Pediatr Radiol (2011) 41:327334
DOI 10.1007/s00247-010-1812-6
ORIGINAL ARTICLE
Diagnostic errors in pediatric radiology
George A. Taylor & Stephan D. Voss &
Patrice R. Melvin & Dionne A. Graham
Received: 29 June 2010 / Revised: 23 July 2010 / Accepted: 3 August 2010 / Published online: 9 September 2010
# Springer-Verlag 2010
Abstract
Background Little information is known about the frequency,
types and causes of diagnostic errors in imaging children.
Objective Our goals were to describe the patterns and
potential etiologies of diagnostic error in our subspecialty.
Materials and methods We reviewed 265 cases with
clinically significant diagnostic errors identified during a
10-year period. Errors were defined as a diagnosis that was
delayed, wrong or missed; they were classified as percep-
tual, cognitive, system-related or unavoidable; and they
were evaluated by imaging modality and level of training of
the physician involved.
Results We identified 484 specific errors in the 265 cases
reviewed (mean:1.8 errors/case). Most discrepancies involved
staff (45.5%). Two hundred fifty-eight individual cognitive
errors were identified in 151 cases (mean = 1.7 errors/case).
Of these, 83 cases (55%) had additional perceptual or system-
related errors. One hundred sixty-five perceptual errors were
identified in 165 cases. Of these, 68 cases (41%) also had
cognitive or system-related errors. Fifty-four system-related
errors were identified in 46 cases (mean = 1.2 errors/case) of
which all were multi-factorial. Seven cases were unavoidable.
G. A. Taylor (*) : S. D. Voss
Department of Radiology, Harvard Medical School,
Childrens Hospital Boston,
300 Longwood Ave,
Boston, MA 02115, USA
e-mail: george.taylor@childrens.harvard.edu
P. R. Melvin : D. A. Graham
The Program for Patient Safety and Quality,
Childrens Hospital Boston,
Boston, MA, USA
D. A. Graham
The Department of Pediatrics, Harvard Medical School,
Boston, MA, USA
Conclusion Our study defines a taxonomy of diagnostic
errors in a large academic pediatric radiology practice and
suggests that most are multi-factorial in etiology. Further
study is needed to define effective strategies for improvement.
Keywords Diagnostic errors . Radiology . Pediatrics
Introduction
Risk-adjustment of medical errors for general medical
subspecialties such as internal medicine and for
procedure-based specialties is well developed, with clearly
defined outcomes and severity measures. Examples include
prevalence of post-operative infection, return to the oper-
ating room, length of stay in the ICU, and mortality for
surgical subspecialties. Yet, despite estimated error rates in
diagnostic radiology of 3.54.0%, similar tools for risk-
adjusting diagnostic imaging errors are not available [13].
The existing literature is largely descriptive, reporting
mainly the frequency and types of diagnostic errors
occurring in adult diagnostic radiology [4]. These studies
have focused on detection of pulmonary nodules, mammo-
graphic studies and overnight interpretation of emergency
cranial CT studies [57]. There is relatively little informa-
tion available regarding the causes of errors, and the
relative contribution of system-level and cognitive factors
in diagnostic imaging [8]. In addition, disease entities and
imaging strategies encountered in pediatric radiology
practice vary considerably from adult practice, and specific
analyses for this subspecialty are not available.
We performed a retrospective study of existing cases
reported as diagnostic errors during a 10-year period at a
large urban academic childrens hospital. Our goals were to
describe the patterns and potential etiologies of diagnostic
error in our subspecialty.
328 Pediatr Radiol (2011) 41:327334

