The
American Statistician, November 2004, Vol. 58, No. 4
ActivEpi
1.1 (2nd ed.).
David
G. KLEINBAUM.
Macintosh
and Windows platforms, ISBN: 0-387-21015-6.
ActivEpi
Companion Textbook: A Supplement for Use with the
ActivEpi
CD-ROM.
DavidG.KLEINBAUM, KevinM.
SULLIVAN, and NancyD. BARKER.
New
Students in introductory epidemiology classes often have a wide range
of
backgrounds
in quantitative methods. Although one can teach epidemiology
without
reference to formulas or even to numbers, at a certain point it becomes
necessary
to resort to mathematical symbols. At this point it is important to remember
that
recent educational theory suggests that students learn best by participating
in
their own education rather than by passively receiving information
(Velleman
and Moore 1996).
It is on this educational principle that Kleinbaum’s
introductory
epidemiology text on CD, ActivEpi,
is based.
The sequence of 15 chapters follows the order in the Kleinbaum,
Kupper,
and
Morgenstern (1982) book Epidemiologic
Research (KKM),
although the
level
of presentation is introductory rather than intermediate as in KKM. Topics
include
study design, measures of effect and impact, validity, and statistical
analysis
of epidemiologic data. Also offered is a review of statistical inference
that
is quite useful to students who have not yet studied statistics or do not
remember
what they learned when they did study it. Information is presented
in
videos, lectures, study questions, interactive quizzes, data analysis
exercises
[using
Data
Desk,
a statistical package also used on the ActivStatsCD(Velleman
2004),
adapted here to the special needs of epidemiological analyses], homework
questions,
and Web links that provide background information on datasets and
refer
students to otherWeb siteswith
epidemiological information. Lectures stop
after
important ideas are presented, and proceed at students’ discretion giving
them
time to consider questions about previous material. For example, within
the
lesson on confounding, after the distinction between confounding and effect
modification
is described, the students are presented with examples that may
include
confounding or effect modification. When the lecture is restarted, the
narrator
discusses each example in detail providing a rationale for each answer.
The
students are thereby allowed to check their understanding of the distinction
between
confounding and effect modification immediately after learning it. It is
this
active partnership between student and imaginary professor that makes the
CD
so useful, particularly for students not accustomed to studying quantitative
topics.
Many chapters contain data-analysis sessions in which the student is
given
a
dataset and required to use Data Desk to apply and interpret a newly learned
analytic
tool. Epidemiology templates allowestimates of
crude and adjusted riskand
odds-based
measures of association, crude and adjusted incidence density
measures,
and etiologic and preventive fractions. Also available is a template
to
analyze matched-pairs, case-control, or cohort studies. Finally, a Data Desk
template
is provided for adjustment of effect measures for exposure or disease
misclassification.
Independent of the format, there are some particularly lucid
expositions that
will
be of great benefit to students. These include an explanation of the
distinction
between
incidence and prevalence and another of the conditions under which
the
rare disease assumption is not required for the odds ratio to approximate
the
rate ratio. Also excellent are Kleinbaum’s
explanations of age adjustment,
selection
bias, confounding involving several risk factors, control of extraneous
factors,
and matching.
One problem with the lectures is that frequent stopping for questions
makes
it
hard to return to a point for later review. This problem is addressed to some
extent
by the extensive index, glossary, and table of contents that places the
reader
at the beginning of the desired section. Another solution to this problem
is
to purchase the accompanying text that closely follows the CD but has all the
advantages
of a book (easy to find specific sections, no computer is needed to
read
it, etc.). I therefore recommend that students buy both the book and the CD.
Another problem results from the need to simplify material for an
introductory
level.
For example, Berkson’s bias is a hospital selection bias based on over
representation of patients with both exposure and disease in hospital
populations.
If the degree of this over-representation is not equal between cases and
controls,
then
an odds ratio from a hospital-based case-control study may be distorted
(Schwartzbaum,
Ahlbom, and Feychting
2003).
However,
Boring
(1989) suggested that Berkson’s bias may not be
present when incident
hospital
cases are used with community controls; yet, in the chapter on selection
bias,
Kleinbaum uses a study of hospital-based lung
cancer cases and community
controls
to illustrate Berkson’s bias. It may be that
cases in Kleinbaum’s example
are
prevalent cases (although prevalent cases are rarely used in case-control
studies)
or that Flanders et al. were not correct and that Berkson’s
bias is actually
present
in this situation. However, even introductory students should be told that
the
presence of Berkson’s bias in the particular
example given is subject to
debate.
The problems cited above, however, are minimal and do not detract from
the
overall richness and creativity of ActivEpi
and
the companion text. The CD
and
book may be used as an adjunct to a class, for self-teaching, or for review.
Overall,
I recommend ActivEpi
because
it is fun to use. A final word to the wise:
check
the ISBN number to make sure that you are buying the most recent version
of
ActivEpi.
I would like to acknowledge assistance in reviewing ActivEpi
from
the following
students:
Hoda Jradi, Ryan
Mayes, Nicholas Miceli, Susanne Scott, and
Melissa
Senter.
Judith
SCHWARTZBAUM
The
REFERENCES
Differential
Hospitalization Rates in Incident Case-Control Studies,” Journal
of
Clinical Epidemiology, 42,
395–401.
Kleinbaum,
D., Kupper, L., and Morgenstern, H. (1982), Epidemiologic
Research:
Principles
and Quantitative Methods,
Schwartzbaum,
J., Ahlbom A., and Feychting
M. (2003), “Berkson’s Bias Reviewed,”
European
Journal of Epidemiology,
18, 1109–1112.
Velleman,
P. (2004), ActivStats
2003–2004 Release,
Wesley.
Velleman,
P., and Moore, D. (1996), “Multimedia for Teaching Statistics:
Promises
and Pitfalls,” The
American Statistician,
50, 217–225.