1688 
A high discrepancy rate has been found in the interpretation of advanced scans among radiologists from the same speciality.
There was a high discrepancy in the interpretation of advanced radiological scans among radiologists who had the same specialist training. This was the unexpected finding from a retrospective analysis of nearly 6 million acute examinations according to researchers from the Department of Radiology and Imaging Sciences, Indianapolis, US. Fellowship or specialist training has been found to be a highly desired attribute when hiring radiologists and provides them with the necessary expertise in a specific area of practice. Nevertheless, despite having undertaking fellowship training, in practice, much of the work of radiologists requires them to be ‘multi-specialists‘, i.e., they are required to be able to interpret scans across a range of specialities such as neurology or abdominal radiology.
Although several studies have examined image interpretation and discrepancy rates between radiologists from different specialities, to date, no studies have specifically considered situations where the radiologists have been concordant, i.e., have undertaken the same sub-speciality training comparing differing specialities, i.e., discordant. For the present study, the US team turned to a large US tele-radiology company database which contracted with community hospitals and provided a service, 24 hour a day, 7 days a week, for a range of examinations including CT, MRI and ultrasound scanning. The company tele-radiologist provided an initial interpretation of a scan and then the second (final) interpretation was undertaken by a radiologist at the hospital.
For the purposes of the analysis, examination scans were classified as common or advanced, the presence of a discrepancy as either major or minor (based on factors such as patient safety and clinical outcomes) and the relationship between the two radiologists as either concordant or discordant.
Findings
The analysis included 5,883,980 examinations performed on patients with a mean age of 50 years. For the whole examinations, 40% were deemed concordant with respect to the interpreting radiologist’s specialist training. The overall discrepancy rate was 0.43%, of which 0.14% were deemed to by major.
Among concordant radiologists, the major discrepancy rate was lower with common compared to advanced examinations (0.13% vs 0.26% common vs advanced). In addition, among both common and advanced examinations, the frequency of major and minor discrepancies was not different between concordant and discordant radiologists.
However, in cases where the two radiologists were concordant, among advanced examinations, the frequency of major discrepancies was significantly higher (0.26% vs 0.18% concordant vs discordant) giving a 45% increased likelihood of a major discrepancy for examinations among concordant vs discordant specialities (hazard ratio, HR = 1.45, 95% CI 1.18 – 1.79). Similarly, the frequency of minor discrepancies was also higher (HR = 1.17) among concordant compared to discordant radiologists (0.34% vs 0.29%, concordant vs discordant).
In their discussion, the authors noted that for the interpretation of both common and advanced examinations, there were no important differences between concordant and discordant radiologists, which was unsurprising given that many of these examinations would be frequently encountered in practice. Nevertheless, the authors were unable to account for the high discrepancy rate for both major and minor discrepancies in advanced examinations for concordant radiologists. In other words, why there were differences in the interpretation when both radiologists had undertaken the same specialist training.
They concluded by suggesting that practice leaders should carefully consider efforts to match sub-specialities when interpreting scans.
Citation
Chong S et al. Interpretations of Examinations Outside of Radiologists’ Fellowship Training: Assessment of Discrepancy Rates Among 5.9 Million Examinations From a National Teleradiology Databank. AJR Am J Roentgenol 2021
