Inferring pandemic growth rates from sequence data

de Silva, Eric; Ferguson, Neil M.; Fraser, Christophe

Indeed, this flattening in the latter portion of the growth curve of the effective population size is visible in BSPs looking at HIV [11] [12] [13] , dengue [14] and hepatitis C [15] , and is often interpreted as being evidence for a slowing of spread. However, the implication that H1N1 transmission may have temporarily been reduced around late April 2009 should be treated with caution. Samples obtained in the USA during spring 2009 will be beset with sampling biases: spatially (more samples from some areas than others, given that US states varied in the level of testing undertaken), temporally (more intensive testing was often undertaken for the first few weeks of cases in a locality, and less thereafter, and there were national changes in testing protocols) and epidemiologically linked (samples taken from individuals in the same local outbreak, and which therefore are not random independent samples from the H1N1 infected population overall). In some analyses, a few of these biases were controlled for by removing sequences collected from patients in known local epidemiological clusters [8] but in general it is not clear the extent to which such biases in sampling affect estimates.