Something wonderful happened today. We officially set in motion the road to free access to the TRIP Database. For various contractual reasons free access won’t start until 1st September 2006. However, we’re hoping to publish a ‘beta’ version of the free site ASAP (probably 3-4 months). The free version will actually be an enhancement on the current version – the next step in the evolution of the site. But, as mentioned, contractually we’re not able to push anything ‘official’ out until September.
When answering clinical questions we give a range of answers. Some are answered very well and others less well. If the evidence is there – great, if not the answers are less complete. Already, today, we’ve answered two questions at the different end of the spectrum.
The ‘good’ answer “Can you prescribe HRT for someone with focal migraines. If not what can be tried for her flushes? Also would HRT be advisable to protect her bones?” . This uses a number of evidence sources, principally PRODIGY. This answer links a number of separate sources of good evidence to ‘weave’ a ‘good’ answer.
The ‘bad’ answer “Can raised uric acid or gout could cause tinnitus?” We only found one reference in Medline and that was from Medical Hypotheses. So as we point our in our answer – it’s a hypothesis not evidence.
On one level the latter answer is bad. It’s bad as it doesn’t resolve the uncertainty for the GP. However, it’s an appropriate answer as that is the only evidence.
“Geriatric Care and Treatment: A systematic compilation of existing scientific literature” by The Swedish Council on Technology Assessment in Health Care was the most viewed paper, from the TRIP Database, in 2005.
Second was “National clinical guidelines for stroke Second edition” by the Royal College of Physicians and third, the NICE guideline on dyspepsia.
More stats… Around 50% of all our searches result in people going off to the filtered Medline articles. Around 5% of the time users personalise their Medline results by restricting to key journals in a specific area. TRIPwire (our search refinement tool) is used just under 2% of the time.
What do those figures suggest? Our users obviously really like the auto-searching of Medline. However, other features such as specialisation and TRIPwire are more ‘luxurious’!!
Currently, on the TRIP Database you do a search and the results are returned hierarchically. Therefore, you need an additional click to view results in, say, ‘Evidence-based synopses’. Perhaps a better way would be to return a page of results straight-away and then if people want to select only a single category – let them do that. Some thought needed but perhaps merge all the results into one and devise a weighting mechanism to ensure the ‘better’ material is at the top. But what might the attributes of ‘better’ be? Type of publication, year of publication….
One of my regular blogs is Dean Giustini’s Google Scholar Blog. It was seeing Dean’s editorial in the BMJ (http://bmj.bmjjournals.com/cgi/content/full/331/7531/1487) that got me thinking about where the TRIP Database might fit in. Behind a subscription login/firewall is no place. As one of my colleagues said ‘no-one pays for search’. Not technically true as we have reasonable subscription levels – but I get the point. Do we continue to stay in our niche getting searched 50,000 times per month? Ordo we change to being free-access? The subscription allowed us to grow significantly in the 2-3 years we’ve been charging. However, our latest upgrade (Dec 2005) has increased my frustration of it ‘hiding’ behind a firewall. Biased though I may be – it’s just too good to hide….
So thank you to Dean for getting me thinking.
More stats to celebrate the new blog! The top search terms used in TRIP are:
- otitis media
These terms are so broad and return lots of results – arguably too many to sift through. It makes the decision to add a prominent ‘refine search’ box appear sensible.
Mis-spelling. One of the new features on the TRIP Database was not a mis-spelling feature (we’ve had one for years) but a counter to accurately measure how often people use it. So we feel it’s a pretty good proxy for number of searches mis-spelt. For the first 3 weeks of February nearly 5% of all searches were mis-spelt.