Quality Donald L. Renfrew, Frank H. Weigelt,
MD MA
#{149} Edmund A. Franken, Monzer M. Abu-Yousef,
Jr, MD MD
#{149}
Kevin
S. Berbaum,
#{149}
Error in Radiology: Classification Lessons in 182 Cases Presented at a Problem Case Conference’ The authors review and classify errors in 182 cases that were presented at problem case conferences between August 1986 and October 1990. Errors were classified by means of a system developed 20 years ago and by means of a system developed within the past several years. The authors found that sources of error have changed very little. Errors usually involved failure to consult old radiologic studies or reports, limitations in imaging technique, acquisition of inaccurate or incomplete clinical history, localion of a lesion outside the area of interest on an image, lack of knowledge, failure to continue to search for abnormalities after the first abnormality was found, and failure to recognize a normal biologic variant. Errors included 126 perceptual errors (64 false-negative, 15 false-positive, and 47 misclassification errors) and 56 mishaps, including 38 complications and 18 communication errors. In seven cases nonperception errors occurred because established departmental routines were not followed, and in nine cases a new departmental routine was established after a cornplication occurred. Departmental poiicy exerts less effect on perception and interpretation errors. Index
terms:
performance partmental Radiology
I
Diagnostic Radiology
management
Table 1 Classification
Cause
and
of Error
Explanation
Complacency
II
Decision-Analysis
Faulty
reasoning
Lack
of knowledge
False-positive
Term error
True-positive reading, but misclassification
category. III
IV
The
finding is seen but is attributed to the wrong cause because of a lack of knowledge on the part of the viewer. The finding is missed. Underreading may result from failure to isolate important material or from satisfaction of search. The lesion is identified and interpreted correctly, but the message fails to reach the clinician. The lesion was not present on the image obtained, even in retrospect. This may be secondary to limitations of the examination or to an inadequate examina-
Underreading
V
Poor
VI
communication
Miscellaneous
True-positive
reading,
but misclassification
False-negative
error
...
False-negative
error
tion. Vii
Complications*
Untoward
events
happened
during
...
the course of examination, which was most frequently encountered during invasive procedures. Source-Adapted * We have added
from Smith (4). this class, which was not in Smith’s
scheme.
radiology, observer and radiologists, de#{149} Quality assurance
1992; i83:i45-i50
From the Department
of Radiology, the University of Iowa College of Medicine, Iowa City. Received May 17, 1991; revision requested June 21; final revision received October 17; accepted
November 4. Address reprint requests to D.L.R., Department of Radiology, Rush-Presbyterian-St Luke’s Medical Center, 1653 W Congress Pkwy, Chicago, IL 60612. 1992
and Examples
Errors of overreading and misinterpretation, in which a finding is appreciated but is attributed to the wrong cause. Errors of overreading and misinterpretation, in which the finding is appreciated and interpreted as abnormal but is attributed to the wrong cause. Misleading information and a limited differential diagnosis are included in this
E
occurs frequently practice of medicine
in the
RROR
(1). In addilion to the mistakes common to all areas of medicine, perceptual error in diagnostic radiology is particularly prevalent, as the human perceptual
RSNA,
PhD
of Error
Class I
Assurance
system
error ogy
is imperfect.
is common has
been
errors
cases
of interpretation
sents
an
and
other
noted images.
analysis
mishaps
for
in over
radiolyears
(2),
one-third
of
of ambiguous
This
article
of perceptual
that
occurred
preerrors
in
radiology over
MATERIALS
AND
Since August 1986, ences have been held
perceptual
recognized
with
diagnostic
That
in diagnostic
the diagnostic at our institution
department a 4-year period.
METHODS
problem case conferby the Department
of Radiology at the University of Iowa; the conferences, on the model of Wheeler (3), are held 10 times per year. Subspecialty divisions
within
radiology)
are
the department assigned
a rotating basis, so that tributes approximately
to present
(pediatric cases
on
each division confour cases per year.
145
182 Errors
Comp’acency Faulty Lack
Reasoning
of
I
Knowledge
126 Perceptual/Cognitive
56 Other
Underr.adng Poor
I
Communcat#{232}on
15 False Positive Tab$e4
UiscelI&ous Complicat#{232}on
47 Misclassification TableS
260
200
r
150 Smiths
100 60 Ssr#{149}s
0
1020 P,ob.m
E
30 40 60 Cass Conf.
I
60
64 False
Negative
is
Ioame
undercalls
19
severity
7 department routine not followed Table6
overcalls
Figure 1. Graph depicts distribution of errors in Smith’s original 437 cases (4) and in our 182 cases. 23
Problems sources.
are identified The review
aminations terpreted
such
as magnetic
imaging
malities
may
that
are
on routine leagues
of ex-
apparent
often
A logbook provides
associated morbidity
about the interpretations
errors, any that resulted
Table
2
Errors
Caused
cause
important
a single cause The radiologists cording
I.
was 12
of error,
they
18
types 77
with
chose
188
aging
with
marizes
newer
most
Smith’s
analysis error.
was
The
a 4-year radiologic
reported
146
#{149}
Radiology
and
therefore
0
These
administration
of three
a transverse
of imaging
colon
the
interpretation fracture
was
an inherently
contrast
material
was missed on first arteriogram because carotid arteriography. of the spine was missed at MR imaging was
not
diagnosed
beam
and the vessel
large
margin
esophogram
enemas
did
of failure
to per-
because
of the
chosen.
disease
the Doppler
water-soluble
tumor.
sequences
venoocclusive
are the case numbers
representative
failed
in our logbook
were
on initial
almost
Doppler
studies
perpendicular,
be-
a situation
of error.
to show
and
a lealc the clinician
are used
in this study, greater than
the
series
incidents
that
were
per-
department of case selection necessarily
does
in the de-
represent
considered
those by
radiologists and by our clinical to be the most serious errors.
the
staff
colleagues
refused
for anonymous
191 cases 182.
were
to have bar-
identification
presented
of patients.
in the conferences,
Figure i presents classification Smith’s data, as well as our own, means tion on methods
sification perceptual
of Smith’s
scheme.
of by
Reclassifica-
the basis of contemporary of perceptual-cognitive
demonstrated error and
tributed sents rors
to other
causes.
clas-
126 cases of 56 mishaps at-
Figure
the classification of our on the basis of contemporary
2 pre182
er-
methods.
Perceptual-Cognitive
RESULTS in
occurred over were
of all error
partment,
er-
herein
after
Although there were i82 errors classified and, thus, the case number is occasionally
not
as
perceptual-cogfalse-negative, and other error
period in which examinations
on plain radiographs; was performed, and
on the such
communication
formed in the radiology our institution. Although arbitrary
we
terminology.
182 cases
was
made
methods
and
meth-
to his
I also indicates the closest to Smith’s classes that exists
decision-analysis during 700,000
1 sum-
to classify perceptualSuch classification in-
(complications
Table equivalent
Table system;
also
of contemporary
we
and im-
“complications”
volved dividing cases into nitive error (false-positive, or misinterpretation error) rors).
of what
classification
of contemporary
decision cognitive
pracnot per-
radiology,
modalities.
added the category scheme. 2. Classification basis
of gen-
his colleagues in private and his colleagues did
medicine,
basis
ium administered, and a leak was missed. Computed tomography (CT) failed to demonstrate an intraabdominal abscess, which was documented with labeled white blood cell study. Salter-Harris undisplaced type I fracture was missed on initial radiograph. Postoperative upper CI examination was interpreted as negative for leak, but abdominal CT demonstrated the leak. Plain radiographs showed normal findings in a neonate with hypoplastic left heart syndrome. Eight-millimeter-thick sections acquired on an ultrafast chest CT scan failed to show a double aortic arch.
