Forecasting Flu Diffusion Taking Account of Avoidance Tactics

Featured in print Digest

More lives are saved and infection rates are reduced when both avoidance behaviors and a vaccination campaign begin before the flu has spread

Novel influenza A or nH1N1, also known as the swine flu, appeared in the spring of 2009. Its first fatality was reported in Oaxaca, Mexico that April, and two months later the World Health Organization declared that it had reached pandemic status worldwide. Unlike seasonal influenza, nH1N1 afflicts children, pregnant women, and young adults, not newborns or the elderly. Infected people do not necessarily present fevers or coughs, frustrating both diagnosis and infection control. According to estimates from the Centers for Disease Control (CDC), nearly 10,000 people around the world had died from nH1N1 infection by December 2009.

The public and the public health community responded to the spread of the virus immediately and globally with voluntary avoidance of public spaces, school closings, and the ubiquitous placement of hand-sanitizer pumps. The widespread avoidance response in 2009 was both unprecedented and significant, but at that time, officials had no flu forecasting models that took avoidance response into account.

In Public Avoidance and the Epidemiology of novel H1N1 Influenza A (NBER Working Paper No. 15752), co-authors Byung-Kwang Yoo, Megumi Kasajima, and Jay Bhattacharya use data on avoidance behaviors in the United States and Australia to build nH1N1 forecasting models that explicitly account for the impact of avoidance on the epidemic's severity. The researchers compare their forecasting models to a model showing the cumulative path of confirmed nH1N1 infected cases - the pandemic's actual course - based on data from the CDC. They find that including a population's avoidance behavior results in a substantially better forecast of the actual spread of the flu.

When 196 million doses of an nH1N1 vaccine became available in the United States between October and December 2009, this nation's avoidance response to the epidemic broadened. The researchers include data on the new vaccine in their forecasting models. They consider a worst-case scenario, with a 50 percent effective vaccine and 50 percent of the population receiving it, and a better scenario, with 50 percent vaccine effectiveness and 90 percent of the population receiving it. In both cases, the vaccine's effectiveness is time-sensitive: the earlier the vaccination campaign begins, the lower the proportion of the population that will be infected. The authors find that more lives are saved and infection rates are reduced when both avoidance behaviors and a vaccination campaign begin before the flu has spread.

Their research also explores how prevalence of the flu influences individuals' decisions about whether to practice or postpone avoidance behaviors. In the first stages of the epidemic, the governments and mass media both here and in Australia released information on nHN1 to the general public. Despite early vaccine shortages in the United States, widespread avoidance behaviors quickly slowed the nH1N1 infection rate. Yet, as the flu grew less prevalent and was less discussed in the media, the number of people receiving nH1N1 vaccinations lessened, even among health care workers and those who received seasonal flu shots.

-- Sarah H. Wright