Interobserver variability in the interpretation of contrast venography, technetium-99m red blood cell venography and impedance plethysmography for deep vein thrombosis.

Contrast venography (CV) is the standard technique for diagnosing deep vein thrombosis (DVT). Newer noninvasive tests have also proven efficacious. However, there is a lack of data on the level of agreement among observers in their interpretation of the results of the various tests. After agreeing on well-defined criteria, three experienced observers assessed, blindly, the results of tests performed over a 4-month period on 117 patients who were suspected clinically of having had a first episode of DVT. The kappa statistic was used to measure the level of agreement beyond chance for CV (69 patients), red blood cell venography (RBCV) (82 patients) and impedance plethysmography (76 patients). The results of CV were assigned to normal, abnormal or inadequate categories, and those of RBCV and IPG to normal, equivocal or abnormal categories. The kappa values for CV, RBCV and IPG ranged from 0.53 to 0.56, 0.42 to 0.56 and 0.90 to 0.91 respectively. Values greater than 0.75 represented excellent agreement beyond chance and those between 0.40 and 0.75 represented fair to good agreement. Excellent kappa values were obtained for IPG because interpretation of the results of this method is entirely objective. Although the values for CV and RBCV showed good to fair agreement, there was a greater degree of observer variation, despite the well-defined criteria, indicating the subjectivity of interpretation of these test results. It is concluded that the kappa statistic can be used to measure observer variation of the results of tests for diagnosing DVT and may serve as a quality control tool for studies in which more than one person interprets the results.