Tuesday, September 26, 2006


I attended the latest DUETs meeting yesterday and was delighted to see the progress being made. Iain Chalmers and Muir Gray refer to the National Therapeutic Ignorance Service - highlighting the significant number of uncertainties in clinical practice.

Perhaps the most significant development will be the move, for DUETs, into the National Knowledge Service and close links with the NLH's specialist libraries. Muir was very strong in his support for highlighting uncertainty/ignorance. It may be too soon to hope that clinicians will enthusiastically embrace uncertainty, but things are beginning to change.

So DUETs will continue , highlighting and recording ignorance as it grows.

My involvement started around 2 years ago with a meeting with Iain to discuss the ATTRACT service and unanswered questions. I highlighted my favourite unanswered question - what is the optimal frequency of vitamin b12 injections in pernicious anaemia? In the UK these are given every 3 months and in North America it is typically monthly. However, there is no trial evidence to support either position. Still no answer to that Q though.....

Monday, September 25, 2006

Top sites part 2

In yesterday's post I showed that eMedicine article were viewed the most. However, I feel that's misleading, as eMedicine has lots of articles so are more likely to have a broader appeal. I suspect eMedicine is viewed more in esoteric areas not much covered by mainstream 'evidence'.

My preferred measure of 'top sites' involves looking at the number of views each source receives then dividing this by the number of articles contained within TRIP. So, eMedicine has 6,386 articles and these articles have been viewed 5,618 times - a ratio of 0.88. Rearranging the articles by this article/view ration changes things considerably:

Saturday, September 23, 2006

Top sites

I like to analyse various statistics associated with the site. One of these is how useful a particular resource is. Below is not an exhaustive list but some of the 'biggies' included in TRIP. The various column headings:

  • Source - should be self-explanatory
  • Views - how many times people have clicked on an actual article from that source
  • Articles - how many articles TRIP contains from that particular source
  • View/Article - simply division!

So what are the results:

So is eMedicine the top site? I don't think so!

Friday, September 22, 2006


Occasionally, this is the 4th time in nearly 4,000 questions, an answer from the NLH Q&A Service got rated ‘poor’. So that’s only 0.1% - but it still upsets me. The reason though for us doing badly?

It was a question about p-Phenylenediamine (PPD), which is often used, in henna tattoos (popular in the UK). The general practitioners (GPs) patient developed a severe allergic reaction and so wanted to know what other products PPD appears in – to help guide the person away from future allergic reactions.

The problem being that the Q&A services answer questions of the type answerable by research evidence, by searching the likes of TRIP, Medline, Cochrane etc. This question I beyond our remit and expertise. So, trying to be helpful, we ‘answered’ the question – badly – just because we tried to help.

The lesson to learn is to know your limits and sometimes trying too hard can backfire.

The outstanding question now is - do we leave our poor answer up there and hope people will help us out?

Wednesday, September 20, 2006

Search algorithm improvements

Less than a month since the launch of the free service and we're already planning our first algorithm improvement. This has come about by the observation that eTextbooks are appearing too high up in the results. TRIP is principally about returning best evidence. As such eTextbooks have a role. However, they should generally appear below other evidence sources.

The problem of them appearing so high up has been caused by the perfect match between search terms and eTextbook document titles . Typically users use a single search word (or phrase) to search TRIP. eTextbooks typically use single words or phrases in their title. So they score highly for title matches. In other words if someone searches for 'Asthma' a document just called 'Asthma' gets a significantly higher score than 'In children with asthma what is the appropriate management'. In the first example its a 100% match, in the 2nd it only matches 11%. we've tried various techniques to dampen this effect but now is the time to get the programmers in! Hopefully the new algorithm will be tested and released by mid-October.

Tuesday, September 19, 2006

FAQs and Ganfyd

There are a number of clinical questions that our services receive that occur with a fair degree of regularity, a few examples:

  • Is there any benefit in prescribing statins in the elderly?
  • What is the frequency of vitamin b12 injections in pernicious anaemia?

We answer these as best as the service can. However, we can only report on the evidence. If the evidence isn't there or is abstractly written our answers reflect that.

So why not 'throw' an FAQ to the willing doctors in Ganfyd and see if they can make them more accessible? Well that's what I've done with the question on overweight women and progestogen-only contraceptive pills

Monday, September 18, 2006

Tuesday, September 12, 2006

A little experiment

Just a trial of some new software my web-people (Sequence) have developed, to make site creation easy. They asked me to try it out so I thought I'd have a go at creating some new mini-sites for TRIP. I've created six sites and added google ads - not sure what, if anything, will happen with these:

Cognitive behaviour therapy
Smoking cessation

The software Sequence have developed was certainly easy to use. Even though I only created a batch of one page sites I've started more complicated multi-page sites and that seems equally straightforward.

Monday, September 11, 2006

More feedback

Two particularly nice bits of feedback today:

1) The publicity about the free availability of the TRIP database is an opportunity for us to say how much we have valued the database as an essential tool for our enquiry work.

We have subscribed to it since it changed to a subscription basis but we are delighted that it has reverted to free access. This also makes it easier to publicise and promote to our enquirers.

2) Just a note to let you all know that I really appreciate the great work that you have done with TRIP Database. As a health professional, I have used TRIP to help a lot of patients.....Please let me know if there is anything that I can do to help TRIP remain such a great resource or to help TRIP continue to develop.