Materials and methods Class IV: Disagreement in diagnosis that, if left un-corrected,
would have led to a major change in management.
Between December 30, 1998, and July 29, 2009, 131,728 (Example: Free intraperitoneal gas not detected on
cases were reviewed as part of our ongoing quality- abdominal radiograph).
assurance process. There are five general mechanisms by
which cases are identified and reported: the first is a Of the 131,728 cases subjected to additional review, a
hospital-wide sentinel event reporting system (SERS) with subset of 4,051 cases was identified in which a diagnostic
which clinicians report cases where there was a discrepancy error had been reported to our departmental Quality
or diagnostic error related to the Radiology Service. The Improvement Committee. Of the 4,051 discrepancies
second is the radiologist double-read process in which each reported, we limited our review to 265 cases that were
radiologist in the department reviews a series of randomly initially classified as Class III or IV disagreements.
selected cases interpreted by other staff radiologists. The Two senior pediatric radiologists reviewed each case
third method identifies discrepancies that arise during individually and categorized the etiology of the diagnostic
weekly multidisciplinary conferences (i.e. solid tumor, bone error according to the method developed by Graber [9] and
tumor, neuro-oncology, general surgical, urology and used widely in evaluating medical errors [1014]. Disagree-
advanced fetal care conferences) in the radiology depart- ments were resolved by consensus.
ment. These represent a random selection of active cases in A diagnostic error was defined as a diagnosis that was
the hospital at any given time. Identified discrepancies are delayed, wrong or missed [12]. Errors were further classified
assessed with regard to clinical progression of disease, as perceptual, cognitive, system-related or unavoidable [9,
surgical or pathological findings. The fourth method is 1214]. Perceptual errors were defined as non-recognition of
reporting by radiologists who either missed cases them-
selves or identified cases that were missed by other
radiologists. These are reported at the monthly morbidity Table 1 Reader status, imaging modality and initial severity rating in
and mortality conference of the department. Finally, there is 265 cases with diagnostic errors
the daily review of images interpreted initially by resident n=265 n %
and fellow trainees during after-hours shifts. Although there
are biases in each of the sampling methods, the combina- Reader status
tion of inputs gives us a broad look at the type and Staff 121 45.7
distribution of cases as well as their severity. These Fellow 57 21.5
discrepancies were initially classified by consensus by our Resident 85 32.1
departmental Quality Improvement Committee according to Technologist/clinician 2 0.8
a four-step ordinal scale: Image modality
X-ray 143 54.0
Class I: Minor disagreement in observation that would Fluoroscopy 21 7.9
have led to no change in management. (Example: US 20 7.5
Small post-operative pneumomediastinum not CT 65 24.5
detected in a child in the intensive care unit MRI 14 5.3
who already has a thoracostomy tube in place). Nuclear medicine 2 0.8
Class II: Disagreement in diagnosis that would have led to Initial severity
no change in management. (Example: Initial Class I (minor event) 6 2.3
interpretation of pre-operative chest radiograph Class II (moderate event) 36 13.5
failed to diagnose cardiomegaly and shunt vascu- Class III (major event) 138 52.1
larity in a child with known VSD being prepared Class IV (severe event) 85 32.1
for surgical correction. Decision to operate and
operative planning would not have been affected). Definitions
Class III: Disagreement in diagnosis that, if left un-corrected, Class I: Disagreement in observation that would have led to no
would have led to a minor change in management. change in therapy
(Example: Initial impression on a skull radiograph Class II: Disagreement in diagnosis that would have led to no change
in therapy
of Luchenshadel skull, with final diagnosis of
Class III: Disagreement in diagnosis that would have led to a minor
normal convolutional markingshad this error
change in therapy
gone unrecognized, the child might have had an
Class IV: Disagreement in diagnosis that, if left un-corrected, would
unnecessary MRI with sedation to assess for Chiari have led to a major change in therapy. (Example: free peritoneal gas
II malformation). not detected on abdominal radiograph)
Pediatr Radiol (2011) 41:327334 329

an imaging abnormality. Errors were defined as cognitive in
nature when due to faulty information processing (over-
interpretation of an imaging finding, misinterpretation of a
finding or failure to consider a different diagnosis for a given
finding [premature closure]); faulty data gathering (poorly
performed imaging examination, inadequate review of
patient history or lack of consideration of a patients
underlying condition), or insufficient knowledge base.
System-related errors were defined as technical (due to
equipment failure) or organizational flaws such as ineffective
policies, inadequate training or supervision, and defective
communication. Repeated instances of the same error
(clustering) were also included as a system-based error. An
error was considered unavoidable when abnormal imaging
findings were absent or masked or so atypical that arriving at
a correct diagnosis would not be expected. An error was
considered multi-factorial when more than one type of error
was identified in a given clinical case.
The individual associated with the error (technologist,
radiologist or clinician) was tabulated along with level of
training of the radiologist involved (resident, pediatric
radiology fellow or staff radiologist).
This study was approved by our institutions Internal Review
Board, and the requirement for informed consent was waived.
Results
Table 1 shows the distribution of discrepancies by reader
status, imaging modality and initial severity score for the 265
cases reviewed in detail. Radiology trainees were responsible
for the majority of discrepancies reviewed (53.6%), followed
by staff radiologists (45.7%). This is consistent with the
workload distribution in our department in which 48% of the
studies are interpreted by staff radiologists alone. A technol-
ogist or clinician was associated with an error in less than 1%
of cases. The two modalities most frequently involved with
discrepancies were radiography (54%) and CT (24.5%).
Fluoroscopy (8%), US (7.5%) and nuclear medicine (0.8%)
were involved less often.

