In addition, analysis of laboratory surveillance data from the WHO and CDC showed that in the 2001-2002 flu season, it took 53 days for flu to spread across the U.S., 60 percent longer than the average time of 33 days.
By contrast, in France, where flight restrictions were not imposed, there was no delay in flu activity during the 2001-2002 flu season.
Brownstein and Mandl, both also of Harvard Medical School, then compared their data on flu spread with monthly estimates of passengers on domestic and international flights, obtained from the U.S. Department of Transportation.
For domestic flights, airline volume in November 2001 was an especially strong predictor of flu spread. With the Thanksgiving holiday, November is typically one of the busiest travel months of the year, but in 2001, many people kept close to home or sought other forms of travel.
"Thanksgiving is when new flu strains often spread across the country," Brownstein notes.
For international flights, volume during September most strongly predicted the U.S. flu peak, suggesting that September is a key month for introduction of foreign flu strains. In September 2001, international flights fell 27 percent (from 4.9 to 3.5 million passengers), and peak flu mortality that winter was delayed by two weeks. In 2002, international travel was still down by 10 percent, and the U.S. peak was again delayed.
"When we first looked at our data we noticed that the 2001-2002 flu season was highly aberrant," Mandl recounts. "At first we thought it was a problem with the data, but then we realized we were seeing the shadow of September 11th cast upon the influenza season."