Diagnostic errors involving imaging include not ordering appropriate imaging, incorrectly reading imaging, and incorrectly applying the results of diagnostic imaging due to its intrinsic limitations. We can order the right image and correctly read it, but if we don’t consider the inherent test characteristics we can still make a diagnostic error.
This especially applies to X-rays which are simple screening tools with very imperfect sensitivities. Scaphoid fractures, for example, are notorious for being radiographically occult—and we are trained to assess snuffbox tenderness in FOOSH patients and immobilize them even if the X-rays fail to reveal a fracture, because of the poor sensitivity of X-rays.
It’s important to know the limitations of X-rays so we can correctly interpret them in clinical context, and consider better imaging—from Point Of Care UltraSound (POCUS) to CT or MRI. Below is a brief overview of X-ray limitations and considerations for high-risk patients.
Cervical spine injury
X-ray is a good screening tool in the low-risk patient, but for high-risk patients we need CT. According to Computed tomography versus plain radiography to screen for cervical spine injuru: a meta-analysis:
- 52% sensitive X-ray
- 98% sensitive CT
X-rays of epiglottitis show the “thumbpring sign”, measuring >8mm in adults, but this is only 90% sensitive. A relative measurement taking into account patient size–the epiglottis width to the AP width of C4>0.33–is more sensitive, at 95% (see A Reappraisal of the Radiologic Findingns of Acute Inflammation of the Epiglottis and Supraglottic Structures in Adults). But for the high-risk patient the gold standard is laryngoscopy
- X-ray sensitivity: 90-95%:
- laryngoscopy: 100%
According to the literature the X-ray sensitivity for pneumothorax is a surprisingly low 50%, perhaps because this includes supine patients in whom the air layors anteriorly rather than superiorly. POCUS is emerging as a far quicker and more accurate study: Diagnosis of pneumothorax by radiography and ultrasonography: A meta-analysis.
- X-ray sensitivity: 50%
- POCUS: 90%
Traditionally we stratify fever and cough into pneumonia (Xray infiltrate) or bronchitis (normal Xray). But Xray sensitivity is only 65%, so we need to be on the lookout for radiographically occult pneumonia in the high-risk patient. Ultrasound can help, with a sensitivity of 80% (Lung Ultrasound May Be Superior to Chest X-ray for Diagnosing Pneumonia), and in the admitted patients radiographic infiltrates often develop over time. Admission chest radiograph lacks sensitivity in the diagnosis of community-acquired pneumonia. “In patients admitted with a clinical diagnosis of CAP, the initial chest radiograph lacks sensitivity and may not demonstrate parenchymal opacifications in 21% of patients. Moreover, greater than half of patients admitted with a negative chest radiograph will develop radiographic infiltrates within 48 hours.”
- X-ray sensitivity: 65%
- POCUS: 80%
The importance of a good clinical exam for CHF is emphasized by the poor performance of CXR. Radiographic signs include cardiomegaly (a chronic sign predisposing to CHF), cephalization, hilar haziness, Kerley B lines, interstitial or alveolar edema and pleural effusion. But the poor sensitivity of these signs in CHF is compounded by the challenge of interpretation. Ability of physicians to diagnose congestive heart failure based on chest X-ray: “Physicians correctly identified the CHF chest X-rays 79% of the time (sensitivity 59%, specificity 96%; positive likelihood ratio 14.6, negative likelihood ratio 0.43)… An accuracy of 95% among radiologists suggests that a negative X-ray does not rule out CHF.” Use of chest radiography in the emergency diagnosis of acute congestive heart failure: “Cardiomegaly was the only radiographic finding with a sensitivity > 50%. However, specificity (71%) was lower than that for cephalisation (93%), hilar haziness (94%), Kerley B lines (96%), interstitial oedema (93%), and alveolar oedema (99%). Accuracy was modest overall and was highest for cardiomegaly (67%). The radiological interpretation regarding CHF was accurate in only 69% of patients. Patients with an incorrect radiological interpretation had a non‐significantly higher rate of hospital admission (91% v 81%, p = 0.078) and longer time to discharge (median 14 v 9 days, p = 0.068) than did patients with a correct interpretation.” Accuracy of chest radiograph interpretation by emergency physicians “The proportion of agreement between emergency physician and radiologist reports for normal, congestive heart failure, and pneumonia cases was 84.3%, 41.4%, and 41.4%, respectively. Emergency department physicians frequently missed specific radiographic abnormalities, and there was considerable discrepancy between their interpretations and those of trained radiologists.” On the other hand, POCUS is emerging as a complement to the physical exam to increase our diagnostic accuracy—looking for B lines. Point-of-care Ultrasonography for the Diagnosis of Acute Cardiogenic Pulmonary Edema in Patients Presenting With Acute Dyspnea: A Systematic Review and Meta-analysis: “The sensitivity of US using B-lines to diagnosis ACPE is 94.1% (95% confidence interval [CI] = 81.3% to 98.3%) and the specificity is 92.4% (95% CI = 84.2% to 96.4%) …in patients with a moderate to high pretest probability for ACPE, an US study showing B-lines can be used to strengthen an emergency physician’s working diagnosis of ACPE.