Table 2 Cognitive and perceptual errors

Errors (cases) Definition Example

Cognitive errors 258 (151)

Faulty information processing
Faulty interpretation of test
Premature closure
Over-interpretation of finding
Faulty context generation
Failure to order follow-up or
appropriate test
Faulty data gathering
Faulty/incomplete test
performance technique
Ineffective review of history/
physical exam
Faulty knowledge
Inadequate knowledge base
Inadequate skills
Perceptual error
Under-interpretation of finding
174
78 Diagnostic finding is correctly identified,
but incorrect conclusion is drawn
71 Failure to consider other diagnostic
possibilities once an initial diagnosis
has been reached
13 Importance or relevance of diagnostic
finding is over-emphasized
4 Lack of awareness/consideration of
aspects of patients situation that are
relevant to diagnosis
8 Diagnostician does not use an appropriate
test or take the appropriate next steps
after test
47
39 Standard diagnostic test is conducted
incorrectly or incompletely
8 Failure to collect appropriate information
prior to initiation of diagnostic test
37
29 Insufficient knowledge of relevant
condition
8 Insufficient skill in performance or
interpretation of diagnostic test
165 (165) Diagnostic finding is noticeable but missed
by diagnostician
Congenital hepatic hemangioma misinterpreted
as hepatic malignancy or metastases
Right mediastinal mass misinterpreted as upper
lobe atelectasis. Failure to consider other
possibilities.
Post-operative edema at surgical site interpreted
as residual sarcoma on CT
Stable liver metastases interpreted as resolved
despite history of no additional treatment
Recurrent lung opacities in same location on
chest radiograph. Failure to suggest CT
resulting in delayed diagnosis of pulmonary
AVM
No voiding phase obtained on cystogram,
resulting in delayed diagnosis of posterior
urethral valves
Failure to review prior images, resulting in
missed migration of catheter fragment from
atrium to pulmonary artery
Internal defibrillator leads malpositioned in left
ventricle interpreted as well-positioned
Enema on premature infant performed
by hand injection leading to perforation
Aneurysmal bone cyst in sacrum missed
on two pelvis radiographs
330 Pediatr Radiol (2011) 41:327334

Initial severity classification
We classified the discrepancies as major events (if left
un-corrected, these would have led to a minor change in
management) in 138 of 265 cases reviewed (52%), and
as severe events (if left un-corrected, these would have
led to a major change in management) in 85 cases
(32.1%). During the study, 42 cases were re-classified as
class I (6) or class II (36) disagreements but were still
included in the overall data and did not affect the overall
distribution of discrepancies in any category.
Etiology of diagnostic error
We identified 484 errors in the 265 cases reviewed (mean
1.8 errors/case, Tables 2 and 3). The most common type of
error was cognitive, accounting for 258 (53.3%) errors.
Within this category, faulty information processing
accounted for the majority of errors, including misinterpre-
tation of a test (78 errors), premature closure (71 errors) and
over-interpretation of a finding (13 errors) (Figs. 1 and 2).
Failures in data gathering were the next most common
source of cognitive error (47 errors), consisting of situations
when an imaging test was performed either incorrectly or
incompletely (39 errors) or in which insufficient information
was collected before starting a diagnostic test (8 errors). In our
environment, faulty knowledge was a relatively uncommon
etiology, accounting for only 37 cognitive errors. An
inadequate knowledge base contributed to most errors in this
category (29 errors), while errors were attributable to
insufficient skills at performing an imaging test in only eight
instances. Cognitive errors were seen in association with
perceptual or system-related errors in 83/151 cases (55%).
However, multiple cognitive errors were present in the
majority of cases with at least one cognitive error.
Perceptual errors were the next most common category,
accounting for 165/484 (34.1% of errors, Table 2, Fig. 3).
This class of error occurred as an isolated cause of error in
97 cases (59%), and was associated with cognitive or
system-related errors or both in 68 cases (41%). Table 3
shows the type and frequency of the 54 system-related errors
we encountered. Organizational errors were most frequent in
this category (48 errors), consisting primarily of repeating
instances of the same error type (clustering, 18 errors), faulty
or incomplete medical history (13 errors, Fig. 4), and
difficulties in communication (6 errors). Inefficient workflow
processes, inadequate policies and procedures, and failure to
supervise a system or trainee were less commonly attribut-
able causes of error. Technical errors consisting of a technical
or equipment failure could be identified in only six cases.
Errors were considered unavoidable in only seven instances.
Discussion
The majority of the diagnostic errors documented in our
study were cognitive and multi-factorial. These errors were
not commonly caused by a lack of medical knowledge but

