Abstract
Purpose :
To evaluate the interobserver agreement between uveitis experts on diagnosis of the specific uveitic disease.
Methods :
The Standardization of Uveitis Nomenclature (SUN) Project has collected 100-250 cases of each of 25 distinct uveitic diseases into a preliminary database using a standardized, informatics-derived, online data collection form (case collection). A final database of cases generally accepted to represent the disease (for each disease) is being selected (case selection) from the preliminary database by disease-specific committees, each of 9 individuals with broad representation from the international uveitis community. This final database will be used to develop and evaluate disease-specific classification criteria using machine learning approaches. Case selection proceeds in 2 steps: 1) independent online case voting by each disease committee member, in which cases are accepted into the final database with a supermajority (>75%) "yes" vote, rejected with a supermajority "no" vote, or tabled otherwise; and 2) consensus conference calls using nominal group techniques and anonymous voting on previously tabled cases. The online voting step provides the opportunity to evaluate disease-specific agreement between "experts" by comparing pairwise "yes/no" votes for each pair of comittee mambers. Kappa (κ) statistics were calcualted for each disease's 35 pairwise comparisons, and a mean κ was calculated for each disease.
Results :
Mean ± SD κ's for independent online voting for the following diseases were: Fuchs uveitis syndrome, 0.44±0.13; varicella zoster anterior uveitis, 0.58±0.10; birdshot chorioretinitis, 0.36±0.09; multifocal choroditis with panuveitis, 0.30±0.13; CMV retinitis, 0.27±0.16; serpiginous-like tuberculous choroiditis, 0.28±0.15; serpiginous choroiditis, 0.37±0.19; and sympathetic ophthalmia 0.31±0.12. After consensus conference calls, supermajority agreement was achieved on ≥99% of cases for each disease.
Conclusions :
These data demnostrate only moderate agreement at best among "experts" on the diagnosis of a specific uveitic entity. The nearly universal agreement after consensus calls suggests that development of disease-specific criteria can be achieved, but that formal classification criteria will be needed for each disease. The development of these criteria is the goal of the SUN Uveitis Classification Criteria Project.
This is an abstract that was submitted for the 2016 ARVO Annual Meeting, held in Seattle, Wash., May 1-5, 2016.