Trip Database Blog

I’ve been involved in clinical uncertainties for many years.  I had the pleasure of helping create the DUETs database (UK Database of Uncertainties about the Effects of Treatments).  Around the same time Trip released the Tag Cloud of Clinical Uncertainty which was a great experiment. In all aspects of my professional life I like to highlight the importance of clinical uncertainty.  Although the term often unnerves people, I think they feel threatened by the notion.  But, it could be worse, the likes of Iain Chalmers and Muir Gray have often used the phrase clinical ignorance – which is far harsher.

My interest in uncertainties stems from my desire to improve the research procurement process.  The main drive with Trip is to help clinicians answer their clinical questions.  Without knowing the gaps in their knowledge and the evidence base how can you procure suitable primary research or even secondary, evidence synthesis?  Myself and my teams have answered over 10,000 clinical questions so we know how frequently the research base is lacking or not focused on clinical care.  So things need to improve and I think recording uncertainties is a great way to help.

I raise all this as, in my business planning, I’ve spoken to one of the UK’s largest research ‘agency’ and they are keen to work with Trip to better understand users questions and gaps in the evidence base.  So, for me, the answer is to try and capture the clinical questions our users have and when Trip has let them down.  I have a few ideas around this, including creating a PICO+ tool; a step-by-step tool to allow users to easily answer their clinical questions.  The user would start by adding the full question and then we would guide them through the PICO steps (e.g. what is the population, what is the intervention).  At the end they can tell us if their question has been answered.  If not, it’s got a good chance of being a clinical uncertainty. 

Seems like a plan!

Trip prides itself as a great tool for answering clinical questions. Over 80% of users find the information they need, all or most of the time.  But that’s still not perfect and one idea I keep coming back to is the ‘answer engine’.  The wonderful Muir Gray said, in relation to finding evidence, that three clicks was two clicks too many.  So, the challenge is, is there a way of getting answers based on a single mouse click?

The answer (or perhaps ‘My answer’) is the answer engine.

This would involve a system to try and understand the search and display a suitable answer.  So, if a user searches for minocycline and acne we can be fairly confident that they’re interested to know if minocycline is effective in treating acne.  Therefore, we could drop in the following answer:

Minocycline is an effective treatment for moderate to moderately-severe inflammatory acne vulgaris, but there is still no evidence that it is superior to other commonly-used therapies. This review found no reliable evidence to justify the reinstatement of its first-line use, even though the price-differential is less than it was 10 years ago. Concerns remain about its safety compared to other tetracyclines.

This has been taken from a recent Cochrane Systematic Review.  The normal search results would appear beneath the ‘answer’.  The user gets a great answer in one click.

There are a few issues with the above and one is scalability.  Parts of this can be automated but much of it will be manual.  Also, relying of sources – such as Cochrane – means it’s led by the evidence producers not the user.  So, the challenge is to have users supply answers.  Which leads to the Journal of Clinical Q&A.

The idea is to set up a brand new journal dedicated to answering real clinical questions.  Based, roughly on the BestBETs site it will follow a similar structure (e.g. Steroids in lateral epicondylitis).  The clinical bottom line will be pulled through into Trip to act as the ‘answer’ and then users can click to see the full article.

Peer-review is problematic on many levels and Richard Smith (former editor of the BMJ) has frequently criticises the current peer-review process (e.g. A woeful tale of the uselessness of peer review and Scrap peer review and beware of “top journals”).  But how can we improve on it?  I’m not sure we can but I’m open to help!  My current proposal is as follows:

• Each answer will be reviewed by an in-house team, a sanity check.  Those that seem reasonable will be released into the answer engine as an ‘answer pending approval’
• We would then ask the wider Trip community to read and rate the answer.  This would borrow from the F1000 approach which uses three classes: Approved, Approved with Reservations and Not Approved.  An article will be considered published when it reaches a certain approval threshold. Note, the F1000 approach is not without criticism (e.g. PubMed and F1000 Research — Unclear Standards Applied Unevenly), hence writing this article with the hope of obtaining help.

Trip would have a good answer and the person who uploaded it will obtain a citation.  The plan is to start slowly, see how it develops but the longer-term view would be to see the answers appear in Medline/PubMed – as currently happens with a large number of the articles in BestBETs.

The above is an idea, a work in progress.  I think there is every chance this can become a reality but a little help in refining the concept would be really good.