94
Smith and tice. Smith
practice
now call interventional
the
or Technique
was not apparent medicine study
a nuclear
Water-soluble
85
171
nuclear
Hepatic
cause
by Smith
in the
of the scaphoid
Intracerebral aneurysm form selected internal Disk abscess and erosion
30
eral radiology, classified 437 errors that occurred during a special 2-year study of
form
Imaging
systems.
developed
errors
Examination
of
qualified;
not show
144 156
scheme,
(4) for
on
demonstrated.
in
One
1967
Limitations
Fracture
on the single
classification
by
of 182 errors
of
Search
Description
5
after discussion of the case. classified the cases ac-
to two
classification
Case*
and/or the error,
according to the single most important cause of error. In the few cases in which most
depicts
11 SatIsfaction
of Knowledge
ex-
and steps that were taken to minimize the chance of recurrence of the error. Two radiologists (D.L.R., E.A.F.) independently classified the i82 cases presented at problem case conferences between August 1986 and October 1990
disagreed
chart
Search
I
I 2 Lack
out
radiologic and
mortality from
the radiologists
Flow
I S Abnormality Outside Area of Primary Exam
col-
in pointing
compiled from each cona brief history of the pa-
tient, information amination with
41 Specific Cause for False Negative
4 Misdirected
11 Poor Films of limitation of exam Table 3
9 new department routine implemented Table 7
retrospect”
Clinical
helpful
2.
18 Communication Errors Table S
ods.
abnor-
“in
No Specific Caine ot False Negative
I S Old Films/Reports Table2
Figure
resonance
demonstrate
radiographs. are
errors. ference
a variety imaging
that occurs when cases are inoften discloses errors. Advanced
modalities
(MR)
from of prior
Complicatmos
38
I
,
Errors
The i26 cases of perceptual-cognitive error included 64 false-negative, is false-positive, and 47 misclassification errors. False-negative
error-A
ple of a false-negative monary
nodule
that
classic
error is visible
exam-
is a pulin retro-
April
1992
t.,:
.
,
* .
,
.
.
:
b. Figure
3.
anatomic intense weeks
Limitation
snuff
of technique.
box. (a) Ulnar-deviation
increased
after
Images
radiotracer
the injury
are
oblique
localization
demonstrates
in the
a fracture
of an
18-year-old
radiograph region
gymnast
who
of the scaphoid of the
scaphoid,
came
to a sports
fails to demonstrate
consistent
with
an acute
3. Misdirected
possible
search
was
hemochromatosis
noted
in
in a patient
with hepatomegaly and The fact that the patient cancer was not noted on tion. A metastatic lesion ulna was missed. 4. In eight
False-negative of the ankle
was obtained was that
after
ankle
error. Lateral in a 22-year-old
sprain.
plain man
Radiograph
interpreted as normal, despite the fact there is an obvious fracture of the anteprocess of the calcaneus (arrows). The
rior
observer may have istence of fractures
spect initial study
been unaware in this location.
of the ex-
but that was not detected on the radiologic evaluation. Further showed
41 cases
in which
specific cause could be identified. 1 . In five cases error occurred cause
old
radiologic
studies
a
be-
or reports
were not consulted. For example, a pulmonary nodule was missed that had been demonstrated on a prior chest radiograph and had been reported correctly. 2. In ii
cases,
limitation
of tech-
nique caused the error (Table 2). In some cases poor quality of images led to the lesion being missed, whereas in others the lesion was not visualized despite performance of an adequate radiologic
Volume
examination
183
Number
#{149}
(Fig
1
3).
clinic
a fracture.
with
exquisite
(b) Nuclear
fracture.
tenderness
medicine
(c) Follow-up
in the
scan
radiograph
shows obtained
2
of the scaphoid.
four cases in which the lesion was missed because of inaccurate, incomplete, or misleading clinical history. For example, radiographs of the wrist were obtained for the evaluation of
Figure 4. radiograph
medicine
cases
the
wrist pain. had lung the requisiof the distal
was
abnormality
difficult
because
of
the inability to separate this cause of error from simple underreading. Figure 4 shows an example of a falsenegative error that may have occurred because of a lack of knowledge or because of simple underreading. 6. “Satisfaction of search” refers to the circumstance in which the radiologist recognizes one abnormality but fails to see a second lesion. This was noted in ii cases. False-positive
error.-This
less
fre-
quent cause of error was noted only is times (Table 3). Figure 5 shows an example of a false-positive error. Misclassification-In these 47 cases the abnormality was recognized but was
given
an incorrect
interpretation.
Errors
were
divided
among
called
lesser
disease
(undercalls)
(n (n
=
18),
=
iO), or a more
disease
of the
severe
same
those severity
ailment
as scarring.
pretation having
An
example
of a misinter-
with the wrong diagnosis an equivalent disease severity
is a herniated nucleus pulposus C5-6 misdiagnosed as a herniated
was outside the area of primary examination; for example, an abnormality of the chest was visible on an abdominal radiograph but was overlooked. 5. Lack of knowledge was a very unusual cause of error, accounting for two cases in our series. Assigning cases to the category of “lack of knowledge”
(overcalls) (n = i9). An example of an undercall is an adenocarcinoma of the mesentery that was seen on a CT scan of the abdomen and was interpreted
of
nucleus pulposus of C4-5. An example of an overcall-in which gynecomastia was misinterpreted as possible malignancy-is presented in Figure 6. Other
Mishaps
Fifty-six mishaps were attributed to problems other than perceptual error. These included 38 complications and 18 communication errors. Complications-In this group of 38 cases,
seven
have
occurred
were
departmental
lowed
(Table
specifically
because
noted
to
established
standards
4). In nine
departmental practice because of the incident
were
not
cases
a new
fol-
was established involved (Ta-
ble5). Communication
errors.-In
18 cases
communication logic examinations wrong patients;
errors included radioobtained of the incorrect examina-
tions
on
performed
the
correct
pa-
tients; delayed diagnosis because radiologic images were allowed to be taken from the radiology department before they were interpreted; and failure to alert attending physicians of important, unsuspected findings on images. Most of these errors could have been prevented if established departmental
routine
had
been
fol-
lowed. Radiology
147
#{149}
Figure
5.
False-positive
a 1-year-old
baby
unsupervised ging. (a) Detail
error.
girl who
Images
are
had been
with buttons and from anteroposterior
began
radiograph
was read
as demonstrating
2-cm-diameter esophagus.
foreign A barium
body swallow
of
playing gagchest
a
in the midstudy was
ob-
tamed, which was interpreted as normal. (b) Oblique spot radiograph of the chest was obtained after placement of a metallic pellet on the manubrium. The pellet was used to demonstrate that the “foreign body” actually represented
a center
of prominent
manubrial
ossification.
a.
b.
DISCUSSION Others
(3) have
shown
problem awareness
case conference of perceptual
mishaps
in
initiate
diagnostic
measures
lems
in the
little
value
radiology
to reduce
future.
of a
other
and
such
for many
have
years;
those of Smith (Fig Mishaps not due
Table
policy
by changes It is of value,
standards
It is less clear tual-cognitive
how errors
required.
to reduce
percep-
(ie,
mistakes
satisfactory
images,
shallow inspiration One-year-old patient was misinterpreted
79
T-5 fracture
by some shadows
99
In a patient
100
duct. In a patient
malrotation surgery
read
swallow
because
had marked
staff members on screening
failed
wedging
point
as normal mammogram
to demonstrate
and
oma.”
of in-
with laboratory
showed
Results
was
found,
104
133
this is actually
true
165
difficult
clinical
truth.
providing
standards
number
the
In principle,
of underreading
result
of overt
should be reduced (5). Kundel (5) distinguished perceptual and cognitive lieved a perceptual error image
features,
though
the are
between errors. He beoccurred when recorded,
were
not appreciated. A cognitive, or reasoning, error occurred when image features,
148
though
Radiology
#{149}
appreciated,
led
to wrong
ultrasound
(US)
suspicious
opacity
in left
upper
lobe
was
identified
on a chest nodule, and
constrast
radiograph. repeat plain
A CT scan radiogra-
material-enhanced
barium
and MR imaging. 170
Nonvisualization ease. Agenesis
by frequently
of comparison, errors that decision making
as showing a fracture. as mass lesion; follow-up
enema study was read as suspicious for tumor. At surgery, no tumor was found. Injection technique resulted in misreading of area of lower attenuation at intravenous cholangiography as a thrombosis. Right hilar mass was seen on CT scan. Repeat CT scan obtained later showed no mass to be present. Ultrafast CT scan showed a column of contrast material adjacent to the aorta that was probably the result of partial volume effect but that was read as dissection. This reading was later excluded on the basis of findings at thoracic aortography
160
is
radi-
Clinicians were informed that findings on a nuclear medicine study confirmed brain death when in fact the scan showed some intracerebral activity. An 8-mm-diameter appendix was interpreted as demonstrating appendicitis, but a normal appendix was noted at surgery and on subsequent pathology reports.
of
to establish. At the very least, group discussion ought to lead to better calibration of the findings of different observers relative to each other and to
at plain
CT scan was
for parathyroid adenlaboratory tests were normaL At surgery, normal as well as a lymph node from the right inferior
radiologists will recognize and others like them when in the radiology reading
Whether
by others were read
mass
i16
room.
body
of hyperparathyroidism,
isoechoic
of preoperative
parathyroid tissue lobe of the thyroid.
