H1N1 spreads like cooties through a school's social networks

A few years back, researchers were excited to hear about the spread of a contagion in World of Warcraft. The excitement came because Blizzard’s logs could potentially provide a complete picture of the spread of the disease, something that’s not generally possible in the real world. But, thanks to a town identified only as “a semirural community in Pennsylvania,” researchers may have their best glimpse yet: every H1N1 flu case that was reported in an elementary school, supplemented by details as fine-grained as classroom seating charts.

As the outbreak swept through the community, the Pennsylvania Department of Health’s H1N1 working group gathered demographic and clinical information on over 80 percent of the students in the school, then extended out into their households, obtaining information on a total of 1,269 individuals (370 of them students). A full 35 percent of the students reported acute respiratory illness during the study period, as did over 15 percent of the general population.

There are some caveats to this information, noted by the authors. For starters, it doesn’t capture the community as a whole; many households wouldn’t have had kids in this school, and transmission from other sources probably played a major role. Another issue is that not everyone with respiratory distress has the flu. Seventy-two percent of suspected cases in their limited sampling of students turned out to be real infections. That’s quite high by most standards, but it does introduce a significant source of potential error into the analysis.

A quick look at the data showed that fourth graders were hit the hardest, so they focused attention there, giving the students a more detailed survey that included social network information, and getting copies of seating charts as well as additional information on in-school activities. With all the data in hand, they modeled the spread of the virus within the population.

A full 74 percent of students got the flu from another student, but sitting next to an infected student doesn’t seem to be a big factor in producing infections, nor did having an infected student in one’s social network. Being in the same classroom, however was a huge factor. A student in the same grade but a different classroom was five times less likely to pick up the flu from another infected child. A six-day closure of the school had no impact on the infection rate; over a quarter of the students had the flu by then anyway.

One aspect of social networks did influence the flu’s spread among fourth graders: children of the same sex tended to socialize together, and infections within those groups tended to be more common.

Parents tend to claim that their kids bring every illness on the planet back home with them, but the majority of adult cases originated from someone other than their children. Although adults were less susceptible to this flu virus, once infected, they were just as good as their kids at transmitting it to the rest of the household.

Overall, the results paint a complicated picture. The flu virus mostly entered a household from someone other than a student, but the students helped maintain a reservoir of infection outside the school that continually replenished the infected population within it; the authors refer to “back-and-forth waves of transmission between the school, the community, and the households.” Within the classroom, infections appear to be very efficient, but classroom size is small relative to the total school population. As a result, the magnitude of the two types of infection (within and outside of class) ended up roughly equal.

As the authors point out, this is the sort of information that we need to incorporate into models we use to predict the spread of new pandemics, and to design measures to limit that spread. For example, closing schools once a substantial fraction of students is already infected appears to be ineffective. And it seems likely that an office might act much like a single classroom when it comes to transmission of a virus.

PNAS, 2011. DOI: 10.1073/pnas.1008895108 (About DOIs).