J Cbron Dii Vol. 39, No. 6, pp. 411415, 1986 0021-9681/86$3.00+ 0.

00
Printed in Great Britain PergamonJournalsLtd

Editorial

CLINICAL EPIDEMIOLOGY*

WALTER 0. SPITZER?

Strathcona Professor and Chairman, Department of Epidemiology and Biostatistics,

McGill University, Montreal, Quebec, Canada H3A IA2

I DEFINE clinical epidemiology as the study of determinants and effects of clinical decisions.
Clinical epidemiology is not merely epidemiology done by clinicians. Nor does it pertain
to the clinical activities of medically-trained epidemiologists. Clinical epidemiology is not
a clearly demarcated field. As many other scientific pursuits that concern more than one
discipline, clinical epidemiology is best characterized by its spans, its overlaps, its gates and
even by its incursions rather than by its borders.
It is more difficult to define a field with multiple components than to do so for a single
or “purer” area of endeavour. Yet the challenge is not unique: Consider these descriptions
in the biomedical sciences: “clinical pharmacology”, “bio-statistics”, “community health”,
“environmental toxicology”, “family medicine”, “clinical psychology” and “health eco-
nomics”; and in other sciences, “electrical engineering”, “physical chemistry”, “experi-
mental psychology” and “marine geology”.
Such double designations are a taxonomic accommodation to the real world. Our
interest in chemistry may require more specificity than the unmodified term depending on
the relevance (biochemistry or physical chemistry) or depending on applications (chemical
engineering or clinical chemistry).
Curiously, the term clinical epidemiology causes resistance; it even raises eyebrows.
Among those having adverse reactions, the phrase evokes, at worst, suggestions of
unworthy compromise or some sort of dilution of pure epidemiology. At best, clinical
epidemiology is regarded as unnecessary, unusual or even bizarre. I am of the opinion that
clinical epidemiology is a sensible way to describe a distinctive area of activity within a very
heterogeneous discipline. I maintain that its two-word identification, far from bizarre, is
the conventional way of designating a bridge science. In the applied sciences, the joining
of two or more methods or fields is less likely to be a sign of compromised rigour than
the mark of judicious focusing.
In what way is clinical epidemiology distinctive? To develop an answer, I considered
taxonomy first. That, in turn, took me to standard definitions:
“Epidemiology is the study of the distribution and determinants of disease frequency
in man” [I].
“Epidemiology is the study of disease occurrence in human populations” [2].
“ . . . the study of the distribution of a disease or a physiologic condition in human
populations and of the factors that influence this distribution” [3].
“Clinical” is used in endless ways. It can be a modifying adjective that downgrades a
rank, as in clinical professor. It is used to describe the research that a clinician does,
*Based on an address given at the inaugural scientific meeting of the American College of Epidemiology.
tReprint requests should be sent to: Purvis Hall, 1020 Pine Avenue West, Montreal, Quebec, Canada
H3A IA2.

(‘” 196 A 411

412 Editorial

without regard, necessarily, of the methods, content or study subjects of the research. It
can be employed simply to identify the setting where patient-activity takes place. Some
schools for health administrators, for example, refer to the postings of their trainees in
hospitals as “clinical”. Some consensus about the meaning of clinical can be distilled from
standard sources:
“having to do with medical study or practice based on actual treatment and
observation of patients, as distinguished from experimental or laboratory study” [4].
“ . denoting the symptoms and course of a disease as distinguished from the
laboratory findings of anatomical change” [5].
“ . . . qui observe directement (au lit des malades) les manifestations de la maladie”
Fl.
Feinstein [7] extends the principle of direct observation to that of direct responsibility
as he defines clinician:

“Let me define a clinician as a member of one of the healing professions-such as

medicine, osteopathy, and clinical psychology-who takes direct responsibility for
the care of living patients, or who has spent substantial amounts of postgraduate time
in developing his skillful knowledge of such activities. The clinician may be in private
practice, academic research, or administrative work, but his distinguishing character-
istic is a background of observational and therapeutic experience in dealing with sick
people.”

The definitions emphasize direct : direct examination, direct observation, direct responsi-
bility for living patients. Such direct bedside and consulting room activities then become
the ultimate frame of reference for the science of clinical epidemiology because such are
the activities to be guided.
What then is the distinctive mark of clinical epidemiology? In my view, it is epide-
miologic research oriented to the improvement of clinical decisions. My own definition is:

Clinical epidemiology is the study of determinants and effects of clinical decisions.

