Objective: Federal regulations mandate that radiologists receive regular albeit limited feedback regarding their interpretive accuracy in mammography. We sought to determine whether radiologists who regularly receive more extensive feedback can report their actual performance in screening mammography accurately.
Subjects and methods: Radiologists (n = 105) who routinely interpret screening mammograms in three states (Washington, Colorado, and New Hampshire) completed a mailed survey in 2001. Radiologists were asked to estimate how frequently they recommended additional diagnostic testing after screening mammography and the positive predictive value of their recommendations for biopsy (PPV2). We then used outcomes from 336,128 screening mammography examinations interpreted by the radiologists from 1998 to 2001 to ascertain their true rates of recommendations for diagnostic testing and PPV2.
Results: Radiologists' self-reported rate of recommending immediate additional imaging (11.1%) exceeded their actual rate (9.1%) (mean difference, 1.9%; 95% confidence interval [CI], 0.9-3.0%). The mean self-reported rate of recommending short-interval follow-up was 6.2%; the true rate was 1.8% (mean difference, 4.3%; 95% CI, 3.6-5.1%). Similarly, the mean self-reported and true rates of recommending immediate biopsy or surgical evaluation were 3.2% and 0.6%, respectively (mean difference, 2.6%; 95% CI, 1.8-3.4%). Conversely, radiologists' mean self-reported PPV2 (18.3%) was significantly less than their mean true PPV2 (27.6%) (mean difference, -9.3%; 95% CI, -12.4% to -6.2%).
Conclusion: Despite regular performance feedback, community radiologists may overestimate their true rates of recommending further evaluation after screening mammography and underestimate their true positive predictive value.