and 4-arm studies (which are never referred to as trials in the title or abstract). Terms for different types of trials were also added recently including "stepped wedge" and "window of opportunity". 

Prior to adding a term permanently, tests are performed by comparing results from the current filter against results from the revised filter. Additional revisions are sometimes made to correct for degree of proximity (how many words separate 2 terms in a phrase). 

Additionally, librarians who have used these filters will, at times, suggest edits. As changes are made, the date of the update is revised as well.

Steps for creating and/or revising a search filter

1. Compare the filter against a list of known items. The search filter should find all items in the list.
2. Compare the filter against results from other filters. The search filter should find all of the relevant items found by other filters and as well as relevant items not found by the other filters. Ideally, I also hope to reduce the number of non-relevant items.
3. Compare the results of the potential new filter against the old filter. We want new terms to improve sensitivity; we do not want them to drastically reduce specificity! In other words, more true positives should be found without greatly increasing the number of false positives. If that happens, then the search is modified to find ways to improve the latter without impacting the former.
4. Review new vocabulary and add terms as appropriate.
   • For example, there are many types of systematic reviews: scoping reviews, umbrella reviews (also called "systematic review of systematic reviews"), and so forth. A search filter for those reviews should include those terms as well as variations that appear in the literature. If you look at the SR review filter, you will see a high degree of complexity as authors do not use a standardized vocabulary when describing their research.
     • A lesson for you: If you want your research found, use standardized vocabulary as much as possible. Appropriate methodology terminology can be found in health research reporting guidelines which are available from the EQUATOR Network.
5. Test the filter with a subject search. Apply the filter on a subject search, then remove those items found with the filter and scan what is left over to determine if items are in the list that the filter should pick up. Examine the terminology used and incorporate in the filter, returning to Step 3.

Using search filters

There are examples throughout the search filters guides demonstrating how to use them. 

Steps to add a search filter to your search

1. Run your search as you normally do in the database of your choice.
   • Search filters are currently available for PubMed and Ovid Medline.
2. After you are done combining your terms, find the filter based on the database you are searching and the research methodology you want.
3. Copy the search filter and go to the database you are searching.
   • If you are searching PubMed, clear the search box.
4.  Paste the filter into the search text box; hit <Enter> to run the search.
5. Combine using AND with your search strategy.

Example PubMed Search

PubMed Advanced Search: History & Search Details

I still don't quite get how to use search filters

If you still have questions, you can use the HSLS Ask A Librarian service if you are part of the Pitt/UPMC community.

If you landed on this page and you are at another college or university, consider getting in touch with a librarian at your institution for assistance. If you have never spoken with one of your academic librarians, now is a good time to give it a try. Librarians know a great deal about database searching and finding just what you need.