Energy lab creates bioterrorism tool

An Energy Department laboratory is developing a simulation tool to help public health officials decide how to deal with a bioterrorism attack.The scenario'a mixture of medical research, computer simulations and graphics'shows participants what kinds of information they might receive during such an emergency and the consequences of their subsequent decisions, said Carolyn Pura, a manager in the Exploratory Systems Technology Department at Sandia National Laboratories in Livermore, Calif.The Sandia researchers began developing the decision tool in the spring last year, several months before an unknown assailant mailed anthrax spores to a senator's Capitol Hill office and other locations.Rob Allen, a systems analyst and principal technical staff member at Sandia, said the lab's software engineers coded the tool in Java in conformance with the IEEE 1516 High-Level Architecture standard for distributed simulations.The tool simulates an airborne release of anthrax spores over part of San Francisco during the January flu season. A group leader, called the analyst, gives the public-health participants a background briefing before kicking off the simulation, which presents snapshots of the attack and poses questions about the response for each succeeding day.In the simulation, victims do not start showing symptoms of anthrax for four to seven days. As the participants begin to see simulated reports dribble in from hospitals, they first have to decide whether they're facing widespread flu or anthrax.'In the beginning, those are going to look an awful lot alike,' Pura said.As the scenario progresses, the officials find out whether their decisions cause more or fewer deaths. For example, they can view hospital admission and discharge and coroners' reports, and use them to decide what statements to release and whether to distribute antibiotics to the public.'It's a big help to them to see the implications of their decisions,' Pura said.The participants must decide how often to restock drug distribution sites and what fraction of the population they would be able to treat, Allen said.The simulated disease continues on its course unless its progression is modified by the treatments that are ordered. In a typical scenario, the simulation tracks hundreds of thousands of people, Allen said.The programmers used Census 2000 data to give their faux San Francisco a real-world distribution of people of varying ages and genders'two factors that affect susceptibility to infectious disease.'The results of any particular decision aren't hard-wired into the system,' Allen said. Each person is treated as an object in the object-oriented programming sense, he said.For example, say a decision-maker decides to distribute drugs from the National Pharmaceutical Stockpile. The antibiotics are made available to distribution centers, and the simulation depicts people trying to get to a distribution center or hospital. The decision-makers see how many people can't reach the centers because of geography or traffic jams.When a health officer orders people in certain census tracts to seek treatment, that changes the behavior of the people-objects, Allen said.In developing the simulation, the researchers spent a lot of time worrying about the length of the time steps, Allen said. In real life, a bioterrorism scenario would take up to months to play out. The Sandia tool makes each day just a few minutes of computational run time. The developers eased their concern about putting excess time pressure on the participants by letting them pause the simulation to review data, Allen said.The model runs on 'a couple of fairly standard Intel Pentium 4-class machines over a standard Ethernet connection,' Allen said. Each two-way server has 2-GHz processors and 1G of memory.One server controls what a participating health official can see, and the other server shows the analyst in charge of the simulation everything that goes on, from the release of anthrax spores to the real-time data arriving from hospitals. But the health official gets only delayed reports and doesn't know the details of the attack.'We call the analyst the God's-eye view,' Pura said.Sandia's Visualization Design Center has full-wall screens, 'so you can really immerse yourself in this exercise,' she said. A 33- by 9-foot screen, divided into three panels depicting maps and data, gives a war-room feel.Sandia researchers are exploring ways to make the bioterrorism simulation more flexible and realistic. For example, decisions could be required on an hourly instead of a daily basis, Pura said.'In modeling real life, you're never done,' Allen said. 'There's always more complexity you can add in.'Changing the geographic area of the simulation is not difficult, he added. The code can easily substitute one census data tract for another.It's more challenging to simulate contagious diseases such as smallpox, particularly if the modeled disease affects people's behavior, Allen said. The software has to be capable of simulating public panic and some persons' refusal to seek medical treatment.The first version of the bioterror simulation focuses solely on the role of public health officials, Pura said. Future versions could tailor the exercise for other types of participants. For example, a police chief could get reports about traffic and looting, whereas an elected official might see evacuation plans but not reports about looting.The real-life anthrax cases in 2001 helped the researchers refine the handling of disease progression in humans. The Sandia group has received positive feedback from several public health officials in the Bay area and would like to make the simulation available to other agencies, Pura said.