Diagnostic performance of breast technologists in reading mammograms in a clinical patient population

Purpose: In the setting of an increasing workload for radiologists, this study focuses on the feasibility of skill mixing in breast imaging in a hospital radiology department. Methods: Two radiological technologists with more than 10 years of experience in performing mammograms were trained in prereading mammograms to select the cases that require further evaluation by a radiologist. Mammograms of consecutive patients were independently evaluated by the technologists, next to the standard clinical interpretation by the radiologist on duty. Mammographic findings were recorded and a BI-RADS classification was assigned for each breast. Different prereading scenarios were analysed using clinical decision rules. Two different cut-off points of BI-RADS classifications were applied to the data. Analysis was performed for the overall clinical patient population as well as for a subgroup of patients with no immediate indication for further work-up. Results: Mammograms of 1994 patients were evaluated. In total, 93 breast cancers were found in 91 patients (prevalence 4.6%). Sensitivity and specificity in selecting mammographic findings (cut-off point between BI-RADS 1 and BI-RADS 0, 2-5 and the radiologist's diagnosis as reference standard) was 98% and 74% for technologist 1 and 98% and 78% for technologist 2. In distinguishing normal and benign mammograms from those with abnormalities that are probably benign, suspicious or highly suggestive for malignancy (cut-off point BI-RADS 1-2 and BI-RADS 0, 3-5 and pathology results as reference standard), sensitivity decreased to 89% and 91% respectively. Specificity increased to 82% for both technologists. In a subgroup of 1389 patients with no immediate indication for additional imaging with the involvement of a radiologist, technologists obtained a mean sensitivity and specificity of 98% and 77% in detecting mammographic findings, and a mean sensitivity and specificity of 78% and 88% in detecting suspicious abnormalities. Conclusions: The employment of technologists in prereading mammograms seems to be an effective working strategy in daily clinical practice. However, its position in clinical practice remains indistinct as a continuous availability of radiologists still needs to be guaranteed. Nevertheless, as a substantial proportion of mammograms could be evaluated without the attention of a radiologist, the employment of technologists in prereading mammograms seems a promising new working strategy.