Tip #43: EMBASE (Ovid) Candidate Terms

The Gist:

  • Candidate terms can uncover relevant articles not retrieved using title, abstract, or keywords.
  • The difference between word indexed and phrase indexed fields is really important, not just with candidate term fields, but with ALL fields. 


In EMBASE, on Ovid, my favored fields are TI (title), AB (abstract), KF (keyword heading word) and DQ (candidate term word). If you are curious about the KF field, Kate wrote about why you might use it rather than KW in Tip #4. That keystroke heard around the world on the day Kate's blog dropped was the sound of searchers changing KW to KF. 

Today's post, however, is about candidate terms, which Ovid defines as new concepts proposed during indexing and judged to be useful additions to Emtree (EMBASE's thesaurus). Over 100,000 candidate terms are proposed each year! 

In addition to having their own fields (DJ & DQ), candidate terms are indexed in SH (EMBASE subject headings) and HW (heading word) fields. Since HW is one of the default fields for unqualified searches, searching for a term without specifying a field in advanced search, or alternatively specifying .mp. (stands for multi-purpose), will return records indexed with candidate terms. However, as they are in candidacy, they have not been added to Emtree, and will not be picked up when browsing or mapping to Emtree.

This is how and where candidate terms appear in an Ovid EMBASE record:


Word Indexed vs Phrase Indexed Terms

There are two fields for candidate terms and the difference makes all the difference in how these and all other fields work. 

DJ is the Candidate Terms field and it is phrase indexed
DQ is the Candidate Term Word field and it is word indexed

Rooted in candidate terms is the need to understand phrase vs word indexing. When you know the difference, your choice of one over the other will be more skilled. 

When a field is phrase indexed, a search query must include the exact phrase, unencumbered by operators, in order to purposefully retrieve articles indexed with that phrase. Essentially, a phrase indexed field does not support operators

Another way to think about phrase indexed fields: you wouldn't alter a subject heading by adding quotation marks, wildcards, operators, or otherwise modifying the heading. You'd be wise to apply this restraint to any phrase indexed field.

Consider Kate's testing in Tip #4 when she searched for (surg* ADJ teach*).kw. which did not result in the keyword heading phrase surgical teaching in any of the records. When you use adjacency in a phrase indexed field, it defaults to Boolean AND among phrases, not within phrases. If you expand the adjacency operator to ADJ10, you'll get the same number of results. If you replace the adjacency operator with Boolean AND, it will continue to result in the same number of records. It's all Boolean AND. If you search (surg* OR teach*).kw., it searches among the list of keyword headings not within any single phrase, which would be an especially unsatisfying search. 

Kate's solution was to use KF, a word indexed field which does support operators both within a phrase and among phrases. 

But, back to candidate terms. Like the KF field, DQ is word indexed, meaning the field is processed at the word and not phrase level. Thus, it allows Boolean, adjacency, frequency, truncation, wildcards, and other symbols.

How do you find candidate terms?

  • In the wild, discovered in the Candidate Terms section of an EMBASE record.
  • In the index for the DJ field using the Search Fields function.

Index for DJ (Candidate Terms)

When do you use which? Very simply:

  • Use the phrase indexed DJ field - without operators - when you know the exact phrase. 
  • Otherwise, use the word indexed DQ field - operators allowed.

Candidate Terms Bring More Results

Why would you use these fields? Consider the following literature search: "correlation between impact of critical event debriefing and processing time for healthcare workers."

A search of Emtree via the Search Tools Map Term feature did not result in a suitable match for debriefing. I might have stopped there if I'd not known about candidate terms. 

Thus, I searched debrief*.dq., which resulted in 213 records. 

Since debriefing is a singular word, it made sense that it would also return results in the phrase indexed DJ field, which it did - 90 records. 

The 100+ difference in results between DQ & DJ indicated that debriefing also would be found as a word within a phrase. And it was. For example, the phrases  "cognitive debriefing interview" and "team debrief" appeared in some records.

I then tried debrief*.dj. and got 99 results, 9 more than debriefing.dj., which implied the lexical root debrief might also be a candidate term. (I know I said that wildcards don't work in phrase indexed fields. In fact, they will work with single-term concepts, like debrief, but not with multiple term phrases like surgical teaching or rather not how you wish them to. It's complicated.) 

Those 9 additional records included two instances of debrief, as suspected, but also debriefings, debriefing session, Debriefing Assessment for Simulation Healthcare Score, debriefing questionnaire, debriefing to learn, and debriefing to treat.

Finally, I compared debrief*.ti,ab,kf. with debrief*.ti,ab,kf,dq., which resulted in 63 more records when including DQ. 

That candidate terms are deemed by indexers to be potentially useful additions to Emtree; plus seeing its use result in additional potentially relevant records are reasons enough to add this field to my searches.

Shout out to my colleague Dana Gerberi who first introduced me to the DQ field. 

Thanks, that's all I've got. 


Comments

Popular posts from this blog

Tip #1: Bulk export from Google Scholar

Tip #23: PubMed's [tiab] vs. [tw]

Tip #4: Ovid MEDLINE Adjacency and Field Tags