THE EFFECT OF RADIOLOGY TRAINING ROUTE AND WORKSTATION LAYOUT ON MAMMOGRAPHY DECISIONS

Purpose: To measure the effect of mammography readers' previous training on diagnostic performance in the transition to digital mammography.

Method: In the UK there are two types of mammography reader; radiologists and radiography advanced practitioners (radiographers trained to read mammograms). Radiologists experience a broader training programme in medicine and radiology. Performance of these two groups was investigated with two different digital mammography workstation layouts: with prior mammograms (from the previous screening round) digitised and adjacent to the current mammograms; or presented in film format and perpendicular to the current mammograms. 160 difficult test cases were read by four radiologists and four radiography advanced practitioners at each workstation. Diagnostic performance was measured using Jackknife Free Response Receiver Operating Characteristic (JAFROC). Whether participants looked at the prior mammograms and time taken was recorded for each case using video equipment.

Result: There was no difference in overall performance between radiologists and radiography advanced practitioners (F(1)=.002, p=.97). There was an interaction between type of mammography reader and both workstation layout (F(1)=10.5, p=.048), and percentage of cases for which the prior mammogram was looked at (F(1,6)=11.6, p=.01). Radiologists outperformed radiography advanced practitioners and looked at the prior mammograms for a greater proportion of cases when using film prior mammograms, but the reverse relationship was apparent when using digitised prior mammograms, see figure 1. Participants spent longer per case in the modalities in which they performed the worst, suggesting performance differences were not due to lack of effort.

Conclusion: The introduction of changes to the breast screening programme such as new technology or workstation changes may affect mammography readers in different ways dependent on the training route they took, even if they do not impact on diagnostic performance overall. Identifying these relationships may help optimise performance.