Abstract
In the previous study the authors investigated the utility of the areal division by the grid square method applied to the geographical distribution of cancer mortality in Tokyo. In the present study, we intended apply it to major causes of death and extended areas which have larger variations in population density. To smooth the variation of confidential intervals of death rates, some grid square areas are grouped into one area. When the expected number of deaths is less than five, the grid square areas are made into one unit to the extent that the expected number is up to five. As a result, 1126 original grid square areas are reproduced to 341 new combined areas. The maximum number of original areas in one combined area is 63 and the average is 3.3. Inspection of the geographical distribution for the various causes of death reveals some differences in the patterns of mortality. However, some patterns are similar and may be grouped into two categories:
1. high rates in the western area, rural and/or hilly, and in the eastern area, industrial, but generally low rates for middle zone, residential; all causes, apoplexia, pneumonia or bronchitis
2. a cluster of high rates in central Tokyo; cancer
