Trip Database Blog

Following the news about the National Guideline Clearinghouse (NGC) having its funding withdrawn we have set about capturing as many USA guidelines as possible (see Guidelines on Trip – moving forward after the demise of NGC).  So, instead of linking to NGC summaries we decided we would link directly to the guidelines on the publishers site. We started migrating all the records last weekend (adding the new links and removing the old ones) and this process is now complete. We now have 3,495 US guidelines.  For comparison NGC has around 1,350 guideline summaries.

At Trip we break guidelines down by geography so the full count is as follows:

• USA – 3,495
• Canada – 1,730
• Australia and NZ – 1,149
• UK – 3,239
• Other – 948

So, across the board Trip links to 10,561 guidelines.

This is a unique FREE collection one we hope to develop in the forthcoming months (more to follow in due course).

This weekend we are migrating our American guidelines from the National Guideline Clearinghouse to our own collection.  For further information click here.

This is ‘non-trivial’ and there may well be some teething issues.  Still, the sooner we start the sooner we get it right…..

We’ve reached a significant milestone….

We’ve been working away, quietly, on the autosynthesis system and today we shared the beta-version with a small number of testers.  There are still issues relating to data quality, but every day it’s getting better.  By allowing others to test it means we’ve reached a significant milestone.

One things for sure though – I’m nervous about the feedback.  If it goes well we’ll extend the testing to a wider group.

Fingers crossed….

This is a follow-up post to Incorporating MeSH in Trip.

Adding MeSH to Trip is not going to happen overnight but we’re starting on the journey.  One key element we’re focusing on at the moment is ensuring all the documents within Trip are assigned MeSH terms. Many of our documents come from PubMed so these will already have MeSH terms assigned to them, we can simply grab these.  However, a large number of documents (synopses, guidelines etc) don’t, so we need to ‘tag’ these documents – and this is what we’ve been doing this week.

We’ve been trying three methods (helped by staff at the NLM), all based around their automated tagging systems that they use in-house:

1. Running the title through the automated system
2. Running the whole document through the system
3. Summarising the document and then putting it through the automated system.

The third option was tried as option 1 gave too few MeSH terms while option 2 gave too many.  And it seems that 2 is pretty good.  Here are two examples:

Source document: Acute management of croup in the emergency department, Canadian Paediatric Society.

Terms assigned:

Child
Humans
Infant
Croup
Antipyretics
Child Health
Consensus
Outpatients
Caregivers
Canada
Dexamethasone
Adrenal Cortex Hormones
Fever
Emergency Service, Hospital
Vomiting
Algorithms
Fungi
Airway Obstruction

Source document: Pimecrolimus for the Treatment of Adults with Atopic Dermatitis, Seborrheic Dermatitis, or Psoriasis: A Review of Clinical and Cost-Effectiveness, CADTH

Terms assigned:

Humans
Adult
pimecrolimus
Dermatitis, Atopic
Dermatitis, Seborrheic
sertaconazole
Cost-Benefit Analysis
Methylprednisolone
Dermatology
Tacrolimus
Imidazoles
Thiophenes
Dermatologic Agents
Psoriasis
Eczema
Adrenal Cortex Hormones
Treatment Outcome

The terms seem plausible and members of the Trip Advisory Board have agreed that the two-step approach (summarising before putting through the automated system) is the best. We’re not claiming it’s perfect, but it seems pretty good.

The next step is to try to make this a seamless, fully automatic, process – to generate the MeSH terms via simply adding a URL (as opposed to summarise first, then add to automated system).

I’m quietly confident that this ‘problem’ has been solved.  This then leaves the more complex problems of:

• How to incorporate it in to Trip (hopefully that’ll be ok)
• Creating an interface that works for everyone.  I suspect most health professionals will not want to bother with MeSH, so we’ll need to hide the complexity. However, information specialists will have opinions on how they’d like to see it work – so that will need some additional work….

The longest journey begins with the first step….

MeSH (full name Medical Subject Headings) is a controlled vocabulary that is widely used in medical information systems.  We’re actively exploring using it in Trip as we believe it can significantly improve our search results.

As far as I can tell it will improve them for two main reasons.  Firstly, it’ll improve our synonyms function as MeSH is great for that.  Secondly, and this is the most exciting aspect for me, is that MeSH is hierarchical. If you do a search for arrhythmia that maps to the MeSH concept of Arrhythmias, Cardiac:

From the image you can see that Arrhythmias, Cardiac is classed as a pathological condition (as you look up the hierarchy).  But as you go down the hierarchy you can see other terms such as atrial fibrillation, tachycardia, long QT syndrome etc.

This is important as by using MeSH in Trip a search for arrhythmia will also return results for atrial fibrillation, tachycardia, long QT syndrome.

No time frame for this work. But we’re starting investigating the various systems we require, namely:

• Annotating articles in Trip that currently don’t have MeSH terms assigned. This will need to be automated but there are already good systems to do this.
• How might it interact with the keyword searches our site currently uses? We need to figure out how MeSH might fit in with our current system.

I’m optimistic that this is not a massively complex piece of work – but we’ll have to see. If you think this would be useful feel free to send:

• Encouragement (which will help speed things up)
• General advice
• Ways you’d like to see it used
• Ways you’d not like to see it used

I recently wrote about the demise of the National Guideline Clearinghouse (NGC).  At the time it was fairly bad news as, by them aggregating the guidelines from multiple guideline publishers, it saves us considerable resource.  However, every cloud has a silver lining!  The NGC was not without challenges, for instance:

• It takes time to translate a guideline into one of their summaries – so it’s not as up to date as it could be.
• It seems to withdraw the summaries after 5 years.  This, seemingly, arbitrary cut-off means they withdraw guidelines which the producer still has them listed on their sites.

So, what are we doing about it:

• We have gathered our own collection of guidelines from the USA.  We have found nearly 3,000 guidelines, considerably more than the 1,342 guideline summaries have on their site (as I type this).  These are not live yet as we’ll need to transition over, but that will happen soon.
• We have refreshed our other collections and identified around 15 guideline publishers from around the globe that we have previously not included.
• Put this all together and it will cement our place as the leading source of guidelines.
• As well as featuring them in the main Trip search we are currently planning a separate guideline site, watch this space.

A massive amount of hard work, but it should be worth it.