Catching the Bug Before It Kills

Pentagon would mine data to fight bio-terrorism

Web sites and a broad sweep of electronic databases are being targeted by the military for a vast surveillance project aimed at detecting incidents of bio-terrorism before its too late.

The project is applauded by people toiling against bio-terrorism, but privacy advocates, while not condemnatory in this early stage of the research, warn against building a system that is overly intrusive.

"The threat of terrorism has replaced the threat of communism as the defining domestic security and national security threat, and weve seen the concern about terrorism being used repeatedly to justify further government surveillance and information collection," said Jim Dempsey, general counsel at the Center for Democracy and Technology and a terrorism expert.

Dempsey, who described the scope of the project as "breathtaking," said that with terrorism, "there are so many potential threats and actors, you are really looking for the classic needle in the haystack, which leads to these kinds of proposals, of Lets watch everything everywhere. "

The idea behind the so-called Bio-Surveillance System, which is sponsored by the Defense Advanced Research Projects Agency, is to build a network linking disparate sources of health-related information that will alert officials if terrorists have clandestinely used biological weapons such as anthrax spores in an attack.

Anthrax infections, for example, show symptoms of the common cold before more serious pneumonia symptoms appear, followed ultimately by death, anywhere from several days to weeks later. In the project, officials would search pharmacy, grocery store, and school and civil service absenteeism databases, as well as information retrieved from or queried on health-related Web sites and reports from hospitals. If data reveals sudden, severe spikes in reference to the common cold, officials could use the information to test for anthrax and, if the virus is found, start prescribing antibiotics.

Time is of the essence in dealing with biological warfare. The attack may begin quietly — with a terrorist spraying biological agents into the air with aerosol equipment — but could suddenly turn disastrous within days or weeks. People who breathe anthrax spores can be successfully treated with antibiotics, but the drugs must be administered within roughly eight hours, "or forget it, youre history," said Jennie Hunter-Cevera, president of the University of Marylands Biotechnology Institute.

Hunter-Cevera described building the bio-surveillance system as "extremely challenging."

"How do you get all of these databases to talk to each other?" she asked. "That will be a big problem. Thats where the computational geniuses will have a lot of fun."

The $24 million, five-year DARPA project will serve as a prototype system for "protecting [Department of Defense] military and civilian personnel from bio-warfare attack," the DARPA announcement said.

DARPA officials were not available for comment, said DARPA spokeswoman Jan Walker.

The Dec. 27, 2000, DARPA program announcement referred to an earlier DARPA study that demonstrated "it was possible to identify an abnormal health event caused by a terrorist release of a pathogenic agent several days before traditional health surveillance methods. The project mined grocery store, pharmacy and absentee databases, and gained access to health-care records, after obtaining and receiving voluntary permission for such access."

The announcement states that the envisioned network would "gather and integrate information from nontraditional health information sources (e.g., school absentee reports, veterinary reports, over-the-counter pharmacy sales, Web page inquiries) and state-of-the-art biosensors for a city with a large DOD population."

Its unclear what the final system would look like. The agency is aware of the potential for privacy problems. In its proposal it says the "monitoring of human health databases must conform to appropriate privacy protection regulations."

DARPA is now asking for formal proposals from industry and universities. Sometime this year, the agency will select contract victors.

Sifting Through Data

In essence, the project represents a vast exercise in data mining — or sifting through data in search of patterns and anomalies — and its precisely what the government should be pursuing, said Jerome Hauer, the former director of the New York City Mayors Office of Emergency Management. Hauer was at the helm during the West Nile virus outbreak last year. New York City, he said, is already working on a system similar to what DARPA is investigating.

"Data mining does work," Hauer said. "Ive been a huge advocate. You have a nonintrusive way of getting at data that is needed to detect patterns."

The alternative, called syndromic surveillance, requires physicians in emergency rooms and elsewhere to fill out extra forms relating to infections and disease in patients. This method, he said, "does not work" and is "very dangerous," because doctors are too busy with patients to attend to more paperwork. As a result, information about disease is not logged, and disease outbreaks can slip through the cracks.

Electronic surveillance of records represents the future of tracking disease outbreaks in general, not just incidents of terrorist attacks, Hauer said.

But first, it will have to be demonstrated to hospitals that patient confidentiality will be revered, and that officials wont use patient information for other purposes, such as tracking how well individual hospitals are performing.

"Hospitals are concerned that their morbidity and mortality data will be reviewed, a Big Brother syndrome, where the data will be used for purposes other than surveillance," Hauer said.

Echoing Hauer, D.A. Henderson, director at Johns Hopkins Universitys Center for Civilian Biodefense Studies and an international luminary on biological warfare, described the current system as "weak" and championed DARPAs efforts.

"The intent here is: Are there ways by which we might get more advanced warning than we now have?" he said. "We all are deeply concerned. What can we do to strengthen the alert system, to make it better?

"If one could tap into some data that is routinely collected, it could be useful," he added. "The only way we will find out if such data is available is to experiment with it, as they suggest here."

One of many struggles researchers will face is calibrating the system to work somewhere in between sending off alarms every time somebody gets a cold and alerting officials well after an epidemic has been established, said Jonathan Tucker, director of the chemical and biological weapons nonproliferation program at the Monterey Institute of International Studies Washington, D.C., office.

"How do you know that it is sufficiently worrisome to do something about it?" Tucker asked. "If the system has too many false positives, then people will disregard it after awhile. What they are looking for is for people to develop new algorithms to extract signals from a lot of noise. Thats challenging."