103
evidence
“9 x 9 x il-cm
obtained 10 days later failed to demonstrate phy also failed to demonstrate nodule. Nodule or polyp of the colon seen on a single
pitfalls reoccur
vertebral
Subsequent
lesion.
Nodular
these they
of the
tenderness.
112
colleagues,
ossification center in performance of
with dilated biiary system, endoscopic retrograde cholangiopancreatograph was read as showing a cutoff in the pancreas indicative of a mass. Surgery demonstrated an accessory pancreatic duct and a normal main pancreatic
scan
images,
volvulus in an infant no abnormality.
examination.
of slight
and on acquisition of appropriate clinical information. It may be that by confronting perceptual errors made by ourselves and our prior
with showed
rather than a pulmonary nodule. possibly swallowed a button. Manubrial as being a foreign body, which resulted
the patient
read Vague
as probable Subsequent
of Case
cancer had a metastatic deposit according to reading of a This finding was subsequently proved to be the result of a
and a barium was
mammography
standards an optimal abnormality. on technion review
was read vomiting.
67
in er-
when
terpretation). Departmental can be developed to ensure environment for detecting Examples include insistence cally
Patient with ovarian chest radiograph.
98
of complications and communication, to reeducate physicians about departmental practice standards or to new
38
ography;
of existing particularly
meal study intermittent
endoscopy
rors
establish
Barium with
in departmental
or by enforcement
policy.
31
parallel
i).
to errors in perception and interpretation represent problems that are more readily and easily addressed
Description
very
our results
Errors
Case6
in di-
changed
3
False-Positive
interest
the category of sources of error
radiology
to
prob-
It is of some
that, other than “complications,” agnostic
the
to create error and
C
These
conclusions.
because tribute tures.
are the case
numbers
Perceptual
of a faulty covert a specific meaning
Cognitive of a faulty overt
of gallbladder of the gallbladder
at US was interpreted was demonstrated
in our logbook
and
error
occurred
decision to atto the fea-
error occurred because decision, as a result of
either response bias or flawed diagnostic logic. In reviewing quantitative stud-
are used
as probable at surgery.
for anonymous
ies of error
Kundel occurs
identification
in chest
(5) noted four
times
gallbladder
of patients.
radiography
that
(6,7),
perceptual
more
dis-
error
frequently
than
cognitive error. For our data, as in all nonlaboratory studies, the distinction between
Most
these
perceptual
subcategories
or cognitive
is difficult.
errors April
1992
Table 4 Complications
That
Resulted
from Failure
to Follow
Description
Case6 117
Patient
suffered
dergoing 122
cardiac
arrest
while
un-
A proper
developed
following
angiogra-
monitor
Inadequate
phy. 123
Policies Standard
was not in use dur-
cuff
compression
lowing
A pneumothorax diagnosed
Departmental Departmental
ing the examination.
MR imaging.
Hematoma
Existing Violated
that had been correctly on a chest radiograph was
was
applied
fol-
the procedure.
Patient’s viewed,
records and studies were not reand heart and lungs were not
.b
125
not noted prior to angiography. Subacute bacterial endocarditis followed angiography in a patient who had un-
auscultated prior to angiography. Adequate antibiotic therapy should be administered to patients who undergo
135
dergone mitral valve replacement. Sepsis resulted after a barium enema in a neutropenic patient.
heart valve replacement. Barium enema study should have avoided, or antibiotic coverage have been provided.
occurred
during
study
136
Hypoxia tion.
esophageal
dila-
142
The cerebellar tonsils were punctured during a tap to obtain cerebrospinal fluid.
These
are the case numbers
in our logbook
and
:
In patients with a vascular ring, anomaly, or tracheomalacia, this complication should
6
been should
be anticipated
prior
Review of the MR image would have prevented
are used
for anonymous
.
6’;i
to dilation.
identification
a.
of patients.
b.
Figure 6. 40-year-old
noticed
That Resulted
8
A colon
stricture
tempted 42
43
Departmental
ruptured
during
at-
Pneumothorax was phy 4 hours after second follow-up obtained until 14
apparent at radiograa lung biopsy, but a radiograph was not h after biopsy.
T-tube injection resulted in septicemia, probably secondary to overdistention
During
109
an epidural
for pain
re-
If the
this injection, and hospitalization required. Transvaginal placement of a drainage
was
catheter was unsuccessful because prior hysterectomy. Insulin reaction occurred in a patient
of
A contrast
emptying
material
Do not
strictures
of a had
in the size of his past
few
weeks.
Mam-
diagnosis
after
surgery
only in the rectum.
Augmentation
A patient
demonstrates pneua second examination
in technique the contrast
dur-
betic
occurred
thecal
was instituted material is allowed
sac is punctured
attempt
patients
after
aging
center
material prosthesis
of compression
during
transvaginal
only
inepi-
Warn
during
the patient
in the morning.
of this
In
prior
difficult
to study
because changes
the behavior when they
treatment
with
ra-
Calculate
to
alleged,
dioactive iodine had a greater dose to the thyroid gland than calculated because of retention of dose as a result of renal failure. These are the case numbers
in our logbook
with the 2-Ci
clearance
diagnostic
view, noted
that lesions prospectively.
of 1-131
re-
explana-
retrospective basic
percep-
methods thresholds
of increasing
of decreasing
former, stimulus until an observer
of patients.
it is fre-
retrospective
psychophysical
the method
partly subjects
should have been There is, however,
tion of lesions. Two discovering perceptual
dose.
identification
errors,
for improved
method
and are used for anonymous
in the labora-
of human are observed.
after
an interesting
of the examination.
whole-body
The
in the laboratory
quently
tion
undergoing
data.
the effects
For false-negative
of contrast possibility
of clinical
tory that could have been expected. Of course, many of these factors are very
reactions.
performance
lack
not produced
catheter
for treatment
and
courts occasionally treat false-negative errors as if they were errors of negligence (13-is). However, false-negative errors occur even when these potential sources of error and others, such as poor viewing environment (eg, glare, noise, distractions) and fatigue, can be discounted. Most of these potential error sources are either untested or have
addition, an intravenous catheter should be in place for rapid administration of glucose or other drugs, if necessary. Have epinephrine available in the MR im-
of gadopentetate mammoplasty
ness,
placement in patients who have undergone a hysterectomy. Perform gastric emptying studies in dia-
study.
reaction
ruptured because mammography.
6
colon
dural injection, discontinue the procedure and reschedule at least 1 wk later.
dimeglumine.
173
the
lignancy. Pathologic was of gynecomastia.
