bars
spacerzimmer topbars
spacercz bottombooksarticlestalksblogcontactsearchspacer

Article Archives

[ 2014 ] [ 2013 ] [ 2012 ] [ 2011 ]
[ 2010 ] [ 2009 ] [ 2008 ] [ 2007 ]
[ 2006 ] [ 2005 ] [ 2004 ] [ 2003 ]
[ 2002 ] [ 2000 ] [ 2001 ] [ 1999 ]
[ 1998 ]      

 

2014

This Week’s Forecast: What Flu Season May Look Like
New York Times, January 16, 2014
Link

Jeffrey Shaman, an environmental health scientist at Columbia University, hopes that he and his colleagues will someday change the nightly news. “The way you get pollution reports and pollen counts on the local weather report, you could also have a flu forecast on there,” said Dr. Shaman.

Each year, the flu season arrives like clockwork. But once it strikes, it can unfold in surprising ways. In 2012, for example, it arrived in November, four weeks ahead of the typical flu season. Some years it can be especially brutal, and in others, very mild. Infection rates may start climbing in some parts of the United States when they are already falling in others.

Scientists like Dr. Shaman are reducing this uncertainty with computer models that make predictions about flu seasons in the United States. Last year, Dr. Shaman and his colleagues carried out their first flu forecasts in real time. They are now making predictions about the current outbreak, and this week they set up a website where you can see their predictions for yourself.

Their interest is greater than curiosity. Hospitals and public health workers could someday use flu forecasts to prepare their vaccine supplies and hospital beds. The advanced warning would be useful not only for the regular seasonal flu, but also for so-called pandemics, when a new strain sweeps across the country and causes higher rates of disease and death.

“In the event of a pandemic, this could become a more important issue,” said Matthew S. Biggerstaff, an epidemiologist at the Centers for Disease Control and Prevention in Atlanta. “We could understand how much lead time we had before a pandemic might peak, and it could help us get resources to the places that are going to need them.”

The spread of the flu is difficult to predict because it depends on many interconnected factors — how fast the virus replicates in people, for example, how well their immune systems wipe it out, and how easily the virus travels from one person to another. Despite decades of research on the flu, scientists are still deciphering some of the most important factors. Dr. Shaman and his colleagues, for example, have found that the drier the air is, the easier it is for viruses to spread from person to person.

Scientists are now using that understanding to make forecasts. To build computer models for predicting the flu, Dr. Shaman and his colleagues have gotten inspiration from meteorologists, whose embrace of mathematical tools has greatly improved their forecasting ability. “They really have the same problems for predicting the weather that we have for predicting influenza,” Dr. Shaman said.

To estimate the current level of the flu, Dr. Shaman and his colleagues look at the number of influenza-related Google searches, a pretty reliable indication of how many people are actually sick with the flu.

Dr. Shaman and his colleagues combine these estimates with other factors, like humidity, to generate a prediction for how the number of flu cases will rise or fall in different cities. Later, as verified cases are reported, the scientists judge how well their predictions turned out.

Initially, the model doesn’t do well. That’s not surprising, given that Dr. Shaman and his colleagues can’t directly measure some of the most important factors behind a flu season. They don’t know how many people in a city have some immunity, for example, or how many are susceptible. To start the model, the scientists make educated guesses.

Once the scientists have hard data about the flu season, they can adjust those guesses. That process leaves them with a more accurate model, which they can then use to make new predictions.

“It’s a little bit like aiming a gun. We fire it off into the future,” said Dr. Shaman.

In November 2012, Dr. Shaman and his colleagues put their model to a nationwide test, making weekly flu predictions for 108 cities. As the researchers reported last month in Nature Communications, they accurately predicted the timing of the peak in 63 percent of those cities, typically two to four weeks in advance. Thati s a much better performance than previous methods, which weren’t based on computer models.

This flu season, Dr. Shaman and his colleagues are making more predictions, broken down by regions or cities.

According to their model, New York, Chicago and some other cities are at the peak of flu season this week. Regions that have already been hit hard, like Texas and Tennessee, can expect declining cases. But other cities, like Boston and Miami, won’t reach their peak until February.

“It’s really exciting and important to get ahead of the flu,” said Lyn Finelli, who runs influenza surveillance for the C.D.C.

Dr. Shaman’s model may not be the one that public health workers rely on in years to come. The C.D.C. is running a contest, the Predict the Influenza Season Challenge, with a $75,000 prize for the best predictions for this year’s flu season. Fifteen teams, including Dr. Shaman’s group, are competing.

In the meantime, Dr. Shaman hopes that the website he and his colleagues have set up will encourage responses that they can use to strengthen their model.

“I expect us to do well in some instances, and not so well in others,” said Dr. Shaman, “the same way the weather forecast is good some days and not so good on others. And hopefully, like the weather forecast, we’ll improve it over time.”

Copyright 2014 The New York Times Company. Reproduced with permission.
Content Management Powered by CuteNews