 
|
|
|
 |
 |
 |
 |
 |
 |
 |
 |
Wednesday, February 06, 2002
By Christopher Snowbeck, Post-Gazette Staff Writer
The computer system that President Bush learned about yesterday at the University of Pittsburgh monitors the symptoms of patients flowing through regional hospitals in hopes of detecting a bioterrorism incident.
Noting that promise, Bush yesterday likened the surveillance system to the DEW line -- the Distant Early Warning line of radar installations erected across northern Canada and Alaska to provide early warning of a missile attack from the Soviet Union.
Since its genesis in 1999, Pittsburgh's DEW line -- called the Real-time Outbreak and Disease Surveillance System -- has detected little more than spikes in the number of diarrhea cases and flu, said Dr. Michael Wagner, a Pitt professor and director of the surveillance system.
But every time the system detects something, it immediately pages a public health doctor, who can then notify the county Health Department. That sort of warning could be crucial if a bioterrorism attack occurred, researchers say.
"If you suddenly get 20 or 30 extra people coming in from across the region complaining of respiratory problems ... and if you suddenly see that they're all from one specific location or all from one occupation, then that raises a really big red flag," said Andrew Moore, a computer science professor at Carnegie Mellon University.
Moore and Wagner contribute work to the Biomedical Security Institute, a joint venture of Pitt and CMU that was a focus of Bush's visit. The institute has received several million dollars in government funding.
Wagner said the surveillance system analyzes 1,200 patient visits from 17 hospitals in Western Pennsylvania each day, looking for increases in the number of cases of flu-like symptoms, respiratory illnesses, diarrhea, rash, paralysis, encephalitis and hemorrhage. The information does not identify individual patients.
The computer system gets some of its data from hospital laboratories, but another key source is the information patients provide when they register at an emergency room. That information isn't always a perfect record of how a patient is feeling, Wagner said, but it still can be useful.
"There are typographical errors, spelling errors, unique abbreviations, but our natural language processing techniques can get a lot of signal out of that noise," Wagner said. "When you're trying to find an outbreak, it's not necessary to accurately diagnose every single case."