Will health databases spot bioterror attacks?

Will health databases spot bioterror attacks?

The challenge in creating a health alert network is finding a way to analyze and disburse data about anomalies as they occur, New York City health official Dr. Farzad Mostashari says.

New York City fine-tunes its warning network

After Sept. 11 and the anthrax scare, health officials nationwide began clamoring for a better data infrastructure so they could view the same health data that the first responders in emergency rooms and ambulances were seeing.

The Centers for Disease Control and Prevention provides a secure e-mail and fax alert system called the Health Alert Network, which Homeland Security director Tom Ridge said could be the basis for a national medical intelligence database [GCN, Jan. 21, Page 31].

But public health officials agree that the existing infrastructure needs work. They are struggling to develop an information infrastructure that will link emergency rooms, physicians and public health departments, and serve as an early warning system for public health crises and bioterrorism, said Elin Gursky, a senior fellow for civilian biodefense strategies at Johns Hopkins Center.

One such early warning system is already in use by the New York City Health Department. Epidemiologist Dr. Farzad Mostashari oversees a citywide health surveillance network that links emergency rooms, 911 dispatch facilities and pharmacies.

New York's system can detect phenomena such as a widespread outbreak of nonspecific symptoms'coughs, sore throats and fevers, for example.

Ambulance chasing

A Microsoft SQL Server database at the Health Department stores ambulance dispatch data. It arrives automatically by File Transfer Protocol as delimited text files over CityNet, New York's data communications backbone.

"There's no human intervention," Mostashari said. "We don't wan to impose any data burdens on anybody."

The database receives about 3,000 records a day. Health Department officials study the records using a data analysis program from SAS Institute Inc. of Cary, N.C. Next, a predictive model compares the number of dispatch calls received in a 24-hour period with the number expected based on call histories.

The department is working on a Web interface for the system so that health officials throughout the city can see the same data simultaneously. For now, Mostashari and his team produce graphs and e-mail them to public health officials.

After Sept. 11, CDC sent 50 officers'mostly physician'from its Epidemic Intelligence Service to 15 emergency rooms throughout the city to collect patient data. These disease detectives worked 24 hours a day in eight-hour shifts, Mostashari said.

After the immediate crises passed, city health officials developed a system for sending emergency room data via e-mail or FTP to the health department database.

"We're moving towards the same kind of system with city employee sick logs and also pharmacy sales of antidiarrheal products," Mostashari said. "If Kaopectate starts flying off the shelf, that could be cause for alarm."

Mostashari's group also has adapted freeware to detect clusters of cancer cases. Called Spatial Scan Statistic, the software is downloadable for free from the National Cancer Institute at www.nci.nih.gov/prevention/bb/satscan.html.

Developed by Martin Kulldorff of the University of Connecticut, it tracks data across time and location.

The Health Department runs the MS-DOS program on a Compaq Computer Corp. PC with 2G of RAM, but "you don't need that much firepower," Mostashari said.

When a cluster of symptoms appears, he maps it using ArcView 8.0 from Environmental Systems Research Institute of Redlands, Calif.

Cluster found

A cluster of emergency room visits and ambulance calls fro respiratory distress did occur in Queens one day after the crash of American Airlines Flight 587 on Nov. 12, Mostashari said.

The cause turned out to be smoke inhalation from the crash and not a bioterrorist attack. But it proved the system could alert officials to a time-space cluster of health anomalies.

Mostashari said the events of the past few months have added urgency to his work. "We used to think the big challenge is getting the data and analyzing it," he said. "But once you detect and increase in symptoms, you have to figure out if it is chance, viral infections or bioterrorism. So we're developing ways to look at data within 10 to 15 hours.

About the Author

Trudy Walsh is a senior writer for GCN.

Stay Connected

Sign up for our newsletter.

I agree to this site's Privacy Policy.