In contrast, classical epidemiology is oriented to the elucidation of cause. Many

examples of classical epidemiologic work demonstrate that such research can also affect
and does affect clinical decisions. I shall cite only one example, the Framingham study [8].
But the questions and hypotheses of clinical epidemiology arise more directly and
immediately from the daily problems of clinical practice. The resulting research is planned
to provide clinicians with answers they can apply directly and readily at the bedside or in
the consulting room. From question to application the loop is tight. These are some
questions that could concern the clinical epidemiologist:

(1) What is the frequency of application that promotes maximum benefit of Papanicolau
smears in the prevention of cancer of the cervix?
(2) Which subgroups of patients with diabetes mellitus are most susceptible to the
benefits of the insulin pump?
(3) Does spirometry have any role in early detection of chronic obstructive pulmonary
disease or in the management of early manifest chronic obstructive lung disease?
(4) Is there any difference in the effectiveness of home dialysis compared to more
conventional renal dialysis?
(5) Under what circumstances do management decisions of gastroenterologists improve
with use of endoscopic devices?
(6) What is the mix and what is the working relationship of health professionals that
enables the most effective primary health care to a defined population?
(7) Can one develop valid measures of quality of life that permit a scientifically
admissible evaluation of palliative care services?
 Editorial 413

One landmark definition of clinical epidemiology was proposed by Sackett in 1969:

“Clinical epidemiology is the application of epidemiologic biometric methods to the
study of diagnosis and therapy by a clinician who provides direct patient care.”

I cannot accept the constraint that clinical epidemiology is done or can only be done
by “a clinician who provides direct patient care.” I do not believe it matters who does the
study as long as the problem concerns the clinical decision-making process and as long
as the results of the research are immediately useful to a clinician who provides direct
patient care. Non-medical biostatisticians, for instance, can initiate and conduct clinical
epidemiologic research and do so in relevant and elegant ways. I am concerned that
Sackett’s definition sets aside disease aetiology which may affect clinical decisions
(communicable diseases still involve clinicians) and it overlooks prevention in the context
of clinical practice. The few available effective measures of secondary prevention or early
detection of pre-symptomatic chronic disease are undertaken almost exclusively by
clinicians.
The list of authorities invoked in this discussion would be incomplete without reference
to John R. Paul, who I believe was the first to combine the terms in a 1938 paper.

“Epidemiology is concerned with measurements of the circumstances under which

diseases occur, where diseases tend to flourish and where they do not” [lo].

In Paul’s 1938 definition, the word epidemiology was not modified or qualified, but the
context and the concepts of his paper were entirely consistent with his own later better
developed ideas. He wrote, for instance, that the clinical epidemiologist may have similar
interests to those of the health department epidemiologist, but “he is to the statistical
epidemiologist what a gardener is to a farmer” [l I].
In a recent editorial Holland records his opinion, “that the term clinical epidemiology
has served its purpose but is now no longer of any use” [12]. He marshalls two main
arguments to support his rejection. He is against the idea that clinical epidemiology is an
undertaking restricted to clinicians who provide direct patient care (Sackett’s phrase). “It
is not helpful”, he writes, “to describe specifically a small group who practices both
epidemiology and medicine” [12]. He also argues that while ward rounds can be a highly
appropriate medium for teaching epidemiology, that fact provides insufficient grounds to
add “clinical” as a means to partition a part of the field which is specifically important
to practising doctors or which is best studied by clinicians as investigators or collaborators.
Historically, he showed that the origin of the term in the United Kingdom was little more
than an expedient administrative ploy.
I agree with Holland’s arguments and concern but not with his conclusions. The
specificity of clinical epidemiology is not determined by the clinical background of the
investigator. Nor should it be influenced by paedagogic strategies that can be used to
sensitize medical students and physicians to the proper place of epidemiology in the array
of relevant biomedical sciences. The specificity of clinical epidemiology is determined
largely by the category of problems under study, clinical decisions, and by unique
methodological challenges.
The main challenges include difficulties of linking numerator clinical events or patients
with corresponding denominators, the critical choice of proper comparison groups for
observational clinical studies, and the calibration of precise, valid useful measures of
exposure and outcome for use both in experiments and observational studies. Consider
again the seven research questions set forth earlier in this paper. They do suggest distinctive
turf and distinctive technique. Neither the turf nor the technique are likely to be of priority
to the classicist whose skills and interest have prepared him to study aetiology.
Holland dismisses the need for clinical epidemiology to define a distinctive area of
activity. Semantics and expediency are not enough, he argues, and I agree. Yet the purpose
of this position paper is to show that the rationale goes much beyond semantics or
expediency.
414 Editorial

I now turn to some practical considerations. Operationally, research and teaching in