Standard
to drip in under the force of gravity stead of being injected. injection
administration 151
Departmental
A change whereby
of
system.
ing a gastric
1 18
New
If the 4-h radiograph mothorax, obtain at 6 h.
lief, the thecal sac was initially punctured. Spinal anesthesia resulted from
59
increase
over
are who
mograms show (a) the left breast and (b) the right breast. The asymmetry present on these images was interpreted as suspicious for ma-
Standards
Dilate
breast
Images patient
dilation.
the biliary 47
in New
Description
Case6
Misclassification. male alcoholic
a small
right
Table 5 Complications
S
,
prior to the tap this problem.
limits limits.
energy reports
for are
and
the
In the
is increased that the stimu-
lus becomes detectable; in the latter, the stimulus energy is decreased until an observer reports that the stimulus has
encountered in this series were the result of underreading or misclassificalion. We found that underreading
lesions in 25%-32% of cases, but a rate of false-positive interpretations of truly negative radiographs of only 1.6%-
become
(false-negative
2.0%.
ergy is required)
errors)
occurred
more
than five times more frequently than overreading (false-positive errors). Our
A variety of explanations for false-negative errors.
are given Many have
findings ies (8-12)
agree that
aura
as faulty
Volume
183
with noted
Number
#{149}
those of prior studfailure to detect
1
ing
of negligence, techniques,
such social
pressure,
method tablished
an
imagcareless-
undetectable.
detectability
established
is invariably with
its, whereas
threshold
with
higher
former
(more
en-
prospectively
a method
viewing
of
the
than the threshold the latter. Viewing
rial radiographs essentially
The
esof se-
is
of increasing
of a series
Radiology
lim-
of radio-
149
#{149}
graphs
retrospectively
is essentially
method of decreasing that stimulus energy lesion sense
size that
and contrast, more lesions
retrospectively.
a
limits. Assuming is analogous to
Studies
it would make are perceived of detection
of
lung nodules support this thesis (i6). There are several primary ways to reduce the number of false-negative diagnostic has been
and
errors. advocated.
adherence
such
as insistence
factory
images,
Increased education Development of
to practice
standards
on technically
review
of prior
satis-
examina-
tions, and acquisition of appropriate clinical information are of obvious importance. Another way is to improve imaging technology so that abnormalities are more easily discovered. To a large extent this recommendation has been accomplished, and we are now considering a new class of errors. For
example, until the newer modalities were used, even large pancreatic abnormalities were seldom discovered. Now we are concerned about the misclassification of recognized small pancreatic lesions. A major but underemphasized approach to error reduction is to conduct further research on the perceptual and intellectual aspects of the human interpretive process as applied to diagnostic images. As we better understand some of the foibles of human perception, we can theoretically design imaging methods that use the more efficient aspects of the human perceptual system. Conversely, we may be able to develop methods such as analysis with artificial intelligence to substitute for defective aspects of the human eye-brain system. Multiple areas of research may be productive in addressing the problem of false-negative readings. In addition to “satisfaction of search,” research could investigate visual time and its effect on accuracy, fixation clusters and dwell time, dual reading, availability of cmical history and comparative images, application of artificial intelligence, and selection of superior observers. The p0tential benefit of tort procedures or tort reform to reduce such error is undetermined; such procedures consume considerable resources. Failure to report multiple lesions has been noted in several studies (17-19). Satisfaction of search is a type of underreading error that occurs when some abnormalities remain unreported when multiple abnormalities are present in an image. Tuddenham (18) believed that a quest for meaning had been satisfied when the first positive finding was rec-
150
Radiology
#{149}
ognized, after which the image was not searched for additional findings. Although plausible, this explanation has recently been found to be inaccurate after laboratory investigations (19). In false-positive errors, the radiographic finding is appreciated but is attributed to disease rather than to a normal physical variant. Despite the considerable increase in knowledge in recent years of what is normal in a radiologic examination, false-positive error occurs with the same frequency. As Smith (4) stated, “More emphasis should be placed on the teaching of normal anatomic and physiologic variants.”
Regarding misclassification of lesions, Smith found that the acquisition of more clinical information and the conscious attempt to expand the differential diagnosis helped decrease such errors. It must be emphasized that there is a limit to the usefulness of expanding the differential diagnosis. For instance, Felson and Reeder (20) list eight common and 30 uncommon disease processes
in
the
differential
diagnosis
of a
solitary than time such More
pulmonary nodule that is less 4 cm in diameter, but most of the it serves no purpose to include a list in the radiographic report. important than the histologic accuracy of diagnosis in a given case is patient outcome. In this regard, radiographic misclassification of disease in our series was often to a diagnosis of similar severity to the patient’s true diagnosis. Misclassification of more or less
severe
disease
occurred
with
References 1.
N, et al. in hospital-
ized patients: results of the Harvard medical study practice II. New EnglJ Med 1991; 324:377-384. 2.
YerushalmyJ,
HarknessJT,
Kennedy BR. mass radiography. 3.
443-464. Wheeler
PS.
surance,
and
ence. 4. 5.
6.
7.
The
role of dual reading in Am Rev TB 1950; 61:
Risk prevention,
as-
the
Smith tic radiology. Springfield, Ill: Thomas, 1%7. Kundel HL. Perception errors in chest radiography. Semin Respir Med 1989; 10: 203-210. Herman PG. Hessel SJ. Accuracy and its relationship to experience in the interpretation of chest radiographs. Invest Radiol 1975; 10:62-67. Rhea JT, Potsaid MS. DeLuca SA. Errors in interpretation as elicited by a quality audit ofan emergency radiology facility.
1978; 132:277-280.
Birkelo
CC,
Schools
PE, Zachs
berculosis effectiveness
Chamberlain case
WE,
Phelps
PS,
D, YerushalmyJ.
finding: of various
and photofluorographic 9.
quality
missed diagnosis conferRadiology 1982; 145:227-228. MJ. Error and variation in diagnos-
Radiology 8.
Tu-
a comparison of the roentgenographic methods. JAMA
1947; 133:359-366. Garland LH. On scientific evaluation of diagnostic procedures. Radiology 1949; 52:309-328.
10.
Garland
LH.
diagnostic 11.
Newell
Studies
on the accuracy
procedures.
of
AJR 1959; 82:25-38.
RR, Chamberlain
WE,
Rigler
L.
Descriptive classification of pulmonary shadows: revelation of unreliability in
12.
roughly
equal frequency. Thus, it appears that there is no systematic bias in over- or undercaffing severity of disease in misclassification. Current methods of evaluation of perceptual error leave certain areas unstudied. Analysis of receiver operating characteristic curves has been the basis for the interpretive process, particularly regarding judgment calls. This technology as currently defined is not helpful in understanding cases of misinterpretation, which is a major source of perceptual error in clinical situations. The data indicate a need to develop better experimental and formal methods of analysis of perceptual error. In conclusion, study of the errors presented at problem case conferences demonstrated that perceptual errors continue to constitute the bulk of errors made by radiologists and that false-negative errors are the most frequently committed perceptual-cognitive mistakes. Errors secondary to complications have resulted in the establishment of several effective departmental policy changes. I
Leape LL, Brennan TA, Naird The nature of adverse events
13.
14.
roentgenographic diagnosis of tuberculosis. Am Rev TB 1954; 69:566-584. YerushalmyJ, Garland LH, Harness JT, et al. An evaluation of the role of serial chest roentgenograms in estimating progress in patients with pulmonary tuberculosis. Am Rev TB 1951; 64:225-248. Berlin L. Does the “missed” radiological diagnosis constitute malpractice? Radiology 1977; 123:523-527. Berlin L. Malpractice and radiologists. AJR 1980; 135:587-591.
15.
Berlin
16.
tice: an ominous 1031-1034. Muhm JR, Miller
L.
Is a radiologic “miss” malpracexample. AJR 1983; 140: WE,
Fontana
RS, et al.
Lung cancer detected during a screening program using 4-month chest radiographs. Radiology 1983; 148:609-615. 17.
18.
19.
Tuddenham WJ. Visual search, image organization, and reader error in roentgen diagnosis. AJR 1962; 78:694-704. Tuddenham WJ. Problems of perception in chest roentgenology: facts and fallacies. Radiol Clin North Am 1963; 1:227-289. Berbaum KS, Franken EA, Dorfman DD, et
al. 20.
Time course
of satisfaction
of search.
Invest Radiol 1991; 26:640-648. Felson B, Reeder MM. Gamuts in radiology: comprehensive lists of roentgen dif-
ferential
diagnosis.
diovisual
Radiology
2nd ed. Cincinnati: of Cincinnati,
Au-
1987.
