View report page

71
Those of us who, frankly, do not like the look of this tertium quid will naturally try to explore
other possibilities of escape. And, first of all, the question may be raised as to the extent to which
reliance can be placed, in such a connection as this, upon "partial correlation." When a scientific
method is newly introduced, it is apt to carry all before it; and yet every student of preventive medicine
who studied Professor Pearson's early papers was perplexed on learning that while coefficients of correlation
encourage a belief that a child may resemble one or other or both its parents, they give but little
added support to many of the hard-won truths of epidemiology; and further, it may be urged, that
while from the ultimate judicial point of view the size of the samples and the range of the enquiries made
are very important, from the point of view of the field enquirer, who is working on new ground, perhaps
even more depends on quality than quantity in the matter of observations. The bearing of these considerations
may be better appreciated by considering, on the one hand, Professor Pearson's somewhat
ambiguous correlation coefficient for vaccination and immunity from smallpox, and, on the other hand,
the quite definite evidence of the value of recent and efficient vaccination in the case of doctors and
nu ses who have been freely exposed to smallpox infection. Moreover, on turning from total to "partial
correlation," there can be no doubt about the fact that the plot thickens. Thus (i.) the strain upon
the reliability of the data is now far greater than before; (ii.) and perhaps the most important of all,
in the case of partial correlation, the very fact of having allowed for possible disturbing influences exerted
by other variable factors may have introduced a new source of error, for these other factors may stand
in such close relationship with one of the factors primarily concerned, that in eliminating the influence
of the other variables, we may have exhausted from our first and second partial correlation coefficients
almost all trace of a correlation which really does spell causation; (iii.) finally, even when correlation
is somewhat close, the fact may merely mean, as Dr. Greenwood says, that some unrecognised tertium
quid is deluding us into supporting a false theory of causation.
In the present instance all three difficulties in interpreting the partial correlation co-efficients
are probably in question.
First as to unreliability of data. The measure of overcrowding (more than 2 per room) is
admittedly untrustworthy; the details concerning the industrial employments are not given, but clearly,
in making comparison between the sexes, the considerations referred to on p. 340 of Dr. Greenwood's
paper render appraisement of the precise influence of industry in a place like London particularly
difficult; then over it all there is the feeling that the items of information scrutinised are, so to speak,
every one of them, on the move—thus, the young persons employed in an overcrowded borough, for
example, do not in many instances live in the borough, and deaths of those who do live or work there,
when in the fulness of time they occur, may possibly (it may perhaps be said, probably), not be registered
in or transferred to the borough.
The second difficulty is that already referred to of the exhaustion of the apparent effect of an
influence when allowance is made for other influences closely associated with it. "Other diseases,"
"overcrowding," and "employment" are all bound up with poverty, and the task of allotting to
each its special share in the total result qua tuberculosis is a well nigh impossible one. Thus in table
13 on p. 339 the total correlation of overcrowding and employment in females, 25-45, is high 8472;
the first partials "Tuberculosis and Overcrowding" (employment constant), and "Tuberculosis and
Overcrowding" (other causes constant), are only 2707 and 2311; the second partial "Tuberculosis
and Overcrowding" (the other two variables constant) is 0519, i.e., practically zero. But does not
this merely mean that when allowance is made for the overcrowding bound up with employment of
these women, and for the overcrowding bound up with other diseases affecting them, we have entirely
exhausted the effect of overcrowding? And, if so, it clearly does not mean that overcrowding and
tuberculosis are not closely correlated in the case of women at these ages.
Now as to the third or "tertium quid" difficulty at the ages at 15-25. If this were "blind
alley employment" one would expect the males to be much more affected than the females; the
actual difference is not, however, great, both partial coefficients being decisively negative. Again the
statistics of Table XII. do not give much support even to the case for males; for the factor of selection
must operate in the earlier ages—a lad of 15-20 in the group "Carmen, carriers, etc.," might be expected
to be better grown and of superior physique to the average young person, and thus less likely to die
from tuberculosis—furthermore, alcohol plays great havoc in some of the employments here in question,
and its influence would, of course, not be apparent in favouring tuberculosis until the later ages; and
yet again the mobility of these young persons, the shortness of time taken up by the employments in
question and the number of years likely to elapse in fatal cases before death supervenes, all these
considerations are likely to vitiate the conclusions tentatively suggested.
But in connection with this question of the tertium quid at the earlier ages (15-25), there is a
special claim to be made I submit for immunisation. In these reports, as pointed out at the commencement
of the present discussion (p. 70), the close correlations between measles and scarlet fever, on the
one hand and poverty and overcrowding, on the other, have during the past seven years, been frequently
emphasised. Hence the earlier attack, in the case of these diseases of children, in the poorer boroughs.
The same phenomenon certainly plays a part in tuberculosis; measles and scarlet fever afford life-long
protection; but the position in tuberculosis is not, of course, so clear as this. It is known, however,
that those who manifest in early life tuberculosis of glands, joints, or spine, the ordinary non-pulmonary
tubercle of the children's hospitals and sanatoria, are not prone to develop pulmonary phthisis in
early manhood and womanhood. There is, in fact, a growing conviction at the present time that
slight forms of tuberculosis in early life protect against later manifestation of the disease. Furthermore,
the heavy death toll from abdominal and cerebral tuberculosis in young children must act as a severe
selective influence.