Table 3 Organizational and unavoidable errors

Errors (cases) Definition Example

System error 54 (46)

Organizational error
Clustering
Faulty medical history
Teamwork/
communication
Inefficient processes
Management/supervision
Policy and procedures
Technical error
Unavoidable errors
48
18 Repeating instances of same error type Missed cortical (buckle) fractures by radiology residents
13 Absent, confusing, incomplete or History of TB exposure given in child with hilar
erroneous medical history calcification and pulmonary opacity. History of missing
tooth not given resulting in misinterpretation of aspirated
tooth and atelectasis as Ghon complex
6 Failure to share needed information Poor communication with clinical team resulting in wrong
or skills, hand off problems imaging test ordered
5 Standardized processes resulting in Requisition missing, resulting in unreported case and
unnecessary delay missed progressive pneumonia
4 Failed oversight of system issues or Trainee not supervised during fluoroscopy, resulting in
of trainees missed vascular ring
2 Policies that fail to account for PET images not available in a timely fashion for comparison
certain conditions or that facilitate with CT images, resulting in missed residual tumor
error-prone situations
6 Technical or equipment failure Inappropriately low milliamperage setting on CT, resulting
in missed liver lesion
7 (7) Imaging findings are absent or CT scan shows no signs of appendiceal inflammation in
masked or present in a very patient with acute appendicitis
atypical fashion
Pediatr Radiol (2011) 41:327334 331

Fig. 1 Faulty interpretation

error. A 12-day-old girl with
rapidly involuting hepatic
hemangioma (RICH),
presenting with hepatomegaly.
Axial (a) and coronal (b)
contrast-enhanced abdominal
CT images show enhancing
nodular hepatic densities
misinterpreted as
hepatoblastoma

rather by a combination of lack of recognition of an
important imaging finding, faulty interpretation of a
finding, and premature closure. A number of cases were
assigned to multiple categories of error. For example, the
cases illustrated in Figs. 1 and 2 and included in Table 2
were included as examples of a specific type of cognitive
error, yet these two cases were categorized as both faulty
interpretation and premature closure.
Radiologists engage in two interrelated processes when
interpreting imaging studies: perception and analysis [15].
Kundel [16] defines perception as the unified awareness of
the content of a displayed image and analysis as
determining the meaning of the perception in the context
of the medical problem that initiated the acquisition.
Failures in perception show up as failure to identify an
important finding on an imaging study, the most common
source of diagnostic error identified in our study (165/265
cases), while failures in the analytic portion of the process
(over- or faulty interpretation of a finding) were not as common
(151/265 cases). These findings are consistent with an earlier
review of 182 radiology errors at a university hospital that
identified 69% of diagnostic errors as perceptual/cognitive in
nature, of which only two were related to lack of knowledge
[8]. Under-interpretation of findings can be related to a
number of perceptual and visual phenomena. The first is
visual isolation, where attention is selectively focused on a
main area of the image while less or no attention is given to
secondary areas. A second cause has been termed "satisfaction
of search," which occurs when additional lesions remain
undetected after detection of an initial lesion [8, 17].
Students of cognitive psychology suggest that cognitive
errors are not the act of a few bad performers or the result
of ignorance but rather a result of systematic factors that are
predictable and occur routinely in clinical medicine [18].
Thus, understanding how these mistakes are made might be
helpful in correcting their underlying causes. Many diag-

