G. D. Maynarp 
281 
the method by which the standard death-rate used was obtained is given in the 
Supplement to the 65th Antmal Report of the Registrar-General, for the years 
1891 — 1900. The standard population is that of England and Wales for the 
same period. It is immaterial what standard is adopted so long as all crude rates 
are reduced to the same base, as they can then be compared inter se, although 
they do not represent the actual rate obtaining in the district considered. 
The only source at my disposal from which it was possible to obtain the 
necessary data to correct the rates for towns and districts, was the United States 
Census of 1900. Fortunately this country covers a large area, and is subject 
to many varying influences of climate and social conditions. It is thus a 
particularly suitable country in which to study the conditions which may pre- 
dispose to the development of this disease. Unfortunately, however, the whole 
country is not under a uniform registration law ; the field is therefore restricted to 
those states and cities which are. The statistics from these Registration States and 
Cities are probably fairly accurate ; and I think the marked differences in rates 
that are observed, cannot be due to faults of registration, and point to real 
differences in predisposing causes. As a check on the accuracy of the death 
returns the Census Department made an enumeration of the deaths occurring in 
the year ending May 31st, 1900. This was carried out in all districts where there 
was any doubt as to the state of the registration. The results show that so far as 
the cities dealt with in this paper are concerned, the error in no case exceeds 
9'5 '7o- 111 oiily 0116 city was the error greater than 6"5 °/^, while the mean error 
is 3'5 % with a coefficient of variation of 2-225 
In addition to the correction for age, a sex correction might also have been 
introduced. In towns of over 100,000 inhabitants — and none with a smaller 
population than this have been used in this paper — the variation in sex distribution 
is not, from our present point of view, of very great importance, while the extra 
arithmetical labour involved in calculating over 150 further correction factors 
is very considerable. In a few cases where the sex distribution seemed to vary 
considerably from the usual proportions, I calculated the sex correction factor, but 
in no case did the alteration in the corrected rate amount to O'o " j ^. It should be 
remembered that the cancer death-rate is highest for females, while the diabetes 
death-rate is highest for males, thus any error resulting from this omission would 
tend to lower the correlation that exists between these diseases. 
In dealing with large areas of mixed urban and rural populations another 
source of disturbance must be borne in mind, viz. the different age distributions in 
the two districts, with its effect on the death-rates in each. The correction factor 
for the rural population of the ten Registration States in 1900 is for cancer 7585, 
while for the urban areas of the same States it is 1'0()17. Now as the proportion 
of the population in the urban and rural districts differs in each State, this in 
itself will cause an appreciable difference in the observed rates. The following 
table shows the change in rate that would be made from this cause alone, were the 
age distribution of rural and urban districts uniform for each State. 
