The Wisconsin Age-related Macular Degeneration grading system: performance in an independent centre

Age-related Macular Degeneration (AMD) is the commonest cause of blindness in developed nations. Despite this, the epidemiology of AMD is poorly understood. A need for the documentation of AMD prevalence and incidence at a population level has stimulated the development of a comprehensive, observer-based, photographic grading method for AMD in Wisconsin.

To independently assess the performance of the Wisconsin method by self-taught graders outside its centre of inception.

The inter-observer variability and confidence limits for detection of change were assessed for two self-taught graders (ophthalmologists). Self teaching was achieved exclusively from documentation and photographs provided by the system developers in Wisconsin. 295 retinal photographs of elderly people were independently assessed for 13 features by each of the two graders.

Weighted and unweighted kappa statistics, % exact and one step apart agreement, and confidence limits for detection of change were calculated for the graded features on a "by eye' basis, and where appropriate, on a "by retinal subfield' basis. Levels of agreement for weighted kappa were moderate to substantial for most features. 95% and 90% confidence limits for significant change beyond measurement error were determined in terms of scale increments.

We conclude that the Wisconsin AMD grading system can be independently learnt from documentation and photographs alone, and that an acceptable level of performance is attainable by self-taught graders.