clinical epidemiology can be done by establishing a university department of clinical
epidemiology, as Sackett did at McMaster University in 1968. Alternatively, clinical
epidemiology programmes within more traditional departments of epidemiology can be
organized. In another model, teaching hospitals of medical schools can create clinical
epidemiology divisions such as those organized at the Royal North Shore Hospital of
Sydney, Australia in 1979 or more recently, in 1983, at my hospital, the Montreal General
Hospital of McGill University. Such units maintain vital links with a clinical department,
usually a department of medicine and a university department of epidemiology. Dual
academic accountability can be accomplished smoothly with simultaneous appointments
of key investigators and teachers in the department of epidemiology and a clinical
department. The precise administrative arrangements vary from faculty to faculty and
hospital to hospital.
Ideally, to achieve fruitful investigation with effective diffusion of research findings to
the clinical community, a clinical epidemiology unit or programme should have a critical
mass of methodologists who spend the majority of their time in epidemiologic research
working on projects with latent clinical objectives. Preferably, the methodologists should
spend a meaningful proportion of time in clinical practice, within a discipline in which they
are formally trained, but I do not believe such a requirement is essential.
In the clinical environment where they work there should be a substantial number of
academic clinicians, whose primary role is patient care and clinical teaching but who have
formal training in epidemiology to the point that they can function as investigators for
an important proportion of their time. The clinically-based colleagues can be principal
investigators, collaborators or sophisticated consumers of epidemiologic research. A fertile
research programme in clinical epidemiology depends on two solid bridgeheads; the
span which creates the bridge occurs naturally when each side is competent, credible and
well-trained in the specialty of the opposite side.
C. M. Fletcher (cited by J. R. Paul) wrote, “If the work of clinicians and epidemiologists
is indeed in continuity then it is essential that they should be in continuous professional
contact” [l 11. I believe that the lack of many functional bridges between clinical disciplines
and epidemiology reflects inadequate contact frequently reinforced by physical barriers in
medical campuses. The result is a very limited understanding of relevant research questions
on the part of epidemiologists and not even awareness of the methodological arma-
mentarium of the epidemiologist on the part of clinicians. Important research questions
cannot fail to arise when frequent contact is promoted. Diffusion and application of useful
findings is facilitated when there is close interaction among methodologists and clinicians.
The impact of epidemiologic inquiry on decision-making for the patient then approaches
efficiency.
Clinical epidemiology is unlikely to be a passing fad that competes unfairly in the short
run with other worthy pursuits of the epidemiologist for resources and which puts other
time-tested strategies in jeopardy. Traditional public health and occupational health still
need classical epidemiology as their basic science. Epidemiological methods and statistical
methods must continue to be developed. Knowledge in the realm of infectious diseases
needs to have the frontiers pushed back further with epidemiology in a strategic role. The
search for cause through population studies should never be encumbered.
All strategies of sound epidemiologic research need no further justification than as an
expression of human creativity. Einstein wrote, “Do not stop to think about the reasons
for what you are doing, about why you are questioning. . . Curiosity has its own reason
for existence. Never lose a holy curiosity” [13]. Even though my personal commitment is
to research and practice in clinical epidemiology, I can support and will support the pursuit
of holy curiosity in any of the subdisciplines of epidemiology, not for any particular
pragmatic reason but because it seems right. But I would like to express a concern that
we as epidemiologists reach out to the clinician less reluctantly and more effectively than
we have done in rheumatology, in family medicine, in nephrology, in neurology and in
several other critically important clinical disciplines. I will use one last definition by
Thomas Addis, a giant among clinicians.
 Editorial 415

“A clinician is complex. He is part craftsman, part practical scientist, and part

historian. . . . It is only if we look at him when he is working with his patients that
we find him single-minded. Then he is a wholly pragmatic and utilitarian. His only
design is to bring relief, and he is not at all scrupulous about how he does it” [14].
Let the designs of some of us in epidemiology, some of the time, be also directed to bring
relief to the patient and even to the clinician. In so doing, we need not surrender our
scientific scruples.

REFERENCES

1. MacMahon B, Pugh TF: Epidemiology. Principles and Methods. Boston: Little, Brown and Company, 1970.
P. 1
2. Friedman GD: Primer of Epidemiology. 2nd edn. Toronto: McGraw-Hill, 1980. p. I
3. Lilienfeld AM: Foundations of Epidemiology. New York: Oxford University Press, 1976. p. 3
4. Webster’s New Twentieth Century Dictionary (unabridged). 2nd edn. Collins World, 1977. p. 339
5. Stedman’s Medical Dictionary. Baltimore: Williams & Wilkins, 1976. p. 288
6. Petit Robert: Dictionnaire. Paris: S.N.L., 1976. p. 293
I. Feinstein AR: Clinical Biostatistics. Saint Louis: C.V. Mosby, 1977. p. 7 (footnote)
8. Dawber TR, Kannell WB, Lye11LP: An approach to longitudinal studies in a community: The Framingham
study. Ann NY Acad Sci 107: 539-556, 1963
9. Sackett DL. Clinical epidemiology. Am J Epidemiol 89: 125-128, 1969
IO. Paul JR: Clinical epidemiology. J Clin Invest 17: 539-541, 1938
II. Paul JR: Clinical Epidemiology. Revised edn. Chicago/London: University of Chicago Press, 1966. pp. I@-1I
12. Holland W: Inappropriate terminology (editorial). Int J Epidemiol 12: 5-7, 1983
13. Strauss MB (Ed.): Familiar Medical Quotation. Boston: Little, Brown and Company, 1968. p. SOS(a)
14. Addis T: Glomerular Nephritis, Diagnosis and Treatment. New York: Macmillan, 1949