This edition of the HILJ club has been prepared by Alan Fricker, NHS Knowledge and Library Hub Manager. @AlanFricker.bsky.social (Twitter RIP)
Paper for Discussion
Development of a validated search filter for Ovid Embase for degenerative cervical myelopathy
Maaz A. Khan BA, Oliver D. Mowforth BA, MB BChir, Isla Kuhn MA, MSc, Mark R. N. Kotter MD, MPhil, PhD, Benjamin M. Davies MRCS, BSc
Volume 40 (2) June 2023 Pages 181-9 https://doi.org/10.1111/hir.12373 (Open Access)
Abstract
Degenerative cervical myelopathy (DCM) is a term for cervical spinal cord compression due to spine degeneration, but inconsistent terminology makes literature searches challenging. The study aimed to adapt an Ovid Medline search filter for the Ovid Embase database to improve literature searches for DCM. The filter was adapted and validated with reference to a previously established “Development gold standard” collection of articles. Direct translation wasn’t possible due to different indexing and fields. The ‘focus’ function was used to enhance precision. The resulting filter achieved 100% sensitivity in testing, providing a reliable tool for more efficient evidence retrieval.
HILJ Club reflections
I picked this article as I am interested in how we can speed up and improve retrieval across different databases. I look after a discovery tool so thinking about how we can move between or simultaneously search databases better is appealing.
I don’t know anything about the condition in question (Degenerative Cervical Myelopathy (DCM)) but can see that it reflects a common issue of shifting terminology and the emergence of terms to attempt to encompass a range of previous descriptors. There is a lack of established MeSH and other subject headings. The original paper was published in 2018 so there is scope for there to have been improvements to the indexing since then.
The previous Medline filter development saw a set of 250 relevant articles identified and I was suprised to hear that only 220 of these were found in EMBASE. I thought Medline had been added as a whole into EMBASE – I assume the researchers excluded the Medline in EMBASE records?
Table 1 shows the iterative process of moving to a 100% sensitive filter. However there is a cost in the precision with not far off 50% more articles retrieved to capture the extra 10 articles (300 to capture the last 2). Quite a high cost (the authors acknowledge this and justify it). The subsequent use of Focus seems to reflect this – an attempt to shrink the result set back down.
This looks like a thorough piece of work though not quite what I was hoping for.
Questions?
It feels like not much is different here than when I used to move searches between databases? Mostly a function of redoing the search taking into account the foibles of the database within the search interface. What other options are available to support this process? Are there automated tools to do this (the current issue of Journal of EAHIL is an AI fest and my reading is generative AI is not good at search translation)? This article might offer a useful test of those tools!
Are we happy with using Focus in the way the authors did? My feeling was that Focus is subjective and potentially variable based on the indexer. Is it reasonable in an area flagged as inconsistent for indexing?
HILJ Club 
Sometimes using Focus is justified but not very often. I recently used it in a search about smoking in pregnancy – as you can imagine a lot of abstracts mentioned smoking almost as an aside. So to cut the search from many thousands to a few thousand, I used .ti. and focus. This worked well but I’m not sure how often it would be best practice.
My feeling about search filters in general is that they can be useful as a starting point but in certain areas, would go out of date. Care should always be taken to ensure any new subjects/scope notes are considered.
LikeLiked by 1 person