Tip #51: MeSH Major Topics across PubMed and Ovid, with a bonus glimpse of changes to NLM indexing practices

You might think that a recent publication indexed in PubMed with

Epilepsy, Temporal Lobe* / pathology 

would be retrieved by this query in Ovid MEDLINE :  

*Epilepsy, Temporal Lobe / pa [Pathology]

 

You would be incorrect.

One example is Xie K, Royer J, Larivière S, et al. Atypical connectome topography and signal flow in temporal lobe epilepsy. Prog Neurobiol. 2024;236:102604. doi:10.1016/j.pneurobio.2024.102604.

 and you can compare for yourself by searching for 38604584.ui in your preferred Ovid Medline segment, where it will show 

This issue, of Major Topic MeSH-with-subheading not being accurately imported into Ovid, seems to affect literally every article since mid-2022 with a Heading/subheading combination marked as a Major Topic.

What’s happening, and why?

At the moment of intake processing, Ovid breaks apart headings consisting of a Major Topic and subheading.

 

NLM Data	Ovid
HIV Infections* / prevention & control	HIV Infections/pc [prevention & control]
	*HIV Infections

This sounds like it’s going to be a lot. What are the (too long; didn’t read) tl;dr takeaways?

• If you search Ovid, or have a saved search in Ovid with a query that combines a Major Topic with a Subheading, it is broken, and has been broken for a while now.
• The recommended workaround is to search with (*MeSH term AND MeSH term/subheading) instead of *MeSH/subheading.
• MEDLINE indexing now seems to require that if any MeSH/subheading pairing is a Major Topic, all other instances of that MeSH in the record are also Major Topics.
  
  

What’s happening, and why? (continued)

This raises questions, as we’ve all seen accurately represented Heading/subheadings in Ovid. 

The missing piece to this puzzle is that NLM indexing allowed for Major Topic to be applied to both headings and subheadings. From Topic Searching in PubMed: 

"an article about sleep initiation and maintenance disorders that discusses drug therapy but where drug therapy is not a main point of the article would be indexed with:
    > Sleep Initiation and Maintenance Disorders* / drug therapy

If the heading/subheading combination describes the main topic, the asterisk is placed after the subheading. For example, an article that is mainly about drug therapy for sleep initiation and maintenance disorders will be indexed with:
    > Sleep Initiation and Maintenance Disorders / drug therapy*"

The good news is that for PubMed, it doesn’t matter. From the same page:

    This placement does not affect PubMed searching. A search of either:
       
        Sleep Initiation and Maintenance Disorders[majr]

        Sleep Initiation and Maintenance Disorders/drug therapy[majr]

will retrieve both records, regardless of where the asterisk is applied.

But in Ovid, it does matter. 

I contacted Ovid.  Their response included:

"A few years ago the NLM stopped making that distinction and they [stopped indexing subheadings as major topics]. As far as we can tell there was no announcement about this change and Ovid thus kept on ingesting the data in the same way."

So… when Ovid takes in a Medline record data with [majr] on the heading, rather than create a [majr]Heading/subheading pair, Ovid splits the line into a [majr] heading and a second Heading/subheading without any [majr]. 

When Ovid imported a record from the NLM where [majr] was on the subheading, Ovid faithfully rendered that with [majr]Heading/subheading.  

But. If the change in NLM indexing, mentioned by Ovid above, where the NLM no longer puts [majr] on subheadings, goes back a few years… it implies we have years of Ovid records that aren’t findable with queries that are faithful to PubMed-displayed indexing. 

This is unfortunately (and surprisingly!) the case.

Show me impacts. 

Records in PubMed showing a MeSH-with-subheading as a major topic are not findable in Ovid with that query.  

A query consisting of a focused MeSH term paired with a subheading in Ovid will retrieve fewer results than the same search in PubMed, even after accounting for divergences like unexpected subheading explosions and unexpectedly-conjoined pharmacologic action trees. 

PubMed	"HIV Infections/prevention and control"[Majr:NoExp]	30,424
Ovid	*HIV Infections/pc [Prevention & Control]	23,434

In all below examples, PubMed years used the date slider, and Ovid ANDed the query with 202#.yr

Year	PubMed	Ovid
	"HIV Infections/prevention and control"[Majr:NoExp]	*HIV Infections/pc [Prevention & Control]
2024	590	0
2023	1873	0
2022	2281	34
2021	1907	367

Year	PubMed	Ovid
	"Pre-Exposure Prophylaxis/statistics and numerical data"[Majr:NoExp]	*Pre-Exposure Prophylaxis/sn [Statistics & Numerical Data]
2024	11	0
2023	6	0
2022	6	3
2021	27	25

Year	PubMed	Ovid
	"COVID-19/epidemiology"[Majr:NoExp]	*COVID-19/ep [Epidemiology]
2024	5355	0
2023	18504	0
2022	24916	747
2021	13655	8429

What is the recommended workaround, and does it work?

Ovid’s recommendation is to build a query as follows:

*Focused MeSH AND MeSH/subheading.

If your PubMed search is as below, the recommended query to get the same results is 

PubMed	Ovid
"COVID-19/epidemiology"[Majr:NoExp]	*COVID-19/ and COVID-19/ep [Epidemiology]

… but the results won’t be exactly the same. The Ovid workaround query will also retrieve records where /epidemiology is present, but a subheading other than Epidemiology is focused. 

Consider
Zhu W, Xie K, Lu H, Xu L, Zhou S, Fang S. Initial clinical features of suspected coronavirus disease 2019 in two emergency departments outside of Hubei, China. J Med Virol. 2020;92(9):1525-1532. doi:10.1002/jmv.25763 

where /diagnosis and /virology are Major, but /epidemiology isn’t. Ovid’s workaround for "COVID-19/epidemiology"[Majr:NoExp] retrieves this even though it does not have /epidemiology as a Major Topic.

The new behaviour of the NLM, as noted by Ovid, appears to be that if a Heading/subheading pair has [majr], any other instances of that heading on the record must also have [majr]. This is intuitively a loss to precision in indexing.

When will Ovid fix this?

From Ovid:

Some clever users pointed us to the differences in results between PubMed and Ovid, and already last year we wanted to make a change and update our ingestion scripts to follow the PubMed logic, but unfortunately it was too late to include this for the 2024 reload.

 With the 2025 reload for Ovid MEDLINE the Duplication in the Ovid indexing will be removed, and Ovid MEDLINE ALL will display the MeSH Subject Headings as [appropriate]. 

 

This post was submitted by Alexandre Amar-Zifkin, Bibliothécaire (librarian) at the Bibliothèque de la santé (Health library) of the Université de Montréal (University of Montreal), with revisions from Zahra Premji (@ZapTheLibrarian), Health Research Librarian from the University of Victoria Libraries, pre-publication.