April
1992
MD MA
#{149} Edmund A. Franken, Monzer M. Abu-Yousef,
Jr, MD MD
#{149}
Kevin
S. Berbaum,
#{149}
Error in Radiology: Classification Lessons in 182 Cases Presented at a Problem Case Conference’ The authors review and classify errors in 182 cases that were presented at problem case conferences between August 1986 and October 1990. Errors were classified by means of a system developed 20 years ago and by means of a system developed within the past several years. The authors found that sources of error have changed very little. Errors usually involved failure to consult old radiologic studies or reports, limitations in imaging technique, acquisition of inaccurate or incomplete clinical history, localion of a lesion outside the area of interest on an image, lack of knowledge, failure to continue to search for abnormalities after the first abnormality was found, and failure to recognize a normal biologic variant. Errors included 126 perceptual errors (64 false-negative, 15 false-positive, and 47 misclassification errors) and 56 mishaps, including 38 complications and 18 communication errors. In seven cases nonperception errors occurred because established departmental routines were not followed, and in nine cases a new departmental routine was established after a cornplication occurred. Departmental poiicy exerts less effect on perception and interpretation errors. Index
terms:
performance partmental Radiology
I
Diagnostic Radiology
management
Table 1 Classification
Cause
and
of Error
Explanation
Complacency
II
Decision-Analysis
Faulty
reasoning
Lack
of knowledge
False-positive
Term error
True-positive reading, but misclassification
category. III
IV
The
finding is seen but is attributed to the wrong cause because of a lack of knowledge on the part of the viewer. The finding is missed. Underreading may result from failure to isolate important material or from satisfaction of search. The lesion is identified and interpreted correctly, but the message fails to reach the clinician. The lesion was not present on the image obtained, even in retrospect. This may be secondary to limitations of the examination or to an inadequate examina-
Underreading
V
Poor
VI
communication
Miscellaneous
True-positive
reading,
but misclassification
False-negative
error
...
False-negative
error
tion. Vii
Complications*
Untoward
events
happened
during
...
the course of examination, which was most frequently encountered during invasive procedures. Source-Adapted * We have added
from Smith (4). this class, which was not in Smith’s
scheme.
radiology, observer and radiologists, de#{149} Quality assurance
1992; i83:i45-i50
From the Department
of Radiology, the University of Iowa College of Medicine, Iowa City. Received May 17, 1991; revision requested June 21; final revision received October 17; accepted
November 4. Address reprint requests to D.L.R., Department of Radiology, Rush-Presbyterian-St Luke’s Medical Center, 1653 W Congress Pkwy, Chicago, IL 60612. 1992
and Examples
Errors of overreading and misinterpretation, in which a finding is appreciated but is attributed to the wrong cause. Errors of overreading and misinterpretation, in which the finding is appreciated and interpreted as abnormal but is attributed to the wrong cause. Misleading information and a limited differential diagnosis are included in this
E
occurs frequently practice of medicine
in the
RROR
(1). In addilion to the mistakes common to all areas of medicine, perceptual error in diagnostic radiology is particularly prevalent, as the human perceptual
RSNA,
PhD
of Error
Class I
Assurance
system
error ogy
is imperfect.
is common has
been
errors
cases
of interpretation
sents
an
and
other
noted images.
analysis
mishaps
for
in over
radiolyears
(2),
one-third
of
of ambiguous
This
article
of perceptual
that
occurred
preerrors
in
radiology over
MATERIALS
AND
Since August 1986, ences have been held
perceptual
recognized
with
diagnostic
That
in diagnostic
the diagnostic at our institution
department a 4-year period.
METHODS
problem case conferby the Department
of Radiology at the University of Iowa; the conferences, on the model of Wheeler (3), are held 10 times per year. Subspecialty divisions
within
radiology)
are
the department assigned
a rotating basis, so that tributes approximately
to present
(pediatric cases
on
each division confour cases per year.
145
182 Errors
Comp’acency Faulty Lack
Reasoning
of
I
Knowledge
126 Perceptual/Cognitive
56 Other
Underr.adng Poor
I
Communcat#{232}on
15 False Positive Tab$e4
UiscelI&ous Complicat#{232}on
47 Misclassification TableS
260
200
r
150 Smiths
100 60 Ssr#{149}s
0
1020 P,ob.m
E
30 40 60 Cass Conf.
I
60
64 False
Negative
is
Ioame
undercalls
19
severity
7 department routine not followed Table6
overcalls
Figure 1. Graph depicts distribution of errors in Smith’s original 437 cases (4) and in our 182 cases. 23
Problems sources.
are identified The review
aminations terpreted
such
as magnetic
imaging
malities
may
that
are
on routine leagues
of ex-
apparent
often
A logbook provides
associated morbidity
about the interpretations
errors, any that resulted
Table
2
Errors
Caused
cause
important
a single cause The radiologists cording
I.
was 12
of error,
they
18
types 77
with
chose
188
aging
with
marizes
newer
most
Smith’s
analysis error.
was
The
a 4-year radiologic
reported
146
#{149}
Radiology
and
therefore
0
These
administration
of three
a transverse
of imaging
colon
the
interpretation fracture
was
an inherently
contrast
material
was missed on first arteriogram because carotid arteriography. of the spine was missed at MR imaging was
not
diagnosed
beam
and the vessel
large
margin
esophogram
enemas
did
of failure
to per-
because
of the
chosen.
disease
the Doppler
water-soluble
tumor.
sequences
venoocclusive
are the case numbers
representative
failed
in our logbook
were
on initial
almost
Doppler
studies
perpendicular,
be-
a situation
of error.
to show
and
a lealc the clinician
are used
in this study, greater than
the
series
incidents
that
were
per-
department of case selection necessarily
does
in the de-
represent
considered
those by
radiologists and by our clinical to be the most serious errors.
the
staff
colleagues
refused
for anonymous
191 cases 182.
were
to have bar-
identification
presented
of patients.
in the conferences,
Figure i presents classification Smith’s data, as well as our own, means tion on methods
sification perceptual
of Smith’s
scheme.
of by
Reclassifica-
the basis of contemporary of perceptual-cognitive
demonstrated error and
tributed sents rors
to other
causes.
clas-
126 cases of 56 mishaps at-
Figure
the classification of our on the basis of contemporary
2 pre182
er-
methods.
Perceptual-Cognitive
RESULTS in
occurred over were
of all error
partment,
er-
herein
after
Although there were i82 errors classified and, thus, the case number is occasionally
not
as
perceptual-cogfalse-negative, and other error
period in which examinations
on plain radiographs; was performed, and
on the such
communication
formed in the radiology our institution. Although arbitrary
we
terminology.
182 cases
was
made
methods
and
meth-
to his
I also indicates the closest to Smith’s classes that exists
decision-analysis during 700,000
1 sum-
to classify perceptualSuch classification in-
(complications
Table equivalent
Table system;
also
of contemporary
we
and im-
“complications”
volved dividing cases into nitive error (false-positive, or misinterpretation error) rors).
of what
classification
of contemporary
decision cognitive
pracnot per-
radiology,
modalities.
added the category scheme. 2. Classification basis
of gen-
his colleagues in private and his colleagues did
medicine,
basis
ium administered, and a leak was missed. Computed tomography (CT) failed to demonstrate an intraabdominal abscess, which was documented with labeled white blood cell study. Salter-Harris undisplaced type I fracture was missed on initial radiograph. Postoperative upper CI examination was interpreted as negative for leak, but abdominal CT demonstrated the leak. Plain radiographs showed normal findings in a neonate with hypoplastic left heart syndrome. Eight-millimeter-thick sections acquired on an ultrafast chest CT scan failed to show a double aortic arch.
