On the -16th day of Xmas, NLM gave to me

- new and welcome changes to the MeSH tree

Triple Negative Breast Neoplasms
Prostatic Neoplasms, Castration-Resistant
Medical Marijuana
Mindfulness
Mobile Applications
Pediatric Obesity :(
Dandruff (really? I had no idea this wasn't in there)
Non-Nutritive Sweeteners
Nutritive Sweeteners

And lots of other good ones, too! However, these are the most likely to make an impact on daily searching.

But hey! Still no ptyps for Systematic Reviews and Protocols? Come on, the sooner you add it, the sooner it'll be useful!

Searching Steroids in Medline

Here's a real pain in the wazoo.

You search in the MeSH database and get the MeSH Steroids. So far, so good. But you get only a few articles on the topic and you have a sneaking suspicion that there should be more.

So, you try some other combinations of MeSH, or some keywords, and lo and behold, there is another MeSH term, Adrenal Cortex Hormones! It's in a completely different part of the tree than Steroids.

Grr. It particularly burns me that articles with the word Steroid in the title are indexed Adrenal Cortex Hormones rather than Steroids. If you're gonna be like that, why not use both?

The same thing happens with corticosteroids and glucocorticoids, which are narrower terms of each, so y'know, be warned.

Google Scholar sort by date removes results

OK, here's a weird one:

Doing a search in Google Scholar. Of course the default is Sort by Relevance.

Whoa! Way too many results, so I selected only articles from the past year (Since 2013)

OK, so now I have 662 results. OK, but I'd like to look through them by date...

I clicked on Sort by Date.

Now I've only got 3 results.

What gives?

Now, it turns out that whether or not you've limitted your results by year, when you click on Sort by Date, it always gives you just the past year. How ridiculous and annoying that is would take up a blog post in itself. But what I don't understand is why even when I'd limitted the results to just the past year, there were more than when I sorted by date. And it wasn't just one or two. It lost over 600 results!

This is not a one-time glitch either - it persists over time and with different browsers. Guess it's supposed to be like that? Man! Google Scholar has its good points, but it is damn useless when it comes to result manipulation.

They keep changing PICO...

...but it never gets better.

PICO was a good first step in trying to describe the process of turning a clinical question into a search strategy. However, it is fundamentally flawed. The problem with PICO cannot be solved by adding more letters onto the end. The problem can only be addressed by understanding what is wrong with PICO, and why so many clinicians find it confusing.

The fundamental problem with PICO is that it de-emphasises the most important question that a person can ask when starting work on their search strategy. That question, as explained to me on the first day of my practicum by my mentor, is: what is the disease or disorder in question?

Medline and other biomedical databases are set up so that the best search results are gained by searching first for the disease or disorder in question. Searching first for the disease or disorder was unquestionably the best advice I ever received for framing a search strategy.

However, every question that I've received from a clinician who had used PICO was missing that fundamentally important piece of information. These omissions resulted in hours of wasted time. In each case, once I finally realised what information I was missing, or queried the clinician and received a clarification, I had to completely revise my search strategy, and immediately began receiving much more relevant results.

A second problem with PICO, or perhaps an extension of the first problem, is that although it is intended as a tool for use by healthcare professionals, it is not written in their language. The very fact that we have to define the word "population" as something other than what a healthcare professional usually means by this term means that PICO fails to communicate with the target audience.

Another problem with PICO is that it assumes that most questions are about interventions, when many are about risk factors, statistics, cost-effectiveness, and other non-intervention topics. Many others do not have a specific intervention in mind, such as the perennial "best practices" request.

The fourth problem with PICO is that the organisation of the components does not correspond well to the order in which the search should be entered into a database. For instance, some of the information falling under population, such as age, species, and setting, are listed first in PICO, but entered last in most search strategies.

On several occasions, when I have mentioned the flaws of PICO to my colleagues who work in various health libraries, many immediately spoke up with great relief, agreeing that it causes more problems than it fixes. Yet it seems that no one wants to be the first to say that the Emperor has no clothes.

The Emperor has no clothes.

But it's not enough to say that there's a problem - that is futile unless it comes with a proposed solution. Let's develop a search frame that emphasises the disease or disorder, uses the target audience's language, and recognises that many questions are not about interventions. Even better, let's develop a search frame that is based on the structure of health databases, so that the clinician using the frame fits their keywords into a structure that readily translates into a database search. Clinicians know what they want to know - what they struggle with is how to tell a database what they want to find. We, as health librarians, are the people with the expertise to develop that translation.

