Radiological diagnostic errors are common and may have severe consequences. Understanding these errors and their possible causes is crucial for optimising patient care and improving radiological training. Recent postmortem studies using an animal model highlighted the difficulties associated with accurate fracture diagnosis using radiological imaging. The present study aimed to highlight the fact that certain fractures are easily missed on CT scans in a clinical setting and that caution is advised. A few such cases were discussed to raise the level of suspicion to prevent similar diagnostic errors in future cases. Records of adult patients from the radiological department at an academic hospital in South Africa were retrospectively reviewed. Case studies were selected by identifying records of patients between January and June 2021 where traumatic fractures were missed during initial imaging interpretation but later detected during secondary analysis or on follow-up scans. Seven cases were identified, and the possible causes of the diagnostic errors were evaluated by reviewing the history of each case, level of experience of each reporting radiologist, scan quality and time of day that initial imaging interpretation of each scan was performed. The causes were multifactorial, potentially including a lack of experience, fatigue, heavy workloads or inadequate training of the initial reporting radiologist. Identifying these causes, openly discussing them and providing additional training for radiologists may aid in reducing these errors.
This article aimed to use case examples of missed injuries on CT scanning of patients in a South African emergency trauma setting in order to highlight and provide insight into common errors in scan interpretation, their causes and possible means of mitigating them.
Traumatic fractures are frequently encountered in both postmortem and clinical settings.
Recent studies
Comparison of percentage of blunt (BFT) and sharp force (SFT) lesions detected by CT, X-ray and Lodox in various body regions in a postmortem context using a pig model. Piglets (black bars) were included to simulate cases of child abuse. (a) Skull, (b) ribs, (c) vertebrae, (d) forelimbs, (e) hindlimbs. Percentages taken from Spies et al.
While the radiological analysis in postmortem and clinical settings may differ, the potential reasons for failing to detect traumatic fractures may largely be the same. In a clinical setting, major detection errors may result in inadequate patient care and management and could have life-threatening results.
While the potential reasons for missing skeletal lesions radiologically in a postmortem context have been explored,
Records of adult patients (18–99 years) who presented to the radiological department at an academic hospital in South Africa, over a 6-month period between January 2021 and June 2021, were retrospectively reviewed by the senior author (D.N.P.). In this department, initial imaging interpretation is performed by the trainee radiologist (registrar) on call, and these initial reports are checked by a consultant radiologist during the following shift. Junior and senior registrars typically have less than and more than two years of experience prequalification, respectively. Junior and senior consultants, respectively, typically have less than and more than five years of training post qualification. Fractures detected during both the initial and secondary imaging interpretations are recorded.
Case studies were selected from patient records that showed fractures not diagnosed upon initial interpretation but identified during secondary analysis of the initial scan or upon analysis of a follow-up scan. Patient age, sex, case history, time of day initial imaging was performed, and the level of experience of the registrars and consultants were recorded. No patient identifying information was recorded, and patient anonymity was ensured throughout the study.
Seven cases with fractures not diagnosed at the initial interpretation of the radiological images, but identified during secondary analysis, were selected. These cases may not represent all of those where fractures were not diagnosed but were selected as they each had specific teaching points that could be used to highlight learning opportunities. They all also demonstrate specific potential pitfalls in detection and interpretation.
A 57-year-old man presented to the emergency department (ED) following a motor vehicle accident (MVA) and had been stabilised with a cervical collar. The patient experienced a loss of consciousness, headache and cervical spine tenderness. A head and cervical spine CT was performed at 03:23 on a Saturday. Initial interpretation by a junior registrar failed to detect any fractures and the cervical collar was removed. Soon thereafter, the patient developed ‘unexplained quadriplegia’. A subsequent CT pan-scan was performed at 10:00 the same day to search for any thoracolumbar spine fractures to explain the symptoms. A junior consultant identified a fracture of the C4 vertebra on the CT pan-scan, which, in retrospect, was visible on the initial CT (
Axial (a and b) and coronal (c) reconstructions of the cervical spine CT showing comminuted fractures of the left superior and inferior facets of C4 (black arrow) and left lamina of C4 (white arrow). Sagittal (d) reconstruction shows a small avulsion fracture of the anterosuperior margin of the C4 vertebral body (black arrow). These fractures represent an unstable cervical spine injury.
