2017 Volume 70 Issue 6 Pages 609-615
In norovirus and Campylobacter food poisonings, the frequencies of the number of patients per incident and that of the number of eaters per incident followed a lognormal distribution, with medians of 12–27 and 23–48 for norovirus and 5–8 and 9–21 for Campylobacter food poisonings, respectively. The lognormal frequency distribution of eaters could be simulated by assuming that people find a dish more appealing if that dish has already been found to be appealing to others. The numbers of patients and eaters per incident were not necessarily inter-correlated; the frequencies of the attack rates (number of patients/number of eaters) were distributed evenly from 0.01 to 1; that is, the attack rates of these food poisonings could not be represented by means and standard deviations. The frequency distributions of the attack rates were nevertheless not entirely disordered; plotting the attack rate against the number of patients in individual incidents produced fingerprint-like patterns that were repeatedly produced at the prefectural and national levels.