Uploaded by al.n.yakovlev

Medical usage and abusage prevalence and incidence

advertisement
THE
LITERATURE
OF
MEDICINE
Medical Usage and Abusage: "Prevalence" and "Incidence"
GARY D. FRIEDMAN, M.D., S.M. in Hyg., F.A.C.P., Oakland, California
In view of the frequent misuse of the term "incidence/' the
concepts of incidence, prevalence, and period prevalence
are defined and illustrated. The importance of restricting
"incidence" to the rate of development of new events in a
population or a group of patients, per unit of time, is
stressed. Clearly distinguishing incidence from prevalence or
simple proportion can help both in the understanding and
the communication of findings in patient groups.
I T HAS BEEN SAID that the easiest way to distinguish a clinician from an epidemiologist is by the clinician's incorrect
use of the term "incidence." Ignoring the highly useful
distinction between the concepts described by "prevalence"
and "incidence" is, however, not restricted to clinicians. A
forceful example of this problem came to me in a set of
galley proofs from the Annals of Internal Medicine. The
copy editor had changed "prevalence" to "incidence"
wherever it appeared in the paper. Thus, even this presumed expert in grammar and usage, whose job is to make
sure that physician-authors use the English language correctly and effectively, had erred about "incidence." This
copy editor may have read too many clinical manuscripts
with misuse of "incidence."
Soon after this occurred, Dr. Huth, the Editor, asked me
to write a brief note reviewing the correct definitions of
these measures of relative frequency. After some years of
reading thousands of manuscripts, he had concluded that
the imprecise use of "incidence" and "prevalence" was
probably the most frequent error in vocabulary among
physician-authors, although it might be rivaled by confusion between "case" and "patient." Apparently the repeated presentation of these concepts in courses in preventive medicine and the pleas by epidemiologists and an
expert World Health Organization committee (1) have not
had much effect.
To verify this impression, I sampled the use of "incidence" in the latest edition of a widely used textbook of
medicine. On every tenth page I looked for subsections
headed "Incidence" or "Prevalence" in relation to specific
diseases. Within the six subsections found, "incidence" was
• From the Department of Medical Methods Research, Kaiser-Permanente
Medical Care Program, Oakland, California.
502
Downloaded From: https://annals.org/ by a University of Texas User on 10/12/2018
used correctly in one, was used probably to mean prevalence in one, was used to mean proportion of hospital
admissions in one, and was not used in the text in the other
three. Where "incidence" was not used in the text, its use
in the subheading seemed to refer to a discussion of some
aspects of the distribution of occurrence of the disease in
the population—for example, "occurs at any age" or "is
common in most wars."
In epidemiology, "incidence" has a very specific meaning, pertaining to the development of new cases of a disease in a population free of the disease. Thus, an incidence
rate is the fraction of the population at risk who develop
or acquire the disease per unit of time (hours, days,
months, years, or whatever period is most convenient).
Or, put in terms of an equation:
Incidence rate =
number of new cases of a disease
per unit of time
total population at risk
If a person already has the disease, he should obviously
not be included in the denominator population at risk.
In contrast to the process of disease development as represented by "incidence," "prevalence" (or "point prevalence") refers to a static picture or "snapshot" of the
number of persons who have a disease in a population at
one point in time, such as a certain day when a survey is
made. This point does not have to be in calendar time but
can be at an event in people's lives, such as at birth, or at
a special screening examination, or at autopsy, each of
which occurs for different members of a population at
different points in calendar time. The prevalence rate
(sometimes called the prevalence ratio) is merely the number of persons with the disease divided by the number in
the population under consideration. It is a simple fraction
or proportion:
number of existing cases
Prevalence rate =
——:
=—;
It may not be strictly correct total
to refer
population
to this simple proportion as either a "rate" or a "ratio" but "prevalence rate"
seems currently to be the most commonly used term (2).
