Intra- and interrater agreement with cumulative defect curves

To examine intra- and interrater agreement when analyzing cumulative defect curves. Cumulative defect (Bebié) curves provide a graphical representation of the visual field and allow a subjective classification of diffuse and localized loss.

We used 75 Humphrey 30-2 visual field tests, randomly chosen from a database of 782 fields of 113 patients with open-angle glaucoma. Cumulative defect curves were generated and randomly arranged into five sets, with each set containing the 75 curves in a unique sequence. Five raters (two experienced and three inexperienced) rated each set and classified each curve as showing diffuse loss, localized loss, both diffuse and localized loss, or no loss. The intra- and interrater agreement in rating the curves was then analyzed.

Intrarater agreement ranged from 73.3-88.0% for perfect rater agreement, with 5 identical ratings. Agreement for experienced versus inexperienced raters gave similar results (means: 84.0% and 77.8%, respectively, for perfect agreement). Interrater agreement for each set, evaluated by the kappa statistic, was substantial for all 5 sets (0.65-0.71). Kappa values for each set were comparable for experienced and inexperienced raters (0.72-0.83 and 0.59-0.69 respectively).

Analyzing the nature of visual field loss using the cumulative defect curve is simple to learn and provides high intrarater agreement as well as substantial interrater agreement.