94
Smith and tice. Smith
practice
now call interventional
the
or Technique
was not apparent medicine study
a nuclear
Water-soluble
85
171
nuclear
Hepatic
cause
by Smith
in the
of the scaphoid
Intracerebral aneurysm form selected internal Disk abscess and erosion
30
eral radiology, classified 437 errors that occurred during a special 2-year study of
form
Imaging
systems.
developed
errors
Examination
of
qualified;
not show
144 156
scheme,
(4) for
on
demonstrated.
in
One
1967
Limitations
Fracture
on the single
classification
by
of 182 errors
of
Search
Description
5
after discussion of the case. classified the cases ac-
to two
classification
Case*
and/or the error,
according to the single most important cause of error. In the few cases in which most
depicts
11 SatIsfaction
of Knowledge
ex-
and steps that were taken to minimize the chance of recurrence of the error. Two radiologists (D.L.R., E.A.F.) independently classified the i82 cases presented at problem case conferences between August 1986 and October 1990
disagreed
chart
Search
I
I 2 Lack
out
radiologic and
mortality from
the radiologists
Flow
I S Abnormality Outside Area of Primary Exam
col-
in pointing
compiled from each cona brief history of the pa-
tient, information amination with
41 Specific Cause for False Negative
4 Misdirected
11 Poor Films of limitation of exam Table 3
9 new department routine implemented Table 7
retrospect”
Clinical
helpful
2.
18 Communication Errors Table S
ods.
abnor-
“in
No Specific Caine ot False Negative
I S Old Films/Reports Table2
Figure
resonance
demonstrate
radiographs. are
errors. ference
a variety imaging
that occurs when cases are inoften discloses errors. Advanced
modalities
(MR)
from of prior
Complicatmos
38
I
,
Errors
The i26 cases of perceptual-cognitive error included 64 false-negative, is false-positive, and 47 misclassification errors. False-negative
error-A
ple of a false-negative monary
nodule
that
classic
error is visible
exam-
is a pulin retro-
April
1992
t.,:
.
,
* .
,
.
.
:
b. Figure
3.
anatomic intense weeks
Limitation
snuff
of technique.
box. (a) Ulnar-deviation
increased
after
Images
radiotracer
the injury
are
oblique
localization
demonstrates
in the
a fracture
of an
18-year-old
radiograph region
gymnast
who
of the scaphoid of the
scaphoid,
came
to a sports
fails to demonstrate
consistent
with
an acute
3. Misdirected
possible
search
was
hemochromatosis
noted
in
in a patient
with hepatomegaly and The fact that the patient cancer was not noted on tion. A metastatic lesion ulna was missed. 4. In eight
False-negative of the ankle
was obtained was that
after
ankle
error. Lateral in a 22-year-old
sprain.
plain man
Radiograph
interpreted as normal, despite the fact there is an obvious fracture of the anteprocess of the calcaneus (arrows). The
rior
observer may have istence of fractures
spect initial study
been unaware in this location.
of the ex-
but that was not detected on the radiologic evaluation. Further showed
41 cases
in which
specific cause could be identified. 1 . In five cases error occurred cause
old
radiologic
studies
a
be-
or reports
were not consulted. For example, a pulmonary nodule was missed that had been demonstrated on a prior chest radiograph and had been reported correctly. 2. In ii
cases,
limitation
of tech-
nique caused the error (Table 2). In some cases poor quality of images led to the lesion being missed, whereas in others the lesion was not visualized despite performance of an adequate radiologic
Volume
examination
183
Number
#{149}
(Fig
1
3).
clinic
a fracture.
with
exquisite
(b) Nuclear
fracture.
tenderness
medicine
(c) Follow-up
in the
scan
radiograph
shows obtained
2
of the scaphoid.
four cases in which the lesion was missed because of inaccurate, incomplete, or misleading clinical history. For example, radiographs of the wrist were obtained for the evaluation of
Figure 4. radiograph
medicine
cases
the
wrist pain. had lung the requisiof the distal
was
abnormality
difficult
because
of
the inability to separate this cause of error from simple underreading. Figure 4 shows an example of a falsenegative error that may have occurred because of a lack of knowledge or because of simple underreading. 6. “Satisfaction of search” refers to the circumstance in which the radiologist recognizes one abnormality but fails to see a second lesion. This was noted in ii cases. False-positive
error.-This
less
fre-
quent cause of error was noted only is times (Table 3). Figure 5 shows an example of a false-positive error. Misclassification-In these 47 cases the abnormality was recognized but was
given
an incorrect
interpretation.
Errors
were
divided
among
called
lesser
disease
(undercalls)
(n (n
=
18),
=
iO), or a more
disease
of the
severe
same
those severity
ailment
as scarring.
pretation having
An
example
of a misinter-
with the wrong diagnosis an equivalent disease severity
is a herniated nucleus pulposus C5-6 misdiagnosed as a herniated
was outside the area of primary examination; for example, an abnormality of the chest was visible on an abdominal radiograph but was overlooked. 5. Lack of knowledge was a very unusual cause of error, accounting for two cases in our series. Assigning cases to the category of “lack of knowledge”
(overcalls) (n = i9). An example of an undercall is an adenocarcinoma of the mesentery that was seen on a CT scan of the abdomen and was interpreted
of
nucleus pulposus of C4-5. An example of an overcall-in which gynecomastia was misinterpreted as possible malignancy-is presented in Figure 6. Other
Mishaps
Fifty-six mishaps were attributed to problems other than perceptual error. These included 38 complications and 18 communication errors. Complications-In this group of 38 cases,
seven
have
occurred
were
departmental
lowed
(Table
specifically
because
noted
to
established
standards
4). In nine
departmental practice because of the incident
were
not
cases
a new
fol-
was established involved (Ta-
ble5). Communication
errors.-In
18 cases
communication logic examinations wrong patients;
errors included radioobtained of the incorrect examina-
tions
on
performed
the
correct
pa-
tients; delayed diagnosis because radiologic images were allowed to be taken from the radiology department before they were interpreted; and failure to alert attending physicians of important, unsuspected findings on images. Most of these errors could have been prevented if established departmental
routine
had
been
fol-
lowed. Radiology
147
#{149}
Figure
5.
False-positive
a 1-year-old
baby
unsupervised ging. (a) Detail
error.
girl who
Images
are
had been
with buttons and from anteroposterior
began
radiograph
was read
as demonstrating
2-cm-diameter esophagus.
foreign A barium
body swallow
of
playing gagchest
a
in the midstudy was
ob-
tamed, which was interpreted as normal. (b) Oblique spot radiograph of the chest was obtained after placement of a metallic pellet on the manubrium. The pellet was used to demonstrate that the “foreign body” actually represented
a center
of prominent
manubrial
ossification.
a.
b.
DISCUSSION Others
(3) have
shown
problem awareness
case conference of perceptual
mishaps
in
initiate
diagnostic
measures
lems
in the
little
value
radiology
to reduce
future.
of a
other
and
such
for many
have
years;
those of Smith (Fig Mishaps not due
Table
policy
by changes It is of value,
standards
It is less clear tual-cognitive
how errors
required.
to reduce
percep-
(ie,
mistakes
satisfactory
images,
shallow inspiration One-year-old patient was misinterpreted
79
T-5 fracture
by some shadows
99
In a patient
100
duct. In a patient
malrotation surgery
read
swallow
because
had marked
staff members on screening
failed
wedging
point
as normal mammogram
to demonstrate
and
oma.”
of in-
with laboratory
showed
Results
was
found,
104
133
this is actually
true
165
difficult
clinical
truth.
providing
standards
number
the
In principle,
of underreading
result
of overt
should be reduced (5). Kundel (5) distinguished perceptual and cognitive lieved a perceptual error image
features,
though
the are
between errors. He beoccurred when recorded,
were
not appreciated. A cognitive, or reasoning, error occurred when image features,
148
though
Radiology
#{149}
appreciated,
led
to wrong
ultrasound
(US)
suspicious
opacity
in left
upper
lobe
was
identified
on a chest nodule, and
constrast
radiograph. repeat plain
A CT scan radiogra-
material-enhanced
barium
and MR imaging. 170
Nonvisualization ease. Agenesis
by frequently
of comparison, errors that decision making
as showing a fracture. as mass lesion; follow-up
enema study was read as suspicious for tumor. At surgery, no tumor was found. Injection technique resulted in misreading of area of lower attenuation at intravenous cholangiography as a thrombosis. Right hilar mass was seen on CT scan. Repeat CT scan obtained later showed no mass to be present. Ultrafast CT scan showed a column of contrast material adjacent to the aorta that was probably the result of partial volume effect but that was read as dissection. This reading was later excluded on the basis of findings at thoracic aortography
160
is
radi-
Clinicians were informed that findings on a nuclear medicine study confirmed brain death when in fact the scan showed some intracerebral activity. An 8-mm-diameter appendix was interpreted as demonstrating appendicitis, but a normal appendix was noted at surgery and on subsequent pathology reports.
of
to establish. At the very least, group discussion ought to lead to better calibration of the findings of different observers relative to each other and to
at plain
CT scan was
for parathyroid adenlaboratory tests were normaL At surgery, normal as well as a lymph node from the right inferior
radiologists will recognize and others like them when in the radiology reading
Whether
by others were read
mass
i16
room.
body
of hyperparathyroidism,
isoechoic
of preoperative
parathyroid tissue lobe of the thyroid.
