|
Historical Keynote
Addresses
Rothman Gives Cassel
Memorial Lecture at SER
Eight Essential
Qualities of Enduring Work in Epidemiology Discussed
Interlacing his presentation
“The Challenge of Enduring Significance” with humor and quotations and
getting a warm reception from the audience, well known epidemiologist
Kenneth Rothman gave the annual Cassel Memorial
Lecture at the SER meeting in Snowbird, Utah in mid-June. Rothman
presented his views on the qualities and characteristics of
epidemiologic work that make it of lasting significance. He was
careful not to claim to have “real knowledge” of this subject;
however, his presentation was provocative and revealed detailed
thinking about some of the issues presented.
Taking John Graunt’s
Observations on the Bills of Mortality published in 1662 as his
jumping off point, Rothman identified eight characteristics which he
believes are associated with enduring work or with epidemiologists who
produce lasting contributions.
Five Features
According to Rothman, the first
five distinguishing features of Graunt’s work were: 1) brevity; 2)
clear reasoning as for example the details Graunt provided for his
estimate of the magnitude of the underestimation of plague deaths in
the Bills; 3) subjecting theories to repeated and varied tests as
evidenced by Graunt’s use of five separate approaches to arrive at his
estimate of 384,000 as the population of London; 4) recognition of the
potential for error as evidenced by Graunt’s invitation to readers to
criticize his work; and 5) avoiding mechanical thinking as Graunt did
by admitting he revised some of his preconceived conjectures in light
of his data.
Rothman particularly criticized
“science by recipe or formula” and urged that epidemiologists be free
to react to and interpret their data “after the fact” rather than
always be expected to state in advance and strictly adhere to what the
questions and parameters will be. In short, Rothman reminded the
audience that science is about discovery, and this cannot, by
definition, be anticipated.
Rothman described three
additional lessons which he said Graunt’s work also illustrates and he
spent the majority of his allotted time in describing these lessons.
In discussing these topics, Rothman was returning to familiar themes
he has struck before in some of his other writings and he is well
known for the views he expressed once again at SER.
Significance Testing
First, Graunt had no knowledge
of and no need for statistical significance testing. Rothman
criticized the use of such testing as “the ultimate in mechanical
thinking” and ascribed it to a kind of “tropism for formula answers to
complex questions.” He elaborated on the use of confidence intervals
and p-value functions as preferred alternatives for the interpretation
of data.
Second, Rothman noted the
“unimportance of credentials” in judging scientific work. Graunt was a
pioneer in epidemiology and demography but was known to his
contemporaries as a merchant. “Judge the work and not the author,”
said Rothman. He noted that this tendency persists today, and cited
the recent editorial in the AJPH by Walter Holland which was critical
of non-physician epidemiologists. Rothman sees the tendency to judge
the author and not the work reflected in recent journal policies on
conflict of interest which require authors to report their funding
sources. He criticized the “frenzy of credential checking as not the
right solution to the problem.” We are getting off the science track
by judging a work on other than its actual content, he said.
To promote objectivity in
science he argued instead for a process of open dialogue. Let’s
presume “the study guilty of bias until proven innocent and the
investigator innocent of bias until proven guilty,” Rothman suggested.
Science vs. Policy
The final lesson Rothman drew
from Graut’s work is the importance of distinguishing epidemiologic
research from policy analysis. In this case, Rothman believes Graunt
took the wrong approach by making a public policy recommendation based
on his data. Graunt had observed that beggars were healthy and he
proposed that a general tax be collected to pay beggars not to go to
work so that they would not displace skilled workers. Rothman was
careful to point out that he is not against advocacy and policy
recommendations made by epidemiologists based on their findings,
however he believes such statements should appear in separate
publications or in a different forum. He believes such recommendations
are not part of etiologic research. “Policies,” he said, “will come
and go. However, knowledge about the causes of disease will be of
enduring significance.”
Published July 1995
|
