Trip Database Blog

A couple of interesting snippets….

First, EBM and Fascism. An article that appeared in the International Journal of Evidence-Based Healthcare ‘Deconstructing the evidence-based discourse in health sciences: truth, power and fascism‘….

Second, the ‘official’ Google blog has reported on the ‘related articles’ feature on Google Scholar. I’ve always liked the related article feature on PubMed and the GS version appears to work as well. Whether it would be as useful on TRIP is another matter. Related article style features have traditionally relied on latent semantic indexing (see wikipedia article for more detail). This approach can also be used for all sorts of other useful semantic ‘tricks’ such as synonym analysis.

I often wondered if the related article feature could be used to keep the Q&A answers up to date. In other words, use the LSI technique to look for new articles in PubMed published since we published our answer. The idea seems plausible but implementing it is a significant issue and quite possibly not worth the potential gain…

As mentioned in previous posts we get lots of positive comments from users of the service. Currently we are experiencing high volumes in the NLH Q&A Service and have had to, temporarily, stop taking new questions. This has resulted in a few e-mails, one in particular caught my eye:

“Pity that you are inundated with questions! This is one fo the few worthwhile services we as GPs have, and has a TREMENDOUS IMPACT ON CLINICAL PRACTICE for busy GPs. It wouold be shame to be constrained by staff shortage. I have disseminated the usefulness in many meetings. Perhaps – we should stop informing GPs!!”

While away the web people have been busy beavering away and getting closer to the finished article. A few graphics and some text need dropping in. I was confident enough to send test URLs to two colleagues whose opinion matters a great deal to me. Fortunately, they got back to me with useful constructive criticism and a general positivity about the site.

Another glimpse, this time of the results page with a new feature the ‘drug box’ which highlights drug interactions, warnings etc. As well as the ‘standard’ results.

The new site is nearly, functionally ready (some design needs dropping in), with my web-people working for a few more hours before I head off on holiday. In other words they have a few more hours to correct one problem. Basically I had the algorithm set-up pretty well on the test harness. However, now we’re using the algorithm on the new site with updated content something’s gone wrong. I’m hoping it’s not a big deal.

When they’ve fixed that I can relax and enjoy the holiday. I’ll send the test URL to a number of tame users – mainly a load of trusted Q&A people from ATTRACT and the NLH Q&A Service as well as the TRIP clinical director (whose also a general practitioner and TRIP co-owner). They’ll give it some heavy use over the week. Then, on my return, I can pass on any issues to my web-people. Hopefully giving them 2 weeks to finalise any changes.

Sounds so simple……

Mixed results for this test of the city-wide wi-fi roll out in Mountain View, California – supplied by Google. As the author of this piece on the BBC News website said:

“After several hours grilling Google’s Mountain View wi-fi network I realised both the power of the service, but also the present-day limitations and youth of the technology.

While the service was ubiquitous throughout the city, it’s not as reliable, as fast, or as easy to use, as my home internet connection or my cell phone. Not yet anyway.

Start-up companies and major manufacturers are working on all these issues. They just take time”

As wi-fi takes off and expands we really must invest in optimising TRIP for searching on laptops, PDAs, mobile phone……

As TRIP moves to free access we are removing our classic ‘title’ or ‘title and text’ search distinction. When TRIP initially started it was a title word search only. Then we started spidering content and we added a text search as well. However, our various search analyses have shown this method to be problematic. For instance if you wanted to search for ‘asthma and steroids’ you would need to search for ‘asthma’ as a ‘title’ word, ‘steroids’ as a ‘title and text’ word and then combine them. This equates to three steps (two searches and then a combined search). As Dean highlighted in his recent blog entry – searchers are using more search terms per query and expressing more complex information needs.

Google, the main general search tool, allows users to add multiple terms and normally returns very good result sets. Google is setting the standard for how users search.

So what’s the coincidence?

The NLH have announced a new search format – adopting our current (but soon to be moved to advanced search) search method – of ‘title’ and ‘title and text’ distinction – see screen shot. I can’t help feeling flattered with them adopting TRIP’s search method. However, I feel its potentially a wrong move…..

I received the test version of the site on Friday and – no real content to search – the indexing of content had failed! We’ve still got problems with the indexing but these should be fixed by the end of the day. Other problems include the test server being woefully underpowered to test the site, no advanced search, no content from the big five journals (NEJM, JAMA, Lancet, BMJ, Annals of Internal Medicine). Irrespective of these issues we should still be live for the start of September!

As means of a progress report I’ve posted the above picture. Missing text and graphics – but no ‘title’ and ‘title and text’, separate tabs for ‘Medical images’ and ‘Patient information’ and a few other ‘bits and bobs’. Roll on September……

A BioMed Central article looks at Cochrane systematic reviews and whether updating them improves the reporting and methodological quality. The biggest surprise to me was the comment:

“The overall quality of Cochrane SRs is fair-to-good.”

Only fair-to-good…..