Does diagnostic accuracy in mammography depend on radiologists' experience?

Abstract

This study was designed to determine if radiologists' experience in mammography is associated with their performance in correctly interpreting mammograms. Study mammograms (n = 150) were chosen by stratified random sampling from those interpreted as normal, abnormal-benign or abnormal-suspicious for cancer, with oversampling of cancer cases. Ten radiologists who had varying amounts of experience were asked to read the mammograms. Associations between the levels of the radiologists' experience and their accuracy in reading mammograms were assessed. Significant associations (p < 0.05) were found between the frequency of immediate workup recommendations in cancer patients and obtaining feedback, total lifetime mammograms read, number of mammography continuing medical education (CME) credits, and practice type. Radiologists with more experience also noted smaller cancer lesions. However, these experience variables were also associated with increased workup recommendations in the noncancer patients (p < 0.10). In multivariable analysis, obtaining regular feedback and the total lifetime number of mammograms read were independently associate with the number of times immediate workup was recommended in the cancer cases. The most experienced radiologist had the highest sensitivity in diagnosing breast cancer. Further studies are needed to assess whether the current requirements of the U.S. Food and Drug Administration for radiologists who read mammograms ensure acceptable levels of accuracy.