Fig. 2 Premature closure

error. A 5-year-old boy with
neuroblastoma presenting with
cough. a Anteroposterior chest
radiograph shows right upper
lobe density interpreted as
right upper lobe atelectasis.
No other diagnostic possibilities
were considered. b Coronal
contrast-enhanced CT image
shows solid right upper
mediastinal mass
332
Fig. 3 Perceptual error. An 11-year-old girl with aneurysmal bone
cyst in left sacrum presenting with hip pain. a Cyst (arrows) not
detected on anteroposterior spine radiograph. b Axial inversion
recovery MRI shows cystic lesion
nostic errors are the result of commonly used heuristics or
shortcuts in reasoning. These include the availability
heuristic in which individuals judge likelihoods based on
memory of a similar case, the framing effect in which
Pediatr Radiol (2011) 41:327334
individuals arrive at a different decision based on how the
information is presented and the anchoring heuristic in
which the initial impression is difficult to change once it is
solidly formed, despite conflicting new information (also
known as premature closure). Another pitfall recognized by
cognitive psychologists is blind obedience, in which a
diagnostician stops thinking when confronted by authority.
This authority can be human (a more senior radiologist) or
technical (reliance on a laboratory value) [18, 19].
Examples of these heuristics can be identified in radiology
practice: the inclusion of a ruptured Meckel diverticulum as
the most likely diagnosis in a child with ruptured
appendicitis because there was such a case recently in the
department (availability heuristic), the tendency to over- or
under-call pulmonary findings in a set of chest radiographs
when given either a history of cancer or a history of
screening for anesthesia (framing effect), or the repeated
missed diagnosis of intestinal lymphoma by more junior
staff because a respected, senior radiologist made the
diagnosis of constipation (blind obedience).
Finally, Berner and Graber [20] suggest in a compre-
hensive review article that diagnostic errors can result from
an attitude of overconfidence. Potchen [21], in a study of
diagnostic accuracy in 95 board-certified radiologists,
found that the top 20 radiologists had an average accuracy
rate of 95%, compared with 75% for the bottom 20. Yet, the
confidence level expressed by the bottom group was higher
than that of the best performers.
Minimization of cognitive errors remains an important
challenge, and a number of strategies have been described
to reduce them. These include developing an increased
awareness of cognitive biases, specific training and simu-
lation of common errors, and cognitive forcing strategies
where predictable biases under particular clinical situations
are identified and avoided. Yet, the validity of these
strategies has been recently called into question. Educa-
tional researchers Eva and Norman [22, 23] suggest that
attempts to be constantly vigilant and eliminate cognitive
biases are neither possible nor desirable because many of
the mental activities in which we engage are outside of
conscious awareness and heuristics used in clinical medi-
cine evolve because they yield better overall outcomes than
more careful or rational approaches. Other strategies involve
creating conditions that systematically minimize the risk of
cognitive errors, such as reducing time pressures to provide
adequate time for quality decision-making and providing rapid
and consistent feedback so that errors are immediately
identified, understood, and corrected [24].
One example of specific training used in an effort to
reduce errors is described by Halsted et al. [25]. This group
analyzed the frequency and types of diagnostic errors made
by radiology residents in interpreting pediatric radiographs
in an emergency setting. The majority of errors in their
Pediatr Radiol (2011) 41:327334
Fig. 4 Organizational
error (faulty medical history).
A 14-year-old boy with history
of exposure to tuberculosis,
presenting with wheezing.
AP (a) and lateral (b) chest
radiographs show aspirated
tooth and pulmonary opacities
misinterpreted as a probable
Ghon complex. History of
recent fight and missing tooth
was not provided
study (69%) involved recurrent under-interpretation of
buckle, Salter II, avulsion and transverse bony fractures.
This information was used to modify their training
programs to emphasize these fractures early in the
residents training and thus reduce the risk of this recurrent
error.
There are limitations to our methodology. The optimal
method of identification of diagnostic discrepancies would be
to perform a prospective study of a large random sample of
imaging studies performed in our department, with represen-
tative distribution among imaging modalities and radiologists
