G. D. Maynard 
287 
Unfortunately no figures were obtainable to allow of a correction being made 
for those deaths which occurred in hospital and did not belong to the district. 
Although this correction is most desirable, yet it is not likely to very greatly 
affect the correlation values for cancer and diabetes, as the latter disease is not 
one that bulks very largely in hospital deaths. It is not a disease very prevalent 
among the class of patient that frequent hospital, and from its nature would 
not be admitted unless combined with some acute affection. This of course will 
occur in a certain number of cases but to a small extent as compared with cancer. 
This objection will not apply to the returns from the States, and all the cities 
here dealt with, except a few eastern ones, will draw from fairly large rural 
areas. 
Before coming to the conclusion that the correlation found between cancer and 
diabetes is of real significance, one must as far as possible exclude all sources 
of error that might lead to false deductions. The question of the inaccuracy 
in the records from incomplete registration has already been discussed, but there 
is another form of inaccuracy that must be considered, viz. that arising from want 
of care and knowledge on the part of the certifying medical practitioner. In 
Mortality Statistics U.S.A. 1907, the following remark occurs: "In Table IV. may 
be found two columns headed ' ill-defined causes ' and ' cause unknown ' which are 
of no special significance in themselves, but which are nevertheless important, 
because they show the margin of understatement that must be considered in 
examining the rates for more definitely stated causes of death." It seemed 
therefore desirable to correlate cancer and diabetes death-rates with a five-yearly 
average rate of these groups, for the 40 cities occurring in the other tables. The 
values found are as follows : 
Cancer and diseases of ill-defined and unknown cause, p = '1020 ± "lOoo. 
Diabetes and diseases of ill-defined and unknown cause, p = -0773 + "1060. 
It does not seem likely therefore that any appreciable correlation will have arisen 
from this cause. 
It has been suggested to me that in towns well equipped with competent 
medical men these diseases would figure less in the death returns, because by 
means of operative treatment in the case of cancer, and dietetic treatment in the 
case of diabetes, death might eventually be due to some other cause ; whereas 
where medical skill was inferior, or the population not sufficiently educated to take 
suitable precautions in regard to diet, or submit to operation, a larger number 
of cases of both diseases would appear in the records. In the first place I do not 
suppose that the average medical skill would vary very greatly in cities of the size 
dealt with in this paper. It is no doubt true that in both diseases, life may be 
prolonged by skilful treatment but a cure is not effected, except in very rare 
instances. When death occurs from an intercurrent disease the older complaint is 
included on the death certificate. At least this should be done and it is less likely 
to be omitted where the medical skill is of a high order. In classifying causes of 
37—2 
