vaccination. The complete spectrum of scenarios is summarized in Figure 3-10. For each of the three architectures of the 6,000-person town model, each of the eight response scenarios was run 35 different times, with a different random seed each time. Note that the model is stochastic; the use of a different random seed for each run ensures that particular realizations of the same model produce different random contact patterns in the population. The random seeds also ensure that model parameters drawn from set distributions (see Longini et al., 2006 and Figure 3-9) vary from individual to individual and run to run. For each combination of town architecture and response scenario, we report here the mean and the standard deviation (over the 35 simulated epidemic runs) for the following key epidemic outcome measurements: total number of cases, total number of deaths, total number of vaccinations, and total epidemic duration.
For the 50,000-person town model runs, the analysis format was similar to that for the 6,000-person towns, except that here 500 initial infected individuals were introduced into the population. Because these runs absorb substantially more computational resources than the comparable 6,000-person town simulations, we explored only two architectures: the single uniform large town of 50,000 and a ring of six districts of equal size. Instead of 35 simulated epidemic runs per scenarioarchitecture pair, we present the statistics for just ten stochastic realizations.
Cases, deaths, vaccinations, and epidemic durations for simulated epidemics under the two “no response” and the eight response scenarios are shown in Tables 3-11 and 3-12 (see Annex 3-1) for populations of 6,000 individuals and 50,000 individuals, respectively.
Simulations under the highly unrealistic “no response” scenarios 1 and 2 gave rise to large and lengthy epidemics. Each index case, on average, initiated an epidemic chain of transmission that subsequently infected hundreds of other individuals.
Response scenarios 3–10 were all examined in model populations with each of the three (single uniform, ring, and hub-and-spoke) town architectures, and 35 simulated epidemics were run for each scenario in each architecture. Results are displayed in Table 3-11 (see Annex 3-1). Although there are some minor differences in the impact of the different response scenarios in different town struc-