- X-ray sensitivity: 50%
- POCUS: 95%
Similarly, diagnosing small bowel obstruction based on Xray is a skill that improves with experience. Accuracy of abdominal radiography in acute small-bowel obstruction: does reviewer experience matter?: “The sensitivity for SBO among the six reviewers ranged from 59% to 93%. The senior staff members were significantly more accurate. The mean sensitivity, specificity, and accuracy for all six reviewers were 82%, 83%, and 83%, respectively. Three radiographic signs were highly significant (p < 0.001): two or more air-fluid levels, air-fluid levels wider than 2.5 cm, and air-fluid levels differing more than 5 mm from one another in the same loop of small bowel.” But even correctly read the X-ray has at best a 70% sensitivity, while POCUS is better at 90% and CT 95%. Bedside ultrasonography for the detection of small bowel obstruction in the emergency department: “Dilated bowel on US had a sensitivity of 91% (95% CI 75 to 98%) and specificity of 84% (95% CI 69 to 93%) for SBO… X-ray had a sensitivity of 46.2% (95% CI 20.4 to 73.9%) and specificity of 66.7% (95% CI 48.9 to 80.9%) for SBO when diagnostic, but was non-diagnostic 36% of the time.”
- X-ray sensitivity: 70%
- POCUS: 90%
- CT 95%
Like the cervical spine, Xrays are a good screen for the low-risk trauma patient with mid to lower back pain. But in high-risk patients we need CT. Are plain radiographs of the spine necessary during evaluation after blunt trauma? Accuracy of screening torso computed tomography in thoracic/lumbar spine fracture diagnosis: “Sensitivity and specificity of CT scan for thoracolumbar fracture were excellent at 100% and 97%, respectively, with a negative predictive value of 100%. Plain radiographs were 73% sensitive, 100% specific, and had a negative predictive value of 92%.”
- X-ray sensitivity: 75%
- CT: 100%
Hip fracture is radiologically apparent in 90% of cases. For the 10% of radiographically occult fractures CT picks up most but not all–so MRI is required for those with very high pre-test probability. Magnetic resonance imaging identifies occult hip fractures missed by 64-slice computed tomography: “Of 235 hip fractures, 211 were visible on initial plain films (90%, 95% CI 85-93%) and 24 (10%, 95% CI 6-15%) were occult. Eighteen occult fractures (7.6%, 95% CI 4.6-11.8%) were identified by CT (MRI not done), one (0.4%, 95% CI 0-2%) by MRI (CT not done), one (0.4%, 95% CI 0-2%) by both CT and MRI, and 4 patients (1.7%, 95% CI 0.5-4.3%) had a positive MRI but negative CT scan”
- X-ray sensivity: 90%
- CT: 98%
- MRI: 100%
In summary, the key to image interpretation starts with a good clinical exam and formulation of pre-test likelihood, which determines the image modality ordered and the interpretation of its results–including injuries that may be radiographically invisible and require more advanced imaging. POCUS is emerging as a helpful complement to the bedside clinical exam, with a higher sensitivity for many emergencies that can help us generate faster and more accurate diagnoses.