in near-real time. This would allow for a detailed root-cause
analysis of factors associated with missed diagnoses [10, 11].
Although we relied on a number of mechanisms to identify
errors, the data collected were non-random and had the
potential for a number of systematic biases, including the
over- or under-reporting of errors in certain modalities or
categories of error and self-reporting biases. The classification
of potential patient harm was based on the theoretical risk to
patient if a diagnostic error remained undetected. It was not
based on documented, actual patient outcomes. In the majority
of these cases, the discrepancy was identified before harm
came to a patient or therapy was changed.
Conclusion
Our study defines a taxonomy of errors encountered in a
large academic pediatric radiology practice and suggests
that most are multi-factorial in etiology. Yet, many
radiologists believe that increasing workload, increasing
numbers of images in CT and MRI studies, interruptions
and multi-tasking during the workday also have an impact
on error rates. Further studies are needed to evaluate the
333
effect of these variables on radiologist performance and to
better define modality and radiologist-specific error rates,
as well as effective strategies for long-term and sustainable
improvement.
References
1. Berlin L (2007) Accuracy of diagnostic procedures: has it
improved over the past five decades? AJR 188:11731178
2. Siegle RL, Baram EM, Reuter SR et al (1998) Rates of
disagreement in imaging interpretation in a group of community
hospitals. Acad Radiol 5:148154
3. Borgstede JP, Lewis RS, Bhargavan M et al (2004) RADPEER
quality assurance program: a multifacility study of interpretive
disagreement rates. J Am Coll Radiol 1:5965
4. Soffa DJ, Lewis RS, Sunshine JH et al (2004) Disagreement in
interpretation: a method for the development of benchmarks for
quality assurance in imaging. J Am Coll Radiol 1:212217
5. Miyakoshi A, Nguyen QT, Cohen WA et al (2009) Accuracy of
preliminary interpretation of neurologic CT examinations by on-
call radiology residents and assessment of patient outcomes at a
level I trauma center. J Am Coll Radiol 6:864870
6. Berbaum KS, Caldwell RT, Schartz KM et al (2007) Does
computer-aided diagnosis for lung tumors change satisfaction of
search in chest radiography? Acad Radiol 14:10691076
7. Miglioretti DL, Haneuse SJ, Anderson ML (2009) Statistical
approaches for modeling radiologists' interpretive performance.
Acad Radiol 16:227238
8. Renfrew DL, Franken EA, Berbaum KS et al (1992) Error in
radiology: classification and lessons in 182 cases presented at a
problem case conference. Radiology 183:145150
9. Graber ML, Franklin N, Gordon R (2005) Diagnostic error in
internal medicine. Arch Intern Med 165:14931499
10. VA National Center for Patient Safety (2002) The basics of
healthcare failure mode and effect analysis. Available at www.
patientsafety.gov/SafetyTopics.html. Accessed 6 July 2006
11. Bagian JP, Gosbee J, Lee CZ et al (2002) The veterans affairs root
cause analysis system in action. Jt Comm J Qual Improv 28:531
545
334
12. Benavidez OJ, Gauvreau K, Jenkins KJ et al (2008) Diagnostic
errors in pediatric echocardiography: development of taxonomy
and identification of risk factors. Circulation 117:29953001
13. Kassirer JP, Kopelman RI (1989) Cognitive errors in diagnosis:
instantiation, classification, and consequences. Am J Med 86:433441
14. Bordage G (1999) Why did I miss the diagnosis? Some cognitive
explanations and educational implications. Acad Med 74:S138S143
15. Kundel HL (2006) History of research in medical image
perception. J Am Coll Radiol 3:402408
16. Kundel HL (1979) Images, image quality and observer perfor-
mance. Radiology 132:265271
17. Berbaum KS, Franken EA Jr, Dorfman DD et al (1992) Satisfaction of
search in diagnostic radiology. Invest Radiol 27:571573
18. Redelmeier DA (2005) The cognitive psychology of missed
diagnoses. Ann Intern Med 142:115120
Pediatr Radiol (2011) 41:327334
19. Tversky A, Kahneman D (1986) Rational choice and the framing
of decisions. J Bus 59:S251S278
20. Berner ES, Graber ML (2008) Overconfidence as a cause of
diagnostic error in medicine. Am J Med 121:S2S23
21. Potchen EJ (2006) Measuring observer performance in chest
radiology: some experiences. J Am Coll Radiol 3:423432
22. Eva KW, Norman GR (2005) Heuristics and biasesa biased
perspective on clinical reasoning. Med Educ 39:870872
23. Eva KW, Norman GR (2010) Diagnostic error and clinical
reasoning. Med Educ 44:94100
24. Croskerry P (2003) The importance of cognitive errors in
diagnosis and strategies to minimize them. Acad Med 78:775780
25. Halsted MJ, Kumar H, Paquin JJ et al (2004) Diagnostic errors by
radiology residents in interpreting pediatric radiographs in an
emergency setting. Pediatr Radiol 34:331336