Q&A: Radiologist, Patient Factors and Mammography Accuracy

August 27, 2015

A look at what radiologist and patient factors play a role in false-positive mammography.

As scrutiny increases on the effectiveness of breast cancer screening guidelines, pressure also mounts on radiologist performance in both screening and diagnostic mammography.

Researchers have found that minimum interpretive volume requirements, as well as minimal requirements for diagnostic interpretation for radiologists, could reduce the number of false-positive work-ups for breast cancer. But while experience does play a major role in image interpretation, other variables outside of the radiologist's control may also have an effect.

Breast density, prior history of mammography, and the presence of symptoms such as a lump are associated with higher interpretive performance by radiologists, according to a study published in the American Journal of Roentgenology. The study looked at patient and radiologist characteristics for two types of diagnostic mammograms that took place between 1998 and 2008: added evaluation of a recent screening mammogram and evaluation of a breast-related problem. Approximately 20% of the radiologists were affiliated with an academic institution.

Seven regions in the U.S. were represented in this study. Two-hundred-forty-four radiologists who interpreted 274,401 diagnostic mammograms performed for additional evaluation responded to the survey. A total of 104,115 examinations were performed for additional evaluation of a recent mammogram, 4663 with cancer. Another 170,286 examinations were performed for evaluation of a breast problem, 7007 with cancer.

The researchers evaluated patient and radiologist characteristics associated with false-positive rate and sensitivity for each diagnostic mammogram type:

Type of examinationFalse-positive RateSensitivityAUC
Additional evaluation of a recent mammogram11.9%90.2%0.894
Evaluation of a breast problem7.6%83.9%0.871

“Multiple patient characteristics were associated with measures of interpretive performance, and radiologist academic affiliation was associated with higher sensitivity for both indications for diagnostic mammograms,” the authors wrote.

The researchers found that several patient characteristics were associated with the false-positive rate, such as older age and breasts that were not heterogeneously dense. Radiologist characteristics included:

Female sex

Academic affiliation

Fellowship training

10–19 years of mammography interpretation

Greater percentage of time spent in breast imaging

Greater screening and diagnostic interpretive volume

The results also showed that if indication for the diagnostic mammogram was a breast lump, compared with nipple discharge or pain, the interpretations were more likely to be false-positives and have higher sensitivity.

Radiologist characteristics associated with higher false-positive rates included:

Younger radiologist age

Female sex

Academic affiliation

Fewer years interpreting mammograms

Higher sensitivity was associated with:

Female sex

Academic affiliation

Fellowship training

10–19 years of mammography interpretation

Having never been named in a malpractice suit

“These results indicate the potential for improved radiologist training, using evaluation of their own performance relative to best practices, and for improved clinical outcomes with health care system changes to maximize access to diagnostic mammography interpretation in academic settings,” the researchers concluded.

Diagnostic Imaging spoke with Sara Jackson, MD, MPH, who led this study. Jackson is a clinical assistant professor in the Department of Medicine, and director of the Adult Medicine Clinic at Harborview Medical Center, in Seattle, Washington.[[{"type":"media","view_mode":"media_crop","fid":"40768","attributes":{"alt":"","class":"media-image media-image-right","id":"media_crop_1364592183399","media_crop_h":"0","media_crop_image_style":"-1","media_crop_instance":"4226","media_crop_rotate":"0","media_crop_scale_h":"0","media_crop_scale_w":"0","media_crop_w":"0","media_crop_x":"0","media_crop_y":"0","style":"height: 201px; width: 200px; float: right;","title":"Sara Jackson, MD, MPH","typeof":"foaf:Image"}}]]

Why did you decide to do this study?

Variability in performance for diagnostic mammography had previously been identified, and we were looking to see if we could elucidate opportunities to improve performance based upon patient or radiologist characteristics.

Did any of the results surprise you?

It was surprising that many patient characteristics are associated with sensitivity and false positive rate, but this wasn’t the case for radiologist characteristics. We expected factors such as radiologists' experience and volume of interpretation to be associated with better interpretive performance. But, while radiologists with academic affiliations were more likely to have greater interpretive sensitivity, little else was identified. This may indicate that the variability is primarily due to inherent patient or breast tissue characteristics.

With the changes in technology, do you think that the differences between the academic-based radiologists and those who aren't in academia will change?

This is speculative and not based upon our study, but I suspect that any technologic changes that alter performance would affect academics and non-academics similarly. Teleradiology, however, could potentially improve access to diagnostic interpretation by academic radiologists, as capacity allows.

What would you like the readers to take away from the study?

Although academic radiologists have higher cancer detection rates, overall, the interpretive accuracy of diagnostic mammograms is high for all.

Also, the patient characteristics, including factors inherent to breast tissue, are associated with most of the variability in diagnostic mammography interpretation, and these are difficult to modify.

What's next for you? What would you like to look at?

It would be interesting to study continuing education opportunities for radiologists who interpret diagnostic mammograms that are designed by academic radiologists, and incorporate the radiologists’ own performance data relative to benchmarks for academic practice. Additionally, study of health system opportunities that maximize access to academic interpretation of diagnostic mammography would be of interest.