2008年1月5日 星期六

search out hospital superbugs

"Researchers in the UK plan to use Google's PageRank algorithm to find how super-bugs like MRSA sprea d甲氧西林耐药金黄色葡萄球菌 in a hospital setting. Previous studies have discovered how particular objects, like doctors' neckties, can harbor infection, but little is known about the network routes by which bugs spread. Mathematician Simon Shepherd plans to build a matrix describing all interactions between people and objects in a hospital ward, based on observing normal daily activity."

メチシリン耐性黄色ブドウ球菌(methicillin-resistant Staphylococcus aureus: MRSA)


search out hospital superbugs

Google tool could search out hospital superbugs

  • 12:35 04 January 2008
  • NewScientist.com news service
  • Tom Simonite


Transmission of hospital acquired infections like the "superbug" MRSA could be cut using the method that Google uses to rank search results, say UK researchers.

Crunching data from wards using Google's PageRank algorithm could help focus preventative measures more accurately by identifying key routes of infection and transmission, they say.

PageRank judges the relative importance of a web page in search results based on how many other pages link to it.

The first response of policymakers faced with rising incidence of hospital-acquired infections, including super-bugs like MRSA (methicillin-resistant Staphylococcus aureus), has been to enforce stricter washing of hands to stem the spread.

But while that has been shown to help, infections in the UK at least are not falling, says Clive Beggs, who leads a research group studying prevention and control of hospital-acquired infections at Bradford University, UK.

Infection networks

"The question is, how do bugs get from A to B?" says Beggs. "We don't really know that much about the epidemiology of these infections." The extent to which infections are carried by hands, are airborne, or travel by other routes is largely unknown, he adds.

While studies have taken samples to discover how particular objects, like doctors' neckties, can harbour infection, little is known about the network routes by which bugs spread across patients, staff, and the hospital environment.

Beggs' colleague, mathematician Simon Shepherd, thinks the PageRank algorithm behind Google's search engine can help, by ranking routes of infection in the same way it ranks search results.

"Something isn't working," Shepherd told New Scientist. "The hand-borne route is the major route, but there are others and we need to know what they are."

"Our new model is based very much on the way Google has achieved number one status among search engines," he explains. "When [Google's] spiders crawl the web they build up a connectivity matrix of links between pages."

Node search

Shepherd's idea is to build a similar matrix describing all interactions between people and objects in a hospital ward, based on observing normal daily activity.

"Obviously nurses move among patients and that can spread infection, but they also touch light switches and lots of other surfaces too," he says., "If you observe a network of all those interactions you can build a matrix of which nodes in the network are in contact with which other nodes."

Shepherd has started testing the technique using data gathered for another study. "We sussed out in one ward that the chief node was a light switch," he says. "It could potentially distribute infection to the rest of the ward very quickly."

The PageRank technique will become most useful, however, after the completion of studies estimating the extent to which different interactions transfer bacteria.

Combining that information with the strength of different interactions within a ward makes it possible to calculate which ties to cut – by, perhaps, tougher cleaning – to maximally disrupt the network and cut infections.

"Ultimately, we would like to produce a software tool so managers of wards can carry out the analysis for themselves," says Shepherd.

Shepherd presented preliminary results of the technique at the International Conference on Internet Technologies and Applications, September 2007

沒有留言: