Week 1 Influenza Forecast for the 2012-2013 U.S. Season

This is part of a series of weekly influenza forecasts made during the 2012-2013 influenza season. Here we present results of forecasts initiated following assimilation of observations for Week 1 (i.e. the forecast begins January 6, 2013) for municipalities in the United States. These forecasts were performed on January 11, 2013. Results from forecasts initiated the six previous weeks (Weeks 47-52) are also presented. The accuracy of these predictions will not be known for certain until the conclusion of the current influenza season; however, at the moment a number of the forecasted peaks appear to be inaccurate. This inaccuracy may be due to the virulence of influenza this season, which appears to be sending more influenza-infected persons to seek medical attention and inflates ILI levels (and possibly the proportion testing influenza positive) relative to years with milder flu strains. New forecasts that adjust, or scale, for this difference and match the two focus cities that appear to have already peaked are identified. These new forecasts will be used, in addition to the previously scaled forms, to make influenza predictions for the remainder of the season. 1 Observation Bias and Scaling It is becoming clear that many of the forecasts archived in previous weeks predicted outbreak peaks that are too early. In part, the model grew overly confident in its own consensus or spread, but also the forecasts had difficulty handling the high levels of influenza incidence this year. Understanding why the forecasts are predicting peaks earlier than observed should help improve predictions going forward, both this year and in the future. Corresponding author address: Jeffrey Shaman, Department of Environmental Health Sciences, Mailman School of Public Health, Columbia University, 722 West 168th Street, Rosenfield Building, Room 1104C, New York, NY 10032. E-mail: [email protected]