Radiologist Performance in the Detection of Pulmonary Embolism: Features that Favor Correct Interpretation and Risk Factors for Errors

Abstract

Purpose: This study aimed to assess the factors contributing toward accurate detection and erroneous interpretation of pulmonary embolism (PE).

Materials and methods: Over 13 months, all computed tomography pulmonary angiography studies were retrospectively rereviewed by a chest radiologist. Two additional chest radiologists assessed cases with disagreement between the first interpretation and rereview. The number, extent, and location of PE and specialty training, experience, time of study, kV, resident prelim, use of iterative reconstruction, signal to noise ratio (SNR), and reports describing the study as "limited" were recorded. Parametric and nonparametric statistical testing was performed (significance P<0.05).

Results: Of 2555 computed tomography pulmonary angiography cases assessed, there were 230 true positive (170 multiple, 60 single PE), 2271 true negative, 35 false-negative (15 multiple and 20 single PE), and 19 false-positive studies. The overall sensitivity, specificity, positive predictive value, negative predictive value and accuracy of radiologists was 86.8%, 99.2%, 92.4%, 98.5%, and 97.9%. Sensitivity for the detection of multiple and central PE was significantly higher than the detection of single and peripheral PE, respectively (P<0.01 for both). The sensitivity of thoracic radiologists (91.7%) was higher than nonthoracic (82.8%) and reached significance for single PE (89.2% vs. 61.4%, P<0.02). Errors were more likely in cases with lower SNR (P=0.04) and those described as limited (P<0.001). Misses occurred more frequently in the upper lobe posterior and lower lobe lateral segments and subsegments (P=0.038).

Conclusions: The accuracy for PE detection is high, but errors are more likely in studies with single PE interpreted by nonthoracic radiologists, especially when located in certain segments and in cases with low SNR or described as limited.