Using Routine Surveillance Data to Estimate the Epidemic Potential of Emerging Zoonoses: Application to the Emergence of US Swine Origin Influenza A H3N2v Virus

Cauchemez, Simon; Epperson, Scott; Biggerstaff, Matthew; Swerdlow, David; Finelli, Lyn; Ferguson, Neil M.

For situations where detection of a case may increase the probability of detecting other cases in the same cluster (surveillance scenario 2), we estimate R from the proportion F of first detected cases in each cluster that were infected by the reservoir. Figure 2 shows the relationship between the reproduction number R and the proportion F for different values of the detection rate r and the overdispersion parameter k (parameter k characterizes case-to-case variation in infectiousness; see [12] and Methods). When the case detection rate r is relatively low, at levels similar to those seen in sentinel systems like the US influenza virological surveillance network, the straight line dependence shown in Figure 2 illustrates that R can be estimated as R = 12F (see Methods and Text S1). We find that it takes relatively high levels of case detection (r) or case-to-case variation in infectiousness (k) to cause substantial deviations from this linear relationship, and even then, such deviation only occurs for values of R close to 1 ( Figure 2 ). We also find that the relationship is only weakly sensitive to having multiple chains of transmission per cluster of human cases ( Figure S1 ; Text S1).