Accuracy of the chest radiograph in diagnosis of pulmonary embolism.

Abstract

In an effort to determine the sensitivity and specificity of the chest roentgenogram for the diagnosis of pulmonary embolism, roentgenograms of 152 patients who were all suspected of having pulmonary embolism were randomized and presented to nine interpreters. One hundred eight patients in the series were proven to have pulmonary embolism on the basis of a positive pulmonary angiogram. Forty-four patients were assumed not to have embolism on the basis of either a normal perfusion isotope scan or a pulmonary angiogram which did not show embolism. The interpreters were requested to indicate whether pulmonary embolism was present or absent, or whether they could not tell from the roentgenogram. Readers had no prior knowledge of the actual disease state. The average true-positive ratio (sensitivity) was 0.33, with a range of 0.52 to 0.88. The average true-negative ratio (specificity) was 0.59, with a range of 0.31 to 0.80. The false-positive and false-negative ratios were respectively, 0.21 (range 0.05 to 0.39) and 0.41 (range 0.15 to 0.70). A predictive index, reflecting the overall accuracy of diagnosis, was calculated for the entire group and was 0.40, with a range of 0.17 to 0.57. There appeared to be no correlation between training or experience and accuracy of performance in this study.