103
evidence
“9 x 9 x il-cm
obtained 10 days later failed to demonstrate phy also failed to demonstrate nodule. Nodule or polyp of the colon seen on a single
pitfalls reoccur
vertebral
Subsequent
lesion.
Nodular
these they
of the
tenderness.
112
colleagues,
ossification center in performance of
with dilated biiary system, endoscopic retrograde cholangiopancreatograph was read as showing a cutoff in the pancreas indicative of a mass. Surgery demonstrated an accessory pancreatic duct and a normal main pancreatic
scan
images,
volvulus in an infant no abnormality.
examination.
of slight
and on acquisition of appropriate clinical information. It may be that by confronting perceptual errors made by ourselves and our prior
with showed
rather than a pulmonary nodule. possibly swallowed a button. Manubrial as being a foreign body, which resulted
the patient
read Vague
as probable Subsequent
of Case
cancer had a metastatic deposit according to reading of a This finding was subsequently proved to be the result of a
and a barium was
mammography
standards an optimal abnormality. on technion review
was read vomiting.
67
in er-
when
terpretation). Departmental can be developed to ensure environment for detecting Examples include insistence cally
Patient with ovarian chest radiograph.
98
of complications and communication, to reeducate physicians about departmental practice standards or to new
38
ography;
of existing particularly
meal study intermittent
endoscopy
rors
establish
Barium with
in departmental
or by enforcement
policy.
31
parallel
i).
to errors in perception and interpretation represent problems that are more readily and easily addressed
Description
very
our results
Errors
Case6
in di-
changed
3
False-Positive
interest
the category of sources of error
radiology
to
prob-
It is of some
that, other than “complications,” agnostic
the
to create error and
C
These
conclusions.
because tribute tures.
are the case
numbers
Perceptual
of a faulty covert a specific meaning
Cognitive of a faulty overt
of gallbladder of the gallbladder
at US was interpreted was demonstrated
in our logbook
and
error
occurred
decision to atto the fea-
error occurred because decision, as a result of
either response bias or flawed diagnostic logic. In reviewing quantitative stud-
are used
as probable at surgery.
for anonymous
ies of error
Kundel occurs
identification
in chest
(5) noted four
times
gallbladder
of patients.
radiography
that
(6,7),
perceptual
more
dis-
error
frequently
than
cognitive error. For our data, as in all nonlaboratory studies, the distinction between
Most
these
perceptual
subcategories
or cognitive
is difficult.
errors April
1992
Table 4 Complications
That
Resulted
from Failure
to Follow
Description
Case6 117
Patient
suffered
dergoing 122
cardiac
arrest
while
un-
A proper
developed
following
angiogra-
monitor
Inadequate
phy. 123
Policies Standard
was not in use dur-
cuff
compression
lowing
A pneumothorax diagnosed
Departmental Departmental
ing the examination.
MR imaging.
Hematoma
Existing Violated
that had been correctly on a chest radiograph was
was
applied
fol-
the procedure.
Patient’s viewed,
records and studies were not reand heart and lungs were not
.b
125
not noted prior to angiography. Subacute bacterial endocarditis followed angiography in a patient who had un-
auscultated prior to angiography. Adequate antibiotic therapy should be administered to patients who undergo
135
dergone mitral valve replacement. Sepsis resulted after a barium enema in a neutropenic patient.
heart valve replacement. Barium enema study should have avoided, or antibiotic coverage have been provided.
occurred
during
study
136
Hypoxia tion.
esophageal
dila-
142
The cerebellar tonsils were punctured during a tap to obtain cerebrospinal fluid.
These
are the case numbers
in our logbook
and
:
In patients with a vascular ring, anomaly, or tracheomalacia, this complication should
6
been should
be anticipated
prior
Review of the MR image would have prevented
are used
for anonymous
.
6’;i
to dilation.
identification
a.
of patients.
b.
Figure 6. 40-year-old
noticed
That Resulted
8
A colon
stricture
tempted 42
43
Departmental
ruptured
during
at-
Pneumothorax was phy 4 hours after second follow-up obtained until 14
apparent at radiograa lung biopsy, but a radiograph was not h after biopsy.
T-tube injection resulted in septicemia, probably secondary to overdistention
During
109
an epidural
for pain
re-
If the
this injection, and hospitalization required. Transvaginal placement of a drainage
was
catheter was unsuccessful because prior hysterectomy. Insulin reaction occurred in a patient
of
A contrast
emptying
material
Do not
strictures
of a had
in the size of his past
few
weeks.
Mam-
diagnosis
after
surgery
only in the rectum.
Augmentation
A patient
demonstrates pneua second examination
in technique the contrast
dur-
betic
occurred
thecal
was instituted material is allowed
sac is punctured
attempt
patients
after
aging
center
material prosthesis
of compression
during
transvaginal
only
inepi-
Warn
during
the patient
in the morning.
of this
In
prior
difficult
to study
because changes
the behavior when they
treatment
with
ra-
Calculate
to
alleged,
dioactive iodine had a greater dose to the thyroid gland than calculated because of retention of dose as a result of renal failure. These are the case numbers
in our logbook
with the 2-Ci
clearance
diagnostic
view, noted
that lesions prospectively.
of 1-131
re-
explana-
retrospective basic
percep-
methods thresholds
of increasing
of decreasing
former, stimulus until an observer
of patients.
it is fre-
retrospective
psychophysical
the method
partly subjects
should have been There is, however,
tion of lesions. Two discovering perceptual
dose.
identification
errors,
for improved
method
and are used for anonymous
in the labora-
of human are observed.
after
an interesting
of the examination.
whole-body
The
in the laboratory
quently
tion
undergoing
data.
the effects
For false-negative
of contrast possibility
of clinical
tory that could have been expected. Of course, many of these factors are very
reactions.
performance
lack
not produced
catheter
for treatment
and
courts occasionally treat false-negative errors as if they were errors of negligence (13-is). However, false-negative errors occur even when these potential sources of error and others, such as poor viewing environment (eg, glare, noise, distractions) and fatigue, can be discounted. Most of these potential error sources are either untested or have
addition, an intravenous catheter should be in place for rapid administration of glucose or other drugs, if necessary. Have epinephrine available in the MR im-
of gadopentetate mammoplasty
ness,
placement in patients who have undergone a hysterectomy. Perform gastric emptying studies in dia-
study.
reaction
ruptured because mammography.
6
colon
dural injection, discontinue the procedure and reschedule at least 1 wk later.
dimeglumine.
173
the
lignancy. Pathologic was of gynecomastia.
