Flu? Disease forecasters look to the sky

"Why Dallas, and not areas with similar ecological conditions? We don't really know," said Roger Nasci of the Centers for Disease Control and Prevention. He is chief of the CDC branch that tracks insect-borne viruses.

Some think flu lends itself to outbreak forecasting — there's already a predictability to the annual winter flu season. But that's been tricky, too.

Seasonal flu reports come from doctors' offices, but those show the disease when it's already spreading. Some researchers have studied tweets on Twitter and searches on Google, but their work has offered a jump of only a week or two on traditional methods.

In the study of New York City flu cases published last month in the Proceedings of the National Academy of Sciences, the authors said they could forecast, by up to seven weeks, the peak of flu season.

They designed a model based on weather and flu data from past years, 2003-09. In part, their design was based on earlier studies that found flu virus spreads better when the air is dry and turns colder. They made calculations based on humidity readings and on Google Flu Trends, which tracks how many people are searching each day for information on flu-related topics (often because they're beginning to feel ill).

Using that model, they hope to try real-time predictions as early as next year, said Jeffrey Shaman of Columbia University, who led the work.

"It's certainly exciting," said Lyn Finelli, the CDC's flu surveillance chief. She said the CDC supports Shaman's work, but agency officials are eager to see follow-up studies showing the model can predict flu trends in places different from New York, like Miami.

Despite the optimism by some, Dr. Edward Ryan, a Harvard University professor of immunology and infectious diseases, is cautious about weather-based prediction models. "I'm not sure any of them are ready for prime time," he said.