A 60-year-old man presented to the ED following a MVA. The patient had an altered level of consciousness, headache, cervical spine tenderness and was unable to move his legs. A head and cervical spine CT was performed at 17:56 on a Monday. The junior registrar noted severe degenerative changes in the cervical spine but no fractures. Upon review, the senior consultant detected a missed C4 spinous process fracture – a stable injury – but no other abnormalities (
Sagittal reconstruction of the initial CT of the cervical spine (a) showing the C4 spinous process fracture (black arrow). The follow-up CT (b) shows worsened retrolisthesis of the C4 on C5 vertebra with an anterior teardrop fracture of the C4 vertebra (white arrow). This fracture is visible in (a) but was called a ‘fractured osteophyte’ (white arrowhead). The axial (c) and coronal (d) reconstructions of the scan show how the degenerative cervical spine changes may make diagnosing fractures challenging – the teardrop fracture is pointed out by the white arrow in (c) and may have been misinterpreted as an osteophyte (white arrowhead). The black arrowhead in (c) demonstrates a right lamina fracture, and the black arrow again points out the spinous process fracture.
A 34-year-old man involved in a MVA, presented to the ED with loss of consciousness, a large head wound, cervical spine tenderness, a fractured left femur and rib fractures. A CT pan-scan was performed at 17:47 on a Saturday. A senior registrar performed the initial interpretation, reporting the cervical spine CT as normal. A junior consultant then reviewed the CT and noted several vertebral fractures that were missed initially (
Paramidline sagittal (a and b) reconstructions of a cervical spine CT showing a fracture of the right inferior facet of the C6 vertebra (black arrow), a fracture of the right superior facet of the C7 vertebra (white arrow) and a fracture of the anterosuperior margin of the body of C7 (black arrowhead). A midline sagittal reconstruction (c) demonstrates reversed cervical spine lordosis but no listhesis. Associated prevertebral soft tissue swelling is present (white asterisk). Note that the patient was ‘scanned skew’ (d).
A 29-year-old man sustained blunt trauma to the head and a CT on the day of injury demonstrated a large, depressed skull fracture which was treated conservatively. Ten days later, a high-resolution temporal bone CT was performed at 12:27 on a Wednesday after the patient complained of a 1-day history of right facial nerve fallout. Initial interpretation by a junior registrar noted the depressed squamous temporal fracture (
Axial CT of the head (a) demonstrating the depressed right squamous temporal bone fracture (black arrow) detected on initial interpretation. An oblique reconstruction of the right temporal bone (b) shows that the fracture (white arrow) extends into the facial nerve canal (highlighted here by the white dots). Specifically, it involves the tympanic segment of the canal. The fracture line extending into the mastoid part of the temporal bone is pointed out by the black arrowhead in the axial image (c) – note the fluid in the mastoid air cells and compare it to the well-aerated left mastoid air cell; a secondary sign of temporal bone fracture. The fracture line (white arrowhead) is much more conspicuous on the axial cut of the bone reconstruction algorithm (d) than the soft tissue algorithm (e), despite both being set to a standard ‘bone window’.
A 65-year-old man with an unknown mechanism of trauma presented with a right pneumothorax and extensive subcutaneous emphysema. A CT pan-scan (
A CT pan-scan showing extensive subcutaneous emphysema of both the head (a) and chest (b). When viewed without straightening the scan using multiplanar reconstruction (c), the right temporal bone otic capsule–sparing fracture is not easy to detect even utilising the bone reconstruction algorithm (black arrow) and nearly invisible (white arrow) on the soft tissue reconstruction algorithm (d). The bone window (e) demonstrates an initially overlooked nondisplaced right 9th rib fracture (black arrowhead). Note the right haemothorax (black star) and pneumothorax (white star) seen on the lung window of the chest CT (f). Also note that the patient was ‘scanned skew’.
An adult man of unknown age, involved in a MVA, presented to the ED on a Sunday. The patient had a head injury and a fractured tibia and fibula confirmed at radiography. A CT pan-scan was performed at 12:42, and initial interpretation by a junior registrar failed to diagnose subtle fractures of the right 10th, 11th and 12th ribs with associated lung contusions (
Axial bone reconstruction of a CT (a and b) of the chest demonstrating subtle buckle-type fractures of the 10th and 11th ribs (black arrows) associated with a small lung contusion (white arrow).
