How to phrase search on PubMed

So, you know the title of an article, you want to search for it on PubMed, you paste in the title, put quote marks round it, maybe specify title field and then you find it? WRONG.

Incredibly, PubMed will not accept this. See information from PubMed here: https://pubmed.ncbi.nlm.nih.gov/help/#phrase-index

If your phrase isn’t a recognised concept in their arbitrary phrase index, you can’t search for it as a phrase. You’ll get a vague error message saying that part of the search was ignored.

To force the matter, you can use a proximity search with a proximity of zero, e.g. [ti:~0].

Furthermore, this can cause trouble when truncating. Suppose you’re searching “endometrial carcinoma*”[tw]

– this is fine, the visible phrase is recognised

You decide to truncate to include, say ‘endometrioid carcinoma’ too, so:

“endometr* carcinoma*”[tw]

NO GOOD! Despite this merely being an expansion of what you’ve just searched, it is not permitted on PubMed because it doesn’t recognise the phrase (due to the truncation).

Worse, you can’t use the Text Word field with proximity, so you can’t even do [tw: ~0]. You might consider [tiab:~0] EXCEPT you can’t use truncation with proximity either so that’s no good either.

In the end I think you’d just need to search for both phrases in full: “endometrial carcinoma*”[tw] OR “endometrioid carcinoma*”[tw]

But it’s a pretty unintuitive situation.

Importing references in XML format from REHABDATA into EndNote

REHABDATA by NARIC is perhaps not top of the databases you’d consider essential to use for a literature search for most topics, but if you are using it (or indeed any other database that only exports in XML), how can you get your results into EndNote?

The following process I’ve come up with is based primarily on this helpful video’s process for converting XML files into an EndNote-friendly tab-delimited format.

Save your results from REHABDATA in XML format.

Open the file in Excel.

Edit the column headings to exactly match the EndNote fields you want the data to go into, and delete all unnecessary columns. So, for example, I’d suggest ending up with: Author, Title, Journal, Year, Volume, Issue, Pages, Abstract, Keywords, ISSN for journal articles.

If you’ve got book results, you’d probably want to cut and paste these out into a new Excel sheet, change the column headings to ones relevant to ‘Book’ reference type and repeat the process with a separate file.

Save the tidied up table as a Text (tab delimited) (*.txt) document.

Open this in Word.

Put the cursor at the start of the document, press return to get a blank line, and type:

*Journal article

(including the asterisk at the start)

Run a Find and Replace (Ctrl-H), for:

Find: | (vertical line – usually shift \ on UK keyboards)

Replace with: // (two forward-slashes)

Replace all

Find: (double-quote-mark symbol – usually shift 2 on UK keyboards)

Replace with: (nothing! leave blank)

Replace all

Save (as text file)

In EndNote, go to File > Import > File

Choose the edited text file.

Select Tab Delimited as the Import Option.

Go!

Success!

If you get an error saying you’ve got the wrong field names, this will be because one or more of your headings you were editing in Excel are not the correct exact wording of a corresponding field name in EndNote. Try again!

You may find some older results appear with all the journal details crammed in the title field. Unfortunately, that’s just how the data comes out of REHABDATA. Obviously you could do some cunning find/replaces (in Excel or Word or EndNote), but that’s not really part of the import process; I’ll leave that to you to work out if you have sufficient results of that format to warrant it!

Sharing searches on Ovid databases (or re-running searches from text)

(thanks to the Ovid support team for info about their search launcher tool)

You’ve done a long and complex search on Medline (or Embase or PsycInfo or whatever) on the Ovid platform, and you’d like to show it to someone else, so they can assess the results or play around with the search?

The ’email jumpstart’ function (from your saved searches) does not permit this unless the person using the jumpstart link is logged in on the same personal account. Sometimes setting up an account for a project may be a useful way forwards, but often it may not be practical.

Solution? Ovid have a really handy ‘search launcher’ tool: http://demo.ovid.com/demo/ovidsptools/launcher/launcher.html

It allows you to simply paste in the text of a search, pick the database you want to run it on (by choosing the appropriate code from their list) and run.

To share a search, you just need to send the search steps to your recipient in a format that can easily be pasted into the search launcher – this needs to be each step of the search on a new line, without set numbers or results numbers.

There are a number of ways to get this from your search – if you select one or more results and choose to export or print results, you can tick to include your search history (which you can then use), using this option, the easiest two routes seem to be:

Choose Print, click to view the print preview, copy the search history and paste into Word as a table, highlight and copy the correct column (without column heading).

Or: Export, choose to do so in XML format, save, open with Excel, highlight and copy the correct column (without column heading).

Send the search lines to whoever needs it, then they can paste into the launcher and start working with it.

Another interesting application of the search launcher is when you come across an existing search (in a review article appendix, for example), this tool means you can relatively easily re-run that search  without typing it in step at a time.

To prepare the search to be pasted into the launcher, you might want to go through this process:

Copy and paste the search into MS Word. I’m assuming each step of the search will be on a separate line, but will probably have the search step number at the start of the line (and this is what will need removing).

Put a blank line at the start of the document (to ensure there’s a paragraph mark before line ‘1’ – this will be needed shortly).

Press Ctrl+H to bring up Word’s Find and Replace function. Click More >>> to see the advanced options, and tick Use wildcards

In the ‘Find’ box put:

(^013)([0-9]@ )

With a space before that final bracket. This is telling Word to look for two things in sequence: a paragraph marker, then a digit 0-9 any number of times (i.e. any number) followed by a space.

In the ‘Replace’ box put:

^013

This is telling Word to replace with just a paragraph marker.

Replace All and that should be that.

If there is any other punctuation in the search, just add that into the ‘Find’ box as required; e.g. if you have dots after your step numbers, use: (^013)([0-9]@. )

You should end up with a tidied-up search with just each step on it’s own line – ready to be pasted into the launcher and run as a search.

***UPDATE***

If you have text you want to remove at the ends of lines, obviously you can add that in as part of the searched-for sequence.

If you want to use a character that would form part of the search syntax (for example, a bracket), use a backlash in front of it, e.g. \(

I found lots of helpful detailed info about what can be done with Word’s wildcard finding and replacing functionality here.

If you find the search text doesn’t paste correctly into the Search Launcher with line breaks, run another Find and Replace – with ‘Use wildcards’ unticked – to find ^013 and replace with ^p and this should definitely ensure regular line breaks are there to copy.

Finally, the link on the Search Launcher page seems out of date so, for convenience, here’s a current link to the list of most Ovid database shortname codes, plus it can be worth checking individual database guides (for e.g. Medline and Embase).

If rerunning a search in order to update a search (run in the past), and you only want to get results added since a date it was previously run, the best fields to use are (generally, currently, though there are other fields worth considering): ‘ed’ (‘Entry Date’) in Medline and ‘dc’ (‘Date Created’) in Embase, both applied using this format:

limit 57 to ed=20170329-20191128

(Limiting results from step 57 to entry dates of 29th March 2017 to 28th November 2019. In Embase, it would just be dc instead of ed)

Though note: ed generally won’t work for Medline E-pub Ahead of Print or In-Process & Other Non-Indexed Citations – to get new results from these sources, you may need to run the search independently on them without the ed limit.

***UPDATE 12/2021***
Ovid now has an incredibly useful function of a button that will create a link to your search that you can send to anyone, thus obviating the need to use the search launcher for this. However, the launcher still remains really handy when you have a long search in text format that you want to run.