As we head into the 2013 flu season, the need for preparation is critical.
The Center for Disease Control (CDC) estimates that the influenza virus is responsible for anywhere from 3,000 to 49,000 deaths every year in the United States. With a unique strain introduced to the population every year, traditional methods to lessen the infections are often ineffectual. In addition, the threat of a widespread pandemic infecting millions is always feared during peak season.
In response to these concerns, researchers at Columbia University developed a method of forecasting influenza outbreaks they hope will result in "flu forecasts" as ubiquitous and widespread as weather forecasts.
Jeffrey Shaman, assistant professor at the department of Environmental Health Sciences for Columbia University, created a model that can predict the spread of the influenza virus up to seven weeks in advance.
The paper outlining the construction of the model, "Forecasting Seasonal Outbreaks of Influenza," was published in the Proceedings of the National Academy of Science.
With the prediction model, it is possible to calculate the path, timing and severity of influenza outbreaks using some of the basic principles used for weather forecasting.
The inspiration for using weather framework was found in the notion that the rate of influenza outbreaks are specific to each region and associated with absolute humidity levels. Shaman explained, "We pursued the sort of methods they use for weather forecasting, but we applied it to [the] flu."
Previous models were only able to predict the spread of influenza around two weeks in advance. "They didn't push it to see how far into the future they could get it and never calibrated it to see how accurate the forecasts were," he explained.
The new method uses three components: a dynamic model that describes the system (the weather forecasting framework), an observational network of real-time influenza outbreaks and a data method to process the information and optimize the forecast. This allowed them to postulate the spread and rate of influenza infections into the future.
For the observational network, the study utilized Google Flu Trends, a program that monitors Google web searches for flu-related terms that have correlated strongly with actual Center for Disease Control (CDC) data. Using Google Flu Trends allowed them to track the infections at the municipal level. Similar statistics from the CDC, due to privacy concerns, are only available on a regional and national level and are not accessible in real-time.
The method resulted in an accurate forecast of predicting the severity and spread of influenza infections across the country. Currently, Shaman says disseminating a flu forecast for the peak flu season is actually "quite achievable. We have the models, we have the statistical method and we have the observations. We need to put them together and do it in real time."
Researchers applied their forecast to the 2012 flu season, and the results of that study are currently in peer review.
Looking forward, "The idea now is to make it operational, to disseminate it and to make it useful for public health officials to provide it so the public can use it. They give pollen counts, they give pollution levels, there's no reason why they can't give an infectious disease forecast," Shaman says.
By distributing a forecast, public health officials can implement stronger health initiatives such as vaccine allocation and distribution. Building on that notion, influenza outbreaks could be dramatically lessened with this forecast. "Unlike the weather, you can actually change the outcome with the flu," Shaman explains, "If we put this up here, and everyone went out and got vaccinated, that could potentially nip the outbreak in the bud."
The authors are confident that, "The forecasts indicate that we will soon reach an era when reliable forecasts of some infectious pathogens are as commonplace as weather predictions."