552 
A Study of the Effects of Diphtheria Isolation 
(d) Economic prosperity as measured by the number of indoor and outdoor 
servants of both sexes per 100 private families. 
Our data are based on the census of 1911 as providing more ample information 
on these points. It will we think be admitted that the list of towns dealt with 
pro.vides a very fair sample of the urban populations of this country. It ranges 
from manufacturing towns* like Preston, Rochdale and Bolton, mining and iron 
towns like Rhondda, Wigan and Middlesbrough, sea-ports like Hull, Liverpool 
and Southampton, to county towns like York and Reading, watering places like 
Brighton and Blackpool, suburban districts like Acton and Hornsey, and residential 
towns like Oxford or Bath. We ought from such a list to be able to throw some 
light on the relation of isolation to incidence under a variety of social conditions, 
if indeed these latter are factors in the problem at all. 
(3) What are the crude correlations between Isolation-Rate, Attack-Rate and 
Mortality-Rate ? The isolation-rate (/) has been measured as the average per 
cent, of cases removed to hospital during a five or four year period. We have two 
such periods, the earlier period 1904-1908 and the later period 1909-1912. The 
attack-rate has been measured per 1000 of the population, uncorrected for age 
distribution. Since diphtheria is largely a disease of infancy and childhood this 
neglect of the age correction — the reduction to a standard population — may seem 
serious. But in the first place we had not the age incidence in the individual 
districts, and in the second place we satisfied ourselves that such correction, if it 
could have been made, would not substantially modify any argument we have 
based on our data. For we calculated the attack-rate {A') on the population 
under 15 years of age, as well as the attack-rate {A) on the total population of the 
districts. We found the correlation between the two methods of measuring the 
attack-rate was -!-'972, which indicates how close is the relation between the two 
methods of measuring the attack-rate and how little influence small variations 
in the proportion of less immune persons in the population due to age differences 
could have on the resultsf- 
The attack-rate {A) has been measured as the number of cases per 1000 of the 
population. The mortality -rate has been measured in two different ways ; first as 
the population mortality, the death-rate in the ordinary sense {M) or the deaths 
per 1000 of the population; and secondly the case death-rate or the mortality (m) 
per 100 attacked. We now give the crude correlations between / and A. 
They are : 
First Period : 1904-1908, r^^, = + '427 + -063, 
Second Period : 1909-1912, = + "290 ± -069. 
* See table, p. 567, for 76 of the 80 towns, tlie four others with full data only for the second period 
being Beading, Stoke, Devvsbury and Edmonton. 
t The formula giving the juvenile attack-rate A' in terms of the crude attack-rate A is : 
^' = 1-3094^1 + -0164 
with a probable error of ±-1369. 
Thirty-three towns were selected at random out of the 80, and gave the following results for A' 
calculated from A and A' as observed. The theoretical mean error=-162; the mean error of the defects 