Standard
to drip in under the force of gravity stead of being injected. injection
administration 151
Departmental
A change whereby
of
system.
ing a gastric
1 18
New
If the 4-h radiograph mothorax, obtain at 6 h.
lief, the thecal sac was initially punctured. Spinal anesthesia resulted from
59
increase
over
are who
mograms show (a) the left breast and (b) the right breast. The asymmetry present on these images was interpreted as suspicious for ma-
Standards
Dilate
breast
Images patient
dilation.
the biliary 47
in New
Description
Case6
Misclassification. male alcoholic
a small
right
Table 5 Complications
S
,
prior to the tap this problem.
limits limits.
energy reports
for are
and
the
In the
is increased that the stimu-
lus becomes detectable; in the latter, the stimulus energy is decreased until an observer reports that the stimulus has
encountered in this series were the result of underreading or misclassificalion. We found that underreading
lesions in 25%-32% of cases, but a rate of false-positive interpretations of truly negative radiographs of only 1.6%-
become
(false-negative
2.0%.
ergy is required)
errors)
occurred
more
than five times more frequently than overreading (false-positive errors). Our
A variety of explanations for false-negative errors.
are given Many have
findings ies (8-12)
agree that
aura
as faulty
Volume
183
with noted
Number
#{149}
those of prior studfailure to detect
1
ing
of negligence, techniques,
such social
pressure,
method tablished
an
imagcareless-
undetectable.
detectability
established
is invariably with
its, whereas
threshold
with
higher
former
(more
en-
prospectively
a method
viewing
of
the
than the threshold the latter. Viewing
rial radiographs essentially
The
esof se-
is
of increasing
of a series
Radiology
lim-
of radio-
149
#{149}
graphs
retrospectively
is essentially
method of decreasing that stimulus energy lesion sense
size that
and contrast, more lesions
retrospectively.
a
limits. Assuming is analogous to
Studies
it would make are perceived of detection
of
lung nodules support this thesis (i6). There are several primary ways to reduce the number of false-negative diagnostic has been
and
errors. advocated.
adherence
such
as insistence
factory
images,
Increased education Development of
to practice
standards
on technically
review
of prior
satis-
examina-
tions, and acquisition of appropriate clinical information are of obvious importance. Another way is to improve imaging technology so that abnormalities are more easily discovered. To a large extent this recommendation has been accomplished, and we are now considering a new class of errors. For
example, until the newer modalities were used, even large pancreatic abnormalities were seldom discovered. Now we are concerned about the misclassification of recognized small pancreatic lesions. A major but underemphasized approach to error reduction is to conduct further research on the perceptual and intellectual aspects of the human interpretive process as applied to diagnostic images. As we better understand some of the foibles of human perception, we can theoretically design imaging methods that use the more efficient aspects of the human perceptual system. Conversely, we may be able to develop methods such as analysis with artificial intelligence to substitute for defective aspects of the human eye-brain system. Multiple areas of research may be productive in addressing the problem of false-negative readings. In addition to “satisfaction of search,” research could investigate visual time and its effect on accuracy, fixation clusters and dwell time, dual reading, availability of cmical history and comparative images, application of artificial intelligence, and selection of superior observers. The p0tential benefit of tort procedures or tort reform to reduce such error is undetermined; such procedures consume considerable resources. Failure to report multiple lesions has been noted in several studies (17-19). Satisfaction of search is a type of underreading error that occurs when some abnormalities remain unreported when multiple abnormalities are present in an image. Tuddenham (18) believed that a quest for meaning had been satisfied when the first positive finding was rec-
150
Radiology
#{149}
ognized, after which the image was not searched for additional findings. Although plausible, this explanation has recently been found to be inaccurate after laboratory investigations (19). In false-positive errors, the radiographic finding is appreciated but is attributed to disease rather than to a normal physical variant. Despite the considerable increase in knowledge in recent years of what is normal in a radiologic examination, false-positive error occurs with the same frequency. As Smith (4) stated, “More emphasis should be placed on the teaching of normal anatomic and physiologic variants.”
Regarding misclassification of lesions, Smith found that the acquisition of more clinical information and the conscious attempt to expand the differential diagnosis helped decrease such errors. It must be emphasized that there is a limit to the usefulness of expanding the differential diagnosis. For instance, Felson and Reeder (20) list eight common and 30 uncommon disease processes
in
the
differential
diagnosis
of a
solitary than time such More
pulmonary nodule that is less 4 cm in diameter, but most of the it serves no purpose to include a list in the radiographic report. important than the histologic accuracy of diagnosis in a given case is patient outcome. In this regard, radiographic misclassification of disease in our series was often to a diagnosis of similar severity to the patient’s true diagnosis. Misclassification of more or less
severe
disease
occurred
with
References 1.
N, et al. in hospital-
ized patients: results of the Harvard medical study practice II. New EnglJ Med 1991; 324:377-384. 2.
YerushalmyJ,
HarknessJT,
Kennedy BR. mass radiography. 3.
443-464. Wheeler
PS.
surance,
and
ence. 4. 5.
6.
7.
The
role of dual reading in Am Rev TB 1950; 61:
Risk prevention,
as-
the
Smith tic radiology. Springfield, Ill: Thomas, 1%7. Kundel HL. Perception errors in chest radiography. Semin Respir Med 1989; 10: 203-210. Herman PG. Hessel SJ. Accuracy and its relationship to experience in the interpretation of chest radiographs. Invest Radiol 1975; 10:62-67. Rhea JT, Potsaid MS. DeLuca SA. Errors in interpretation as elicited by a quality audit ofan emergency radiology facility.
1978; 132:277-280.
Birkelo
CC,
Schools
PE, Zachs
berculosis effectiveness
Chamberlain case
WE,
Phelps
PS,
D, YerushalmyJ.
finding: of various
and photofluorographic 9.
quality
missed diagnosis conferRadiology 1982; 145:227-228. MJ. Error and variation in diagnos-
Radiology 8.
Tu-
a comparison of the roentgenographic methods. JAMA
1947; 133:359-366. Garland LH. On scientific evaluation of diagnostic procedures. Radiology 1949; 52:309-328.
10.
Garland
LH.
diagnostic 11.
Newell
Studies
on the accuracy
procedures.
of
AJR 1959; 82:25-38.
RR, Chamberlain
WE,
Rigler
L.
Descriptive classification of pulmonary shadows: revelation of unreliability in
12.
roughly
equal frequency. Thus, it appears that there is no systematic bias in over- or undercaffing severity of disease in misclassification. Current methods of evaluation of perceptual error leave certain areas unstudied. Analysis of receiver operating characteristic curves has been the basis for the interpretive process, particularly regarding judgment calls. This technology as currently defined is not helpful in understanding cases of misinterpretation, which is a major source of perceptual error in clinical situations. The data indicate a need to develop better experimental and formal methods of analysis of perceptual error. In conclusion, study of the errors presented at problem case conferences demonstrated that perceptual errors continue to constitute the bulk of errors made by radiologists and that false-negative errors are the most frequently committed perceptual-cognitive mistakes. Errors secondary to complications have resulted in the establishment of several effective departmental policy changes. I
Leape LL, Brennan TA, Naird The nature of adverse events
13.
14.
roentgenographic diagnosis of tuberculosis. Am Rev TB 1954; 69:566-584. YerushalmyJ, Garland LH, Harness JT, et al. An evaluation of the role of serial chest roentgenograms in estimating progress in patients with pulmonary tuberculosis. Am Rev TB 1951; 64:225-248. Berlin L. Does the “missed” radiological diagnosis constitute malpractice? Radiology 1977; 123:523-527. Berlin L. Malpractice and radiologists. AJR 1980; 135:587-591.
15.
Berlin
16.
tice: an ominous 1031-1034. Muhm JR, Miller
L.
Is a radiologic “miss” malpracexample. AJR 1983; 140: WE,
Fontana
RS, et al.
Lung cancer detected during a screening program using 4-month chest radiographs. Radiology 1983; 148:609-615. 17.
18.
19.
Tuddenham WJ. Visual search, image organization, and reader error in roentgen diagnosis. AJR 1962; 78:694-704. Tuddenham WJ. Problems of perception in chest roentgenology: facts and fallacies. Radiol Clin North Am 1963; 1:227-289. Berbaum KS, Franken EA, Dorfman DD, et
al. 20.
Time course
of satisfaction
of search.
Invest Radiol 1991; 26:640-648. Felson B, Reeder MM. Gamuts in radiology: comprehensive lists of roentgen dif-
ferential
diagnosis.
diovisual
Radiology
2nd ed. Cincinnati: of Cincinnati,
Au-
1987.
April
1992