A 30-year-old man involved in a MVA had a loss of consciousness with an open midshaft fracture of the tibia and fibula confirmed on plain radiographs. A junior registrar performed the initial interpretation of the CT pan-scan at 02:05 on a Monday, reporting no additional fractures. Secondary analysis by a junior consultant, however, revealed a nondisplaced fracture of the right second rib and a buckle-type fracture of the right third rib (
Axial bone reconstruction CT of a the chest demonstrating a minimally displaced second rib fracture (black arrow) associated with a small lung laceration and contusion (white arrow). As with cases 5 and 6, lung injuries such as contusions or lacerations should prompt careful search for associated rib fractures in the trauma patient.
A growing body of literature shows that there are many radiological misdiagnoses of fractures of the ribs, vertebrae and cranial region, both in clinical and postmortem contexts,
While some of the fractures missed in the present study are subtle (cases 5, 6 and 7), others are very evident on the initial scans (cases 1, 2 and 4). That these fractures were not detected or were misinterpreted as other abnormalities are likely due to a combination of factors termed ‘observer errors’ or other extenuating circumstances. It could even suggest inadequate radiology training and/or inadequate image interpretation techniques.
Three types of observer error have been described.
In a postmortem setting, missed fractures were shown to be partially due to a lack of radiological experience and training.
Lack of experience with and knowledge of normal skeletal anatomy and skeletal growth, development and degeneration may result in misdiagnosing fractures as normal anatomical variants or other pathological conditions.
However, diagnostic errors are also commonly made by more experienced radiologists. The cervical spine fractures in case 2 were also initially misdiagnosed by a consultant as a stable injury, having only detected the spinous process fracture and missing the unstable teardrop fracture (recognition error). One potential reason for this could be that as experience level increases, the speed with which the observer interprets images increases, and as a result, so does the number of detection errors.
Errors of speed may also occur due to increased workloads and a reduction in the time available for radiological reporting.
Additionally, error rates may be related to the level of alertness of the radiologist.
Another type of diagnostic error is a satisfaction of search error, which is the result of one abnormality causing the attention of the radiologist to be diverted away from another, such as a fracture, resulting in this abnormality being overlooked.
In cases of polytrauma, CT pan-scans are often requested and include noncontrast head, cervical spine, contrast-enhanced chest and multiphase abdomen and pelvis imaging. Occasionally, these scans will also include peripheral or cervical CT angiograms, if clinically indicated. Trauma patients, for various reasons, are sometimes placed on the CT table in ways that result in the scans not being true axial cuts such as in cases 3 and 5. The resultant asymmetry when the patient is ‘scanned skew’ can make fractures of the skull and spine difficult to detect (cases 3 and 5). This was also noted in the postmortem studies using pig models, where some of the X-ray and Lodox images were ‘skew’, and some scans were not true-lateral or true-frontal images.
As in the postmortem contexts,
Like the imaging of trauma in postmortem contexts, the causes of diagnostic errors in detecting traumatic fractures in a radiology department are multifactorial and may include lack of radiologist training, knowledge and experience; fatigue and heavy workloads; and inadequate image interpretation techniques. Understanding these errors and their root causes is crucial to improving the efficacy of radiological departments. Additional training and open discussions of these errors and their causes, treating them as learning opportunities, can aid in reducing the prevalence of reporting errors. In particular, a high index of suspicion is important, especially when injuries to the chest and skull are concerned, as these are the most commonly misdiagnosed regions.
We acknowledge the chief executive officer (CEO) and head of the Department of Radiology at the academic hospital in South Africa for allowing access to and analysis of patient radiological records.
The authors declare that they have no financial or personal relationships that may have inappropriately influenced them in writing this article.
All authors contributed to the study’s conception and design. Data collection was performed by D.N.P. The first draft of the manuscript was written by A.J.S. with support from D.N.P. All authors commented on all subsequent versions of the manuscript, read and approved the final manuscript.
Permission was obtained from the Human Research Ethics Committee (Medical) at the University of the Witwatersrand (ref. no. M220114), and permission to access patient records was obtained from both the CEO and the head of the Department of Radiology at the academic hospital in South Africa.
This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.
The data that support the findings of this study are available on request from the corresponding author, A.J.S., subject to ethical clearance. The data are not publicly available due to ethical considerations.
The views and opinions expressed in this article are those of the authors and do not necessarily reflect the official policy or position of any affiliated agency of the authors.