| |
Epi Wit & Wisdom Resources
The Lay Epidemiologist’s
Required Reading List (1 of 2)
By Karyn Pomerantz
In February, a reader of the
epidemiology listserv administered from Montreal asked list members
what books they would recommend to patients and others interested in
health research. Many people could ask this question as they sift
their way through conflicting news stories about the dangers of
electromagnetic fields; the benefits of beta carotene; and the
uncertainties of prostate screening. None of these stories are
supported with neat, orderly evidence and conclusions. As New England
Journal of Medicine editor Marcia Angell has written,
“Health conscious Americans increasingly find themselves beset by
contradictory advice. No sooner do they learn the results of one
research study than they hear of one with the opposite message. They
substitute margarine for butter, only to learn that margarine may be
worse for their arteries. They are told to eat oat bran...but later
learn that the bran they dutifully ate may be useless...(NEJM 1994;
331(31):189 - 190).
For some readers, these stories
are merely interesting. For others who are considering participating
in clinical trials, in environmental advocacy in their community, or
on a health advisory panel, these stories pack a personal wallop.
Epidemiologists themselves may be involved in advocacy movements or
research in which they can educate the public on epidemiological
principles. Issues of data sources, measurement bias, and
generalizations are not only the concerns of methodologists. Consumers
and policy makers also need to understand these issues which often
have an impact on policy and personal health practices.
Epidemiologists can recommend
the following books and articles to their friends who want to
understand epidemiological studies. This is by no means a
comprehensive list, and many readers may want to offer their favorites
to the Epi Monitor. As noted below, these books and articles vary by
content and scope, sophistication, style, and organization.
1. “Non-Trivial Pursuits,”
Nutrition Action, October 1994: 1 - 7
The shortest and most visual of
the articles, this piece depicts epidemiological studies (non-trivial
pursuits) as a modern chutes-and-ladders board game. “Welcome to that
zany, madcap world of epidemiology, where you, the ambitious
scientist, attempt to unequivocally prove your theory of diet and
disease. But watch out for those nasty Pitfalls! They’re out to
undermine your work at every turn!” Our ambitious scientist advances
along the game board as he attempts population studies, surveys, case
reports, case-control studies, and cohort studies. Along the way lurk
the “Common Pitfalls” of confounding, chance, misclassification, and
bias. This is a very clever and well illustrated explanation of the
different types of study designs with their likely weaknesses. It is a
wonderful snapshot of epidemiology for people who “feel like a
ping-pong ball” in the face of contradictory health reports.
2. “How Numbers Can Trick You,”
Barnett A. Tech Review, October 1994: 39 - 45
Barnett has written a short
simple guide to the common “Six Deadly Sins of Statistical
Misrepresentation” reported in the popular press. Covering the diverse
fields of health, aviation, and criminology, he demonstrates with
charts and simple calculations the problems when researchers
generalize from non-random samples, report overall averages without
considering confounders such as age-groups, cite odds and
probabilities inconsistently and improperly, and use data sources
inappropriate for the question being asked.
3. “Secondhand Smoke: Is it a
Hazard?” Consumer Reports, January 1995 26-33
While responding to the tobacco
industry’s attempts to debunk EPA’s condemnation of secondhand smoke,
Consumer Reports interviews Charles Hennekens,
Julie Buring and other epidemiologists to explain the
concepts of statistical significance, dose-response relationships,
confounding factors, study design, and causality. The most important
contribution in this article, however, is its attention to
non-epidemiological bias, the bias inherent in studies paid for and
promoted by the tobacco industry to protect its profits. Attempting to
control the anti-smoking activists, the Imperial Tobacco Limited,
Canada’s largest tobacco company, conducted a study in which they
drafted an attack plan on the issue of secondhand smoke. “The
challenge will be to find a sympathetic doctor who can be demonstrated
to take a largely independent stance,” they wrote. Lisa Bero,
a University of California, San Francisco health policy analyst, also
relates her studies of tobacco industry organized symposia through
which non-peer-reviewed articles enter the scientific literature. This
is a wonderful article for demonstrating the application of
epidemiological studies to a specific issue (tobacco) as well as
illustrating how corporate interests can deliberately misrepresent
scientific evidence to deny the risks of confirmed health hazards.
4. News and Numbers: A Guide to
Reporting Statistical Claims and Controversies in Health and Other
Fields
(A project of the Center for
Health Communication, Harvard School of Public Health), V. Cohn, Iowa
State University Press, 1989
Cohn, the health reporter for
the Washington Post for many years, writes for journalists who report
the news on health and medicine to the public. It is a wonderful
introduction to critical appraisal skills for the public as well. Cohn
helps the reader appreciate the scope of scientific endeavors, from
the exploratory study to a conclusive body of evidence, as well as the
details of statistical significance and power, rates, and sensitivity
and specificity. As a journalist, he explains how easy it is for
reporters to write stories that overestimate positive results, omit
the context of a study such as its size or sponsorship, or simplify
complex relationships. Throughout the book, he poses numerous
questions that journalists and their readers can ask: What makes a
study honest? What is the evidence? Who were the subjects? Where did
the money to support the study come from? Do the conclusions make
sense to me?
5. How to Lie With Statistics,
D. Huff, NY: WW Norton & Co., Inc., 1954
This is certainly an oldie but
goodie. The book is written in a breezy style with such chapter titles
as “The Sample with the Built in Bias” and “How to Talk Back to a
Statistic,” and illustrated by many irreverent cartoons. Huff writes
for the general public who read statistical results in newspaper and
magazine stories and advertisements. As he explains in the
introduction, “This book is a sort of primer in ways to use statistics
to deceive. It may seem altogether too much like a manual for
swindlers. Perhaps I can justify it...the crooks already know these
tricks; honest men (sic) must learn them in self-defense.” The
“tricks” he covers include sampling bias, the perils of graphs and
charts (the “gee-whiz graph), imprecise measurements, deceptive drug
advertising, incomparable comparison groups, and faulty assumptions of
causality. While these issues are covered in many other books, this
book has the clearest and most engaging explanations, illustrated by
numerous, commonplace examples.
6. A Mathematician Reads the
Newspaper, J. A. Paulos, NY: Basic Books, 1995
Writing for the same audience
and in the same humorous style as Huff, Paulos explains the misuses
and misinterpretations of statistics in the popular press. His subject
range is just as broad, from economics to health, but obviously more
contemporary and illustrated more by quantitative examples. Paulos
structures his book like a newspaper, progressing from national news
through business and lifestyle to health, science and sports. In the
section on health, he cautions readers about the “availability error”
in which people remember more sharply very emotional and vivid news,
such as the number of annual deaths from cocaine (8,000) versus the
more commonplace annual deaths from tobacco (400,000). Also covered
are the statistical concepts of Type I and II errors, sampling,
statistical significance and confidence intervals.
Published April 1996
|
|
