Descriptive features of gastric ulcers: do endoscopists agree on what they see?

BACKGROUND Little is known about the interobserver variation between endoscopists on descriptive morphologic features. METHODS This study describes the agreement among 10 endoscopists on their description of 12 morphologic features, using 10 photographs of gastric ulcers, and on their eventual interpretation. The endoscopists used a form with predefined options for description. RESULTS Kappa value was on average 0.36 for descriptive features and 0.31 for interpretation. The proportion of endoscopists agreeing on descriptive features was on average 84%, and 81% on interpretations. The chance of an endoscopist describing all 12 morphologic features of an ulcer on a photograph exactly the same as a colleague ranged from 4% to 46% (average 15%). A positive correlation between agreement in description and interpretation (0.75, p < 0.05) was found. CONCLUSIONS These results indicate a poor agreement between endoscopists in their translation of visual observations into descriptive terms. The positive correlation between agreement in description and interpretation suggests disagreement in description as an important cause for disagreement in interpretation. We believe that the use of more explicit descriptive terms will improve agreement in description and in subsequent interpretation.