The purpose of this paper is to study the 1957 epidemic of Asian influenza in Nagoya and its environs as viewed a spatial diffusion process. The study area is Aichi prefecture and Gifu prefecture except Hida area, the core portion being Nagoya metropolitan area. The epidemic in each city was identified from the first closing date of a class or a school caused by the influenza mass outbreaks in the elementary school and the junior high school.
The influenza breaks out in 103 cities out of 192, and the epidemic period (from May 24th to July 17th) is about nine weeks. According to the time lag of five sub-divided epidemic periods in each city the influenza tends to spread from the densely populated plain area to the mountain area, and an explanation that the outbreak pattern has been produced by the hierarchy effect (Fig. 4) and neighbourhood effect seems to be possible.
Then in order to examine the outbreak pattern from the aspect of process, the author com-pared the actual outbreak pattern with the outbreak patterns simulated by random process model and interurban diffusion model (Model I and II) by the aid of the Monte Carlo technique. Firstly, with A. D. Sorensen's coefficient of spatial association which is a kind of the nearest neighbour method, the output of random process model was compared with the actual. The coefficients of spatial association of the period II_??_V were 0.18, 0.19, 0.28, 0.33 respectively, and consequently there is a low possibility that the epidemic is resulted from random process. In addition as the outbreak in the period I is given for the model, both random process model and Model I and II, the simulation of the outbreak in the period I is not possible.
Secondly the author considered the outbreak pattern with Model I having the following main operating rules ;
1) It is assumed that the epidemic spreads from Nagoya, Kasugai, Seki and Tsukechi where the influenza broke out in the period I.
2) Muge and Utsumi where the influenza was introduced from the other regions are given for the model, and inserted in the generation corresponding to the actual outbreak period.
3) The probability of the contact of a city is determined by the contact field based on the number of commuters and externs which seems to fit the gravity model.
4) The generating frequency of the contact of the infected city is one time in one generation.
5) The city which has received the contact breaks out the influenza immediately, and continues to transmit the contact to the other cities from the next generation.
Fig. 6 is the output simulated by Model I. Though the author supposed that Model I contained the hierarchy effect and the neighbourhood effect, the neighbourhood effect emerges strongly in the simulated output compared with the actual. The coefficients of spatial association of the period II_??_V were 0.31, 0.28, 0.44 and 0.47 respectively.
Further the author attempted the simulation by Model II which is a revision of Model I in two operating rules; firstly by changing the generating frequency of the contact in proportion to the city size; and secondly by containing the density effect together with by increasing the hierarchy effect. Especially the reason why the second alternation is made needs a comment. From epidemiology it is said that the epidemic of the infectious disease is caused by the infectious chance and the susceptibility. In Model I the former is expressed by the contact field, but the latter is not considered. Calculating the coefficients of corre-lation between the outbreak date and five variables which seem to indicate the suscepti-bility, the author judged two varibales relating to the density the population density and the number of students per class in the elementary school to be comparatively signi-ficant.
View full abstract