I myself am not yet experienced enough to develop such a frame on my own. However, by pointing out the need for it, and giving a suggestion of what it might look like, I hope to spark the idea for the next tool in the minds of more experienced librarians. Here are some of the elements I think would be worth considering:

1. The first element could be the letter D standing for disease or disorder.
2. An I for specific interventions could be included although optional
3. The next element, also optional, could bring in the idea of risk factors, statistics, etc... unfortunately, the only words I can think of to describe this idea are Information and Demographics, both of which start with letters already taken
4. At this point there could be a P for population, including such concepts as age, gender, species, setting, type of patient (inpatient, outpatient, postop, etc...), chronic vs acute, etc..
5. An L for limits could be next - this would refer to publication limits (year, type, and language of publication, methodology, peer reviewed, etc...)

If the healthcare professional is going to submit that information to a librarian to do the search, then they can use that framework to create a sentence or paragraph explaining what they want.

However, if they're going to do their own search, the search frame has to indicate what to do with all this data.

1. Separate the search terms and search each one individually
2. Modify search terms with subheadings
3. combine synonyms with OR
4. combine concepts with AND
5. Refine with limits
6. Evaluate current search results
7. Look for new search strategies
8.  Iterate

Unfortunately that spells SMOARELI which is not exactly memorable!

I looked through my literature search requests from the past 6 months, and couldn't find much to add there, but I'm sure there are other and better suggestions out there in the minds of my fellow librarians!

My purpose in creating this post is not to downplay the value of PICO as a starting point. Like the Lamarck's theory or the Bohr atom, it was the first step in describing something very complex. What I'm saying is that PICO is not the end point. Let us not, by adhering too closely to the offering of our Lamarck, fail to inspire our Darwin.

Antibiotics

When searching for Antibiotics in Medline, the correct MeSH is:

Antibiotic Prophylaxis

not

Anti-infective Agents/therapeutic use

(Apparently).

Cone Rod Dystrophies, or a rare epic failure of MeSH

If you search Cone Rod Dystrophy in MeSH, you'll find that it is an entry term for Retinitis Pigmentosa, and (at least in the PubMed Mesh database), it maps only to that one term. I stupidly assumed that this meant it was the same thing. (Bad librarian, no cookie!)

My search was: "Retinitis Pigmentosa/therapy"[Majr]

Luckily, one of the articles on the topic set me straight. The abstract clearly stated that cone rod dystrophy was different from retinitis pigmentosa. I immediately stopped collecting articles on RP, and investigated.

It turns out that some people consider cone rod dystrophy to be a sub-type of retinitis pigmentosa. The more prevalent form of RP is rod cone dystrophy, where the rods are affected before the cones, resulting in night blindness. In Cone Rod Dystrophy, the cones are affected before the rods, resulting in a loss of central vision and sometimes in colour blindness. The causes of Cone Rod Dystrophy are mostly genetic, unlike with standard RP.

Armed with this new data, my new search was: ("cone rod dystrophy"[all fields] OR "cone rod dystrophies"[all fields] OR "cone rod degeneration"[all fields]) AND "Retinitis Pigmentosa/therapy"[Mesh]

However, I noticed that combining with the RP term reduced the number of articles to less than half of the results found with the keywords. It seemed odd - surely if CRD was a subtype of RP, most of the articles should have been included?

Experimental search: ("cone rod dystrophy"[all fields] OR "cone rod dystrophies"[all fields] OR "cone rod degeneration"[all fields]) NOT "Retinitis Pigmentosa/therapy"[Mesh]

Whoo-ey! It turns out that many articles about Cone Rod Dystrophy are indexed under "Retinal Diseases"[Mesh], or even more obscure terms. I noticed that the MeSH "Retinal Dystrophies" was hardly used at all, when I would have guessed that it would be the closest MeSH to Cone Rod Dystrophies. It's strange, because Cone Rod Dystrophies is so clearly an entry term for Retinitis Pigmentosa, and yet the indexers do not seem to be treating it as such.

It turned out that the best search was: "cone rod dystrophy"[all fields] OR "cone rod dystrophies"[all fields] OR "cone rod degeneration"[all fields]

I ended up going through all of the results, which was good because there was very little on therapy, and fortunately a couple of articles mentioned that fact (but of course it wasn't reflected in the indexing since the article wasn't about that). But that's a problem that had nothing to do with the indexing.

Normally MeSH terms are truly awesome, and too often taken for granted by those lucky enough to work in health librarianship. We only remember how brilliant they are when we end up searching some other subject and see how inaccurate and inconsistent the use of controlled vocabulary can truly be. However, sometimes there is a term within our own literature that shows us how messy things can be if indexing is not up to Medline's usual high standards.

Home birth

I searched home birth in the MeSH database and got no results.

I searched home birth in PubMed Reminer and found Home Childbirth.

MeSH database, can I get an "oops"?

D'as right.