March 6, 2018

Flu season — What are we learning?


KPWHRI infectious disease expert Dr. Michael Jackson reports on the latest influenza news and how we’re learning to improve future predictions.

by Michael Jackson, PhD, MPH, Kaiser Permanente Washington Health Research Institute associate investigator

You’ve probably been hearing a lot about influenza lately. This winter, the United States has experienced the most severe influenza season since the 2009 influenza pandemic. As of February 24, 2018, the Centers for Disease Control and Prevention (CDC) report more than 39,000 laboratory-confirmed cases of influenza in the United States, mainly influenza A. More than 114 children have died of the flu this season. Although reports of hospitalization rates that are similar to severe flu seasons such as 2013–2014 are alarming, we may be over the worst at this point, as the number of reported cases of the flu are declining slightly,

One of the big challenges with influenza is that influenza epidemics are unpredictable. Although they happen every winter, we don’t yet have the ability to forecast when the season will peak, how severe it will be, and which strains of the influenza virus will dominate. This uncertainty makes it difficult for hospitals and emergency departments to plan for how many flu patients to expect. It also causes difficulties in choosing which virus strains to put into the flu vaccine.

Predicting the flu season with the Flucaster

My research group is developing a computer model — the Flucaster — in response to a call from the CDC for ways to predict the timing and intensity of a flu season. Models like the Flucaster are an important public health tool to guide our annual influenza preparations.

Flucasting week 18 | KPWHRI
by Michael Jackson, PhD, MPH, Kaiser Permanente Washington Health Research Institute (KPWHRI) associate investigator In…

Since the flu season is usually November through April in our part of the world, here’s a midway report on the national flu situation and what the Flucaster is learning:

  • In October, the Flucaster model predicted many cases of influenza because of data from countries in the southern hemisphere such as Australia. Their flu season always precedes ours and because they had a severe season, we braced for one, too.
  • As the Flucaster predicted, we had an initial wave of influenza A and are now getting a second, smaller wave of influenza B.
  • The Flucaster initially predicted about 11,000 to 17,000 cases of influenza A in the United States at the season’s peak. Near the probable actual peak in February, we had about 20,000, so the model underestimated the severity of the season.
  • The biggest inaccuracy with the Flucaster so far is around the dominant virus strains. The model initially predicted that the influenza A(H1N1) virus would dominate. Instead, the season has been dominated by influenza A(H3N2).
  • In Washington state, the Flucaster predicted about 350 to 500 cases of influenza A at the peak and we will have a little more than 500.
  • In both our state and the country, the season arrived and appears to be peaking a week or two earlier than the Flucaster predicted.

Improving the Flucaster for future flu season preparations

Every week, my research group collects CDC data that we’ll use to improve the algorithms of our Flucaster model. Training our model on these real-world data will make it a more reliable predictor that could help public health departments make decisions when planning for the annual flu season.

My colleagues and I have spent the winter focused on the news and numbers about flu season for personal and professional reasons. I hope you’ll join us in watching the weekly Flucaster updates while you follow health care providers’ advice about staying well during flu season.

Learn more about Kaiser Permanente Washington Health Research Institute. Sign up for our free monthly newsletter.