Table 1 Sample of published studies of radiological error
From: Error and discrepancy in radiology: inevitable or avoidable?
Year | Author | Ref | Material | Key points | Comments |
|---|---|---|---|---|---|
2001 | Goddard et al. | [11] | Various | Clinically significant error rate of 2–20%, depending on radiological investigation | |
1981 | Forrest et al. | [12] | Retrospective review of previous chest x-rays (CXRs) in patients subsequently diagnosed with lung cancer | False-negative rate of 40% | Lesions visible but not reported on prior studies |
1983 | Muhm at al | [13] | Lung cancers detected by plain radiography screening | 90% of cancers detected visible in retrospect on prior radiographs going back months or, in some cases, years (53 months in one case) | |
1993 | Harvey et al. | [14] | Review of prior mammograms in patients in whom impalpable breast cancer subsequently diagnosed by mammography | Evidence of carcinoma identifiable on prior studies in 41% when blindly reinterpreted, and in 75% when reviewers were aware of subsequent findings | |
1999 | Quekel et al. | [15] | Non-small cell lung cancer diagnosed on plain CXR | 19% missed diagnosis rate | 16-mm median diameter of missed lesions, median delay in diagnosis of 472 days |
1949 | In Robinson (1997) | [3] | CXR in patients with suspected TB | Interpreted differently by different observers in 10–20% | |
1990, 1994 | Markus et al., Brady et al. | Barium enema | Average observer missed 30% of visible lesions | Supposed gold standard of colonoscopy also subject to error | |
1999 | Robinson | [18] | Emergency dept. plain radiographs | Major disagreement between two observers in 5–9% of cases | Estimated error incidence per observer of 3–6% |
1997 | Tudor et al. | [19] | Plain radiographs | Mean accuracy: 77% without clinical information, 80% with clinical information. Modest improvements in sensitivity, specificity and inter-observer agreement with clinical information | Five experienced radiologists reported mix of validated normal and abnormal studies 5 months apart. No clinical information on first occasion, relevant clinical information provided on second occasion |
2008 | Siewert et al. | [20] | Oncologic CT | Discordant interpretations in 31–37%, with resultant change in radiological staging in 19%, and change in patient treatment in up to 23% | |
2007 | Briggs et al. | [21] | Neuro CT & MR | 13% major & 21% minor discrepancy rates (undercalls, overcalls & misinterpretations) | Specialist neuroradiologist second reading of studies initially interpreted by general radiologists |