Bloomington Herald-Times Articles
May 15, 2009
IU researchers' predictions about H1N1 flu's spread on target
By Mike Leonard
May 14, 2009
One of Alessandro Vespignani's research specialties at Indiana University is modeling the worst-case scenarios for the spread of disease in epidemic conditions.
For the past 5-6 years, the work has been academic in nature, he acknowledged this week. With the outbreak H1N1 influenza, also known as Mexican flu or swine flu, Vespignani's team in the School of Informatics has been like a combat-trained unit thrown into its first firefight.
"In general, we work on scenarios. This time we've been called into the real world. It's a bit different," the IU professor said with a chuckle. "The reaction time is much more limited. Everything is hectic."
Vespignani's lab is one of only a handful in the world trying to predict the geographic spread of the disease and the number of people likely to be infected. In the early going, the work has been very successful, with Vespignani's group and a similar research team at Northwestern coming up with similar and, relatively speaking, highly accurate predictions, using different models.
"We've had remarkably good projections for the hot spots geographically," the IU researcher said. "The numbers have been very good, too."
The IU group has used transportation statistics as the basis of its modeling. Driving patterns, air travel patterns, food delivery routes and other transportation data figure into complex mathematical calculations that are recalibrated every two days.
The Northwestern group has received a certain amount of attention because its model includes data from the "Where's George" Web site, on which people report the movement of dollar bills. Though the site was established by a computer programmer more interested in creating a databased interactive Web site, it's turned into a valuable data base that's tracked more than 100 million dollar bills over nearly a decade.
The IU and Northwestern flu projections remain very similar in their predictions, despite the differences in the data and algorithms used. "At first, there were models projecting more than 100,000 cases in the first two weeks and we were like, wow, wow, wow. Either we are very wrong, or they are very wrong," Vespignani said. "Fortunately, we were not the ones who were very wrong."
The IU researcher is quick to point out that the mathematical modeling is just one component in the research into the flu epidemic that many feared would turn into a global catastrophe. "There are so many variables," he said. "The fact that the disease has been very mild to this point means there is no huge containment effort, and that could mean big things down the road."
The U.S. is moving from spring into summer, when flu incidence goes down. At the same time, South America is moving into its cold weather months, so the potential for the rapid spread in the southern hemisphere is unknown, but potentially great.
So far, the outbreak of the disease has been less than many expected -- currently at about 2,600 cases in the U.S. -- the numbers are going up. Vespignani said his latest projection is for the U.S. rate to climb to 20,000 cases by the end of May. And while that might sound like a lot, you have to recognize that's still a very small number in a country of 300 million people.
"I tell people, this really is a lot like forecasting the weather in that you really can't make very good projections very far out," he said. "I can't tell if it will rain 155 days from now, that would be silly, but if I were a weather forecaster, I could give you some pretty good ideas of what will happen in the next week."
As it is, Vespignani and his team are crunching numbers at a rate that would have been impossible even 10 years ago. "Here we are very lucky," the research leader said. "When this thing started, IU gave us the use of its Big Red supercomputer, and that makes all of the difference in the world. We can run overnight computations and do the analysis in the morning. We simply could not do that without that kind of computing power."
Although Vespignani emphasizes that fresh data and short-term predictions are going to be the most accurate, he said he tends to agree with those who say that, before this strain of flu runs its course, it will likely have stricken a third of the world's population. "I think in this country we will see a rise again next winter," he said. "As for the whole thing, it may take a year, a year and a half to play out."
The IU team is named GLEaM, which is the acronym for the Global Epidemic and Mobility modeler. Its Web site, which includes national and international modeling and data, can be found at http://www.gleamviz.org/.