Comparison of interpretations of CT angiograms in the evaluation of suspected pulmonary embolism by on-call radiology fellows and subsequently by radiology faculty

OBJECTIVE. Our objective was to evaluate interobserver variability in interpretations performed by on-call radiology fellows and subsequently by attending radiologists of CT angiograms obtained for clinically suspected pulmonary embolism and to evaluate factors contributing to discrepancies. MATERIALS AND METHODS. Written interpretations made by on-call fellows were compared with reports approved by attending radiologists for all CT angiograms obtained for suspected pulmonary embolism after work hours and on weekends in a recent 19-month period. Interpretations were stratified as positive, negative, or equivocal for pulmonary embolism. In cases of discordant interpretations, those CT angiograms were rereviewed by two thoracic radiologists; then patient medical records were reviewed for evidence of clinical effect. Technical and patient-related reasons for discordant interpretations of CT angiograms were recorded. RESULTS. Six hundred fifty-eight oncology patients were examined on CT angiography; five were examined twice. The fellows reported 137 CT angiograms (21%) as positive, 498 (75%) as negative, and 28 (4%) as equivocal for pulmonary embolism. Interpretations of the fellows and attending radiologists agreed in 93% (615/663) of CT angiograms (kappa = 0.80). The concordance rates for CT angiograms interpreted by fellows as positive (89%, 122/ 137), negative (96%, 479/498), and equivocal (50%, 14/28) were significantly different from each other (p < 0.001 for each). A significantly greater proportion of CT angiograms with discordant interpretations was reported to be technically limited (p < 0.01). No clear adverse clinical events were attributed to discordant interpretations of CT angiograms, although the death of one patient in that subgroup was of indeterminate cause. CONCLUSION. In the evaluation of CT angiograms obtained for suspected pulmonary embolism, on-call fellows showed good agreement with attending radiologists. CT angiograms with discordant interpretations often were limited by technical or patient-related factors.