How fast can flu spread in a household?

How fast flu spreads is related to how many susceptible people an infectious person can infect (a measure called R0) and also something called the serial interval. The serial interval is the average length of time between the start of one infection and the start of the infection of that case's infected contacts. The horter the serial interval the faster a virus can spread. So what is the serial interval for this virus and how can we determine it? The answer to the first question is the usual. We don't know yet. The answer to the second one tells us a little about why we don't know and why we are even somewhat uncertain about the serial interval for seasonal influenza.

A paper just published in the journal Epidemiology illustrates some of the difficulties (Cowling, Benjamin J.; Fang, Vicky J.; Riley, Steven; Malik Peiris, J S.; Leung, Gabriel M., "Estimation of the Serial Interval of Influenza," Epidemiology: May 2009 - Volume 20 - Issue 3 - pp 344-347, doi: 10.1097/EDE.0b013e31819d1092). It reports a household transmission study from Hong Kong in 2007. 123 index cases of laboratory confirmed influenza who had presented to a primary care provider and had least 2 symptoms from a list of 5 characteristic of influenza-like illness (ILI): fever, cough, headache, sore throat, muscle and joint aches or pains. Index cases were required to have had symptom onset no longer than 48 hours previous and lived in a household where no one had an ILI for a minimum of the previous 2 weeks. The households were visited 4 times in the next 10 days and nose and throat swabs of all household members taken. The serial interval was estimated by looking at the time from onset of symptoms in the index case (by history) and onset of any symptoms in household members with infection verified by viral culture or PCR.

This looks like a pretty good design and something similar is probably being done with the new H1N1/2009 virus. But notice this isn't exactly the serial interval but rather the clinical-onset serial interval. In this case it could also be called the household serial interval. Not measured is the latent period, the time from infection to the time of infectiousness, a relevant quantity. Moreover the real serial interval is the sum of the time from exposure to infection, and the incubation period, the time from infection to the onset of symptoms. So what is measured as the clinical-onset serial interval is shorter than the serial interval.

In addition the authors faced another difficulty. There is a built in bias to the way the interval is being measured here. If an index case is 48 hours post symptom onset and we require that prior to this the household be ILI free, it cannot contribute any data on serial intervals less than 48 hours, and similarly for index cases found within 1 day of onset. This effect is called left censoring and there are statistical methods to deal with it, which is what Cowling et al. did, but it required using a model for time to appearance. Already we can see that determining this quantity is not completely straightforward.

The results for this seasonal influenza virus gave an estimate for clinical onset serial interval of 3 to 4 days, somewhat longer than other estimates in the literature. There were 350 household contacts, but only 21 confirmed infections among them. Of these, a third (7) didn't experience any flu symptoms. A frequently used estimate of 1 - 2 days for incubation period comes from a classic investigation of 36 virologically confirmed cases from a single infector on an airplane (Moser et al., Am J Epidemiol. 1979 Jul;110(1):1-6; we've mentioned it in other posts). So Cowling et al. observe that their estimate of serial estimate of 3.6 days minus an incubation period of 1.5 days would suggest an "average time from symptom onset in the index case to secondary infection in the household" of around 2 days.

There are other difficulties, as well, as the authors point out. The number of cases is small and represent disease dynamics in a non pandemic situation. A pandemic may have very different characteristics because the relationship of the virus to host factors will be different. Moreover there may have been some "co-primaries" among the household contact, i.e., flu infections not from the index case but roughly concurrent with it or cases contracted outside the household.

So this estimate takes its place among others of 1 - 2 days, 2 - 3 days and now 3 - 4 days. Whether the differences are due to biases or imperfections or differing definitions of the interval or true variability for different flu viruses is not clear. While we will likely soon get some numbers for H1N1/2009, the inherent problems in making these serial interval estimates should be kept in mind.

Science is hard.

More like this

If there was ever a graphic illustration of how global interconnectedness affects public health, it's the swine flu affair. Wherever it started, the current crop of cases seems related to Mexico, either as the epicenter or via travelers. Four US states have cases. Those not on the Mexican border…
If you were infected with seasonal influenza virus, how long before you start having symptoms (the incubation period)? How about how long before you start shedding virus so that you can infect others (the latent period)? When are you most infectious? Expert opinion. based on experience and the…
I have a reflexive skepticism about some conventional flu wisdom. There's so much about flu we don't know and even more we think we know that we find out we're wrong about. But skepticism is an occupational hazard of epidemiologists. Our training and practice focusses on detecting subtle biases…
When I tell people I am an epidemiologist, most of them think it means I'm a skin doctor. I'm not (although the skin disease specialty is much more lucrative). Instead I study patterns of disease in populations and use what I see to try to figure out why the observed pattern rather than another.…

One other thing springs to mind...how much virus (not just how long in contact with an infected person) but actually how MANY virus particles are required to catch a bug.

The research here suggests just one.
http://www.sciencedaily.com/releases/2009/03/090313150254.htm

Of course, it needs to be in the right place at the right time ...and hence the chances surely go up the more virus particles that a family member is exposed to and the longer the time that the individual remains exposed.

I think that it's important to ascertain how many particles are required to cause infection as well as how long before the onset of symptoms,or, after "infection", is one shedding sufficient quantities of virus to be infectious.

It seems to me that Cowlings study can only be conducted under VERY controlled circumstances, and then what would be a satisfactory sample size ? does 36 do it? That result would then be reliable for that given strain, and you would then model the other potential variables.

It seems any definitive discoveries then need to be compared with common sense man in the street observations,after the fact, in the wild, "like 10 days ago we had a an index case "in town", now everyone I ask knows someone who's sick, or whatever...with the same strain.

While on the topic of natural history of flu, there's another recent paper from Lancet Infectious Diseases that combines historical data to give an estimate of the incubation period for a host of respiratory viral infections, including influenza A and B. While there are some potential complications with combining data from a bunch of different studies that looked at different flu viruses, this paper does give some more precision to our understanding of the incubation period of influenza and of the utility of incubation periods in general.

crap!!! lol :)