Interobserver and intraobserver variability in lung scan reading in suspected pulmonary embolism.

The interobserver and intraobserver variability in lung scan reading and interpretation was tested using a consecutive series of 96 lung scans done for patients clinically suspected to have pulmonary embolism. Lung scans were also done on 37 reference patients who were matched to a suspect case and consented to the procedure. Two experienced readers blindly read all films twice. Intraobserver agreement was low, especially for lung scans showing minor defects. Kappa, a measure of agreement beyond chance, was consistently less than .500. Interobserver agreement was lower. Only 4.3% of the suspect cases had lobar defects on the lung scan. The distribution of abnormalities in suspect and reference cases was similar. The results indicate that the reproducibility of lung scan readings is probably to low to be clinically useful, for readings involving defects that are segmental or smaller. The reproducibility for lobar defects could not be determined adequately because of their low frequency in this study.