The concept of "period prevalence" is somewhat different. If the ascertainment of existing cases spans a long
period, such as a year, the cases detected will include those
prevalent at the beginning of the period plus those new
"incidence cases" developing during the period. When
period prevalence is used, the time period (for example,
Annals of Internal Medicine 84:502-504,1976
one year, 10 years) should be specified, for example:
all cases existing during the year
1-year period prevalence =
—
;
;—:
total population
If the period is long and the population denominator
changes in size during that time it is customary to use the
actual or estimated population size at the midpoint of the
interval. This applies also to determinations of incidence
over long time periods.
Now, it so happens that clinicians (and pathologists)
have other important groups to talk about beside populations, which are the chief concern of the epidemiologist.
Examples of such groups may be patients with a particular
disease, or patients treated with a certain drug, or all persons seen in a clinic during a certain year, or patients who
come to the emergency room on Monday mornings, or all
patients in a particular hospital who had postmortem examinations in the past 20 years. In describing such groups
it is frequently desired to refer to a proportion or fraction
who have a particular characteristic, who developed it over
a period of time or who are going to develop it in the
future. Here, the indiscriminate use of the word "incidence"
in referring to all these proportions is inappropriate. It
would be helpful to the clinician to adopt the epidemiologist's distinction and reserve "incidence" for the rate of
development per unit of time and "prevalence" for describing that fraction of the group with the characteristic at a
particular point in time.
Let us apply these concepts to a clinical situation. Suppose a physician has been following up 150 women with
asymptomatic bacteriuria detected during a health checkup.
He first obtained further information and ordered additional studies on these women. Forty had had previous
episodes of symptomatic cystitis. Intravenous pyelography
showed that 20 women had congenital anomalies of the
urinary tract, such as double ureters. Creatinine clearance
tests showed that 10 women (including 5 of those with
urinary tract anomalies) had normally low values. For
simplicity let us assume that all women were not treated
initially but were carefully followed up at regular intervals
for 2 years to assess the development of complications and
changes in renal function. During the 2 years, 30 women
had one or more episodes of frank symptomatic urinary
tract infection, which were treated. There were 40 episodes
in all—1 each among 20 of the women and 2 each among
the other 10. At the end of the 2 years, abnormally low
creatinine clearance values were found in 15 women. Of
those, 8 were among the 10 with low values originally,
and 7 were previously normal. Two of the 10 with low
values originally were now normal.
With the present widespread indiscriminate use of "incidence," this physician in reporting his findings at a conference is likely to stand up and state that, "Among this
group of women with asymptomatic bacteriuria, the incidence of congenital anomalies of the urinary tract was
. . . and so forth and so on."
How should these data be summarized? Table 1 shows
how these observations can be described, the numbers to
be used, and correct terminology for the findings. Even
though prevalence or incidence is being presented, it is not
always necessary to use either term. As will be seen in
some of the examples in Table 1, it may be simpler and
less clumsy to refer merely to a fraction, proportion, percentage, or a rate that is explicitly described. If, as often
happens, follow-up durations for members of a study group
are not the same, the investigator can take advantage of all
available follow-up data and compute incidence rates using
person-years or person-months in the denominator instead
of persons ( 3 ) .
It should be stressed that the time unit is an essential
component of the incidence rate. Without specification of
the time period covered the use of "incidence" to refer
merely to a fraction of patients who develop an episode
or finding has little meaning and can be quite misleading.
For example, a paper on implantation of prosthetic cardiac
valves may report this observation: "Among patients who
survived through the first 2 postoperative weeks, the incidence of bacterial endocarditis was 2 % . " Without a
statement as to the time through which this observation
was made, we cannot evaluate this rate or compare it
to the findings of others. Two cases of endocarditis per
hundred patients in 2 months would be a relatively high
incidence rate—12 cases per hundred patients per year,
on an annual basis; whereas 2 cases per hundred patients
in 2 years would be a relatively low incidence rate—only
1 case per hundred per year. If it is necessary to report
such clinical observations without reference to a period of
time the facts should be stated simply without using the
word "incidence," to avoid implying that a rate per unit of
time has actually been measured. A more accurate version
of the above statement would be: "Among patients who
survived through the first 2 postoperative weeks, 2 %
later developed bacterial endocarditis."
