Screening mammography facilities that have their ducks in a row are significantly more likely to interpret images accurately than institutions with looser attributes. Higher accuracy of interpretation was associated with facilities that offered only screening mammography, had a breast imaging specialist to interpret mammogram, and performed single reading, said Stephen Taplin, M.D., of the National Cancer Institute.
Such institutions also conducted two or more audit reviews a year. "Identifying facility structures and process that influence interpretive performance could be a foundation for improving the quality of mammography interpretive performance and choices among mammography facilities," Dr. Taplin and colleagues concluded in the June 18 issue of the Journal of the national Cancer Institute. Patient and radiologist characteristics have been shown to affect mammographic interpretive performance. However, the characteristics examined account for only 10% of the variation in performance, the authors said. Variation among mammography facilities has not been examined carefully, if at all, Dr. Taplin and colleagues continued. Identification of facility-specific factors that affect interpretive performance could help patients and physicians make better informed decisions about choosing a mammography facility and could inform the facilities about changes in practice that could improve interpretive performance.
Dr. Taplin and colleagues examined factors influencing interpretive performance at 44 mammography facilities that performed 484,463 screening mammograms on 237,669 women from 1996 to 2002. Breast cancer was diagnosed in 2,686 women during follow-up. The 44 facilities had a mean sensitivity of 79.6% and a mean specificity of 90.2%. Mean positive predictive value was 4.1%. Investigators calculated positive predictive value on the basis of the likelihood that cancer would be found among women referred for biopsy, and the mean among the facilities was 38.8%. The facilities' interpretive performance varied significantly with respect to specificity (P<0.001), positive predictive value (P<0.001), and the biopsy-defined positive predictive value (P=0.002). Additionally, the authors identified four facility characteristics that significantly influenced interpretative performance, as defined by area under the curve:
Screening mammograms only, versus screening and diagnostic mammograms (0.943 versus 0.911, P=0.006). Interpretation by a breast imaging specialist versus not (0.932 versus 0.905, P=0.004). Single reading versus independent double reading versus consensus double reading (0. 925 versus 0.915 versus 0.887, P=0.034). Audit reviews at least twice annually versus annually or unknown (0.929 versus 0.904 versus 0.900, P=0.018). Neither facility volume nor method of audit review influenced performance.
The study had a number of limitations including missing data, possible unaccounted differences among women and radiologists, and inability to assess differences in double-reading methods. The finding of poorer interpretive performance in facilities doing double readings is not consistent with other studies. "Understanding how facility characteristics influence interpretive accuracy is important because it could allow women and physicians to choose a mammography facility based on characteristics that are more likely to be associated with higher quality," the authors said. "Radiologists could also change the facilities' structure or processes to include practices that improve interpretive accuracy."