Dianette

I like to analyse what people look at via TRIP. Given the Q&A services we run (ATTRACT and NLH Q&A service) I decided to see what articles users follow via TRIP - so restricted the analysis to just those two services:

- What is the duration of treatment when adding clopidogrel to routine aspirin short-term post CABG and post coronary artery stenting in IHD.
- Are influenza vaccinations effective in older people?
- *What is the length of time a patient can remain on Dianette provided that there are no contraindications?
- What are the recent guidelines/evidence regarding the treatment of prostate cancer?
- What treatments are available for someone with a ruptured Bakers cyst?
- *What is the reason for stopping dianette a few months after skin condition resolve?
- Can the drug amlodipine cause atrial fibrillation?
- What causes a yellow tongue, and what is the best available treatment?
- *What are the risks of using Dianette as an oral contraceptive, over and above the risks associated with a 3rd generation oral contraceptive pill?
- What skin preparation if any is recommended when taking blood or giving an injection?
- *What therapies are available for women with PCOS with hirsutism other than Dianette?
- *What is the recommended length of time a woman should be on Diannette for acne, should any blds be checked, should there be any break in-between treatments?

What immeadiately struck me was that of the top 12 questions 5 related to Dianette. What does that tell us about Dianette?

Top resources

For the first 3 months of 2006 the following are the top ten most popular 'destinations':

eMedicine
Clinical Answers
DARE
Guidelines.gov
ATTRACT
NHS EED
Cochrane
HTA
PRODIGY
Bandolier

Improving the search algorithm

The new search algorithm is still performing well and I actually use it as my default search for answering current clinical questions. But, even before this is released, I'm thinking of the next set of improvements, I have two in mind:

1) Popularity. Giving papers an additional score based on how often they have been viewed. In effect our users choices will help decide the ranking. Suggested by Dean Giustini.

2) Location. I've been toying with this idea for a while and saw a nice blog entry on the subject earlier today. If you're searching from the Australia you'd probably want to see Australian guidelines before UK or Canadian guidelines. So the general idea is to see where the searcher comes from and give their countries guidelines a scoring boost.

So, the new algorithm will be released (when we go free in September) and then we can try and incorporate the above improvements sometime early 2007.

TRIP and Google Health

I had an interesting chat with Ben from the NHS eLibraries blog. He suggested trying to tie-in with Google, as part of their Google Health 'initiative'. One option in Google Health is 'For health professionals' - TRIP could form the backbone of a new tab 'Evidence-based medicine'.

Q&A Feedback

We frequently get very positive praise via our Q&A feedback mechanism. Every now and then we get comments which stand out (for one reason or another):

"Very impressed, very useful. I was however, quite alarmed at some of the other questions posted i.e people asking for guidance on issues which you would have hoped as health professionals they would be extremely familiar with, I suppose you could see that as a good thing, as if this service didn't exist, presumably they would stay uninformed and at least they are asking questions which will improve their practice."

Test search - results of first formal evaluation

In a trial of 8 'proper' clinical questions we rated which search located results best:

Current live system - 1
New test system - 4
No difference - 3

I think the first round goes to the new test system.

ClinicalQuestions Collection

Not 100% sure why the authors chose to ignore convention and remove the space between 'clinical' and 'questions'. I'm sure it seemed a good idea at the time!

However, that doesn't detract from the site (http://clinques.nlm.nih.gov/JitSearch.html) which is a collection on 4,500+ clinical questions. No answers, just the questions. Not sure why I like it so much. It'll be useful in testing the new search algorithm!

When testing a search appliance it seems sensible to try 'real' searches. We've received a question about SLE (Systemic Lupus Erythematosus) and statins - should all patients with SLE receive statins.

I entered the search 'sle AND statins' and the first result was what I was looking for. On the current 'live' system the same search would return one result. However, the user would be required to click on the 'Clinical Questions' link. It's only one click but the new search is showing great promise..

Test search

At last, the final installments of the potential new TRIP algorithm have been delivered! We've got 35 volunteers testing it alongside a number of staff...

What is TRIP for?

The testing of the new search algorithm is having implications far beyond those expected. The current system of 'title' and 'title and text' distinction allows for a very accurate location of material. However, the use of these features can be cumbersome. The abolition of 'title'/'title and text' distinction creates a compromise.

Perhaps the solution is to adopt the new algorithm, which will give us a great new challenge of always trying to improve it. We could then create an advanced search page for our 'advanced' users who take advantage of more of the current search features. The advanced search could look like the Google advanced search with fields for 'title' words and 'title and text' words....

But, going back to the title of this post, what is TRIP for? Our general view is that its a tool to allow clinicians to rapidly identify high-quality research evidence with which to base a clinical decision on. But that's our view, our users might have different ideas.

Search algorithm test

Before going free-access in September we're experimenting with a new search algorithm, which was delivered 2 days ago. We've got around 25 volunteers to play with it to see how they think it works. It's difficult for me to be objective but my initial trials are promising. A few weird results, but hopefully these are small issues relative to the improvements.

It's made me appreciate the 'work in progress' all search systems must have in order to improve. I feel we're at the start of a very long journey - exciting times.

Google Health - arrived?

Been doing some work around google health and yesterday I searched for 'tonsillitis' and got the image on the left, this didn't work in Canada! Just did a search for 'prostate cancer' and got the second image - note the refine function. Interestingly, or perhaps disappointingly, it didn't give the same results with 'prostatic carcinoma' - perhaps synonyms will be plugged later.

Although is this google health? Wouldn't that be stand alone like google finance - something like http://health.google.com?

One of the refine functions is 'For health-professionals' which when clicked gives the following search syntax 'more:for_health_professionals'. Alas, for tonsillitis the first page of results are for patient information sites! For prostate cancer it did slightly better - but not hugely!

Popular Papers - April 2006

Below are the top ten papers reached via the TRIP Database for April 2006.

1) Acute Pain Management (National Health and Medical Research Council)
2) Geriatric Care and Treatment (SBU)
3) National clinical guidelines for stroke 2nd edition (Royal College of Physicians)
4) A review of the effectiveness and appropriateness of peer-delivered health promotion interventions for young people (eppi)
5) Anemia (eMedicine)
6) Antenatal care Routine care for the healthy pregnant woman (RCOG/NICE)
7) Management of chronic pain in adults (NHS Quality Improvement Scotland)
8) COPD (PRODIGY)
9) Contraception (PRODIGY)
10) COPD (NICE)

Microsoft QnA

Microsoft discuss their new QnA service (click here).

Recommendations from HubMed (click here)

EBMsources evaluate the TRIP Database

EBMsources have evaluated the TRIP Database (click here for evaluation). TRIP did reasonably well, scoring 87/100.

I've been working with EBMsources to try and improve their scoring system, a new version is currently being validated.

TRIP lost score for not being bilingual - betraying EBMsources origins - French Canadian, making bilingualism more of an issue. It also losts points for not being free (soon to be fixed) as well as a few other bits and bobs.

Google health

USA Today reports on the possible release of Google Health next week.