Improved search

When Paul Glasziou evaluated the TRIP Database around 12 months ago he reported that of all the searches 50% resulted in a user following any of the links on the site. Another, pessimistic, way of looking at this is that 50% of users found nothing of interest. So, potentially, a poor result but one that reflects the lack of high quality evidence for all clinical scenarios.

However, an anlysis of the first 10 days of searching using the new interface has shown a dramatic improvement in 'link outs'. It has now increased to 73%. One could almost say that the move to the new search algorithm has resulted in a 50% improvement in the search experience.

Friday, September 08, 2006

Pubmed Wizard

I like this site (click here).

I've wanted to introduce similar features to TRIP, such as comments, rank etc.

E-Health Insider

Yesterday I did a brief interview for the above publication. One obvious mistake is them mis-spelling my name - such is life. Still, nice they've taken notice. You can read the article here.

Wednesday, September 06, 2006

Top of the heap

An article from last week's Guardian - Top of the heap.

A very interesting look at user preferences for search results. Apparently, on Google the top result gets 42% of the clickthroughs while the second result gets 11%. As the article points out the top two results account for over half the click throughs. The article proceeds to expand on this phenomenon.

Is this replicated in TRIP? I have no idea and no mechanism, as yet to measure it. I would like to think our users are more 'discerning'. However, it does emphasise the need to continually improve on our search algorithm. Getting the best results at the top. We do a pretty good job but I would not say perfect, so there is still work to do.

When I worked, on a daily basis with GPs (family physicians) one particularly cynical one told me that if he searched he'd pick the first article that supported his hunch. Since hearing that (probably 8+ years ago) I have come to appreciate the problems faced by clinicians searching for clinical literature. They generally haven't got time to do a full literature search, they haven't time to appraise etc. This theme "please no more training on searching and appraising, just answer our questions for us" is still as strong today as it was when ATTRACT started. However, it also reinforces that if s/he does search then they need the 'good stuff' at the top.

How one defines the 'good stuff' needs considering another day.

Tuesday, September 05, 2006

Steve Irwin

Steve's tragic death (click here for more details) has had a dramtic impact on our search records. For the first four and a half days of September the article most visited from TRIP, by a factor of 400%, is the eMedicine article on Stingray Envenomations


Yesterday we added the links to 2880 articles in ganfyd. Ganfyd is a collaborative medical reference by medical professionals and invited non-medical experts. As the site says:

Ganfyd.org was set up to act as a free repository for the vast amount of knowledge that doctors carry around in their heads, both of a factual nature (for example, the causes of Infective endocarditis) and of a practical nature (how to take blood from somebody with difficult veins, or how to deal with an abusive patient, or even how to ensure you don't become dehydrated on your first day as a doctor).

All contributors are volunteers and come from either the UK, Canada, Ireland, Australia or New Zealand. As with any wiki ganfyd is 'work in progress'.

It may not meet the strict definition of being 'evidence based'. However, the value of ganfyd cannot be stated highly enough. Clinical knowledge has an immense value and much of what is written is owned, and controlled, by people/organisations whose first interest is frequently not patient care.

By including ganfyd records in TRIP we hope that there will be a mutual benefit. We will get good content and ganfyd will get increased exposure and therefore volunteers, making the resource even better. If you read a ganfyd article and think it can be improved - make the effort!
Growth predicted in clinical decision support market

Sunday, September 03, 2006

Health Wikis

Two pieces of wiki news:

1) Dean Giustini has started the UBC HealthLib-Wiki (A Knowledge-Base for Health Librarians):

The idea is to provide a forum for best practice for health librarians, and other information professionals around the world.......We believe this open-model will more accurately reflect the "wisdom of the field" and bring significant value to the wiki's knowledge-base.

2) Ganfyd, the clinical wiki, is still going strong and an amazing amount of content is now in there. Due to the distractions of getting the TRIP Database re-designed and free I've neglected my input. However, as well as being able to contribute I have, at last, found a way of importing data from ganfyd into TRIP. It's not a hugely robust way but it should end up covering the vast majority of the content fairly quickly.

Saturday, September 02, 2006

Quick feedback

We've had an awful lot of e-mails thanking us for moving to free-access. One of the first (within 3 hours) was from a UK health librarian:

"This is good news! It's already helped me to find a piece of guidance I hadn't found by any other route."
Protocol for the Quick Clinical study: a randomised controlled trial to assess the impact of an online evidence retrieval system on decision-making in general practice . Although I feel more research is needed in this area I’m not hugely convinced about the meaningfulness of any results from this trial. One can predict that those that used ‘Quick Clinical’ will help change clinician behaviour and support clinical decisions. But, more importantly, it won’t check on patient outcomes.

If we want doctors, nurses etc to incorporate ‘evidence’ into their practice we need them to use it routinely, make it very easy to use and supply 'nuggets of evidence'. I really don’t think that is controversial.

On an unrelated note and following on from earlier blog-entries about USA about citywide wireless access, Norwich, UK is doing the same (click here for BBC News story). I’m still convinced that wireless access to the internet will be widespread (at least in the ‘developed’ nations) in the not too distant future.

Friday, September 01, 2006

Live and trouble free (so far)

The re-indexed material was added at 9am this morning and so far things are working well with no obvious problems.