If an author or speaker would make careful distinctions
among "incidence," "prevalence," and related measures of
relative frequency, this would not only improve communiction with others, but would help assure his own full understanding of exactly what it is he is trying to say.
ACKNOWLEDGMENTS: Received 26 September 1975; revision
accepted 12 January 1976.
• Requests for reprints should be addressed to Gary D. Friedman,
M.D., 3779 Piedmont Avenue, Oakland, CA 94611.
References
1. Terminology and nomenclature: prevalence and incidence. Bull
WHO 35:783-784, 1966
2. ELANDT-JOHNSON RC: Definitions of rates: some remarks on their
use and misuse. Am J Epidemiol 102:261-21 \y 1975
3. FRIEDMAN GD: Primer of Epidemiology. New York, McGrawHill, 1974, pp. 107-108, 116
Friedman • Literature of Medicine
Downloaded From: https://annals.org/ by a University of Texas User on 10/12/2018
503
Table 1 . Measurements of Relative Frequency from the Hypothetical Study of Women with Asymptomatic Bacteriuria Described in Text
Type of
Measurement
Numbers to be Used
Past history of cystitis
Prevalence
40 with positive history _ Q 27 o r 21°/
150 women
Twenty-seven percent gave a history of
symptomatic cystitis.
Congenital urinary
anomalies
Prevalence
20 with anomalies n . . -~
—
= 0.13 or 13%
150 women
The prevalence of congenital anomalies
was 13%. Or 13% had congenital
anomalies.
Diminished renal function
initially among all
the women
Prevalence
10 with low creatinine
The prevalence of abnormally low
creatinine clearance was 7%. Or 7%
had abnormally low creatinine
clearance.
Diminished renal function
initially among women
without congenital
anomalies
Prevalence
Women developing
symptomatic urinary
tract infection
Incidence
Development of episodes
of symptomatic urinary
tract infection
Incidence
40 episodes . —^
in 2 years
150 women
= 27 episodes/100 women in 2 years
or 13.5/100 per year
The incidence of episodes of urinary
tract infection was 13.5 episodes per
100 women per year. (Note that episodes are not a fraction or proportion
of women and that there can be more
episodes than women.)
Development of diminished renal function
Incidence
7 women with normal creatinine
clearance initially and low
values at the end of follow-up
140 women with normal creatinine
clearance initially (and at risk
for developing a low value)
Among women with normal renal function initially, the incidence of abnormally low renal function was 5% in 2
years or 2.5% per year.
Area of Concern
clearance
Suitable Description of Results
0.07 or 7%
150 women
5 with low creatinine
clearance and without
congenital anomalies
130 women without
congenital anomalies
30 women developing
1 or more episodes
150 women
= 0
04 0 r 4 ^
„
—u.z or MV ye m
2 years or
10% per year
Of the women with asymptomatic bacteriuria and no congenital anomalies,
4% had low creatinine clearance.
One or more episodes of symptomatic
urinary tract infection occurred in
20% of the women in the 2-year period
for an incidence rate of 10% per year.
— 0.05 or 5% in 2 years or
2.5% per year
Improvement of diminished renal function to
normal
Incidence
2 women with low creatinine clearance
initially and normal values at the
end of follow-up
10 women with low creatinine
clearance initially (and at risk for
developing a. normal value)
- 0.2 or 20% in 2 years or 10%
per year
Of the women with diminished renal
function initially, 20% improved to
normal levels in 2 years. The rate
of return to normal was 10% per year.
Presence of diminished
renal function at any
time during the 2-year
period
Period prevalence
17 women with low
creatinine clearance
(10 initially plus
Of all the women, 11% showed evidence
of diminished renal function at some
time during the 2-year period.
,7newcases)
150 women
0.11 or 11%
(Note that this cannot be divided by two
to give the 1-year period prevalence
because the prevalence at the beginning of the period was already 7%.)
504
Apr//1976 • Annals of Internal Medicine • Volume 84 • Number 4
Downloaded From: https://annals.org/ by a University of Texas User on 10/12/2018
Download