New computer program can track bioterror

Published: Feb. 3, 2003 at 5:39 PM

BOSTON, Feb. 3 (UPI) -- A new quartet of computer filters soon could help medical authorities spot covert bioterrorism attacks, scientists reported Monday.

By monitoring hospital data, the novel algorithms may not only prevent the explosion of germ warfare-triggered epidemics, but should yield significant peacetime uses also in tracking the spread of infections like West Nile virus.

"The research that we've done can help update all the monitoring systems operational now by today," researcher Ben Reis, a medical informatician at Harvard Medical School in Cambridge, Mass., told United Press International.

Early detection of disease outbreaks is paramount. "Knowing early is often as important as knowing at all," Reis explained.

"If it were an anthrax attack, for example, you only have a few days to give those infected antibiotics and try to limit the spread. If you wait longer, antibiotics are no longer effective. There's a very narrow window of opportunity to nip things in the bud."

Anthrax symptoms are flu-like in the beginning. Although a single doctor might not consider one more patient a day with such symptoms significant, computers scouring hospital data could recognize if the medical center saw 100 patients more than normal with such warning signs.

Such surveillance systems are in place nationally to monitor hospitals for suspicious increases in visit rates. These filters seek to detect outbreaks by comparing daily hospital visit rates with estimates based off historic models for each day. However, in simulated attacks these filters only prove about 30 percent effective at catching outbreaks, Reis explained. The problem is each attack may have a different profile when it comes to how many patients come in over time.

For instance, infections such as anthrax are spread mostly by agents such as powders or aerosols and not by people. Therefore the spread of any outbreak would most likely be a one-time group of people, leading to a steady linear growth of victims from day to day, "say 10 extra daily," Reis said. For extremely infectious agents such as smallpox, which are spread by people, you would see exponential growth -- "three the first day, then 9, then 27, then 81, then 243," he explained.

Reis and his team developed a set of four filters to chart as many kinds of outbreaks as possible. While one looked at daily rates, the other three looked at hospital visits over weeklong spans to take a disease's progression over time into account. These filters were tested with 10 years of data from the Children's Hospital of Boston against a variety of simulated bioterrorist attacks.

In findings published online Feb. 3 in the Proceedings of the National Academy of Sciences, Reis reported, "We were able to see dramatic improvements in both the reliability and timeliness of detection."

Each of the weeklong filters proved more effective than the daily filter. "In some cases, we were even able to more than double the performance," he explained. "If you're using all four combined, you basically get 100 percent detection."

Reis said it should not prove hard to update the software in all existing surveillance systems, "basically on the order of a few days to a few weeks." When in place, medical scientists can also use it to track natural outbreaks of diseases such as West Nile virus or the flu. "In setting it up against bioterrorists, you have public health applications for it as well," Reis said.

The next step is to develop surveillance systems that analyze hospital data over space as well as time, to localize outbreaks to particular areas.

Marc Overhage, a medical informatician at the Regenstrief Institute of Health Care in Indianpolis, said the findings showed great promise.

"The danger right now is that this is an unproven technology," he said. "What's particularly good about this work is that they're approaching it very carefully to test how effective it might be."

Reis agreed with Overhage, noting there are very few cases of bioterrorist attacks to test the system on. "These have not been tested in a real bioterrorist attack, and our hope is that they never get tested that way," Reis said.

(Reported by Charles Choi, UPI Science News, in New York)

© 2003 United Press International, Inc. All Rights Reserved.